# ADVANCED IMMUNIZATION TECHNOLOGIES FOR NEXT GENERATION VACCINES

EDITED BY : Donata Medaglini, Rino Rappuoli, Peter Andersen and David J. M. Lewis PUBLISHED IN : Frontiers in Immunology

### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-931-1 DOI 10.3389/978-2-88963-931-1

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# ADVANCED IMMUNIZATION TECHNOLOGIES FOR NEXT GENERATION VACCINES

Topic Editors: Donata Medaglini, University of Siena, Italy Rino Rappuoli, GlaxoSmithKline (Italy), Italy Peter Andersen, Statens Serum Institut (SSI), Denmark David J. M. Lewis, Imperial College London, United Kingdom

Citation: Medaglini, D., Rappuoli, R., Andersen, P., Lewis, D. J. M., eds. (2020). Advanced Immunization Technologies for Next Generation Vaccines. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-931-1

# Table of Contents

*06 Editorial: Advanced Immunization Technologies for Next Generation Vaccines*

Donata Medaglini, Peter Andersen and Rino Rappuoli


Birgit Weinberger, Mariëlle C. Haks, Roelof A. de Paus, Tom H. M. Ottenhoff, Tanja Bauer and Beatrix Grubeck-Loebenstein

*35 Heterologous Prime-Boost Combinations Highlight the Crucial Role of Adjuvant in Priming the Immune System*

Annalisa Ciabattini, Elena Pettini, Fabio Fiorino, Simone Lucchesi, Gabiria Pastore, Jlenia Brunetti, Francesco Santoro, Peter Andersen, Luisa Bracci, Gianni Pozzi and Donata Medaglini

*48 Overcoming the Neonatal Limitations of Inducing Germinal Centers Through Liposome-Based Adjuvants Including C-Type Lectin Agonists Trehalose Dibehenate or Curdlan*

Maria Vono, Christiane Sigrid Eberhardt, Elodie Mohr, Floriane Auderset, Dennis Christensen, Mirco Schmolke, Rhea Coler, Andreas Meinke, Peter Andersen, Paul-Henri Lambert, Beatris Mastelic-Gavillet and Claire-Anne Siegrist

*60 Molecular Signatures of a TLR4 Agonist-Adjuvanted HIV-1 Vaccine Candidate in Humans*

Jenna Anderson, Thorunn A. Olafsdottir, Sven Kratochvil, Paul F. McKay, Malin Östensson, Josefine Persson, Robin J. Shattock and Ali M. Harandi


Sven Kratochvil, Paul F. McKay, Amy W. Chung, Stephen J. Kent, Jill Gilmour and Robin J. Shattock

*96 Integrase Defective Lentiviral Vector as a Vaccine Platform for Delivering Influenza Antigens*

Alessandra Gallinaro, Martina Borghi, Roberta Bona, Felicia Grasso, Laura Calzoletti, Laura Palladino, Serena Cecchetti, Maria Fenicia Vescio, Daniele Macchia, Valeria Morante, Andrea Canitano, Nigel Temperton, Maria Rita Castrucci, Mirella Salvatore, Zuleika Michelini, Andrea Cara and Donatella Negri

### *107 Age and Influenza-Specific Pre-Vaccination Antibodies Strongly Affect Influenza Vaccine Responses in the Icelandic Population Whereas Disease and Medication Have Small Effects*

Thorunn A. Olafsdottir, Kristjan F. Alexandersson, Gardar Sveinbjornsson, Giulia Lapini, Laura Palladino, Emanuele Montomoli, Giuseppe Del Giudice, Daniel F. Gudbjartsson and Ingileif Jonsdottir

*118 Mouse Models of Influenza Infection With Circulating Strains to Test Seasonal Vaccine Efficacy*

Helen T. Groves, Jacqueline U. McDonald, Pinky Langat, Ekaterina Kinnear, Paul Kellam, John McCauley, Joanna Ellis, Catherine Thompson, Ruth Elderfield, Lauren Parker, Wendy Barclay and John S. Tregoning

*129 Respiratory Disease Following Viral Lung Infection Alters the Murine Gut Microbiota*

Helen T. Groves, Leah Cuthbertson, Phillip James, Miriam F. Moffatt, Michael J. Cox and John S. Tregoning


Christina W. Obiero, Augustin G. W. Ndiaye, Antonella Silvia Sciré, Bonface M. Kaunyangi, Elisa Marchetti, Ann M. Gone, Lena Dorothee Schütte, Daniele Riccucci, Joachim Auerbach, Allan Saul, Laura B. Martin, Philip Bejon, Patricia Njuguna and Audino Podda

*164 Alum/Toll-Like Receptor 7 Adjuvant Enhances the Expansion of Memory B Cell Compartment Within the Draining Lymph Node* Hoa Thi My Vo, Barbara Christiane Baudner, Stefano Sammicheli,

Matteo Iannacone, Ugo D'Oro and Diego Piccioli

*176 Src Family Kinases Regulate Interferon Regulatory Factor 1 K63 Ubiquitination Following Activation by TLR7/8 Vaccine Adjuvant in Human Monocytes and B Cells*

Lorenza Tulli, Francesca Cattaneo, Juliette Vinot, Cosima T. Baldari and Ugo D'Oro

*188 High Antigen Dose is Detrimental to Post-Exposure Vaccine Protection Against Tuberculosis*

Rolf Billeskov, Thomas Lindenstrøm, Joshua Woodworth, Cristina Vilaplana, Pere-Joan Cardona, Joseph P. Cassidy, Rasmus Mortensen, Else Marie Agger and Peter Andersen

*199 Protective Effect of Vaccine Promoted Neutralizing Antibodies Against the Intracellular Pathogen* Chlamydia trachomatis Anja Weinreich Olsen, Emma Kathrine Lorenzen, Ida Rosenkrands,

Frank Follmann and Peter Andersen

*211 Seasonal Influenza Split Vaccines Confer Partial Cross-Protection Against Heterologous Influenza Virus in Ferrets When Combined With the CAF01 Adjuvant*

Dennis Christensen, Jan P. Christensen, Karen S. Korsholm, Louise K. Isling, Karin Erneholm, Allan R. Thomsen and Peter Andersen

### *221 AS03- and MF59-Adjuvanted Influenza Vaccines in Children*

Amanda L. Wilkins, Dmitri Kazmin, Giorgio Napolitani, Elizabeth A. Clutterbuck, Bali Pulendran, Claire-Anne Siegrist and Andrew J. Pollard

*238 An Unexpected Major Role for Proteasome-Catalyzed Peptide Splicing in Generation of T Cell Epitopes: Is There Relevance for Vaccine Development?*

Anouk C. M. Platteel, Juliane Liepe, Willem van Eden, Michele Mishto and Alice J. A. M. Sijts

*244 Molecular Signatures of Immunity and Immunogenicity in Infection and Vaccination*

Mariëlle C. Haks, Barbara Bottazzi, Valentina Cecchinato, Corinne De Gregorio, Giuseppe Del Giudice, Stefan H. E. Kaufmann, Antonio Lanzavecchia, David J. M. Lewis, Jeroen Maertzdorf, Alberto Mantovani, Federica Sallusto, Marina Sironi, Mariagrazia Uguccioni and Tom H. M. Ottenhoff

*256 Improved Immune Responses in Young and Aged Mice With Adjuvanted Vaccines Against H1N1 Influenza Infection*

Susan L. Baldwin, Fan-Chi Hsu, Neal Van Hoeven, Emily Gage, Brian Granger, Jeffrey A. Guderian, Sasha E. Larsen, Erica C. Lorenzo, Laura Haynes, Steven G. Reed and Rhea N. Coler

*270 Efficacy Testing of H56 cDNA Tattoo Immunization Against Tuberculosis in a Mouse Model*

Anouk C. M. Platteel, Natalie E. Nieuwenhuizen, Teresa Domaszewska, Stefanie Schürer, Ulrike Zedler, Volker Brinkmann, Alice J. A. M. Sijts and Stefan H. E. Kaufmann

*285 Nanoporous Microneedle Arrays Effectively Induce Antibody Responses Against Diphtheria and Tetanus Toxoid*

Anne Marit de Groot, Anouk C. M. Platteel, Nico Kuijt, Peter J. S. van Kooten, Pieter Jan Vos, Alice J. A. M. Sijts and Koen van der Maaden

# Editorial: Advanced Immunization Technologies for Next Generation Vaccines

### Donata Medaglini <sup>1</sup> \*, Peter Andersen<sup>2</sup> and Rino Rappuoli 3,4,5

*<sup>1</sup> Department of Medical Biotechnologies, University of Siena, Siena, Italy, <sup>2</sup> Department of Infectious Disease Immunology, Statens Serum Institut, Copenhagen, Denmark, <sup>3</sup> GSK, Siena, Italy, <sup>4</sup> vAMRes Lab, Toscana Life Sciences, Siena, Italy, 5 Imperial College, London, United Kingdom*

Keywords: ADITEC, immunization technologies, vaccines, adjuvants, vaccine vectors, systems vaccinology, human immunology

**Editorial on the Research Topic**

### **Advanced Immunization Technologies for Next Generation Vaccines**

The spectacular increase of life expectancy in many countries of the world is mainly due to the conquest of infectious diseases by vaccines, hygiene, and antibiotics. Vaccines have done and continue to do an excellent job in eliminating or reducing the impact of infectious diseases (1). Vaccines prevent 2.5 million deaths per year: every minute five lives are saved by vaccines worldwide. However, despite the huge progress made in past decades, there is still work to be done: for some important diseases we do not have a vaccine yet, and for others, currently available vaccines are not good enough. Thanks to the new technologies, vaccines can do much more for the needs of modern society. Several challenges and unmet needs still remain and require an urgent effort in vaccine research and development: emerging infectious diseases, infectious diseases linked to poverty, bacteria resistant to antibiotics, and non-communicable diseases.

To address these challenges new immunization technologies that allow for the development of safe and more effective vaccines are needed. New technologies and big data analysis, including genomics and systems biology, are fueling advances in our understanding of human immunology, transforming the old field of vaccinology, and shaping the future of medicine. Thanks to new technologies, vaccines have the potential to make an enormous contribution to the health of modern society by preventing and treating not only communicable diseases in all ages, but also non-communicable diseases such as cancer. The sophisticated science behind the development of modern vaccines has become so complex that scientists and policy-makers need to develop a new model for vaccine research (2–4), and joint research efforts that combine academic research with the vaccine development and manufacturing expertise found in commercial vaccine companies are essential.

To this aim, the high Impact Project ADITEC (Advanced Immunization Technologies; www. aditecproject.eu/) was funded by the 7th Framework Programme for Research and Innovation of the European Union (2). The project has contributed to accelerate the development of novel and powerful immunization technologies for the next generation of human vaccines. Scientists from 42 partner institutions in 13 different European countries and USA have joined forces in the ADITEC project. With a budget of about 30 million euros over 6 years, ADITEC has made significant advances in the development of novel immunization technologies, adjuvants, vectors and delivery systems, formulations, and vaccination methods optimized for different age groups and for different routes of administration. ADITEC has conducted 12 vaccine clinical trials and contributed to international regulation and standards for these novel technologies. Along with regularly setting up and running European training programmes, ADITEC has also created synergies and cross-

### Edited and reviewed by:

*Denise Doolan, James Cook University, Australia*

> \*Correspondence: *Donata Medaglini donata.medaglini@unisi.it*

### Specialty section:

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

> Received: *08 April 2020* Accepted: *16 April 2020* Published: *06 May 2020*

### Citation:

*Medaglini D, Andersen P and Rappuoli R (2020) Editorial: Advanced Immunization Technologies for Next Generation Vaccines. Front. Immunol. 11:878. doi: 10.3389/fimmu.2020.00878* fertilization between different research areas that have the potential to fill existing gaps and advance this knowledge well into the future. To date, ADITEC has a track record of over 325 scientific publications in international peerreviewed journals (https://www.aditecproject.eu/publications/) and has generated innovations with high socio-economic impact (https://www.aditecproject.eu/category/reports/impactreports/). New technologies have been developed that can effectively advance these innovations to the clinic and make a real difference for future health.

In this Research Topic are assembled a series of original research articles and reviews, which highlight the progresses made by the ADITEC project in the development of advanced immunization technologies.

Novel insights have been obtained in the field of adjuvants and delivery systems (Anderson et al.; Baldwin et al.; Christensen et al.; Obiero et al.; Olsen et al.; Santoro et al.; Tulli et al.; Vo et al.; Vono et al.), vectors (Gallinaro et al.; Nieuwenhuizen and Kaufmann), vaccine formulations (Billeskov et al.), delivery devices (deGroot et al.; Platteel, Nieuwenhuizen et al.), antigen design (Platteel, Liepe et al.), and prime–boost immunization strategies (Ciabattini, Pettini et al.). Together with structurebased design of immunogens this wealth of new information contribute to development of new immunization technologies with the potential to radically transform the field of vaccinology. Moreover, better infection models using clinically relevant influenza strains to test vaccine induced protection (Groves, McDonald et al.) have been developed and the potential impact of host microbiota on vaccine immune responses has been investigated [(5); Groves, Cuthbertson et al.].

Recent omics and systems biology big data platforms have yielded valuable insights and a systems vaccinology approach, integrating clinical, immunological, and omics data, has the potential to contribute to identification of markers that will guide the rational development of vaccines (6, 7). Here are reported the human molecular signatures of immunity and immunogenicity in infection and vaccination identified in the context of the ADITEC clinical studies and technologies developed to assess the human immune response at the level of the transcriptomic and proteomic response, T-cell and B-cell memory, cellular trafficking, and key molecular pathways of innate immunity, with an emphasis on mechanisms of protective immunity (Anderson et al.; Haks et al.).

Human challenge models, in which volunteers are experimentally infected with a pathogen of interest, offer

### REFERENCES


the opportunity to identify both natural and vaccine-induced correlates of protection. In this Research Topic, we also highlight how the application of transcriptomics to human challenge studies allows for the identification of novel correlates and gives insight into the immunological pathways required to develop functional immunity (Barton et al.).

Vaccines need to be tailored for diverse age and target groups to optimally stimulate the host immune system. Indeed, the changes in the immune system of the elderly, where some immunological components are weakened (immunosenescence) while others such as inflammation are increased, are the basis of the reduced response to vaccination in the elderly (8). There is the need to design vaccine formulations, including vaccine adjuvants, that optimally stimulate the elderly immune system taking in consideration the need to avoid excess inflammation. Similarly, vaccination in infants, the other important target population of vaccination, face the challenge of an immature immune system that impose specific requirements on vaccines to be able to prime the necessary immune response in small children. In this issue are reported studies on the characterization of human vaccine immune responses in different age groups, including adults (Anderson et al.; Kratochvil et al.; Obiero et al.), elderly (Weinberger et al.), and children (Wilkins et al.), through the latest generation of advanced methodologies. The effect of age, gender, underlying disease, and medication in vaccine responses has also been investigated (Olafsdottir et al.).

With its broad portfolio of complementary expertise from both academia and the private sector integrated by a truly collaborative spirit, ADITEC represent a valuable model for future large integrated projects to tackle complicated problems for the benefit of society.

### AUTHOR CONTRIBUTIONS

All authors conceived the outline of the manuscript. DM wrote the manuscript. RR and PA critically revised and approved the final version of the manuscript.

### ACKNOWLEDGMENTS

We acknowledge the financial support from the Commission of the European Communities, Seventh Framework Programme, contract HEALTH-2011-280873 Advanced Immunization Technologies (ADITEC).

vision for the vaccines of tomorrow. Vaccine. (2018) 36:1136–45. doi: 10.1016/j.vaccine.2017.11.069


humans. Semin Immunol. (2013) 25:209–18. doi: 10.1016/j.smim.2013. 05.003


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Medaglini, Andersen and Rappuoli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Transcriptomics of the Vaccine immune response: Priming With adjuvant Modulates recall innate responses after Boosting

*Francesco Santoro1 \*† , Elena Pettini1†, Dmitri Kazmin2 , Annalisa Ciabattini1 , Fabio Fiorino1 , Gregor D. Gilfillan3 , Ida M. Evenroed3 , Peter Andersen4 , Gianni Pozzi1 and Donata Medaglini1 \**

*<sup>1</sup> Laboratorio di Microbiologia Molecolare e Biotecnologia (LA.M.M.B.), Dipartimento di Biotecnologie Mediche, Università di Siena, Siena, Italy, 2Emory Vaccine Center, Emory University, Atlanta, GA, United States, 3Department of Medical Genetics, Oslo University Hospital, University of Oslo, Oslo, Norway, 4Department of Infectious Disease Immunology, Statens Serum* 

### *Edited by:*

*Jeffrey K. Actor, University of Texas Health Science Center at Houston, United States*

### *Reviewed by:*

*Robert Jeffrey Hogan, University of Georgia, United States Katie Louise Flanagan, RMIT University, Australia Shaper Mirza, Lahore University of Management Sciences, Pakistan*

### *\*Correspondence:*

*Francesco Santoro santorof@unisi.it; Donata Medaglini donata.medaglini@unisi.it*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 22 December 2017 Accepted: 18 May 2018 Published: 05 June 2018*

### *Citation:*

*Santoro F, Pettini E, Kazmin D, Ciabattini A, Fiorino F, Gilfillan GD, Evenroed IM, Andersen P, Pozzi G and Medaglini D (2018) Transcriptomics of the Vaccine Immune Response: Priming With Adjuvant Modulates Recall Innate Responses After Boosting. Front. Immunol. 9:1248. doi: 10.3389/fimmu.2018.01248*

Transcriptomic profiling of the immune response induced by vaccine adjuvants is of critical importance for the rational design of vaccination strategies. In this study, transcriptomics was employed to profile the effect of the vaccine adjuvant used for priming on the immune response following re-exposure to the vaccine antigen alone. Mice were primed with the chimeric vaccine antigen H56 of *Mycobacterium tuberculosis* administered alone or with the CAF01 adjuvant and boosted with the antigen alone. mRNA sequencing was performed on blood samples collected 1, 2, and 7 days after priming and after boosting. Gene expression analysis at day 2 after priming showed that the CAF01 adjuvanted vaccine induced a stronger upregulation of the innate immunity modules compared with the unadjuvanted formulation. The immunostimulant effect of the CAF01 adjuvant, used in the primary immunization, was clearly seen after a booster immunization with a low dose of antigen alone. One day after boost, we observed a strong upregulation of multiple genes in blood of mice primed with H56 + CAF01 compared with mice primed with the H56 alone. In particular, blood transcription modules related to innate immune response, such as monocyte and neutrophil recruitment, activation of antigen-presenting cells, and interferon response were activated. Seven days after boost, differential expression of innate response genes faded while a moderate differential expression of T cell activation modules was appreciable. Indeed, immunological analysis showed a higher frequency of H56-specific CD4+ T cells and germinal center B cells in draining lymph nodes, a strong H56-specific humoral response and a higher frequency of antibody-secreting cells in spleen of mice primed with H56 + CAF01. Taken together, these data indicate that the adjuvant used for priming strongly reprograms the immune response that, upon boosting, results in a stronger recall innate response essential for shaping the downstream adaptive response.

Keywords: vaccine adjuvant, RNA sequencing, adaptive immunity, recall innate response, CAF01

# INTRODUCTION

*Institut, Copenhagen, Denmark*

Vaccines based on purified antigens are often poorly immunogenic and need to be formulated with adjuvants or delivery systems, to increase the amount, quality, and duration of the immune response to vaccination as well as to ensure long-lived immunological memory and protection (1). Adjuvants are substances capable of enhancing and properly skewing the immune responses to the vaccine

**9**

antigen, and their choice can dramatically affect the type and the magnitude of the adaptive immune response to the vaccine antigen, by impacting on the innate response starting signal (2).

Profiling the mode of action of different adjuvants is of critical importance for the rational design of vaccination strategies, based on heterologous combinations of vaccine formulations for priming and boosting, and for predicting the protective potential of an adjuvant for a given vaccine antigen (3–8). The immune profile and efficacy of five different adjuvants, combined with vaccine antigens from *Mycobacterium tuberculosis*, influenza, and *Chlamydia*, were tested in murine infection models within the ADITEC project on advanced immunization technologies (6, 9). We have also characterized the antigen-specific T and B cell responses after both parenteral and mucosal priming with vaccine formulations including different adjuvants or delivery systems, and combining heterologous prime-boost schedules, demonstrating the crucial role of both the immunization route and vaccine composition (3, 5, 10–13).

A systems biology approach, integrating gene expression data with immunological results, has been used to study the human response to different vaccines (14–26). Systems biology is an interdisciplinary approach to analyze multiple data types related to complex biological interactions by using computational analysis and mathematical modeling. The systems biology approach was first applied to characterize the immune response elicited by the yellow fever vaccine YF-17D in humans (23) and, more recently, to a variety of other vaccines including the study of adjuvanted and non-adjuvanted influenza vaccines in adults (21) and children (22, 27).

This approach has been recently applied to investigate the priming properties of different vaccine adjuvants at early time points after priming in the mouse model, using genome wide microarrays, and identified shared blood gene modules enriched in T follicular helper and germinal center responses (28). Mathematical models have been also developed to predict the probability of antigen-specific CD4+ T cells to expand and disseminate *in vivo* into other secondary lymphoid organs following primary immunization with adjuvanted vaccine formulations (29, 30).

CAF01 is a promising vaccine adjuvant that has been tested in five phase I clinical trials, administered in combination with four different antigens including the H56 tuberculosis (TB) vaccine antigen (Clinical trial no. NCT00922363), to evaluate its safety, tolerability, and immunogenicity. CAF01 is a liposomal adjuvant system composed of cationic liposome vesicles [dimethyldioctadecylammonium (DDA)] combined with a synthetic variant of cord factor of the mycobacterial cell wall [trehalose 6,6-dibehenate (TDB)], which was shown to activate macrophages and dendritic cells (DCs) through a specific innate activation program *via* Syk–Card9–Bcl10–Malt1 (31). CAF01 promotes vaccine depot formation, prolonging the release of antigens and stimulating the induction of T follicular helper cells into the draining lymph nodes (dLN), together with combined Th1 and Th17 responses and the generation of a robust, long-lived memory response in mice (32–34). CAF01 has been used as a component of a promising TB vaccine candidate in combination with the chimeric antigen H56 of *M. tuberculosis*

consisting of the antigen Ag85B fused to the 6-kDa early secretory antigenic target and the latency-associated protein Rv2660c (35, 36). The phase I clinical trials showed that CAF01 is safe and induces a strong cell-mediated immune response in addition to antibodies in humans (37, 38).

In this work, we have analyzed, through a systems biology approach, how the CAF01 adjuvant, combined with the H56 antigen used for priming, programs the immune response to downstream re-exposure to the same antigen. Mice were primed with H56 + CAF01 or H56 alone and boosted with the H56 antigen. A very low antigen dose was used for the boost to select antigen-specific clones of T and B cells and to mimic the challenge with the pathogen. RNA sequencing was used to characterize the blood transcriptome at several time points (1, 2, and 7 days) both after priming and after boosting, allowing us to follow the transcriptomic profile of the same animals over time. Transcriptomic data were analyzed together with immunological data on both cellular (antigen-specific CD4+ T cells and germinal center B cells in dLN) and humoral responses (quantification of H56-specific IgG up to 7 weeks after boost). These studies characterize, for the first time, using a systems biology approach, the modulation of the response elicited after prime-boost vaccination with the CAF01 adjuvant.

# MATERIALS AND METHODS

### Mice

Seven-week-old female C57BL/6 mice, purchased from Charles River (Lecco, Italy), were housed under specific pathogen-free conditions in the animal facility of the Laboratory of Molecular Microbiology and Biotechnology (LA.M.M.B.), Department of Medical Biotechnologies at University of Siena, and treated according to national guidelines (Decreto Legislativo 26/2014). All animal studies were approved by the Italian Ministry of Health with authorization no. 1004/2015-PR on September 22 2015.

### Experimental Design

Mice were subcutaneously primed with the chimeric TB vaccine antigen H56 administered alone or in combination with the liposome system CAF01, and boosted after 4 weeks with H56 antigen alone (**Figure 1**). Both the innate and adaptive immune responses elicited by the two vaccine formulations were explored. T and B cells responses were characterized 10 days after boosting (day 38) within the local dLN and spleen, while the induction of H56 specific IgG serum response was followed at different time points, up to 7 weeks post boost (day 77). In parallel, other groups of mice were immunized with the same vaccine schedule described earlier, and the molecular signature of the vaccine formulations was explored by sequencing RNA from blood at early time points (1, 2, and 7 days) after primary and secondary immunizations, and thus following the transcriptomic profile over time without sacrificing the animal (**Figure 1**).

### Immunizations

Groups of four to five mice were immunized by the subcutaneous route at the base of the tail. Vaccine formulations consisted of

the chimeric TB vaccine antigen H56 (2 μg/mouse for priming, 0.5 μg/mouse for boosting; Statens Serum Institut, Denmark), administered alone or combined with the adjuvant CAF01 (250 µg DDA and 50 µg TDB) per mouse (Statens Serum Institut, Denmark). Vaccine formulations containing CAF01 were injected in a volume of 150 μl/mouse of 10 mM Tris buffer, while formulations containing H56 alone in a volume of 100 μl/mouse of PBS. Mice primed with both H56 + CAF01 or H56 alone were boosted with H56 antigen 4 weeks later.

### Blood Sample Collection

For antibody analysis, blood samples were taken by temporal plexus bleed on days 0, 14, 28, 38, 56, and 77 after primary immunization. Blood was incubated for 30 min at 37°C, centrifuged at 1,200 × *g* at 4°C for 15 min, and then serum was collected and stored at −80°C until further analysis.

For transcriptomic analysis, blood samples were collected from individual mice on days 1, 2, 7, 28, 29, 30, and 35 after priming. Blood (200 µl) was mixed in a 1:2.8 ratio with PaxGene reagent, in a 2 ml Biosphere Eppendorf tube (39, 40). The tube was inverted five times and incubated at room temperature for 2 h before freezing at −80°C. Total RNA was extracted using the PaxGene blood RNA kit (Qiagen, Germany) following manufacturer's instructions and resuspended in 60 µl of BR5 buffer. RNA was quantified using (i) the total RNA nano kit on a 2100 Bioanalyzer (Agilent Technologies, USA), (ii) the RNA broad range kit on a Qubit 2.0 (Thermo Fisher, USA), and (iii) the Nanophotometer (Implen, Germany).

### Illumina Sequencing

Libraries were prepared using TruSeq stranded RNA reagents (Illumina, USA) according to manufacturer's instructions, using 500 ng of total RNA input per sample with dual indexing to enable all libraries to be sequenced in the same run. Pooled libraries were sequenced on four runs of an Illumina NextSeq 500 instrument with 75 bp single end reads. Image analysis and base calling were performed using Illumina's RTA software version 2.4.11 and bcl2fastq version 2.18.0.12. Reads were filtered to remove those with low base call quality using Illumina's default chastity criteria. Raw sequence data were deposited in Sequence Read Archive under Bioproject number PRJNA437839.

### Multiparametric Flow Cytometric Analysis

B and T responses were analyzed in dLN (sub iliac, medial, and external). Samples were mashed onto 70 µm nylon screens (Sefar Italia, Italy) and washed two times in complete RPMI medium (cRPMI, Lonza, Belgium) containing with 100 U/ml penicillin/ streptomycin and 10% fetal bovine serum (Gibco, USA). Samples were treated with red blood cells lysis buffer according to manufacturer instruction (eBioscience, USA). Cells were incubated for 30 min at 4°C in Fc-blocking solution [cRPMI medium with 5 µg/ml of CD16/CD32 mAb (clone 93; eBioscience, USA)]. To evaluate germinal center B lymphocytes from dLN, cells were stained with AF700-conjugated anti-CD45R (anti-B220, clone RA3-6B2; BD Biosciences), BV421-conjugated anti-GL7 (clone GL7; BD Biosciences), and PE-Cy7-conjugated anti-CD95 (clone Jo2; eBioscience). To evaluate H56-specific CD4+ T lymphocytes, cells from dLN were stained for 1 h at RT with PE-conjugated I-A(b) *M. tuberculosis* Ag85B precursor 280–294 (FQDAYNAAGGHNAVF) tetramer (kindly provided by NIH MHC Tetramer Core Facility, Emory University, Atlanta, GA, USA), washed and surface stained with HV500-conjugated anti-CD4 (clone RM4-5; BD Biosciences), and BV786-conjugated anti-CD44 (clone IM-7; BD Biosciences). Samples were labeled with Live/Dead Fixable Near-IR Stain Kit according to manufacturer's instructions (Invitrogen, USA). All antibodies and tetramers were titrated for optimal dilution. Approximately 5–10 × 105 cells were acquired on a LSRFortessaTM X-20 flow cytometer (BD Biosciences). Data analysis was performed using FlowJo v10 (TreeStar, USA).

### IgG ELISpot

Antibody-secreting cells (ASC) were evaluated in the spleen by ELISpot PLUS for mouse IgG kit (Mabtech) 10 days post boost. Multiscreen filter (PVDF) plates (Millipore, USA) were pre-wet for 2 min with EtOH 70%, washed with sterile dH2O and coated with H56 (5 µg/ml) diluted in PBS (Sigma-Aldrich) for the detection of antigen-specific IgG. After incubation overnight at 4°C, the coated wells were washed with sterile PBS and blocked for 30 min with serum free CTL-Test B culture medium (CTL, USA) supplemented with 1% l-glutamine (Sigma-Aldrich, USA). After removal of the blocking medium, 1 × 106 cells/well were added in a volume of 100 µl of CTL-Test B medium for the analysis of H56-specific IgG ASC. Each sample was assayed in triplicate, and the plates were incubated overnight for 20 h at 37°C in a humidified atmosphere with 5% CO2. After incubation, cells were removed by washing with PBS and 100 µl/well of anti-IgG biotinylated detection antibody (Mabtech, Sweden), diluted to 1 µg/ml in PBS containing 0.5% fetal calf serum (FCS, Gibco), was added. After incubation for 2 h at room temperature followed by washing steps, ELISpot plates were incubated with 50 µl/ well of streptavidin–horseradish peroxidase (Mabtech) diluted 1:1,000 for 1 h at room temperature. Plates were washed again with PBS, and then 100 µl/well TMB substrate solution (Mabtech) was added for approximately 10 min. The reaction was stopped by extensive washing in dH2O, and plates were then dried in the dark at room temperature. The number of spots was determined by plate scanning and analysis services performed at Cellular Technology Limited (Germany).

## Enzyme-Linked Immunosorbent Assay (ELISA)

Serum H56-specific IgG were quantified by ELISA at different time points. Flat bottomed Maxisorp microtiter plates (Nunc, Denmark) were coated with H56 (0.5 µg/ml) for 3 h at 37°C and overnight at 4°C in a volume of 100 μl/well. Plates were washed and blocked with 200 µl/well of PBS containing 1% BSA (Sigma-Aldrich) for 2 h at 37°C. Serum samples were added (100 μl/well) and titrated in twofold dilution in duplicate in PBS supplemented with 0.05% Tween 20 and 0.1% BSA (diluent buffer). After incubation for 2 h at 37°C, samples were incubated with the alkaline phosphatase-conjugate goat anti-mouse IgG (100 μl/well diluted 1:1,000 in diluent buffer, Southern Biotechnology, USA) for 2 h at 37°C and developed by adding 200 μl/well of 1 mg/ml of alkaline phosphatase substrate (Sigma-Aldrich). The optical density was recorded using a Multiskan FC Microplate Photometer (Thermo Scientific). Antibody titers were expressed as the reciprocal of the highest dilution with an OD value ≥ 0.2 after background subtraction.

### Bioinformatic Analysis

Raw reads were aligned to mouse genome (UCSD mm10 annotation), and counts were generated using STAR (41). Differential gene expression was determined using edgeR (42) with the generalized linear model (GLM) fitting approach for both longitudinal analysis and pairwise adjuvanted *versus* non-adjuvanted vaccine comparisons at each time point. For each comparison, edgeR generated a list of genes associated with a log fold-change and to a false discovery rate (FDR, i.e., the *P* value after multiple test correction). Genes were considered significantly differentially expressed when FDR was <0.05. For enrichment analysis, lists of genes were sorted by FDR, then mouse gene identifiers were converted to human gene identifiers using an in house script utilizing the BiomaRt Bioconductor package. Enrichment analysis was performed using the tmod R package (43) with blood transcription modules (BTMs) developed by Li and coworkers (19). Significance of module enrichment was assessed using the CERNO statistical test (a modification of Fisher's combined probability test) and corrected for multiple testing using the Benjamini–Hochberg correction.

### Statistical Analysis

Analysis was performed on individual samples, and data were reported in box plots encompassing minimum and maximum values. The Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical differences among the number of (i) Ag85B-specific CD4+ T cells, (ii) germinal center B cells, and (iii) H56-specific ASC in different groups of mice (naïve, primed with antigen H56 alone or primed with H56 + CAF01). H56-specific IgG were reported in graphs as log-transformed geometric mean titers with 95% confidence intervals and compared using *t*-test. Statistical significance was defined as *P* < 0.05, and analyses were performed using Graph Pad Prism version 7 (GraphPad Software, USA).

# RESULTS

# Antigen-Specific CD4**+** T-Cell Response

Antigen-specific CD4+ T cell responses were analyzed in mice primed with the chimeric TB vaccine antigen H56 administered alone or in combination with the liposome system CAF01, and boosted after 4 weeks with H56 antigen alone (**Figure 1**). T cell responses were characterized 10 days after boosting (day 38) within the local dLN. CD4+ T cells specific for the immunodominant epitope Ag85B, part of the H56 fusion protein, were identified using Ag85B280–294-complexed MHC class II tetramers. Staining specificity of Ag85B280–294-complexed MHC class II tetramers was determined using a control tetramer complexed with an unrelated antigen, which showed a level of staining below 0.02% (data not shown). Ten days after booster immunization with H56 antigen, the number of tetramer-binding CD4+ T cells in dLN was about 66,000 (±39,000) in mice primed with H56 + CAF01. This was significantly higher than the amount of antigen-specific CD4+ T cells present in mice primed with the H56 antigen alone (8,260 ± 6,000, *P* < 0.01, **Figure 2A**). The percentage of antigen-specific CD4+ T cells in respect to total activated CD4+ T cells was 0.55% and 0.07% in mice primed with H56 + CAF01 and H56 antigen alone, respectively (*P* < 0.05, data not shown). These data show that inclusion of CAF01 in the priming formulation induces an antigen-specific proliferation of CD4+ T cells which is not observed with the antigen alone.

### Characterization of the B-Cell Response

B cells responses were characterized 10 days after boosting (day 38) within the local dLN and spleen. Similar to the CD4+ T recall response, the reactivation of the B cell response was significantly higher in mice that had been primed with H56 + CAF01 compared with H56-primed mice (**Figure 2B**). Indeed, a significant expansion of the B220+ GL7+ CD95+ germinal center B cells (GC-B cells) 10 days following booster immunization was measured in mice primed with the CAF01 adjuvant (*P* < 0.05 *versus* H56-primed mice, *P* < 0.01 *versus* naïve mice), while the amount of cells detected in mice primed with H56 alone was not significantly higher compared with naïve mice (**Figure 2B**). The observed GC-B cells response is clearly a recall response of cells elicited by the primary immunization, since we have previously demonstrated that primary immunization with H56 alone does not stimulate the germinal center reaction, while the CAF01 adjuvant is a strong promoter of this reaction (4).

The induction of antigen-specific IgG antibody responses was also assessed at different time points after priming and boosting (**Figure 2C**). The IgG response elicited by both H56 alone or combined with CAF01 was very similar at 12 days after priming, while a significantly higher response was observed at day 28 and

Figure 2 | Adaptive immune response following booster immunization. C57BL/6 mice were subcutaneously immunized as reported in Figure 1. dLN were collected 10 days after booster immunization and analyzed for the T and B recall responses. (A) Number of Ag-specific CD4+ T cells, identified as CD4+ CD44+ tetramer Ag85B-specific cells. Mean values ± SEM of five mice per group are reported. (B) Number of germinal center B cells, identified as GL7+ CD95+ among B220+ B cells in dLN. (C) H56-specific serum IgG titers determined on days 0, 14, 28, 38, 56, and 77 following priming, by ELISA. Antibody titers values are reported as GMT ± 95% CI of five mice per group. Arrow indicates day 28 that is the time of boosting. (D) H56-specific antibody-secreting cells (ASC) detected in spleens by ELISPOT assay. Kruskal–Wallis test was used to assess statistical differences between groups of mice primed with H56 + CAF01 *versus* H56 alone (\**P* ≤ 0.05, \*\**P* ≤ 0.01, and \*\*\**P* ≤ 0.001).

maintained 10 days after boosting in mice primed with antigen and adjuvant compared with animals primed with antigen alone (day 38, *P* < 0.001). This was further confirmed by assessing antigen-specific ASC in the spleen collected 10 days after booster immunization (**Figure 2D**). The number of H56-specific ASC was higher in mice primed with H56 + CAF01 compared with animals primed with H56 alone with mean values of 51 and 20, respectively.

### Blood Gene Expression Compared With Pre-Immunization and Pre-Boost Baselines

The molecular signature of the vaccine formulations was analyzed in mice primed with H56 alone or in combination with CAF01, and boosted with H56 antigen alone (**Figure 1**). The transcriptomic profile was followed over time (without sacrificing the animals), sequencing blood RNA samples collected at early time points (1, 2, and 7 days) after primary and secondary immunizations (4 samples/each time point). An average of 7.2 µg total RNA/sample (range 1.473–60.552 μg/sample, as quantified by fluorimetry) was obtained. The isolated RNA was generally of high quality, with an average RNA Integrity number of 9.05 (range 5.8–9.9), as measured by the Bioanalyzer, and a mean 260/280 absorbance ratio of 2.17.

Gene expression in the blood of mice primed either with H56 + CAF01 or H56 alone was first compared with gene expression in blood collected before immunization. In **Figure 3** the left column reports bivariate plots showing differences in gene expression between the H56 + CAF01 and H56 alone

Figure 3 | Comparisons of gene expression changes elicited by H56 and H56 + CAF01 vaccine formulations. Each bivariate plot illustrates the comparison at one time point, as indicated. For the post-prime time points, the comparisons are represented relative to the naïve baseline. For post-boost time points, the baseline is the pre-boost time point. Axes represent log 2 of baseline-normalized expression values. Each dot represents a gene, and colored lines indicate the density of dots on the graph. Only genes significantly regulated (FDR < 0.05) by at least one vaccine were included and plotted. Numbers indicate the number of genes included in areas indicated by solid black lines. Dashed red lines indicate 0 values (no change in gene expression). Functional analysis was done using Reactome database, and representative significantly enriched pathways are included in captions under each plot.

immunized mice at each time point. Differentially expressed genes were subjected to functional analysis using the Reactome database.1

At day 1 after priming, 39 genes were found to be upregulated, while 1,147 genes were downregulated by both formulations. The gene modulation was essentially limited to pathways of platelet function adhesion, activation and degranulation, and clotting, as expected for injection and repeated bleeding of mice. At day 2 after priming, 143 genes were upregulated, while 1,031 genes were downregulated. Among the upregulated genes there was an activation of pathways related to innate immunity, neutrophils degranulation and IL4-IL13 signaling. At day 7, 124 genes were upregulated, and 588 were downregulated by both vaccines. Overall, the transcriptomic response was very consistent between the two vaccine formulations.

Four weeks after priming, both groups of mice received a booster immunization with H56 antigen alone. The day before booster immunization, blood was collected from both groups of mice, and gene expression analysis performed at this time point was then used as a baseline. Gene expression in the blood of mice primed with H56 + CAF01 or H56 alone and boosted with H56 was indeed compared with gene expression in blood collected in the same group of mice before booster immunization. In the right column of **Figure 3** bivariate plots show differential gene expression at 1, 2, and 7 days after H56 boosting.

Among the genes upregulated 1 day after H56 boost in mice primed with H56 + CAF01, we observed an over representation of genes relevant to innate immune response, centered primarily on the interferon response, as well as IL4 and IL13 signaling. Importantly, these genes were regulated almost exclusively in the H56 + CAF01 primed group, but not in the group that received H56 alone as priming. Two days after boost, only a few genes were found to be differentially expressed, and, again, as at day 1 post boost, gene regulation was observed primarily in the H56 + CAF01 primed group, but not in the group of mice primed with the antigen alone. Interestingly, according to the Reactome analysis, the upregulated genes were enriched in *M. tuberculosis* response genes; however, due to the very small number of genes regulated at this time point, these results should be taken with caution. By 7 days after the boost, the two groups displayed a similar gene expression pattern, mainly related to genes involved in cell cycle and proliferation, suggesting the activation of an adaptive response and clonal expansion of responding antigen-specific cells (**Figure 3**, right column).

### Effect of CAF01 Adjuvant on Gene Expression

To identify the CAF01-specific transcriptomic response, the blood gene expression in mice primed with H56 + CAF01 was compared with that of mice primed with H56 alone. All comparisons were performed at 1, 2, and 7 days after priming and after H56 boost. **Table 1** reports the number of differentially expressed genes identified in edgeR by the genewise Negative Binomial Generalized Linear Models algorithm. Lists of genes, sorted by FDR, were converted to human IDs and tested for the significance in the enrichment of BTMs (19). BTMs include 346 sets of genes which are coordinately expressed and exert a specific function, described by the module title, including innate and adaptive immunity or general biological processes such as cell cycle, transcription or signal transduction. **Figure 4** reports modules that were found to be significantly enriched (FDR < 0.001) in at least one tested time point. Data analysis at days 1, 2, and 7 after priming indicated that the major differences between the two vaccine formulations were detected at day 2, with 2.2% of differentially expressed genes and upregulation of modules related to innate immunity by the CAF01 adjuvanted vaccine formulation. In particular, modules related to antigen-presenting cells (monocytes and DCs) were found to be enriched, together with neutrophils and TLR inflammatory signaling. Differential gene expression analysis found no significant difference between H56 + CAF01 and H56 alone vaccinated groups at day 1, and only 17 significant genes at day 7 (**Table 1**). Nevertheless, module enrichment analysis identified seven enriched modules at day 1 and one module at day 7. These modules were also related to innate immune response, indicating that CAF01 modulates the transcriptomic response also at those time points.

The most significant differences between the two vaccine formulations could be appreciated after antigen boost, thus demonstrating the key role of the boost in highlighting the skewing of the immune system induced by the adjuvant used for priming. In fact, at day 1 after boost, 34 out of the 44 significant modules were modulated by CAF01, with 8.1% of genes being differentially expressed (**Figure 4**; **Table 1**). In particular, the innate immunity compartment was found to be affected: 10 modules regarding monocytes and antigen presentation, three modules associated with neutrophil recruitment and six modules related to interferon response were upregulated, while one module associated with natural killer (NK) cells was downregulated. No enrichment of modules related to B- and T-cell populations was observed at this early time point. At day 2 after boost, the overall difference between the two formulations was lower: 13 modules were significantly enriched, mainly related to innate response with a minor modulation of B cells. Seven days after the boost,

<sup>1</sup>https://reactome.org/ (Accessed: November 20, 2017).


Figure 4 | Activation of blood transcription modules (BTMs) by the CAF01 adjuvant. Each column represents a pairwise comparison between blood RNA samples from mice primed with H56 + CAF01 *versus* H56 alone. Both groups were boosted after 4 weeks with H56 alone. Blood samples were collected 1, 2, and 7 days after priming and after boosting. Activation of modules was tested using the FDR-ranked lists of genes generated by edgeR generalized linear model fitting and applying the CERNO test. Rows indicate different BTMs, which were significantly (FDR < 0.001) activated in at least one time point. Each module is represented by a pie in which the proportion of significantly upregulated and downregulated genes is shown in red and blue, respectively. The gray portion of the pie represents genes that are not significantly differentially regulated. The significance of module activation is proportional to the intensity of the pie, while the effect size (area under the curve) is proportional to its size.

despite the relatively high number of differentially expressed genes (**Table 1**), only eight modules were affected (**Figure 4**). In particular, these were regarding adaptive immune response with three upregulated modules regarding T cell activation and one downregulated relevant to B cells activities.

# DISCUSSION

Adjuvants have been extensively employed in vaccinology to enhance the immune responses to the antigen and promote the strength and persistence of the resulting immunity. Transcriptomic profiling of the immune response induced by vaccine adjuvants can contribute understanding the mechanism of action of adjuvants and guide the rational design of prime-boost vaccination strategies. A systems biology approach has been recently applied to profile priming properties of different vaccine adjuvants in a preclinical model using genome wide microarrays (24) and to analyze the response to adjuvanted and unadjuvanted influenza vaccines in human studies (18, 19). A clinical study performed on young children vaccinated with TIV vaccine administered with or without MF-59 adjuvant demonstrated that the inclusion of the oil-in-water adjuvant results in a stronger transcriptional response at the early time point post vaccination, including a higher interferon response, and higher resulting HAI titers (22). However, little is understood about the way adjuvants program the immune system for the response to re-exposure to the antigen.

In this study, we sought to investigate how the inclusion of the adjuvant in the vaccine formulation modifies the way the immune system responds to the same antigen, administered at a later time without the adjuvant, thus modeling the re-exposure scenario. We used mRNA sequencing to profile the mice immune response to the vaccine adjuvant CAF01 used in a primary immunization followed by a booster immunization with a low dose of H56 (0.5 μg/mouse) to select antigen-specific clones of T and B cells. While we found almost no difference in the transcriptomic profile after priming with or without the CAF01 adjuvant, strikingly, the effect of the CAF01 adjuvant, used in the primary immunization, could be readily appreciated after a booster immunization with the vaccine antigen alone. Indeed, we observed a significantly stronger upregulation of multiple genes and modules related to the innate immune response following boost in mice that received the adjuvant with the primary immunization compared with the mice that did not. This indicates the potent reprogramming of the immune response to re-exposure by the adjuvant included with the primary immunization. Transcriptomics of the vaccine immune response highlights therefore that priming with adjuvant modulates recall innate responses after boosting. Recent studies have reported the capacity of innate immune cells such as NK cells, monocytes and macrophages to mount a differential immune response upon a secondary contact with the same or distinct stimuli (44, 45). This concept of "memory" related to cells of the innate immune system is revolutionizing our knowledge of the immune system, and could represent an important goal for future vaccination strategies, based on the interaction between adjuvants and innate cells. Our analysis, performed on the whole blood level, does not permit us to dissect the underlying mechanisms of the observed differential response, nevertheless, a new study that will include sorting of the innate cell subpopulations and transcriptional profiling of the isolated populations, is currently in the development phase.

In this study, we extracted total RNA from small volumes of whole blood to make a longitudinal analysis of the transcriptomic response in the same animals bled multiple times during the course of the experiment. This study design and technical approach has been previously employed only in few studies (39, 40) and has proved the benefit of longitudinal observation within the same animal allowing analysis of gene modulation over time. An unexpected limitation introduced by this approach related to the fact that transcripts that we found to be regulated in response to the primary immunization were strongly enriched in genes relevant to platelet activation and aggregation, suggesting that the response to wounding may have masked the vaccinespecific response. Nevertheless we identified day 2 after priming and day 1 after boost as optimal time points to analyze the gene modulation induced by priming with CAF01 adjuvant compared with unadjuvanted vaccine formulation.

While we observed only a moderate differential expression of T cell activation modules 7 days after boost, the immunological analysis performed 10 days after boosting showed a higher frequency of H56-specific CD4+ T cells and germinal center B cells in dLN, a strong H56-specific humoral response and a higher frequency of ASC in spleen of mice primed with H56 + CAF01. We and others have previously shown that the presence of the CAF01 adjuvant combined with the antigen in the vaccine formulation used for priming is crucial for the induction of the germinal center reaction, assessed by the presence of the T follicular and germinal center B cells (4, 46), as well as for a strong T cell mediated immune response with a Th1 and Th17 profile (4, 6, 47).

The induction of germinal center reaction and the antigenspecific T cell effector function could only be detected in dLN or in the spleen, while the corresponding gene networks were not detected in whole blood. In fact, the transcriptome perturbation detected in peripheral blood is thought to reflect the overall response of the immune system to vaccine stimulation, even if it is not possible to detect finely tuned responses or phenomena localized in specific lymph nodes.

Taken together, our data suggest that the reprogramming of the recall immune response by the adjuvant used for priming impinges on both innate and adaptive arms resulting, upon boosting, in a stronger recall innate response essential for shaping the downstream adaptive response.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Italian national guidelines (Decreto Legislativo 26/2014). The protocol was approved by the Italian Ministry of Health with authorization no. 1004/2015-PR on September 22 2015.

### AUTHOR CONTRIBUTIONS

EP, DK, FS, AC, GP, and DM conceived and designed the experiments; EP, FF, and FS performed the experiments; GG and IE performed library preparation and sequencing; EP, DK, FS, FF, AC, GP, and DM analyzed the data; FS, EP, DM, and DK: wrote the paper; PA: provided reagents. All the authors read, critically revised, and approved the final manuscript.

### ACKNOWLEDGMENTS

We thank Dr. Gabiria Pastore for technical support for flow cytometry and Prof. Andrea De Maria for helpful discussion of the data. The authors acknowledge the NIH Tetramer Core Facility (contract HHSN272201300006C) for provision of MHC class II tetramers. This study has been carried out with financial support from the Commission of the European Communities, Seventh Framework Programme, contract HEALTH-2011-280873 "Advanced Immunization Technologies" (ADITEC). High throughput sequencing was performed by the

### REFERENCES


Norwegian Sequencing Centre (www.sequencing.uio.no), a national technology platform hosted by the University of Oslo and Oslo University Hospital, and supported by the "Functional Genomics" and "Infrastructure" programs of the Research Council of Norway and the Southeastern Regional Health Authorities.


**Conflict of Interest Statement:** PA is co-inventors on patent applications covering CAF01. As employee, PA has assigned all rights to Statens Serum Institut, a Danish non-profit governmental institute.

*Copyright © 2018 Santoro, Pettini, Kazmin, Ciabattini, Fiorino, Gilfillan, Evenroed, Andersen, Pozzi and Medaglini. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Birgit Weinberger <sup>1</sup> \*† , Mariëlle C. Haks 2†, Roelof A. de Paus <sup>2</sup> , Tom H. M. Ottenhoff <sup>2</sup> , Tanja Bauer <sup>3</sup> and Beatrix Grubeck-Loebenstein1*

*<sup>1</sup> Institute for Biomedical Aging Research, Universität Innsbruck, Innsbruck, Austria, 2Department of Infectious Diseases, Leiden University Medical Center, Leiden, Netherlands, 3 Institute of Virology, Technische Universität München/Helmholtz Zentrum München, Munich, Germany*

### *Edited by:*

*Peter Andersen, State Serum Institute (SSI), Denmark*

### *Reviewed by:*

*Anita S. Iyer, Harvard Medical School, United States Sarah Rowland-Jones, University of Oxford, United Kingdom*

> *\*Correspondence: Birgit Weinberger*

*birgit.weinberger@uibk.ac.at*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 13 December 2017 Accepted: 25 April 2018 Published: 15 May 2018*

### *Citation:*

*Weinberger B, Haks MC, de Paus RA, Ottenhoff THM, Bauer T and Grubeck-Loebenstein B (2018) Impaired Immune Response to Primary but Not to Booster Vaccination Against Hepatitis B in Older Adults. Front. Immunol. 9:1035. doi: 10.3389/fimmu.2018.01035*

Many current vaccines are less immunogenic and less effective in elderly compared to younger adults due to age-related changes of the immune system. Most vaccines utilized in the elderly contain antigens, which the target population has had previous contact with due to previous vaccination or infection. Therefore, most studies investigating vaccine-induced immune responses in the elderly do not analyze responses to neo-antigens but rather booster responses. However, age-related differences in the immune response could differentially affect primary versus recall responses. We therefore investigated the impact of age on primary and recall antibody responses following hepatitis B vaccination in young and older adults. Focused gene expression profiling was performed before and 1 day after the vaccination in order to identify gene signatures predicting antibody responses. Young (20–40 years; *n* = 24) and elderly (>60 years; *n* = 17) healthy volunteers received either a primary series (no prior vaccination) or a single booster shot (documented primary vaccination more than 10 years ago). Antibody titers were determined at days 0, 7, and 28, as well as 6 months after the vaccination. After primary vaccination, antibody responses were lower and delayed in the elderly compared to young adults. Non-responders after the three-dose primary series were only observed in the elderly group. Maximum antibody concentrations after booster vaccination were similar in both age groups. Focused gene expression profiling identified 29 transcripts that correlated with age at baseline and clustered in a network centered around type I interferons and pro-inflammatory cytokines. In addition, smaller 8- and 6-gene signatures were identified at baseline that associated with vaccine responsiveness during primary and booster vaccination, respectively. When evaluating the kinetic changes in gene expression profiles before and after primary vaccination, a 33-gene signature, dominated by IFN-signaling, pro-inflammatory cytokines, inflammasome components, and immune cell subset markers, was uncovered that was associated with vaccine responsiveness. By contrast, no such transcripts were identified during booster vaccination. Our results document that primary differs from booster vaccination in old age, in regard to antibody responses as well as at the level of gene signatures.

clinical Trial registration: www.clinicaltrialsregister.eu, this trial was registered at the EU Clinical Trial Register (EU-CTR) with the EUDRACT-Nr. 2013-002589-38.

Keywords: hepatitis B virus, vaccine, primary vaccination, booster vaccination, elderly, gene expression profiling

# INTRODUCTION

Life expectancy is increasing worldwide, and the number of persons older than 60 years of age is expected to double, reaching 2.1 billion by 2050 (1). The incidence and severity of many infectious diseases is high in the elderly compared to that in younger adults (2). Vaccination is one of the most effective measures to prevent infections, but most current vaccines are less immunogenic and less efficient in older adults. Age-related changes of the immune system include a decline of naïve T and B cells (3, 4), which potentially hampers immune responses to neo-antigens. It has been shown that primary immune responses to vaccines against tick-borne encephalitis (5), Japanese encephalitis (6), hepatitis A (7), and pandemic influenza strains (8) are lower in the elderly. Reduced immunogenicity in old age has also been shown for booster vaccinations against tetanus, diphtheria (9, 10), and tick-borne encephalitis (11). However, the molecular mechanisms underlying age-related hyporesponsiveness to vaccination remain unclear. Genome-wide RNA expression profiling has identified a clear association between chronological age and progressive changes in the transcriptional landscape of peripheral blood cells. Significant age-related changes were found in the transcript levels of markers involved in, e.g., immunosenescence, inflammation, and oxidative stress (12). Moreover, many of the identified genes were highly and preferentially expressed in naïve and memory T and B cells and may thus reflect age-related changes in immune function (13, 14). Hence, pre-immunization transcriptomic profiles and/or changes in gene expression patterns in blood, resulting from the elicited innate and adaptive immune responses after vaccination, could potentially be used as biomarkers to classify and predict vaccine responsiveness and be key to a better understanding of (hypo)responsiveness to vaccination in the elderly population.

Immune responses after influenza and pneumococcal vaccination—the most studied vaccines in the elderly—are usually a mixture of primary and recall responses, as natural contact with various influenza strains and pneumococcal serotypes is frequent, but vaccines might also contain neo-antigens. It is therefore rare that primary responses and purely vaccine-induced recall responses (without natural exposure) to the same antigens are investigated in the elderly. We have chosen hepatitis B virus surface antigen (HBsAg) as a model antigen since natural exposure to hepatitis B virus (HBV) is relatively rare in Austria (15), and primary as well as booster vaccinations can be performed in young and older adults following national recommendations.

In addition to the value of HBsAg as a model antigen, HBV is also of clinical relevance for the older population. Acute infection with HBV is mainly recognized in young adults with high-risk behaviors, but is also relevant in old age (16). Older adults with viral hepatitis have a higher mortality rate than younger patients, which can be partially explained by underlying comorbidities, but also by a diminished immune response, metabolic and nutritional deficiencies, and age-related anatomic changes of the liver (17–19). Progression of acute HBV infection to chronicity occurs in less than 5% of young adults, but is observed more frequently in the elderly. In an outbreak in a nursing home in Japan, almost 60% of the infected elderly became HBsAg carriers (20). The prevalence of HBsAg (>6 months), which indicates chronic infection, is higher in nursing home residents compared with non-institutionalized populations, and the risk of transmission is higher in these facilities (21).

Childhood vaccination against hepatitis B is recommended in many countries. Most European countries and the USA also recommend primary vaccination of adults and booster vaccinations for persons with an increased risk for infection. In addition to persons with high-risk lifestyles (use of injected drugs, high-risk sexual behavior) or occupational risk factors (e.g., health-care personnel, police and other emergency service personnel, persons working with refugees), these risk groups include household contacts of chronically infected patients, patients with chronic liver disease, persons undergoing dialysis or regularly receiving plasma products, and persons under immunosuppressive therapy. All of these risk groups are also composed of older persons (22–27). In addition, vaccination is recommended for persons traveling to areas with a high prevalence of chronic HBV infection. Due to increased life expectancy as well as improved health status and mobility of elderly persons, the number of older long-distance travelers rises. The exact data on the extent of travel by older adults are limited, but data from several countries indicate that a substantial fraction of long-distance travelers are over the age of 65 (28, 29). Primary immunization against hepatitis B is performed relatively late in life, e.g., in travelers, household contacts of chronic hepatitis B carriers, or persons at risk in long-term care facilities. For this reason, the goal of this study was to compare the outcome of primary versus booster vaccination against hepatitis B in young and elderly persons from the same geographical and social background at the level of antibody response and gene signatures.

# MATERIALS AND METHODS

### Study Cohort

Two study cohorts of young and old healthy adults were recruited. Group A had never been vaccinated against hepatitis B and received a primary immunization series (three doses, 0–1–6 months). Serum was collected at days 0, 7, and 28 after each vaccination as well as 6 months after the last dose. Group B comprised persons who had received a full primary vaccination against hepatitis B at least 10 years ago and now received a single booster vaccination. Serum was collected at days 0, 7 and 28, and 6 months after the vaccination. Whole blood was collected in PAXgene blood RNA tubes (PreAnalytiX GmbH, Switzerland) before and 1 day after the first vaccination for expression analysis. Each group included young and old donors (**Table 1**). Persons with chronic viral infection (human immunodeficiency virus, hepatitis B virus, hepatitis C virus), transplant recipients, and patients under immunosuppressive or chemotherapy were not included in the study. Twinrix® (Glaxo Smith Kline, UK), which contains 720 ELISA units of inactivated hepatitis A virus and 20-µg recombinant HBsAg adjuvanted with Al(OH)3 and AlPO4, was used for all vaccinations. The objective of this study was to measure the hepatitis B-specific immune response. The protocol was approved by the ethics committee of the Innsbruck Medical

### Table 1 | Characteristics of study cohort.


*a One person was excluded from the analysis because HBsAg-specific antibodies were detectable at day 0 indicating previous contact with HBV.*

*bTwo persons were excluded from the analysis because HBsAg-specific antibodies* 

*were detectable at day 0 indicating previous contact with HBV.*

*c CMV serology was not available for one person.*

University. All participants gave their written informed consent in accordance with the Declaration of Helsinki. This trial was registered at the EU Clinical Trial Register (EU-CTR) with the EUDRACT-Nr. 2013-002589-38.

### Detection of HBsAg-Specific and Cytomegalovirus (CMV)-Specific Antibodies

Serum levels of HBsAg-specific antibodies (anti-HBs) were quantified using the ARCHITECT® chemiluminescence microparticle immunoassay (Abbott Diagnostics, Wiesbaden, Germany). Samples with anti-HBs levels below the limit of detection (0.9 IU/l) were set to 0.45 IU/l and samples with antibody levels above the dynamic range of the assay (10,000 IU/l) were diluted and re-tested. Serum levels of CMV-specific antibodies were quantified by ELISA using the Serion ELISA classic CMVs IgG Kit (Virion/Serion GmbH, Würzburg, Germany). Individuals with antibody concentrations below 25 PEI-U/ml are considered to be seronegative and values above 40 PEI-U/ml indicate seropositive individuals. Values between 25 and 40 PEI-U/ml are considered borderline, but were not seen in this study.

### RNA Isolation

Total RNA from venipuncture PAXgene blood collection tubes was extracted using the PAXgene Blood mRNA kit (PreAnalytiX GmbH, Switzerland) according to the manufacturer's protocol. The RNA yield was determined by a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), while the quality and integrity of the RNA were assessed on an Agilent 2100 BioAnalyzer (Agilent Technologies, Amstelveen, The Netherlands) using the RNA 6000 Nano Chip kit. The average RNA integrity number of the total RNA samples obtained from PAXgene tubes was 9.6 ± 0.05.

## Dual-Color Reverse-Transcription Multiplex Ligation-Dependent Probe Amplification (dcRT-MLPA)

Focused gene expression profiling using dcRT-MLPA was performed in triplicate as described in detail elsewhere (30). Briefly, for each target-specific sequence, a specific reverse transcription (RT) primer was designed located immediately downstream of the left- and right-hand half-probe target sequence. RNA was reverse transcribed using an RT-primer mix and MMLV reverse transcriptase (Promega Benelux, Leiden, The Netherlands). Transcriptase activity was inactivated by heating at 98°C for 2 min. Following RT, the left- and right-hand half-probes were hybridized to the cDNA at 60°C overnight. Annealed half-probes were ligated using ligase 65 and subsequently amplified by PCR (33 cycles of 30 s/95°C, 30 s/58°C, 60 s/72°C, followed by 1 cycle of 20 min/72°C). Primers and probes were from Sigma-Aldrich Chemie (Zwijndrecht, The Netherlands) and MLPA reagents from MRC-Holland (Amsterdam, The Netherlands). PCR amplification products were diluted 1:10 in HiDi-formamide containing 400HD ROX size standard, denatured at 95°C for 5 min, ice-cooled, and analyzed on an Applied Biosystems 3730 capillary sequencer in GeneScan mode (BaseClear, Leiden, The Netherlands).

Reverse transcription primers and half-probes were designed by Leiden University Medical Center (Department of Infectious Diseases, Leiden, The Netherlands) (30) and comprised sequences for four housekeeping genes (including GAPDH) and 144 selected genes to profile innate, adaptive, and inflammatory immune responses (Table S1 in Supplementary Material).

Trace data were analyzed using GeneMapper software 5 package (Applied Biosystems, Bleiswijk, The Netherlands). The areas of each assigned peak (in arbitrary units) were exported for further analysis in Microsoft Excel. Data were normalized to GAPDH, and signals below the threshold value for noise cutoff in GeneMapper (log2-transformed peak area of 7.64) were assigned the threshold value. Finally, the normalized data were log2-transformed for statistical analysis.

RNA expression values were visualized by principal component analysis (PCA) plots and heatmaps using the hierarchical clustering algorithms of Clustvis.1 String network analysis2 or Ingenuity Pathway Analysis of genes that were differentially expressed was performed to identify relevant signaling pathways and regulatory networks.

### Statistical Analysis

Statistical analyses of antibody concentrations for the two age groups were performed for each time point using the nonparametric Mann–Whitney *U*-test. Triplicate measurements of log2-transformed dcRT-MLPA-derived gene expression data were analyzed for statistical significance between two populations (young versus old adults or persons with ≥10,000 IU/l versus <10,000 IU/l anti-HBs) using the Mann–Whitney *U*-test. Statistical tests were two-sided, and *p*-values were adjusted for multiple testing using the Benjamini–Hochberg correction. Findings were regarded positive when meeting the criterion on the false discovery rate (FDR) of <10% and an optional factorial change (FC) of 1.25. Since relatively few samples were included in our study, leave-one-out cross-validation (LOOCV) was performed in the PCA analysis to estimate the generalization ability of the gene signatures as classifiers, and Wilks' Lambda test was used to test for differences between groups. The chi-squared test was used to test the independence of the two categorical variables' age and vaccine responsiveness. Significant differences were

<sup>1</sup>http://biit.cs.ut.ee/clustvis/ (Accessed: November 26, 2017).

<sup>2</sup>https://string-db.org/ (Accessed: December 5, 2017).

evaluated using the Fisher's exact test. Statistical computation was performed using SPSS 23 software.

### RESULTS

### Antibody Responses After Primary Vaccination

Young and old healthy adult volunteers who had never received a hepatitis A or B vaccination were recruited to receive a primary vaccination series (months 0–1–6) with a combined hepatitis A/B vaccine. HBsAg-specific antibody concentrations were measured before, 1 week, and 4 weeks after each dose, as well as 6 months after the last dose. None of the old participants mounted antibody responses after the first dose of the vaccine, whereas low levels of HBsAg-specific IgG were detectable in 40% of the young group, 4 weeks after the first dose (range 5.70–18.75 IU/l). After the second dose, antibody concentrations increased in all but one of the young participants (median 28.8-fold, range 1.8–418.8), but only in two (29%) of the older vaccines (40.3- and 68.5 fold, respectively). As expected, a further increase in antibody concentrations was observed after the third vaccination in both the young (median 31.2-fold, range 12.2–6,052.6) and the old group (median 29.3-fold, range 1–268.2). The median drop of antibody concentrations was 2.8-fold (range 0.7–9.1) within the first 6 months after the last vaccination with no age-related differences (Mann–Whitney *U*-test; *p* = 0.833). **Figure 1A** depicts the geometric mean antibody concentrations for both age groups.

Two of the old participants (29%) did not develop detectable antibodies (<0.9 IU/l) at any time point following primary vaccination and were classified as non-responders (NRs), whereas antibody concentrations were above 100 IU/l for all young participants. **Figure 1B** depicts the geometric mean antibody concentrations for the young (same as in **Figure 1A**), as well as the old responders (R) and old NR separately. When only the old responders were considered, their antibody concentrations were still lower than in the young group. This difference was only significant at week 8, probably due to the relatively low sample size.

### Antibody Responses After Booster Vaccination

Young and old healthy adults, who had received a complete primary series of hepatitis vaccination more than 10 years ago, received one dose of hepatitis A/B vaccine. Anti-HBs concentrations were measured before, 1 week, 4 weeks, and 6 months after the vaccination. Antibody concentrations before the booster vaccination ranged from below the limit of detection (<0.9 IU/l) to 3270 IU/l and were similar for both age groups (young versus old, *p* = 0.910, Mann–Whitney *U*-test). Within the first 4 weeks after vaccination, anti-HBs concentrations increased in young (median 469.4-fold, range 1–1,120.5) and old (median 328.1 fold, range 1–2,325.6) participants. No age-related differences were seen at any time point (young versus old, *p* = 0.678–1.000, Mann–Whitney *U*-test). The median decline of antibodies from weeks 4 to 26 was 4.4-fold (range 1.4–14.9). **Figure 2A** depicts the geometric mean antibody concentrations for all participants.

Anti-HBs concentrations were below 10 IU/l in four young (33%) and four old (50%) participants at the time of enrollment. In three of the four young individuals, no anti-HBs were detectable at all. Primary vaccination had been documented for all individuals, but anti-HBs concentrations after the primary series were not available. **Figures 2B,C** depict the antibody concentrations for these individuals and show the geometric mean titers for young and old participants with anti-HBs concentrations of ≥10 IU/l before the booster vaccination in comparison. Two of the young individuals (young-1 and young-2) did not develop adequate antibody concentrations after the single booster shot. The primary vaccination series of these persons was performed in early childhood and therefore dated back 30 and 33 years, respectively. By contrast, the other individuals with low anti-HBs concentrations before the vaccination (young-3, young-4, and old-1–4) showed an anamnestic response to the booster dose and reached sufficient antibody concentrations. Their last hepatitis B vaccination dated back 10–13 years. A strong correlation of anti-HBs concentrations before (w0) and after the booster vaccination (w4) was observed for young (*rS* = 0.875; *p* < 0.001) and old (*rS* = 0.786; *p* = 0.021) vaccines.

Figure 1 | HBsAg-specific antibody concentrations after primary vaccination series. Geometric mean concentrations and 95% confidence intervals of anti-HBs are depicted. Dashed vertical lines indicate the time points of vaccination (weeks 0–4–26). The limit of detection is 0.9 IU/l. Samples below this value were set to 0.45 IU/l. (A) This graph depicts data for young (blue; *n* = 11) and old (red; *n* = 7) participants. Mann–Whitney *U*-test young versus old; \**p* < 0.05; \*\**p* < 0.01. (B) This graph depicts data for young (blue; *n* = 11) participants, old responders (red; *n* = 5) and old non-responders (light red; *n* = 2). Mann–Whitney *U*-test young versus old responders; \**p* < 0.05.

shown for four young (light blue) participants, with low (<10 IU/l) anti-HBs at the time of enrollment. The numbers in brackets represent the time since the last vaccination (years) for the individual donors. (C) Geometric mean concentrations and 95% confidence intervals of anti-HBs are depicted for old (dark red; *n* = 4) participants with anti-HBs concentrations of ≥10 IU/l prior to the booster vaccination. Individual antibody concentrations are shown for four old (light red) participants, with low (<10 IU/l) anti-HBs at the time of enrollment. The numbers in brackets represent the time since the last vaccination (years) for the individual donors.

### Correlation Between Pre-Immunization Transcriptomic Profiles and Vaccine Response

Before exploring whether pre-immunization gene expression levels correlated with different responses to the hepatitis B vaccine between young and old adults, we first investigated the impact of age on the pre-immunization transcriptomic profiles. Mann–Whitney *U*-test was used to identify transcripts differentially expressed between young and old participants, and *p*-values were adjusted with the Benjamini–Hochberg method to correct for multiple testing. Using an FDR of <10%, we identified 29 transcripts whose pre-immunization expression levels correlated with age. This signature strongly clustered in a network centered around type I interferons (IFNα and IFNβ) and proinflammatory cytokines (IL-12, GM-CSF, and type II interferons TNFα and IFNγ) and was able to generally separate young and old adults' pre-immunization following hierarchical cluster analysis and PCA (**Figure 3**). Applying an FC cutoff of >1.25 reduced the 29-gene signature to a 10-gene signature with improved accuracy (LOOCV of 70%) and discriminatory power (*p*= 0.031) (Figure S1 in Supplementary Material). Our finding that pro-inflammatory pathways predominate in elderly persons is in agreement with the study by Fourati et al., where genome-wide RNA expression profiles were compared between 30 participants aged 25–40 years and 144 participants aged ≥65 in Canada, which validates the results observed in our smaller-sized cohort (31).

To investigate whether pre-immunization gene expression profiles correlated with different responses to the hepatitis B vaccine, participants were categorized independently of age, but based on their anti-HBs concentrations 4 weeks after the third dose of the primary series (week 30) or 4 weeks after the booster vaccination. Individuals reaching anti-HBs concentrations of ≥10,000 IU/l were considered as high responders. Based on this criterion, five out of eight (62.5%) young adults were considered high responders during primary vaccination while the large majority of old adults [six out of seven (85.7%)] did not reach anti-HBs concentrations of ≥10,000 IU/l, suggesting a strong correlation between chronological age and lower vaccine responsiveness during primary vaccination (chi-squared test: *p*= 0.036). By contrast, 6 out of 12 (50.0%) young adults and 3 out of 8 (37.5%) old adults were considered high responders during booster vaccination (chi-squared test: *p* = 0.582), corroborating the observation that the increase in anti-HBs concentration was only significantly different between the age groups after primary vaccination, but not after booster vaccination (**Figures 1** and **2**).

Subsequently, Mann–Whitney *U*-test was used to identify transcripts differentially expressed between individuals with anti-HBs concentrations below or above 10,000 IU/l independent of age, and *p*-values were adjusted with the Benjamini–Hochberg method to correct for multiple testing. Using an FDR of <10% and an FC of >1.25, we uncovered eight transcripts during primary vaccination and six transcripts during booster vaccination whose pre-immunization expression levels correlated with vaccine responsiveness (**Figures 4A,C**), and the identified gene signatures displayed an LOOCV accuracy of 81 and 72%, respectively (**Figures 4B,D**). In **Figure 5**, plots of the single genes encompassing the identified eight- and six-gene signatures are depicted. During primary vaccination, higher basal expression levels of CD8A, CD14, IFITM1/3, LYN, SEC14L1, and TNIP and lower basal transcriptomic levels of IL9 and TNF were associated with high anti-HBs concentrations (**Figure 5A**), while during booster vaccination, increased basal expression levels of BLR1, CCR7, CD19, CD274, CTLA4, and IL6 were correlated with higher anti-HBs concentrations (**Figure 5B**). Since BLR1, CCR7, CD19, CD274, and CTLA4 are all transcripts highly and preferentially expressed on T cells and/or B cells, these data suggest that during booster vaccination, vaccine responsiveness is mainly determined by differences in the representation and/or function of peripheral blood lymphocytes.

### Correlation Between the Kinetic Changes in Gene Expression Profiles and HBsAg-Specific Antibody Responses During Vaccination

Before exploring whether FCs in gene expression profiles before and after vaccination correlated with distinct responses to the hepatitis vaccine between young and old adults, we first investigated the impact of age on the FCs of the transcriptomic profiles between d0 and d1 during primary and booster vaccination. Using an FDR of <10%, we identified 16 and 11 transcripts during the primary and booster vaccination, respectively, that correlated with age and displayed an FC of >1.25 (**Figure 6**). While inflammasome markers, IFN-inducible genes, and genes encoding proinflammatory cytokines dominated the 16-gene signature that discriminated young and old adults during primary vaccination, the 11-gene signature that distinguished young and old adults during booster vaccination primarily encompassed genes coding pattern recognition receptors, immune cell subset markers as well as IFN-inducible genes. The capacity of the 16- and 11-gene signature to separate young and old adults is depicted by PCA analysis in **Figure 6**.

Next, to investigate whether FC in gene expression profiles before and after vaccination correlated with different responses to the hepatitis B vaccine, participants were again categorized into high responders (anti-HBs of ≥10,000 IU/l) and individuals not reaching these antibody levels, independent of age. FCs of transcriptomic profiles between d0 and d1 were calculated, and Mann–Whitney *U*-test was used to identify transcripts that were differentially regulated between the two groups independent of age. *P*-values were adjusted with the Benjamini–Hochberg method to correct for multiple testing. Using an FDR of <10% and an FC of >1.25, we uncovered 33 transcripts during primary vaccination that displayed an association with anti-HBs concentrations with excellent classifying capacity (**Figures 7A,B**). String network analysis revealed a network dominated by IFN-inducible genes, pro-inflammatory cytokines, inflammasome components, and immune cell subset markers (**Figure 7C**). In **Figure 7D**, plots of eight single genes that are part of the 33-gene signature are depicted, whose change in expression levels after vaccination not only correlated with anti-HBs concentrations after primary vaccination but also showed a significant correlation with chronological age (Mann–Whitney *U*-test: CD14 *p* < 0.001; NLRP1 *p* = 0.021; NLRP12 *p* = 0.042; RAB33A *p* = 0.044; TGFB1 *p*< 0.001; TNFRSF1A *p*< 0.001; TNFRSF1B *p*< 0.001; and TNIP1 *p* = 0.021). By contrast, none of the 144 transcripts displayed an association with anti-HB concentrations following booster vaccination. This finding is in concordance with the observation that the increase (fold-change) of anti-HBs concentrations after the booster vaccination (weeks 0–4) did not differ between high responders and individuals with anti-HBs of <10,000 IU/ml at week 4 (median 469.4-fold, range 72.5–743.8 versus median 467.1-fold, range 51.4–2325.6; *p*= 0.863, Mann–Whitney *U*-test). For this calculation, the two NRs were excluded.

# DISCUSSION

In most cases, older adults are not immunologically naïve regarding the vaccines they receive. The most frequently studied vaccines in the elderly are against influenza and pneumococcal disease. Natural exposure to various influenza strains and pneumococcal serotypes induces immunological memory; therefore, even the first vaccination against those pathogens does not trigger a pure primary immune response. Primary, as well as booster immune responses against various pathogens, has been studied in the elderly (5–11), but a direct comparison using the same antigen is rarely done. We used HBsAg as a model antigen to investigate the impact of age on primary and booster immune responses following immunization in the absence of natural exposure.

Blood samples were taken at several time points after each dose of vaccine in order to evaluate not only the maximal response but also the development of antibody concentrations over time. Despite small sample sizes, which are a limitation of this study, we could observe a number of age-related differences in the response to primary hepatitis B vaccination. Firstly, the antibody response to primary vaccination was delayed in the older group compared to that in the younger adults. No antibody responses were detectable in the older vaccines after the first vaccination, and only 30% of the older participants showed an increase in antibodies after the second dose. By contrast, all but one young participant had mounted protective antibody levels 4 weeks after the second dose. In order to achieve protection, three doses of vaccine are needed for the elderly. This finding is of particular relevance when hepatitis B vaccine is administered as a travel vaccine, as usually only two doses of vaccine are administered before the travel activity. An age-associated delay in primary immune responses has also been described for yellow fever vaccine, with only 50% seroprotection in the older compared to 77% in the younger age group at

transcripts whose pre-immunization expression levels correlated with vaccine responsiveness independent of age is represented using a blue to yellow to red color scale. Rows and columns correspond to the genes and the profiled samples, respectively. The vaccine responder groups are presented in colored squares above each sample. (A) It depicts the eight-gene signature identified after primary vaccination and (C) the six-gene signature identified after booster vaccination, respectively. (B,D) PCA analysis of the gene expression profile of the eight-gene signature identified during primary vaccination (B) and the six-gene signature identified during booster vaccination (D). Blue and red spheres represent individuals with anti-HBs concentrations of <10,000 and ≥10,000 IU/l, respectively.

expression levels of the individuals with anti-HBs concentrations of <10,000 IU/l (<) and ≥10,000 IU/l (≥), and the area in between is indicated in gray. In case basal gene expression levels are significantly increased in high responders compared to persons with anti-HBs concentrations of <10,000 IU/l, the blue shading indicates the area above the median gene expression of the high responders and above an anti-HBs concentration of 10,000 IU/l. The area below the median gene expression levels of the persons with anti-HBs concentrations of <10,000 IU/l and below the anti-HBs concentration of 10,000 IU/l is shown in red. By contrast, if basal gene expression levels are significantly decreased in high responders compared to persons with anti-HBs concentrations of <10,000 IU/l, the opposite areas of the vertical lines representing the median gene expression levels are tinted using a similar color coding. Blue and red spheres represent young and old adults, respectively.

day 10 after vaccination (32). Age has been reported to be a risk factor for being an NR to HBV vaccination in individual studies (33–36), as well as in meta-analyses (37, 38). In concordance with these studies, we observed NRs in the old, but not in the young age group. It has been shown previously that only 34% of an old and potentially frail (median age 72 years, range 57–95) cohort in a long-term care facility developed anti-HBs concentrations of >10 IU/ml after three doses of hepatitis B vaccination (39). The decline of antibody concentrations over time (weeks 30–52) was similar for both age groups.

By contrast, both young and older adults respond similarly to booster vaccination against HBV, and the decline of antibodies

Figure 7 | Correlation between factorial changes (FCs) of transcriptomic profiles following primary vaccination and vaccine responsiveness. FCs of transcriptomic profiles between d0 and d1 were calculated following primary vaccination independent of age. Mann–Whitney *U*-test was used to identify transcripts that were differentially regulated between individuals with anti-HBs concentrations of ≥10,000 and <10,000 IU/l independent of age, and *p*-values were adjusted with the Benjamini–Hochberg method to correct for multiple testing. Using an FDR of <10% and an FC of >1.25, 33 transcripts were found to be differentially expressed between the two groups. (A) The median-centered gene expression of the 33-gene signature is represented using a blue to yellow to red color scale. Rows and columns correspond to the genes and the profiled samples, respectively. The vaccine responder groups are presented in colored squares above each sample. (B) PCA analysis of the gene expression profile of the 33-gene signature. Blue and red spheres represent individuals with anti-HBs concentrations of <10,000 and ≥10,000 IU/l, respectively. (C) String network analysis of the proteins represented in the 33-gene signature. Individual proteins are displayed as nodes. Lines represent protein–protein interactions, and the thickness of the lines indicates confidence. (D) Plots of the eight single genes that showed both a correlation between FCs of transcriptomic expression levels before and after vaccination (d0 and d1) (log2-transformed) and anti-HBs concentrations at week 30 (log10 transformed) following primary vaccination (4 weeks after the third dose) as well as a correlation with age. The horizontal line indicates an anti-HBs concentration of 10,000 IU/l used as a cutoff value to categorize participants, and the area above the cutoff value (>10,000 IU/l) is highlighted in gray. Blue and red spheres represent young and old adults, respectively.

over 6 months did not show age-related differences. We have previously reported that recall responses to other vaccine antigens are lower in the elderly compared to younger adults (9, 11), but this was not the case in the current study. A prerequisite for successful booster vaccination is adequate priming. It can be speculated that primary vaccination (e.g., against tetanus and diphtheria), which is administered in early childhood, might not have been complete for all older adults contributing to low responses to booster vaccination in old age. All participants of the current study had completed a three-dose primary series at least 10 years prior to enrollment and at least for the older cohort, primary vaccination had been administered during adulthood. Studies indicate that immunological memory remains intact for at least 20 years among healthy vaccinated individuals who initiated hepatitis B vaccination >6 months of age (40). Even after the loss of anti-HBs, T cells may still confer protection (41). In a long-term follow-up of two clinical trials performed in the early 1990s, almost all participants still had detectable anti-HBs 20 years after primary vaccination (42). It has been reported that approximately 20% of health-care workers did not possess detectable HBsAg-specific antibodies more than 10 years after the primary vaccination. Booster vaccination led to an increase of antibody concentrations in almost all participants (43). Notably, all participants of these two studies received the primary vaccination series as adults. Primary vaccination of adolescents (12–15 years) resulted in 80% of the vaccines retaining specific antibodies 15 years later. Upon booster vaccination of 19 individuals without residual antibodies, all but one mounted a robust anamnestic response (44). By contrast, several studies showed that only 20–30% of adolescents who received primary vaccination in infancy still had protective anti-HBs levels around age 18 (45–47). The age when primary vaccination is received appears to play an important role in the long-term maintenance of anti-HBs. Among vaccinated cohorts who initiated hepatitis B vaccination at birth, long-term follow-up studies are ongoing to determine the duration of vaccine-induced immunity (40). Leuridan and Van Damme have summarized several studies from different countries in which booster vaccination was administered to individuals who had received primary vaccination in infancy, but had antibody concentrations below 10 IU/l several years later. Anamnestic responses to the booster vaccination were less frequent with increasing time since the primary vaccination and ranged from 100%, 5 years after the primary vaccination to 75.6% after 18–23 years (48). In our small cohort, two young participants had received primary vaccination in infancy more than 30 years ago. Both individuals did not have any residual antibodies and they did not mount sufficient antibody responses after the booster vaccination. No information was available on their anti-HBs responses after the primary vaccination. Therefore, we cannot rule out the possibility that the NRs never had sufficient antibody responses after primary vaccination, as it has been reported that approx. 5% of vaccinated children do not develop antibodies after primary vaccination (49). By contrast, the individuals with anti-HBs concentrations of <10 IU/ml before the booster, who had received primary vaccination as adults, responded adequately to the booster vaccination. It can be speculated that the rate of successful booster vaccination declines with increasing time since the primary vaccination and that age at the time of primary vaccination also plays a role. Further studies in larger cohorts would be necessary to determine optimal booster schedules, which take into account the age at the time of primary vaccination and the response to this primary series as well as residual antibodies and time since the last vaccination.

Latent infection with CMV has a substantial impact on the composition of the T cell pool (50–53), and an impact of CMV seropositivity on vaccine-induced antibody responses has been suggested (54–56), but was not confirmed in other studies (57, 58). CMV serostatus was determined for the participants of the current study, but the sample size was not sufficient to reliably determine the impact of CMV on primary and booster antibody responses in this cohort.

Because the molecular mechanisms underlying age-related hyporesponsiveness to vaccination remain unclear, whole blood samples for focused gene expression profiling were collected before (d0) and after (d1) receiving the first dose of primary vaccine or booster vaccine. Prior to vaccination, significant age-related changes were found that strongly clustered in a network centered around type I interferons and pro-inflammatory cytokines, and these pathways have been implicated in promoting immunosenescence in older adults (**Figure 3**) (59). Our observation that pro-inflammatory pathways prevail in the elderly is consistent with a larger study in Canada, where genome-wide RNA expression profiles were compared between young and old adults (30). Importantly, several studies have shown that naïve T cells decrease, while highly differentiated effector and memory T cells accumulate with chronological age (60, 61). This change is not only reflected by an alteration in cell numbers but also reflected by a change in the expression levels of transcripts that are highly and preferentially expressed in T and B lymphocytes (13, 14). In agreement with this view, expression levels of T cell and B cell subset markers (CD3E, IL7R, CD19, BLR1) were altered in old adults compared to younger participants, likely reflecting a shift in the relative abundance of peripheral blood lymphocytes with aging. Moreover, the expression of PTPRCv1 (CD45RA) located on naïve T cells was higher in younger participants, while the expression of PTPRCv2 (CD45RO) located on memory T cells was higher in older participants. This finding is consistent with a progressive increase of memory T cells with chronological age and may therefore reflect age-related changes in immune function.

A correlation between pre-immunization expression levels and vaccine responsiveness during primary and booster vaccination could only be established for a few genes (**Figures 4** and **5**). However, during booster vaccination, five of the six genes encompassing a six-gene signature encoded transcripts that were highly and preferentially expressed on T cells and/or B cells (BLR1, CCR7, CD19, CD274, and CTLA4), indicating that specifically during booster vaccination, vaccine responsiveness is largely determined by variances in the representation and/or function of naïve and memory B cell and T cell subsets.

In contrast to pre-immunization transcriptomic profiles, FCs in gene expression profiles before and after primary vaccination identified a larger 33-gene signature that correlated with different responses to the HBV vaccine (**Figure 7**), revealing a network dominated by IFN-inducible genes, pro-inflammatory cytokines, inflammasome components, and immune cell subset markers. Type I interferons have been shown to enhance the development of CD4 and CD8 central memory T cells as well as CD4 and CD8 effector memory T cells (62) and to improve B cell function by amplifying the B cell receptor signal (63), providing a plausible explanation for the observed positive association between genes induced by type I interferons and antibody titers. By contrast, a negative association between the induction of regulatory T cell-associated genes (CD3E, IL7R, TGFB1) and antibody titers was found, speculating that in individuals with antibody concentrations of <10,000 IU/l, the accelerated induction/ recruitment of regulatory T cells compromises the magnitude and duration of inflammatory responses that are required for optimal antibody production (64). Another prominent group of genes identified in the network were inflammasome components. Reports about the role of these components during vaccination or B cell function are scarce, but it has been documented that the stimulation of NLRP3 by fungal antigens modulated IgM production (65). In addition, NLRP3<sup>−</sup>/<sup>−</sup> mice demonstrated a

### REFERENCES


decreased vaccine efficacy as measured by antibody production (66). Together, these data suggest that those signaling networks that change with progressive age, such as inflammation status, contribute to hyporesponsiveness in the elderly during primary vaccination. In contrast to primary vaccination, no correlation between FCs of transcriptomic profiles and vaccine responsiveness could be identified during booster vaccination. This finding is in concordance with the fact that the increase (fold-change) of anti-HBs concentrations after the booster vaccination did not differ between individuals with high and low antibody concentrations at week 4. Our results document that primary differs from booster vaccination in old age, regarding antibody responses as well as at the level of gene signatures.

# DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors upon request.

### ETHICS STATEMENT

The protocol was approved by the ethics committee of the Innsbruck Medical University. All participants gave their written informed consent in accordance with the Declaration of Helsinki. This trial was registered at the EU Clinical Trial Register (EU-CTR) with the EUDRACT-Nr. 2013-002589-38.

## AUTHOR CONTRIBUTIONS

BW, BG-L, MH, and TO designed the study. BW, MH, and TB performed experiments. BW, MH, and RP analyzed data. and BW, MH, TO, and BG-L wrote the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

### FUNDING

The research leading to these results has received funding from the European Union's Seventh Framework Programme [FP7/2007- 2013] under Grant Agreement No: 280873 ADITEC.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01035/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Weinberger, Haks, de Paus, Ottenhoff, Bauer and Grubeck-Loebenstein. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Annalisa Ciabattini1 \*, Elena Pettini1 , Fabio Fiorino1 , Simone Lucchesi1 , Gabiria Pastore1 , Jlenia Brunetti2 , Francesco Santoro1 , Peter Andersen3 , Luisa Bracci2 , Gianni Pozzi1 and Donata Medaglini1*

*<sup>1</sup> Laboratory of Molecular Microbiology and Biotechnology, Department of Medical Biotechnologies, University of Siena, Siena, Italy, 2U&E PreMed Laboratory, Department of Medical Biotechnologies, University of Siena, Siena, Italy, 3Department of Infectious Disease Immunology, Statens Serum Institute, Copenhagen, Denmark*

### *Edited by:*

*Fabio Bagnoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Evelina Angov, Walter Reed Army Institute of Research, United States Arun Kumar, Linköping University, Sweden*

> *\*Correspondence: Annalisa Ciabattini annalisa.ciabattini@unisi.it*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 30 November 2017 Accepted: 12 February 2018 Published: 12 March 2018*

### *Citation:*

*Ciabattini A, Pettini E, Fiorino F, Lucchesi S, Pastore G, Brunetti J, Santoro F, Andersen P, Bracci L, Pozzi G and Medaglini D (2018) Heterologous Prime-Boost Combinations Highlight the Crucial Role of Adjuvant in Priming the Immune System. Front. Immunol. 9:380. doi: 10.3389/fimmu.2018.00380*

The induction and modulation of the immune response to vaccination can be rationally designed by combining different vaccine formulations for priming and boosting. Here, we investigated the impact of heterologous prime-boost approaches on the vaccine-specific cellular and humoral responses specific for a mycobacterial vaccine antigen. C57BL/6 mice were primed with the chimeric vaccine antigen H56 administered alone or with the CAF01 adjuvant, and boosted with H56 alone, or combined with CAF01 or with the squalene-based oil-in-water emulsion adjuvant (o/w squalene). A strong secondary H56-specific CD4+ T cell response was recalled by all the booster vaccine formulations when mice had been primed with H56 and CAF01, but not with H56 alone. The polyfunctional nature of T helper cells was analyzed and visualized with the multidimensional flow cytometry FlowSOM software, implemented as a package of the R environment. A similar cytokine profile was detected in groups primed with H56 + CAF01 and boosted with or without adjuvant, except for some clusters of cells expressing high level of IL-17 together with TNF-α, IL-2, and IFN-γ, that were significantly upregulated only in groups boosted with the adjuvants. On the contrary, the comparison between groups primed with or without the adjuvant showed a completely different clusterization of cells, strengthening the impact of the formulation used for primary immunization on the profiling of responding cells. The presence of the CAF01 adjuvant in the priming formulation deeply affected also the secondary humoral response, especially in groups boosted with H56 alone or o/w squalene. In conclusion, the presence of CAF01 adjuvant in the primary immunization is crucial for promoting primary T and B cell responses that can be efficiently reactivated by booster immunization also performed with antigen alone.

Keywords: prime-boost regimens, adjuvants, computational flow cytometry, T cells, intracellular cytokines, CAF01, priming

# INTRODUCTION

A key aspect for the generation of efficacious vaccines is the optimization of vaccine schedules capable of eliciting the more adequate immune response for a specific pathogen, balancing between humoral and cell-mediated immunity. The design of prime-boost vaccine combinations based on the selection of the vaccine formulation, the dose, the route, and the intervals between doses is therefore of critical importance for optimally shaping the immune response.

With the exception of very few antigens, such as certain toxins, almost all the purified proteins used as vaccine antigens generally induce a modest antibody response with little or no T cell response (1). Adjuvants have proven to be key components in vaccines, providing danger signals and triggering a sufficient activation of the innate system. The presence of the adjuvant allows to enhance and appropriately skew the immune responses toward a vaccine antigen (1, 2) and promotes the induction of long-lived immunological memory and protection. Profiling the mode of action of different adjuvants is of critical importance for the rationale design of vaccination strategies (3–5). One of the immunological events that play a pivotal role in the generation of a vaccine-specific immune response is the primary activation of T helper cells, due to its close relationship with long-term humoral immunity and induction of protective antibodies (6). Antigen-specific T helper clonal expansion, differentiation, and dissemination toward distal sites are regulated by different factors, such as the route of the primary immunization, the dose of the antigen, and the vaccine formulations (7–15). We have demonstrated that an efficient antigen-specific CD4<sup>+</sup> T cell priming, generating cells capable of responding to booster immunization, is preferentially elicited by the subcutaneous, and not by the nasal route of immunization (3). Nevertheless, the nasal route can be efficiently used for booster immunization, when a local effector cellular response is aimed, since it promotes the recruitment of activated T cells into the lungs (3). Mathematical models can be used as a tool to estimate *in vivo* the probability of antigen-specific CD4<sup>+</sup> T cell expansion and dissemination upon immunization with adjuvanted vaccine formulations (16). Clonally expanded CD4<sup>+</sup> T cells exert the effector function producing cytokines. On the basis of the simultaneous expression of specific pattern of cytokines, Th cells are classified into functionally defined effector subpopulations. This fate is strongly affected by factors such as the local pro-inflammatory environment, the dose and the route of the vaccine used, and the adjuvant included in the vaccine formulation (17, 18).

Since the priming event impacts the type and quality of the induced immune response, we have recently characterized the mode of action of four different adjuvants, alum, a squalenebased oil-in-water emulsion (structurally similar to the licensed MF59 adjuvant), CpG ODN1826 (19), and the liposome system CAF01 (20), after a single immunization (4). Comparative analysis showed that CAF01 and o/w squalene were the strongest adjuvants capable of activating cellular response, with a Th1/Th2 and Th1/Th17 profile, respectively. O/w squalene rapidly induced the release of antigen-specific IgG in serum while CAF01 stimulated the germinal center (GC) reaction within the draining lymph nodes. A strong GC reaction was also observed in the presence of alum, even if an early humoral response was not detected. On the contrary, CpG ODN adjuvant elicited a rapid humoral response, but not a CD4<sup>+</sup> T cell activation and only a mild GC reaction, suggesting a T-independent activation of the B cell response, due to the direct stimulation of TLRs on B cells (21). With these information, rationale combination of adjuvants can be exploited for designing vaccination approaches capable of eliciting the most adequate immune response for a specific pathogen. The strategy of generating a toolbox of adjuvants, with a well-defined profile to shape the immune response, has also been recently identified as a key priority in vaccine research and development in Europe1 (22).

When many parameters are combined in a flow cytometric analysis for studying the phenotype, the effector function, and the polyfunctionality of activated cells, as is the case of the characterization of an immune response elicited by vaccination, classical two-dimensional scatter plots analysis cannot be sufficient for the multidimensional nature of the data. To overcome this problem, novel computational techniques have been developed in the recent years, and computational flow cytometry has become a novel discipline useful for providing a set of tools to analyze, visualize, and interpret large amounts of cell data in a more automated and unbiased way (23). FlowSOM is an advanced visualization technique in which more information are provided than in the traditional two-dimensional scatter plots (24). A self-organizing map (SOM) is an unsupervised technique for clustering and dimensionality reduction, in which a discretized representation of the input space is trained. With FlowSOM, cells are grouped into cell type clusters that are then represented in a lower-dimensional space. This approach allows to visualize in the same picture information regarding the frequency of cells coexpressing different markers, and to compare different groups.

In this work, we have assessed different prime-boost combinations, using the CAF01 and o/w squalene adjuvants, in order to dissect their role in shaping the secondary immune response to the chimeric vaccine antigen H56, a promising vaccine candidate against *Mycobacterium tuberculosis*, consisting of the antigens Ag85B fused to the 6-kDa early secretory antigenic target and the latency-associated protein Rv2660c (25), already tested in four phase I and II clinical trials.2 The analysis of the serum IgG strength of binding to the vaccine antigen performed by surface plasmon resonance, and the computational analysis of the polyfunctional nature of reactivated CD4<sup>+</sup> T cells, have been used to highligh the impact of the priming event in the induction of the adaptive immune response.

### MATERIALS AND METHODS

### Mice

Seven-week-old female C57BL/6 mice, purchased from Charles River (Lecco, Italy), were housed under specific pathogen-free conditions in the animal facility of the Laboratory of Molecular Microbiology and Biotechnology, Department of

<sup>1</sup>www.iprove-roadmap.eu.

<sup>2</sup>http://ClinicalTrials.gov.

Medical Biotechnologies at University of Siena. This study was carried out in accordance with national guidelines (Decreto Legislativo 26/2014). The protocol was approved by the Italian Ministry of Health (authorization no. 1004/2015-PR, 22 September 2015).

### Immunizations

Mice were immunized by the subcutaneous route at the base of the tail, with vaccine formulations including the chimeric tuberculosis vaccine antigen H56 (2 μg/mouse for priming, 0.5 μg/mouse for boosting; Statens Serum Institut, Denmark), administered alone or combined with the adjuvants CAF01 (250 µg dimethyldioctadecylammonium and 50 µg trehalose dibehenate/mouse; Statens Serum Institut, Denmark), or a squalene-based oil-in-water adjuvant [50 μl/mouse, sorbitan trioleate (0.5% w/v) in squalene oil (5% v/v), and Tween 80 (0.5% w/v) in sodium citrate buffer (10 mM, pH 6.5), Invivogen, USA]. Groups of five mice were primed with H56 or H56 + CAF01 and boosted with H56, H56 + CAF01, or H56 + o/w squalene, 4 weeks later, as reported in **Figure 1**. Formulations containing CAF01 were injected in a volume of 150 μl/mouse of Tris 10 mM, while formulations containing o/w squalene or H56 alone in a volume of 100 μl/mouse of PBS. Groups of mice were sacrificed 7 and 28 days after priming, and 3 or 10 days after boosting (**Figure 1**).

Figure 1 | Study design and sample collection. Groups of C57BL/6 mice were subcutaneously primed with H56 alone (H56) or combined with CAF01, and boosted at day 28 with H56 + CAF01, or H56 + o/w squalene or H56 alone. Blood samples were collected at day 0, 12, 28, 31, and 38 following priming; draining lymph nodes (dLN) and spleens (Spl) were collected at day 7, 28, 31, and 38. The abbreviations of the different prime-boost combinations are here reported and used in all the figures.

### Sample Collection and Cell Preparation

Blood samples were taken from the temporal plexus of mice 0, 12, and 28 days after priming and 3 and 10 days after boosting. Samples were incubated for 30 min at 37°C and centrifuged at 1,200 × *g* at 4°C for 10 min for collecting sera that were stored at −80°C. Draining lymph nodes (sub iliac, medial, and external) and spleens were collected 7 and 28 days after priming, and 3 or 10 days after boosting. Samples were mashed onto 70 µm nylon screens (Sefar Italia, Italy) and washed two times in complete RPMI medium [RPMI (Lonza, Belgium), 100 U/ml penicillin/ streptomycin, and 10% fetal bovine serum (Gibco, USA)]. Samples were treated with red blood cells lysis buffer (eBioscience, USA) and counted with cell counter (Bio-Rad).

### Multiparametric Flow Cytometric Analysis

Samples from draining lymph nodes (dLN) and spleens were incubated for 30 min at 4°C in Fc-blocking solution (cRPMI with 5 µg/ml of CD16/CD32 mAb [clone 93; eBioscience, USA]). Cells from dLN were stained for 1 h at RT with PE-conjugated I-A(b) *M. tuberculosis* Ag85B precursor 280–294 (FQDAYNAAGGHNAVF) tetramer (kindly provided by NIH MHC Tetramer Core Facility, Emory University, Atlanta, GA, USA), washed and surface stained with HV500-conjugated anti-CD4 (clone RM4-5; BD Biosciences) and BV786-conjugated anti-CD44 (clone IM-7; BD Biosciences). Samples were labeled with Live/Dead Fixable Near-IR Stain Kit according to the manufacturer instruction (Invitrogen, USA). Intracellular cytokine production was assessed on splenocytes cultured for 6 h in the presence of anti-CD28, anti-CD49d (both 2 µg/ml, eBioscience), and H56 protein (2 µg/ml). Unstimulated or PMA and ionomycin calcium salt (50 ng/ml and 1 µM respectively, Sigma-Aldrich) treated cells were used as negative and positive controls, respectively. Brefeldin A (BFA, 5 µg/ml, Sigma-Aldrich) and monensin solution (1×, eBioscience) were added to all samples for the last 4 h of incubation. Cells were washed twice in PBS and labeled with Live/Dead Fixable Yellow Stain Kit according to the manufacturer instruction (Invitrogen, USA). Fixation and permeabilization were performed using BD Cytofix/Cytoperm kit according to the manufacturer instruction (BD Biosciences) before Fc-blocking and stained with HV500-conjugated anti-CD4 (clone RM4-5; BD Biosciences), BV786-conjugated anti-CD44 (clone IM-7; BD Biosciences), PerCP Cy5.5-conjugated anti-IFN-γ (clone XMG1.2; BD Biosciences), AF700-conjugated anti-TNF-α (clone MP6-XT22; BD Biosciences), APC-conjugated anti-IL-17A (clone eBio17B7; eBioscience), AF488-conjugated anti-IL-4 (clone 11B11; eBioscience), AF488-conjugated anti-IL-13 (clone eBio13A; eBioscience). All antibodies and tetramers were titrated for optimal dilution. About 5–10 × 105 cells were stored for each sample, and acquired on BD LSRFortessa X20 flow cytometer (BD Biosciences). Data analysis was performed using FlowJo v10 (TreeStar, USA).

### Computation Analysis of Flow Cytometric Data with FlowSOM

Data from restimulated splenocytes were first analyzed with FlowJo. Live lymphocytes expressing CD4 and CD44 were manually gated, concatenated within the same immunization group, randomly downsampled to 15,000 cells and exported as uncompensated cells. Data were then compensated, logically transformed and scaled with FlowSOM (24). FlowSOM is a package of R, an opensource environment for statistical analysis, computation, and visualization,3 available as a Bioconductor package. Clustering analysis of data was performed following the FlowSOM function pipeline. The algorithm considers each cell in an *n*-dimensional space (where *n* = the number of cytokines considered for the analysis); cells with similar position in *n*-dimensional space are clustered together. After clustering, a SOM is built, where all clusters represent nodes and the nodes closely connected to each other resemble each other more than nodes that are only connected through a long path. The resulting clustering of the SOM is visualized in a minimal spanning tree. Three different sets of FCS files, named flowSet, were analyzed: (a) FCS files from the groups primed with H56 + CAF01 and boosted with H56, H56 + CAF01, and H56 + o/w squalene, (b) groups primed with H56 + CAF01 or H56 alone and boosted with H56 + CAF01, and (c) groups primed with H56 + CAF01 or H56 alone and boosted with H56 + o/w squalene. For each flowSet, a SOM (function "FlowSOM") was built. Functions "CountGroups" and "PlotGroups" were used to visualize the flowSets as minimal spanning tree, to color the nodes depending on the expression of five cytokines and to highlight changes higher than twofold in each node.

### Enzyme-Linked Immunosorbent Assay (ELISA)

Serum H56-specific IgG were determined by ELISA. Flat bottomed Maxisorp microtiter plates (Nunc, Denmark) were coated with H56 (0.5 µg/ml) for 3 h at 37°C and overnight at 4°C in a volume of 100 µl/well. Plates were washed and blocked with 200 µl/well of PBS containing 1% BSA (Sigma-Aldrich) for 2 h at 37°C. Serum samples were added and titrated in twofold dilution in duplicate in PBS supplemented with 0.05% Tween 20 and 0.1% BSA (diluent buffer) in 100 µl/well. After 2 h at 37°C, samples were incubated with the alkaline phosphatase-conjugate goat anti-mouse IgG (diluted 1:1,000 in diluent buffer, Southern Biotechnology, USA) for 2 h at 37°C in 100 µl/well and developed by adding 1 mg/ml of alkaline phosphatase substrate (Sigma-Aldrich) in 200 µl/well. The optical density was recorded using Multiskan FC Microplate Photometer (Thermo Scientific). Antibody titers were expressed as the reciprocal of the highest serum dilution with an OD value ≥0.2, after subtraction of background values measured with diluent buffer alone.

# Binding of Anti-H56 Antibodies by Surface Plasmon Resonances

H56 antigen was immobilized on CM4 sensor chip (GE Healthcare) following standard amine coupling procedures. Antigen diluted at the concentration of 25 µg/ml in sodium acetate pH 3.5 was injected for 300 s at the flow rate of 5 µl/min over the sensor chip surface, previously activated with a 1:1 mixture of EDC-NHS. After immobilization, ethanolamine-HCl was injected for 7 min over all the surface to block any remaining active site on sensor chip surface. A blank immobilization was performed for the reference flow cell. Sera were diluted 200-fold in HBS-EP+ (10 mM Hepes, 150 mM NaCl, 3.4 M EDTA, 0.05% polysorbate 20, pH 7.4) and injected for 180 s at a flow rate of 30 µl/min onto immobilized H56, and dissociation phase was allowed for 300 s. Surface regeneration was achieved with a 45-s pulse of 10 mM glycine pH 2.0 at the same flow rate. For antibody isotyping, serum samples diluted 1:200 in HBS-EP+ were injected for 180 s across the H56-immobilized surface at 30 µl/ min, allowing sample binding to the surface. Then anti-mouse IgM and anti-mouse IgG antibodies, respectively, diluted 1:250 and 1:500 in HBS-EP+ were sequentially injected for 120 s each at 30 µl/min. For anti-H56 monoclonal antibody binding, Hyb76-8 was diluted at different concentrations (40, 20, 10, 5, 2.5, 1, and 0.5 µg/ml) in HBS-EP+ and then injected for 180 s at the flow rate of 30 µl/min onto immobilized H56, and dissociation phase was allowed for 450 s. Surface regeneration was achieved with a 30-s pulse of 10 mM glycine pH 2.0 at 30 µl/min.

The experiments were performed on a Biacore T100 instrument (GE Healthcare). The actual binding response of samples (RU) was obtained by subtracting the background response, recorded by injecting the sample through the reference flow cell. Kinetic of anti-H56 mAb was analyzed with the "Biacore T100 evaluation 1.1.1" software using the 1:1 Langmuir model for fitting the curves. Dissociation rates were calculated by curvefitting analysis to a dissociation model.

### Statistical Analysis

Mann–Whitney test for multiple pairwise comparisons was used for assessing statistical difference between groups receiving the same booster and primed with H56 alone or H56 + CAF01 adjuvant. Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical difference among all groups. IgG titers were reported as geometric mean titers (GMT) with 95% CI, and statistical analysis was performed on log-transformed data. Statistical significance was defined as *P* ≤ 0.05. Analysis was performed using GraphPad Prism v7 (GraphPad Software, USA).

# RESULTS

## Primary Ag-Specific CD4**+** T Cell Expansion and Effector Function

Groups of mice were parenterally primed with the chimeric tuberculosis vaccine antigen H56 administered alone or combined with the liposome system CAF01. Four weeks later, mice were boosted with three different formulations: the H56 antigen alone, the H56 antigen combined with CAF01, or mixed with an oil-in-water (o/w) squalene-based emulsion, hereafter o/w squalene (**Figure 1**). The induction of antigen-specific CD4<sup>+</sup> T cell expansion into the draining iliac lymph nodes was assessed 7 and 28 days after priming, and 3 and 10 days after booster immunization. Antigen-specific CD4<sup>+</sup> T cells were identified using Ag85B280–294-complexed MHC class II tetramers, specific for the immunodominant epitope of Ag85B, that is part of the chimeric

<sup>3</sup>www.r-project.org.

H56 protein. Staining specificity was determined using a control tetramer complexed with an unrelated antigen that showed a level of staining below 0.02% (data not shown). Representative dot plots showing the frequencies of tetramer-positive (Tet<sup>+</sup>) T cells elicited by the different vaccine formulations 10 days after boosting are shown in **Figure 2A**. As clearly observed, all the groups primed with the vaccine formulation containing the CAF01 adjuvant, developed a recall response of the CD4<sup>+</sup> T cells significantly higher compared to groups primed with H56 antigen alone (**Figure 2B**, *P* ≤ 0.05). The frequency of antigen-specific CD4<sup>+</sup> T cells was higher in mice primed with H56 + CAF01 and boosted with H56 + o/w squalene versus all groups primed with H56 (*P* ≤ 0.05) while no significant differences were observed versus groups boosted with H56 alone or H56 + CAF01 (CAF/ H56 and CAF/CAF). The secondary response clearly reflects the reactivation of cells activated by the primary immunization with H56 + CAF01, observed 7 days after priming (**Figure 2B**).

Since most of T helper cells activated into the draining lymph nodes exit and recirculate, the effector function of H56-specific CD4<sup>+</sup> T cells was assessed in the spleen, analyzing the intracellular production of different cytokines using multicolor flow cytometry. Dot plots, representative of the different groups, showing the frequencies of H56-specific CD4<sup>+</sup> T cells producing TNF-α, IFN-γ, IL-17, or IL-4/IL-13 cytokines versus IL-2, a cytokine indicative of the proliferative response and activation program of antigen-specific T cells, are reported in **Figure 3A**. The analysis of the effector function clearly confirmed the importance of the priming event to elicit cells capable of reactivation upon antigen restimulation. Indeed, only groups of mice that had been primed with H56 + CAF01 were able to reactivate, upon boosting, cells co-expressing IL-2 with TNF-a (55%, 38%, and 31% in CAF/ SQL, CAF/H56, and CAF/CAF groups, respectively), with IFN-γ (47%, 32%, and 30%, respectively) or with IL-17 (9%, 3%, and 6%, respectively), with percentages of effector cells significantly higher compared to the respective groups primed with H56 (**Figure 3B**; Mann–Whitney test for multiple pairwise comparisons, *P* ≤ 0.05). Interestingly, the booster immunization with o/w squalene, an adjuvant capable of stimulating a primary Th1 and Th2 mixed response (4), increased the frequency of cells releasing IL-2 with IL4 and IL13 (7%; *P* ≤ 0.05 versus the respective group H56/ SQL), even though these cytokines were not observed following priming with H56 + CAF01 (**Figure 3B**). It was also clearly shown that mice primed with H56 alone did not respond to the booster immunization, also when CAF01 or o/w squalene adjuvants were used (**Figure 3B**).

# Computational Analysis of Intracellular Cytokines Production

In order to have a global picture of the polyfunctional profiles of T cells elicited by the different vaccine formulations, a computational analysis of data was performed using the multidimensional software FlowSOM. A minimal spanning tree was built connecting the nodes that were most similar to each other in minimal branches. Each cell was then classified to the nearest node that was coded as a pie chart with information about the expression of the five different cytokines (**Figure 4**). The mean fluorescence intensity (MFI) values of each cytokine are visualized inside each node. The height of each sector indicates the intensity, therefore when it reaches the border of the circle, this indicates that these cells have a high expression for that cytokine. The co-expression

Figure 2 | Ag85B-specific CD4+ T cell response. C57BL/6 mice were subcutaneously immunized as summarized in Figure 1. Draining lymph nodes were collected 7 and 28 days after priming, and 3 and 10 days after booster immunization and analyzed for the frequency of Ag-specific CD4<sup>+</sup> T cells, identified by staining with Ag85B-specific MHC class II tetramers (Tet-Ag85B). (A) Scatter plot of CD44 versus Tet-Ag85B, gated on live CD4+ lymphocytes, shown from a single animal representative of each immunization group, collected 10 days after boosting. Activated tetramer<sup>+</sup> cells are gated and the frequency reported. (B) Time course analysis of the frequencies of tetramer+ CD4+ T cells, detected in each group, reported as mean ± SEM of five mice per group. Mann–Whitney test for multiple pairwise comparisons was used for assessing statistical difference between groups receiving the same booster and primed with H56 alone or H56 + CAF01 adjuvant. Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical difference among all groups (*P* ≤ 0.05).

intensity of more cytokines by a group of cells clustered into the same node can be easily visualized. Moreover, the size of the nodes indicates the number of cells assigned to each node, so it is indicative of the frequency of activated CD4<sup>+</sup> T cells expressing that pattern of cytokines.

Trees obtained in the groups primed with H56 + CAF01 and boosted with H56 + CAF01 (CAF/CAF) or H56 + o/w squalene (CAF/SQL) are shown in **Figures 4A,B**, respectively. Forty nine nodes were created by the FlowSOM algorithm. The upper half of the trees shows nodes co-expressing more cytokines, in particular TNF-α, IL-2, and IFN-γ, with a high MFI. On the contrary, cells expressing very few cytokines with low MFI are shown in the lower half of the tree. When the frequency of cells within a node increased or decreased by at least twofold with respect to mice boosted with H56 alone, nodes were colored in blue or pink, respectively (**Figures 4A,B**). Mice boosted with CAF01 increased the frequency of cells expressing the pattern of cytokines shown in nodes 6, 7, 14 (upper half of the tree), and 10, 16, 41, 48 and 49 (lower half of the tree). Nodes 6, 7, and 14 include cells with a high expression of IL-17 together with TNF- α, IL-2 (node 7), and also IFN-γ (nodes 6 and 14) as shown in the magnification reported on the right of **Figure 4A**. The other nodes (10, 16, 41, 48, and 49) include cells with a weak expression of some cytokines, as indicated by the height of the sectors. At the same time, CAF01 induced the reduction of nodes 4, 11, and 33 (pink), which express TNF-α, IL-2 (nodes 11 and 33), and also IFN-γ (node 4). Interestingly, the cluster of cells co-expressing IL-17 (nodes 6, 7, and 14) was significantly increased also with o/w squalene with respect to H56 alone (**Figure 4B**). The amount of cells co-expressing TNF- α, IL-2, IFN-γ with an intermediate MFI of IL-4/IL-13 (nodes 1, 27, and 28) was also increased (**Figure 4B**). The frequency of cells clustered within single nodes for each immunization schedule is reported in **Figure 5**. Here, we can observe how many nodes the two plotted groups have in common, with a similar percentage of cells clustered inside (black dots), and how many nodes include an amount of cells ≥2-fold with respect to the compared immunization group (colored dots). In **Figure 5** panels A and B, the comparison of FlowSOM clustering between groups primed with H56 + CAF01 and boosted with adjuvanted formulations or H56 alone is reported, reflecting the differences shown in **Figure 4**. Panels C and D show the comparison between groups receiving different priming (H56 alone or with CAF01) but boosted with the same formulation (either H56 + CAF or H56 + o/w squalene). In this case, the amount of common nodes between the compared formulations drastically diminished while more than 40 nodes were differently clustered among compared groups (**Figures 5C,D**).

As a whole, we can conclude that there are some clusters of cells, mainly oriented toward the secretion of IL-17 together with other cytokines, that significantly increase due to the presence of the adjuvants in the booster immunization, while the rest of the nodes do not significantly change between different boosting formulations. On the contrary, a drastical change of most of the nodes is observed comparing groups receiving a priming with or without the adjuvant, confirming again that the priming setting defines the generation of cells capable of secondary reactivation. The overall CD4<sup>+</sup> T cell response analysis clearly underlines the different roles of the priming and boosting events with respect to the adaptive immune response.

### Humoral Immune Response

The induction of antigen-specific IgG antibody response was assessed at different time points after priming and boosting (**Figure 1**). The primary response elicited by both H56 alone or combined with CAF01 was very similar at 12 and 28 days after priming, while the secondary IgG response showed significant differences according to the formulation used for the booster immunization (**Figure 6**). Mice primed with H56 + CAF01 and boosted with H56 + o/w squalene, or with H56 alone developed the highest humoral response compared to all the other groups (*P* ≤ 0.05 of CAF/H56 versus H56/CAF). A significant difference was also detected between the two groups boosted with H56 alone but differently primed (*P* ≤ 0.05), indicating that a consistent recall antibody response can be elicited also by the antigen alone when an efficient priming had been performed.

Using surface plasmon resonance, we further compared sera collected after boosting for the capacity of binding to H56 immobilized on the sensor chip surface and data were expressed as resonance units (RU). The antibody isotype characterization indicated a complete absence of IgM in all the analyzed samples, thus allowing to correlate RU with IgG concentration (Figure S1 in Supplementary Material). Sensorgrams from single animals primed with or without the adjuvant and boosted with the same vaccine formulations showed higher binding responses to H56 in groups that had received a primary immunization with CAF01, compared to sera from mice primed with H56 alone (**Figure 7A**). The mean values of RU calculated for each immunization group are reported in **Figure 7B**. H56-specific IgG binding response in groups primed with H56 + CAF01 and boosted with H56 + o/w squalene or with H56 alone was of 732 and 700 RU respectively, while the group boosted with CAF01 showed a lower RU (190), even if it was significantly higher compared to the corresponding group primed with H56 alone (*P* ≤ 0.05). The dissociation kinetic rates (koff) calculated for each curve did not show significant differences among the differently immunized groups (Figure S2B in Supplementary Material) and were similar to the koff value of a monoclonal anti-H56 antibody (Hyb76-8) used as reference (Figure S2A in Supplementary Material).

The analysis of the humoral response again confirms the importance of the correct combination of vaccine formulations to be used for priming and boosting, and strengthens the importance of the presence of the CAF01 adjuvant in the formulation used for priming rather than for boosting, in line with the cellular data.

### DISCUSSION

Heterologous prime-boost combinations including two different vaccine adjuvants were used to study the role of the CAF01 adjuvant in generating primary immune response. The analysis of the secondary antigen-specific immune response upon booster immunization has been instrumental for evaluating (i) the impact of the priming and boosting events on the immune response to vaccination, (ii) the role of the adjuvant component in the

Figure 3 | Intracellular cytokines production. C57BL/6 mice were subcutaneously immunized as summarized in Figure 1. Spleens were collected 7 and 28 days after priming, and 3 and 10 days after boosting. Splenocytes were restimulated for 6 h with H56 protein. (A) Dot plots showing the production of TNF-α, IFN-γ, IL-17, IL-4/IL-13 versus IL-2 assessed on live CD4+ CD44+ lymphocytes in each group, collected 10 days after boosting. (B) Percentages of T cells positive for both IL-2 and the indicated cytokines, with respect to total CD4+ CD44+ cells, elicited by different vaccine formulations reported as mean ± SEM of five mice per group. Mann–Whitney test for multiple pairwise comparisons was used for assessing statistical difference between groups receiving the same booster and primed with H56 alone or H56 + CAF01 adjuvant. Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical difference among all groups (*P* ≤ 0.05).

primary and booster immunization, and (iii) the impact of the vaccine formulation on the functional profile of the cellular response elicited.

The prime-boost study reported here demonstrates that a primary immunization performed with the vaccine antigen and the CAF01 adjuvant deeply impacts the immune response

of the host, generating T and B cells capable of responding to recall immunization also when it is performed with the vaccine antigen alone. On the contrary, when the H56 vaccine antigen alone is used for priming, the secondary T immune response is completely abolished by the booster immunization, even in the presence of formulations including two potent adjuvants, such as CAF01 or o/w squalene, and the serum humoral response is much lower. The properties of these two adjuvants in eliciting an H56-specific primary immune response, have been previously characterized by our group (4). We demonstrated that the CAF01 adjuvant promotes a Th1 and Th17 primary immune response, stimulates the GC reaction inside the draining lymph nodes, and promotes a slower response in terms of early serum antibodies compared to other adjuvants, probably due to its mechanism of action that entraps the antigen slowing down its release. The o/w squalene adjuvant stimulates a primary cellular response characterized by the release of TNF-α, IFN-γ, and IL-4/IL-13 indicative of a mixed Th1/Th2 response, and elicits a rapid and significant humoral response in serum (4). The mixed combination of CAF01 for priming and o/w squalene for boosting tested here was extremely immunogenic both in terms of cellular and humoral response. A reactivation of the H56-specfic T helper response was elicited into the draining lymph nodes, and the effector function of reactivated cells reflected the profile elicited by the priming event, as shown by the production of TNF-α, IFN-γ, IL-2, and IL-17. Nevertheless, at the same time, there was a significant increase also of cells producing the cytokines IL-4 and IL-13, that were not elicited by the priming with CAF01, but were observed in mice primed with o/w squalene adjuvant (4). As shown by the computational analysis of cytokine production, cells expressing IL-4/IL-13 produced also TNF-α, IFN-γ, and IL-2. The analysis of the secondary immune response elicited in the group primed with H56 + CAF01 demonstrated that the use of the antigen alone for booster immunization was extremely efficient. In this case, there was a significant reactivation of the T helper response, also in terms of effector cells. We observed that antigen alone stimulates the reactivation of cells releasing TNF-α, IFN-γ, and IL-2, but not high levels of IL-17, that was observed only in groups boosted with each of the adjuvanted formulations.

The induction of cells secreting these cytokines is critical in the host immune response to *M. tuberculosis*. IFN-γ is crucial for the activation of macrophages, which in turn inhibit *M. tuberculosis* growth *via* induction of inducible isoform of nitric oxide synthase and autophagy (26), while TNF-α promotes the formation of mature granulomas and also activates infected

number of nodes is reported in each sector.

analyzed on day 0, 12, 28, 31, and 38 following priming, by enzyme-linked immunosorbent assay. Antibody titers are expressed as the reciprocal of the highest serum dilution with an OD value ≥0.2 after subtraction of background value (diluent buffer). Values are reported as geometric mean titers (GMT) ± 95% CI of 8–10 mice per group from two independent experiments. Mann-Whitney test for multiple pairwise comparisons was used for assessing statistical difference between groups receiving the same booster and primed with H56 alone or H56 + CAF01 adjuvant, # *P* ≤ 0.05. Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical difference between all groups. \**P* ≤ 0.05.

macrophages. IL-2 stimulates the expansion and maintenance of the T cell responses, therefore contributes to the host defense, and loss of IL-2-producing CD4<sup>+</sup> T cells is associated with loss of protection (27). The requirement for cells producing IL-17 to control the pathogen is less absolute, even if vaccine promoted Th17 cells can improve mycobacterial control in animal models, promoting early Th1 cell recruitment to the lung following aerosol *M. tuberculosis* infection and reduce bacterial burden (28, 29).

The H56-specific humoral response was induced by formulations containing the antigen alone or combined with adjuvants, with higher antibody titers observed in mice primed with H56 + CAF01 and boosted with H56 + o/w Squalene, or with H56 alone. The latter formulations also elicited antibodies with the highest binding properties for the vaccine antigen. The analysis performed in the groups primed with H56 alone showed that also in the presence of adjuvants in the booster immunization, the secondary antibody response was lower, with a very low binding capacity to the vaccine antigen. Previous studies have shown a complete absence of GC B and follicular T helper cells when the primary immunization was performed with antigen alone, while short-lived plasma cells were induced (4). Interaction of Tfh cells with B cells drives the GC reaction, a dynamic micro-anatomical structure that supports the generation of B-cell activation, antibody class switch recombination, and affinity maturation (30, 31). The lack of the GC reaction, together with the induction of short-lived plasma cells capable of secreting low affinity antibodies (32), can explain the quality and the quantity of the humoral secondary response observed in mice primed with H56 alone. To note, a completely opposite result in terms of vaccine antigen binding capacity was obtained in mice primed with H56 and boosted with the H56 + CAF01, versus mice primed with the H56 + CAF01 and then boosted with the H56 alone.

The strong difference between the impact of the priming and boosting event on the immune response was clearly visualized with the computational analysis of the cytokine profile of reactivated CD4<sup>+</sup> T cells. The use of software capable of managing the huge amount of data produced by flow cytometry for each cell has become a necessity. Multiparametric data can no longer be adequately analyzed using the classical, mostly manual, analysis techniques, in which two parameters are combined in twodimensional scatterplots, and therefore require the use of novel computational techniques (23). Among many softwares now available (33), we employed the FlowSOM software (9), which is a platform of analysis available as an open-source package for R, an open-source environment for statistical analysis, computation, and visualization. A SOM is an unsupervised technique for clustering and dimensionality reduction, in which a discretized representation of the input space is trained. The graphical output generated is helpful for visually displaying T cell polyfunctionality. This analysis clearly allowed to visualize a clusterization of cells producing different patterns of cytokines and to compare the polyfunctional activity of CD4<sup>+</sup> T cells elicited by the different prime-boost combinations. Six different clusters were observed among groups primed with the CAF01 adjuvant and boosted with the same formulation, or with the o/w squalene adjuvant or with the vaccine antigen alone. Both adjuvants used for boosting increased the amount of cells producing IL-17 together with TNF-α, IL-2, and IFN-γ, while only o/w squalene increased the frequency of cells co-expressing also IL-4 and IL-13. This analysis highlighted that also in the presence of the antigen alone we could

Figure 7 | Surface plasmon resonance analysis of H56-specific sera. A Biacore assay was performed on sera collected 10 days after booster immunization. Sera were diluted 200-fold and injected over H56 previously immobilized on CM4 sensor chip. (A) Sensorgrams of H56-specific antibodies binding (RU) from sera of single animals primed with the CAF01 adjuvant (blue lines) or with H6 alone (dashed lines). (B) Mean RU values ± SEM of five sera for group; filled bars indicate groups primed with CAF01 adjuvant, open bars with antigen alone. Mann–Whitney test for multiple pairwise comparisons was used for assessing statistical difference between groups receiving the same booster and primed with H56 alone or H56 + CAF01 adjuvant, # *P* ≤ 0.05. Kruskal–Wallis test, followed by Dunn's post test for multiple comparison, was used to assess statistical difference between all groups. \*\**P* ≤ 0.01.

reactivate a polyfunctional response, but the use of the adjuvant can be instrumental for modulating a specific type of effector cells. The number of clusters significantly different between groups primed with or without the CAF01 adjuvant, and boosted with the same formulation increased to 42, confirming again that the priming formulation defines the generation of cells capable of secondary reactivation.

The role of vaccine-induced polyfunctional CD4<sup>+</sup> T cells in the protection from *M. tuberculosis* infection is not completely clear, and results obtained in preclinical and clinical studies are sometimes contradictory (34). In the mouse model, the magnitude of polyfunctional CD4+ T cells often correlates with vaccine-induced protection, generally assessed as vaccine-induced control of bacterial replication following challenge, thus making polyfunctional T cells a good candidate for a mechanistic correlate of protection (34). It has been demonstrated that immunity elicited by H56 + CAF01 vaccination is associated with the maintenance of circulating polyfunctional CD4 T cells, that selectively home to the lung parenchyma, and confer durable protection (35). It is likely that cells expressing multiple effector functions may be more effective in controlling infection than those producing a single pro-inflammatory cytokine. Nevertheless, in humans, same cases have been reported in which there was a progress to disease, also in the presence of a strong Th1 responses (25, 28), and very little data on the role of IL-17 are available (36). Taken together, polyfunctional CD4<sup>+</sup> T cells deserve to be assessed and characterized, but other functional immunological components, including the humoral responses (37) should be analyzed and considered when assessing vaccine candidates to *M. tuberculosis* (38).

The data from the current study in mice emphasizes the role of adjuvant for the priming of an optimal immune response, whereas the adjuvant seems to be of less importance during the boosting. In particular, CAF01 seems to have the ability to prime a response that results in a qualitatively and quantitatively superior antibody response. Ongoing clinical trials with CAF01 adjuvanted priming vaccines will soon demonstrate if these data can be translated into humans. If we in the future should implement heterologous immunization protocols in humans will eventually rely not only on immunogenicity but also on a number of practical considerations of great importance for vaccine implementation.

In conclusion, the primary stimulation of the immune system is crucial for the generation of cells capable of a recall response. The choice of an immunogenic vaccine formulation for priming event paves the path for the subsequent secondary response, while the choice of the formulation for boosting can be a tool for modulating the quality, more than the quantity, of the secondary response. The heterologous prime-boost approach for vaccination appears as an excellent strategy for the generation of vaccines specifically designed for specific pathogens.

### ETHICS STATEMENT

This study was carried out in accordance with national guidelines (Decreto Legislativo 26/2014). The protocol was approved by the Italian Ministry of Health (authorization no. 1004/2015-PR, 22 September 2015).

# AUTHOR CONTRIBUTIONS

AC, EP, GPo, and DM contributed conception and design of the study; AC, EP, FF, and GPa performed the experiments; AC, EP, FF, and DM analyzed the data; SL, FS and AC performed computational analysis of flow cytometric data; JB performed and analyzed surface plasmon resonance data; PA provided reagents; AC wrote the manuscript; EP, FF, and JB wrote sections of the manuscript; GPo, DM, PA and LB critically revised the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

### REFERENCES


# ACKNOWLEDGMENTS

The authors acknowledge the NIH Tetramer Core Facility (contract HHSN272201300006C) for provision of MHC class II tetramers, and Prof. Yvan Saeys for the helpful suggestions on the use of the FlowSOM software. This study has been carried out with financial support from the Commission of the European Communities, Seventh Framework Programme, contract HEALTH-2011-280873 "Advanced Immunization Technologies" (ADITEC).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00380/ full#supplementary-material.

mucosal immunization. *Eur J Immunol* (2011) 41:2642–53. doi:10.1002/eji. 201041297


**Conflict of Interest Statement:** PA is a co-inventor on patent applications covering CAF01. As an employee, PA has assigned all rights to Statens Serum Institut, a Danish non-profit governmental institute. All other authors that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Ciabattini, Pettini, Fiorino, Lucchesi, Pastore, Brunetti, Santoro, Andersen, Bracci, Pozzi and Medaglini. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*, Floriane Auderset1*

*, Andreas Meinke6*

*,* 

*, Peter Andersen3*

*,* 

# Overcoming the neonatal limitations of inducing germinal centers through liposome-Based adjuvants including c-Type lectin agonists Trehalose Dibehenate or curdlan

### *Edited by:*

*Dennis Christensen3*

*Paul-Henri Lambert1*

*Fabio Bagnoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Michael Schotsaert, Icahn School of Medicine at Mount Sinai, United States Katie Louise Flanagan, Monash University, Australia*

*\*Correspondence:*

*Claire-Anne Siegrist claire-anne.siegrist@unige.ch*

*† These authors have contributed equally to this work.*

*‡ These two authors are co-authors.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 23 November 2017 Accepted: 12 February 2018 Published: 28 February 2018*

### *Citation:*

*Vono M, Eberhardt CS, Mohr E, Auderset F, Christensen D, Schmolke M, Coler R, Meinke A, Andersen P, Lambert P-H, Mastelic-Gavillet B and Siegrist C-A (2018) Overcoming the Neonatal Limitations of Inducing Germinal Centers through Liposome-Based Adjuvants Including C-Type Lectin Agonists Trehalose Dibehenate or Curdlan. Front. Immunol. 9:381. doi: 10.3389/fimmu.2018.00381*

*1WHO Collaborative Center for Vaccine Immunology, Department of Pathology-Immunology, University of Geneva, Geneva, Switzerland, 2WHO Collaborative Center for Vaccine Immunology, Department of Pediatrics, University of Geneva, Geneva, Switzerland, 3Vaccine Adjuvant Research, Department of Infectious Disease Immunology, Statens Serum Institut, Copenhagen, Denmark, 4Department of Microbiology and Molecular Medicine, University of Geneva, Geneva, Switzerland,* 

*<sup>5</sup> Infectious Disease Research Institute, Seattle, WA, United States, 6Valneva Austria GmbH, Vienna, Austria*

*, Rhea Coler5*

*, Beatris Mastelic-Gavillet1‡ and Claire-Anne Siegrist1,2\*‡*

*Maria Vono1†, Christiane Sigrid Eberhardt1,2†, Elodie Mohr1*

*, Mirco Schmolke4*

Neonates and infants are more vulnerable to infections and show reduced responses to vaccination. Consequently, repeated immunizations are required to induce protection and early life vaccines against major pathogens such as influenza are yet unavailable. Formulating antigens with potent adjuvants, including immunostimulators and delivery systems, is a demonstrated approach to enhance vaccine efficacy. Yet, adjuvants effective in adults may not meet the specific requirements for activating the early life immune system. Here, we assessed the neonatal adjuvanticity of three novel adjuvants including TLR4 (glucopyranosyl lipid adjuvant-squalene emulsion), TLR9 (IC31®), and Mincle (CAF01) agonists, which all induce germinal centers (GCs) and potent antibody responses to influenza hemagglutinin (HA) in adult mice. In neonates, a single dose of HA formulated into each adjuvant induced T follicular helper (TFH) cells. However, only HA/CAF01 elicited significantly higher and sustained antibody responses, engaging neonatal B cells to differentiate into GCs already after a single dose. Although antibody titers remained lower than in adults, HA-specific responses induced by a single neonatal dose of HA/CAF01 were sufficient to confer protection against influenza viral challenge. Postulating that the neonatal adjuvanticity of CAF01 may result from the functionality of the C-type lectin receptor (CLR) Mincle in early life we asked whether other C-type lectin agonists would show a similar neonatal adjuvanticity. Replacing the Mincle agonist trehalose 6,6′-dibehenate by Curdlan, which binds to Dectin-1, enhanced antibody responses through the induction of similar levels of TFH, GCs and bone marrow high-affinity plasma cells. Thus, specific requirements of early life B cells may already be met after a single vaccine dose using CLR-activating agonists, identified here as promising B cell immunostimulators for early life vaccines when included into cationic liposomes.

Keywords: T follicular helper cells, germinal centers, neonates, vaccines, adjuvants

# INTRODUCTION

Neonates and young infants are particularly vulnerable to infectious diseases and providing protection at that early time in life remains challenging (1). One example is influenza, against which currently available vaccines elicit weak responses. Newborn and infant protection against influenza may currently only be achieved by maternal immunization and transplacental transfer of maternal antibodies to the fetus. However, maternal antibodies wane rapidly after birth. Between 6 and 25 months of life, trivalent influenza vaccines (TIV) have limited immunogenicity and protective efficacy (2, 3), which may be enhanced in part by MF59® adjuvantation (4). In contrast, influenza vaccines for infants younger than 6 months are lacking: TIV showed poor efficacy (3) and the live attenuated intranasal vaccine appeared too reactogenic in this age group (5). MF59®-adjuvanted vaccines have not yet been tested in young infants. In infant mice, MF59® induced adult-like antibody titers, T follicular helper (TFH) cells, germinal centers (GCs) and protection against influenza challenge but failed to do so in neonatal mice (6), indicating the existence of different immunological requirements in newborns.

The mechanisms underlying the limitations of early life B cell responses are multiple and not well understood yet. Preclinical murine models suggest that the pattern of early life antibody responses, hallmarked by low antibody titers with limited persistence, reflects the restricted induction of GCs-derived B cells (1, 7). So far, only one adjuvant, LT-K63, was shown to enhance the GC reaction and antibody responses in neonatal mice (8) but its clinical development has been stopped due to transient adverse reactions in humans and its mechanisms of action remain unknown. We and others have previously identified a critical role for TFH cells in the impaired development of GC reactions following neonatal immunization with the current aluminumcontaining vaccines (9, 10). Hence, new adjuvants targeting these specific neonatal requirements are needed.

Several novel candidate adjuvants in advanced clinical development are being assessed within the Advanced Immunization Technologies (ADITEC) collaborative research program (11). Within this consortium, we initially selected three promising adjuvants to explore their neonatal adjuvanticity. Glucopyranosyl lipid adjuvant (GLA)-squalene emulsion (SE) is a SE combined with the TLR4 agonist GLA. In adult mice, GLA-SE elicited potent TH1 responses and protective antibody titers to influenza (12, 13). The induction of strong antibody responses in adults was confirmed in a human phase 1 trial (14). IC31® consists of the cationic membrane interacting peptide KLK (KLKL5KLK) and of a single-stranded DNA-phosphodiester oligo-d(IC)13 (ODN1a), a TLR9 agonist. IC31® induced strong TH1, and substantial murine B cell responses in adult mice (15) and improved influenza vaccine responses in adult and aged mice (16). An IC31®-containing tuberculosis (TB) vaccine was shown to induce potent TH1 responses in humans (17). In neonatal mice, IC31®-containing vaccines elicited adult-like TH1 responses to TB antigens (18, 19) and enhanced TH1 responses, antibody responses, and protection against pneumococcal challenge (20). The combined TH1-driving and B cell supporting functions of GLA-SE and IC31® could thus potentially address some key requirements for neonatal adjuvantation.

CAF01 is an adjuvant composed of a liposomal delivery vehicle formed by the cationic surfactant dimethyldioctadecylammonium (DDA) incorporating the immunostimulator trehalose 6,6′-dibehenate (TDB) (21). CAF01 signals *via* the C-type lectin receptor (CLR) Mincle, activating the Syk/Card9 pathway to increase the production of pro-inflammatory cytokines (22, 23). In adult mice, CAF01 elicited strong TH1/TH17 responses but moderate antibody responses to influenza hemagglutinin (HA) (12). In neonates, CAF01 elicited mixed TH1/TH17 responses against TB antigens (24). Its neonatal B cell adjuvanticity had not yet been assessed.

Here, we used these three novel adjuvant formulations to explore the capacity of the neonatal and adult murine immune system to elicit GC B cell responses to influenza HA. Our findings identified for the first time CAF01 as a potent neonatal adjuvant able to strongly enhance neonatal B cell responses and thus the protective efficacy of early life vaccines. Interestingly, formulating Curdlan, a different CLR agonist, in DDA similarly increased primary neonatal B cell responses to HA, revealing the great potential of CLR agonists as B cell adjuvants for early life vaccines.

### MATERIALS AND METHODS

### Mice

Adult CB6F1/OlaHsd females were purchased from Harlan (Horst, The Netherlands) together with BALB/c OlaHsd females and C57BL/6 OlaHsd males. The latter were crossed to produce F1 CB6F1 mice. All mice were bred, kept in pathogen-free animal facilities in accordance with local guidelines and used at 1 week (neonates) or 6–8 weeks (adults) of age. All animal experiments were approved by the Geneva veterinary office and conducted under relevant Swiss and European guidelines.

### Antigens, Adjuvants, and Immunization

We used an experimental monovalent purified subunit influenza vaccine composed of HA from the influenza strain H1N1 A/California/7/2009 [Novartis Vaccines (a GSK company), Siena, Italy]. Groups of 5–8 CB6F1 neonatal (1-week-old) and adult mice were immunized subcutaneously (s.c.) with 100 µl of the plain HA (1 µg) or in combination with either CAF01 (250 µg DDA/50 μg TDB, Statens Serum Institut, Copenhagen, Denmark), IC31® (KLK/ODN1a = 100 nmol/4 nmol, Valneva Austria GmbH), GLA-SE (5 µg GLA and 2% v/v squalene, Infectious Diseases Research Institute, Seattle, WA, USA), or DDA-Curdlan (250 µg DDA/50 μg Curdlan, Statens Serum Institut, Copenhagen, Denmark) produced according to the protocol previously described for DDA-TDB (25). Curdlan was purchased from Sigma-Aldrich.

Mice were immunized at the base of the tail and inguinal draining lymph nodes (LNs) were harvested, except for the experiments with GLA-SE in which mice from both age groups were injected s.c. (100 µl) at the scruff of the neck and brachial draining LNs harvested. This use of the base of the tail as injection site was required to comply with the new local animal welfare guidance to limit procedures requiring anesthesia. We carefully checked that this change did not affect our results (Figure S1 in Supplementary Material).

## Influenza Challenge

Viral challenge was performed as recently described (6) using a mouse-adapted H1N1 Influenza strain (A/Netherlands/602/2009, passage 2 in mice). Virus was grown on MDCK cells (ATCC). Fifty-six days post-immunization, 2 × 103 PFU of virus in 20 µl sterile PBS were instilled intranasally under anesthesia induced by Ketasol® (Graeub) and Rompun® (Bayer). Mice were observed daily to monitor body weight and survival. Mice showing more than 20% of body weight loss were humanely euthanized.

## Enzyme-Linked Immunosorbent Assay (ELISA) and Avidity of HA-Specific Antibodies

Mice were bled from the tail vein at the indicated time points except for neonatal mice at day 0 that were bled by decapitation. Titration of HA-specific total IgG, IgG1, and IgG2a titers was performed by ELISA on individual serum samples as previously described (6).

Avidity was measured by ELISA as the overall strength of binding between antibody and antigen, using plates incubated for 15 min with increasing concentration of ammonium thiocyanate (NH4SCN) from 0 to 1.5 M. Antibody avidity was defined as the amount of antibody eluted for each increment of NH4SCN concentration. Arbitrarily, the percentage of antibody eluted between 0 and 0.25 M of NH4SCN was assigned as low-avidity antibody fraction, between 0.25 and 0.75 M as intermediate and >0.75 M as high-avidity antibody fraction.

### Enzyme-Linked Immunospot (ELISpot) Assay for HA-Specific Plasma Cells

Hemagglutinin-specific plasma cells were quantified by direct *ex vivo* ELISpot assay as previously described (6).

# Flow Cytometric Analysis

Cells from the two draining LNs of each individual mouse were pooled and stained with fluorescently labeled antibodies to GL7, CD8, B220, TCR-β, CD95 (Fas) (all from BD Biosciences), PD-1, Ter119, GR1, CD11c (all from eBioscience), and CD4 (all from BioLegend). CXCR5 staining was performed using purified anti-CXCR5 (BD Biosciences), followed by FITC anti-rat IgG (Southern Biotech), and normal rat serum (eBioscience). The stained cells were analyzed using a Gallios cytometer (Beckman Coulter) and the generated data analyzed using FlowJo Software (Tree Star).

# Immunohistochemistry

Germinal centers in the draining LNs of immunized mice were stained and quantified as previously described (6). Sections were visualized and photographed with a Zeiss LSM700 confocal microscope (objective: 20×) or a Mirax scan microscope (Zeiss). Images were acquired with Zeiss LSM image browser software (Zeiss) or the Pannoramic Viewer software (3DHistec).

# Statistical Analysis

Data were analyzed using Prism 6.0 (GraphPad Software) and presented as mean ± SEM of at least three independent experiments. Difference between groups was analyzed as described in figure legends. *P*-values less than 0.05 were considered statistically significant.

# RESULTS

### Novel Adjuvants Exert Distinct Effects on B Cell Responses to Vaccination in Adult and Early Life

We first compared the antibody titers elicited by a subunit monovalent influenza vaccine containing HA administered alone or formulated with GLA-SE, IC31®, or CAF01. Neonates (referred to as "1 week") and adult mice were immunized twice at days 0 and 21. In adults, a single dose of unadjuvanted HA was sufficient to elicit HA-specific antibody responses (**Figure 1**). These responses were strongly enhanced by each tested adjuvant (**Figure 1**), as recently reported (12). The highest primary IgG responses were induced by HA/GLA-SE (**Figure 1A**), followed by HA/IC31® (**Figure 1B**), and HA/CAF01 (**Figure 1C**). HA/GLA-SE induced antibodies with a more pronounced IgG2a profile, while HA/ IC31® and HA/CAF01 primarily induced IgG1 antibodies.

Consistently with our previous findings (6), neonates did not raise detectable primary IgG responses to unadjuvanted HA. They also poorly responded to a single dose of HA/GLA-SE or HA/IC31®. In contrast, HA/CAF01 strongly increased primary HA-specific IgG and particularly IgG1 responses (**Figure 1C**). Neonatal responses to HA/CAF01 were significantly higher than to HA/IC31® (*P*≤ 0.004) and HA/GLA-SE (*P*≤ 0.0004). Neonatal responses to the second dose of HA/CAF01 were also significantly higher than those elicited by unadjuvanted HA, although neither primary nor secondary responses reached the titers elicited by adjuvanted vaccines in adults (**Figure 1**).

Thus, adult and neonatal requirements for B cell adjuvanticity differ: HA/GLA-SE is the strongest antibody inducer in adults, whereas only CAF01 adjuvantation succeeds in eliciting potent primary responses in neonates.

# The Three Adjuvants Successfully Induce T Follicular Helper Cells in Neonates

TFH cells are critical for GC formation, controlling the number of GC B cell divisions, and are thus essential for the generation of high-affinity matured antibodies (26, 27). The induction of TFH cells in early life is challenging: we previously reported that both alum-adsorbed (10) and MF59®-adjuvanted (6) vaccines failed to generate TFH cells in neonates. CD4<sup>+</sup> CXCR5highPD-1high TFH cells were thus measured in the draining LNs after neonatal or adult (control) immunization. TFH cell responses increased significantly upon HA/CAF01 immunization in both age groups, reaching similar numbers in adults and neonates (**Figures 2A,B**).

Unexpectedly, despite the poor neonatal primary antibody responses to HA/GLA-SE and HA/IC31®, a significant increase in CD4<sup>+</sup>CXCR5highPD-1high TFH cells, reaching adult levels, was

non-adjuvanted groups using the Mann–Whitney *U* test. \*HA/Adj vs HA statistics in adult mice: \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001. # statistics in neonatal mice: # *P* < 0.05, ##*P* < 0.01, ###*P* < 0.001.

also observed in HA/GLA-SE- and HA/IC31®-immunized neonates (**Figures 2A,B**). Neonatal TFH cell responses elicited by HA/GLA-SE peaked at day 10 but dropped down relatively rapidly over time (Figure S2A in Supplementary Material)*.* In contrast, less than 0.05% of CD4<sup>+</sup> CXCR5highPD-1high TFH cells were repeatedly observed in naïve infant mice (not shown). Thus, all tested adjuvants passed the challenge of inducing TFH cells in neonates.

## Only HA/CAF01 Elicits *Bona Fide* GC Responses in Neonates

As GC reactions are essential determinants of the magnitude and the quality of antibody responses and humoral memory, we next assessed whether these adjuvants differed in their capacity to induce GCs. B220<sup>+</sup>GL7<sup>+</sup>CD95<sup>+</sup> GC B cells were quantified by flow cytometry in draining LNs at day 10 or 12 upon immunization, as indicated in figure legends. In adults, the three adjuvants significantly increased the numbers of GC B cells when compared with unadjuvanted HA (**Figures 3A,B**). The few GC B cells in the draining LNs of neonates that received unadjuvanted HA were comparable with that observed in nondraining LNs (not shown), representing background levels. In neonates, adjuvantation with CAF01 strongly augmented both the number (**Figure 3B**) and proportion of GC B cells compared with unadjuvanted HA (HA/CAF01: 1.99% ± 0.22 vs HA: 0.30% ± 0.05, *P* < 0.0001). In contrast, only few GC B cells were observed following HA/IC31® or HA/GLA-SE neonatal immunization. Neonatal GC B cell responses induced by HA/GLA-SE remained low even later after immunization (Figure S2B in Supplementary Material)*.* In contrast, in adult mice higher GC B cell responses were observed at day 12 following vaccination with HA/GLA-SE (Figure S2B in Supplementary Material). Thus, a single dose of HA/CAF01-induced significantly higher numbers of neonatal GC B cells than unadjuvanted HA, HA/ GLA-SE, or HA/IC31®, in consistency with the observed higher antibody titers (**Figure 1**).

To explore whether the GC B cells identified by flow cytometry were organized in *bona fide* GC structures, we imaged follicular B cells (IgD, green), GC B cells [peanut agglutinin (PNA), red] and CD4<sup>+</sup> T cells (CD4, purple) in draining LN sections post-prime (**Figure 3C**). In adults, highly organized IgD<sup>−</sup>PNA<sup>+</sup> GC structures were observed. These GCs were very few in adult mice that received unadjuvanted HA and increased following immunization with the adjuvanted HA, irrespective of which adjuvant (**Figure 3C** and data not shown). Vaccination of neonatal mice with unadjuvanted HA, HA/

HA/Adj vs HA

IC31®, or HA/GLA-SE did not generate *bona fide* GC structures. In contrast, well-organized GCs were observed in HA/ CAF01-immunized neonates (**Figure 3C**), albeit PNA<sup>+</sup> GCs induced by HA/CAF01 remained fewer (1 week: 1.85 ± 0.29 vs adults: 3.44 ± 0.46, *P* ≤ 0.01 per LN section) and smaller (1 week: 23,883 ± 9,973 µm2 vs adult: 42,888 ± 9,810 µm2 ) than in adults. Thus, a single dose of HA/CAF01 elicited *bona fide* GCs even in neonates.

# A Single Dose of HA/CAF01 Elicits High-Affinity Sustained B Cell Responses

Early life immunization in general induces reduced titers of antibodies and these are of lower avidity compared with responses achieved by immunization in adults (1, 28). We thus asked whether a single dose of HA/CAF01 also affected these hallmarks of neonatal B cell responses. First, serum IgG antibody responses were measured for up to 9 weeks post-prime. In adults, HA/CAF01 induced significantly higher antibodies than unadjuvanted HA and these responses reached a plateau already 3 weeks after prime (**Figure 4A**). A similar enhancement of antibody responses was observed in neonates, although with slower kinetics (peak at 5 weeks, **Figure 4A**) and, as previously reported, lower magnitude (**Figure 1C**). HA/CAF01-induced antibodies persisted for at least 9 weeks post-prime (last time point assessed, **Figure 4A**) in both neonates and adults.

CAF01 (29) and IC31® (15) both induce a depot effect which may allow a slow antigen release from the injection site and contribute to sustained antibody responses. However, antibodies induced by HA/IC31® in neonates remained very low over time (Figure S3 in Supplementary Material). Thus, a depot effect is not sufficient to eventually elicit antibody responses in early life.

The avidity of HA-specific IgG antibodies was measured early (3 weeks) and late (9 weeks) after a single dose of HA/CAF01 (**Figure 4B**). Avidity is shown as percentages of eluted HA-specific antibodies after treatment with increasing concentrations of ammonium thiocyanate (NH4SCN) (30). At 3 weeks post-prime, a higher proportion of low-avidity antibodies (eluted with a low concentration of ammonium thiocyanate) was observed in neonates than in adults (**Figure 4B**). However, avidity increased rapidly following neonatal immunization with HA/CAF01, reaching adult-like levels at 9 weeks (**Figure 4B**).

We previously reported that neonatal immunization with aluminum adjuvants elicits abortive and rapidly terminated GC responses (10). In contrast, a single dose of HA/CAF01-induced sustained GC activity: HA/CAF01-induced GC B cells persisted in both age groups for at least 5 weeks after a single immunization, albeit at significantly lower numbers than in adults in agreement with previous results (Figure S4 in Supplementary Material). Thus, the neonatal B cell adjuvanticity of CAF01 results into a potent and sustained induction of *bona fide* GC B cells, which rapidly generates high-affinity and persistent primary antibody responses.

### A Single Dose of HA/CAF01 Protects Mice against Influenza Virus Challenge

We next asked whether the HA-specific antibody responses elicited by a single dose of HA/CAF01 in neonates were sufficient to confer protection against influenza viral challenge, which is difficult to achieve. Neonatal mice received a single dose of unadjuvanted HA, HA/CAF01, or PBS (control) and were challenged intranasally 8 weeks later with a lethal dose of matching influenza A/H1N1 virus. As additional controls, we challenged HA/CAF01- (positive control) or PBS- (negative control) injected adult mice. The body weight of mice that received PBS or unadjuvanted HA declined to 80% on average within 6 days (**Figures 5A,B**). In contrast, the body weight of HA/CAF01-immunized mice only transiently declined by less than 10% (4–6%) in both age groups (**Figures 5A,B**). The protection mediated by HA/CAF01 translated into a survival rate of 100% in both age groups (**Figures 5C,D**). In neonates, CAF01 adjuvantation was required for protection, as only 3/8 mice that received unadjuvanted HA survived up to day 8 (**Figure 5C**). Among these survivors, two mice had signs of severe infection and lost considerable body weight (17.1 and 19.9% of weight loss, respectively). Similarly,

\*\*\**P* < 0.001, \*\*\*\**P* < 0.0001.

in the group of neonates that received PBS, one of two surviving mice had signs of severe infection on day 8 (18.3% of weight loss). Only one mouse in each group either was not fully infected or recovered spontaneously. Thus, a single dose of HA/CAF01 is sufficient to confer protection against influenza viral challenge even when given to neonates.

Mann–Whitney *U* test: \**P* < 0.05, \*\*\*\**P* < 0.0001.

# Primary Responses to HA/DDA-Curdlan

The potent neonatal B cell adjuvanticity of the Mincle-activating CAF01 adjuvant suggests that CLR targeting may be critical in neonates rather than signaling through TLRs. To explore this possibility, we tested the neonatal B cell adjuvanticity of Curdlan, which binds the CLR Dectin-1. Thus, by replacing the CAF01 immunostimulator TDB with Curdlan in the cationic surfactant DDA, we generated DDA-Curdlan.

We first confirmed the adjuvanticity of DDA-Curdlan in adult mice: a single dose of HA/DDA-Curdlan strongly enhanced HA-specific IgG antibodies compared with unadjuvanted HA [anti-HA IgG (log10) 4.4 ± 0.14 vs 2.1 ± 0.07, *P* < 0.001, 4 weeks post-prime], a similar effect to that induced by CAF01. Then, we vaccinated neonatal mice with unadjuvanted HA or formulated with either CAF01 or DDA-Curdlan. HA/DDA-Curdlan strongly increased primary HA-specific neonatal antibody responses, with no differences to HA/CAF01 (**Figure 6A**). Both adjuvants significantly increased HA-specific antibody titers when compared with unadjuvanted HA, at all assessed time points for up to 10 weeks post-prime. This reflected the enhancement of both TFH and GC B cell responses by DDA-Curdlan, with no differences compared with CAF01 (**Figure 6B**).

Next, we evaluated the capacity of both formulations to induce antibody-secreting cells (ASCs) able to home into the bone marrow (BM), another hallmark of potent GC reactions which is rarely reached in early life (28). Neonatal mice were immunized s.c. at day 0 with HA/CAF01 or HA/DDA-Curdlan. Adult mice receiving HA/CAF01, HA only or naïve were included as control groups. Except for naive animals, all groups of mice were boosted 10 weeks later with HA only and ASCs were measured 1-week post-boost by ELISpot assay. High numbers of HA-specific ASCs were retrieved from the neonatal BM after vaccination with either HA/CAF01 or HA/DDA-Curdlan (**Figure 6C**). Neonatal responses were higher than those observed in adults immunized with HA only, but did not reach the levels observed in adults immunized with HA/CAF01. High numbers of ASCs were also observed in the spleens of immunized neonates (**Figure 6D**). Thus, the neonatal B cell adjuvanticity of CAF01 and DDA-Curdlan results in a potent and sustained induction of *bona fide* GC B cells, which rapidly generates high-affinity and long-lived ASCs.

# DISCUSSION

In this study, we explored the neonatal potency of three adjuvants in clinical development given their strong B-cell promoting activity in adults, and we identified the Mincle agonist-containing adjuvant CAF01—but not the TLR-based adjuvants GLA-SE and IC31®—as capable of inducing *bona fide* GC responses and thus robust and prolonged primary humoral responses in murine neonates. Our major observations with CAF01 were extended to an agonist targeting a distinct CLR (Curdlan/Dectin-1), revealing the potency of CLR agonists over TLR-based adjuvants in circumventing the limitations of neonatal B cell responses to current early life vaccines.

TLR agonists are a group of arising adjuvants, some of which have already been approved for human use or are currently in clinical trials. Monophosphoryl Lipid A (MPL), a potent agonist of TLR4, is currently in use in combination with alum in vaccines against hepatitis B and papilloma virus in adults. AS01, containing MPL and the saponin QS-21, was included in the candidate RTS,S malaria vaccine in infants and children affected by HIV and improved responses to vaccination (31). TLR7/8 agonists enhanced responses to vaccination in neonatal mice and rhesus macaques (32) and their ability to activate APCs from human cord

blood *in vitro* is well documented (33–37). CpG oligonucleotides, agonists of TLR9, partially circumvented the TH2 polarization of neonatal responses to vaccines and increased antibody responses to distinct antigens (8, 10, 38). Considering the growing evidence of adjuvant activity of TLR agonists on neonatal APCs and T cells, and the strong GC-inducing capacity of both GLA-SE and IC31® adjuvants in adult mice, their weak neonatal B cell adjuvanticity was unexpected.

Graphs show survival rates post-challenge. Data are representative of two independent experiments.

Reduced TLR-mediated responses have been reported in early life (39, 40) and may limit the functionality of GLA-SE and IC31® in this age group. However, their lack of early life B cell adjuvanticity did not concur with T cell unresponsiveness to the TLR4 or TLR9-agonists: both HA/GLA-SE or HA/IC31® increased TFH cell responses compared with unadjuvanted HA, suggesting sufficient APC activation to initiate the TFH cell differentiation process and induce adult-like TFH cell numbers. In adults, TFH cell responses directly translate into GC induction and strong antibody responses. This is not the case in neonates, indicating the existence of additional requirements for the optimal induction of GC responses by TLR agonists. In humans, adult-like TLR4- and TLR9-mediated APC/T cell responses are typically achieved during the first year of life (7, 41). Should the induction of TFH cell responses not directly translate into B cell responses in humans as observed in mice, the youngest age at which HA/ GLA-SE or HA/IC31® might induce potent B cell responses to primary vaccination may thus not be predicted.

CAF01 includes the C-type lectin agonist TDB, which signals through Mincle and the Syk-Card9 pathway (23). A recent report showed that TDB activates human newborn DCs and greatly enhanced TH1 polarizing cytokine production by DCs when given in combination with a TLR7/8-ligand (42), extending previous preclinical reports of the unique efficacy of CAF01 to induce TH1/TH17 responses in murine neonates (43). In adults, the B cell adjuvanticity of CAF01 is lower than that of GLA-SE or IC31® (12, 44). Its greater capacity to trigger the differentiation of neonatal B cells into *bona fide* GCs was thus unexpected. It does not merely result from its DDA-associated depot effect (29), also exhibited by IC31® (15), but likely from the activation of a CLR-mediated pathway—as shown by the similar GC-promoting capacity of CAF01 and DDA-Curdlan in neonates. Curdlan was shown to enhance TH1 responses to a subunit TB vaccine in neonatal mice (45) and neonatal human monocyte-derived DCs readily responded to Dectin-1 and TLR7/8 agonists by producing IL-12p70 (45). Importantly, this study is the first evidence of its potent neonatal B cell adjuvanticity *in vivo*.

What may explain the higher neonatal B cell adjuvanticity of CLR- over TLR-based adjuvants?

The recognition by TLRs mainly triggers intracellular signaling cascades that result in APC maturation and the induction of inflammatory cytokines, leading to T cell activation (46, 47). In contrast, CLRs are known to perform as efficient endocytic receptors for antigen on the surface of APCs, especially on DCs where they are highly expressed (48). CLRs main function is to internalize their ligand antigens for degradation in lysosomal compartments and to enhance antigen processing and presentation by DCs and other APCs (49, 50). Efficient APC maturation in turn may provide better T cell activation. CLRs do not only function as antigen uptake receptors, they

Figure 6 | A single dose of hemagglutinin (HA)/dimethyldioctadecyl-ammonium (DDA)-Curdlan elicits primary responses similar to HA/CAF01. (A,B) 1-week-old mice were immunized subcutaneously (s.c.) at day 0 with HA alone, HA/CAF01, or HA/DDA-Curdlan. (A) Sera samples were drawn at the indicated time points to measure HA-specific IgG titers. Values represent mean logarithmic titers (log 10) of more than eight mice per group ± SEM. Both CAF01 and DDA-Curdlan significantly increased antibody titers when compared with unadjuvanted HA at all assessed time points (statistics not shown in the graph). (B) Graphs report the total CXCR5highPD-1high TFH cell numbers and the total GL7+CD95+ GC B cell numbers in draining lymph nodes at day 12 post-immunization. (C,D) 1-week-old and adult mice were immunized s.c. at day 0 as indicated and boosted 10 weeks later with HA only; 1-week post-boost antibody-secreting cells (ASC) were measured by enzyme-linked immunospot assay. The graph shows the proportions of ASCs in the bone marrow (C) and spleen (D) in the vaccinated groups and naïve mice. Data pool of three independent experiments. Mann–Whitney *U* test: \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001.

also facilitate efficient loading of antigen on MHC class I and II molecules and induce both antigen-specific CD8 and CD4 T cell responses (51, 52).

Although their APC activation capacities may differ, all tested adjuvants efficiently triggered the induction of TFH cells, which is in contrast to what we observed with aluminum salt-based or MF59®-adjuvanted influenza vaccines in neonates (6, 10). However, the quality of TFH cells elicited by CLR- vs TLR-based adjuvants might differ. TFH cell functionality, hallmarked by high-expression levels of Bcl6, ICOS, and secretion of cytokines such as IL-21 and IL-4, is critical for optimally cognate TFH/B cell interactions in GCs (53, 54) and subject to current studies. Another hypothesis is that CLR- and TLR-based activation may essentially differ at the B cell level. Both GLA-SE and IC31® induce a small number of GC B cells in neonates, although these fail to develop into *bona fide* GC structures. Efficiently activated APCs by CLR agonists may provide early GC B cells with the amount of antigen required for development and persistence of GC.

Moreover, it would be interesting to study whether CLRbased adjuvants share common mechanisms with LT-K63, the first adjuvant shown to induce early and persistent antibodies responses in neonatal mice (8). All these questions are now open for studies focusing on the relative GC B cell inducing capacity of adjuvants considered for use in early life.

Despite the strong adjuvanticity of CAF01 in neonates, GC and antibody responses did not reach the levels elicited by adjuvanted vaccines in adults. This suggests that additional neonatal limiting factors may be addressed to further improve B cell adjuvants for early life.

The identification of CAF01—a safe adjuvant currently in clinical development—as a potent neonatal adjuvant, the definition of its mode of adjuvanticity through the induction of *bona fide* GC responses and the demonstration that this property is shared by a distinct CLR agonist are major steps forwards. This paves the way to a large area of investigation to identify CLR agonistcontaining adjuvants or combination-derivatives thereof, that are able to induce the most appropriate and effective responses to vaccination in early life.

## ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the Geneva veterinary office and conducted under relevant Swiss and European guidelines. The protocol was approved by the Geneva veterinary office.

# AUTHOR CONTRIBUTIONS

MV, CSE, EM, FA, DC, P-HL, BM-G, and C-AS contributed to formulation of theory and prediction. MV, CSE, P-HL, BM-G, and C-AS designed research. BM-G, MV, CSE, EM, FA performed the experiments and analyzed and/or interpreted the data. MV, CSE, BM-G, and C-AS wrote the manuscript. DC, MS, RC, AM, and

### REFERENCES


PA provided reagents and critically revised the manuscript. All authors reviewed the manuscript.

### ACKNOWLEDGMENTS

The adjuvants were obtained from partners involved in the EU-funded large collaborative project ADITEC. We thank Novartis Vaccines (a GSK company, Siena, Italy) for providing the HA antigen. We thank Florian Krammer for providing the A/Netherlands/602/2009 isolate. We thank Steven G. Reed for his support to the evaluation of GLA-SE in neonates within the ADITEC consortium, and David Pejoski for discussions; Stephane Grillet, Chantal Tougne, Paola Fontannaz, and Anne Rochat for their key contribution to the complex experimental work required by this study; Anthony Joubin for excellent assistance with animal care; and the colleagues of the Bioimaging and FACS facilities at the University of Geneva.

# FUNDING

This study was supported by grants to C-AS from the European Commission of the Seventh Framework Programme (Advanced Immunization Technologies, 280873) and from the Swiss National Science Foundation (grant number 310000-111926/1 and 310030-165960).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00381/ full#supplementary-material.

cells and poor persistence of both protein- and polysaccharide-specific antibody-secreting cells in neonatal mice. *J Immunol* (2012) 189(3):1265–73. doi:10.4049/jimmunol.1200761


dendritic cells via a humanized anti-DC-SIGN antibody. *Blood* (2005) 106(4):1278–85. doi:10.1182/blood-2005-01-0318


**Conflict of Interest Statement:** PA and DC are co-inventors on patent applications covering CAF01. As employees, DC and PA have assigned all rights to Statens Serum Institut, a Danish non-profit governmental institute. RC is an employee at the Infectious Disease Research Institute, Seattle. AM is an employee at Valneva Austria GmbH. Other authors have no conflict of interest to declare.

*Copyright © 2018 Vono, Eberhardt, Mohr, Auderset, Christensen, Schmolke, Coler, Meinke, Andersen, Lambert, Mastelic-Gavillet and Siegrist. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Molecular signatures of a Tlr4 agonist-adjuvanted hiV-1 Vaccine candidate in humans

*Jenna Anderson1†, Thorunn A. Olafsdottir1†, Sven Kratochvil2 , Paul F. McKay2 , Malin Östensson1 , Josefine Persson1 , Robin J. Shattock2 and Ali M. Harandi1 \**

*1Department of Microbiology and Immunology, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, 2Department of Medicine, Mucosal Infection and Immunity Group, Section of Virology, Imperial College London, London, United Kingdom*

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Thorsten Demberg, Immatics Biotechnologies, Germany Randy A. Albrecht, Icahn School of Medicine at Mount Sinai, United States*

*\*Correspondence:*

*Ali M. Harandi ali.harandi@microbio.gu.se These authors have contributed* 

*†*

### *Specialty section:*

*equally to this work.*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 07 November 2017 Accepted: 02 February 2018 Published: 26 February 2018*

### *Citation:*

*Anderson J, Olafsdottir TA, Kratochvil S, McKay PF, Östensson M, Persson J, Shattock RJ and Harandi AM (2018) Molecular Signatures of a TLR4 Agonist-Adjuvanted HIV-1 Vaccine Candidate in Humans. Front. Immunol. 9:301. doi: 10.3389/fimmu.2018.00301*

Systems biology approaches have recently provided new insights into the mechanisms of action of human vaccines and adjuvants. Here, we investigated early transcriptional signatures induced in whole blood of healthy subjects following vaccination with a recombinant HIV-1 envelope glycoprotein subunit CN54gp140 adjuvanted with the TLR4 agonist glucopyranosyl lipid adjuvant-aqueous formulation (GLA-AF) and correlated signatures to CN54gp140-specific serum antibody responses. Fourteen healthy volunteers aged 18–45 years were immunized intramuscularly three times at 1-month intervals and whole blood samples were collected at baseline, 6 h, and 1, 3, and 7 days post first immunization. Subtle changes in the transcriptomic profiles were observed following immunization, ranging from over 300 differentially expressed genes (DEGs) at day 1 to nearly 100 DEGs at day 7 following immunization. Functional pathway analysis revealed blood transcription modules (BTMs) related to general cell cycle activation, and innate immune cell activation at early time points, as well as BTMs related to T cells and B cell activation at the later time points post-immunization. Diverse CN54gp140-specific serum antibody responses of the subjects enabled their categorization into high or low responders, at early (<1 month) and late (up to 6 months) time points post vaccination. BTM analyses revealed repression of modules enriched in NK cells, and the mitochondrial electron chain, in individuals with high or sustained antigen-specific antibody responses. However, low responders showed an enhancement of BTMs associated with enrichment in myeloid cells and monocytes as well as integrin cell surface interactions. Flow cytometry analysis of peripheral blood mononuclear cells obtained from the subjects revealed an enhanced frequency of CD56dim NK cells in the majority of vaccines 14 days after vaccination as compared with the baseline. These results emphasize the utility of a systems biology approach to enhance our understanding on the mechanisms of action of TLR4 adjuvanted human vaccines.

Keywords: HIV vaccine, systems vaccinology, transcriptomics, adjuvant, glucopyranosyl lipid adjuvant-aqueous formulation, TLR4

# INTRODUCTION

Since the onset of the epidemic, more than 70 million people have been infected with human immunodeficiency virus (HIV), and about 35 million people have died of HIV (1). An efficacious vaccine, alone or in combination with other interventions, is widely considered to hold promise for controlling the spread of HIV infection worldwide (2). Despite numerous efforts over the past three decades, an effective anti-HIV vaccine remains elusive. As exemplified in previous vaccine trials [reviewed in Ref. (3)], including the recent encouraging RV144 trial in Thailand (4), inter-individual variation in the immune response to HIV-1 is one of the several challenges for developing a broadly protective anti-HIV vaccine (5).

Antibodies directed against the viral *Env* protein are largely accepted to contribute to protection against HIV infection (6), though the nuanced involvement of different antibody classes or subclasses is unclear. Recent studies have shown that immunization with the stable trimeric recombinant HIV-1 envelope glycoprotein, CN54gp140, induces strong systemic and mucosal immune responses following different immunization routes and strategies (7–9). In particular, immunization with this candidate antigen has induced potent humoral immune responses when adjuvanted with TLR4 agonist adjuvants, such as monophosphoryl lipid A (9) or GLA-AF (glucopyranosol lipid adjuvant-aqueous formulation) (10). These adjuvants are deemed to exert their adjuvanticity, at least in part, by activating the myeloid differentiation factor 88 (MyD88) and toll-interleukin 1 receptor domain-containing adapter inducing interferon-β (TRIF) pathways (11–13). Nevertheless, the magnitude of elicited responses to CN54gp140 adjuvanted with GLA-AF are variable between individuals, the degree to which such variability could be contributed to difference in responsiveness to the adjuvant are unknown.

Early transcriptomics profiling has recently provided new insights into the mode of action of human vaccines and adjuvants (14–16), and has the potential to help elucidate mechanisms underlying diverse individual responses to identical vaccinations. Systems biology was first used to identify gene signatures correlating with immune responses in humans following vaccination with the yellow fever vaccine YF-17D (17). Other studies have since then employed whole blood or peripheral blood mononuclear cell samples to evaluate gene expression signatures of human vaccines and correlate them with vaccine responses [for example (14, 18–20), and as reviewed in Ref. (15)]. Here, we employed whole genome transcriptomics combined with a systems biology approach with the aim of characterizing early molecular signatures induced in the whole blood of healthy volunteers vaccinated with an HIV-1 subunit vaccine candidate, composed of CN54gp140 adjuvanted with the TLR4 agonist GLA-AF. We herein report early transcript signatures and blood transcription modules (BTMs) in the whole blood within 7 days following vaccination. Further, we identified BTMs that were differentially enriched in high vs low serum CN54gp140-specific antibody responders. These results provide new insights into the early blood transcript signatures of TLR4 agonist-adjuvanted HIV-1 envelope glycoprotein vaccine candidates in humans.

# MATERIALS AND METHODS

### Study Subjects and Vaccine

Healthy male (*n* = 8) or female (*n* = 6) volunteers aged between 18 and 45 and with no history of HIV-1 and HIV-2 infection were enrolled in an open label randomized clinical trial at Imperial College London, UK (study registered at ClinicalTrials.gov under registration no. NCT01966900). Informed written consent was obtained for all volunteers as part of the enrollment process. A recombinant uncleaved clade C HIV-1 envelope gp140 protein (CN54gp140) produced by Polymun Scientific (Klosterneuburg, Austria) to GMP specification, which has previously been reported to be immunogenic in a number of preclinical and clinical studies (7, 8, 21, 22), was used for the clinical trial. All subjects were immunized in the deltoid muscle three times at 1-month intervals with a vaccine candidate composed of 100 µg CN54gp140, adjuvanted with 5 µg GLA-AF. Of the 14 recruited participants, 12 went on to receive additional boosted immunizations at 6 (*n* = 6) or 12 (*n* = 6) months post their third priming immunization. More detailed information concerning study participants and design can be found in Ref. (23). Whole blood samples were taken for RNA analysis at the baseline (day 0) as well as at 6 h and 1, 3, and 7 days following first immunization (reported for this study only). This work was performed in compliance with UK Clinical Trial Regulations and all procedures were approved by the recognized Research Ethics Committee and by the Medicines and Healthcare Products Regulatory Authority.

### RNA Samples

Whole blood samples were collected in PAXgene RNA stabilization tubes (Qiagen, USA) from each participant before vaccination (day 0, 0 h) as well as at 6 h and 1, 3, and 7 days following first vaccination. RNA was extracted using the PAXgene Blood RNA kit, according to the manufacturer's protocols (Qiagen, USA). Spectrophotometry was used to measure RNA concentration (ND-1000 spectrophotometer, NanoDrop Technologies Inc., USA) and the RNA integrity number (RIN), an indicator of sample quality, was determined using an Agilent 2200 TapeStation and 2100 Expert Software (Agilent Technologies, USA). The RIN for the mRNA extracted from the remaining 66 whole blood samples ranged from 6.2 to 9.0 with a mean of 7.9 ± 0.6 for all extracted samples. Only samples with good quality mRNA, defined as having a 260/280 ratio approximating 2 and RIN >7, were used in microarray analyses.

### Microarray Analysis and Data Acquisition

Whole genome microarray analysis was performed at the Genomics Core Facility, University of Gothenburg, Sweden using an Illumina platform according to the manufacturer's protocols. Briefly, the Fluorescent Linear Amplification Kit was used for RNA labeling, and quantity and labeling efficiency were verified prior to hybridization of samples to whole genome 8x60k human expression arrays (HumanHT-12 v4 Expression BeadChip kit; Illumina, USA). Post-hybridization scans were performed using an Agilent scanner at 5 µm and image analysis to generate raw data was completed using Feature Extraction software (version 11.5.1.1, Agilent Technologies). The raw data was pre-processed and normalized using the limma package and corrected for background using the "normexp" method in the statistical program R (24). Differentially expressed genes (DEGs) compared to baseline were identified by performing the moderate Student's *t*-test at each of the time points (*p*-value) followed by further adjustments for multiple testing using the Benjamini–Hochberg method (adjusted *p*-value). Changes in expression at the four time points compared to baseline were calculated by subtracting the normalized expression value of each individual transcript at baseline from that at the specified time point. Following microarray analysis, four samples were excluded from further analyses due to failed cRNA synthesis. In total, 62 high quality mRNA samples were included in further analyses (Figure S1A in Supplementary Material).

### Gene Set Enrichment Analyses

Gene set enrichment analysis (GSEA) using the blood transcription modules identified in Ref. (14) was applied to the data at each individual time point compared to baseline, using the GSEA software from the Broad Institute (25, 26).

Serum samples were obtained during the course of the study and concentrations of CN54gp140-specific serum IgM, IgA, and IgG (including subclasses IgG1, IgG2, IgG3, and IgG4) antibodies were determined as previously described (23). Individuals were classified either as low or high responders based on early CN54gp140-specific IgM (14 days), IgG (28 days), and IgA (28 days) antibody concentrations, as well as on late antigen-specific serum antibody concentrations (168 days for IgA, IgG, IgG1, IgG2, and IgG4; 84 days for IgG3) (Figure S2 in Supplementary Material). These groupings were statistically validated using an unpaired Mann–Whitney test (at least *p* < 0.05 for all). GSEA using BTMs was applied as described above, to individual-level gene expression data for high or low antibody responders compared to baseline, per antibody class or subclass.

### Antibodies

For phenotypic analysis of lymphocyte subsets by multi-color flow cytometry, PBMCs were stained with fluorochrome-conjugated antibodies for the surface markers CD3 (OKT3, Biolegend), CD4 (SK3, Biolegend), CD8 (SK1, Biolegend), CD14 (M5E2, Biolegend), CD19 (HIB19, Biolegend), CD56 (NCAM16.2, BD Biosciences), CD94 (DX22, Biolegend), CD161 (HP-3G10, Biolegend), CRACC (235614, R&D Systems) NKG2A (Z199, Beckman Coulter), NKp80 (MA152, Beckman Coulter), and TCRγδ (IMMU510, Beckman Coulter) as well as for intracellular EOMES (WD1928, eBioscience), granulysin (GNLY) (DH2, Biolegend), NKG7 (2G9, Beckman Coulter), and PLZF (R17-809, BD Biosciences). Intracellular HOPX was detected with an unconjugated primary antibody (rabbit polyclonal, Proteintech) followed by a fluorochrome-labeled F(ab)2 anti-rabbit IgG secondary antibody (goat polyclonal, ThermoFisher). Dead cells were identified using a fixable live/dead stain (ThermoFisher).

### Flow Cytometry

Cryopreserved PBMCs from 0 h to 14 days were thawed, washed with cold RPMI-1640 (Gibco) supplemented with 10% FCS (Sigma-Aldrich) and labeled on ice with fluorochromeconjugated antibodies (Table S1 in Supplementary Material) and live/dead stain in FACS buffer (PBS supplemented with 2% FCS and 2 mM EDTA; all Sigma-Aldrich). Cells were then washed with FACS buffer, fixed at room temperature in PBS containing 2% formaldehyde (Polysciences Inc.), permeabilized at room temperature in PBS with 0.1% Triton X-100 (Sigma-Aldrich) followed by intracellular staining performed in FACS buffer at room temperature. Samples were acquired on an LSRII Fortessa (BD Bioscience) and data analyzed and visualized using FlowJo software (v9.9.4; FlowJo, LLC), and GraphPad PRISM (v5.0). Gating strategy is shown in Figure S3 in Supplementary Material.

### RESULTS

### Overall Changes in Gene Expression in Response to Vaccination

Changes in expression at any time point following vaccination compared to baseline (0 h) were identified for 529 genes (adjusted *p*-value < 0.05, Table S2 in Supplementary Material), for which the peak number of uniquely expressed DEGs was observed at 1 day post-first immunization (206 transcripts; **Figure 1A**). The majority of total DEGs were observed within 24 h post vaccination (329) compared to later time points (59 at 3 days and 24 DEGs at 7 days post-immunization). Of the transcripts uniquely and differentially expressed at the different time points compared to baseline, the lowest number was observed at 7 days post-immunization (24), reflecting a return to baseline within the observational period*.* The majority of DEGs identified at 6 h were downregulated compared to baseline (92/141, 65%), while at all other time points, the majority of DEGs were predominantly upregulated compared to baseline (220/323, 68% at 1 day; 114/152, 75% at 3 days; and 79/93, 85% at 7 days) (**Figures 1B–E**).

The magnitude of observed group level transcriptional changes varied between time points compared to baseline (**Figures 1B–F**), and ranged between −1.132 log2FC and 2.220 log2FC with the largest spread at 3 days (−0.696–2.220 log2FC). A total of 108 DEGs were identified to illustrate the largest fold change compared to baseline (>|0.500| log2FC, 67 upregulated and 41 downregulated DEGs) (**Figure 1F**). Of note, 21 genes appeared to switch from downregulated at 6 h to upregulated compared to baseline at later time points for most individuals. These included genes presumably involved in general response to vaccination, such as those related to tissue regeneration and hemoglobin (e.g., *TMSB4X* and *HBE1*), calcium signaling (e.g., *CAMK2A* and *CAMK1G*), and vesicle-mediated transport (e.g., *NECAP2* and *NRSN2*) as well as specific immune-related genes (e.g., *IL17REL*, *GDF1,* and *HLA-C*) (27, 28).

### Functional Differences in Transcriptional Profiles following Vaccination

We next identified functional pathways perturbed by the vaccination by performing a two-class comparison in GSEA to identify enriched blood transcription modules (BTMs), first presented in Ref. (14). Transcripts were ranked according to a fold change difference relative to the 0 h baseline, and only BTMs with an adjusted

Figure 1 | Transcriptional changes in gene expression over time following vaccination. Subjects were intramuscularly immunized with CN54gp140 + GLA. In the Venn diagram, (A) the number of unique and shared differentially expressed genes (DEGs) in whole blood at 6 h, 1 day, 3 days, and 7 days after immunization (compared to baseline, 0 h) are shown, while the numbers and magnitudes (*y*-axis) of upregulated (red dots) or downregulated (blue dots) DEGs at 6 h (B), 1 day (C), 3 days (D), and 7 days (E), are depicted in volcano plots. An overview of all DEGs hierarchically clustered and arranged by time, are shown in a heat map (F), according to the color legend (upregulated DEGs in red, downregulated in blue, log2FC scale).

*p*-value < 0.05 were considered significant. BTMs with weighted average of logFC for the different BTM forming transcripts being upregulated or downregulated are defined as enhanced or repressed, respectively. The enriched BTMs were divided into the following three main categories: cell cycle regulation and signaling; innate immune cells and activation, and lymphocyte activation and signaling (**Figure 2**; Table S3 in Supplementary Material). Some of the BTMs involved in cell cycle regulation and signaling were enhanced as early as 6 h following vaccination and remained enhanced until 7 days post vaccination, including regulation of transcription, transcription factors (M213), and cell cycle ATP binding (M144) demonstrating the immediate systemic impact of vaccination on general cell cycle pathways. BTMs representing cell movement, adhesion, and platelet activation (M30) were repressed at all time points (**Figure 2A**). BTMs related to early immune responses, including toll-like receptors (TLR) and inflammatory signaling (M16), enrichment in neutrophils (I) (M37.1), and several BTMs related to enrichment in monocytes (M11.0, M118.0, S4) were significantly enhanced at 6 h compared to baseline (0 h), whereas most of these BTMs were not found to be significantly enriched at later time points (**Figure 2B**). BTMs belonging to lymphocyte signaling and activation, including T cell activation and signaling (M7.0, M19, S0), were mainly identified as enhanced at 1 and 3 days post vaccination. By comparison, the B cell modules (M47.0 and M69) were most highly enriched at 6 h followed closely by 1 and 7 days post vaccination (**Figure 2C**). These results indicate enrichment of BTMs related to cell cycle regulation and signaling as well as those related to innate and adaptive immune responses following a single vaccination with the GLA-adjuvanted CN54gp140 vaccine candidate.

### Early BTM Signatures Induced in Whole Blood Post Vaccination Correlate with Later Serum Antibody Responses

Next, we sought to study whether early transcript signatures correlated with later serum antibody responses. This was achieved by using GSEA to identify enriched BTMs in the transcriptomes of individuals grouped by their CN54gp140-specific antibody responses, at different time point post vaccination relative to the corresponding baseline. Individuals were divided statistically into high and low CN54gp140-specific serum IgM, IgG, or IgA antibody responders based on their serum antibody concentrations and separation into groups was validated using statistical tests as described above (Figure S2 in Supplementary Material).

Individuals categorized as high IgM responders (day 14) exhibited a gene signature marked by a repressed enrichment of NK cell-related genes (M7.2) at 3 and 7 days post-vaccination (**Figure 3A**). Genes including *KLRB1*, *KLRD1*, *NKG7*, *SLAMF7*, *TBX21,* and *IL2RB* contributed to the core NK cell enrichment and displayed downregulation compared to the baseline. In contrast, gene expression levels for low IgM responders at 6 h and 1 day exhibited enhancement in the BTM for "plasma cell surface signature" (S3) (**Figure 3A**), including genes *SLC44A1*, *KCNH1*, *CAV1*, *TXNDC15*, and *KRTCAP2*. The repression of two BTMs related to respiratory electron transport chain (mitochondrion) (M219, M238) marked early antigen-specific IgG and IgA responders, while enhancement of BTMs enriched for myeloid cells and monocytes (M81) at 6 h and 1 day post-vaccination were identified only in low IgG or IgA serum responders. Enhancement in the BTM, "integrin cell surface interactions" (M1.0) was also identified as another early BTM signature for low antigen-specific IgA responders.

Vaccinees were also categorized into high and low responders at later time points (up to 6 months post vaccination) based on their individual CN54gp140-specific serum antibody concentrations (Figure S2 in Supplementary Material). Using this analysis, no specific BTMs were identified using the gene expression data from individuals classified as high or low serum IgG or IgG1 responders. However, BTM-based gene signatures were identified which differentiated late high serum IgA, IgG3, and IgG4 responders from low responders (**Figure 3B**). Notably, three NK cell-related enriched BTMs (M7.2, M61.0, and S1) were significantly repressed in the gene expression profiles from individuals categorized as either late high serum IgA or IgG4 responders, including downregulation of many of the same genes indicated for early high IgM responders. Different BTMs were also significantly enriched post-vaccination for individuals classified as late low IgA, IgG2, and IgG4 responders (**Figure 3B**). BTM module "Enrichment in myeloid cells and monocytes (M81)", which was identified as significantly enhanced for low early serum IgG and IgA responders, was also identified as significantly enhanced for late low serum IgG2 antibody responders.

### Enhanced Frequency of CD3– CD56dim NK Cells following Vaccination

Given tendencies for vaccination-induced enrichment in NK cell-related BTMs and significant associations between such NK cell-related BTMs modules and serum antibody responses, we next examined peripheral blood NK cells by flow cytometry in samples from subjects at 0 h baseline and at 14 days post first vaccination. Samples from intermediate time-points were not available. To determine whether observed changes in NK cellrelated BTMs could be attributed to change in NK cell numbers or expression levels of proteins upon vaccination, we stained PBMC for lineage markers, NK cell receptors, and an array of proteins within regulated BTMs. Such proteins included transcription factors EOMES and HOPX, granule constituent's *GNLY* and NKG7, as well as surface receptors CD161 (*KLRB1*), NKG2A (*KLRC1*),

Figure 2 | Enrichment of blood transcription modules (BTMs) regulated upon vaccination. Spider web chart indicating enriched BTMs at 6 h (red), 1 day (yellow), 3 days (blue), and 7 days (green). BTMs are functionally divided into: cell cycle regulation and signaling (A), innate immune cells and activation (B), and B and T cell activation and signaling (C). Only BTMs with FDR *q* values < 0.05 at at least one of the four time points (those time points are marked with asterisk) are considered. Points inside the dashed blue circle are negative enrichment and points outside the blue circle are positive enrichment.

NKG2D (*KLRD1*), NKp80 (*KLRF1*), and CRACC (*SLAMF7*). Results showed an overall trend of increase in the frequency of CD3<sup>−</sup>CD56dim NK cell population at 14 days post vaccination compared to the 0 h baseline (**Figure 4**). Levels of EOMES, HOPX, GNLY, NKG7, CD161, NKG2A, NKG2D, NKp80, and CRACC were not increased in CD3– CD56dim NK cells at 14 days post vaccination (data not shown). In the limited number of analyzed samples, frequency of CD3– CD56dim NK cell population in the blood of high antibody responder subjects was increased on 14 days post vaccination compared to the 0 h baseline (**Figure 4A**). This was observed for all of limited number of the early IgM (**Figure 4B**), IgG, and IgA (**Figure 4C**) high responders, whereas three out of four (IgM), two out of four (IgG), and three out of five (IgA) of the low responders demonstrated a decrease of the CD3– CD56dim NK cell phenotype. Similar trends were observed when changes in NK cells frequencies between baseline and 14 days post vaccination were studied in vaccines defined as high versus low responders at later time points (**Figure 4D**).

### DISCUSSION

Herein, we employed a whole genome transcriptomics analysis combined with a systems biology approach to pinpoint early transcriptional signatures in whole blood of healthy volunteers

Figure 4 | Frequency of CD56dim NK cell was enhanced in high antibody responders 14 days post vaccination. FACS plots showing NK cell populations of representative low (no IgM 14 days and low IgG 168 days) and high (IgM response 14 days and high IgG 168 days) responder before (D0) and 14 days following vaccination (A). Average frequency of the CD56bright and CD56dim NK cell populations at 0 h and 14 days (B). Log2 fold change of CD56bright and CD56dim NK cell population divided into IgM responders (Y) and non-responders (N) Day 14 following vaccination (C). Log2 fold change of CD56bright and CD56dim NK cell populations divided into early (84–168 days) high and low responders (IgG, IgG1, IgG2, IgG3, IgG4, and IgA) (D).

vaccinated with the TLR4 agonist GLA-AF adjuvanted CN54gp140 HIV-1 vaccine candidate. We identified DEGs and BTM signatures of the whole blood response to the vaccine candidate up to 7 days following vaccination, and the BTMs that showed correlation with early and late CN54gp140-specific antibody responses.

We observed an increase in the number of DEGs compared to the baseline as early as 6 h post-vaccination, which peaked at 1 and 3 days followed by a decline to the expression levels comparable to that of baseline by 7 days. The observed kinetics of the transcriptional changes in whole blood is in line with previous studies (15, 29). Notably, transcriptional analysis of the TLR4 agonist GLA adjuvant in mice revealed minimal gene regulation in blood compared with the site of injection (muscle) and draining lymph nodes (13), highlighting one of the major limitations of our and most human studies. A group of 21 genes were identified that demonstrated the greatest fold changes compared to baseline, and additionally appeared to switch from downregulated compared to baseline at 6 h to upregulated compared to baseline at later time points, for most vaccinated individuals. Several of these identified genes (such as *TMSB4X* and *HBE1*) were likely involved in the general response to vaccination, including hemoglobin genes and genes related to tissue regeneration. Of interest, all except three of the individuals had significantly increased HLA-C expression at D1, D3, and D7 relative to the baseline. HLA-C belongs to the classical MHC molecules with least variability and recently it has been shown to interact with the Env protein of HIV (30), the antigen component of the vaccine used in our study. HLA-C is known to interact with two main ligands, the CD8 T cell receptor and even more efficiently with NK cell receptors in an allele-specific manner, including: NKG2A/C immunoglobulin receptors (KIRs) and, to a lower extent, leukocyte immunoglobulin receptors (31, 32).

The functional pathway analysis of transcriptomics data with GSEA revealed BTMs related to general cell cycle activation, and innate immune cell activation at early time points, as well as BTMSs related to T cells, and B cell activation at the later time points. BTM offers an unprecedented possibility to functionally assess transcriptomic changes related to immunological responses in human blood (14). By statistically categorizing individual vaccines into high or low responder groups based on their CN54gp140-specific serum antibody levels, we identified several BTMs that were significantly enriched in either high or low responders for serum antibody classes and subclasses. In particular, we identified a repression in BTM modules related to NK cells, especially at 3 and 7 days post-vaccination, for high serum IgM, IgA, and IgG4 antibody responders. Our finding is somewhat similar to a recent report, where the BTMs enriched in NK cells and NK cell surface signatures (M7.2, M61.0, and M61.2) were negatively associated with antibody titers in subjects that received the malaria vaccine candidate RTS,S (33) at day 56 of their prime-boost regiment. Collectively, these data may suggest an involvement of NK cells in negatively regulating humoral responses (34).

Notwithstanding, the flow cytometry analyses of the subjects' PBMCs showed an increase in the frequency of CD3– CD56dim NK cells for 14 days post vaccination relative to the 0 h baseline. Unfortunately, the paucity of cellular samples from the subjects for 1–7 days post vaccination prevented our further study of this phenomenon. It is tempting to speculate that the repression of BTMs related to NK cells observed in the first 7 days post-vaccination reflects NK cells leaving the circulation early in the response. Given that NK cells are short lived, the enhanced frequency of NK cells for 14 days post vaccination is presumably attributed to secondary induction of NK cell differentiation processes in response to vaccination. NK cells have been shown to inhibit generation of long-lived memory B cells by suppressing follicular helper T cells during the first days of infection in mice (34). NK cells have previously been suggested to play an important role in the adjuvanticity of GLA-AF and GLA in squalene emulsion (SE) *via* rapid production of IFN-γ (35). It has also been hypothesized that vaccine-induced HIV-specific antibody-dependent cellular cytotoxicity mediated by healthy NK cell activity could help in preventing HIV infection (29). The vast majority of NK cells in the circulation is CD56dim NK cells and considered cytotoxic, while around 10% are CD56bright NK cells and known to produce cytokines (36, 37). Vaccination with the inactivated influenza vaccine has previously been shown to downregulate membrane surface NKp46 cells on CD3<sup>−</sup>CD56dim NK cells and increase NK cell-derived IFN-γ expression (38). Additionally, CD56bright NK cells have been shown to increase following vaccination with a hepatitis B DNA vaccine, with significantly increased expression of CD244 and NKG2D, and correlating with specific T cell responses (39). However, NK cells are associated with both positive and negative regulation of humoral responses (40) and hence it is unclear if the observed repression in NK cells related BTMs has a direct causal role in enhancing humoral response to CN54gp140 adjuvanted with GLA, or is a surrogate of other regulatory events.

In this respect, it is intriguing that low early serum IgG and IgA responders were associated with the BTMs "Enrichment in myeloid cells and monocytes (M81)." The observed enhancement of the BTMs for myeloid cells over those of NK cells in low responders may suggest potential differences in antigen handling and processing between low and high responders. Further, the enhancement in BTMs related to plasma cell enrichment in the low responders warrants further studies to pinpoint whether short-lived plasmablasts are being generated in preference to long-lived plasma cells in the low antibody responders.

In line with the enriched upregulation of early plasma cell signature at 6 h and 1 day observed in the IgM antibody low responders in our study, the yellow fever vaccine showed an inverse correlation of antibody responses with plasma cell and B cell BTMs. Conversely, the immune responses to the trivalent influenza vaccine (TIV) and the quadrivalent meningococcal conjugate vaccine showed a positive correlation to those BTMs. Recently, significant correlations between the vaccine-induced antibody responses and plasmablast signatures following influenza vaccination (20) and interferon signaling following yellow fever and live attenuated influenza vaccinations (17, 20) were reported. These transcript signatures, however, did not show a significant correlation with the vaccine-induced antibody responses in our dataset. This is presumably explained, at least in part, by different mechanisms of action of the live attenuated vaccines and the TLR4 agonist-adjuvanted protein vaccines. In keeping with this notion, Li et al. reported different transcript signatures associating with vaccine-induced immune responses when five human vaccines of different modalities were compared (14). Therefore, it is plausible that while some of the blood transcript signatures are shared among different human vaccines, vaccines with different mechanisms of action may possess differential transcript signatures. Notwithstanding, we could not nevertheless rule out the possibility that some of the identified BTMs, e.g., the IgA correlating BTMs "expression of respiratory electron transport chain" and "elevated integrin cell surface interaction" merely play a surrogate rather than a causal role in the antibody responses. Further studies are needed to pinpoint the possible role of correlating BTMs in the development and/or durability of antibody responses induced following vaccination in humans.

Altogether, we herein report potential transcript and BTM signatures of human whole blood in response to vaccination with the TLR4 agonist-adjuvanted CN54gp140 anti-HIV vaccine candidate. However, a few caveats should be applied to our findings. First, the small sample size suggests interesting observations that would require the execution of larger clinical studies to provide a greater statistical power. Second, PBMC samples from days 1–7 were not available for detailed analysis of changes in the number and phenotype of circulating cells, thus observed transcriptomic changes may reflect differences in mRNA in circulating cells and/ or influx/efflux of cellular populations in the systemic circulation. Nevertheless, these results provide new information on the candidate transcript signatures that may have potential as early biomarkers of antibody responses induced by TLR4 agonistadjuvanted HIV-1 Env vaccine candidates, and support the potential of systems approaches in providing new insights into underlying mechanisms of individual variation in response to vaccines in humans.

### REFERENCES


### ETHICS STATEMENT

This work was performed in compliance with UK Clinical Trial Regulations and all procedures were approved by the recognized Research Ethics Committee and by the Medicines and Healthcare Products Regulatory Authority. The clinical trial study is registered at ClinicalTrials.gov under registration no. NCT01966900.

# AUTHOR CONTRIBUTIONS

AH and RS conceived the project and designed the experiments. JA, TO, and JP conducted the whole blood RNA sample preparation and quality control, and together with MÖ and AH analyzed the transcriptomics data, and composed the manuscript. SK and PM executed the serology experiments and interpreted the data. All the authors provided critical feedback on the manuscript prior to publication and have agreed to the final content.

### ACKNOWLEDGMENTS

We would like to thank Heinrich Schlums and Yenan T. Bryceson of Department of Medicine, Center for Hematology and Regenerative Medicine, Karolinska Institute and Karolinska University Hospital, Sweden for the execution of flow cytometry experiments and data analysis. The authors are grateful to the Genomics Core Facility at Sahlgrenska Academy, University of Gothenburg for running the samples for microarray analysis.

# FUNDING

This project was supported by the FP7 European Commission ADITEC project (grant agreement no. 280873) and the Fondation Dormeur, Vaduz. Conduct of the clinical study was supported by the International AIDS Vaccine Initiative and the NIHR at Imperial College Healthcare NHS Trust. Provision of CN54gp140 and GLA-AF was supported through core funding from the Wellcome Trust *via* UKHVC (083844/Z/07/Z). RS is supported by the European Union's Horizon 2020 research and innovation program under grant the EAVI2020 consortium, agreement no. 681032. AH is supported by the Innovative Medicines Initiative, European Commission under the BioVacSafe (grant agreement no. 115308), VSV-EBOVAC (grant agreement no. 115842), and VSV-EBOPLUS (grant agreement no. 116068) consortia.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00301/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Anderson, Olafsdottir, Kratochvil, McKay, Östensson, Persson, Shattock and Harandi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Next-Generation Vaccines Based on Bacille Calmette–Guérin

*Natalie E. Nieuwenhuizen and Stefan H. E. Kaufmann\**

*Max Planck Institute for Infection Biology, Berlin, Germany*

Tuberculosis (TB), caused by the intracellular bacterium *Mycobacterium tuberculosis* (Mtb), remains a major health threat. A live, attenuated mycobacterium known as Bacille Calmette–Guérin (BCG), derived from the causative agent of cattle TB, *Mycobacterium bovis*, has been in clinical use as a vaccine for 90 years. The current incidence of TB demonstrates that BCG fails to protect sufficiently against pulmonary TB, the major disease manifestation and source of dissemination. The protective efficacy of BCG is on average 50% but varies substantially with geographical location and is poorer in those with previous exposure to mycobacteria. BCG can also cause adverse reactions in immunocompromised individuals. However, BCG has contributed to reduced infant TB mortality by protecting against extrapulmonary TB. In addition, BCG has been associated with reduced general childhood mortality by stimulating immune responses. In order to improve the efficacy of BCG, two major strategies have been employed. The first involves the development of recombinant live mycobacterial vaccines with improved efficacy and safety. The second strategy is to boost BCG with subunit vaccines containing Mtb antigens. This article reviews recombinant BCG strains that have been tested against TB in animal models. This includes BCG strains that have been engineered to induce increased immune responses by the insertion of genes for Mtb antigens, mammalian cytokines, or host resistance factors, the insertion of bacterial toxin-derived adjuvants, and the manipulation of bacterial genes in order to increase antigen presentation and immune activation. Subunit vaccines for boosting BCG are also briefly discussed.

Keywords: tuberculosis, *Mycobacterium bovis* bacille Calmette–Guérin, vaccine, recombinant *Mycobacterium bovis* bacille Calmette–Guérin, subunit vaccine, mycobacteria

# INTRODUCTION

The bacterium *Mycobacterium tuberculosis* (Mtb) remains one of the most difficult pathogens to control, and caused 10.4 million recorded cases of tuberculosis (TB) and 1.7 million recorded deaths in 2016 (1). Just under a quarter of the world's population is estimated to be latently infected with Mtb, with the highest prevalence in Africa and Asia (2). The risk of developing active disease for those with latent TB infection (LTBI) is greatest within the first 2 years and approximately 10%

*Edited by:* 

*Rino Rappuoli, GlaxoSmithKline (Italy), Italy*

### *Reviewed by:*

*Juraj Ivanyi, King's College London, United Kingdom Sima Rafati, Pasteur Institute of Iran, Iran*

### *\*Correspondence:*

*Stefan H. E. Kaufmann kaufmann@mpiib-berlin.mpg.de*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 27 October 2017 Accepted: 15 January 2018 Published: 05 February 2018*

### *Citation:*

*Nieuwenhuizen NE and Kaufmann SHE (2018) Next-Generation Vaccines Based on Bacille Calmette–Guérin. Front. Immunol. 9:121. doi: 10.3389/fimmu.2018.00121*

**71**

**Abbreviations:** Ag85B, antigen 85B; BCG, *Mycobacterium bovis* bacille Calmette–Guérin; CFP-10, culture filtrate protein-10; CT, cholera toxin; DCs, dendritic cells; ESAT-6, early secretory antigenic target-6; ESX-1, ESAT-6 secretion system 1; IL, interleukin; LC3, microtubule-associated protein light chain 3; LLO, listeriolysin O; LT, *Escherichia coli* heat labile toxin; LTBI, latent TB infection; MCP-3, monocyte chemotactic protein 3; MDR, multi-drug resistant; Mtb, *Mycobacterium tuberculosis;* NK, natural killer; NOX2, NADPH oxidase; Pfo, perfringolysin O; rBCG, recombinant BCG; RD(1-16), region of difference (1-16); ROS, reactive oxygen species; SCID, severe combined immunodeficiency; TB, tuberculosis; TCM, central memory T; TEM, effector memory T; Th1, T helper cell type 1; TRM, resident memory T; XDR, extremely drug resistant.

over a lifetime, or 5–10% per year in HIV infected individuals (3). Treatment with drugs requires 6–9 months of antibiotics, and not only multi-drug resistant (MDR) strains but also extremely drugresistant (XDR) strains continue to emerge, leading to extended drug treatment regimens with considerable side effects (4, 5). The HIV pandemic and socio-economic factors are the two major drivers of TB disease, with factors such as poor living conditions and sanitation, crowded housing, poor air quality, malnutrition, stress, and co-infections all increasing susceptibility to developing active TB disease (6). Improvement of socio-economic conditions along with development of a more effective vaccine against TB will be critical in controlling this devastating disease.

Almost 100 years ago, in 1921, the first newborn was immunized with a live attenuated strain of the bovine *Mycobacterium* species, *Mycobacterium bovis* bacille Calmette–Guérin (BCG), followed by mass vaccination campaigns (7). BCG is partially protective against TB and has immunostimulatory effects that reduce general mortality during the first years of life by enhancing responses to other infectious diseases such as respiratory viruses (8–10). However, the efficacy of BCG against TB varies geographically and BCG does not provide adequate protection against pulmonary disease, the main form of disease manifestation and the cause of transmission (1). The development of a more effective TB vaccine is therefore likely to play a profound role in controlling this disease. As a live vaccine, BCG can also cause local or systemic infection in immunocompromised individuals (11) and is thus contraindicated in individuals who stand to benefit most from vaccination, such as HIV-positive individuals who are at high risk of developing active TB. Hence, the development of a vaccine that is safer for use in immunocompromised individuals is also a high priority.

A number of TB vaccine candidates are under clinical development, and many more have been pre-clinically tested in animal models (12–15). Pre-clinical evaluation of novel vaccine candidates has improved our knowledge of protective responses against TB and has shown that as a standalone vaccine BCG is at least as effective as novel subunit vaccines (16). BCG continues to be used in countries where TB is endemic due to its partial efficacy and has an established safety record. Hence, two major strategies in TB vaccine development have been to generate live mycobacterial vaccines with improved efficacy and safety, such as recombinant BCG (rBCG) vaccines, or to boost BCG with subunit vaccines containing Mtb antigens. This review provides an update on the latest knowledge on BCG and summarizes the rBCG candidates that have been tested against TB in animal models or clinical trials.

### BCG AS A VACCINE AGAINST TB

Meta-analyses have found that BCG provides on average 50% protection against TB and is effective for 10–20 years, but efficacy varies between countries and is much lower in adults than in children (17–21). Absence of sensitization to environmental mycobacteria or prior Mtb infection is associated with higher efficacy of BCG against TB (18). BCG is particularly effective against TB meningitis and disseminated TB in infants, with protection against pulmonary TB being much lower (22). The original BCG developed at the Pasteur Institute in Lille, France, was distributed around the world, and continuing passaging led to accumulating genetic mutations and the divergence of numerous substrains (23). These substrains appear to vary in efficacy in animal models, which has been reviewed previously (23). It has been suggested that this could contributes to the variable efficacy seen in different studies; however, a meta-analysis suggests that the type of BCG substrain does not significantly affect efficacy (18). More strikingly, analyses found higher efficacy in colder countries such as UK and Norway and lower efficacy in warmer countries such as India and Indonesia (18, 19, 22, 24, 25). This variation in efficacy seems to be due to increased exposure to environmental mycobacteria, which appears to reduce reactivity to BCG (18, 26, 27). Prior infection with Mtb also reduces the efficacy of the BCG vaccine (18). People living in TB endemic countries are more frequently exposed to Mtb, which raises the risk of individuals being infected (28). The HIV pandemic has contributed to increasing the burden of TB (3). Other risk factors for TB disease include diabetes, smoking, alcoholism, indoor air pollution, chronic corticosteroid treatment, malignancy, and malnourishment (29, 30). Therefore, these factors probably also contribute to the failure of BCG to protect against disease in some individuals.

Humans are not the only species at risk of TB, as wildlife and farmed animals are also susceptible to infection with various mycobacterial strains. Two species of agricultural importance include *M. bovis* and *Mycobacterium caprae*, closely related species of the same clade that cause TB in cattle and goats, and can also be transmitted to humans (31). BCG was first tested and proven effective against virulent *M. bovis* infection in cattle by Calmette and Guerin in 1911 (32), 10 years before its delivery to a human newborn; however, it is not routinely given to cattle to avoid interference with diagnostic tests for *M. bovis* (32). Recently, it has been shown feasible to distinguish vaccinated from infected animals (33, 34). Trials vaccinating large animals with BCG and subunit vaccine boosters have demonstrated that BCG is more protective when administered to neonates and that subunit vaccines can boost protection after BCG prime [reviewed in Ref (32)]. Cows and goats vaccinated with BCG and exposed to natural infection were protected compared to those without vaccination (32). Efficacy was approximately 55–70%, similar to the estimated efficacy of BCG in humans in Norway and the UK (19, 35, 36).

Vaccines rely on the generation of memory responses, which result from the clonal expansion and differentiation of antigen-specific lymphocytes (37). T cells can differentiate into effector memory T (TEM) cells that have effector functions such as cytokine production and migrate to affected tissues, and central memory T (TCM) cells that home to secondary lymphoid organs, where they can proliferate and differentiate into new effector cells upon re-exposure to antigen. Immunity to Mtb requires CD4<sup>+</sup> and CD8<sup>+</sup> T cells producing type I effector cytokines such as IFN-γ and TNF-α and their recruitment to the lung (38–43). T cells that home to the lung tissue and accumulate there, known as tissue resident memory (TRM) cells, are particularly important for immunity to Mtb (38, 43). Mucosal immunization with BCG induces more TRM cells than parental immunization (43). In addition, it was recently shown that interleukin (IL)- 21-dependent memory-like natural killer (NK) cells generated after BCG vaccination were protective against Mtb challenge (44). Furthermore, innate immune responses are undoubtedly important in resistance against Mtb (45) and increasing evidence suggests a role for antibodies in protection (46–48). There is a fine balance between protective immune responses and excess immunopathology, and the type of immune responses induced is critical, particularly as Mtb can infect recruited myeloid cells such as neutrophils and myeloid-derived suppressor cells (49, 50). The immunology of Mtb infection is illustrated in simplified form in **Figure 1**. The strengths of BCG as a vaccine are that it induces immune responses against a broad range of mycobacterial antigens, boosts innate immunity by stimulating monocytes, persists for a relatively long time compared to non-live vaccines, and requires no adjuvants (51, 52). Its failure to provide sufficient protection against TB may be related to insufficient generation of CD8<sup>+</sup> T-cell responses and CD4<sup>+</sup> TCM cell responses, which are required for long-term protection against Mtb (41, 53–57). It has been proposed that BCG fails to provide long-term protection, as the rate of TB increases in early adulthood and some studies have shown waning of protection after 10–15 years (25, 58). However, recent meta-analyses suggest that BCG can be effective for 20 years or longer in some populations (19). This is similar to, or even better than, the duration of protection of other commonly used vaccines, which often require boosters every 10 years. However, boosting BCG with repeat doses does not seem to be effective (59). In individuals infected with Mtb or constantly exposed to environmental mycobacteria, vaccine-induced TCM cells are under increased pressure and constant exposure to antigens may deplete pools of TCM cells by stimulating them to differentiate into effector cells (56, 58).

In both mice and humans, the T-cell responses to BCG vaccination are dominated by effector or TEM cells rather than TCM

cells (57, 58). BCG resides in the phagosome of host cells, and its antigens are therefore primarily processed by major histocompatibility complex (MHC) class II pathways, stimulating CD4<sup>+</sup> T-cell responses (60). BCG is also a poor inducer of apoptosis, a process of controlled cell death, which promotes the induction of both CD4<sup>+</sup> and CD8<sup>+</sup> T cells (53, 55, 61). In mice, the loss of BCG-mediated protection in the chronic phase coincides with a loss of CD4+ TCM cells and an increase in terminally differentiated, dysfunctional T cells (41, 54). Mtb is a slow growing bacterium, and the chronic nature of the disease may lead to T-cell exhaustion—a progressive loss of T-cell function. Thus, an effective vaccine should induce large pools of memory cells that can replenish TEM cells. In humans, increasing mycobacterial load coincides with progressive impairment of Mtb-specific CD4<sup>+</sup> T-cell responses (62) (41, 54, 63, 64). As exhausted T cells can be restored by inhibiting programmed cell death protein (PD)-1 or stimulating toll-like receptor 2, it is possible that host-directed therapy to improve T-cell function could be used therapeutically in TB patients, but studies are still ongoing (65). Recently, it was shown that memory CD4<sup>+</sup> T cells recruited to the lung attenuated Mtb growth in the early stages of disease, but their interaction with Mtb-infected macrophages did not promote their continued proliferation, resulting in only transient protection followed by waning immunity (66). Furthermore, Mtb-infected dendritic cells (DCs) cannot efficiently present antigens and instead transfer antigens to bystander DCs in the lymph nodes, which present the antigens to T cells (67–69). This causes a delay in activation of memory T cells and their recruitment to the lung. Overcoming such obstacles to sterilizing immunity against Mtb would improve the efficacy of TB vaccines markedly (68).

Novel vaccines aim to increase the number and quality of TRM and TCM cells generated (13, 70–72). While it was first thought that only live vaccines could generate good TCM responses, novel adjuvants administered with Mtb antigens have also shown success in this regard (41, 70). In clinical trials of novel vaccine candidates, it will be important to monitor long-term protective efficacy in populations with different levels of exposure to Mtb (58).

### rBCGs AGAINST TB

A number of rBCGs have been generated and tested for immunogenicity and/or efficacy in animal models. To narrow it down, we will discuss primarily those that have been tested for protective efficacy against Mtb (see **Table 1**). This includes BCG strains that have been engineered to induce increased immune responses by insertion of genes for Mtb antigens, mammalian cytokines or host resistance factors, insertion of bacterial toxin derived adjuvants, and manipulation of bacterial genes in order to increase antigen presentation and immune activation.

### rBCGs Expressing Mtb Antigens

Analysis of the genetic differences between BCG and Mtb has determined that 16 genomic regions of difference (designated RD1–RD16) have been deleted from BCG substrains, although some substrains do contain RD2, RD8, RD14, and RD16 (104–106). RD1 to RD3 were the first to be identified and are present in virulent *M. bovis* as well as Mtb (104). RD1 is a 9.5-kb segment deleted from all BCG substrains but conserved in all virulent isolates of *M. bovis* and Mtb, and it regulates multiple genetic loci. The RD1 segment contains the genes for the immunodominant antigens early secretory antigenic target (ESAT)-6 and culture filtrate protein-10 (CFP-10), as well as components of the type VII ESAT-6 secretion system (ESX)-1 required to secrete them (73, 104, 107, 108). RD1 deletion and the loss of the ESX-1 secretion system was a major factor in the original attenuation of BCG (73, 104, 107). At least five additional T-cell antigens are encoded by RD1 (PE35, PPE68, Rv3871, Rv3878, and Rv3879c), suggesting that RD1 constitutes an immunogenicity island (108). RD2 is a 10.7-kb segment deleted only from substrains derived from the original BCG Pasteur strain after 1925, and it is conserved in *M. bovis* and Mtb. RD2 contains novel repetitive elements and the *mpt64* gene, encoding the protein MPT64 which elicits T-cell responses and delayed hypersensitivity reactions in Mtb-infected patients (109). Finally, RD3 is a 9.3-kb segment absent from BCG, present in virulent laboratory strains of *M. bovis* and Mtb, but absent from 84% of virulent clinical isolates. The loss of RD1 to RD16 means that BCG lacks a number of the antigens of Mtb, and attempts have been made to improve the efficacy of BCG against TB by generating rBCGs expressing antigens particular to Mtb, such as ESAT-6 (74, 75, 80, 81, 110). ESAT-6 not only acts as an antigen but also can induce IL-18-dependent IFN-gamma secretion by Mtb antigen-independent memory CD8<sup>+</sup> T cells and NK cells (111). Furthermore, rBCGs over-expressing antigens that are found in both BCG and Mtb, such as antigen 85B (Ag85B), have been generated in an attempt to boost immune responses against shared mycobacterial antigens. In support of the fact that including Mtb antigens improves the protective efficacy of mycobacterial strains, removing CFP-10 and ESAT-6 from the attenuated Mtb-derived MTBVAC strain reduced its protection against Mtb in mice to that of BCG levels (112). MTBVAC is an Mtb strain attenuated by independent deletions of the *phoP* and *fadD26* virulence genes.

An rBCG over-expressing ESAT-6 induced stronger IFN-γ responses than parental BCG but did not improve protection against aerosol Mtb challenge in guinea pigs (75). In contrast, rBCG30, over-expressing the shared immunodominant Ag85B, improved protection against Mtb in a guinea pig model relative to BCG (76). Ag85B is one of the three similar secreted mycolyltransferases that are important for bacterial wall synthesis (113). A Phase I clinical trial was completed where 35 adults were randomized to receive rBCG30 or parental BCG in a doubleblind fashion (77). The vaccine was well tolerated, and expansion and IFN-gamma production of Ag85B-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cells were increased in the rBCG30 immunized individuals compared to those immunized with parental BCG (77). To improve the safety of rBCG30 for use in immunocompromised individuals, a new construct, rBCG(mbtB30) was developed, which has disrupted synthesis of the siderophore mycobactin, preventing normal iron acquisition (78). This strain is mycobactin-dependent but can undergo limited replication if sufficient ferric mycobactin is provided. It was shown to be safer than BCG in immunocompromised severe combined immunodeficiency

Nieuwenhuizen and Kaufmann

5

Frontiers in Immunology


www.frontiersin.org

Table 1 | Recombinant bacille Calmette-Guérin (rBCG) vaccine candidates and their protective efficacy against *Mycobacterium tuberculosis* (Mtb) challenge.


(*Continued*)

Next-Generation Vaccines Based on BCG



(SCID) mice and was slightly more protective against Mtb than parental BCG in guinea pigs (78).

Other rBCGs expressing Mtb antigens that provided increased protection against Mtb in mice include rBCG expressing Ag85A (79), rBCG expressing Ag85B and latency antigen HspX (rBCG:XB)(80), and rBCG expressing MPT64 fused to the PE domain of the PE\_PGRS33 protein of Mtb that localizes to the cell wall ((H)PE-ΔMPT64-BCG) (81). In another approach, co-administration of an rMTb72F fusion protein (composed of Mtb32 and Mtb39 antigens) in an AS02A adjuvant together with BCG increased survival and decreased lung pathology in guinea pigs (110). A small number of antigens feature frequently in the makeup of vaccine candidates generated for pre-clinical and clinical evaluation, with six of the eight subunit vaccines in clinical trials containing an Ag85 protein (14). Current research therefore aims to identify novel antigens and broaden antigen selection for vaccine design.

### rBCGs Complemented with ESX-1 Variants

As mentioned in the previous section, BCG lacks the RD1 and thus a functional ESX-1 secretion system and the antigens ESAT-6 and CFP-10, which form a heterodimeric complex and act as virulence factors (73, 104, 107, 108). ESX-1 is involved in host–pathogen interactions, as access of bacterial antigens to the cytosol influences both bacterial virulence and host immune recognition (73, 83, 114). An rBCG complemented with the complete RD1 locus (BCG:RD1, also known as BCG:ESX-1) provided better protection against Mtb challenge in both mice and guinea pigs than parental BCG, with reduced lung pathology and dissemination of bacteria (74). However, insertion of the Mtb *esx-1* locus into BCG leads to increased virulence in immunodeficient mice and prolonged persistence in immunocompetent mice (74). This issue was approached by introducing mutations into the *esxA* gene (encoding ESAT-6) of BCG:ESX-1 (82) or by inserting the *esx-1* locus of *Mycobacterium marinum*, which has reduced virulence, into BCG (83). A BCG:ESAT-L28A/L29S strain carrying modifications at residues Leu(28)-Leu(29) of the ESAT protein was strongly attenuated in mice and demonstrated protective efficacy against Mtb challenge in mice and guinea pigs, with similar lung bacterial burdens but a strong reduction in spleen bacterial burdens (82). BCG:ESX-1Mmar increased protective efficacy compared to BCG in mice, demonstrating similar immunogenicity and efficacy to BCG:ESXMtb but was as safe as parental BCG (83). Compared to BCG Danish, BCG:ESX-1Mmar reduced bacterial loads in lungs and spleens by an additional log after a challenge with Mtb HN878 (Beijing family) and Mtb strain M2 (Harlem family). Expression of ESX-1Mmar induced the cGas/STING/TBK1/IRF-3/type I interferon axis and enhanced AIM2 and NLRP3 inflammasome activity, leading to increased proportions of mycobacteria-specific cytokine-producing CD4<sup>+</sup> Th1 cells and CD8<sup>+</sup> T cells.

# rBCGs Expressing Host Immunomodulatory Molecules

In an effort to increase host responses to mycobacterial antigens following BCG vaccination, a number of rBCGs have been generated expressing functional mammalian cytokines or other host molecules (84, 85, 87, 93). Cytokine-producing rBCG strains have also been generated with the aim of testing them in intravesical BCG immunotherapy for bladder cancer, which seems to require the activation of Th1 responses for efficacy (86, 115, 116). Most murine cytokines tested could be produced and secreted by BCG and were functional (84). Splenocytes stimulated with IL-2 secreting BCG produced increased IFN-γ compared to those stimulated with BCG (85). A study tested rBCGs expressing murine IL-4, IL-6, GM-CSF, IFN-γ, and IL-2 and found that BCGs secreting IL-2, IFN-γ, or GM-CSF were more potent stimulators of responses to PPD than BCG, with splenocytes from mice immunized with these strains producing large amounts of IFN-γ (84). In addition, two independent studies demonstrated that rBCG strains expressing IL-18, a cytokine that acts in synergy with IL-12 to induce IFN-γ (88), promoted IFN-γ production by splenocytes in response to PPD (86, 87). In one of these studies, mice infected with rBCG-mIL-18 had decreased bacterial loads compared to mice infected with parental BCG (86), whereas in the other, they did not (87). None of the aforementioned strains were tested for protective efficacy against Mtb challenge; however, rBCGs expressing IL-2 or IL-18 were tested against infection with virulent *M. bovis* (88). In this study, rBCG/IL-18 induced less IFN-γ than BCG and had lower protective efficacy against *M. bovis* challenge compared to BCG. Despite inducing increased Th1 responses and protecting against intranasal challenge with BCG, immunization with rBCG/IL-2 did not increase protective efficacy against the virulent *M. bovis* challenge. Similarly, when IFN-γ-deficient mice were infected via aerosol with rBCG secreting murine IFN-γ, they had reduced bacterial loads and more differentiated granulomas compared to mice infected with the control rBCG containing vector only, demonstrating the potential to influence disease outcomes (89). However, subsequent challenge of IFN-γ−/<sup>−</sup> mice with Mtb demonstrated that rBCG secreting IFN-γ did not provide additional protection against Mtb compared to BCG-vector. In this study, IFN-γ was only produced locally and local production was insufficient for improving systemic immune responses. Overall, conferring upon BCG, the ability to secrete cytokines does not seem to enhance its protective efficacy, probably because of the amount of cytokine secreted and its locality.

In another strategy, rBCGs expressing combinations of cytokines and antigens, sometimes in the form of fusion proteins (90–92), have also been generated. The rBCG strain expressing the fusion protein Ag85B-ESAT-6-IFN-γ only slightly reduced bacterial burdens after Mtb challenge compared to BCG (92). rBCG-Ag85B-ESAT-6-TNF-α increased IFN-γ secreting cells, but protection was not measured (91). More promising was an rBCG expressing a fusion protein of Ag85B and IL-15, a cytokine important for the proliferation and survival of memory CD8<sup>+</sup> T cells (rBCG-Ag85B-IL-15), which was found to increase numbers of IFN-γ-producing CD44<sup>+</sup>CD4<sup>+</sup> and CD44<sup>+</sup>CD8<sup>+</sup> T cells and decrease bacterial burdens after intratracheal Mtb challenge, compared to BCG-Ag85B, although comparison to BCG was not performed (90).

Genes for other host-derived immunoregulatory molecules have also been inserted into BCG. In one study, the *Ipr1* (intracellular pathogen resistance 1) gene was inserted into BCG to produce rBCGi (93). *Irp1* is an IFN-regulated gene that enhances macrophage resistance to intracellular pathogens such as Mtb and *Listeria monocytogenes* and is not expressed in susceptible C3HeB/ FeJ mice (117). Vaccination of C3HeB/FeJ mice with rBCGi was more protective against Mtb infection than BCG, with decreased bacterial loads in the lungs and spleens of BCGi-vaccinated mice. Gene expression analysis of 113 immune-related genes demonstrated 20 differentially expressed genes with greater than twofold change between rBCGi and BCG-vaccinated groups. In another study, insertion of the gene for the chemokine monocyte chemotactic protein 3 (MCP-3) into BCG did not improve its efficacy, but increased its safety, since immunodeficient mice infected with rBCG(MCP-3) survived longer than mice infected with the parental BCG (94).

### VPM1002 and Second-Generation Derivatives

In previous reviews, we have discussed the development of VPM1002 (BCG *ΔureC*:*hly*) from the laboratory through to clinical trials in detail (118, 119). Essentially, VPM1002 is an rBCG that has been engineered to express the *L. monocytogenes* protein listeriolysin O (LLO) instead of urease C. Urease C inhibits acidification of the phagosome by converting urea to ammonia, preventing phagosome maturation (120–122). This activity inhibits trafficking of MHCII to the cell surface resulting in a reduced MHCII expression and antigen presentation (120). LLO is a cholesterol-binding, pore-forming protein that allows escape of *L. monocytogenes* from the phagosome (123). It requires acidic pH for optimal activity, which can be achieved by deletion of urease C (95). Expression of LLO in the rBCG causes perturbation to the phagosome (**Figure 2**) and leakage of bacterial DNA into the cytosol, triggering activation of the AIM2 inflammasome, and increased autophagy and apoptosis (124). Access of bacterial antigens to the cytosol increases its availability to the MHC class I antigen presentation pathway and promotes cross-presentation (69, 125). In mice, VPM1002 induces better protection, providing a strong increase in efficacy compared to BCG (42, 96). It also provides protection as a post-exposure vaccine (97). Furthermore, it is attenuated, with increased safety in both immunocompetent and immunodeficient mice (95, 96). Phase I and Phase II clinical trials have demonstrated its safety and immunogenicity in humans, including neonates, and a Phase II/III efficacy trial as a vaccine against recurrent TB has commenced (119).

While VPM1002 has progressed through the development pipeline, next-generation derivatives of this promising candidate have been developed and tested in pre-clinical studies, with the aim of further enhancing immunogenicity and/or safety (126). The most successful of these so far is VPM1002 *ΔnuoG*, which increased protection compared to its parental strain rBCG *ΔureC*:*hly* in mice while maintaining safety (96). The *nuoG* gene codes for a subunit of the non-essential respiratory enzyme complex NADH dehydrogenase type I and was identified as an anti-apoptotic virulence gene after gain-of-function screening using an *M. smegmatis* mutant library following by generation of Mtb H37Rv mutants (61). Deletion of *nuoG* from Mtb decreased its virulence and lead to increased apoptosis at day 14 post-infection. The mechanism was TNF-α dependent and *ΔureC:hly*.

relied on NADPH oxidase (NOX2)-dependent reactive oxygen species (ROS) (127). Subsequently, it was found that deletion of the *nuoG* gene from both BCG and VPM1002 enhanced protective efficacy against Mtb in mice, with VPM1002 *ΔnuoG* having strongly increased efficacy compared to BCG-vaccinated mice (96). Intriguingly, as well as increasing apoptosis, deletion of *nuoG* also increased the recruitment of the autophagy protein microtubule-associated protein 1A/1B-light chain 3 (LC3) to the live vaccine strains, suggesting that type I NADH dehydrogenase has an additional role in inhibition of autophagic pathways. LC3-associated phagocytosis, which involves conjugation of LC3 to phagosomes and subsequent phagolysosomal fusion, requires NOX2 activity and the production of ROS. Therefore, as type I NADH dehydrogenase inhibits ROS, it is possible that it also inhibits LC3-associated phagocytosis. However, similar persistence of VPM1002 and VPM1002 *ΔnuoG* strains in mice suggested that inflammasome or apoptosis-mediated mechanisms induced by both strains were primarily responsible for the strain attenuation. The increased efficacy of VPM1002 *ΔnuoG* compared to VPM1002 was associated with increased CD4<sup>+</sup> TEM, follicular T helper cells and germinal center B cells (96).

Another approach to improve the efficacy of VPM1002 was to express the latency antigens Rv2659c, Rv3407, and Rv1733c in this strain (98). Rv2659c, encoded by Mtb and not BCG, is expressed during nutrient deprivation, while Rv1733 and Rv3407, present in both BCG and Mtb, are expressed during hypoxia and reactivating disease in a mouse model of TB, respectively. Expression of the three latency antigens in the rBCGΔ*ureC*∷*hly* (pMPIIB01) strain improved long-term efficacy against challenge with a clinical Mtb strain, with reduced bacterial burdens compared to the parental strain in both lung and spleen at day 200 post-infection after intra-dermal vaccination of mice, as well as decreased lung pathology. The strain was not compared to BCG but showed a strong reduction in bacterial loads compared to unvaccinated mice.

Another second-generation construct, BCG *ΔureC*:*hly Δpdx1*, or VPM1002 *Δpdx1*, was generated in an attempt to further improve the safety of the vaccine so that it might be suitable for immunization of immunocompromised individuals such as HIV<sup>+</sup> infants and adults and HIV-exposed neonates, who are at higher risk of developing TB (99). VPM1002 *Δpdx1* is deficient in pyridoxine synthase, an enzyme required for synthesis of the essential vitamin B6, and therefore is auxotrophic for vitamin B6 in a concentration dependent manner. VPM1002 *Δpdx1* was profoundly attenuated, being safer in immunocompromised SCID mice compared to its parental strain. In addition, it demonstrated reduced dissemination in wild-type mice, which was partially reversed by dietary supplementation with vitamin B6. Immune responses to the strain were also dependent on vitamin B6 supplementation. Protective efficacy against Mtb was similar to BCG at day 30 but was lost by day 180; however, a homologous prime–boost regimen afforded similar protection to BCG. Protection depended on a dietary source of vitamin B6 at early time points following Mtb challenge, but protection at 180 days post-challenge with Mtb H37Rv and 160 days post challenge with Mtb Beijing/W remained independent from vitamin B6 supplementation, suggesting that it relied on immune responses generated at early time points.

As discussed earlier, there are several examples of rBCGs expressing mammalian cytokines, which promote increased immune responses (116). Therefore, Rao et al. attempted to further increase the immunogenicity of BCG *ΔureC*:*hly* by generating BCG *ΔureC*:*hly* derivatives expressing the cytokines IL-7 and IL-18 (100). IL-7 and IL-18 play a role in immunity to Mtb infection (128–130). IL-7 is required for T-cell development (131), and rIL-7 influences recall T-cell responses to Mtb infection (132). IL-18 induces Th1 responses (including IFN-gamma and TNF-α secretion), together with IL-12 (133). IL-18-deficient mice are susceptible to TB (128, 134) and BCG infection (134). Previously, splenocytes of mice vaccinated with rBCG expressing IL-18 produced higher amounts of Th1 cytokines after stimulation with mycobacterial antigens than splenocytes of mice vaccinated with parental BCG (87). Furthermore, increased IL-18 mRNA was detected after vaccination with BCG *ΔureC*:*hly* and BCG *ΔureC*:*hly ΔnuoG*, which have increased efficacy compared to BCG (96, 124).

Growth assays demonstrated that expression of IL-7 or IL-18 did not compromise intracellular fitness of BCG Δ*ureC*:*hly\_*hIL7 and BCG Δ*ureC*:*hly*\_hIL18 (100). Secretion of pro-inflammatory cytokines including IL-6, TNF-α, and G-CSF was increased in DCs infected with BCG Δ*ureC*:*hly*\_hIL18 compared to BCG, although in co-cultures T-cell activation was not influenced. Similarly, BCG Δ*ureC:hly*\_hIL18 immunized mice showed up-regulation of proinflammatory cytokines IL-6, KC, CCL5, IL-2, and G-CSF compared to those vaccinated with BCG. At day 60, all strains on the BCG *ΔureC*:*hly* background induced similar numbers of CD40Lexpressing CD4<sup>+</sup> T cells in the lungs, but BCG Δ*ureC*:*hly*\_hIL18 and BCG Δ*ureC*:*hly*\_hIL7 induced increased CD40L<sup>+</sup>TNF-α+ and CD40L<sup>+</sup>TNF-α+IFN-γ+ CD4<sup>+</sup> T cells compared to BCG Δ*ureC*:*hly* between 30 and 60 days post-vaccination. Efficacy, measured by bacterial loads, was comparable to the parental strain. Therefore, expression of hIL-7 or hIL-18 by VPM1002 did not further improve protection. Suggested reasons were poor secretion of hIL-7 and hIL-18 and overload of the mycobacterial export machinery due to the use of the same export system (PgroEL2-Ag85BSS) for both LLO and the cytokines.

VMP1002, VPM1002 *ΔnuoG*, and VPM1002 *Δpdx1* are currently being tested for safety and protective efficacy against *M. caprae* infection in goats by the Friedrich Loeffler Institute in Germany (Menge et al., unpublished data). Results should be available in 2018.

### AEREAS-401 AND AERAS-422

AERAS-401 is an rBCG expressing the cholesterol-binding cytoslysin perfringolysin O (Pfo), a pore-forming protein normally secreted by *Clostrididium perfringens* (101, 135). Pfo interacts with membranes at both low and neutral pH, although low pH enhances Pfo membrane binding, oligomerization, and pore formation (136). Generation of this BCG strain (BCG1331 Δ*ureC*:Ω*pfoAG137Q*) was accomplished by replacing the *ureC* gene with the *PfoAG137Q* gene under the control of the Ag85B promoter (101). AERAS-401 secreted biologically active Pfo, associated with lysis of the endosome compartment, and had a good safety profile in immunocompromised SCID mice. A second-generation derivative of AEREAS-401, AREAS-422 (research strain AFRO-1), was then generated by incorporating genes coding for Ag85A, Ag85B, and TB10.4, into AERAS-401 in order to combine increased access of antigens to the cytosol with over-expression of Mtb antigens (101). AERAS-422 enhanced immune responses in both mice and guinea pigs compared to BCG. A short-term challenge experiment with a laboratory strain of Mtb in mice revealed no differences in bacterial loads in lungs and spleen after immunization with AERAS-422 compared with BCG, but challenge of vaccinated mice with the hypovirulent Mtb strain HN878 demonstrated increased survival after immunization with AERAS-442 compared to BCG (101). AERAS-422 was subsequently tested in Phase I clinical trials (137). It induced more potent immune responses than BCG, but immunization with AERAS-422 at the highest inoculum was associated with the development of shingles (varicella-zoster virus reactivation) in two of the eight healthy volunteers, and the study was discontinued. Whole blood stimulation with BCG demonstrated that both of the volunteers who developed shingles displayed five- to tenfold higher IFN-γ responses compared to the other recipients, and it was suggested that the effects of IFN-γ on type I IFN responses (required for antiviral immunity) should be investigated. In the trial, earlier and more robust NK and cytotoxic T-cell responses correlated with increased mycobacterial growth inhibition, suggesting that NK cells and cytotoxic T cells may serve as a target for improved vaccines against TB. In contrast, increased expression of myeloid and pro-inflammatory genes was negatively associated with mycobacterial growth inhibition.

# rBCGs Expressing Bacterial Toxins

Bacterial toxins and toxin derivatives possess immunostimulatory properties and activate immune responses to bystander antigens when present simultaneously, but their toxicity renders them unsuitable for use as adjuvants in humans (138). An example is cholera toxin (CT), which is a potent mucosal adjuvant but is not suitable for use in humans as its inflammatory nature induces adverse events (68). Mice immunized with rBCG expressing CT B developed increased levels of anti-BCG IgA and IgG responses compared to those immunized with parental BCG, associated with increased TGF-β production (139). In another study, CT enhanced IL-17 when administered together with BCG, and this was associated with increased protection against Mtb challenge (68). While CT cannot be used in humans, the study highlighted the potential role of mucosal adjuvants in protection against TB.

*Escherichia coli* heat labile toxin (LT) is another mucosal adjuvant, shown to promote antigen presentation, T-cell proliferation, cytokine production, and mucosal IgA and IgG responses (140). Detoxification of the A subunit by genetic modification results in a potent, non-toxic mucosal adjuvant with no toxicity in mice, guinea pigs, and macaques (138, 141–144). Detoxified LT has a good clinical safety record after oral and percutaneous administration, but nasal administration is not recommended because it is associated with an increased risk of transient peripheral facial nerve palsy (145, 146). Additional safety studies for other routes of administration would be prudent, considering the adverse effects after intranasal immunization were not detected in initial trials (144).

Recently, an rBCG (rBCG-LTAK63lo) was generated to express low levels of a non-toxic derivative of LT (LTAK63) (102). Vaccination with rBCG-LTAK63lo induced increased Th1 cytokines and IL-17 in the lung compared to BCG. After intratracheal challenge with a laboratory strain of Mtb, mice had greatly reduced bacterial burdens compared to BCG at day 30 post-challenge, and at a high challenge dose, mice immunized with rBCG-LTAK63lo had reduced bacterial loads and increased survival. rBCG-LTAK63lo also increased protection against challenge with a virulent Mtb Beijing isolate.

### BCG **Δ***zmp1*

Interleukin-1β is a major pro-inflammatory cytokine that is activated by cleavage of a pro-IL-1β precursor by caspases activated by assembly of the inflammasome, an inflammatory caspase-activating multi-protein complex (147). Mtb inhibits inflammasome activation and IL-1β processing by a mechanism involving the product of the virulence gene *zmp1*, a putative Zn(2+) metalloprotease (147). Accordingly, infection with Mtb deficient in *zmp1* triggers activation of the inflammasome, increased IL-1β

Table 2 | Boosters to bacille Calmette-Guérin (BCG).


(*Continued*)

### TABLE 2 | Continued


*The table illustrates some of the BCG boosters that have been tested against tuberculosis (TB) in animal models. Decreases in bacterial burdens were estimated from graphs if not specified and listed as follows for comparative purposes: up to 0.5 log decrease: slight; 0.5 to 1.0 log decrease: moderate; over 1.0 log decrease: strong. i.m., intramuscular; i.n., intranasal; s.c., subcutaneous.*

secretion, enhanced maturation of phagosomes, and improved mycobacterial clearance by macrophages. Furthermore, *zmp1* deficient Mtb was attenuated compared to the *zmp1<sup>+</sup>* parental strain, showing reduced bacterial burdens in the lungs after aerosol challenge. A BCG mutant strain deficient in *zmp1* also showed increased phagosome maturation and phagolysosome fusion (147, 148). This was demonstrated to facilitate antigen presentation and to increase mycobacteria-specific CD4<sup>+</sup> and CD8<sup>+</sup> T-cell responses, emphasizing that phagolysosome fusion is important for generating immune responses to BCG antigens (148). Two *zmp1*-deficient strains, BCG Pasteur Δ*zmp1*::*aph* and BCG Danish Δ*zmp1*, induced increased IFN-γ+CD4<sup>+</sup> T-cell responses in cattle compared to BCG (149). Efficacy testing in guinea pigs demonstrated that *zmp1*-deficient BCG strains were more protective than BCG, with BCG Pasteur SmR *zmp1*::*aph* and BCG Denmark Δ*zmp1*, further reducing lung bacterial loads compared to their parental BCG strains (103). Furthermore, the *Δzmp1* mutants showed increased safety in immunocompromised SCID mice compared to BCG.

### BOOSTING BCG WITH SUBUNIT VACCINES

Originally, it was thought that subunit protein vaccines had the potential to replace BCG, but evaluation of an extensive number of subunit vaccine candidates in animal models has demonstrated that at best, these match the protection afforded by BCG (150–152). In general, a survey of the literature shows that protection afforded by subunit vaccines is not as effective as that induced by live attenuated mycobacterial strains. However, subunit vaccines can increase protective efficacy when administered as boosters to a BCG prime and, thus, have an important role to play in TB vaccination strategies. Furthermore, testing does not necessitate withholding the BCG vaccine, which has been shown to be partially effective. Most people are vaccinated against BCG in infancy, and protection wanes after approximately 20 years (19). **Table 2** summarizes the results of preclinical efficacy testing of some of the subunit vaccines that have been tested as BCG boosters. In some cases, subunit vaccine boosters have been used with an rBCG prime. This is by no means an exhaustive list but serves to illustrate the type of subunit vaccines being tested. The majority of the subunit vaccine boosters that reduce bacterial burdens compared to BCG in mice or guinea pigs only reduce them slightly, but some have also shown beneficial effects on survival and lung pathology. The ability to induce high levels of cytokine-producing CD4<sup>+</sup> and CD8<sup>+</sup> T cells does not necessarily correlate with protection, as in some cases vaccine candidates were very immunogenic but did not reduce bacterial loads or pathology (75, 153).

## CONCLUSION

Bacille Calmette-Guérin can contribute to the control of TB, being most effective in children and those not previously infected with Mtb or sensitized to environmental mycobacteria (17–19, 22). However, BCG fails to adequately protect those in high-risk environments, where the prevalence of Mtb is high and socioeconomic conditions are poor. A wide variety of rBCG strains and subunit vaccines have been tested in pre-clinical trials. Only one rBCG is now in clinical trials (VPM1002), while other vaccine candidates in clinical trials include inactivated whole cell vaccines, attenuated mycobacterial strains and subunit vaccine boosters focusing mostly on a narrow range of antigens (Ag85 family, ESAT-6) (14, 15). Head-to-head testing of vaccine candidates in pre-clinical models may be useful for identifying the most promising candidates worth moving forward to clinical trials (152). A number of studies have revealed that the immunogenicity parameters measured often do not translate to increased efficacy, and hence pre-clinical trials should always include Mtb challenge (75, 153). Statistical analysis should be performed comparing novel candidates against BCG, the current clinical vaccine with proven partial efficacy in humans. In summary, it seems likely that an improved vaccination regimen against TB can be achieved, but overcoming the current limits in protective efficacy will require novel approaches.

### AUTHOR CONTRIBUTIONS

Both authors wrote the manuscript and approved it for publication.

### ACKNOWLEDGMENTS

We thank Christian Goosman for transmission electron microscopy and Diane Schad for graphic design. We are grateful for the administrative support of Souraya Sibaei and Katja Grunow. This work was supported by The European Union's Seventh Framework Programme (EU FP7) ADITEC (HEALTH-F4-2011-280873), by the EU Horizon 2020 project TBVAC 2020 (grant no 643381), and the Bundesministerium für Bildung und Forschung (BMBF) project "Infect Control 2020" (grant no 03ZZ0806A) to SK.

# REFERENCES


since vaccination in Norway: a retrospective population-based cohort study. *Lancet Infect Dis* (2016) 16(2):219–26. doi:10.1016/S1473-3099(15)00400-4


DeltaureC::hly improves protection against tuberculosis. *MBio* (2016) 7(3):e00679-16. doi:10.1128/mBio.00679-16


via the constitutively active prfA\* regulon boosts BCG efficacy against tuberculosis. *Infect Immun* (2017) 85(9):e00245-17. doi:10.1128/IAI.00245-17


**Conflict of Interest Statement:** SK is co-inventor/patent holder of VPM1002. NN has no conflicts of interest.

*Copyright © 2018 Nieuwenhuizen and Kaufmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Immunoglobulin G1 Allotype Influences Antibody Subclass Distribution in Response to HIV gp140 Vaccination

*Sven Kratochvil1†, Paul F. McKay 1†, Amy W. Chung2 , Stephen J. Kent2,3,4, Jill Gilmour5 and Robin J. Shattock1 \**

*<sup>1</sup> Imperial College London, Medicine, London, United Kingdom, 2Department of Microbiology and Immunology, Peter Doherty Institute for Infection and Immunity, University of Melbourne, Melbourne, VIC, Australia, 3ARC Centre of Excellence in Convergent Bio-Nano Science and Technology, University of Melbourne, Melbourne, VIC, Australia, 4Melbourne Sexual Health Centre, Department of Infectious Diseases, Alfred Health, Central Clinical School, Monash University, Melbourne, VIC, Australia, 5 IAVI Human Immunology Laboratory, Imperial College London, London, United Kingdom*

### *Edited by:*

*David J. M. Lewis, Imperial College London, United Kingdom*

### *Reviewed by:*

*Johannes S. Gach, University of California, Irvine, United States Raffael Nachbagauer, Icahn School of Medicine at Mount Sinai, United States*

*\*Correspondence:*

*Robin J. Shattock r.shattock@imperial.ac.uk*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 October 2017 Accepted: 11 December 2017 Published: 20 December 2017*

### *Citation:*

*Kratochvil S, McKay PF, Chung AW, Kent SJ, Gilmour J and Shattock RJ (2017) Immunoglobulin G1 Allotype Influences Antibody Subclass Distribution in Response to HIV gp140 Vaccination. Front. Immunol. 8:1883. doi: 10.3389/fimmu.2017.01883*

Antibody subclasses exhibit extensive polymorphisms (allotypes) that could potentially impact the quality of HIV-vaccine induced B cell responses. Allotypes of immunoglobulin (Ig) G1, the most abundant serum antibody, have been shown to display altered functional properties in regard to serum half-life, Fc-receptor binding and FcRn-mediated mucosal transcytosis. To investigate the potential link between allotypic IgG1-variants and vaccine-generated humoral responses in a cohort of 14 HIV vaccine recipients, we developed a novel protocol for rapid IgG1-allotyping. We combined PCR and ELISA assays in a dual approach to determine the IgG1 allotype identity (G1m3 and/or G1m1) of trial participants, using human plasma and RNA isolated from PBMC. The IgG1-allotype distribution of our participants mirrored previously reported results for caucasoid populations. We observed elevated levels of HIV gp140-specific IgG1 and decreased IgG2 levels associated with the G1m1-allele, in contrast to G1m3 carriers. These data suggest that vaccinees homozygous for G1m1 are predisposed to develop elevated Ag-specific IgG1:IgG2 ratios compared to G1m3-carriers. This elevated IgG1:IgG2 ratio was further associated with higher FcγR-dimer engagement, a surrogate for potential antibodydependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) function. Although preliminary, these results suggest that IgG1 allotype may have a significant impact on IgG subclass distribution in response to vaccination and associated Fc-mediated effector functions. These results have important implications for ongoing HIV vaccine efficacy studies predicated on engagement of FcγR-mediated cellular functions including ADCC and ADCP, and warrant further investigation. Our novel allotyping protocol provides new tools to determine the potential impact of IgG1 allotypes on vaccine efficacy.

Keywords: allotype, G1m3, G1m1, HIV, vaccines

### INTRODUCTION

Antibodies are generally accepted to contribute to vaccine induced protection against many infectious diseases including HIV (1). However, the involvement of different antibody classes or subclasses is less clear, where their different structural properties affect functional immunity. While antibody binding fragments (Fabs) are critical to determining binding specificity and neutralization, the Fc

**87**

domain is the primary determinant for a wide spectrum of immunological functions mediated by the engagement of Fc-gamma receptors (FcγR) on a range of effector cells. These functions are regulated during an immune response through Immunoglobulin (Ig) subclass composition tailoring the selective interaction with FcγR on effector immune populations. Indeed, it has been suggested that Ig subclass composition may influence a wide range of Fc-mediated effector functions including antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) (2). In this respect Fc-mediated effector functions likely augment the potency of broadly neutralizing antibodies (3) and are critical to the function of non-neutralizing antibodies. The observed modest efficacy of the RV144 HIV vaccine trial showing vaccine elicited protection in the absence of neutralizing antibodies (4) has driven an intense interest in the role of FcγR effector functions in protection and control of HIV infection (5, 6). Here the antibody subclass distribution is likely to play a critical role where IgG1/IgG3 interact efficiently with most FcγR, while IgG2/IgG4 show reduced affinity for many FcγR. Indeed, divergent antibody subclass profiles have been associated with variable antibody effector functions among HIV-1 controllers, where levels of HIV-specific IgG1/3 were the major distinguishing factor (7), while IgG2/IgG4 HIV specific antibodies were associated with poorer overall antibody activity (8). In this respect, inter-individual variation in the antibody subclass response profiles to HIV-1 infection and/or vaccination provides significant challenges in the development of a globally effective vaccine.

In addition to sequence diversity of variable Fab domains and isotypic variation, IgG-subclass antibodies have been shown to exhibit polymorphic epitopes (IgG-allotypes), which can differ between individuals and ethnic groups (2). Inherited in a codominant Mendelian fashion, IgG-heavy chain allotypes are designated as natural genetic marker (Gm) together with the antibody subclass (e.g., G1m) and the allotype number (e.g., G1m3 or G1m1) (9). So far, a total of 4 G1m human allotypes: G1m17, G1m3, G1m1, and G1m2; two G1m alloallotypes: G1m27 and G1m28; and two G1m isoallotypes: nG1m17 and nG1m1 have been identified *via* serological typing (10). These define 7 G1m alleles: G1m17,1; G1m3; G1m17,1,27; G1m17,1,28; G1m17,1,27,28; G1m17,1,2; and G1m3,1; where the G1m1 allotype is common to all alleles except G1m3. The prevalence of these alleles broadly differs according to European, African or Asian ancestry. Most Gm allotypes are located in the Fc-region (CH2 or CH3) of antibodies, with the exception of G1m3 which is linked to amino acid changes in the CH1-region: expressing Arg rather than Lys at position 120 (2, 10). G1m3 also expresses unique amino acids at positions 356 (Glu) and 358 (Met) in CH3 as opposed to Asp/ Leu common to all G1m1 allotypes. While allotypes are encoded by one given Ig gene, some amino acid variations can be found in antibody chains of other isotypes (isoallotypes). For example, the amino acid residue Arg120, which corresponds to G1m3, is also found in antibodies belonging to the IGHG3 and IGHG4 allele family (10, 11).

Prior work has linked Gm allotypes in the Ig constant heavy G chain (IGHG) to augmented antibody responses against certain diseases (12–15). For example, IgG antibody responses against the hepatitis C virus envelope proteins E1E2 in a cohort of infected subjects with GM 1,17 5,13 and KM 1 phenotypes exhibit fourfold higher levels of E1E2-specific antibodies (13). Another study showed that IgG1 antibody levels to malaria vaccine-antigens were significantly higher in subjects with the GM 3 23 5,13,14 phenotype when compared to subjects lacking this phenotype (16).

Similar trends have been reported for IgG-subclass and specificity profiles in a cohort of elite HIV-1 controllers where HIV-specific IgG1 levels correlated with Fc-dependent effector functions and total plasma IgG1. Subsequently, it was argued that Gm allotypes could be responsible for variations in IgG1 concentrations and it was suggested that future studies should incorporate Gm allotyping protocols to account for this possibility (7). Furthermore, previous studies have demonstrated the involvement of Gm alleles in ADCC of cancer cells (17, 18). Taken together these studies indicate that Gm allotypes impact on antibody functionality (19) and provided a strong rationale for investigating the impact of IgG1-allotypes on the magnitude and functionality of vaccine-induced IgG-subclasses responses in the context of an HIV-vaccine trial.

Traditionally Gm phenotype has been determined *via* hemagglutination inhibition assays (HAI) using anti-Rh IgG antibodies of known allotypy, and polyclonal IgG of a selected allotype-specificity (e.g., antihuman G1m3). However, access to such reagents can be rate limiting and the approach is less amenable to high volume screening. Given the dominant role of IgG1 responses to HIV-1 envelope immunogens (20), we sought to determine the impact of IgG1-allotypy on the magnitude of induced responses. Here, we combined PCR and ELISA assays in a dual approach to determine the IgG1 allotype identity (G1m3 and/or G1m1) of clinical trial participants, using human plasma and RNA isolated from PBMC. Subsequently, the distribution of IgG1-allotypes formed the framework for assessing the effect of IgG1-allotypes on the magnitude and functionality of vaccine-induced antibody responses. Understanding how IgG1 allotype influences IgG subclass distribution in response to vaccination may prove an important consideration in the design and evaluation of vaccines strategies across ethnic groups.

### MATERIALS AND METHODS

### HIV Vaccine Trials

This study mainly builds upon findings from the previously published X001 clinical trial (21), registered at http://ClinicalTrials. gov under no. NCT01966900, EudraCT 2013-001032-22. In brief, a recombinant clade C HIV-1 envelope gp140 protein (CN54gp140) produced by Polymun Scientific (Klosterneuburg, Austria) to GMP specification, which has been reported to be immunogenic in a number of preclinical and clinical studies (22, 23), was used. The vaccine antigen CN54gp140 was administered intramuscularly into the deltoid muscle of the upper arm at a dosage of 100 µg CN54gp140 formulated with 5 µg GLA-AF [Glucopyranosyl Lipid A—Aqueous Formulation, Infectious Disease Research Institute, Seattle, USA (24)] in a total volume of 0.4 ml at weeks 0, 4, and 8 with a boost inoculation with the same material at either month 6 or 12. The trial was performed at the NIHR/Wellcome Trust Imperial Clinical Research Facility, Imperial College, London. The trial population was predominantly Caucasian, although subject ethnicity was not recorded as part of the trial. Samples were also obtained from the MUCOVAC2 study (23) registered with the UK Clinical Research Network (UKCRN) Number 11679, EudraCT 2010-019103-27. In brief, recombinant CN54gp140 was administered by intramuscular (IM), intranasal (IN), or intravaginal routes of administration in HIV negative female volunteers. Sera was obtained for this study from subjects receiving IM immunizations administered at the same dosage as the X001 trial (100 μg with 5 μg GLA-AF) with the same schedule (0, 4, and 8 weeks). The trial population was also predominantly Caucasian.

### Ethics Statement

The clinical trials generating serum and PBMC samples were conducted in compliance with UK Clinical Trial Regulations and any amendments, which include compliance with the principles of Good Clinical Practice, and the study abided by the principles of the Declaration of Helsinki. All volunteers provided written informed consent to participate in the trials on the basis of appropriate information and with adequate time to consider the information and discuss the trial with the Principal Investigators or their delegate. The trial proposal, the trial-specific information provided to volunteers, the consent form and substantial protocol amendments (if applicable) were reviewed by a recognized Research Ethics Committee and by the Medicines and Healthcare products Regulatory Authority (see EudraCT numbers above). All volunteers were made aware that they were free to withdraw without obligation at any time and that such an action would not adversely affect any aspect of their medical care or legal rights.

# PBMC Isolation

PBMC were isolated using density gradient separation from heparinized whole blood, used fresh (within 8 h of blood collection) or frozen in a mixture of fetal bovine serum (Sigma-Aldrich, St. Louis, MO, USA) and DMSO at a 90:10 ratio using a Kryo 560-16 rate controlled freezer (Planer, Sunbury-On-Thames, UK). PBMC were stored in vapor phase liquid nitrogen.

### Determining the IgG1-allotype Identity

Human RNA was isolated from PBMC with the RNeasy Mini Kit for RNA (Qiagen, UK) and transcribed into cDNA *via* oligo(dT)18 primers, using the maxima first strand cDNA synthesis kit (Cat: K1672, Thermo Scientific, UK). Subsequently, the template cDNA was used for a primary PCR for both G1m3 and G1m1 allotyping. While the primary PCR is sufficient to account for G1m1,17 alleles, G1m3-allotyping required a secondary PCR to take account of (and exclude) isoallotypes that can be present in IgG3 and IgG4 regions. Primers and PCR programs are detailed in the supplementary section (Tables S1–S5 in Supplementary Material).

# Agarose Gels

DNA products were separated by 1.2% agarose gel electrophoresis (100 V, 1 h) in 1× Tris acetate EDTA buffer. The agarose gel was stained with SYBR Safe (1 in 20,000), a suitable DNA ladder was loaded and 5× loading dye was added to each sample prior to loading the gel. DNA was visualized on a transilluminator.

### ELISA Protocol for IgG1-Allotyping G1m3-ELISA

Serum antibodies from clinical trial participants were assessed for the presence of the IgG1-allotype G1m3 *via* a novel ELISA protocol adapted from a previously published ELISA platform (21). This antibody recognizes both the G1m3 allele prevalent in Caucasian populations and the G1m1,3 allele prevalent in those with Asian ancestry. In brief, 96-well high binding MaxiSorp plates (Nunc) were coated with 100 μl/well anti-G1m3 (Cat: I5385-0.2ML, Sigma, UK), at a 1:5,000 dilution in PBS, overnight at 4°C. As reference material, standard human Igs, which were captured with a combination of anti-human kappa (Southern Biotech, Cat: 2060-01) and lambda light chain (Cat: 2070-01) specific mouse antibodies, were used. These capture antibodies were coated onto the plates overnight at 4°C and coated plates were washed four times with PBS-T before blocking with PBS supplemented with 1%BSA and 0.05% Tween-20. Following further washing, diluted serum samples were added to the precoated wells (generally between 1:10,000 and 1,000,000) and titrations of Ig standards were added to the kappa/lambda capture antibody coated wells at 50 μl/well and incubated for 1 h at 37°C. Plates were washed four times prior to the addition of antihuman IgG-HRP and incubated for 1 h at 37°C. Plates were washed four times and developed with 50 μl/well of KPL SureBlue TMB substrate (Insight Biotechnology, UK). The reaction was stopped after 5 min by adding 50 μl/well of 1 M H2SO4, and the absorbance read at 450 nm on a KC4 spectrophotometer.

### G1m1 ELISA

For the G1m1 ELISA the commercially available detection antibody anti-IgG1-Hinge-HRP (Cat: 9052-08, Southern Biotech) was used at a 1:5,000 dilution. Except for the detection antibody, the G1m1-ELISA is identical to the anti-IgG1-ELISA protocol previously published (21). This antibody does not bind to G1m3 allele, but does show cross reactivity with G1m1,3 allele prevalent in those with Asian ancestry, suggestive of recognition of the common G1m1 allotype.

## Customized Multiplex Dimer Assay for the Assessment of FcR-Binding

A customized multivariate multiplex assay was developed using a panel of gp140 antigens (Clade C: CN54, 1086, Clade A: UG37, Clade D: UG21–NIH AIDS Reagents) covalently conjugated to different magnetic fluorescent multiplex beads (Bio-Rad, AU) as described previously (25). Biotinylated dimeric Fc-gamma-Receptors (FcγRIIa-H131, FcγRIIIa-V158) were produced as previously described (26). The dimeric FcγR multiplex method has been previously published (21). Briefly, gp140 coupled microspheres (minimum 500 of each individual antigen bead set per well) was added to 1:100 plasma diluted in PBS + beads, incubating overnight at 4°C. HIVIG was used as a positive control to normalize across multiple replicates. Beads were washed using Kratochvil et al. IgG1 Allotype Influences Humoral Response

a Bio-Rad magnetic plate-washer (Bio-Plex Pro Wash station) and incubated for 2 h with biotinylated dimeric FcγR (1.0 μg/ml), then subsequently washed and incubated 1 h with streptavidin PE (1.0 μg/ml), washed again before resuspending in sheath fluid. A Bio-plex MAGPIX and Bio-Plex Manager software (Bio-Rad) was used to detect the median fluorescence intensity (MFI) for each bead set.

### Statistical Methods

Immunological analyses were based on the per protocol population that received all vaccinations. Appropriate comparative statistics are annotated in the text and/or figure captions. Statistical analysis was carried out using Prism 7.0a (GraphPad, CA, USA) or the R software (R3.3.2) for statistical programming (27). The non-parametric Mann–Whitney test was used to compare two groups and the non-parametric Spearman's rank correlation coefficient was used to interrogate correlative relationships between the distributions of HIV-specific IgG1 levels and Fc-receptor binding in X001 study participants. *p*-Values ≤0.05 were considered significant (\**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001).

### RESULTS

### Cross-Validation of a Novel ELISA Protocol for Rapid IgG1-Allotyping against a PCR Protocol

We wished to assess whether IgG-allotypes might be linked to differences in the IgG-subclass profile of Ag-specific antibody responses generated in the context of HIV-1 vaccination (28, 29). To pursue this, we developed novel PCR (Figure S1 in Supplementary Material) and ELISA protocols and combined them in a dual approach to determine the IgG1 allotype identity (G1m3 and/or G1m1) of clinical trial participants, using both human serum and mRNA. The optimized protocol was applied to determine the IgG1 allotype identity of 14 clinical trial participants from an HIV vaccine study utilizing a Clade C gp140 envelope protein (**Figure 1**; Figure S2 in Supplementary Material) (21). IgG1-allotyping X001 study participants revealed that 3/14 (21%) were homozygous for G1m1 and 6/14 (43%) homozygous for G1m3, with 5/14 (36%) carrying both alleles (heterozygous). The IgG1 allotype abundance mirrors previously reported most frequent alleles representative of Caucasian population enrolled in this study (G1m3; G1m 17, 1; and G1m17,1,2), the concurring results demonstrate the interchangeability of the PCR and ELISA protocol for IgG1-allotyping. The clear cross-validation of both assays formats, allowed for the determination of the IgG1-allotype abundance in two additional clinical trials (23, 30), from which only serum samples were available (Figure S3 in Supplementary Material).

### Implications of IgG1-Allotypes for the Analysis of HIV-Vaccine Induce IgG-Subclass Responses

The IgG1-allotype identities determined for the X001 study (21) provided a distinctive framework for the re-analysis of

HIV vaccine-induced IgG-subclass profiles of the 11 per protocol individuals included in the full immunological analysis (**Figure 2**). Most participants, homo- or heterozygous for G1m1, exhibited a trend for higher Ag-specific IgG1 concentrations when compared to homozygous G1m3-carriers (**Figure 2A**). There was also a potential trend for homozygous G1m3-carriers to have greater Ag-specific IgG2 responses following the fourth immunization (**Figure 2B**). By contrast there were no identified differences in Ag-specific IgG3 and IgG4 responses according to G1m1 allotype (data not shown). Although the X001 study (21) was not powered sufficiently to demonstrate a statistical significance between homologous G1m1- and G1m3-allele carriers, the apparent differences in Ag-specific IgG1/IgG2 levels warranted further investigation.

All participants with at least one G1m1-allele were then analyzed as a single group and directly compared with homozygous G1m3-carriers. The subsequent area under curve (AUC)-analysis across all time points helped to further unravel the effect of G1m1-allotypy on the magnitude of Ag-specific IgG1/IgG2 responses by revealing trends and potential differences between G1m1-carriers and homologous G1m3-carriers (*p* = 0.0823, **Figure 2C**). Furthermore, the highest Ag-specific IgG1 responses that occurred 14 days after the second IM and 14 days after the fourth IM were in G1m1-carriers (**Figure 2A**).

# Association of IgG1-Allotypes and Differences in Ratios of Antigen-Specific IgG1/IgG2-Levels following Serial Immunizations with CN54gp140

Following these initial observations, additional serum samples from the related MUCOVAC2 trial [EudraCT 2010-019103-27 (23)] were made available for IgG1-allotyping (Figure S2 in Supplementary Material). MUCOVAC2 is a predecessor study to

X001 and was designed to establish the optimal route and dosage of immunization with the candidate HIV-1 clade C CN54gp140 envelope glycoprotein vaccine (23). The timing (third IM, week8), immunogen (CN54gp140), and dose (100 μg) in the MUCOVAC2 trial was identical to week 8 (third IM) in the X001 study. Thus, it was possible to pool data from the two clinical trials for this selected time point, allowing for a follow-up analysis of the differences in Ag-specific IgG1/IgG2 ratios mediated by G1m1 (**Figure 3A**: homozygous, *n* = 4; **Figure 3B**: homo- and/or heterozygous, *n* = 6) and G1m3 (homozygous, *n* = 7).

Following the 3-immunization priming phase, volunteers homozygous for G1m1 (*n* = 4) had fivefold higher Ag-specific IgG1/IgG2 ratios in comparison to homozygous G1m3-carriers (*n*= 7, *p*= 0.0242, **Figure 3A**). However, no significant differences in Ag-specific IgG1/IgG2 ratios were observed when comparing both hetero- and homozygous G1m1-carriers with homozygous G1m3-carriers when using this larger data set (*p* = 0.1807, **Figure 3B**). It is important to note that the difference in IgG1/ IgG2 ratios when comparing homozygous G1m1 to G1m3 carriers likely reflects differences in magnitude of IgG1 responses, given there was little evidence for higher IgG2 responses at this time point (**Figure 2B**).

### Correlational Relationships between Fc**γ**R-Binding (Fc**γ**RIIa/Fc**γ**RIIIa) and Ag-Specific IgG1 Levels Were Determined with Respect to Different Combinations of G1m1 and/or G1m3 Alleles

Despite statistical limitations in study power, evidence was found to suggest that Ag-specific IgG1/IgG2 levels varied according to the IgG1-allotype of the HIV-vaccine recipients. To further elucidate the role of allotypic variations in antibody responses, the impact of IgG1-allotypes on the magnitude of Fc-mediated functions was investigated. Investigating these observations in the context of an HIV-vaccine trial was facilitated by the use of a novel assay, using FcγR-ectodomains for probing Fc-mediated functions (26, 31). This assay was chosen based on the ease of standardization across laboratories in comparison to the varied cellular models of ADCC and ADCP function. Correlational

FIGURE 4 | HIV-specific IgG1 levels (*x*-axis) correlate with Fc-receptor binding in X001 study participants. Correlation analysis for X001 study participants across all study time-points. (A) homozygous G1m1-carriers, (B) homozygous G1m3-carriers, and (C) individuals heterozygous for G1m1/G1m3. Below the graph Spearman's rank correlation coefficient *r*-values are shown (\*\*\*\**p* < 0.0001). MFI, median fluorescence intensity.

relationships between FcγR-binding (FcγRIIa/FcγRIIIa) and Ag-specific IgG1 levels were determined with respect to different combinations of G1m1 and/or G1m3 alleles in X001 study participants (**Figure 4**; Figure S4 in Supplementary Material). The engagement of FcγRIIa/FcγRIIIa dimers *via* CN54gp140 sepcifc serum antibodies correlated significantly (*p* < 0.0001) with Ag-specific IgG1 levels, irrespective of the IgG1-allotype combinations.

### DISCUSSION

A novel allotyping protocol was developed and employed to determine the abundance of IgG1-allotypes in HIV vaccine studies (X001 and MUCOVAC1 clinical trials). The allotype abundance was found to mirror those previously reported for Caucasian populations, in which the G1m3-allele is known to be predominant (10, 15, 32, 33). Interestingly, the frequency of donors homozygous for G1m3 (45.4%) reported for a larger study (570 community blood donors) coincided with the G1m3 distribution determined for the X001 study (29). It is important to note that abundance of different G1m phenotypes and allotype differ in other ethnicities (e.g., Asian and African populations) for which many of the larger HIV vaccine trials are currently conducted, and these protocols will need to be confirmed in these ethnicities in the future.

Immunoglobulin G1 allotype analysis provided a framework for regrouping and reanalyzing the X001 Ag-specific IgGsubclass data. Initial data analysis of individual Ag-specific IgG1 responses in study participants suggested a link between G1m1 carriers (homozygous or heterozygous) and elevated Ag-specific IgG1 responses when compared to homozygous G1m3-carriers. Thus, Ag-specific IgG1 concentrations from both homo- and heterozygous G1m1 carriers were grouped and compared with homozygous G1m3-carriers, revealing differences in Ag-specific IgG1 levels between the two study groups. The results align well with a previous study, in which different lower serum IgG and IgG-subclass levels were associated with G1m3, G3m5 allotypes in a study population of 157 Caucasian blood donors (34) and mirror the observations made by Lai et al. (7).

In contrast to this finding, HIV-1 vaccine recipients with at least one G1m1 allele appeared to have lower Ag-specific IgG2 levels in comparison to homozygous G1m3-carriers following four immunizations. Subsequently, the Ag-specific IgG1:IgG2 ratio was compared for time- and dose-matched samples from two additional clinical HIV-vaccine trials. The data suggest that vaccinees homozygous for G1m1 have elevated Ag-specific IgG1:IgG2 ratios compared to G1m3-carriers. It is likely that this difference is predominantly driven by higher IgG1 levels associated with the G1m1 allele, any additional contribution of enhanced IgG2 levels associated with G1m3-carriers would need to be confirmed in larger studies. Interestingly, homozygosity for the G1m3-allele is strongly linked to the G2m23-allele in Northern Europe (9). Previous studies have suggested that individuals homozygous for this allotype (G2m23) exhibit higher serum IgG2 titers and decreased antipolysaccharide IgG2 responses than those homozygous for G2m(..), heterozygotes having intermediate levels (9, 12, 19, 35). Further work would be required to determine any additional contribution of G2m23 to the differences in IgG1:IgG2 ratios observed here.

Immunoglobulin G2 antibodies are generally associated with responses to carbohydrate antigens, which may be advantageous for recognizing ENV glycans (2), however, IgG2 has also been shown to recognize protein antigens. Indeed, two studies have demonstrated a link between IgG2 antibodies to HIV Gag proteins and natural control of HIV infection (36, 37). Nevertheless, a related study failed to detect differences in Gag-specific IgG2 levels between HIV controllers and chronic progressors (20). However, in contrast to the two studies, that used viral lysates in Western blot assays, Banerjee et al. (20) used recombinant HIV antigens in ELISAs, implying that conformational variations in HIV antigen could have affected IgG2-detection (20).

Observations in this study that elevated levels of HIVspecific IgG1 and decreased IgG2 levels were associated with the G1m17-allele, support the hypothesis that allotypes could present useful Gms for the assessment of HIV-1 acquisition risk in vaccinated individuals. This concept was first tested on samples from the Step Study (17), that failed to show protective efficacy. This phase IIb proof-of-concept study, designed to assess the efficacy of the MRK Ad5 gag/pol/nef HIV vaccine, was terminated prematurely on the grounds of futility. However, there was an observed increased risk of HIV-1 acquisition in the vaccine group when compared to the placebo group (38, 39). Pandey et al. (17) went on to investigate whether this observed increased risk of HIV-1 acquisition could be linked to Ig allotype and revealed that the risk of HIV-1 acquisition was significantly increased in individuals positive for a combination of homozygous G1m17 and Km1 allotypes. Furthermore, the researchers found that subjects homozygous for FcγRIIIa (F-version) in the absence of G2m23 were more likely to become infected with HIV-1 (17). The significance of these findings in terms of immune function are unclear given the vaccine approach was predicated on eliciting cellular responses designed to control viral replication rather than protective antibodies. Nevertheless, they serve to highlight the potential association of Ig allotype with HIV acquisition.

In the present study, we applied a novel assay for probing the engagement of FcγRIIa/FcγRIIIa dimers to investigate potential links between IgG1-allotypy and Fc-effector binding profiles (26). This has potential implications for HIV vaccine research since the significance of Fc-effector functions has been highlighted by a large phase III HIV-vaccine trial (RV144), in which HIV-specific ADCC was associated with enhanced protection against HIV-1 acquisition (40, 41). A similar vaccine approach is now being pursued in the HVTN 702 trail evaluated in South Africa. The potential for IgG1 allotype to influence ADCC function is not without precedent, previously reported in prostate cancer where the capacity of NK cells to mediate ADCC against prostate cancer cells is influenced by interactions between different IgG1-allotypes and the corresponding FcγRIIIa variants (28). Similar to the RV144, we determined the interactions of serum antibodies specific to different HIV-1 clades, with FcγRIIa/FcγRIIIa-dimers and with respect to the IgG1-allotypes G1m1 and G1m3. Higher titers of FcγR-dimer binding by CN54gp140-specific serum antibodies were detected in G1m1-carriers as opposed to homozygous G1m3 carriers. Increased IgG1-titers that were associated with G1m1 are likely responsible for augmented binding of FcγR dimers given the higher affinity of IgG1 over IgG2. Indeed, IgG1 titers directly correlated with FcγR-binding irrespective of allotype (**Figure 4**). This trend for increased FcγR-dimer binding in G1m1-carriers by vaccine-induced serum antibodies was generally preserved against envelope proteins of different HIV-1 clades, implying that epistatic interactions between Fc-domains and FcγR could play an important role in HIV-1 vaccines. This notion is supported by another study that investigated the abundance of FcγR and Gm allotypes among HIV-1 controllers and non-controllers. The major finding was that among Caucasian Americans negative for the FcγRIIa allele, Gm21-positive individuals were seven times more likely to be HIV-1 controllers than non-carriers of Gm21, whereas this trend was not observed in the African American cohort (42). It would be interesting to investigate at the sequence level, if similar links could be found for G1m3 and G1m1 allotypes in HIV-1 vaccine recipients.

Although the results of this study are preliminary, they suggest that individuals homozygous for G1m3 exhibit lower levels of Ag-specific IgG1 and as a consequence lower FcγR-engagement in response to HIV-vaccination. If FcγR function is important for antibody-mediated protection then these individuals would be less well protected than those homozygous for G1m1. These results have important implications for the two ongoing efficacy studies of HIV vaccines (HVTN 702 and 705) predicated on engagement of FcγR-mediated cellular functions including ADCC and ADCP. Our novel allotyping protocol provides new tools to determine the potential impact of IgG1 allotypes on vaccine efficacy.

### ETHICS STATEMENT

The clinical trials generating serum and PBMC samples were conducted in compliance with UK Clinical Trial Regulations and any amendments, which include compliance with the principles of Good Clinical Practice (GCP), and the study abided by the principles of the Declaration of Helsinki. All volunteers provided written informed consent to participate in the trials on the basis of appropriate information and with adequate time to consider the information and discuss the trial with the Principal Investigators or their delegate. The trial proposal, the trial-specific information provided to volunteers, the consent form and substantial protocol amendments (if applicable) were reviewed by a recognized Research Ethics Committee (REC) and by the UK Medicines and Healthcare products Regulatory Authority (MHRA). All volunteers were made aware that they were free to withdraw without obligation at any time and that such an action would not adversely affect any aspect of their medical care or legal rights.

### AUTHOR CONTRIBUTIONS

RS, PK, and JG conceived the project. SK and PK designed and performed experiments, analyzed the data, and together with RS composed the manuscript. AC and SK designed and performed a customized Fc-dimer multiplex assay, and preprocessed the data, which was analyzed by SK and AC.

### ACKNOWLEDGMENTS

We thank the patients for their willingness to participate in this study.

### REFERENCES


### FUNDING

This study was funded by the International AIDS Vaccine Initiative (IAVI), and conduct of the clinical study was supported by the NIHR at Imperial College Healthcare NHS Trust. Provision of CN54gp140 and GLA-AF was supported through core funding from the Wellcome Trust *via* UKHVC (083844/Z/07/Z). We gratefully acknowledge Dormeur Investment Service Ltd. for providing funds to purchase equipment used in these studies. The Fc-functional antibody work was funded by the European Union's Horizon 2020 research and innovation program under grant agreement no. 681032 and an Australian NHMRC-EU collaborative grant #1115828.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01883/ full#supplementary-material.


routes in female volunteers; MUCOVAC2, a randomized two centre study. *PLoS One* (2016) 11(5):e0152038. doi:10.1371/journal.pone.0152038


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor declared a past coauthorship with the authors.

*Copyright © 2017 Kratochvil, McKay, Chung, Kent, Gilmour and Shattock. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Alessandra Gallinaro1 , Martina Borghi2 , Roberta Bona1 , Felicia Grasso2 , Laura Calzoletti2 , Laura Palladino3 , Serena Cecchetti4 , Maria Fenicia Vescio2 , Daniele Macchia5 , Valeria Morante2 , Andrea Canitano1 , Nigel Temperton6 , Maria Rita Castrucci2 , Mirella Salvatore7 , Zuleika Michelini1 , Andrea Cara1 \* and Donatella Negri2 \**

*1National Center for Global Health, Istituto Superiore di Sanità, Rome, Italy, 2Department of Infectious Diseases, Istituto Superiore di Sanità, Rome, Italy, 3VisMederi S.r.l., Siena, Italy, 4Confocal Microscopy Unit NMR, Confocal Microscopy Area Core Facilities, Istituto Superiore di Sanità, Rome, Italy, 5Center for Animal Research and Welfare, Istituto Superiore di Sanità, Rome, Italy, 6Viral Pseudotype Unit, Medway School of Pharmacy, University of Kent, Kent, United Kingdom, 7Department of Medicine, Weill Cornell Medical College, New York, United States*

### *Edited by:*

*Donata Medaglini, University of Siena, Italy*

*Reviewed by: Thorsten Demberg, Immatics Biotechnologies, Germany Rong Hai, University of California, Riverside, United States*

### *\*Correspondence:*

*Andrea Cara andrea.cara@iss.it; Donatella Negri donatella.negri@iss.it*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 14 October 2017 Accepted: 19 January 2018 Published: 05 February 2018*

### *Citation:*

*Gallinaro A, Borghi M, Bona R, Grasso F, Calzoletti L, Palladino L, Cecchetti S, Vescio MF, Macchia D, Morante V, Canitano A, Temperton N, Castrucci MR, Salvatore M, Michelini Z, Cara A and Negri D (2018) Integrase Defective Lentiviral Vector as a Vaccine Platform for Delivering Influenza Antigens. Front. Immunol. 9:171. doi: 10.3389/fimmu.2018.00171*

Viral vectors represent an attractive technology for vaccine delivery. We exploited the integrase defective lentiviral vector (IDLV) as a platform for delivering relevant antigens within the context of the ADITEC collaborative research program. In particular, Influenza virus hemagglutinin (HA) and nucleoprotein (NP) were delivered by IDLVs while H1N1 A/ California/7/2009 subunit vaccine (HAp) with or without adjuvant was used to compare the immune response in a murine model of immunization. In order to maximize the antibody response against HA, both IDLVs were also pseudotyped with HA (IDLV-HA/HA and IDLV-NP/HA, respectively). Groups of CB6F1 mice were immunized intramuscularly with a single dose of IDLV-NP/HA, IDLV-HA/HA, HAp alone, or with HAp together with the systemic adjuvant MF59. Six months after the vaccine prime all groups were boosted with HAp alone. Cellular and antibody responses to influenza antigens were measured at different time points after the immunizations. Mice immunized with HA-pseudotyped IDLVs showed similar levels of anti-H1N1 IgG over time, evaluated by ELISA, which were comparable to those induced by HAp + MF59 vaccination, but significantly higher than those induced by HAp alone. The boost with HAp alone induced an increase of antibodies in all groups, and the responses were maintained at higher levels up to 18 weeks post-boost. The antibody response was functional and persistent overtime, capable of neutralizing virus infectivity, as evaluated by hemagglutination inhibition and microneutralization assays. Moreover, since neuraminidase (NA)-expressing plasmid was included during IDLV preparation, immunization with IDLV-NP/HA and IDLV-HA/HA also induced functional anti-NA antibodies, evaluated by enzyme-linked lectin assay. IFNγ-ELISPOT showed evidence of HA-specific response in IDLV-HA/HA immunized animals and persistent NP-specific CD8+ T cell response in IDLV-NP/HA immunized mice. Taken together our results indicate that IDLV can be harnessed for producing a vaccine able to induce a comprehensive immune response, including functional antibodies directed toward HA and NA proteins present on the vector particles in addition to a functional T cell response directed to the protein transcribed from the vector.

Keywords: lentiviral vector, influenza, vaccine, antibody, T cell response

# INTRODUCTION

Improvements in existing delivery systems are a significant aspect to be considered in order to obtain more effective vaccines. Viral vectors represent an attractive platform for vaccine development due to their ability to effectively deliver antigens of interest into cells and to generate humoral and cellular mediated immune responses against the encoded transgenes. Integrase defective lentiviral vectors (IDLVs) represent a promising platform for immunogen delivery for vaccination purposes (1, 2). IDLVs are self-inactivating (SIN), non-integrating, and non-replicating vectors with high transduction efficiency both *in vitro* and *in vivo*. In contrast to parental lentiviral vector (LV), IDLVs are produced by incorporating a mutated form of the integrase (IN) protein in the recombinant LV, preventing integration and overcoming the risk of insertional mutagenesis. The loss of integration has been demonstrated in several murine models *in vivo* and *in vitro* (3). In the absence of integration, transgene expression is due to the unintegrated circular forms of the vector, which are maintained episomally in the target cells in the absence of cell division (4, 5). Only the transgene of interest is expressed from episomal IDLV in the absence of any other parental viral product. Dendritic cells and macrophages, the main cell types mediating the immune response, are non-dividing cells that are readily transduced by IDLV, eliciting the expansion of antigen-specific T cells (6, 7).

Over the course of the past decade, several reports have shown that a single immunization with IDLV-vectored antigens induces a persistent immune response both in murine and in simian models of immunization (1, 2, 8). Antigen presentation persisted for at least 30 days from immunization (9), suggesting that prolonged expression might be a unique characteristic of IDLV *in vivo*. In preclinical challenge experiments, a single immunization with IDLV expressing human papillomavirus-E7 tumor-specific antigen resulted in eradication of established tumors in mice, validating the ability to induce an effective T cell response (10). Administration of IDLV encoding antigens in murine models of West Nile Virus and malaria induced protective antibodies when challenged with the respective pathogens (11, 12). IDLV expressing the influenza virus nucleoprotein (NP) was protective against homologous and heterosubtypic influenza virus challenge (13). More recently we showed that immunization of *rhesus macaques* with IDLV expressing HIV-Env 1086.C gp140 induced broad and sustained immune responses up to 1 year from the immunization (8). Importantly, IDLV is under evaluation in clinical trials for cancer immunotherapy (ClinicalTrials.gov Identifier numbers: NCT02609984, NCT02122861, NCT02387125).

In addition to the potential for inducing a prolonged immune response due to expression of the vectored transgene from episomal DNA circles, IDLV can be harnessed as a cargo for delivering immunogens after incorporation into the vector's particles. This can be accomplished by fusion of foreign antigens with proteins incorporated into the lentiviral particles during particle assembly (14, 15) or *via* pseudotyping. Pseudotyping with heterologous viral glycoprotein envelopes is always used during LV production for allowing transduction of target cells or tissues (16). LV particles can be pseudotyped with a wide range of heterologous viral envelope proteins, including Influenza virus hemagglutinin (HA) (17–19). Recovered particles gain the tropism of the virus from which the envelope glycoprotein was derived. The most widely used envelope glycoprotein for pseudotyping LV is the vesicular stomatitis G glycoprotein (VSV.G), which allow broad and efficient transduction of target cells *in vitro* and *in vivo* (20). Importantly, the envelope glycoprotein displayed on the surface of the particles can elicit humoral immune responses that can be protective in animal models of immunizations (21–23).

Seasonal influenza A virus (IAV) infections cause significant morbidity and mortality worldwide and remain a major public health concern (24, 25). Currently licensed influenza vaccines elicit neutralizing antibodies (Abs) targeting HA, preventing influenza virus entry into cells (26). In particular, HAs from influenza A (H1N1) pdm09 virus circulating in humans are a major antigenic component contained in the annual vaccine formulations (27). However, seasonal vaccines do not protect against new mismatched strains and require frequent reformulation based on the prediction of strains that may circulate (27). Conversely, cell-mediated immunity targeting conserved antigens, such as influenza NP, is cross reactive and, although T cell immunity is unable to prevent disease, may contribute to improve clearance and decreased symptoms (28–30). NP is 90% conserved among influenza virus strains (31), and it is the major target of the cross-protective T cell response against influenza virus in the mouse model (32–35). However, while protection from influenza challenge in mice can be achieved in presence of NP-specific T cell responses (36), an efficient influenza vaccine for humans should be able to generate a more comprehensive and durable immune response in terms of both protective antibodies and effective T cells.

To this aim, in the present study, we evaluated the immune response induced by a multi-antigen IDLV-based influenza vaccine pseudotyped with HA protein from A/California/7/09 (H1N1 pdm09) virus and expressing either influenza HA or NP proteins as transgenes. Results indicated that a single immunization with multivalent IDLVs induced persistent and effective antiviral Abs and T cell responses directed toward HA or NP transgenes.

### MATERIALS AND METHODS

### Vector Construction

The SIN lentiviral transfer vector plasmid pTY2-NP expressing the nucleprotein (NP, GenBank: AAM75159.1) from Influenza virus A/PR/8/1934 (H1N1) has been already described (13). For construction of SIN lentiviral transfer vector plasmid pGAE-HA expressing the hemagglutinin (HA, GenBank: ACP44189.1) from influenza virus A/California/7/2009 (H1N1), the codon optimized HA open reading frame (ORF) was chemically synthesized, inserted into pUC57 plasmid (TwinHelix, Milan, Italy), excised with AgeI/SalI restriction enzymes and cloned into pGAE-green fluorescent protein (GFP) lentiviral transfer vector (37) by replacing the GFP coding sequence. The IN defective packaging plasmid, producing the viral proteins necessary for vector particle production, and the phCMV-VSV.G plasmid, encoding the pseudotyping vesicular stomatitis virus envelope glycoprotein G (VSV.G) from Indiana serotype, necessary for IDLV entry into target cells, have been previously described (38, 39). Plasmids pCAGGS-TMPRSS2 and pCAGGS-HAT, expressing the transmembrane protease serine 2 (TMPRSS2, GenBank: U75329) and the human airway trypsin (HAT, GenBank: AB002134) proteins, respectively, were described previously (40). Plasmids pCMV3-H1N1-C-NA (Sino Biological Inc., North Wales, PA, USA), expressing the neuraminidase (NA, GenBank: ACP41107.1) derived from influenza virus A/California/4/2009 (H1N1), and pCMV-HACal09, expressing the codon optimized HA protein from the CMV promoter, were used during production of HA-pseudotyped IDLV-NP/HA.

### Production of Lentiviral Vectors

Lenti-X human embryonic kidney 293T cell line was obtained from Clontech (Mountain View, CA, USA) and was used for IDLV production following established protocols (41). Cells were maintained in Dulbecco's modified eagles medium (Gibco, Life Technologies Italia, Monza, Italy) supplemented with 10% fetal calf serum (Corning, Mediatech Inc., Manassas, VA, USA) and 100 units/ml penicillin/streptomycin/glutamine (PSG) (Gibco, Life Technologies Italia, Monza, Italy). For production of IDLVs, pseudotyped with HA and expressing either HA (IDLV-HA/ HA) or NP (IDLV-NP/HA), 3.5 × 106 293 T Lenti-X cells were seeded on 10-cm Petri dishes (Corning Incorporated—Life Sciences, Oneonta, NY, USA) and incubated overnight. Cells were transiently transfected with lentiviral transfer vector expressing influenza HA or NP, along with IN defective packaging and VSV.G-envelope plasmids by Calcium Phosphate using the Profection Mammalian Transfection System (Promega Corporation, Madison, WI, USA) as previously described (13, 39, 41). Plasmids pCAGGS-TMPRSS2 or pCAGGS-HAT and plasmid pCMV3-H1N1-C-NA were included to express protease and NA during IDLV production. To pseudotype IDLV-NP/HA with HA protein, pCMV-HACal09 plasmid was added during IDLV-NP/HA production. After 48 and 72 h post-transfection, cell culture supernatants were collected, cleared from cellular debris by low-speed centrifugation and passed through a 0.45-µm pore size filter (Millipore Corporation, Billerica, MA, USA). To produce IDLV stocks for mouse immunization, vector containing supernatants were concentrated by ultracentrifugation (Beckman Coulter, Fullerton, CA, USA) on a 20% sucrose gradient (Sigma Chemical Co., St. Louis, MO, USA) at 23.000 rpm for 2.5 h at 4°C using a SW28 swinging bucket rotor (Beckman). Pelleted vector particles were resuspended in 1× phosphate-buffered saline (PBS, Gibco, Life Technologies Italia, Monza, Italy) and stored at −80°C until use. Each IDLV-HA/HA or IDLV-NP/HA stock was titred by the reverse transcriptase (RT) activity assay and the corresponding transducing units (TU) were calculated by comparing the RT activity to the one of IDLV-GFP virions with known infectious titers, thus allowing for the determination of their infectious titer units (42).

# Flow Cytometry and Confocal Laser Scanning Microscopy (CLSM)

293 T Lenti-X cells were plated in six-well microplates for flow cytometry analysis or seeded in 24-well cluster plates onto 12-mm cover glasses previously treated with l-polylysine (SIGMA) for CLSM. Cells were then transfected by calcium phosphate with pCMVHACal09 using the Profection Mammalian Transfection System (Promega). Twenty-four hours following transfection, non-permeabilized cells were stained with antiserum from CB6F1 mice immunized with purified H1N1 subunit vaccine (0.1 µg of HA per dose) from influenza strain H1N1 A/ California/7/2009 (provided by Novartis Vaccine & Diagnostics Srl, Siena, Italy) in the presence of MF59 adjuvant (provided by Novartis). After 45 min, cells were washed in PBS 1× and then incubated for 30 min with secondary antibody PE Goat antimouse IgG (Biolegend, San Diego, CA, USA), for flow cytometry analysis, or Alexa Fluor-488-F(ab′)2 fragments of goat antimouse (Molecular Probes, Life Technologies, Carlsbad, CA, USA), for CSLM analysis. For flow cytometry, cells were fixed with 1% paraformaldehyde and fluorescence was measured using the FACSCalibur (BD Biosciences, Milan, Italy), and data were analyzed using CellQuest (BD Biosciences). For CLSM, cells were fixed with methanol and the coverslips were mounted with Vectashield® antifade mounting medium containing DAPI (Vector Labs, Burlingame, CA, USA) on the microscope slides. CLSM observations were performed on a Leica TCS SP2 AOBS apparatus (Leica Microsystems, Wetzlar, Germany), using excitation spectral laser lines at 405 and 488 nm, using the confocal software (Leica, Wetzlar, Germany) and Photoshop CS5 (Adobe Systems, San Jose, CA, USA). Signals from different fluorescent probes were taken in sequential scanning mode, several fields were analyzed for each labeling condition, and representative results are shown. Images represented a single central optical section taken in the center of each cell nucleus and a 3D reconstruction.

### Western Blot

To evaluate HA presence on IDLV particles, pellets of IDLV concentrated preparations were resuspended in SDS loading buffer. Lysed virions were separated on 12% SDS polyacrylamide gel under reducing conditions and transferred to a nitrocellulose membrane (Sartorius Stedim Italy). Filters were saturated for 2 h with 5% nonfat dry milk in PBST (PBS with 0.1% Tween 20) and then incubated with HA antiserum for 1 h at room temperature followed by incubation for 1 h at room temperature with an antimouse horse radish peroxidase (HRP)-conjugated IgG (Sigma Aldrich, Milan, Italy). The immunocomplexes were visualized using chemiluminescence ECL detection system (Luminata Forte Western HRP Substrate, Millipore). H1N1 subunit vaccine from A/California/7/2009 virus was used as positive control.

### Mice and Immunization Schedule

CB6F1 female mice were purchased from Harlan (Harlan Laboratory, Srl, San Pietro al Natisone, Italy). Six- to eightweek-old CB6F1 mice, four mice per group, were injected once intramuscularly in the thigh with (i) 1 × 107 RT units/mouse of IDLV-HA/HA, (ii) 1 × 107 RT units/mouse of IDLV-NP/HA, (iii) H1N1 A/California/7/2009 (0.1 μg/mouse of HA protein, henceforth referred to as HAp) alone, and (iv) 0.1 μg/mouse of HAp in presence of MF59 adjuvant (HAp + MF59). Naïve, nonimmunized mice were kept for parallel analysis. All immunized mice were boosted with HAp alone 24 weeks after the prime. Antibodies (Abs) were measured in serum at different time points, starting from 2 weeks after the prime and up to 18 weeks after the boost. The cellular immune responses were analyzed at 4, 12, and 24 weeks after the prime in blood samples and in splenocytes at sacrifice (18 weeks after the boost).

Serum samples were obtained from blood collected from the retro-orbital plexus of mice with glass Pasteur pipettes and stored at −20°C until assayed. Heparin-treated glass Pasteur pipettes were used to collect blood in order to perform IFNγ ELISPOT assay. Leukocytes, obtained after ammonium chloride potassium (ACK) treatment of whole blood, were counted, suspended in RPMI 1640 (Gibco) containing 10% fetal bovine serum (Lonza, Treviglio, Milan, Italy), 100 units/ml of PSG (Gibco), non-essential aminoacids (Gibco), sodium pyruvate 1 mM (Gibco), HEPES buffer solution 25 mM (Gibco), and 50 mM 2-mercaptoethanol (Sigma Chemicals). Splenocytes were prepared by mechanical disruption and passage through cell strainers (BD Biosciences) and resuspended in complete RPMI medium, as previously described (39).

# IFN**γ** ELISPOT

The IFNγ ELISPOT assay was performed using the BD ELISPOT kit reagents and protocol (BD Biosciences). Briefly, blood or spleen derived cells were seeded at a density of 2.5 × 105 /well in 96 well plates and stimulated overnight either with 2 µg/ml of the H-2Kd restricted Influenza NP147–155 (TYQRTRALV) epitope or with 10 µg/ml of concanavalin A (Sigma Chemicals) used as a positive control. A 139 peptide array (15mers with 11 amino acid overlaps) spanning the entire HA from influenza virus A/ California/7/09 protein (BEI Resources, Manassas, VA, USA; Catalog No. NR-15433) was distributed in ten pools of 14 peptides each and used to identify the reactive epitopes on splenocytes. Complete medium treated cells were used as negative controls. Spot forming cells (SFC) were counted with an ELISPOT reader (A.EL.VIS, Hannover, Germany) and results expressed as number of IFNγ secreting cells (SFC)/106 cells. The samples were scored positive when a minimum of 50 spots per 106 cells were present and twofold higher than unstimulated sample.

### Measurement of Binding and Functional Antibodies

Sera were tested for the presence of binding Abs by a standard ELISA. Ninety-six well plates (Greiner bio-one, Kremsmünster, Austria) were coated with H1N1 A/California/4/09 subunit vaccine (0.2 μg/well of HA) overnight at 4°C. After washing and blocking, serial dilutions of serum from individual mice were added to wells in duplicate and incubated for 2 h at room temperature. The plates were washed and biotin-conjugated goat antimouse IgG (Southern Biotech, Birmingham, AL, USA) was added to the wells for 2 h at room temperature. The plates were washed before the addition of HRP-conjugated streptavidin (AnaSpec, Fremont, CA, USA) for 30 min at room temperature, followed by the 3,3′5,5-tetramethylbenzidine substrate (SurModics BioFX, Edina, MN, USA). Endpoint titers were determined as the reciprocal of the highest dilution giving an absorbance value at least equal to twofold that of background (biological sample from naive mice). For each group of immunization, results were expressed as mean titer with confidence interval.

Hemagglutination inhibition (HAI) Abs to A/California/7/09 virus were measured, according to standard procedures (43). Briefly, all sera were treated with receptor-destroying enzyme (Sigma Aldrich) to remove non-specific inhibitors of hemagglutination. Serial twofold dilutions of treated sera were mixed with 4 hemagglutinin units of the A/California/7/09 virus and, after 1 h incubation at room temperature, with 0.5% Turkey red blood cells. The HAI titers were expressed as the reciprocal of the highest dilution of serum that inhibited virus-induced hemagglutination.

The sera were tested by MDCK cell-based microneutralization (MN) assay (44) to titrate influenza virus-specific neutralizing antibodies. Briefly, all samples were heat-treated at 56°C for 30 min. Twofold serial dilutions of samples, starting from 1:10 dilution, were performed across the 96-well plates and then the viral working solution of A/California/07/2009 (H1N1) live virus was added (200 TCID50/100 μl). After incubation of mixture sera-virus at 37°C for 1 h, a cell suspension of 2 × 105 was added and the plates were incubated at 37°C for 5 days. Each plate was then checked under an optical microscope to assess the presence of local lesions and cytopathogenic effect. The neutralization titer (NT) for each serum was calculated according to the Spearman-Kärber formula.

Since NA plasmid was included during the IDLV preparation and NA protein may be present in subunit vaccine preparations, the heat-treated sera were also tested by enzyme-linked lectin assay (ELLA) (45) to measure the functional anti-NA Abs. Briefly, serial twofold dilutions of treated sera were mixed with antigen A/California/07/2009 (H1N1) treated with Triton X-100 as described elsewhere, in 96-well plates coated with fetuin. The plates were incubated at 37°C overnight (16–18 h). After washing the plates, the peanut agglutinin conjugate with peroxidase was added and the plates were incubated at room temperature for 2 h in the dark. Then the o-phenylenediamine dihydrochloride substrate was added and the plates were incubated at room temperature for 10 min in the dark. The reaction was stopped by adding 1 N H2S04. The optical density of all the test plates was read at 490 nm. The NA inhibition (NI) titer was expressed as the reciprocal of the last dilution that results in at least 50% inhibition of the maximum signal.

### Statistical Analysis

The temporal trend of antibody response (i.e., ELISA, HAI, MN, and ELLA) was analyzed using a system of piecewise linear equations in order to jointly evaluate the relationships of each outcome at each time point, allowing for correlated errors. The analysis was conducted in STATA13 (StataCorp LP, College Station, TX, USA) within a structural equation modeling frame work (46, 47). The log likelihood estimation procedure was used to fit the model to the data.

# RESULTS

### Production of IDLV Delivering Influenza HA

To produce IDLV expressing HA, 293 T LentiX cells were transfected with IN defective packaging plasmid, expressing the proteins required for IDLV assembly and release, the pseudotyping VSV.G-expressing plasmid, essential for IDLV entry into target cells, and HA-expressing lentiviral transfer vector plasmid, with the aim of pseudotyping the recombinant IDLV particles with the membrane tethered HA. Importantly, the presence of HA protein on the surface of the transfected cells was confirmed by flow cytometry and confocal microscopy (**Figures 1A,B**). However, the recovery of IDLV in the supernatant of the transfected cells was very low, indicating inhibition of release of IDLV pseudotyped with HA envelope (**Figure 2A**, left column). This was expected since, as with wild type influenza virus, NA protein is required for release of lentiviral vectors pseudotyped with influenza HA (17, 18, 48). To improve production of HA-pseudotyped IDLV, we cotransfected plasmids expressing NA protein, required for cleavage of surface sialic acid molecules on producer cells allowing release of the vector in the supernatant, and human serine transmembrane protease TMPRSS2 or HAT proteins, mediating proteolytic activation of influenza HA (18).

As shown in **Figure 2A**, recovery of IDLV-HA increased an average of eightfold in presence of TMPRSS2 but only twofold in the presence of HAT. Importantly, NA was fundamental for IDLV-HA release, increasing the amount of IDLV recovered in the supernatant an average of 24-fold over the IDLV produced in the absence of NA expressing plasmid. The inclusion of TMPRSS2 or HAT in addition to NA protein further improved, although not significantly, vector production (27-fold and 25-fold, respectively, compared to IDLV produced in the absence of NA and proteases). For producing concentrated IDLV preparations for immunization, TMPRSS2 was chosen over HAT protease expressing plasmid since IDLV production was generally higher and more consistent.

To produce IDLV expressing NP, vector was produced as described above and in the Section "Materials and Methods," by including a NP-expressing lentiviral transfer vector plasmid and a plasmid expressing the HA protein for pseudotyping the IDLV produced in the 293 T cells. In **Figure 2B**, is shown a representative western blot (WB) analysis of lysates from concentrated stocks of IDLV showing the presence of HA protein in all purified IDLV preparations.

### A Single Immunization with IDLV Induces High and Persistent Levels of Binding and Antiviral Neutralizing Antibodies

To mimic the immunization protocol of the seasonal influenza in humans, mice were given a single immunization and the immune responses were analyzed at various time points up to 24 weeks. As positive controls, groups of mice were immunized with H1N1 subunit vaccine alone (HAp) or in combination with MF59 (HAp + MF59). The schedule of immunization is described in **Figure 3A**.

Humoral response to H1N1 was assessed initially by ELISA. As shown in **Figure 3B**, 2 weeks after the prime all vaccinated animals developed anti-H1N1 IgG antibodies in serum. The

Figure 1 | Analysis of hemagglutinin (HA) expression. 293 T Lenti-X cells were transfected with HA expressing plasmid and stained at 24 h post-transfection for detection of HA on the plasma membrane as described in Section "Materials and Methods." Cells were fixed and expression of HA was quantitatively measured by flow cytometry (A) or observed by confocal laser scanning microscopy (B). (A) The percentage of HA-expressing cells is indicated. The overlay line (green) represents the fluorescence distribution of cells stained only with the secondary antibody. (B) Images represent single central optical sections (a–c) and a 3D reconstruction (d). Nuclei are colored in blue by DAPI staining and green color represents membrane associate HA protein. Scale bars, 8 µm. Untransfected cells (a) were used as negative control. Results from one representative experiment are shown for each analysis.

IDLV expressing HA. Vectors were produced as described in Section "Materials and Methods" and in the presence of plasmids expressing transmembrane protease serine 2 (TMPRSS2), human airway trypsin (HAT), or NACal09, as indicated in the graph. Data are expressed as the mean result from three independent experiments. The error bars represent the standard errors of the mean. (B) Western blot (WB) of lysates from concentrated stocks of IDLV pseudotyped with HA showing incorporation of HA into IDLV. Note that IDLVs were produced in the presence of TMPRSS2 protease, resulting in the cleavage of HA0 to produce HA1 (not visualized here) and HA2. HA protein (HAp\*, purified hemagglutinin vaccine subunit from influenza virus H1N1 A/California/7/2009) and IDLV-GFP were used as positive and negative control, respectively.

titers increased and persisted in all groups starting from 8 weeks after immunization. The highest titers were recovered in animals vaccinated with MF59, while the lowest levels were detected in the HAp group. No significant difference was observed between HAp + MF59 group and IDLV groups. Importantly, in both groups of IDLV vaccinated animals sustained Abs were induced at significantly higher titers than HAp alone (*p* < 0.01).

To assess the recall response, mice were boosted with HAp alone 6 months after the prime. Two weeks after the immunization with HAp all groups showed a boost in terms of binding Ab titers (**Figure 3B**). In particular, IDLV vaccinated animals were able to increase the anti-H1N1 Abs when boosted with HAp, indicating that immunization with IDLV enables the animals to respond to the HA antigen delivered in a different way, as a subunit vaccine in this case. The responses were persistent up to 18 weeks after the boost in all vaccinated animals. After the boost the differences between IDLV immunized animals and HAp group remained statistically significant (*p* < 0.01). Again, no significant difference in Ab titers between both IDLV groups and the adjuvanted group was seen.

To assess the functional antibodies, HAI assay, and microneutralization (MN) assay were performed at 8 and 19 weeks post priming and 2 weeks and 18 weeks post boost (**Figure 4**). HAI titers were present in all groups at 8 weeks postimmunization and were still detectable at 19 weeks post priming (**Figure 4A**). The response significantly increased after the boost in all groups. The kinetics of HAI titers mirrored the kinetics of binding Abs, showing the highest titers in the adjuvant treated group and the weakest ones in the HAp immunized animals. Both groups of IDLV vaccinated animals showed similar levels of HAI titers (*p* > 0.05), significantly lower than HAp + MF59 (*p* < 0.05), but significantly higher compared to HAp vaccinated animals (*p*< 0.01). Interestingly, the MN assay showed absence of neutralizing activity in the HAp alone group of mice after the prime, while MN titers were always present in both groups of IDLV vaccinated animals (**Figure 4B**). Again the highest activity was present in serum samples from HAp + MF59 immunized group. The boost increased the MN titers in all groups, including the HAp immunized animals.

### IDLV Induces Anti-NA Response

Since NA plasmid was included during the IDLV preparation and NA protein is present in subunit vaccine preparations (49, 50), we investigated the anti-NA response in all groups of immunized animals. Serum samples were thus assayed for NI activity. As shown in **Figure 5**, both IDLV groups showed a strong NI activity at all indicated time points after the prime that was significantly boosted after the immunization with HAp. A similar response was detected in the adjuvant vaccinated group, while the animals vaccinated with HAp alone did not generate detectable anti-NA response after the prime, showing low NI activity only after the boost.

### IDLV Vaccination Induces T Cell Response to HA and NP Delivered as Transgenes

In order to assess the NP specific CD8-restricted cellular response in mice immunized with IDLV-NP/HA, INFγ ELISPOT was performed using blood cells collected at 4, 12, and 24 weeks after the prime, stimulated with MHC Class I-restricted NP peptide (**Figure 6A**). As expected, a high number of IFNγ producing T cells was detected in animals vaccinated with IDLV-NP/ HA overtime, confirming a strong and persistent CD8+ T cell response directed to the transgene delivered by IDLV. The NP-specific T cell response was also analyzed in splenocytes at 42 weeks after a single IDLV-NP/HA immunization further confirming the persistence of transgene-specific IFNγ producing CD8<sup>+</sup> T cells (**Figure 6A**).

Figure 3 | Immunization schedule and kinetics of anti-H1N1 binding antibodies (Abs). (A) Immunization schedule. CB6F1 mice (four mice/group) were primed once intramuscularly with integrase defective lentiviral vector (IDLV) expressing hemagglutinin (HA) and pseudotyped with HA (IDLV-HA/ HA), IDLV expressing NP and pseudotyped with HA (IDLV-NP/HA), HA protein (\*purified hemagglutinin vaccine subunit from influenza strain H1N1 A/ California/7/2009) in combination with MF59 as an adjuvant (HAp + MF59), HA protein alone (HAp) or left untreated (Naive). All groups except for naive were boosted with HAp at 24 weeks after the prime. Blood was collected at several time points in order to perform ELISA and IFNγ ELISPOT assays. (B) Kinetics of serum anti-H1N1 IgG Abs. Serum samples from all groups were collected at the indicated time points after the prime and were assayed for the presence of anti-H1N1 IgG by ELISA. Results are expressed as predicted mean endpoint titers. The predicted mean values at each time point were estimated by a system of piecewise linear regressions including all antibody measurements within a structural equation modeling (SEM) frame work. Error bars indicate the 95% confidence interval. Asterisks indicate significant differences between groups at all the analyzed time points; \*\**p*-value <0.01.

In order to evaluate the HA specific response, INFγ ELISPOT was also performed in splenocytes from all groups of mice at sacrifice 18 weeks after the boost (42 weeks after the prime). Ten pools of 15mer peptides spanning the entire HA protein were used and the cumulative mean response against the HA pools is shown in **Figure 6B**. HA-specific IFNγ producing cells were detected only in mice primed with IDLV-HA/HA, while no positive response was observed in mice immunized with HAp with or without the adjuvant. In particular, pool 1, pool 9, and pool 10 (mean of 165.4, 666.2, 158.1 SFC/106 cells, respectively) were responsible for the HA-specific response.

### DISCUSSION

In this study, we assessed the feasibility of improving the strength of immune response for influenza vaccination by administering IDLV engineered to express either HA or NP proteins as transgenes, for induction of T cell responses and to carry HA on the surface of IDLV particles for a concomitant induction of functional antibodies against influenza virus.

Immunization with either HA-pseudotyped IDLV induced functional and durable Ab responses that were further increased after boosting with H1N1 purified subunit vaccine. This suggests that for induction of HA-specific Abs, HA protein does not need to be expressed from the IDLV transgene. Titers were lower or comparable to those obtained after immunization with MF59 adjuvanted H1N1 purified subunit vaccine, which we used as a gold standard for the induction of a strong and effective antibody response against influenza virus in mice. In particular, persistent HAI titers and anti-HA neutralizing Abs against homologous influenza virus strain were detected throughout the time course after the prime and were further increased after the boost. Of note, mice immunized with IDLV-HA/HA and IDLV-NP/HA induced functional anti-NA Abs, which were detected after the prime and increased after the boost, as measured by the NI assay. Titers were comparable to those obtained after immunization with MF59-adjuvanted H1N1 purified subunit vaccine, but significantly higher compared to HAp immunized animals. NA activity is required for the release of HA-pseudotyped IDLV from producer cells and a plasmid expressing the NA protein was included during preparation of IDLV for enabling vector production, as described in other settings (18). Although NA-specific Abs may not effectively prevent viral infection in humans, they may inhibit virus spread and reduce the severity of disease (51, 52), and a recent report provided strong evidence that NI titers correlated more significantly with reduced disease severity in a healthy volunteer challenge study performed with a wild-type influenza A challenge virus (53). Additional characterization of IDLV vaccine-induced functional antibodies, such as mapping of broadly neutralizing antibodies directed to the stem region of HA, will provide further information on the value of our platform.

While induction of HA-specific antibodies represents the primary strategy for prevention and control of influenza after vaccination in humans (36), the highly conserved viral NP has become an important focus for the development of broad, cross-protective or "universal" influenza vaccines. NP-specific cell-mediated responses have been shown to be protective against homologous and heterosubtypic influenza virus challenge in animal models (13, 54–57) and may help in reducing disease severity through enhancing viral clearance in humans (30, 58). In our previous work (13), we assessed the ability of IDLV-NP to induce protective immunity using different routes of immunization. In particular, we demonstrated that intranasal administration of IDLV-NP was more efficient than intramuscular immunization in protecting mice from influenza virus challenge. Although the protection from influenza challenge in mice can be achieved in presence of NP-specific T cell responses, this correlation has not been confirmed in humans, where the T cell responses directed toward conserved proteins may instead help in controlling the disease symptoms (28–30). To further improve the strength of our vaccine and to render the platform more suitable for human use, in the present work we focused on the development of a new vector strategy for expressing influenza antigens not only as transgenes but also as surface molecules. Here we demonstrated that the multi antigen IDLV-based influenza vaccine induced

Figure 4 | Kinetics of functional antihemagglutinin antibodies (Abs). Serum samples from all groups were collected at the indicated time points after the prime and were assayed for the presence of neutralizing Abs against A/California/7/09 virus, by hemagglutination inhibition assay (A) and microneutralization assay (B). Results are expressed as predicted mean endpoint titers. The predicted mean values at each time point were estimated by a system of piecewise linear regressions including all antibody measurements within a structural equation modeling frame work. Error bars indicate the 95% confidence interval. Asterisks indicate significant differences compared to the HAp group, at the indicated time points; \*\**p* < 0.01; \**p* < 0.05.

durable and functional immune responses in terms of NP or HA transgene-specific T cell responses and protective antibodies directed toward the surface proteins HA and NA. In particular, NP-specific CD8-restricted cellular responses, as measured by INFγ ELISPOT, were present and maintained up to 42 weeks after a single immunization in mice immunized with IDLV-NP/ HA. Similarly, HA-specific IFNγ producing cells were detected in mice immunized with IDLV-HA/HA, but not in mice immunized with HAp with or without MF59 adjuvant.

Other approaches have been tested for delivering multiple antigens for inducing cellular and humoral immunity. As an example, a recent report using Adenovirus and MVA delivering Influenza NP, M1 and HA antigens in a prime-boost regimen showed induction of immune responses against the delivered antigens and protection after challenge (59). The approach described in our report shows that multivalent IDLVs efficiently induce a prolonged cellular immune response due to expression of the vectored HA and NP transgenes from episomal DNA circles,

Figure 6 | Analysis of antigen-specific T cells response. (A) Kinetics of nucleoprotein (NP)-specific T cell response in mice immunized with IDLV-NP/HA. IFNγ ELISPOT was performed using blood cells collected at 4, 12, and 24 weeks post-prime and splenocytes collected at 42 weeks post-prime. Results are expressed as mean spot forming cells (SFC) per 106 cells. Cells were stimulated overnight with the MHC I-restricted epitope derived from NP protein sequence (black bars) or left untreated (white bars). Error bars indicate the SD among mice of the same group. (B) Analysis of HA-specific T cell response in immunized mice. IFNγ ELISPOT was performed using splenocytes collected at 42 weeks post-prime, using 10 pools of 15mers spanning the full length of HA protein, as described in Section "Materials and Methods." Results are expressed as cumulative mean SFC per 106 cells for each pool among mice of the same indicated group.

as described in several settings (1, 2) and functional humoral responses to HA and NA proteins.

Previous work has shown that recombinant influenza vaccines containing insect cell-expressed virus-like particles (VLPs) displaying H1N1 may constitute a promising vaccination approach (60). Pseudotype-based influenza genes delivery represents an alternative to successfully express HA in mammalian cells providing an efficacious vaccine when tested in chickens and mice (61, 62). More recently, Venereo-Sanchez et al. showed that HIV-Gag-based VLP displaying H1N1 induced a sustained immune response which provided full protection after lethal challenge with the homologous virus strain in mice (63). Of note, IDLVs used in our report were pseudotyped with VSV.G glycoprotein which may also contribute to the strength of the IDLV-induced immune responses by increasing the tropism of the vector. In fact, it has been shown that HIV VLP pseudotyped with VSV-G are more immunogenic compared to VLP that lacked VSV-G, in a monkey model of immunization (64) and that VSV-G-pseudotyped LV can adhere to transduced cells for a substantial amount of time, thus leading to additional cycles of transduction (65–67). These are potentially important advantages for exploiting recombinant IDLV pseudotyped with VSV.G and/ or other heterologous viral proteins.

In conclusion, our study highlights the potential for IDLV to be developed for use as a novel multiantigen vaccine platform against influenza virus. Combination of NP, HA, and NA antigens in the same IDLV results in a more comprehensive and functional immune response, which may be beneficial to prevent and/or control influenza virus infection.

### ETHICS STATEMENT

Animals were maintained under specific pathogen-free conditions in the animal facilities at the Istituto Superiore di Sanità (ISS) and treated according to European Union guidelines and Italian legislation (Decreto Legislativo 26/2014). All animal studies were authorized by the Italian Ministry of Healthy and reviewed by the Service for Animal Welfare at ISS (Authorization n. 314/2015-PR of 30/04/2015). All animals were euthanized by CO2 inhalation using approved chambers, and efforts were made to minimize suffering and discomfort.

### AUTHOR CONTRIBUTIONS

AG, AC, and DN designed the experiments, analyzed the data, and wrote the article; AG, MB, RB, FG, LC, LP, SC, DM, VM, AC, and ZM performed experiments; MV performed statistical analysis; MC analyzed the data and critically edited the manuscript; NT provided technical knowhow on pseudotype production and critically edited the manuscript; MS provided key reagents and critically edited the manuscript. All authors have contributed to the drafting of the manuscript, have revised the work, and have approved the final version.

### ACKNOWLEDGMENTS

We are grateful to Massimo Spada for his excellent work on mice, Emanuele Montomoli, Elisa Llorente Pastor and Giulia Lapini for their support and advice on serological influenza assays, Patrizio Pezzotti for statistical analysis, Marina Franco and Stefania Donnini for secretarial assistance, Ferdinando Costa and Patrizia Cocco for technical support. We thank Giuseppe Del Giudice for providing H1N1 and MF59 and for his expert advice. The following reagent was obtained through BEI Resources, NIAID, NIH: Peptide Array, Influenza Virus A/California/04/2009 (H1N1) pdm09 Hemagglutinin Protein, NR-15433.

### FUNDING

This project has received funding from the European Union's Seventh Programme for Research, Technological Development and Demonstration under grant agreement no. 280873 (ADITEC Project) and from NIH (grant no. 1R21AI124141-01 to MS).

## REFERENCES


HAT from human airway epithelium. *J Virol* (2006) 80:9896–8. doi:10.1128/ JVI.01118-06


alveolar macrophages is important for heterosubtypic influenza virus immunity. *PLoS Pathog* (2013) 9:e1003207. doi:10.1371/journal.ppat.1003207


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be interpreted as a potential conflict of interest.

*Copyright © 2018 Gallinaro, Borghi, Bona, Grasso, Calzoletti, Palladino, Cecchetti, Vescio, Macchia, Morante, Canitano, Temperton, Castrucci, Salvatore, Michelini, Cara and Negri. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Age and Influenza-Specific Pre-Vaccination Antibodies Strongly Affect Influenza Vaccine Responses in the Icelandic Population whereas Disease and Medication Have Small Effects

*Thorunn A. Olafsdottir 1,2, Kristjan F. Alexandersson1 , Gardar Sveinbjornsson1 , Giulia Lapini3 , Laura Palladino3 , Emanuele Montomoli <sup>3</sup> , Giuseppe Del Giudice4 , Daniel F. Gudbjartsson1,5 and Ingileif Jonsdottir1,2\**

*1deCODE Genetics, Amgen Inc., Reykjavik, Iceland, 2 Faculty of Medicine, School of Health Sciences, University of Iceland, Reykjavik, Iceland, 3Vismederi srl, Siena, Italy, 4GSK Vaccines, Siena, Italy, 5School of Engineering and Natural Sciences, University of Iceland, Reykjavik, Iceland*

### *Edited by:*

*David J. M. Lewis, Imperial College London, United Kingdom*

### *Reviewed by:*

*Michael Schotsaert, Icahn School of Medicine at Mount Sinai, United States Raffael Nachbagauer, Icahn School of Medicine at Mount Sinai, United States*

### *\*Correspondence:*

*Ingileif Jonsdottir ingileif.jonsdottir@decode.is*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 13 October 2017 Accepted: 08 December 2017 Published: 08 January 2018*

### *Citation:*

*Olafsdottir TA, Alexandersson KF, Sveinbjornsson G, Lapini G, Palladino L, Montomoli E, Del Giudice G, Gudbjartsson DF and Jonsdottir I (2018) Age and Influenza-Specific Pre-Vaccination Antibodies Strongly Affect Influenza Vaccine Responses in the Icelandic Population whereas Disease and Medication Have Small Effects. Front. Immunol. 8:1872. doi: 10.3389/fimmu.2017.01872*

Influenza vaccination remains the best strategy for the prevention of influenza virus-related disease and reduction of disease severity and mortality. However, there is large individual variation in influenza vaccine responses. In this study, we investigated the effects of gender, age, underlying diseases, and medication on vaccine responses in 1,852 Icelanders of broad age range who received trivalent inactivated influenza virus vaccination in 2012, 2013, or 2015. Hemagglutination inhibition (HAI) and microneutralization (MN) titers were measured in pre- and post-vaccination sera. Of the variables tested, the strongest association was with level of pre-vaccination titer that explained a major part of the variance observed in post-vaccination titers, ranging from 19 to 29%, and from 7 to 21% in fold change (FC), depending on the strain and serological (HAI or MN) analysis performed. Thus, increasing pre-vaccination titer associated with decreasing FC (*P* = 1.1 × 10−99– 8.6 × 10−30) and increasing post-vaccination titer (*P* = 2.1 × 10−159–1.1 × 10−123). Questionnaires completed by 87% of the participants revealed that post-vaccination HAI titer showed association with repeated previous influenza vaccinations. Gender had no effect on vaccine response whereas age had a strong effect and explained 1.6–3.1% of HAI post-vaccination titer variance and 3.1% of H1N1 MN titer variance. Vaccine response, both fold increase and seroprotection rate (percentage of individuals reaching HAI ≥ 40 or MN ≥ 20), was higher in vaccinees ≤37 years of age (YoA) than all other age groups. Furthermore, a reduction was observed in the H1N1 MN titer in people ≥63 YoA, demonstrating a decreased neutralizing functionality of vaccine-induced antibodies at older age. We tested the effects of underlying autoimmune diseases, asthma and allergic diseases and did not observe significant associations with vaccine responses. Intake of immune modulating medication did not show any association. Taken together, our results show that previous encounter of influenza vaccination or infection, reflected in high HAI and MN pre-vaccination titer has the strongest negative effect on vaccine responses measured as FC and the strongest positive effect on post-vaccination titer. Increasing age had also an effect but not gender, underlying disease or medication.

Keywords: influenza vaccine, pre-vaccination antibody titer, age effect, underlying diseases, medication

# INTRODUCTION

The influenza virus causes 3–5 million cases of severe illness each year resulting in 250,000–500,000 deaths, most of which occur in elderly people [≥65 years of age (YoA)] (1). Vaccination is the best preventative measure against influenza illness and the World Health Organization (WHO) recommends annual vaccinations for high-risk groups; pregnant women, children 6 months–5 YoA, elderly individuals (≥65 YoA), individuals with chronic medical conditions and health-care workers (1). There is large individual variation in influenza vaccine responses. Factors that have been associated with impaired immune responses to influenza vaccinations include age, gender, health status of vaccine recipients, prior influenza vaccinations, and obesity (2, 3), and various immunomodulators have been reported to influence immune responses to vaccines (4–7). We therefore decided to investigate influenza vaccine responses in unselected Icelandic vaccinees of a broad age range and health conditions.

The trivalent inactivated influenza vaccine (TIV) used in this study contains hemagglutinin (HA) surface glycoprotein from two influenza A strains (H1N1 and H3N2) and one influenza B strain (either Yamagata or Victoria lineage). Vaccine-induced hemagglutin (HA) titers are widely accepted as a correlate of protection against influenza illness and are measured by the ability of HA-specific antibodies to block *N*-acetylneuraminic acid mediated viral agglutination of red blood cells using a hemagglutination inhibition (HAI) assay (8, 9). Based on this, seroprotection has been defined as HAI antibody titers ≥1:40 post-vaccination and the proportion of vaccinees achieving this titer are referred to as seroprotection rate. Following vaccination, the seroprotection rate should be >70% for adults 18–60 YoA and >60% for adults >60 YoA. In contrast to HAI titers that only measure the capacity of blocking the receptor binding of the virus to its host cell, microneutralization (MN) assays are based on the use of infectious doses of influenza virus *in vitro* thereby measuring functional antibodies that block entry of the virus into cells, fusion of the virus HA to the host cell membrane, and internalization of the virus. In addition to measuring antibodies capable of neutralizing the strain specific and immunodominant head domain, MN have been shown to detect antibodies directed to the conserved stalk of HA that could give rise to a broad protection against different strains of influenza A virus (10). Anti-stalk antibodies have been shown to be superior inducers of cytotoxicity of infected cells compared with anti-head antibodies in a mouse model and this effect was dependent on interaction with Fc receptors for IgG (FcγRs) (11). However, due to the higher cost and labor of MN compared with HAI measurements, HAI is still more widely used than MN. In this study, we measured both HAI and MN for H1N1 in all pre- and post-vaccination samples, as well as in a subset of the vaccinees (*n* = 336) for the other strains and compared the two readouts.

Due to the high-mutation rate of the influenza virus, the vaccine components need to be frequently changed to match the circulating virus strains (12). However, the potential antigenic mismatch does not account for all of the observed differences in influenza vaccine efficacy between years. In addition to reported risk factors for poor vaccine responses, variation in human leukocyte antigen (13) and other host genetic factors may play a role (14).

In addition to following WHO's recommendations regarding the annual influenza vaccination in Iceland, many companies offer their staff annual influenza vaccinations, even if they are not in any of the risk groups. In this study, volunteers scheduled for influenza vaccination were recruited at over 70 work places and two nursing homes in the Reykjavik area when the annual influenza vaccination was offered, securing a broad age range independent of health status, and thus both risk and non-risk groups. Furthermore, we received information regarding prescribed medicines as well as diagnosis of asthma allergic and autoimmune diseases for all the study participants.

The main objective of this study was to evaluate the association of age, gender, influenza-specific pre-vaccination immune status, underlying diseases, and medication on immune responses to a seasonal TIV in unselected Icelanders of a broad age during three influenza seasons 2012–2013, 2013–2014, and 2015–2016.

We found that previous influenza virus encounter measured by high-pre-vaccination titer showed strong negative association with fold change (FC) levels whereas it showed positive association with the post-vaccination titer. Furthermore, increased age associated with vaccine responses, although to less extent than pre-vaccination titer. However, gender, underlying diseases, and immunomodulatory medication evaluated in this heterogeneous group of vaccinees did not significantly affect vaccine responses.

### MATERIALS AND METHODS

### Participants and Study Design

A total of 1,852 volunteers agreed to participate in the study and were eligible during October–November in the years 2012 (*n* = 565), 2013 (*n* = 711), and 2015 (*n* = 577). Age of the vaccinees ranged from 20 to 103 YoA, 46% men and 55% women. Informed consent and pre-vaccination blood samples were obtained by staff of the Patient Recruitment Center (PRC) just before the participants received their annual influenza vaccination (administered by health-care workers) that they had signed up for either at their work places or nursing homes. Four weeks later (day 28 ± 3 days), the staff of the PRC went back to the workplaces and nursing homes to collect a postvaccination blood sample. All participants signed informed consent. This study was carried out in accordance with the recommendations of National Bioethics Committee (NBC) of Iceland with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the NBC of Iceland (Approval no. VSN-12-153\_VSNb2012090016- 03-12). The study was reported to the Data Protection Authority of Iceland (ref. S5936/2012) that also approved access to data from the participant's medical records (PV\_2012091015T) from Landspitali, the National University Hospital of Iceland (1990–2016), and existing data at deCODE genetics. Diagnosis of autoimmune diseases we searched for included: ankylosing spondylitis, inflammatory bowel disease, multiple sclerosis, myasthenia gravis, primary biliary cirrhosis, psoriasis, psoriatic arthritis, rheumatoid arthritis, systemic lupus erythematosus, systemic sclerosis, type 1 diabetes, vitiligo, Sjögren's syndrome, and autoimmune thyroiditis. Diagnosis of asthma and allergic diseases we searched for included: asthma, allergic rhinitis, anaphylaxis, angioedema, chronic sinusitis, nasal polyps, urticaria, and atopic dermatitis. 1,611 of the participants (87%) answered questionnaires on general health and lifestyle. Information of drug prescription for each of the participants was retrieved from the Directorate of Health Prescription Database (2003–2016) with approval of the NBC. ATC codes of drugs tested for are listed in Table S1 in Supplementary Material. The personal identities of the participants data and biological samples were encrypted using the Identity Protection System, a third-party encryption system approved, and monitored by the Icelandic Data Protection Authority.

### Vaccines

Study participants were vaccinated with Vaxigrip® (Sanofi-Aventis) in the years 2012 and 2013 containing 15 μg/strain of the A/California/7/2009 (H1N1)pdm09-like virus, A(H3N2) virus antigenically like the cell-propagated prototype virus A/Victoria/361/201 (H3N2)-like virus, and the B strains: B/Wisconsin/1/2010-like virus and B/Massachusetts/2/2012-like virus strains, respectively. In 2015, the vaccine contained the same H1N1 strain together with A/Switzerland/9715293/2013 (H3N2)-like virus and B/Phuket/3073/2013-like virus.

### Clinical Samples

Serum was obtained from whole blood collected both prevaccination (day 0) and post-vaccination day 28 ± 3 days for serological assays. Briefly whole blood was collected in 8 ml Vacuette (Z Serum Sep Clot Activator) tubes, allowed to clot for 30 min at room temperature and kept at +4°C for up to 4 h before centrifugation at 2500 RCF við +4°C. Serum was collected, aliquoted, and kept at −80°C until analyzed by VisMederi (Siena, Italy).

# MN Assay

The MN assay was modified from a previously described procedure (15); and carried out in VisMederi laboratories. This method is based on capability of live virus to infect and replicate in cells, producing cytopathic effect (CPE) in the cell culture substrate, which is prevented by neutralizing antibodies contained in serum of vaccine subjects.

Influenza live virus A/California/07/2009 was egg propagated by VisMederi and used in the assay at the concentration of 200 TCID50/100 μl [50% tissue culture infective dose (TCID50)].

Positive and negative control sera were included in each run, as well as a back titration plate for virus titration check. In particular, antisera used were specific for each strain tested, purchased from NIBSC; depleted serum was supplied by Sigma Aldrich (Serum minus IgA/IgM/IgG, S5393).

Each heat-inactivated serum twofold diluted in microtiter plates, starting from a 1:10, was incubated with a same volume of virus solution (200TCID50/100 μl) for 1 h at 37°C and 5% CO2. Then 100 µl of MDCK (Madin-Darby Canine Kidney) cell suspension was added to the virus-sera mixture at the concentration of 2 × 105 /ml; then plates were incubated at 37°C and 5% CO2 for 5 days. Under optical microscope each wells was assessed for the presence of CPE (complete destruction of the cell layer in the well or the presence of holes in the cell layer, surrounded by destroyed cells), discriminating "infected" and "protected" wells. The total number of infected wells of each serum duplicate was used to calculate the MN titer of each serum sample by Spearman−Karber formula (16). Indicative seroprotection rate was defined as percentage of vaccine recipients with serum MN titer ≥20 after vaccination.

### HAI Assay

The HAI measurement was carried out in agreement with Vis-Mederi procedures; using viral antigens, provided by NIBSC, diluted at the standard concentration of 160 hemagglutinating units (HAU)/ml and correctness of antigen dilution was checked out through a back titration in every test. Serum samples were treated with receptor destroying enzyme provided by Denka Seiken, in a ratio of 1:3, during overnight incubation at 37°C, and heat inactivated for 1 h at 56°C. Twofold serial dilutions starting from 1:10 were performed for each serum in duplicate in "V" bottomed 96-well plates. The antigen solution (4 HAU/25 μl) was added to each serum dilution and plates were incubated for 1 h at room temperature. A 0.35% solution of turkey red blood cells was added to each wells and plates were incubated for 1 h at room temperature (17). The HA protein is able to agglutinate red blood cells due to its binding affinity to surface glycoprotein of erythrocytes, and antibodies may interfere with this binding recognizing the virus antigen; this phenomenon produces an inhibition of the hemagglutination resulting in a change in the appearance of the well (18). The read out was performed by naked eye, distinguishing between the presence of hemagglutination and inhibition of it. The HAI titer was calculated as the reciprocal value of the highest serum dilution in which the hemagglutination was still inhibited. Seroprotection rate was defined as percentage of vaccine recipients with serum HAI titer ≥40 after vaccination.

# Statistical Analysis

Generalized linear regression was used to test the associations of log-transformed post-vaccination HAI or MN titers, FC, and seroprotection with various traits. The post-vaccination titer and FC were corrected for age, gender, pre-vaccination titer, year of immunization, and vaccination status (see post-vaccination titer model in Table S2 in Supplementary Material). Seroprotection was corrected for age, gender, year of immunization, and vaccination status (Table S5 in Supplementary Material). Age was split into five equal-sized groups. Vaccination status was split into three groups; one previous influenza vaccination, more than one previous influenza vaccination, and no previous influenza vaccination or information missing. To account for heteroscedasticity weighted least squares was used. Correlations between HAI and MN derived log-transformed titers were performed using linear regression. Statistical analysis was performed using the computing environment R (19).

# RESULTS

### Overall Influenza Vaccine Responses

A total of 1,852 individuals (46% men and 54% women) at the age of 20–103 years (**Table 1**) that received influenza vaccination and participated in the study had both pre- and post-vaccination HAI titers for all three vaccine strains available for analysis. In addition, MN titer was measured for the whole study group for H1N1 and for a subset of 336 for H3N2 and B strains. Overall seroprotection (HAI ≥ 40) rates post-vaccination for the three study years (2012, 2013, and 2015) were 93% (H1N1), 95% (H3N2), and 65% (B strain), with the lowest seroprotection rate observed in 2015 for all three strains 90% (H1N1), 93% (H3N2), and 33% (B strain). Highest median (25th–75th quantiles) HAI post-vaccination titer across the entire study group was observed for the H3N2; 226 (113–453) followed by H1N1; 160 (80–320) and B; 40 (6–113). MN titer of 20 measured by CPE has been suggested to be predictive of protection and correspond to an HAI titer of 40 (20). Using MN ≥ 20 as definition of indicative MN seroprotection, we observed overall seroprotection rate of 63%, ranging from 60% (year 2013) up to 67 (year 2012). There was a significant difference in HAI post-vaccination titer between the three vaccination years for all three strains (H1N1 *P* = 1.4 × 10<sup>−</sup>20, H3N2 *P* = 0.011, B *P* = 4.7 × 10<sup>−</sup>95). Similarly, the H1N1 MN post-vaccination titer differed between years (*P* = 1.6 × 10<sup>−</sup><sup>4</sup> , **Table 1**).

# Previous Humoral Influenza Virus Immunity Strongly Affects Influenza Vaccine Responses

We tested the association of influenza virus-specific pre-vaccination immunity status with vaccine responses. Pre-vaccination HAI titers associated with FC of post-vaccination HAI titers for all strains tested, i.e., high-pre-HAI titer resulting in less fold increase in post-vaccination HAI titer (H1N1 *P* = 1.1 × 10<sup>−</sup>99, H3N2 *P* = 1.5 × 10<sup>−</sup>89, B *P* = 2.1 × 10<sup>−</sup>57; **Figure 1A**), although post-vaccination HAI titer itself was positively associated with pre-vaccination titer (H1N1 *P*= 1.4 × 10<sup>−</sup>147, H3N2 *P*= 1.0 × 10<sup>−</sup>123, B *P* = 4.5 × 10<sup>−</sup>157; Figure 1B; Table S2 in Supplementary Material). Pre-vaccination titer of HAI explains 27, 24, 19% of the variance in HAI post-vaccination titer and 21, 18, and 12% of the variance in FC for H1N1, H3N2, and B strains, respectively (Table S3 in Supplementary Material). Furthermore, FC of MN titers (H1N1) associated strongly with pre-vaccination titers (*P* = 8.6 × 10<sup>−</sup>30; **Figure 1C**). Similar association was observed for the pre-vaccination titer and FC of MN titer of the small subsets (*n* = 336); Table S4 in Supplementary Material measured for H3N2 (*P* = 1.4 × 10<sup>−</sup><sup>8</sup> ) and B strains (*P* = 2.6 × 10<sup>−</sup><sup>4</sup> ). Similar to the HAI titer, pre-vaccination MN titer was positively associated with post-vaccination MN titer (*P* = 2.1 × 10<sup>−</sup>159; **Figure 1D**) with the pre-vaccination titer explaining 29% of the H1N1 postvaccination titer variance (Table S3 in Supplementary Material).

Vaccinees reporting to have had more than one previous influenza vaccination (*n* = 1,418) had a lower post-vaccination titer for all strains than those who reported to have only had one (*n* = 106) previous influenza vaccination [HAI for H1N1 *P* = 2.32 × 10<sup>−</sup><sup>8</sup> , H3N2 *P* = 3.26 × 10<sup>−</sup>15, B *P* = 9.39 × 10<sup>−</sup><sup>7</sup> (**Figure 2A**), and MN for H1N1 *P* = 3.1 × 10<sup>−</sup>12; **Figure 2B**; Table S2 in Supplementary Material]. There was no significant difference in HAI seroprotection rate between those receiving one or more than one vaccination for H1N1 (*P* = 3.6 × 10<sup>−</sup><sup>1</sup> ) and H3N2 (*P* = 6.1 × 10<sup>−</sup><sup>1</sup> ). However, those that received more than one vaccination had reduced HAI seroprotection rate for the B strain (3.9 × 10<sup>−</sup><sup>3</sup> ) as well as for MN seroprotection rate for the H1N1 (4.6 × 10<sup>−</sup><sup>6</sup> ) (Table S5 in Supplementary Material), using 1:20 MN titer as a cut-off for seroprotection (20).

Taken together, prior influenza virus-specific antibody levels, induced upon influenza virus infection and/or vaccination, strongly affect the vaccine-induced humoral response with the pre-vaccination titer explaining roughly 19–29% of the variance in post-vaccination titer.

### Age Strongly Affects Influenza Vaccine Responses, But Gender Does Not

We tested the effect of age and gender on response to vaccination, after adjusting for pre-vaccination titers and year of immunization. Post-vaccination HAI titer went down with age for all three vaccine strains (H1N1 *P* = 1.5 × 10<sup>−</sup>15, H3N2 *P* = 1.1 × 10<sup>−</sup><sup>9</sup> , B *P* = 2.2 × 10<sup>−</sup>21; **Figure 3A**). Age was found to explain 3, 2, and 2% of the variance in post-vaccination titers for H1N1, H3N2, B strains, respectively (Table S3 in Supplementary Material).

Hemagglutination inhibition seroprotection rate following vaccination was also reduced with increasing age for two out of three strains (H1N1 *P* = 5.9 × 10<sup>−</sup><sup>7</sup> , H3N2 *P* > 0.05, B *P*= 4.6 × 10<sup>−</sup><sup>9</sup> ; **Figure 3B**). The youngest age group (20–37 YoA) had higher post-vaccination titers than any of the other age groups and the difference was most evident when compared with the oldest age group (63–103 YoA) (Table S2 in Supplementary Material).

Microneutralization titer for H1N1 also showed that increasing age associated with decrease in both post-vaccination titer (*P* = 5.4 × 10<sup>−</sup>20; **Figure 3C**) and indicative seroprotection rate (*P* = 8.3 × 10<sup>−</sup>31; **Figure 3D**), similar to what was observed for the HAI titers. Age explained 3% of the variance in H1N1 post-vaccination MN titers. Similar association was observed with age for the subset (*n* = 336; Table S4 in Supplementary Material) measured for H3N2 (*P* = 1.9 × 10<sup>−</sup><sup>4</sup> ) and B (*P* = 9.4 × 10<sup>−</sup><sup>4</sup> ) strains (data not shown). Interestingly, the effect of age on the MN titer was even more pronounced than on HAI titer indicating that the MN titer more specifically captures the age-related reduced neutralizing functionality of antibodies than the HAI titer (**Figures 3A,C**). No significant effects of gender on either post-vaccination HAI or MN titers or seroprotection rates were observed among all vaccinees for any of the strains tested (Figure S1 in Supplementary Material) nor among the elderly vaccinees (data not shown).

### Medication and Underlying Diseases Have Small Effects on Vaccine Responses in the Whole Group of Vaccinees

Underlying health conditions as well as medications with known immunomodulatory effects have been proposed to affect influenza vaccine responses (4–7, 21). We received data from



the Directorate of Health Prescription Database on prescriptions/dispensing of various immunomodulating agents for all vaccinees from 2003 until the time of vaccination. The medication included lipid modifying agents (statins both fermented and synthetic), non-steroidal anti-inflammatory drugs, and asthma/allergy drugs (Table S1 in Supplementary Material). We used strict criteria for the prescribed medicines, dispensed 0–3 months before vaccination, to increase the likelihood of the subjects actually taking the medication at the time of vaccination. Vaccine responses did not show significant association with any of the medications studied here (**Table 2**). Neither was there a significant effect of any of the medications on vaccine responses among vaccinees over 65 YoA (data not shown).

We received discharge diagnosis from Landspitali, the National University Hospital of Iceland for all the vaccinees (dated from 1990), which were used in addition to the existing data at deCODE. The effect of autoimmune disease and asthma and/or allergic diseases on vaccine responses was tested. Subjects with asthma/allergic diseases had lower post-vaccination HAI titer for H3N2 strain than the rest of the vaccinees, which was nominally significant (β = −0.167, *P* = 0.024). No effects of autoimmune diseases on the post-vaccination HAI titers were observed (**Table 3**). Neither did autoimmune or asthma/allergic diseases affect vaccine responses among vaccinees over 65 YoA (data not shown). Self-reported general health status including information on: hospitalization due to infections, underlying diseases, and use of prescribed medicines was assessed through questionnaire data completed by 87% of vaccines. However, none of those parameters associated significantly with post-vaccination HAI titers when corrected for number of statistical tests (data not shown).

corresponding pre-vaccination titer (*x*-axes) (B). Log H1N1 microneutralization (MN) fold increase (*y*-axes) vs. pre-vaccination MN titer (*x*-axes) (C) and log H1N1 MN post-vaccination titer vs. H1N1 pre-vaccination titer (D). Line within the boxplots indicate median value and the top and the bottom correspond to the 25th (Q1) and 75th (Q3) quantiles. The whiskers of the box plots are located at Q1 − 1.5 interquartile range (IQR) and Q3 + 1.5 IQR. *P*-values for association between fold change and pre-vaccination titer (A,C) and post-vaccination titers and pre-vaccination titer (B,D) are shown. Age, measurement date, previous influenza vaccination status, and gender were included as covariates.

Therefore, our data indicate that for this large heterogeneous group of vaccinees of a broad age range medications and underlying diseases had no or minimal effects on influenza vaccine responses although asthma/allergic disease warrants further investigation in a controlled clinical study.

### DISCUSSION

Here, we report seasonal influenza vaccine-induced antibody responses of Icelanders of wide age range and health status. The strongest association with vaccine-induced HAI responses was observed for the level of pre-vaccination HAI titer, thus those with high-pre-vaccination HAI titer did not show fold increase in post-vaccination HAI titer to the same extent as those with lowpre-vaccination HAI titer. The same was observed for the more functional MN titer. Large part of the FC variance (12–21%) was explained by the pre-vaccination titer. However, post-vaccination titer itself strongly associated positively with pre-vaccination titer, meaning that even if those with high-pre-vaccination titer showed less fold increase than those with low-pre-vaccination titer, their post-vaccination HAI/MN titer was overall higher. Pre-vaccination titer explained a large part of the variance in post-vaccination titer, or 19–29% depending on the viral strain and assay used for antibody measurements.

Our results on this population-based heterogeneous group of vaccinees are in line with previous reports from smaller studies showing correlation between pre-existing antibody levels to a given strain and reduced humoral immune responses upon vaccination with the same strain (22–25). The most plausible explanation as to how pre-existing strain-specific antibodies adversely affect subsequent vaccine responses is that they bind and mask viral epitopes in a vaccine containing a homologous strain, supported by the fact that high-pre-existing antibody levels also correlate with activation of fewer strain-specific plasmablasts,

Figure 2 | Multiple previous influenza virus vaccination associate with reduced hemagglutination inhibition (HAI) titer. Log HAI for H1N1, H3N2, and B strain (A) and log H1N1 microneutralization (MN) (B) post-vaccination titer for vaccinees that reported to have had one previous influenza vaccination or more than one previous influenza vaccination. Line within the boxplots indicate median value and the top and the bottom correspond to the 25th (Q1) and 75th (Q3) quantiles. The whiskers of the box plots are located at Q1 − 1.5 interquartile range (IQR) and Q3 + 1.5 IQR. *P*-values for association between post-vaccination titers and previous influenza vaccinations are shown. Pre-titer, age, measurement date, and gender were included as covariates.

and vaccine-induced memory B cells (22). However, our results demonstrate that when effect on vaccine response is evaluated the readout, fold increase vs. post-vaccination titer, must be considered. The effect of lower HAI/MN titer fold increase upon repeated vaccinations on vaccine efficacy has been debated, ranging from less efficacy (26) to reduced serious influenza disease and better efficacy (27–29) reported for individuals receiving repeated influenza vaccinations. It has been proposed that antigenic distances between annual vaccine strains on one hand and the epidemic strains on the other hand might explain those discrepancies (30). Pre-existing immunity has also been shown to affect the type of antibodies induced upon vaccination with low-pre-existing immunity to H1N1 pandemic strain inducing more broadly protecting HA stalk-specific antibodies, whereas with high-pre-vaccination immunity mainly strainspecific antibodies aimed at the HA head were induced (31). We could not distinguish between stalk- or head-specific antibodies in our MN measurements and based on previous publications claiming that antibodies toward the stalk domain appear to neutralize less potently than antibodies directed to the HA head domain (32, 33) it is likely that our MN measurements are mainly picking up head-specific antibodies. Future studies addressing the role of pre-existing immunity on antibody-dependent cellmediated cytotoxicity (ADCC) of cells infected with homologous or heterologous influenza virus strains would be highly interesting as protective role of stalk binding antibodies has been linked with FcγR-mediated ADCC (11). Our study was not designed to evaluate vaccine efficacy. However, given that almost one-third of the post-vaccination variance (both HAI and MN) is explained by the level of pre-vaccination titers in our study and the fact that the current H1N1 strain used in seasonal influenza vaccines did not change for years following the 2009–2010 pandemic despite considerable antigenic drift in the epidemic strains of this same period (34), we believe it is high time to re-evaluate whether repeated vaccinations with the same strain are beneficial for the vaccine efficacy or not. However, it should be noted that there was no difference in HAI seroprotection rate for the H1N1 and H3N2 strains between those that had received only one compared with more than one previous vaccinations, although reduced HAI and indicative MN seroprotection levels were observed for the B and H1N1 strain, respectively, despite higher post-vaccination titer for all strains. The arbitrary cut-off we used here for indicative MN seroprotection is based on a previous publication using the same cytopathic lab test for H5N1 (20), which has not been validated for other influenza virus strains and might not reflect seroprotection levels for the H1N1 strain used in this study.

Fold change is most commonly used to evaluate vaccine responses to influenza viruses and many other pathogens. Our results clearly show that FC can be misleading for evaluation of vaccine responses in vaccine/virus experienced individuals and strongly suggests that overall post-vaccination HAI or MN titers adjusted for pre-vaccination titer are more relevant and should also be considered. This is in line with new guidelines from the European medicines agency on licensing of novel influenza vaccines in Europe that emphasize the importance of quantifying functional antibodies in addition to the HA antibody response. Furthermore, the guidelines no longer rely on the pre-defined

and age (B,D) are shown. Pre-titer, age, measurement date, previous influenza vaccination status, and gender were included as covariates.

Table 2 | Effect of medication 0–3 months before vaccination on post-hemagglutination inhibition (HAI).


are located at Q1 − 1.5 interquartile range (IQR) and Q3 + 1.5 IQR. *P*-values for association between post-vaccination titer and age (A,C) and seroprotection rate

Table 3 | Effect of underlying disease on post-hemagglutination inhibition (HAI).


*Significant values presented in bold.*

protective threshold based on serological assays (GMT increase, seroconversion, and seroprotection based on HAI ≥ 1:40), declaring that this was not the most informative approach for different subgroups of vaccinees (35). The youngest age group (20–37 YoA) had the highest HAI post-vaccination titer (adjusted for pre-vaccination titers, gender, year of immunization, and vaccination status) and seroprotection rates for all three vaccine strains. Previously, age-associated vaccine responses have primarily been reported for young (<65 YoA) vs. elderly (≥65 YoA) subjects, whereas differences in responses within the younger adult vaccination groups has not been extensively studied. Our data are in line with a recent report of significantly different influenza vaccine-induced transcriptomic responses of people older than 35 YoA compared with younger adults (36), indicating that changes in influenza vaccine responses occur much earlier in life than frequently reported. MN that measures functional antibodies seems to better capture the effect of age on vaccine responses than HAI titer, observed in a substantial decrease in the MN titer in vaccinees at 56–62 YoA, and again in those aged 63–103 years old. Still, age explains only 2–3% of the overall variance in postvaccination titers, corresponding to 10% of what is explained by pre-vaccination antibody levels. It was recently reported that only two previous influenza vaccinations were needed to account for the entire differences in influenza vaccine responses observed between the young and elderly groups (37). However, we observed significant effects of age on vaccine responses even when we corrected for vaccination status (if subjects had received only one or more than one previous vaccination) indicating that in our study the age effect is not entirely based on vaccine history.

The strength of this study is presented by the broad age range and the fact that vaccinees were not selected into the study, but all individuals offered vaccination at two nursing homes and over 70 workplaces, including health institutions and schools, were invited to participate. The vaccinees thus represent a snapshot of the Icelandic population that gets an annual influenza vaccination, irrespective of age and health status. This, unfortunately, also means that the study is underpowered for detecting small effects in various subgroups, such as specific disease groups, but rather gives indications for interesting future research.

Overall, our data do not point toward major effects of underlying diseases and/or medication (based on prescription and dispensing within 3 months) on influenza vaccine responses. Out of the underlying diseases tested, vaccinees with diagnosed asthma and/or allergic diseases were the only ones showing a trend toward lower post-vaccination HAI titers. The effects of influenza vaccination in asthma patients has mainly been evaluated on safety and efficacy with inconsistent results that might be due to the different cohorts, vaccines, and methodology applied (38, 39).

Chronic intake of fermentation derived and synthetic statins have previously, been linked with lower influenza vaccine responses and reduced effectiveness in an elderly cohort (21, 40). Therefore, we looked at the association of synthetically and fermentation derived statins on vaccine-induced HAI postvaccination titer but found no significant association of vaccine responses with either class of statins or both classes combined. This could be due to the low number of elderly people among our vaccinees (*n*= 261) compared with the previous study of vaccinees ≥65 YoA (*n* = 6,961) (21). We also tested the effect of statins in our ≥65 YoA vaccinees separately where 26% of the 261 subjects were on statins but we found no significant association with vaccine responses there either (data not shown). Furthermore, our analysis is based on prescribed/dispensed statins and we did not have as detailed information on statin intake as in the previous publication (daily statin intake ≥28 days before through 22 days after vaccination).

Taken together, our results show that in the whole study group influenza virus-specific pre-vaccination immune status is the major factor affecting TIV vaccine responses, followed by age. None of the other variables tested had significant effects, considering the number of tests performed, although vaccine responses in asthma/allergic patients tended to be reduced and could be an interesting future research area.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of National Bioethics Committee (NBC) of Iceland with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the National Bioethics Committee (NBC) of Iceland (Approval no. VSN-12- 153\_VSNb2012090016-03-12). The study was reported to the Data Protection Authority of Iceland (ref. S5936/2012) that also approved access to data from the participant's medical records (PV\_2012091015T) from Landspitali, the National University Hospital of Iceland (1990–2016), and existing data at deCODE genetics. The personal identities of the participants data and biological samples were encrypted using the Identity Protection System, a third-party encryption system approved, and monitored by the Icelandic Data Protection Authority.

### AUTHOR CONTRIBUTIONS

IJ, TO, GS, and DG designed the study. GS, KA, TO, GG, DG, and IJ analyzed and/or interpreted the data. LP, GL, and EM performed HAI and MN measurements. TO and IJ drafted the manuscript. All authors contributed to and approved the final version of the manuscript.

### ACKNOWLEDGMENTS

We would like to thank all the participants in the study, Helga Norland, and the staff at the Patient Recruitment Center for organizing the recruitment of participants and sample collection at work places and nursing homes, and Eva Yr Gunnlaugsdottir and her vaccination team at Vinnuvernd for their assistance.

### FUNDING

This research was funded in part by the European Union's Seventh Framework Program "ADITEC" (EU FP7/2007-2013, grant agreement no: 280873).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01872/ full#supplementary-material.

## REFERENCES


40. Omer SB, Phadke VK, Bednarczyk RA, Chamberlain AT, Brosseau JL, Orenstein WA. Impact of statins on influenza vaccine effectiveness against medically attended acute respiratory illness. *J Infect Dis* (2016) 213(8): 1216–23. doi:10.1093/infdis/jiv457

**Conflict of Interest Statement:** TO, GS, KA, DG, and IJ are employees of deCODE Genetics/Amgen, Inc. GG is employee of GSK vaccines (Siena, Italy). LP, GL, and EM are employees of Vismederi srl.

*Copyright © 2018 Olafsdottir, Alexandersson, Sveinbjornsson, Lapini, Palladino, Montomoli, Del Giudice, Gudbjartsson and Jonsdottir. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Mouse Models of Influenza Infection with Circulating Strains to Test Seasonal Vaccine Efficacy

*Helen T. Groves1†, Jacqueline U. McDonald1†, Pinky Langat1 , Ekaterina Kinnear1 , Paul Kellam1 , John McCauley2 , Joanna Ellis3 , Catherine Thompson3 , Ruth Elderfield4 , Lauren Parker <sup>4</sup> , Wendy Barclay <sup>4</sup> and John S. Tregoning1 \**

*1Mucosal Infection and Immunity Group, Section of Virology, Department of Medicine, St Mary's Campus, Imperial College London, London, United Kingdom, 2 The Crick Institute, London, United Kingdom, 3Respiratory Virus Unit, Public Health England, London, United Kingdom, 4Molecular Virology, Section of Virology, Department of Medicine, St Mary's Campus, Imperial College London, London, United Kingdom*

### *Edited by:*

*Donata Medaglini, University of Siena, Italy*

### *Reviewed by:*

*Carole Henry, University of Chicago, United States Ji Wang, Harvard Medical School, United States*

### *\*Correspondence:*

*John S. Tregoning john.tregoning@imperial.ac.uk*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 13 October 2017 Accepted: 16 January 2018 Published: 31 January 2018*

### *Citation:*

*Groves HT, McDonald JU, Langat P, Kinnear E, Kellam P, McCauley J, Ellis J, Thompson C, Elderfield R, Parker L, Barclay W and Tregoning JS (2018) Mouse Models of Influenza Infection with Circulating Strains to Test Seasonal Vaccine Efficacy. Front. Immunol. 9:126. doi: 10.3389/fimmu.2018.00126*

Influenza virus infection is a significant cause of morbidity and mortality worldwide. The surface antigens of influenza virus change over time blunting both naturally acquired and vaccine induced adaptive immune protection. Viral antigenic drift is a major contributing factor to both the spread and disease burden of influenza. The aim of this study was to develop better infection models using clinically relevant, influenza strains to test vaccine induced protection. CB6F1 mice were infected with a range of influenza viruses and disease, inflammation, cell influx, and viral load were characterized after infection. Infection with circulating H1N1 and representative influenza B viruses induced a dose-dependent disease response; however, a recent seasonal H3N2 virus did not cause any disease in mice, even at high titers. Viral infection led to recoverable virus, detectable both by plaque assay and RNA quantification after infection, and increased upper airway inflammation on day 7 after infection comprised largely of CD8 T cells. Having established seasonal infection models, mice were immunized with seasonal inactivated vaccine and responses were compared to matched and mismatched challenge strains. While the H1N1 subtype strain recommended for vaccine use has remained constant in the seven seasons between 2010 and 2016, the circulating strain of H1N1 influenza (2009 pandemic subtype) has drifted both genetically and antigenically since 2009. To investigate the effect of this observed drift on vaccine induced protection, mice were immunized with antigens from A/California/7/2009 (H1N1) and challenged with H1N1 subtype viruses recovered from 2009, 2010, or 2015. Vaccination with A/California/7/2009 antigens protected against infection with either the 2009 or 2010 strains, but was less effective against the 2015 strain. This observed reduction in protection suggests that mouse models of influenza virus vaccination and infection can be used as an additional tool to predict vaccine efficacy against drift strains.

Keywords: Influenza Vaccines, mouse models, Infection, Antibodies, Viral, vaccine drift

# INTRODUCTION

Influenza infection is a significant cause of morbidity and mortality worldwide; the WHO estimates that there are 3–5 million severe influenza cases every year, causing 250,000–500,000 deaths globally (1). There is also a considerable economic burden from influenza epidemics, which cost the European economy approximately €6–€14 billion and the US economy \$87.1billion annually (2, 3). There are

**118**

currently several different vaccines available including trivalent or quadrivalent inactivated vaccines and live attenuated vaccines. Seasonal influenza vaccination is considered the most effective intervention strategy for reducing the burden of influenza disease (4, 5). However, influenza vaccines have highly variable rates of efficacy, ranging from 10% in 2004–2005 (6) to 60% in 2010–2011 (7). The main cause of vaccine failure is mismatch between the vaccine and circulating strains. The cause of these mismatches is change in the circulating strains, either through antigenic drift (small mutations in hemagglutinin sequence) or antigenic shift (major replacements of circulating virus).

Due to the changing nature of the circulating influenza virus strain, vaccine strain selection mismatches can and do occur (8). In the autumn of 2014, increased rates of influenza activity were observed in the United States and this was attributed to poor vaccine effectiveness as a result of a mismatch between the H3 component of the current influenza vaccine and circulating strains (8). The overall effectiveness of the 2014–2015 influenza vaccine for preventing medically attended laboratory confirmed influenza virus was 23% (9). Early studies of influenza infections during the 2014/2015 season found that 100% of lab confirmed influenza A infections were A (H3N2) and of those 67% were antigenically drifted from A/Texas/50/2012, the reference strain used for the 2014/2015 vaccine in the northern hemisphere (9). A similar report from Canada found that 91% of the isolates were found to be genetically and antigenically distinct from the A/Texas/50/2012 vaccine strain (10). The same time period saw the emergence of a new lineage of H3N2 viruses (3C.2a and 3C.3a), which showed poor reactivity with antisera raised against A/Texas/50/2012 leading to its replacement with A/ Switzerland/9715293/2013 in the next vaccine season (11).

Part of the decision process about which strains should be used for vaccines is hemagglutination inhibition (HI) using ferret sera, complemented with virus neutralization data. However, mice are widely used in the preclinical development and evaluation of potential vaccines and antiviral compounds and have the potential to inform decisions. In this paper, we develop models of influenza infection in CB6F1 mice and evaluate the effect of vaccination on disease outcome after infection with matched and mismatched strains of virus.

### RESULTS

### Recent Clinical Isolates of H1N1 Subtype and Influenza B, but not H3N2 Cause Disease in Mice

Mice were infected with escalating doses of viruses reflective of recent circulating influenza viral strains and or widely used laboratory strains. For H1N1, mice were infected with influenza A/England/195/2009 (12) (clinical isolate: **Figure 1A**) or PR8 (Lab strain: **Figure 1B**). Animals infected with both the seasonal and laboratory strains of H1N1 lost weight proportionally to the infectious dose of virus. Comparing the response by dose of plaque forming units would suggest PR8 causes more disease per PFU used, but there may be limitations in using PFU for comparisons. For H3N2, mice were infected with A/England/691/2010 (Clinical isolate: **Figure 1C**) or A/X31 (Lab strain), which consists of HA and NA molecules from A/Hong Kong/1/68 (H3N2) on a PR8 background (**Figure 1D**). While animals infected with the laboratory H3N2 strain, X31, lost weight after infection, mice infected with the current seasonal H3N2 virus (A/England/691/2010) did not lose weight at the doses used. To test responses to influenza B, mice were infected with virus isolates that are close to current circulating strains, B/Florida/04/06 representing the Yamagata lineage (**Figure 1E**) and B/Brisbane/60/2008 (**Figure 1F**) representing the Victoria lineage. Infection with the Yamagata but not the Victoria lineage influenza B led to weight loss, but a larger dose of virus may be required for the Victoria lineage virus.

### Infectious and Immunological Characterization of Influenza Infection in Mice

Having observed that infection with some strains of influenza virus caused signs of disease, we wished to confirm that these viruses were able to replicate in mouse lungs and wanted to investigate the histological and immunological correlations of disease. Mice were challenged intranasally with representative H1N1 (Eng/195), Flu B (Flo/04), and H3N2 (A/X-31) strains and monitored over 7 days. A control group of mice were given sterile PBS intranasally. All influenza challenged mice lost significant amounts of weight compared to the control group (**Figure 2A**). Temperature was also measured, but no significant differences were observed (**Figure 2B**). Lung viral load was assessed *via* plaque assay (all groups) or influenza A M gene RNA qPCR (H1N1, X31, and control) (**Figures 2C,D**). Virus was detected in the lungs *via* plaque assay on day 4 for all infected mice (**Figure 2C**). Viral RNA was quantified for the influenza A infected groups and was detected on day 4 for both H1N1 and A/X-31 (**Figure 2D**). At day 7, virus and viral RNA was only detected in the H1N1 infected mice (data not shown).

Lung inflammation was investigated as a measure of disease pathology. There were no significant differences in lung inflammation at day 4 after challenge in either the upper and lower airways (data not shown). However, 7 days after challenge there was significantly more inflammation in both the upper (**Figure 2E**) and lower (**Figure 2F**) airways of infected animals. This lower airway inflammation was reflected by an increased cell recovery on day 7 after infection (**Figure 2G**). Airway inflammation correlated with lung cell counts (**Figure 2H**). The composition of the lung cellular infiltrate was assessed by flow cytometry. Four days after influenza challenge, there was a significant increase in the percentage of NK cells in infected mice compared to PBS controls (**Figure 2I**). The infiltrate at day 7 was predominantly made up of CD8 T cells, with no differences in the number of CD4 T cells (**Figure 2J**) but significantly higher levels of CD8 T cells in the lungs of infected mice than controls (**Figure 2K**).

### Mice Are Protected against Homologous Challenge Infection after Inactivated or Live Attenuated Vaccine

Having developed infectious challenge models, we wished to determine the efficacy of seasonal influenza vaccines in mice. Mice

were intramuscularly immunized with purified surface antigens from A/California/7/2009, which was the (H1N1)pdm09 strain used in the trivalent vaccine from 2010 to 2016. The aim of the study was to find the lowest protective dose of vaccine, mice were given increasing doses from 0.02 to 1.5 µg A/California/7/2009 influenza hemagglutinin (as part of a mixture of viral surface antigens), for reference the human vaccine dose is 15 µg. The antibody response was proportional to the immunization dose, with most in the 1.5 µg immunized group (**Figure 3A**). Mice received a single dose of vaccine and were challenged with 2.5 × 105 PFU A/California/7/2009 H1N1 4 weeks later. Mice immunized with 1.5 or 0.5 µg lost up to 15% body weight peaking day 5 after infection. These mice were partially protected compared to the naïve animals, losing significantly less weight than naïve animals on day 6 after infection (**Figure 3B**). Mice immunized once with 0.02 µg did not produce antibodies and were not protected against challenge. To test whether repeat immunization affected the dose required, mice were immunized with 0.02, 0.01, or 0.005 µg (20, 10, or 5 ng) of A/California/7/2009 H1N1 hemagglutinin on days 0 and 21. Antibody responses were significantly greater in

mice immunized with 0.02 µg than 0.005 µg or the naïve animals (**Figure 3C**). Mice were protected against infection when mice were vaccinated twice in a prime boost regime with a dose of 0.02 or 0.01 µg and partial protection was seen after immunization with 0.005 µg (**Figure 3D**). From these studies, we observe that immunization with a very low dose of protein can protect mice against homologous influenza challenge.

## Influenza Drift Reduces the Efficacy of the Inactivated Vaccine Antigen

The biggest recent change in vaccine strains occurred with the emergence of the H1N1 pandemic strain in 2009. Because the strain recommendation preceded the emergence of the virus in 2009, the 2009–2010 vaccine did not contain the (H1N1)pdm09 like strain. However, from 2010 to 2016, the H1N1 subtype strain included in the virus was A/California/7/2009 (H1N1pdm09). By comparison, in the seven seasons since the emergence of the strain of H1N1 influenza (2009 pandemic strain) to the winter of 2017, the H3N2 component was changed four times (**Table 1**). In the same time period, the B component has changed between representative Yamagata and Victoria lineage reference strains in trivalent vaccines (13); after 2012 quadrivalent vaccines with two B strains were recommended. We wished to determine the genetic and antigenic drift of the (H1N1)pdm09 viral strains since its emergence in 2009.

We performed an integrated phylogenetic and antigenic cartography analysis (14) using hemagglutinin sequence data and HI titers for 61 (H1N1)pdm09 viruses collected between 2009 and 2016, comprising 53 viruses collected from England, 2 vaccine strains, and 6 other WHO reference viruses (Table S1 in Supplementary Material). Analysis of these genetic and antigenic data showed gradual genetic drift (**Figure 4A**) as well as gradual antigenic change (**Figure 4B**) of (H1N1) pdm09 viruses since 2009. The minimum antigenic distinction for when an influenza vaccine update is recommended is generally a difference between a vaccine strain and circulating strains of 2 antigenic map units, representing a fourfold drop in heterologous HI titer (15). Viruses circulating in England with at least 3 antigenic units (>8-fold drop in HI titer) difference from A/California/07/2009 only emerged from 2015 onward. A similar pattern was seen using multidimensional scaling (MDS) (Figure S1 in Supplementary Material). These recent viruses are antigenically similar to the updated H1N1 subtype component vaccine strain, A/Michigan/45/2015. Additionally, the recently circulating viruses include the emergence of one genetically


TABLE 1 | Recommended vaccine strains (Northern Hemisphere) 2010–2017.

distinct group of viruses, which are also genetically similar to the A/Michigan/45/2015 vaccine strain.

Since the H1 component of vaccine in use was unchanged from the initial wave of the pandemic, we wished to see whether the protection efficacy changed as the virus changed. Mice were immunized with 0.5 µg A/California/7/2009 antigens and then challenged either with a matched isolate from the initial wave of the pandemic in 2009 (A/England/195/2009), or drift isolates from 2010 (A/England/672/2010) or 2015 (A/England/336/2015). Immunized mice lost significantly less weight than control mice when infected with the 2009 (**Figure 5A**) or 2010 (**Figure 5B**) isolates. However mice infected with a 2015 isolate were not initially protected compared to the control animals, but they recovered slightly more rapidly than the unimmunized mice (**Figure 5C**). There was no significant difference in the antibody response to the immunizing antigen in the mice, suggesting that viral escape from this antigen drives the reduction in protection (**Figure 5D**).

### DISCUSSION

In the current study, we have successfully developed mouse models of seasonal H1N1 influenza infection to test vaccine efficacy. Infection with current seasonal H1N1, but not H3N2 virus, led to disease in mice. Immunization of mice with a vaccine homologous to the challenge strain, protected against infection with the same strain. However, immunization of mice with A/California/7/2009 was not protective against challenge with an H1N1 strain from 2015. This may recapitulate the situation in humans where key changes in the clade 6B H1N1 viruses, not detected by classical serological tests, reduced protection in individuals exposed to an earlier H1N1 strain (16, 17).

Influenza virus infection in mice was characterized by a large percentage of the total body weight lost at the peak of disease, in some of the animals necessitating humane culling. There was not a noticeable change in appetite, so the most likely factor is increased effort in breathing driven by the very high levels of inflammation in the lungs. Previous studies have shown that blocking TNFα blocks reduces disease by reducing cell infiltration into the lower airways (18). We have recently demonstrated that cytokine release after influenza infection is localized to the lungs (19), suggesting that the inflammation is not systemic. It was notable that while the H1N1 and B viruses caused weight loss, nothing was seen after infection with a current H3N2 strain. The most likely reason for this is differences in receptor binding by the hemagglutinin molecules of the different viruses, though we do not have data on whether the H3N2 virus infection took in the mice. The current H1N1 subtype is able to bind both avian (α2,3-linked sialic acid) and human (α2,6-linked sialic acid), whereas the current H3N2 is more human adapted and only able to bind α2,6 (20, 21). Mice, like birds, only express the α2,3-linked receptor (22). Likewise influenza B can bind both α2,3 and α2,6 sialic acids (23). It should be noted that the isolation and propagation of B viruses used in this study may have introduced key mutations leading to the loss of glycosylation sites at 196 or 197 making them antigenically different from the circulating viruses.

Mice are widely used for preclinical vaccine studies. In the current study, we demonstrate that a very low dose of protein is protective against viral infection; this was especially the case when mice were immunized twice in a prime boost regime, where a 0.02 µg dose which was not protective after a single immunization was protective when given twice. One question is why they can be protected with such small doses of protein. One consideration is the dose to size ratio; the average mouse used in these studies is 25 g, the average human 62 kg a 2,500-fold scale up. The human formulation of flu vaccine normally contains 15 µg of each hemagglutinin, so an equivalent dose for a mouse would be 6 ng, which we saw was protective in the prime boost studies. Whether body mass is the best comparison is not clear, another consideration could be muscle size, with the human muscle approximately 400 times larger. The other consideration is that the immune response amplifies signal, especially when it is boosted with the same antigen; so potentially the consideration is not about the amount of protein rather how the cells involved in the response get to the site of immunization.

Putting vaccine dosing aside, a question is whether mice are easier to protect against infection than humans. It should be noted that the viruses used in these studies were not mouse adapted strains. In our study, we were using between 104 and 106 viruses; however in human deliberate challenge studies, a similar dose is used and gives varying levels of disease (24). Both mouse and human challenge studies may not reflect the situation in natural infection where the human infectious dose is believed to be between approximately 100 and 200 infectious virions (25), and a reanalysis of the same data set suggests that disease correlates with infectious dose (26). Studies in ferrets would suggest the dose is even lower, possibly between 3 and 10 virions (27). Since more virus is required to induce disease, it may be that less antibody is required to neutralize the virus and because of the smaller size of

the mouse lung antibody may be more concentrated. While we didnot dissect the correlates of protection in the current study, in other studies the strongest correlate protection is IgG and we

titer. Study viruses (gray outline) and vaccine viruses (red outline) are highlighted.

observed that immunization induced an influenza-specific IgG response. In addition to IgG, we have recently observed a role for IgA in both human (24) and mouse (28) challenge studies.

FIGURE 5 | Antigenic drift in H1N1 strains is seen in mouse models. Mice were immunized with A/Cal/7/2009 (black squares) and challenged with A/ England/195/2009 (A), A/Eng/672/2010 (B), A/England/336/2015 (C), responses were compared to naïve animals (white circles). Antibody response to Cal/09 antigen prior to challenge (D). Points mean of *n* = 4 mice ± SEM (A–C), or individual animals (D). \**p* < 0.05, \*\**p* < 0.01.

TABLE 2 | Influenza strains used in study.


Vaccination with a protein antigen may restrict the specificity of the response to the immunizing antigen. This may especially be the case when the vaccine strain is unchanged over several rounds of immunization as was the case with the H1 antigen. Our data show a clear antigenic and immunogenic drift of the (H1N1)pdm09 virus from 2009 to 2016. Critically, a lack of protection against infection from the vaccine strain was observable with a virus isolated from the season before the vaccine strain was changed. Based on this, we would suggest that modeling in the mouse could be used to contribute to decisions about the efficacy of vaccination against the currently circulating strains of influenza H1N1.

### MATERIALS AND METHODS

### Viruses

Seasonal influenza viruses (**Table 2**) were isolated by Public Health England (UK). The England strains of H1N1, A/England/195/2009, A/England/672/2010, and A/England/336/2015 were isolated in SIAT-MDCK cells (12). B viruses were expanded in eggs prior to being grown in Madin-Darby Canine Kidney (MDCK) cells. Prior to use in mice, viruses were propagated in MDCK cells, in serum-free DMEM supplemented with 1 µg/ml trypsin. The virus was harvested 3 days after inoculation and stored at -80°C. Viral titer was determined by plaque assay as described below.

### Mouse Immunization and Infection

6–10-week-old female CB6F1 mice were obtained from Harlan UK Ltd. (Horsham, UK) or from an internal breeding colony and kept in specific-pathogen-free conditions in accordance with the United Kingdom's Home Office guidelines and all work was approved by the Animal Welfare and Ethical Review Board (AWERB) at Imperial College London. Studies followed the ARRIVE guidelines. Mice were immunized intramuscularly (i.m.) with purified surface antigens from influenza strain H1N1 A/California/7/2009 (GSK Vaccines, Siena, Italy) in 50 µl, either once (prime only) or twice (prime boost). For infections, mice were anesthetized using isoflurane and infected intranasally (i.n.) with 100 µl influenza virus or sterile PBS. Body surface temperature was taken from the xiphoid process using a handheld infrared thermometer.

### Tissue and Cell Recovery and Isolation

Mice were culled using 100 µl intraperitoneal pentobarbitone (20 mg dose, Pentoject, Animalcare Ltd., UK) and tissues collected as previously described (29). Blood was collected from femoral veins and sera isolated after clotting by centrifugation. Lungs were removed and homogenized by passage through 100-µm cell strainers, then centrifuged at 200 × *g* for 5 min. Supernatants were removed and the cell pellet treated with red blood cell lysis buffer (ACK; 0.15 M ammonium chloride, 1 M potassium hydrogen carbonate, and 0.01 mM EDTA, pH 7.2) before centrifugation at 200 × *g* for 5 min. The remaining cells were resuspended in RPMI 1640 medium with 10% fetal calf serum, and viable cell numbers determined by trypan blue exclusion.

### Histology

Upper and lower regions of paraformaldehyde-fixed left lung lobes were processed and embedded in paraffin. Sections of 3 µm were stained with hematoxylin and eosin and the entire section was scanned at ×20 magnification so that the area with the greatest inflammation could be assigned the inflammation score. The degree of airway inflammation was assessed in a blinded manner using a modified system described previously (30). Briefly, the degree of inflammation in the peribronchiolar, perivascular, and interstitial regions of both the upper and lower airways was assessed. A value of 0 (none), 1 (minimal), 2 (mild), 3 (moderate), or 4 (severe) was given to each histological site and the sum of these scores was used as the total upper/lower respiratory inflammation score.

# Influenza Viral Load

### Viral RNA Quantification

Viral load *in vivo* was assessed by Trizol extraction of RNA from frozen lung tissue disrupted in a TissueLyzer (Qiagen, Manchester, UK). RNA was converted into cDNA and quantitative RT-PCR was carried out using bulk viral RNA, for the influenza M gene and mRNA using 0.1 µM forward primer (5′-AAGACAAGACCAATYCTGTCACCTCT-3′), 0.1 µM reverse primer (5′-TCTACGYTGCAGTCCYCGCT-3′) and 0.2 µM probe (5′-FAM-TYACGCTCACCGTGCCCAGTG-TAMRA-3′) on a Stratagene Mx3005p (Agilent technologies, Santa Clara, CA, USA). M-specific RNA copy number was determined using an influenza M gene standard plasmid.

### Plaque Assays

Plaque assays were performed using a modified protocol previously described (31). Briefly, confluent monolayers of MDCK cells in 12-well plates were inoculated with 200 µl of viral or sample dilutions and incubated for 1 h. The inoculum was removed then the cells were overlaid with 0.6% agarose (Oxoid) in MEM including 1 µg/ml trypsin and incubated at 37°C with 5% CO2. After 3 days, the agarose was removed and the cells stained with crystal violet dissolved in methanol and water.

### Flow Cytometry

Live cells were suspended in Fc block (Anti-CD16/32, BD) in PBS-1% BSA and stained with surface antibodies: CD3-FITC (BD, Oxford UK), CD4-APC (BD), CD8-APC Alexa75 (Invitrogen, Paisley, UK), and NK1.1 PerCP-Cy5.5 (BD, Oxford UK). Analysis was performed on an LSRFortessa flow cytometer (BD). FMO controls were used for surface stains.

### Semi-Quantitative Antigen-Specific ELISA

Antibodies specific to influenza H1N1 were measured using a standardized ELISA (32). IgG responses were measured in sera. MaxiSorp 96-well plates (Nunc) were coated with 1 µg/ml surface proteins or a combination of anti-murine lambda and kappa light chain specific antibodies (AbDSerotec, Oxford, UK) and incubated overnight at 4°C. Plates were blocked with 1% BSA in PBS. Bound IgG was detected using HRPconjugated goat anti-mouse IgG (AbD Serotec). A dilution series of recombinant murine IgG was used as a standard to quantify specific antibodies. TMB with H2SO4 as stop solution was used to detect the response and optical densities read at 450 nm.

### Integrated Analysis of Antigenic and Genetic Evolution

A Bayesian multidimensional scaling (BMDS) model (14) of antigenic cartography (15) was implemented in BEAST v1.8.4 (33) to jointly infer antigenic and phylogenetic relationships, as previously described (14, 34). Briefly, an antigenic dataset composed of available HA gene sequences and corresponding HI measurements was assembled for 61 H1N1pdm09 viruses collected between 2009 and 2016: 53 viruses circulating in England, 2 vaccine strains, and 6 WHO reference viruses. Viruses were originally isolated from clinical specimens either by WHO NICs or by the WHO Collaborating Center and the corresponding HI measurements were obtained either from the Francis Crick Institute Influenza Interim Reports or were kindly provided by John McCauley. HA sequences were downloaded from either the Influenza Research Database (35) or the EpiFlu database (36). A phylogenetic tree of the HA sequences was estimated using BEAST (33) which incorporated the HKY substitution model, a coalescent model with constant effective population size and a strict molecular clock. Markov chain Monte Carlo (MCMC) was run for 15 million steps and trees were logged every 1,500 steps, with a burn-in of 5 million steps, resulting in 10,000 trees. This posterior set of 10,000 trees was used with the HI data to implement the full BMDS model infer virus and serum locations in two antigenic dimensions, as well as virus avidities, serum potencies, MDS precision, and virus and serum location precisions in BEAST. MCMC chains were run for 500 million states with sampling every 200,000 states with 10% burn-in, and run convergence was checked in Tracer v1.6 (http://tree.bio.ed.ac.uk/ software/tracer/). A maximum clade credibility tree was summarized in TreeAnnotator v1.8.4 (33) and visualized using FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/). Antigenic map plots were generated using custom Python scripts with the matplotlib library (37).

### Statistical Analysis

Calculations as described in figure legends were performed using Prism 6 (GraphPad Software Inc., La Jolla, CA, USA).

# ETHICS STATEMENT

Work was performed in accordance with the United Kingdom's Home Office guidelines and all work was approved by the Animal Welfare and Ethical Review Board (AWERB) at Imperial College London. Studies followed the ARRIVE guidelines.

# AUTHOR CONTRIBUTIONS

HG, JM, and EK performed the experimental studies; PL and PK performed the data analysis of flu strain drift; JM provided HI data and analysis; JE, CT, RE, LP, and WB provided and grew influenza strains, JT designed the studies and wrote the paper.

## ACKNOWLEDGMENTS

The authors thank Stephen Reece for advice on manuscript. The authors have no commercial conflicts of interest in this study. Giuseppe Del Giudice (GSK Vaccines, Sienna) provided the influenza vaccine antigens. The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. [115308] Biovacsafe, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA members'

### REFERENCES


in kind contribution. This work was supported by the European Community's European seventh Framework Program ADITEC (HEALTH-F4-2011-18 280873).

# FUNDING

The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. [115308] Biovacsafe, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA members' in kind contribution. This work was supported by the European Community's European Seventh Framework Program ADITEC (HEALTH-F4-2011-18 280873). The work done at the Crick Worldwide Influenza Centre, a WHO Collaborating Centre for Reference and Research on Influenza, was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001030), the Medical Research Council (FC001030), and the Wellcome Trust (FC001030).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00126/ full#supplementary-material.


with enhanced histopathological disease in bonnet monkeys (*Macaca radiata*) pre-immunized with a formalin-inactivated RSV vaccine. *J Gen Virol* (2001) 82:2663–74. doi:10.1099/0022-1317-82-11-2663


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Groves, McDonald, Langat, Kinnear, Kellam, McCauley, Ellis, Thompson, Elderfield, Parker, Barclay and Tregoning. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Helen T. Groves1 , Leah Cuthbertson2,3, Phillip James 2,3, Miriam F. Moffatt 2,3†, Michael J. Cox 2,3\*† and John S. Tregoning1 \*†*

*1Mucosal Infection and Immunity Group, Department of Medicine, Section of Virology, St. Mary's Campus, Imperial College London, London, United Kingdom, 2National Heart & Lung Institute, Imperial College London, London, United Kingdom, 3Respiratory Biomedical Research Unit, Royal Brompton & Harefield NHS Trust, Imperial College London, London, United Kingdom*

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Randy A. Albrecht, Icahn School of Medicine at Mount Sinai, United States Raffael Nachbagauer, Icahn School of Medicine at Mount Sinai, United States*

### *\*Correspondence:*

*Michael J. Cox michael.cox1@imperial.ac.uk; John S. Tregoning john.tregoning@imperial.ac.uk*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 12 October 2017 Accepted: 22 January 2018 Published: 12 February 2018*

### *Citation:*

*Groves HT, Cuthbertson L, James P, Moffatt MF, Cox MJ and Tregoning JS (2018) Respiratory Disease following Viral Lung Infection Alters the Murine Gut Microbiota. Front. Immunol. 9:182. doi: 10.3389/fimmu.2018.00182*

Alterations in the composition of the gut microbiota have profound effects on human health. Consequently, there is great interest in identifying, characterizing, and understanding factors that initiate these changes. Despite their high prevalence, studies have only recently begun to investigate how viral lung infections have an impact on the gut microbiota. There is also considerable interest in whether the gut microbiota could be manipulated during vaccination to improve efficacy. In this highly controlled study, we aimed to establish the effect of viral lung infection on gut microbiota composition and the gut environment using mouse models of common respiratory pathogens respiratory syncytial virus (RSV) and influenza virus. This was then compared to the effect of live attenuated influenza virus (LAIV) vaccination. Both RSV and influenza virus infection resulted in significantly altered gut microbiota diversity, with an increase in *Bacteroidetes* and a concomitant decrease in *Firmicutes* phyla abundance. Although the increase in the *Bacteroidetes* phylum was consistent across several experiments, differences were observed at the family and operational taxonomic unit level. This suggests a change in gut conditions after viral lung infection that favors *Bacteroidetes* outgrowth but not individual families. No change in gut microbiota composition was observed after LAIV vaccination, suggesting that the driver of gut microbiota change is specific to live viral infection. Viral lung infections also resulted in an increase in fecal lipocalin-2, suggesting low-grade gut inflammation, and colonic Muc5ac levels. Owing to the important role that mucus plays in the gut environment, this may explain the changes in microbiota composition observed. This study demonstrates that the gut microbiota and the gut environment are altered following viral lung infections and that these changes are not observed during vaccination. Whether increased mucin levels and gut inflammation drive, or are a result of, these changes is still to be determined.

Keywords: influenza, respiratory syncytial virus infections, gut microbiota, *Bacteroidetes*, *Firmicutes*, Mucin 5ac

# INTRODUCTION

The bacteria that colonize the gastrointestinal tract, known collectively as the gut microbiota, play many roles in maintaining human health, such as promoting the development of the mucosa and aiding nutrient metabolism. Importantly, the microbiota protects against enteropathogen colonization through a range of mechanisms including the production of antimicrobial peptides, competition

**129**

for resources, and the induction of local immune responses (1). The gut microbiota also has systemic influences outside the gastrointestinal tract, from the wide-ranging anti-inflammatory effects of bacterial metabolites like short-chain fatty acids to the alteration of neurotransmitter production in the central nervous system (2, 3). One of the most studied areas is the effect of the gut microbiota on immune responses. Germ-free mice, which lack a microbiota, have reduced expression of antimicrobial peptides, fewer antibody-secreting cells, and deficiencies in T cell function (4, 5) which result in reduced responses to influenza virus infection and vaccination (6).

Given its impact on health, research has focused on understanding what factors influence the composition of the gut microbiota. Potentially, the largest contributor to gut microbiota composition is diet. It is thought that the distinct enterotypes into which most human gut microbiotas fall are shaped by diet (7), and everything from the amount of coffee consumed to bread-type preference has been linked to gut microbiota composition (8). Similarly, medication, particularly antibiotic use, can significantly alter gut microbiota composition, and even a short course of antibiotics can have long-lasting effects (9). Infection also shapes the gut microbiota, but the majority of research has focused on gastrointestinal infections or infections that have an impact on the immune response, such as HIV (10, 11). Despite their very high prevalence, little is known about how lung infections affect the gut microbiota.

Lung infections are the leading cause of death in lower-income countries (12) and are the single biggest cause of death in children under the age of 5 (13). Respiratory syncytial virus (RSV) in particular is the most common cause of bronchiolitis and pneumonia in infants (14), and global seasonal influenza virus epidemics are thought to result in three to five million severe infections every year (15). The importance of respiratory infections on global health has led many microbiota researchers to investigate how the gut microbiota might influence lung infection; for example, the depletion of the gut microbiota in mice with antibiotics has been associated with reduced influenza virus-specific T cell and antibody generation (16). Likewise, alveolar macrophages from germ-free mice have been shown to have reduced a phagocytic capacity leading to an increased susceptibility to pneumonia (17). As the high morbidity and mortality associated with lung infections are often due to lack of effective vaccines, there is significant interest in whether the gut microbiota could be manipulated to improve vaccine efficacy as well as response to infection (18).

Whether the "gut-lung axis" is bidirectional and lung infections influence the gut microbiota is currently under debate. Several studies have been published in the last few years on the impact of influenza virus infection on the gut microbiota, but the mechanisms presented behind the changes are conflicting, and further, more in-depth, characterization studies are required. In the present study, we investigated the impact of viral lung infection on the gut microbiota using both RSV and influenza virus as infection models. We also compared this to the effect of protective live attenuated influenza virus (LAIV) vaccination on the gut microbiota as many studies looking to alter the gut microbiota before or during vaccination do not take the effect that vaccination itself may have into account.

# MATERIALS AND METHODS

### Animal Experiments

Specific pathogen-free 10- to 12-week-old female BALB/c mice were purchased from Charles River Laboratories (Margate, UK). Mice were maintained in individually ventilated autoclaved cages, with Tapvei Eco Pure Premium Aspen Chips for bedding (Datesand) and Sizzle Pet for nesting material (LBS), in groups of five animals per cage. Mice were fed irradiated SDS RM3 pellets (LBS) and received reverse osmosis, autoclaved water *ad libitum*. The same specific pathogen-free room was used to house all mice and was maintained on a 12-h light/dark cycle at 20–24°C with 55 ± 10% humidity. For infection studies, mice were anesthetized *via* isoflurane inhalation and infected intranasally with 100 µl of 2 × 106 PFU/ml RSV-A2, 100 µl of 4 × 105 PFU/ml A/Eng/195/2009 influenza virus, or 100 µl phosphate-buffered saline (PBS). For vaccination studies, mice were anesthetized and infected intranasally with 1 × 106 PFU/ml LAIV vaccine (Fluenz® Tetra, MedImmune, 2016/2017 season), a dose shown to be protective in mice against challenge with 4 × 105 PFU/ml A/Eng/195/2009 influenza virus. Mice were weighed daily after infection/vaccination, placed into individual disinfected pots, and feces were collected using autoclaved tweezers and stored in sterile tubes at −80°C. All animal experiments were performed in accordance with the United Kingdom's Home Office guidelines under animal study protocol number one, and all work was approved by the Animal Welfare and Ethical Review board at Imperial College London. Studies followed the ARRIVE guidelines, and all animal infections and infectious work were carried out in biosafety level-two facilities.

# RSV L Gene qPCR

Viral load was assessed by extracting RNA from frozen lung and colon tissue disrupted in a TissueLyzer (Qiagen, Manchester, UK) using Trizol and converting it into cDNA using Omniscript RT Kit (Qiagen, Manchester, UK). RT-PCR was carried out using bulk viral RNA, for the RSV L gene and mRNA using the following primers: 5'-GAACTCAGTGTAGGTAGAATGTTTGCA-3', 5'-TTCAGCTATCATTTTCTCTGCCAA-3' and probe: 5'-FAM-TTTGAACCTGTCTGAACAT-TAMRA-3' on a Stratagene Mx3005p (Agilent technologies, Santa Clara, CA, USA). RNA copy number was determined using an RSV L gene standard.

### 16S rRNA Gene qPCR

Bacterial DNA was extracted from 30-mg feces/mouse using the FastDNA® Spin Kit for Soil (MP Biomedicals). The 16S rRNA gene was quantified using the SYBR Fast qPCR Kit Master Mix (2X, KAPA Biosystems, KK4601) and the following primers: 5'-AYTGGGYDTAAAGNG-3', 5'-TACNVGGG TATCTAATCC-3' (19). Reactions were performed in triplicate alongside a cloned *Vibrio natriegens* full-length 16S rRNA gene standard. Reaction plates were run on the ViiA7 Real-Time PCR System using the following run parameters: 90°C for 3 min (95°C for 20 s, 50°C for 30 s, 72°C for 30 s), 40 cycles, 10 min at 12°C.

### 16S rRNA Gene Sequencing

The V4 variable region of the 16S rRNA gene was amplified by PCR using 1 µl of sample DNA, Q5® Hot Start High-Fidelity 2X Master Mix (NEB, M0494S), and the universal bacterial primers S-D-Bact-0564-a-S-15: 5' AYT GGG YDT AAA GNG 3' and S-D-Bact-0785-b-A-18: 5' TAC NVG GGT ATC TAA TCC 3' which were uniquely barcoded for each sample [barcodes: Illumina Nextera indexes version 2, primers: Klindworth et al. (19)]. The 16S rRNA gene library was amplified using the following run parameters: 95°C for 2 min (95°C for 20 s, 50°C for 30 s, 72°C for 5 min), 32 cycles. Each PCR was run in quadruplicate with a negative and positive control in each run. The library was purified using AMPure® XP beads (Beckman Coulter, A63880) and quantified using the PicoGreen® quantification assay for double-stranded DNA (Thermo Fisher, P11496). Samples were equi-molar pooled to 45 ng/sample, and the pooled library was purified again and concentrated using AMPure® beads followed by agarose gel purification. Prior to sequencing, library quality was accessed by profiling using the High Sensitivity DNA Kit (Agilent Technologies, 5067-4626) on an Agilent 2100 Bioanalyzer and quantified using the illumina Library Quantification Kit. Paired-end sequencing of an 8-pM denatured library, spiked with 8 pM of PhiX, was performed using the Illumina MiSeq platform (20).

### Bioinformatics

16S rRNA-sequencing data were processed using QIIME 1.9.0 software suite (21). Sequences were trimmed, forward and reverse reads were paired, demultiplexed, and any Phix contamination removed using Burrows–Wheeler Aligner (22). Operational taxonomic units (OTUs) were clustered at 97% sequence identify using the UCLUST OTU-clustering tool (23) using open reference clustering, and representative OTUs were picked using the SILVA 115 rRNA database (24). Sequences were aligned using PyNAST (25). Chimeric sequences were identified and removed using ChimeraSlayer (26). Taxonomy was assigned using the RDP classifier (27) and the SILVA 115 rRNA database for reference sequences (24). An approximately maximumlikelihood phylogenetic tree was built using FastTree 2.1.3 (28). Microbiota analysis was conducted in R 3.3.0 (29) with RStudio (30) using the phyloseq package (31) unless otherwise specified. Beta diversity was analyzed using both the phyloseq and the vegan package (32). Beta diversity measures the difference in overall bacterial community composition between different samples. To do this, a distance matrix using the Bray–Curtis dissimilarity index, which calculates differences in bacterial OTU abundance between samples, was created and then analyzed using non-metric multidimensional-scaling (NMDS) ordination. NMDS ranks the order of inter-sample distances which is then represented by the position of samples in the twodimensional ordination map; the closer together the two sample points are, the more similar their microbiota composition is (33). For each grouping variable, 95% confidence ellipses were calculated using the vegan package; overlapping ellipses generally indicate that microbiota composition is not significantly different although this was formally tested using Permutational Multivariate Analysis of Variance (PERMANOVA). Differences in OTU abundance were calculated using the DESeq2 package on unrarefied/untransformed data (34).

## Data Availability

Sequencing data will be uploaded to the European Nucleotide Archive under the accession number PRJEB21782. Metadata, mapping files, OTU tables, phylogenetic trees, and codes used for analysis will be uploaded to BioStudies at EMBL-EBI.

### Histology

Transverse 4-µm sections of colon and longitudinal 4-µm section of lung were cut and stained with hematoxylin and eosin (35). Airway inflammation was assessed in a blinded manner using a system similar to that used by Ponnuraj et al. (36). The degree of colonic inflammation was assessed by counting and measuring the length and width of lymphoid aggregates across the entire section at ×20 magnification using a 10-µm eyepiece graticule.

# Cytokine ELISA

Airway lavage and colon lavage fluid were collected during culling by flushing the airways and colon with 1,000 and 100 µl, respectively, of PBS. Cytokine levels in the airway and colon lavage were assessed using Mouse IFN-y, IL-13, or IL-17 DuoSets (R&D Systems, Abingdon, UK).

# Mucin ELISA

ELISAs to measure the level of Mucin 5 ac (Muc5ac) and Muc2 in the airway and colon lavage fluid were adapted from previously published protocols (37, 38). Briefly, plates were pre-incubated with carbonate buffer (pH 9.5). Airway and colonic lavage samples were then added to the carbonate buffer and incubated at 37°C overnight. Plates were blocked with 1% BSA PBS. Mucin was detected with anti-Muc5ac antibody (45M1, Thermo Fisher Scientific, UK) or anti-Muc2 antibody (ab76774, Abcam, UK). For Muc5ac, goat α-mouse HRP secondary antibody was used; for Muc2, goat α-rabbit HRP secondary antibody was used. ELISAs were developed using TMB (Thermo Fisher Scientific, UK) and stopped using 2 N H2SO4. Plates were read at 540 nm using a Fluostar (Omega).

# Fecal Lipocalin-2

Levels of fecal lipocalin-2 are considered a more sensitive marker of low-grade intestinal inflammation compared to histological analysis (39). Lipocalin-2 levels were measured in 100-mg/ml feces (reconstituted in PBS 0.1% Tween 20) using Mouse Lcn-2 DuoSet ELISA kit (R&D Systems, Abingdon, UK) as described previously (40).

# Statistics

Statistics were performed using either GraphPad Prism V6 or R 3.3.0. Two-way repeated measures analysis of variance (ANOVA) with Dunnett's correction was used to test for significant differences in bacterial load, alpha diversity, and phyla/family abundance before infection/dosing and after. Significant changes in airway/colon inflammation and cytokine/mucin ELISAs were measured using one-way ANOVA to compare infected with PBS controls with Dunnett's correction. Significant changes in beta diversity were calculated using PERMANOVA on the Bray–Curtis distance matrix. Using the DESeq2 package, only OTUs, which significantly changed in abundance by *p* ≤ 0.01, were selected with Benjamini–Hochberg multiple-inference correction.

### RESULTS

### The Composition of the Gut Microbiota Is Altered following Lung Infection

The aim of the study was to investigate the effect of viral lung infection on the gut microbiota. Mice were intranasally dosed with RSV, PBS, or left naïve. RSV-infected animals lost weight on days 1, 5, 6, and 7 following infection, while PBS and naïve animals experienced no significant weight loss (**Figure 1A**). Viral load in the airways was highest at day 4 after infection and was almost undetectable by day 7 as has been shown previously (41), despite day 7 being associated with peak weight loss for RSV-A2 infection (42) (**Figure 1B**). No RSV RNA was detected in the colon of infected mice at either time point (**Figure 1C**). To confirm that feces were a reasonable substitute for sampling the colonic environment (43), we compared the microbiota composition of colonic and fecal samples taken from the same mice (**Figure 1D**) and found no significant differences. Therefore, for further subsequent experiments, feces were used to monitor changes in the gut microbiota. Lung infection did not alter the total fecal bacterial load estimated using either 16S rRNA gene copy number (**Figure 1E**), total observed OTU (**Figure 1F**), or alpha diversity (**Figures 1G,H**). While the total bacterial load remained constant, the composition significantly changed on days 4 and 7 after RSV infection compared to day 0 (*p*= 0.006, **Figure 1I**). Composition changes were not observed among PBS-dosed or -naïve animals. From this, we conclude that RSV infection results in global changes to the gut microbiota.

Having seen overall diversity changes, we wished to dissect these changes at a phyla level. The dominant phyla in all mice, before and after infection, were *Bacteroidetes* and *Firmicutes* (~97–99% combined total relative abundance; data not shown). Other phyla detected were *Tenericutes*, *Actinobacteria*, *Proteobacteria*, and *Deferribacteres* but as these individually never exceeded ~2% relative abundance before or after infection, we decided to focus on changes in the dominant phyla. We observed a significant increase in the relative abundance of *Bacteroidetes* and a corresponding decrease in *Firmicutes* from days 0 to 7 after RSV infection (**Figure 2A**). The increase in *Bacteroidetes* after lung infection was driven by a significant increase in the relative abundance of the *Bacteroidaceae* family at day 7 (*p* ≤ 0.001; **Figure 2B**). Significant changes in individual bacterial OTUs belonging to the *Bacteroidaceae* and S24\_7 families were also observed (**Figure 2C**). The decrease in *Firmicutes* was associated with a significant decrease in the relative abundance of both the *Lachnospiracea*e (*p* ≤ 0.05) and the *Lactobacillaceae* (*p* ≤ 0.01) families. While no changes were seen in naïve mice, mice dosed with PBS had an increase in the *Firmicutes* phylum and a decrease in *Bacteroidetes*, driven by an increase in the *Lachnospiraceae* family (*p* ≤ 0.01) (**Figure 2D**). Therefore, the composition of the gut microbiota was significantly altered following lung infection with a specific enrichment for *Bacteroidetes*.

### Changes in the Microbiota after Lung Infection Are Not Influenced by Cage Effect or Passed on to Uninfected Mice

The microbiota can be affected by a wide range of environmental stimuli and, although housing conditions were the same for all groups in this study, one potential variable was that since the infected animals were housed separately from controls, changes might be driven by cage-specific effect (44). While RSV is not passed between mice, effects on the microbiota could be transferred by coprophagy. To test whether there were cage effects and whether RSV infection-associated changes in the gut microbiota could be passed to non-infected cage mates, RSV-infected mice were mixed with PBS-dosed and -naïve cage mates. Housing infected mice separately or together with control animals had no effect on weight loss; infected animals lost weight while control animals did not (**Figures 3A,B**). The beta diversity of the fecal microbiota significantly changed after RSV infection for mice either separately (*p* = 0.012) or co-housed (*p* = 0.029; **Figure 3C**). Co-housing control mice with infected mice had no effect on the beta diversity of the control mice. Following RSV infection, there was an increase in the relative abundance of *Bacteroidetes* and a corresponding decrease in *Firmicutes* at day 7, and this was unaffected by co-housing with control animals (**Figure 3D**). Therefore, cage effect did not have an impact on the microbiota changes seen after RSV infection.

Interestingly, while there was an increase in the relative abundance of families belonging to the *Bacteroidetes* phylum in both the separately and co-housed mice after RSV infection, this increase was driven by the significant expansion of different families within *Bacteroidetes*. RSV-infected animals in separate cages had a significant increase in the *Porphyromonadaceae* family (*p* ≤ 0.05; **Figure 3E**) while infected animals co-housed in mixed groups had an increase in the S24\_7 family (*p* ≤ 0.05). The reduction in *Firmicutes* phylum was associated with a significant decrease in *Lachnospiraceae* and *Lactobacillaceae* in both cases.

Respiratory syncytial virus-infected mice took 28 days after infection to return to their original starting weight (**Figure 3F**). Changes in microbiota were transient, preceded weight recovery, and by day 14, there was no difference in the relative percentage abundance of either phylum *Bacteroidetes* or *Firmicutes* when compared to preinfection, demonstrating the resilience of the gut microbiota (**Figures 3G,H**).

### Influenza Virus Infection but Not Vaccination Alters the Composition of the Gut Microbiota

To establish whether changes observed in the gut microbiota after lung infection were RSV-specific, mice were intranasally infected with H1N1 influenza A and feces collected before and

Figure 1 | Gut microbiota diversity is altered following viral lung infection. Adult BALB/c mice were intranasally dosed with 2 × 106 PFU/ml respiratory syncytial virus (RSV)-A2, sterile phosphate-buffered saline (PBS) or untreated (naïve). Feces were collected under sterile conditions before infection (D0) and at days 4 (D4) and 7 (D7) after infection. (A) Weight was measured after dosing. (B) Viral load in the lungs and colon (C) was estimated using RSV L gene qPCR at D4 and D7 after infection (limit of detection *LD* for the assay was 2,800 copies; not detected *ND* 0 copies/no CT value). (D) Colonic microbiota composition (red) was compared to the fecal microbiota composition (black) of the same mice (shapes represent individual mice). (E) Bacterial load in the feces was estimated using 16S rRNA qPCR. (F) Number of operational taxonomic units (OTUs) in feces before and after infection. (G,H) Alpha diversity of the gut microbiota was analyzed using the phyloseq package in R v3.4.1. (I) Beta diversity of the fecal microbiota was analyzed using non-metric multidimensional scaling (NMDS) on a Brays–Curtis distance matrix. *N* = 5 mice. (E,H) Colored points represent indicial mice. Two-way repeated measures Analysis Of Variance with Dunnett's correction was used to test for significant differences in viral and bacterial load. Significant changes in microbiota diversity were tested for using Permutational Multivariate Analysis of Variance. \**p* ≥ 0.05, \*\**p* ≥ 0.01, \*\*\**p* ≥ 0.001.

cutoff for significance). Two-way repeated measures Analysis of Variance with Dunnett's correction was used to test for significant differences in phyla and family abundance. \**p* ≥ 0.05, \*\**p* ≥ 0.01.

after (D7) infection. To compare the effect of infection versus vaccination on the gut microbiota, a separate group of mice were intranasally vaccinated with a dose of LAIV known to confer protection against infection in mice (**Figure 4A**). Mice infected with influenza had significantly altered gut microbiota diversity (*p*= 0.008; **Figure 4B**), whereas there was no change in gut microbiota diversity after LAIV vaccination. Influenza virus infection also led to an increase in the ratio of *Bacteroidetes* to

Figure 3 | Respiratory syncytial virus (RSV) infection—associated changes in the gut microbiota are not due to cage effect or passed on to cage mates. BALB/c mice were intranasally infected with 2 × 106 PFU/ml RSV-A2, dosed with phosphate-buffered saline (PBS) or untreated, animals were either housed (A) separately by treatment regime or (B) co-housed with mice receiving a different treatment. (A,B) Weight was measured after treatment. (C) Feces were collected before (D0) and after infection (D7), and beta diversity of fecal microbiota assessed. (D) Relative abundance in microbiota at the phyla and (E) family levels. (F) BALB/c mice infected with RSV and housed separately were allowed to recover their lost weight. (G) Further fecal samples were taken at D14, D21, and D28, and relative abundance at phyla (H) and family levels assessed. *N* = 5, points represent the mean ± SEM. Two-way repeated measures Analysis Of Variance with Dunnett's correction for multiple comparisons was used to test for significant weight loss and changes in phyla and family abundance. \**p* ≥ 0.05. Non-metric multidimensional scaling (NMDS) on a Brays–Curtis distance matrix was used to visualize diversity, and significant changes in diversity were analyzed using Permutational Multivariate Analysis of Variance, comparing D0–D7 for all groups and D0 between groups.

*Firmicutes* (**Figure 4C**). This increase was driven by an increase in the relative abundance of the S24\_7 family (*p* ≤ 0.01) and the *Porphyromonadaceae* family (*p* ≤ 0.05; **Figure 4D**). No changes in phyla or family abundance among vaccinated or control animals were observed (**Figures 4C,D**). Therefore, following both RSV and influenza virus infection, there is an increase in the abundance of the *Bacteroidetes* phylum while LAIV vaccination does not have a significant impact on the gut microbiota.

# Lung Infection Is Associated with an Increase in Low-Grade Gut Inflammation and Colonic Muc5ac Levels

Previous studies investigating the connection between the gut microbiota and lung infections have focused on the exploration of potential immunological mechanisms (16, 17, 45). In the present study, significant upper airway inflammation after RSV infection and significant upper and lower airway inflammation after influenza virus infection were observed (**Figure 5A**).

infection. Feces were collected before (D0) and after primary infection/immunization (D7). (A) Weight loss was recorded for 7 days after primary infection/ immunization and for an additional 7 days after immunized mice were challenged. (B) Beta diversity of the fecal microbiota before and after influenza virus infection, LAIV immunization, and PBS dosing. (C) The relative abundance of the *Firmicutes* and *Bacteroidetes* phyla. (D) The relative abundance of *Bacteroidetes* and *Firmicutes* splits into the most abundant families (>1% total abundance) before and after influenza virus infection, LAIV immunization, and PBS dosing. *N* = 5, points represent the mean ± SEM. Two-way repeated measures analysis of variance with Dunnett's correction for multiple comparisons was used to test for significant weight loss and changes in phyla and family abundance. \**p* ≥ 0.05, \*\**p* ≥ 0.01, \*\*\**p* ≥ 0.001. Non-metric multidimensional-scaling (NMDS) on a Brays–Curtis distance matrix was used to visualize diversity, and significant changes in diversity were analyzed using Permutational Multivariate Analysis of Variance comparing D0–D7.

However, there was no evidence of any significant histological colonic inflammation after either RSV or influenza infection (**Figures 5B,C**). Fecal levels of lipocalin-2 are considered a more sensitive marker of gut inflammation that histological analysis (39); we observed significantly higher fecal lipocalin-2 levels after RSV infection, suggesting that viral lung infections may result in low-grade gut inflammation (**Figure 5D**). To explore this further, levels of cytokines in airway and colonic lavage were measured after RSV infection. IFN-γ was elevated after RSV infection in the airways but not in the colonic lavage, and no significant increases in levels of IL-13 or IL-17 at either site were found (**Figure 5E**). RSV and influenza viruses, similar to most respiratory pathogens, cause elevated mucus secretion in the airways. As mucus is nutrient rich, raising its level in the guts either by swallowing airway mucus or by systemic hypersecretion could explain the bloom in certain

Colonic inflammation was scored by counting and measuring the number (B) and size (C) of lymphoid aggregates in the colonic epithelium. (D) Low-grade gut inflammation was assessed by measuring fecal lipocalin-2 levels. (E) IL-17, IL-13, and IFN-y cytokine levels were measured in the bronchoalveolar lavage fluid (airways) and colonic lavage fluid (colon) after respiratory syncytial virus (RSV) infection or phosphate-buffered saline (PBS) dosing. (F) Muc5ac levels were measured in the airway and colon after RSV infection, H1N1 infection, and PBS dosing. *N* = 5–10 mice/group ± SEM, two-way analysis of variance. \**p* ≥ 0.05, \*\**p* ≥ 0.01, \*\*\**p* ≥ 0.001.

microbiota members. Many members of the gut microbiota, including members of the *Bacteroidetes* phylum, use mucus as an energy source (46). Interestingly, Muc5ac levels were significantly increased in both the airways and the colon of RSV or influenza virus-infected mice but not in those of control mice (**Figure 5F**). These findings suggest that the changes in gut microbiota composition observed after both lung infections could be due to the mucus hypersecretion induced by both respiratory viruses.

### DISCUSSION

In the current study, we demonstrate that the composition of the gut microbiota changes after viral lung infection. This study also contributes to understanding the complex relationship between the gut microbiota and influenza infection and, for the first time, characterizes how the gut microbiota is altered following RSV infection. Viral lung infection led to an increase in the phylum *Bacteroidetes* with a corresponding decrease in the *Firmicutes* phylum. The constituent bacteria that drove these changes varied at the family and OTU levels, suggesting that viral lung infection creates conditions favorable to support a *Bacteroidetes* bloom, but that the species that make up this bloom are selected by currently unknown factors. In addition, we observed no change in gut microbiota diversity or composition after LAIV vaccination, emphasizing that the observed changes are dependent upon infection and providing a preliminary baseline for future studies investigating the impact of vaccination on the gut microbiota and *vice versa*.

As in humans, the majority of the murine gut and fecal microbiota belongs to either the *Firmicutes* or *Bacteroidetes* phyla (47, 48), and a change in the balance between the two has been previously implicated in many diseases and disorders (49). We observed an increase in *Bacteroidetes* and a decrease in *Firmicutes* in the gut microbiota of mice following lung infection. A decrease in the relative abundance of *Firmicutes* has been reported in a chronic mild stress model in mice, with a specific reduction of gut *Lactobacillus* (50), suggesting that changes in gut microbiota composition observed by us and others may be more reflective of a stress response. A decrease in *Lactobacillus*, detected by PCR, has previously been observed after influenza infection (51), and in this study, we consistently saw a decrease in the *Lactobacillaceae* family after viral lung infection. Other studies investigating the effect of influenza virus in mice have highlighted an increase in the *Proteobacteria* phylum after infection (52, 53). In our study, the relative abundance of *Proteobacteria* never exceeded ~0.01%, and there was no change after infection. Although no significant increase in the relative abundance of the *Bacteroidetes* phylum was observed in these previous studies, the abundance of OTUs belonging to *Bacteroidetes* correlated most significantly with weight loss after influenza infection (52). No comparative study in humans investigating the direct effect of lung infections on the gut microbiota has currently been published. However, one group, which profiled the gut microbiota of infants with and without bronchiolitis, found that there was a strong positive association between a high *Bacteroidetes* abundance in the gut microbiota and bronchiolitis, and, interestingly, 65% of these bronchiolitis cases were RSV-positive (54).

The changes that have been observed in the gut microbiota occurring after lung infection remain unclear. Studies in mice have suggested that the immune response to influenza virus infection in the lungs shapes gut microbiota composition, with both type I (53) and type II interferons (51) proposed to have a role. Our observation that different families within the *Bacteroidetes* phylum increase in abundance suggests that rather than the specific targeting of certain microbiota members by the immune system, respiratory infection causes a change in the gut environment, favoring the expansion of *Bacteroidetes*, and whichever *Bacteroidetes* family gains the advantage first increases in abundance. We speculate that one factor contributing to these changes is mucus. We observed increased levels of mucin Muc5ac in the colon, where it is not normally expressed (55, 56). This increase may be due to mice swallowing the excess mucus produced in the airways after infection or it may be that Muc5ac expression in colonic goblet cells is stimulated by the viral infection. The main role of mucus is defense against microbial exposure, but mucus is also utilized by these same microbes to gain an advantage in the extremely competitive microbiota environment (57). Bacteria, which can use mucins as an energy source, may have an ecological advantage if mucus composition changes. This has been seen elsewhere; changes in vaginal microbiota composition were associated with increased cervical Muc5B and Muc5ac levels (58) but the impact of respiratory mucus on the gut microbiota has not been previously studied.

An alternative explanation for the changes in microbiota composition observed after lung infection is infection-induced weight loss. Diet is the biggest contributor to microbiota composition (59), and reduced calorific intake in humans has been associated with a significant increase in *Bacteroidetes* abundance over *Firmicutes* (60, 61), similar to what was observed in this study after both RSV and influenza virus infection. In addition, calorie reduction in conjunction with influenza infection has been shown to enhance the gut microbiota changes observed after influenza infection alone (52). *Bacteroidetes* are considered very metabolically flexible (62), and mouse models of nutrient deprivation have been shown to switch their gene expression profile from enzymes capable of metabolizing dietary polysaccharides to enzymes which break down host mucus glycans (63). Therefore, it may be that the increase in *Bacteroidetes* observed after lung viral infection is due to a reduced food intake, and the increase in mucus observed in the gut may be compensatory for increased mucus metabolism by the gut bacteria.

One interesting feature of this study was the increase in S24\_7 abundance observed after lung infection. Despite being a very common constituent of the murine gut microbiota, only one member of this strictly anaerobic family has been isolated and cultured: *Muribaculum intestinale* (64). There is some controversy about whether S24\_7 is mouse-specific (65) or part of the gut microbiota in humans (66). Equally, the impact of S24\_7 on health is unclear; while some studies have associated increased S24\_7 levels with inflammation, there are conflicting theories on whether S24\_7 is the cause of (67) or the response to (40) the inflammation. Supporting our hypothesis that changes in gut microbiota after lung infection are driven by elevated mucus, the increased abundance of both S24\_7 and the mucin-degrading bacteria *Akkermansia muciniphilia* has been observed in a gut infection/inflammation model (40), suggesting that S24\_7 has a mucin-degrading ability. Likewise, the genetic analysis of S24\_7 has revealed that some members of S24\_7 have a tropism for host glycans (such as mucin) (66). In future studies, it will be interesting to determine the role of this family.

While the changes observed in the present study are robust and reproducible, the functional implications of the shift in microbiota after viral lung infection remain unclear. Changes in the gut microbiota have been associated with, and may in some circumstance amplify, disease. The decrease in *Lactobacillus* seen in the chronic stress model in mice (50) was associated with increased kynurenine, which is associated with depression. Changes in microbiota may also drive the gastrointestinal symptoms associated with influenza infection (68). We did observe an increase in lipocalin-2 in the feces which is associated with gut inflammation. These could be attributed to viral infection of the gastrointestinal tract, but in our study, and in others, no viral RNA was detected in the colonic tissue (51). If changes in microbiota were found in future studies to be associated with increased disease, therapeutic restoration of the preinfection balance may reduce disease. The majority of studies in humans and mice looking to improve the immune response and ameliorate disease in influenza virus infection have used various *Lactobacillus* spp. as probiotics with some success (18). Enriching the gut microbiota for *Lactobacillus* has also been shown to protect against airway inflammation in RSV infection (69). Overall, demonstrating that viral lung infection changes the gut microbiome is an important first step to investigating how these changes might have an impact on both respiratory and gut health, both in an infection setting and during vaccination.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the UK Home Office guidelines. The protocol was approved by the Imperial College London Animal Welfare and Ethics committee and followed the Animal Research: Reporting of *In Vivo* Experiments.

# AUTHOR CONTRIBUTIONS

HG performed experiments, analyzed data, and wrote the manuscript. LC and PJ analyzed data. MM designed studies and wrote the manuscript. MC designed studies, analyzed data, and wrote the manuscript. JT designed studies and wrote the manuscript.

# ACKNOWLEDGMENTS

HG is supported by an MRC-DTP award to the Imperial College. This work was supported by the European Community's European 7th Framework Program ADITEC (HEALTH-F4-2011-18 280873) and the Wellcome Trust. We thank Prof. S. Kroll, Dr. K. Sim, and Dr. A. Shaw for support with initial studies and

### REFERENCES


Prof. W. Cookson for input regarding the role of T cells. Lorraine Lawrence (NHLI, Imperial College London) performed the sectioning and staining for histology.


intestinal inflammation. *PLoS One* (2012) 7(9):e44328. doi:10.1371/journal. pone.0044328


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Groves, Cuthbertson, James, Moffatt, Cox and Tregoning. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Transcriptomics in Human Challenge Models

### *Amber J. Barton† , Jennifer Hill† , Andrew J. Pollard and Christoph J. Blohmke\**

*Oxford Vaccine Group, Department of Paediatrics, University of Oxford and the NIHR Oxford Biomedical Research Centre, Oxford, United Kingdom*

Human challenge models, in which volunteers are experimentally infected with a pathogen of interest, provide the opportunity to directly identify both natural and vaccine-induced correlates of protection. In this review, we highlight how the application of transcriptomics to human challenge studies allows for the identification of novel correlates and gives insight into the immunological pathways required to develop functional immunity. In malaria challenge trials for example, innate immune pathways appear to play a previously underappreciated role in conferring protective immunity. Transcriptomic analyses of samples obtained in human challenge studies can also deepen our understanding of the immune responses preceding symptom onset, allowing characterization of innate immunity and early gene signatures, which may influence disease outcome. Influenza challenge studies demonstrate that these gene signatures have diagnostic potential in the context of pandemics, in which presymptomatic diagnosis of at-risk individuals could allow early initiation of antiviral treatment and help limit transmission. Furthermore, gene expression analysis facilitates the identification of host factors contributing to disease susceptibility, such as *C4BPA* expression in enterotoxigenic *Escherichia coli* infection. Overall, these studies highlight the exceptional value of transcriptional data generated in human challenge trials and illustrate the broad impact molecular data analysis may have on global health through rational vaccine design and biomarker discovery.

Keywords: transcriptomics, vaccines, functional genomics, biomarkers, microarray, human challenge, expression, experimental infection

### INTRODUCTION

Human challenge studies, the deliberate infection of volunteers with a pathogen of interest, have been used to interrogate disease pathogenesis and vaccine efficacy since the smallpox challenge of James Phipps by Edward Jenner in 1796 (1). Human challenge studies have facilitated exploration of many aspects of infectious diseases, ranging from factors affecting susceptibility (2) and basic immune mechanisms to diagnostic biomarkers (3) and vaccine-mediated protection (4). Since some of the early challenge studies, such as those led by Theodore Woodward (5) and Myron Levine (6–8), the number of measurable parameters has extended far beyond vaccine efficacy, clinical symptoms, and antibody titers, now routinely including detailed measurements of molecular parameters such as cytokines, gene expression, and pathogen load (9, 10).

Gene expression studies have become a widely used tool to interrogate human host responses to infection and immunological perturbation. The evolution of gene expression technologies has allowed changes in the expression of thousands of genes in the peripheral blood and tissues to be

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Paola Massari, Tufts University School of Medicine, United States Giuseppe Lofano, Ragon Institute of MGH, MIT and Harvard, United States*

### *\*Correspondence:*

*Christoph J. Blohmke christoph.blohmke@ paediatrics.ox.ac.uk*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 12 October 2017 Accepted: 05 December 2017 Published: 18 December 2017*

### *Citation:*

*Barton AJ, Hill J, Pollard AJ and Blohmke CJ (2017) Transcriptomics in Human Challenge Models. Front. Immunol. 8:1839. doi: 10.3389/fimmu.2017.01839*

**141**

measured in response to vaccination or experimental infection. Starting in the 1990s with serial analyses of gene expression (11), two methods have come to dominate contemporary transcriptomics: microarrays and RNA sequencing (RNA-seq). Whereas microarrays measure the hybridization of fluorescently labeled transcripts (mRNA) to nucleotide probes on a bead chip, RNAseq involves sequencing of each transcript followed by alignment to a reference genome. Although cheaper and less labor intensive, microarrays are now largely being superseded by RNA-seq (12).

Because technologies to measure gene transcription are sensitive methods to broadly assess responses using small amounts of blood (≤3 ml), these approaches have inevitably been applied to the field of controlled human infection models. These models provide the unique opportunity to follow the host transcriptional response from the resting state (prechallenge baseline) through overt clinical disease to convalescence. Although most gene expression studies to date have aimed to deepen understanding at the level of basic mechanisms and pathways, it is hoped the understanding of immune responses will facilitate applications ranging from rational vaccine design to identification of selective and specific prognostic and diagnostic biomarkers of infection. In this review, we summarize the application of transcriptomics to human challenge models, providing examples of work exploring (i) host factors affecting susceptibility to disease, (ii) host–pathogen interactions, (iii) factors associated with symptom severity, (iv) modulation of immune responses by vaccination, and (v) early exposure signatures with diagnostic potential. Although human challenge studies are being deployed with increasing frequency, only a handful of studies exploring the transcriptomic response were identified (summarized in **Figure 1**; Table S1 in Supplementary Material). The paucity of such publications indicates that there may be considerable missed opportunity for exploitation of these unique clinical models, in order to understand the biology of disease and the interventions used to prevent or treat them (13).

Figure 1 | A summary of identified studies which have investigated changes in the transcriptome following controlled human infection. The area of each circle represents sample size, the color the inoculation route, and the line length indicates the number of samples taken for analysis.

### Search Strategy

References were identified through Scopus using search terms ("expression signature" OR transcriptional OR signatures OR profiling OR "host response" OR "rna-seq" OR "RNA sequencing" OR "gene expression" OR transcriptome OR microarray OR "gene signature" OR illumina OR host OR profile OR profiling) AND ("human challenge" OR volunteers OR "experimental infection"). The names of known human challenge agents (13) were also searched in functional genomics databases ArrayExpress and Gene Expression Omnibus, filtering for human gene expression data. Relevant papers cited by the database were included.

# HOST FACTORS AFFECTING SUSCEPTIBILITY

Challenge studies are well suited to analysis of host factors affecting susceptibility to infection, due to elimination of confounding factors such as variability between pathogen strains, or level and route of exposure to the pathogen. Although genome wide association studies (GWAS) are key in determining single nucleotide polymorphisms (SNPs) associated with susceptibility, the small number of subjects in challenge studies compared with studies of natural infection precludes identification of significantly associated genomic loci *via* this approach (14). However, the dramatic reduction in external variables afforded by human challenge studies, and the smaller number of genes assayed in microarrays compared with the millions of SNPs in GWAS, raises the possibility of identifying gene expression signatures present prior to experimental infection associated with susceptibility and protection. Such signatures may be influenced not only by genetic factors but also by the environment and epigenetic state, two factors with critical involvement in resistance and resilience. While the relationship between genetic variability, transcription, and disease susceptibility could be addressed by expression quantitative trait locus mapping (15), correlating genomic and transcriptomic data and applicable to sample sizes smaller than 100, to our knowledge this analysis has not yet been applied in the context of human challenge.

Studies examining the correlation between baseline characteristics (such as cell subset frequencies, antibody titer, or antibody functionality) with challenge outcome are numerous (16–19). In contrast, studies in which gene expression measured at baseline or shortly after challenge are correlated with clinical outcome are relatively few (2, 20). In a study by Yang et al., baseline gene expression was compared between equal numbers of asymptomatic individuals, and those who developed severe symptoms following oral challenge with enterotoxigenic *Escherichia coli* (ETEC) (2). Although there were only subtle differences in gene expression at this time point, several factors which plausibly might affect disease outcome were identified. For example, probes associated with presentation of antigen by major histocompatibility complex (MHC) molecules were enriched in those who did not develop infection, while baseline expression of complement inhibitor *C4BPA* was significantly higher in those who did. In another study, participants underwent multiple consecutive challenges with the Gram-negative bacterium *Haemophilus*  *ducreyi,* responsible for the sexually transmitted infection chancroid (20). Comparison of the transcriptome from lesions sampled 48 h after inoculation between study subjects who consistently resolved infection and those who formed infected pustules supported the hypothesis that the balance of dendritic cell (DC) phenotypes at the inoculation site is of critical importance. In particular, signatures suggested that the presence or absence of regulatory DCs is crucial in determining whether lesion formation will occur. *In vitro* infection of monocyte-derived DCs from participants combined with subsequent gene expression analysis mirrored these findings, supporting the baseline distribution of host DC phenotypes as a major susceptibility factor. Interestingly, both studies demonstrated that either components of the innate immune system involved in the earliest contact with the pathogen or aspects linking the innate and the adaptive immune system played an important role. Overall, the lack of studies examining whether the host transcriptome at time of challenge is predictive of disease outcome represents a major gap in the literature; however, the accumulation of data sets from challenge studies presents an opportunity to further interrogate baseline status in the context of disease outcome following infection.

## ELUCIDATING HOST–PATHOGEN INTERACTIONS

In those who do become infected, postchallenge transcriptomics can be used to identify pathways previously not known to be involved in disease pathogenesis. An overview of pathways highlighted as differentially regulated after challenge is shown in **Figure 2**. In addition to reflecting the human response to a pathogen, such pathways may provide insight into pathogendirected manipulation of the host response. A pathway seemingly ubiquitously upregulated after challenge with intracellular pathogens and parasites is interferon (IFN) signaling, with evidence of induction following challenge with rhinovirus (21), influenza A (22), *Plasmodium falciparum* (23), *Plasmodium vivax* (24), and *Salmonella enterica* serovar Typhi, particularly during periods of detectable bacteremia in typhoid fever (9). By providing a global snapshot of the molecular immune response, transcriptomic data are also well-suited to interrogating response patterns downstream of major immune molecules, as exemplified by studies exploring IFN-γ-induced pathways. By determining which genes had their expression highly correlated with IRF-1 and also contained IRF-1 binding sites in their promoter regions, 12 potential IRF-1 targets upregulated during *P. falciparum* infection were identified (23). Moreover, genes linking IFN-signaling with tryptophan metabolism were found to be upregulated during acute typhoid fever, raising the possibility that IFN-induced tryptophan metabolism plays a role in the host immune response, either by depriving *S.* Typhi of tryptophan or impacting the activation of regulatory T cells (9). Importantly, while transcriptomics provides a cross-sectional profile of the immune response at a given point in time, validation of the results in complementary model systems is pivotal. Thus, multiple validation experiments supported the findings from the typhoid challenge, with the same IFN-γ and tryptophan pathways induced in both a murine

model for *Salmonella* infection and *in vitro* infection of human macrophages with *S.* Typhi (9).

As well as confirming the important role of known cornerstones of immunity, analysis of the transcriptional response to pathogenic challenge has highlighted fundamental biological pathways more peripherally associated with immune responses, including apoptosis and the cell cycle. Changes in gene expression indicative of apoptosis take place following experimental infection with rhinovirus (21), *S.* Typhi (9), and influenza A (22), with greater expression of death receptor signaling in those experiencing more severe influenza (25). Furthermore, apoptosis is one of the few pathways overrepresented in natural *P. falciparum* infection compared with experimental infection, possibly as a result of greater parasite load in those naturally infected (23). Another fundamental process often observed in the host transcriptional response to infection is a signature representing the cell cycle, seen to be significantly upregulated following challenge of naïve participants with *S.* Typhi as well as following vaccination (4, 9, 26). While the origin of this signature is unclear, antigenspecific B cell and T cell proliferation for the generation of adaptive immunity is hypothesized to be a factor (4, 9).

Although upregulated pathways are generally those focused upon and pursued in follow-up studies, gene expression changes in both directions during acute infection. One of many drivers downregulating immune components is host manipulation by the pathogen to evade clearance (27). For example, in a small study involving bladder colonization by an asymptomatic strain of *E. coli*, the innate immune response was suppressed, with *in vitro* experiments suggesting that suppression was mediated by inhibition of polymerase II phosphorylation (28). Downregulation of genes in peripheral blood may also represent migration of immune cells to lymphoid tissue and thus changes in the cell composition rather than true expression changes, an important consideration when measuring gene expression in whole blood. Indeed at typhoid diagnosis, transcriptional modules associated with B, T, and natural killer cells are underrepresented (9), possibly reflecting the leukopenia observed in typhoid fever (29) or migration of cells to the intestinal mucosa (16, 30, 31). Moreover, inflammatory genes are downregulated after challenge with *P. vivax* (24), although no change was observed in natural infection with *P. falciparum*. With neutrophil counts accounting for 34% of variance in the inflammation-associated axis (32), this discrepancy might be explained by the neutropenia observed in *P. vivax* but not *P. falciparum* infection (33). Finally, downregulation may represent tolerance by the host to limit immunemediated damage (22, 34).

Beyond advancing basic scientific understanding, insight into immune pathways elicited by pathogens can facilitate identification of novel therapeutic targets and guide vaccine development toward stimulation of protective immune pathways. The transcriptional changes following ETEC challenge were compared with those induced by a database of small molecule drugs, with a hypothesis that those which correlate closely could be used to augment or modulate antibacterial inflammation (2). This work identified a similar collection of transcriptomic changes for the antibiotic rifampicin, suggesting that the drug might encourage an appropriate immune response in addition to a direct antibacterial effect (2). Thus, by acting as a convenient means to gain an overview of pathogenesis-related pathways, transcriptomics has the potential to identify those amenable to manipulation by drugs. Furthermore, interventions acting on the host response rather than microbial pathways are at a lower risk of evasion through the development of antimicrobial resistance, a factor that will likely be of huge importance in the postantibiotic era.

### FACTORS ASSOCIATED WITH SYMPTOM SEVERITY

The controlled nature, extensive sampling, and close monitoring of participants in human challenge studies provide an invaluable opportunity to correlate changes in gene expression with clinical parameters such as C-reactive protein, temperature, and symptom scores. Expression patterns associated with symptoms are consistently related to innate immunity and inflammation and thus may play an important part in the host response. For example, after rhinovirus challenge expression of the gene *viperin* correlated with rhinorrhea, sneezing, and chills (21). In follow-up experiments on a primary culture of human bronchial epithelium, *viperin* siRNA knockdown increased viral replication, suggesting a protective role. In a recent influenza study aiming to use transcriptomics to predict symptom scores, 16 of the 19 probes identified as predictive were within the antiviral response and innate immunity gene ontogenies (35). Similarly, comparing the transcriptomes of those with mild and more severe laboratory-confirmed influenza demonstrated that signaling pathways involving IFN, pattern recognition receptors, and IRF were significantly enriched in those with moderate/ severe disease (25). Moreover, expression of six genes related by IFN signaling distinguished between the two groups with 100% accuracy. During acute typhoid fever, transcriptional modules relating to innate immunity and inflammation were positively correlated with symptom severity and negatively correlated with time to diagnosis (9), suggesting that pathways associated with symptoms are not always protective.

While genes associated with increased symptom scores following malaria challenge have not been directly identified, two studies have compared transcriptomes between groups with differing symptom profiles: experimentally versus naturally infected (23) and naïve versus pre-exposed (24). After challenge with *P. falciparum*, transcriptomes from naïve volunteers, 73% of whom were afebrile at diagnosis due to early detection, were compared with naturally infected febrile participants (23). *CASP-1* and *IL-1*β were upregulated, and IL-1β suppressor *pyrin* downregulated in naturally infected febrile volunteers compared with challenge study participants, suggesting involvement of this inflammatory pathway in fever. In a *P. vivax* challenge study, the transcriptomes of naïve and pre-exposed volunteers were compared to investigate the greater symptom severity in naïve infection, despite equal parasite load and time to diagnosis (24). Many of the transcriptional changes in genes and pathways were more pronounced in the naïve group at diagnosis, suggesting that the pre-exposed group may suffer fewer symptoms due to greater tolerance for the parasite (24).

Whereas many studies focus on gene signatures relating to symptoms, the transcriptional response driving active suppression of symptoms during infection is less well characterized. In an influenza challenge study where gene expression profiles were followed longitudinally in both symptomatic and asymptomatic participants, it was found that asymptomatic participants were indeed successfully infected with influenza, as evidenced by viral shedding, seroconversion, and active perturbation of the transcriptome (22). While upregulated in those with symptoms, genes associated with the inflammasome were downregulated over time in those without symptoms, as were IFN signaling inhibitors *SOCS1* and *SOCS3*. Ribosomal genes associated with lymphocytes, genes which reduce oxidative stress, and Th1 response inducing *SOCS5* were all upregulated. Therefore, the asymptomatic state does not simply represent absence of a pathogen or a response; rather, the subjects appear to mount a regulated immune response, which contains the infection without causing illness. Hypothetically, augmenting these pathways could be used therapeutically to ameliorate symptoms.

Overall, it is not clear which pathways are associated with symptoms as well as protection, which pathways cause symptoms but are non-protective, and which are correlated with, but not causative of, symptomatic disease. Thus, in order to identify targets for the treatment of symptoms, follow-up experiments employing animal and *in vitro* models will be necessary to establish cause-and-effect relationships. For example, in a candidiasis human challenge model polymorphonuclear cell (PMN) infiltration was associated with symptomatic infection (36). A follow-up RNA-seq experiment on mice challenged with *Candida albicans* showed upregulation of phagocyte infiltration and migration pathways, validating observations from the human challenge model (37). In addition, expression of the NLRP3 inflammasome was increased. Although inflammasome inhibition by diabetes drug glibenclamide did not affect colonization in the mice, PMN recruitment was reduced, suggesting that symptoms could feasibly be pharmacologically attenuated. Thus, the use of transcriptomics to identify symptom-associated pathways after human challenge has huge potential to inform therapeutic development.

## MODULATION OF THE IMMUNE RESPONSE BY VACCINATION

Transcriptional responses following human vaccination have been explored extensively, with major studies conducted on vaccines for yellow fever, influenza, and shingles (38–41). Postvaccination challenge studies present the opportunity to rapidly and efficiently test vaccine-mediated protection, as in contrast to Phase III vaccine trials all participants are exposed to the pathogen during the study period. The identification of transcriptomic signatures associated with immunological responses to vaccination (correlates of immunogenicity) and the development of protection (correlates of protection) may facilitate development of vaccines which achieve higher levels of protection against a greater variety of pathogens. Moreover, such signatures could possibly be established as non-serological biomarkers predictive of vaccine-conferred protection (42).

The increasing intricacy of studies combining vaccination and challenge with transcriptomics is demonstrated in work exploring candidate malaria vaccines, which is summarized in **Figure 3A**. A study comparing two formulations of the RTS,S vaccine identified genes in the immunoproteasome pathway as associated with protection following immunization and were hypothesized to influence immunity through processing of antigenic peptides for presentation on MHC molecules (43). Five years later, a transcriptomic analysis combining two small cohorts receiving different prime-boost vaccine combinations was described, contrasting an RTS,S and a DNA vaccine prime, followed by a boost with two modified vaccinia virus Ankaras expressing different malaria antigens (44). Compared with those who did develop malaria after challenge, the transcriptomes of restimulated PBMCs from the three protected participants showed enrichment of modules associated with IFN induction and antigen presentation. This work illustrates the value of assessing different vaccines together to identify common protective pathways and the advantage of partial vaccine protection for assisting in the identification of protective responses.

Transcriptomics has also been used to investigate factors driving the differences in correlates of protection between two regimens: one where participants received three doses of RTS,S (RRR) and one where participants received the Ad35. CS.01 vaccine followed by two doses of RTS,S (ARR) (4). Although equally protective, antibody titers correlated with protection in the former group and CD4<sup>+</sup> T cell responses in the latter. Despite largely similar responses at the level of BTMs, with common features including IFN responses at day 1 after vaccination, the BTMs correlating with protection differed between the two regimens. While modules relating to cell cycle, plasma, and B cells after each vaccination correlated with protection in the RRR group, modules relating to DCs and antigen presentation after prime correlated with protection in the ARR group. However, in both groups and at multiple time points, modules related to natural killer cells were negatively associated with protection. Validated on the 2010 RTS,S study data set, the RRR protective signature was capable of predicting whether a participant was protected from malaria with 80% accuracy (4).

More recently, postvaccination transcriptomics has been applied to a typhoid vaccination study, where two oral, liveattenuated vaccines (Ty21a and M01ZH09) were given in conjunction with subsequent pathogenic challenge (45). While NK cell signatures were enriched by Ty21a, the vaccine with the higher protective efficacy, cell cycle modules were strongly induced following vaccination with M01ZH09, inducing higher levels of anti-H and anti-LPS antibodies (46). These data further highlight the association of transcriptional cell cycle markers with humoral responses, as was observed during acute infection (9). Furthermore, whereas gene signatures associated with amino acid and transmembrane transport were correlated with time to diagnosis in Ty21a vaccine recipients, in M01ZH09-vaccinated participants CD28 costimulation was associated with protection (**Figure 3B**).

Beyond a binary readout of vaccine success, challenge offers the potential to explore vaccine-mediated alterations in the host response, even when complete protection is not achieved. Although identification of differences using this approach is challenging and depends on the vaccine in question (25), both malaria and BCG challenge studies have shown greater changes in global expression patterns in those vaccinated prior to challenge compared with unvaccinated controls (43, 47). These differences can be striking in their magnitude; an over 10-fold difference in the number of differentially regulated genes between vaccinated and naïve participants was observed 5 days postmalaria challenge. Vaccination thus appears to amplify the magnitude of the host response to infection, negatively correlating with mycobacterial growth in the BCG challenge model but positively correlating with scarring at the site of inoculation (47). To conclude, by elucidating key pathways involved in conferring protection, post-challenge transcriptomic studies have the potential to direct rational vaccine development: either by design of antigens which stimulate certain pathways or through the use of adjuvants which enhance protective responses.

## IDENTIFICATION OF DIAGNOSTIC BIOMARKERS

Most reliable diagnostic tests for infectious diseases rely on direct detection of a pathogen by culture or PCR. Both methods are labor-intensive and time consuming. Diagnosis on the basis of an antigen-specific host response, however, has been possible for many years, with examples including the tuberculin skin test for tuberculosis (48), the Widal agglutination test for typhoid (49), and ELISA-based antibody assays for dengue (50). Although easy, such tests can have low sensitivity and specificity, are limited in their ability to distinguish between acute infection and previous exposure, and can be confounded by vaccination or immunodeficiency (51).

represent a positive correlation with time to diagnosis and pink squares a negative correlation.

Recently, transcriptomic analysis has been identified as a useful tool to identify expression signatures specific for a particular condition of interest (52–57). For example, analysis of a gene expression data set derived from a case–control study identified a 27-gene signature discriminatory between acute and latent tuberculosis, regardless of HIV status (57). However, where there is no gold standard for comparison, in the field it is impossible to ascertain the sensitivity and specificity of novel diagnostic tests. The clear and consistent case definitions within human challenge models, and the certainty over with which pathogen a participant is infected, provide a unique opportunity to develop biomarkers and accurately validate diagnostics.

Work by researchers at Duke University on respiratory viral infection has demonstrated the potential use of transcriptomic signatures from challenge models to distinguish between infected and healthy individuals (3). An analysis of microarray data from individuals infected with human rhinovirus, respiratory syncytial virus (RSV), and influenza A identified a group of antiviral response genes which distinguish between symptomatic and asymptomatic participants (**Figure 4A**). This

gene signature predicted whether participants in an independent influenza data set were symptomatic with 100% accuracy. Despite the identification of several genes unique to RSV and human rhinovirus, the signatures for each were similar. More recently, however, transcriptomic analyses of naturally infected patients have identified signatures that can be used to distinguish RSV and influenza from other respiratory infections (54, 56). While challenge studies provide an environment with clear cases useful to identify gene expression signatures indicative of the infection of interest, a limitation is that often these analyses are performed by comparing cases with controls or asymptomatic with symptomatic participants. The question arises whether such signatures can successfully identify a specific infection in a setting where a differential diagnosis with other competing infections has to be reached, in which case the selection of endemic control samples is pivotal for biomarker discovery.

While of limited application in real-world settings, with the exceptions of limiting spread of pandemic infections (58) and detecting individuals exposed to biological warfare agents (60), a series of studies have explored the potential of transcriptomics for presymptomatic diagnosis. Following influenza challenge, factor scores in those developing infection deviated significantly from asymptomatic participants prior to the peak of symptoms (3), with the gene signatures for those developing infection starting to diverge from uninfected participants at 35–40% of the elapsed time between infection and peak symptoms, a point at which symptoms were still very mild [Ref. (58); **Figure 4B**. The use of these early diagnostic signatures could inform therapeutic interventions which might ultimately influence the outcome of disease. Thus, H3N2 influenza A challenge participants were randomized to either receive early antiviral treatment 36 h postchallenge based on time of divergence in the previous study (58), or "standard" antiviral treatment 120 h after challenge [Ref. (59); **Figure 4C**. Viral shedding declined quicker in the early treatment group, while symptoms peaked earlier and declined quicker.

Although the main application of transcripts as biomarkers is for diagnosis, a recent challenge study instead aimed to identify a panel of genes which could be used as a method of assessing symptom severity [Ref. (35); **Figure 4D**]. Distinct from studies focusing on the biological basis of symptoms, discussed earlier in this review, the main aim was to identify biomarkers capable of predicting symptom score. A 19-gene signature was found to be predictive of actual symptom scores, as self-reported by the participants. Such a group of genes could be used to objectively assess the degree to which novel vaccines and therapeutics attenuate symptoms following challenge. Overall, gene signatures have demonstrated great potential as biomarkers for infection and symptom severity, with potential to be used clinically as a convenient and accurate diagnostic tool.

## STRENGTHS AND CURRENT LIMITATIONS OF THE CHALLENGE MODEL TRANSCRIPTOMICS APPROACH

Controlled human infection models have a number of marked advantages over classical field studies of natural infection, many of which have been alluded to in earlier sections of this review. Here, we examine the strengths and limitations of the application of transcriptomics to challenge studies, which are outlined in **Table 1**. For a more general discussion of challenge studies readers are directed to a review by Darton et al. (13).

The limitation of small sample sizes in challenge studies resulting from practical and cost restraints, in some cases as low as three individuals (28), is pertinent to many types of analysis. Although expression profiling is highly sensitive and capable of detecting small changes between samples from an individual, the power to identify whether such changes are of statistical significance across a group relies on a sufficient sample size. Small numbers are particularly problematic for studies which investigate the link between baseline gene expression and challenge outcome, where high levels of interindividual variation could mask genuine factors affecting susceptibility. Connected with this is the need

Table 1 | Strengths and limitations of transcriptomics in human challenge studies.


for appropriate inclusion of non-challenged control subjects, to allow changes unrelated to infection to be accounted for, such as circadian rhythms, variations in sample processing, and random fluctuations in gene expression.

Another major limitation of many of the studies reviewed here is the type of sample available for RNA extraction and profiling. With several exceptions (20, 21), sites of infection such as the intestinal mucosa in enteric infections and respiratory tract in influenza challenges are frequently inaccessible. Furthermore, in the majority of studies RNA is extracted from samples containing a mixed population of cell types (e.g., whole blood). Although pathway and gene set enrichment analyses can identify pathways generally associated with certain cell types, for many pathways it is impossible to tell which cell population is accountable.

An issue common to many transcriptomic studies, and encountered often in their application to human challenge models, is the paucity of validation work and experimental follow-up performed on the genes and pathways highlighted by this approach, either in replicated human experiments, *in vitro* or *in vivo*. Given that challenge participants might not be representative of naturally infected populations, and sample sizes are small, validation on data sets from field studies are critical. For example, although the greatest burden of RSV is in babies, RSV challenge can only be carried out using consenting adults; cohort studies, however, have found that the transcriptomic response to RSV infection changes markedly with age (54). Furthermore, although transcriptomic analyses can identify enriched pathways in response to infection, it is not possible to dissect cause-and-effect relationships from transcriptomic data alone. Identifying which changes fall upstream of others, which are directly related to parameters of interest, and which are merely bystander effects is highly challenging. Therefore, followup experiments employing *in vitro* or animal models are extremely valuable in supporting hypotheses, as exemplified by the *in vitro* experiments carried out in the *H. ducreyi* (20), urinary tract *E. coli* (28), and *S.* Typhi (9) challenge studies.

### CHANGING TECHNOLOGY AND FUTURE DIRECTIONS

While microarrays dominated as the transcriptomic technology of choice in the 2000s as a result of their ease and high throughput, the decreasing cost and increasing sensitivity of next generation sequencing has led to a gradual switch to the use of RNA-seq since 2008 (12, 61). However, only one of the studies reviewed here profiled gene expression by RNA-seq (24), and many of the unique advantages of RNA-seq have not yet been exploited by challenge studies. For example, while microarrays are limited to detecting transcripts represented by the probes on the chip, RNA-seq is capable of detecting non-coding RNAs thought to play an significant role in the antiviral and antibacterial immune responses (62, 63) and thus may act as novel biomarkers or correlates of protection. RNA-seq can also distinguish between different splice variants, of known importance in immunity: for example, in modulating toll-like receptor signaling (64). More recently, differential splicing in Mycobacterium-infected macrophages has been found to produce truncated transcripts that ultimately disrupt macrophage function (65). Thus, in the context of human challenge models, analysis of alternative splicing could give insight into important host–pathogen interactions. Finally, advances in single cell sequencing could be used to characterize transcriptomic changes in different immune cell populations and investigate the heterogeneity within each population. For example, following *ex vivo* infection of macrophages with *Salmonella,* single cell RNA-seq has been used to show that those cells containing growing bacteria possessed M2 expression markers, whereas those containing non-growing bacteria were more proinflammatory (66). Thus, the problems associated with averaging expression in the whole blood samples taken in human challenge studies could be eliminated (67).

Although system approaches to immunology have mainly focused on nucleic acid profiling, decreasing costs of mass spectrometry-based technologies such as proteomics (68), metabolomics (69), and lipidomics (70) have resulted in their increased use to characterize infection-induced changes downstream of gene expression. Challenge studies which integrate transcriptomic data with other high parameter data sets will begin to build a more detailed image of the biology of infection and generate a better understanding of how the regulation of gene expression affects functional molecules of the immune system.

Furthermore, the requirement for published transcriptomic data to be deposited in publically available databases has increased opportunities for researchers to compare transcriptomic responses toward a range of pathogens, in various species, and between challenge and natural infection. While one obvious application lies in the identification of specific diagnostic signatures, such approaches stand to deepen our understanding of common and disease-specific elements of immune response.

Although the focus of this review is the human transcriptome, monitoring changes in the transcriptome of the challenge agent itself could substantially benefit our understanding of host–pathogen interactions. For example, using *in vitro* models and clinical samples, changes in *Plasmodium* gene expression during its lifecycle have been characterized by RNA-seq and microarrays (71–73). Selective capture of transcribed sequences (SCOTs), in which bacterial cDNAs are hybridized to biotinylated genomic DNA and captured by streptavidin beads, has emerged as a particularly useful technique for bacterial transcriptomic profiling, as it allows detection of gene expression *in vivo* even at low bacterial titers (74). In the context of human challenge models, this method has successfully been used to identify *H. ducreyi* genes expressed in the pustules of experimentally infected

### REFERENCES


volunteers (75). SCOTs has also facilitated detection of *S.* Typhi and *S.* Paratyphi A transcripts in clinical samples from enteric fever patients (76, 77) and could thus be applied to monitor temporal changes in gene expression in the typhoid human challenge model.

### CONCLUDING REMARKS

The application of transcriptomics to human challenge models has given major insight into the postinfection host response, allowing diagnostic biomarkers, correlates of vaccine-induced protection, and indicators of severity to be identified. With increasing availability of data in the public domain and decreasing costs for high-throughput technologies, it is now the time to further investigate host factors associated with susceptibility, exploit the full capabilities of RNA-seq, and integrate transcriptomic data with other "omic" data sets to fully utilize the unique advantages offered by human challenge. It is our ethical obligation to gain as much valuable information as possible from study samples, and transcriptomic profiling is indisputably a key approach which we argue should be considered for application in all future challenge studies. Finally, we would like to emphasize the importance of validation in the field, ensuring that genes or pathways identified as significant are clinically relevant.

### AUTHOR CONTRIBUTIONS

AB carried out the literature search and created the figures. AB and JH drafted the manuscript. CB, JH, and AP reviewed and made revisions to the manuscript and figures.

### FUNDING

The authors would like to thank ADITEC for sponsorship of this research topic and acknowledge the support from the National Institute for Health Research Oxford BRC and the Wellcome Trust (Strategic Award no. 106158/Z/14/Z). JH gratefully acknowledges support from the George and Susan Brownlee Fellowship at Linacre College, and AB the funding provided by the St Cross Paediatrics Scholarship.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01839/ full#supplementary-material.


CVD 103 and CVD 103-HgR. *Lancet* (1988) 332(8609):467–70. doi:10.1016/ S0140-6736(88)90120-1


biology study of five human vaccines. *Nat Immunol* (2014) 15(2):195–204. doi:10.1038/ni.2789


placebo and Ty21a. *PLoS Negl Trop Dis* (2016) 10(8):e0004926. doi:10.1371/ journal.pntd.0004926


splicing complex. *PLoS Genet* (2015) 11(2):e1004932. doi:10.1371/journal. pgen.1004932


**Conflict of Interest Statement:** AP has previously conducted clinical trials of vaccines on behalf of Oxford University funded by vaccine manufacturers, but did not receive any personal payments from them. His department received unrestricted educational grants from Pfizer/GSK/Astra Zeneca in July 2016 and from Gilead/MSD/GSK/Astra Zeneca in June 2017 for a course on Infection and Immunity in Children. AP is chair of the UK Department of Health's (DH) Joint Committee on Vaccination and Immunisation (JCVI), and the scientific advisory group on vaccines of the European Medicines Agency, and a member of WHO's Strategic Advisory Group of Experts, but the views expressed in this manuscript do not necessarily represent the views of JCVI, DH, EMA, or WHO.

*Copyright © 2017 Barton, Hill, Pollard and Blohmke. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# A Phase 2a Randomized Study to Evaluate the Safety and Immunogenicity of the 1790GAHB Generalized Modules for Membrane Antigen Vaccine against *Shigella sonnei* Administered Intramuscularly to Adults from a Shigellosis-Endemic Country

### *Edited by:*

*David J. M. Lewis, Imperial College London, United Kingdom*

### *Reviewed by:*

*Anita S. Iyer, Harvard Medical School, United States William D. Picking, University of Kansas, United States*

### *\*Correspondence:*

*Audino Podda audino.p.podda@gsk.com*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 11 October 2017 Accepted: 11 December 2017 Published: 22 December 2017*

### *Citation:*

*Obiero CW, Ndiaye AGW, Sciré AS, Kaunyangi BM, Marchetti E, Gone AM, Schütte LD, Riccucci D, Auerbach J, Saul A, Martin LB, Bejon P, Njuguna P and Podda A (2017) A Phase 2a Randomized Study to Evaluate the Safety and Immunogenicity of the 1790GAHB Generalized Modules for Membrane Antigen Vaccine against Shigella sonnei Administered Intramuscularly to Adults from a Shigellosis-Endemic Country. Front. Immunol. 8:1884. doi: 10.3389/fimmu.2017.01884*

*Christina W. Obiero1 , Augustin G. W. Ndiaye2 , Antonella Silvia Sciré2 , Bonface M. Kaunyangi1 , Elisa Marchetti2 , Ann M. Gone1 , Lena Dorothee Schütte3 , Daniele Riccucci2 , Joachim Auerbach2 , Allan Saul2 , Laura B. Martin2 , Philip Bejon4 , Patricia Njuguna1,5 and Audino Podda2 \**

*1KEMRI-Wellcome Trust Research Programme, Clinical Research Department, Kilifi, Kenya, 2GSK Vaccines Institute for Global Health, Siena, Italy, 3GSK Vaccines Clinical Laboratory Sciences, Marburg, Germany, 4Clinical Tropical Medicine, Nuffield Department of Medicine, University of Oxford, Headington, United Kingdom, 5Department of Public Health, Pwani University, Kilifi, Kenya*

Shigellosis is a mild-to-severe diarrheal infection, caused by the genus *Shigella*, and is responsible for significant morbidity and mortality worldwide. We evaluated the safety and immunogenicity of an investigational *Shigella sonnei* vaccine (1790GAHB) based on generalized modules for membrane antigens (GMMA) in Kenya, a *Shigella*-endemic country. This phase 2a, observer-blind, controlled randomized study (NCT02676895) enrolled 74 healthy adults aged 18–45 years, of whom 72 were vaccinated. Participants received, in a 1:1:1 ratio, two vaccinations with the 1790GAHB vaccine at doses of either 1.5/25 μg of O antigen (OAg)/protein (group 1.5/25 μg) or 5.9/100 μg (group 5.9/100 μg) at day (D) 1 and D29, or vaccination with a quadrivalent meningococcal vaccine at D1 and tetanus, diphtheria, and acellular pertussis vaccine at D29 (control group). Solicited and unsolicited adverse events (AEs), serious AEs (SAEs), and AEs of special interest (neutropenia and reactive arthritis) were collected. Anti-*S. sonnei* lipopolysaccharide (LPS) serum immunoglobulin G (IgG) geometric mean concentrations (GMC) were evaluated at D1, D29, and D57 and compared to anti-*S. sonnei* LPS antibody levels in convalescent patients naturally exposed to *S. sonnei*. The percentages of participants with seroresponse were also calculated. The most frequently reported solicited local and systemic AEs across all groups were pain and headache, respectively. Only one case of severe systemic reaction was reported (severe headache after first vaccination in group 5.9/100 μg). Seven and three episodes of neutropenia, assessed as probably or possibly related to vaccination respectively, were reported in the investigational and control groups, respectively. No other SAEs were reported. Despite very high baseline anti-*S. sonnei* LPS serum IgG levels, the 1790GAHB vaccine induced robust antibody responses. At D29, GMC increased 2.10- and 4.43-fold from baseline in groups 1.5/25 and 5.9/100 μg, respectively, whereas no increase was observed in the control group. Antibody titers at D57 were not statistically different from those at D29. Seroresponse was 68% at D29 and 90% at D57 in group 1.5/25 μg, and 96% after each vaccination in group 5.9/100 μg. The 1790GAHB vaccine was well tolerated and highly immunogenic in a population of African adults, regardless of the GMMA OAg/protein content used.

Keywords: *Shigella sonnei*, 1790GAHB vaccine, generalized modules for membrane antigen, safety, immunogenicity, *Shigella*-endemic settings

## INTRODUCTION

Diarrheal diseases are a leading cause of morbidity and mortality among all age groups, and particularly among young children (1). With 164,000 deaths in 2015, *Shigella* is one of the major causes of overall diarrheal mortality (1), second only to rotavirus (2). Although a decline in mortality due to diarrheal diseases has been observed in the last decade in children less than 5 years of age, yearly deaths still ranged between 499,000 (1) and 525,000 (3) in 2015. Most of these fatalities occurred in Sub-Saharan Africa and Asia, and *Shigella* accounted for approximately 11% of them (1). Additionally, in a recent study conducted in these continents, *Shigella* was identified as a significant cause of moderate-to-severe diarrhea in children (4), and its relevance was reinforced when analyses were repeated using molecular diagnostic tests (5).

Among the four species of the genus *Shigella*, the 15 serotypes of *Shigella flexneri* are mostly isolated in developing countries, while the single serotype of *Shigella sonnei* was traditionally encountered in high-income settings. However, this serotype has emerged lately as one of the dominant species also in many regions of Asia, Latin America, and the Middle East (4, 6–8).

An additional element of concern about shigellosis is the decreased susceptibility to a large range of antibiotics observed over the last decades, with most of the *Shigella* serotypes becoming multi-drug resistant (9). This reinforces the need for a widely available vaccine against shigellosis.

Several candidate vaccines, developed using different technologies, are currently under investigation (10). Inactivated vaccine candidates based on O Antigen (OAg), which is recognized as a key target antigen for *Shigella*, including conjugates, bioconjugates, and live-attenuated vaccine strains have already been tested in clinical trials (10–14). Recently, generalized modules for membrane antigens (GMMA) have been proposed as a delivery system for *S. sonnei* OAg (15). GMMA are optimally sized for immune stimulation and have self-adjuvanting activity, delivering innate signals through toll-like receptor ligands and other pathogen-associated molecular patterns. Although alum is not needed as an adjuvant, the vaccine has been formulated with Alhydrogel, which was shown to reduce the pyrogenicity in rabbits (15).

This *S. sonnei* GMMA vaccine has been shown to be highly immunogenic and well tolerated in phase 1 clinical trials, when administered by intramuscular route to healthy European adults (16).

The current study aimed to further evaluate the safety and immunogenicity profile of the 1790GAHB vaccine in healthy adults from Coastal Kenya, a *Shigella*-endemic setting, and assessed two different GMMA OAg/protein doses. As no serologic correlates of protection are established for *S. sonnei* and as the presence of anti-*Shigella* lipopolysaccharide (LPS) antibodies was previously associated with acquired immunity to the pathogen (17), vaccine-induced immunogenicity was compared to anti-*S. sonnei* LPS antibody levels in a naturally infected, convalescent population.

A summary contextualizing the results and potential clinical research relevance and impact is displayed in the Focus on Patient Section (**Figure 1**) for the benefit of Health Care Professionals.

# MATERIALS AND METHODS

### Study Design and Participants

This phase 2a, observer-blind, randomized, single-center, controlled study was conducted at the KEMRI-Wellcome Trust in Kilifi, Kenya, between August 2016 and March 2017. The study enrolled healthy adults aged 18–45 years, fulfilling protocol inclusion and exclusion criteria, willing to comply with study procedures and signing, or thumb printing, the informed consent form for study participation. Females of child bearing potential were enrolled only if they agreed to use an effective birth-control method prior to and during the study. However, any potential pregnancies during the trial were to be reported and, if possible, their outcome was to be monitored. Any condition potentially interfering with the ability to participate in the study or with the study results, or causing additional risk by participation in the trial was an exclusion criterion. Individuals were also excluded if they had any of the following conditions: progressive or severe neurological disorders, seizures, previous Guillain–Barré syndrome, history of reactive arthritis, hepatitis B infection, HIV or HIV-related disease, autoimmune disorders, known or

Figure 1 | Focus on patient section.

suspected impairment/alteration of the immune system, known bleeding diathesis (or any condition that may be associated with prolonged bleeding time), serious chronic or progressive disease, malignancy or lympho-proliferative disorder, history of allergy to vaccine components or of substance or alcohol abuse within the past 2 years, participation in clinical trials with other investigational product within 28 days prior to screening, receipt of vaccines containing meningococcus A, C, W, Y, tetanus, diphtheria or pertussis antigens within 12 months before screening, receipt of any other vaccine within 4 weeks prior to screening or plan to do so during the study, receipt of blood/plasma products within 12 weeks prior to first study vaccination, body mass index >30 kg/m2 , laboratory confirmed case of disease by *S. sonnei*, and breast-feeding.

Individuals with an absolute neutrophil count (ANC) <1.8 × 109 /L for the initial 18 participants (first study cohort) or <1.0 × 109 /L, if recommended by an independent Data Safety Monitoring Board (DSMB) for the remaining study population (second cohort) at screening or with previous history of benign ethnic neutropenia/drug-related neutropenia, prior use or likelihood to use neutropenic drugs were not enrolled. Inclusion criteria were re-assessed for all participants, prior to each study vaccination.

Enrolled individuals were randomized in a 1:1:1 ratio, to receive two vaccinations with either the investigational or the control vaccines at day (D) 1 and D29. Investigational groups received the *S. sonnei* 1790GAHB vaccine, used at two different OAg/protein doses, while participants in the control group were randomized to receive a quadrivalent meningococcal conjugate vaccine (MenACWY; *Menveo*, GSK) at D1 and a vaccine against tetanus, diphtheria, and acellular pertussis (Tdap; *Boostrix*, GSK) at D29 (**Figure 2**).

Randomization was performed by a validated internet-based system. The study was observer-blind, due to the different presentations of the investigational and control vaccines. Designated unblinded trained and qualified staff prepared or administered the study vaccines, but was not involved in the evaluation of the participants for safety or in the collection of study data.

The study was monitored by the DSMB. Eighteen participants were initially enrolled in the first study cohort, received first

vaccination and were followed up, according to study procedures, for 7 days. A summary of all safety data and listings including hematology, blood chemistry and urine dipstick/urinalysis test values were provided to the DSMB. After reviewing all safety data, the DSMB recommended that enrollment could be completed using the published African consensus ANC threshold of 1.0 × 109 /L (18) as inclusion criterion and the severity grading system proposed by the Division of Acquired Immunodeficiency Syndrome (19), which takes into consideration also the ethnic differences in ANC, be used for classification of postvaccination neutropenia. The DSMB was also consulted for any potential safety issue reported during the trial.

The informed consent form and the study protocol were reviewed and approved by the KEMRI Scientific and Ethics Committee, the Kenyan Pharmacy and Poisons Board, and the Oxford Tropical Research Ethics Committee prior to study start. The trial was designed and conducted in agreement with the ICH Harmonized Tripartite Guidelines for Good Clinical Practice, applicable local regulations and the Declaration of Helsinki and was registered at ClinicalTrials.gov (NCT02676895).

### Study Objectives

The primary objective was to evaluate the safety profile of two vaccinations in healthy adults with two different dose levels of 1790GAHB in a *Shigella*-endemic country. The secondary objective was to assess the immunogenicity of the investigational vaccine, as measured by anti-*S. sonnei* LPS serum immunoglobulin G (IgG) levels, at 28 days after each vaccination.

### Study Vaccines

The 1790GAHB vaccine consisted of *S. sonnei* 1790-GMMA (approximately 11.8 µg OAg/200 μg total protein per milliliter) adsorbed to Alhydrogel (0.7 mg Al3<sup>+</sup>/mL) in tris-buffered saline, was available as a liquid formulation in single-dose vials with 0.7 mL of injectable suspension and did not contain any preservative. A 0.5-mL dose containing 1.5/25 μg of OAg/protein was obtained by bed-side mixing, by dilution with Alhydrogel in tris-buffered saline (0.7 mg Al3<sup>+</sup>/mL).

Each 0.5 mL dose of MenACWY contained 10 µg of serogroup A oligosaccharide and 5 µg of each of serogroups C, W, and Y oligosaccharides conjugated to 32.7–64.1 µg of CRM197.

Each 0.5 mL dose of Tdap contained 2.5 Lf diphtheria toxoid, 5 Lf tetanus toxoid, 8 µg pertussis toxoid, 8 µg filamentous hemagglutinin, 2.5 µg pertactin, and 0.5 mg aluminum hydroxide.

At each vaccination, a 0.5-mL vaccine dose was administered intramuscularly into the deltoid area of the non-dominant arm.

### Safety Assessments

Participants were observed at 30 and 60 min after each vaccination for any adverse event (AE). Local (injection site pain, erythema, and induration) and systemic (arthralgia, chills, fatigue, headache, malaise, myalgia, and fever) solicited AEs were recorded on diary cards by study personnel performing daily home visits for 7 days following each vaccination. Unsolicited AEs occurring within 28 days following vaccination were reported by the participants and documented by the investigator, during follow-up clinic visits carried out at 7 and 28 days after each vaccination or in the course of unscheduled visits. Solicited AEs continuing beyond the 7-day period following each vaccination were reported as unsolicited AEs.

All AEs were graded for severity by the investigator. Erythema, induration, and swelling of 25–50 mm, 51–100 mm, and >100 mm and fever as axillary temperatures of ≥38.0–38.9, ≥39.0–39.9, and ≥40.0°C were graded as mild, moderate, and severe, respectively. All other local and systemic AEs, if present, were classified as: mild (present but not interfering with activity), moderate (interfering with activity), and severe (preventing daily activity).

Serious AEs (SAEs), AEs of special interest (AESIs), and AEs leading to withdrawal from the study were collected for the entire duration of the study.

The relationship between study vaccination and unsolicited AEs, medically attended AEs, any new onset of chronic disease, AEs leading to withdrawal, and SAEs were also assessed by the investigator. As defined in the clinical protocol, reactive arthritis and neutropenia were AESIs (the former being a general concern for enteric pathogen vaccinations and the latter due to occurrence of similar events during the phase 1 trials evaluating the 1790GAHB vaccine) and, if present, were to be reported as SAEs.

Blood samples for hematology (white blood cells, red blood cells, hemoglobin, hematocrit, platelets, eosinophils, basophils, neutrophils, monocytes, and lymphocytes) and clinical chemistry (total bilirubin, aspartic aminotransferase, alanine aminotransferase, γ-glutamyl transferase, lactic dehydrogenase, alkaline phosphatase, glucose, blood urea nitrogen, creatinine, sodium, and potassium) testing, and urine dipstick samples (for the assessment of glucose, proteins, pH, ketones, nitrites, and blood levels) were collected at 7 and 28 days post-each vaccination. Urinalysis (white blood cells, red blood cells, casts, and bacteria) was performed if urine dipstick showed deviations from normal values. Laboratory measurements were assessed by the investigators and any abnormality considered as clinically significant was reported as an AE.

All study participants with a neutropenia (ANC < 1.8 × 109 /L for adults in the first cohort and ANC < 1.0 × 109 /L for individuals from the second cohort), occurring at 7 days after each vaccination, had additional blood draws for repeated complete blood count on a weekly basis until resolution of neutropenia. Occurrence of ANC < 0.5 × 109 /L, after the first vaccination, was an exclusion criterion for second vaccination.

### Immunogenicity Assessments

Blood samples for immunogenicity assessments were collected prior to first vaccination and 28 days post-each vaccination (**Figure 2**). The sera were kept frozen at −80°C at the KEMRI-Wellcome Trust laboratory until shipment to the GSK Clinical Laboratory Sciences (Marburg, Germany), for serologic testing; for each participant, one aliquot of serum was stored in the clinical site laboratory and one aliquot was used for immunogenicity analyses.

Anti-*S. sonnei* LPS serum IgG was measured by an enzymelinked immunosorbent assay (ELISA), as previously described (15).

Seroresponse to vaccination was defined as an increase in the anti-*S. sonnei* LPS serum IgG level of at least 50% for participants with pre-vaccination levels >50 ELISA units (EU) or an increase of at least 25 EU for participants with pre-vaccination levels ≤50 EU.

In the absence of a correlate of protection, the median anti-*S. sonnei* LPS serum IgG following vaccination was compared to the median level in convalescent patient sera from individuals infected with *S. sonnei*, as previously reported (20). Postvaccination levels of 121 EU were estimated to correspond to the median end point titer of 1:800 measured for the convalescent sera with an ELISA method by Cohen et al. (16).

### Statistical Analyses

A total of 72 participants were planned to be enrolled in the study. No formal statistical sample size was calculated, due to the descriptive nature of the study objectives.

Safety analyses were carried out in all participants from the exposed full analysis set who received at least one study vaccination and had safety data. The number and percentage of participants with AEs, SAEs, AESIs, and deviations from normal ranges of safety laboratory data after vaccination was calculated.

Serologic analyses were performed on participants from the full analysis set who had available ELISA data. The ELISA antibody concentrations were logarithmically transformed (base 10). For each group, geometric mean concentrations (GMC) and their 95% confidence intervals (CIs) were computed by exponentiating (base 10) the mean and 95% CIs of the log10 ELISA concentration. ELISA concentrations below the limit of detection were set to half that limit for the purposes of analysis.

The number and percentage of participants with seroresponse and with postvaccination levels ≥121 EU for anti-*S. sonnei* LPS serum IgG at 28 days post-each vaccination was calculated together with 95% Clopper–Pearson CIs.

Additionally, geometric mean ratios (GMR) were computed for GMC at 1 month after first and second vaccination versus baseline levels (D1). The GMR and 95% CIs were constructed by exponentiating the mean within-subject differences in logtransformed titers and the corresponding 95% CIs.

### RESULTS

### Demographics

A total of 152 adults were screened, 74 were enrolled and randomized, 72 received at least one study vaccination, and 66 completed the study. Main reasons for the 78 screening failures were: not fulfillment of inclusion/exclusion criteria (*n* = 52) and lack of interest to further trial participation, despite being eligible (*n* = 15). Of the two individuals who were not vaccinated after randomization, one declined vaccination and the other was erroneously randomized, after expiry of the allowed 28-day window for screening. The primary reasons for study discontinuation were AEs (*n* = 4), administrative reason (*n* = 1), lost to followup (*n* = 1), protocol deviation (*n* = 1), and consent withdrawal (*n*= 1) (**Figure 2**). The overwhelming majority of the participants were male (88–91% in each group) and of Black origin (≥95% across all groups). Baseline characteristics were well-matched across all vaccine groups (**Table 1**).

### Safety

Following the first vaccination, pain was the only reported solicited local AE, in 20 (91%), 25 (96%), and 10 (42%) participants in the 1.5/25 μg, 5.9/100 μg, and control groups, respectively (**Table 2**). Post-second vaccination, pain was reported by 15 participants in each of the 1.5/25 μg (68%) and 5.9/100 μg (65%) groups, compared with 17 (81%) in the control group, while induration was only reported by 1 (5%) participant in the control group. All reported pain was mild to moderate (**Figure 3**). No severe local reactions were recorded (**Table 2**).

Table 1 | Baseline characteristics of vaccinated study participants (full analysis set).


*N, number of exposed participants in each group; n (%), number (percentage) of participants in each category; BMI, body mass index; 1.5/25* μ*g, 5.9/100* μ*g, participants who received the 1790GAHB vaccine with an O antigen/protein content of 1.5/25 and 5.9/100* μ*g, respectively, at days 1and 29; Control, participants who received meningococcal vaccine against serogroups A, C, W, and Y at day 1 and tetanus, diphtheria, and acellular pertussis vaccine at day 29.*

The most frequently reported solicited systemic AE was headache, reported by 5 (23%), 10 (38%), and 8 (33%) participants following first vaccination and 4 (18%), 5 (22%), and 4 (19%) participants following the second vaccination, in the 1.5/25 μg, 5.9/100 μg, and control groups, respectively (**Table 2**). The incidence of all systemic AEs seemed to decrease following the second vaccination and severe reactions were only reported in 1 (4%) participant in the 5.9/100 μg (severe headache occurring 6 h after the first vaccination and only lasting for that day) (**Table 2**).

Unsolicited AEs following any vaccination were reported by 19 (86%), 24 (92%), and 19 (79%) participants in the 1.5/25 μg, 5.9/100 μg, and control groups, respectively. Possibly or probably related unsolicited AEs were reported by 13 (59%) participants in the 1.5/25 μg and 16 (62%) in the 5.9/100 μg groups, compared to 12 (50%) in the control group. Most of these AEs were post-immunization reactions continuing beyond the 7-day collection window following vaccination (**Table 3**). Fever was very seldom observed (**Table 3**) and was never ≥39°C; no participant reported the use of analgesics/antipyretics within 24 h prior to each vaccination.

During the trial, 10 episodes of neutropenia fulfilling the protocol definition of AESI were reported in two (9%), three (12%), and one (4%) participants in the 1.5/25 μg, 5.9/100 μg, and control groups, respectively. These episodes were all considered probably or possibly related to vaccination, two of them occurred in the 1.5/25 μg group (one mild and one moderate), five in the 5.9/100 μg group (four mild and one moderate), and three in the control group (one mild, one moderate, and one severe). All cases were transient (i.e., recovery within 7 days and

Table 2 | Number and percentage of participants with solicited local and systemic adverse events (AEs) (full analysis set).


*N, number of exposed participants in each group; n (%), number (percentage) of participants in each category; 1.5/25* μ*g, 5.9/100* μ*g, participants who received the 1790GAHB vaccine with an O antigen/protein content of 1.5/25 and 5.9/100* μ*g, respectively, at days 1 and 29; Control, participants who received meningococcal vaccine against serogroups A, C, W, and Y at day 1 and tetanus, diphtheria, and acellular pertussis vaccine at day 29.*

triangles represent the maximum individual pain reported after each vaccination at different doses. The black line represents the average pain score as a function of dose.

Table 3 | Number and percentage of participants with possibly or probably related unsolicited adverse events (AEs) following any vaccination (full analysis set).


*N, number of exposed participants in each group; n (%), number (percentage) of participants in each category; 1.5/25* μ*g, 5.9/100* μ*g, participants who received the 1790GAHB vaccine with an O antigen/protein content of 1.5/25 and 5.9/100* μ*g, respectively, at days 1 and 29; Control, participants who received meningococcal vaccine against serogroups A, C, W, and Y at day 1 and tetanus, diphtheria, and acellular pertussis vaccine at day 29.*

by study end) and, except for one participant in the 5.9/100 μg group, who experienced upper respiratory tract infection (D1) and 38.1°C fever (D8), were also asymptomatic, as confirmed by daily home visits during the 7 days postvaccination. Eight of these episodes occurred in 5 of the 18 individuals from the first cohort (screened and monitored using the ANC Western threshold of 1.8 × 109 /L), while the remaining two episodes occurred in one of the 54 participants in the second cohort (screened and monitored using the local threshold of 1.0 × 109 /L). No SAEs occurred in the study. However, as defined in the study protocol the AESIs were reported as SAEs. No case of reactive arthritis was recorded.

Four participants were prematurely withdrawn due to unsolicited AEs: three in the 5.9/100 μg group (two cases of neutropenia and one case of bone tuberculosis) and one in the control group (γ-glutamyl transferase increase). One participant, who should have been excluded from the second vaccination due to neutropenia, was inadvertently not withdrawn and experienced moderate neutropenia at D57.

Few laboratory abnormalities were considered as clinically significant by the investigators. These were: an increase in alkaline phosphatase levels in one participant in the control group; γ-glutamyl transferase increase in one individual in the control group, and one in the 5.9/100 μg group; decreased hematocrit and hemoglobin levels in one participant in the 5.9/100 μg group; low ANC for one participant in group 5.9/100 μg group; a low platelet count for one participant in the control group; and one increase in platelet count in the 5.9/100 μg group. Following urinalysis, increased leukocyte levels for two, three, and one participants in the 1.5/25 μg, 5.9/100 μg, and control groups, respectively, and high erythrocyte level for one individual in each of the 1.5/25 and 5.9/100 μg groups were considered clinically significant. There were no pregnancies, hospitalization or deaths reported in the study.

### Immunogenicity

Pre-vaccination anti-*S. sonnei* LPS IgG GMC varied between 971 and 1,196 EU across all groups (**Figure 4A**). At 28 days post-first vaccination, antibody levels increased 2.10- and 4.43-fold from baseline values in the 1.5/25 and 5.9/100 μg groups, respectively, but no increase was observed in the control group (**Figures 4A,B**). At 28 days after the second vaccination, anti-*S. sonnei* LPS IgG GMC further increased in the 1.5/25 μg group, but not in the 5.9/100 μg group; however, changes observed from D29 to D57 were not statistically significant (**Figure 4A**).

In the 1.5/25 μg group, the seroresponse was 68% after the first vaccination and 90% after the second vaccination, whereas in the 5.9/100 μg group, seroresponse was 96% after both the first and second vaccination (**Figure 5A**).

At baseline, the percentages of participants with anti-*S. sonnei* LPS IgG ≥ 121 EU were 95, 92, and 88% for the 1.5/25 μg, 5.9/100 μg, and control group, respectively. At 28 days following

Figure 4 | Anti-*Shigella sonnei* LPS IgG geometric mean concentrations (A) and geometric mean ratios (B), by timepoint (full analysis set for immunogenicity). LPS, lipopolysaccharide; IgG, immunoglobulin G; EU, enzyme-linked immunosorbent assay units; 1.5/25 μg, 5.9/100 μg, participants who received the 1790GAHB vaccine with an O antigen/protein content of 1.5/25 μg and 5.9/100 μg, respectively, at days 1 and 29; Control, participants who received meningococcal vaccine against serogroups A, C, W, and Y at day 1 and tetanus, diphtheria, and acellular pertussis vaccine at day 29.

each vaccination, all participants in the investigation groups achieved anti-*S. sonnei* LPS IgG ≥ 121 EU, while in the control group 86 and 90% of participants had this level at 28 days postfirst and second vaccination, respectively (**Figure 5B**).

Reverse cumulative distribution curves for anti-*S. sonnei* LPS antibody levels pre-vaccination and following each vaccination with the 1790GAHB vaccine were compared to antibody levels in convalescent patients as shown in **Figure 6**. The review of individual immunogenicity results showed that data from five serum samples (out of the 210 collected during the study and shipped for testing), all obtained at D29, were clinically implausible. Most likely, the root cause was human error during sample labeling, executed in the same days for 2 and 3 of the samples, respectively; however, as no definite proof of an error could be established, the original data have been used for the analyses. Additional analyses were performed excluding the potentially invalid results and the interpretation of immunogenicity results did not change.

### DISCUSSION

This is the first study to provide clinical data for a GMMA-based *Shigella* vaccine in a country endemic for shigellosis. For both assessed vaccine strengths, the *S. sonnei* 1790GAHB vaccine was well tolerated; confirming safety results previously shown in age-matched European adults and supporting further potential testing of GMMA-based vaccines in younger individuals from developing countries.

The overall incidence of solicited local and systemic reactions was comparable between the groups receiving the 1790GAHB vaccine and the control vaccines, and very few severe reactions were observed. No increase in the reporting rates was observed following the second vaccination, even in the group receiving the formulation with a higher OAg/protein dose level (5.9/100 μg). Mild to moderate pain at injection site was the only solicited local AE reported in recipients of the investigational vaccine, however, as illustrated in **Figure 3**, pain intensity appeared to be lower than that previously reported in European age-matched adults (16). The majority of unsolicited AEs were local and systemic reactions continuing beyond the 7-day period following each dose.

Based on prior experience with phase 1 clinical trials (21), neutropenia was followed as an AESI and during this study, 10 episodes occurred in six participants. All but two occurred in the 18 participants from the first cohort, evaluated based on Western ANC normality ranges. Had local ranges been used for the whole study, there would have been only two cases of neutropenia, one mild and one moderate, in one single individual vaccinated with 1790GAHB and one mild episode in one participant from the control group.

These data support previous observations that populations of African descent have a lower ANC than other ethnicities (22, 23) and that in these populations, targeted clinical laboratory reference intervals should be used (18).

The investigational vaccine was highly immunogenic at both assessed OAg/protein contents. Following the first vaccination, a higher increase in anti-*S. sonnei* LPS serum IgG was observed in the 5.9/100 μg group than in the 1.5/25 μg group. Following a second vaccination, anti-*S. sonnei* LPS serum IgG levels further increased in the group receiving the 1.5/25 μg dose, but not in the 5.9/100 μg group.

Of note, a very high level of preexisting antibodies was observed among the study participants, and baseline GMC in all groups were much higher than those from the European study (16) or than median antibody titer established by Cohen et al. in Israeli individuals naturally exposed to *S. sonnei* (20). This finding may be explained by prior and repeated exposure to *S. sonnei* as previously theorized (24, 25). In fact, according to the Global Enteric Multicentre Study, *S. sonnei* is among the predominant *Shigella* species in Kenya (6), and, overall, its prevalence in Africa has increased in the last decades (8). Additionally, *S. sonnei* was recently identified as the main pathogen in young children from Western Kenya hospitals presenting with acute diarrhea between 2011 and 2014, accounting for ~54% of *Shigella* infections (26), so a prior and repeated exposure of the study participants to this pathogen in an hyper endemic country is a strong possibility.

Patients who have a high level of baseline antibody are generally less likely to have further significant increase in antibody levels after immunization, due to masking of the vaccine epitope and/or other mechanisms of specific B cells inhibition (27, 28).

By contrast, in our study, we observed a robust specific antibody response in both groups receiving the *Shigella* vaccine, although we also found a reduced fold increase in those subjects with the highest baseline antibodies. Overall the magnitude of the response was much greater than that observed in the European population (16) and this outcome can be considered a strength of the 1790GAHB vaccine. Additionally, compared to vaccines exclusively containing the OAg, GMMA have the advantage of presenting multiple outer membrane antigens to the immune system and induce immunological responses through targets other than OAg. We performed a proteomic analysis of *S. sonnei* GMMA (29) and identified a total of 434 proteins with similar composition and relative abundance to the outer membrane and periplasm of the parent bacteria. The four most abundant proteins by mass were OmpA, OmpC, Entericidin B, and then OmpX.

The lack of information on the potential contribution of protein antigens to the immunological response against *Shigella* is a limitation of this trial and should be further investigated in future studies. In addition, antibody levels were not evaluated beyond 28 days after the second immunization, which does not allow conclusions to be drawn on the persistence of responses to vaccination, and only the quantity but not the quality of antibody response was determined; this latter limitation will be addressed by further testing trial serum samples in a *S. sonnei* serum bactericidal assay. Finally, the sample size of the trial was relatively small and a formal statistical interpretation was not planned.

### CONCLUSION

The GMMA-based 1790GAHB vaccine against *S. sonnei* displayed good safety and immunogenicity profiles in healthy adults from a shigellosis-endemic country in Africa, thus paving the way for the future testing of multivalent GMMA-based *Shigella* vaccines in young children and infants, the age group with the highest burden of shigellosis in resource-poor populations.

### ETHICS STATEMENT

The informed consent form and the study protocol were reviewed and approved by the KEMRI Scientific and Ethics Committee, the Kenyan Pharmacy and Poisons Board and the Oxford Tropical Research Ethics Committee prior to study start. The trial was

### REFERENCES


designed and conducted in agreement with the ICH Harmonized Tripartite Guidelines for Good Clinical Practice, applicable local regulations and the Declaration of Helsinki and was registered at ClinicalTrials.gov (NCT02676895).

# AUTHOR CONTRIBUTIONS

PN, PB, CO, AS, LM, AP, JA, ASS, and EM were involved in the design of the study. CO, PN, BK, AG, and PB performed the study and participated in the collection or generation of the study data. LS was responsible for generation of the immunogenicity data. All authors were involved in the analyses and interpretation of the data.

### ACKNOWLEDGMENTS

The DSMB [Drs. Bernard Fritzell (Chair), Paolo Bonanni, Joseph Mbuthia and Simonetta Viviani] are thanked for their support in reviewing study data. This manuscript is published with the permission of the Director of KEMRI. The contribution of study participants, nurses, and other KEMRI staff members is gratefully acknowledged. Likewise, all GVGH personnel, who contributed with their work to make clinical studies possible, are acknowledged and thanked. Authors would also like to thank Petronela M. Petrar and Botond Nagy for medical writing support and Susana Montenegro Gouveia for manuscript development and editorial support (XPE Pharma & Science c/o GSK).

### FUNDING

This study was funded in part by a FP7 grant (280873, ADITEC) and GlaxoSmithKline Biologicals SA. GlaxoSmithKline Biologicals SA also took responsibility for all costs associated with the development of the present manuscript. The costs associated with the publishing of the manuscript were covered by funds from the ADITEC grant.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01884/ full#supplementary-material.


**Conflict of Interest Statement:** AN, ASS, EM, JA, AS, LM, and AP are all employees of the GSK group of companies and report grants from the EU FP7 (Grants 261472 and 280873) during the conduct of the study and grants from Bill and Melinda Gates Foundation, outside the submitted work. DR and LS are employees of the GSK group of companies. EM, JA, AS, LM, and AP own GSK shares. AS has two patents pending (US2016289632 and US2015202274) and one issued (WO2016202872) to GlaxoSmithKline Biologicals SA. LM has one patent issued (WO2016202872) to GlaxoSmithKline Biologicals SA. CO, PN, BK, AG, and PB declare no conflict of interest.

The handling editor declared a past co-authorship with the authors.

*Copyright © 2017 Obiero, Ndiaye, Sciré, Kaunyangi, Marchetti, Gone, Schütte, Riccucci, Auerbach, Saul, Martin, Bejon, Njuguna and Podda. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Alum/Toll-Like Receptor 7 Adjuvant Enhances the Expansion of Memory B Cell Compartment Within the Draining Lymph Node

*Hoa Thi My Vo1 , Barbara Christiane Baudner <sup>2</sup> , Stefano Sammicheli 3,4, Matteo Iannacone3,4, Ugo D'Oro1 \* and Diego Piccioli1 \**

*1Preclinical Research, GSK Vaccines, Siena, Italy, 2Global Discovery Support and New Technologies, GSK Vaccines, Siena, Italy, 3Dynamics of Immune Responses, Division of Immunology, Transplantation and Infectious Diseases, Experimental Imaging Center, IRCCS San Raffaele Scientific Institute, Milano, Italy, 4Vita-Salute San Raffaele University, Milano, Italy*

### *Edited by:*

*Donata Medaglini, University of Siena, Italy*

### *Reviewed by:*

*Ali M. Harandi, University of Gothenburg, Sweden Rhea Coler, Infectious Disease Research Institute, United States*

### *\*Correspondence:*

*Ugo D'Oro ugo.x.doro@gsk.com; Diego Piccioli diego.x.piccioli@gsk.com*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 11 October 2017 Accepted: 14 March 2018 Published: 09 April 2018*

### *Citation:*

*Vo HTM, Baudner BC, Sammicheli S, Iannacone M, D'Oro U and Piccioli D (2018) Alum/Toll-Like Receptor 7 Adjuvant Enhances the Expansion of Memory B Cell Compartment Within the Draining Lymph Node. Front. Immunol. 9:641. doi: 10.3389/fimmu.2018.00641*

Vaccination is one of the most cost-effective health interventions and, with the exception of water sanitization, no other action has had such a major effect in mortality reduction. Combined with other approaches, such as clean water, better hygiene, and health education, vaccination contributed to prevent millions of cases of deaths among children under 5 years of age. New or improved vaccines are needed to fight some vaccine-preventable diseases that are still a threat for the public health globally, as reported also in the Global Vaccine Action Plan (GVAP) endorsed by the World Health Assembly in 2012. Adjuvants are substances that enhance the effectiveness of vaccination, but despite their critical role for the development of novel vaccines, very few of them are approved for use in humans. Aluminum hydroxide (Alum) is the most common adjuvant used in vaccines administered in millions of doses around the world to prevent several dangerous diseases. The development of an improved version of Alum can help to design and produce new or better vaccines. Alum/toll-like receptor (TLR)7 is a novel Alum-based adjuvant, currently in phase I clinical development, formed by the attachment of a benzonaphthyridine compound, TLR7 agonist, to Alum. In preclinical studies, Alum/TLR7 showed a superior adjuvant capacity, compared to Alum, in several disease models, such as meningococcal meningitis, anthrax, staphylococcus infections. None of these studies reported the effect of Alum/TLR7 on the generation of the B cell memory compartment, despite this is a critical aspect to achieve a better immunization. In this study, we show, for the first time, that, compared to Alum, Alum/ TLR7 enhances the expansion of the memory B cell compartment within the draining lymph node (LN) as result of intranodal sustained proliferation of antigen-engaged B cells and/or accumulation of memory B cells. In addition, we observed that Alum/ TLR7 induces a recruitment of naïve antigen-specific B cells within the draining LN that may help to sustain the germinal center reaction. Our data further support Alum/TLR7 as a new promising adjuvant, which might contribute to meet the expectations of the GVAP for 2020 and beyond.

Keywords: adjuvant, vaccination, B cell, immunological memory, lymph node, Alum/toll-like receptor 7

### INTRODUCTION

Immunization is one of the most powerful and cost-effective health interventions. It prevents debilitating illness and disability, and saves millions of lives every year (1, 2). For this reason, immunization has a tremendously beneficial impact on the socioeconomic development of a country and helps to reduce the development gap between high-income and low-income countries. Access to immunization program should, therefore, be recognized as both a core component of the human right to health and an individual, community and governmental responsibility (1, 2).

Thus, research and development for new or improved vaccines together with the efforts to accelerate their market release are considered by the World Health Organization as part of a strategic approach to prevent diseases globally (1, 2). From this point of view, the development of new technologies for vaccine design is recommended as a health priority (1, 2).

Adjuvant technology is one of the leading technologies to design novel, safe, and effective vaccines (1, 3–6). Adjuvants are substances that help a vaccine to enhance its clinical effectiveness (1, 3–6). They can reduce the time the body takes to mount a protective response and can make the immune response more broadly protective against several related pathogens (1, 3–6). Despite the large amount of research in novel adjuvant discovery and the growing pipeline of in development adjuvants, up to now, only five adjuvants have been licensed for use in vaccines administered to humans (1, 3–6).

The physicochemical characteristics of different adjuvants can vary greatly as well as their mechanisms of action (3–6). However, the two main general features of an adjuvant are to facilitate the antigen delivery to antigen-presenting cells (delivery system) or to stimulate immune cells (immune-potentiator) (3–6). This last function can be exerted through targeting various receptors working as sensors of danger signals, such as tissue damage, or recognizing molecular patterns not belonging to the host body, like lipopolysaccharide, and that can discriminate self from non-self (3–6). Toll-like receptors (TLRs) belong to this last type of receptors termed pattern recognition receptors (3–8). There are several TLRs that are able to recognize different pathogenspecific molecular patterns (3–8). For example, TLR4 is able to recognize the lipid A part of the lipopolysaccharide of Gramnegative bacteria, whereas TLR7 is able to recognize the single strand RNA typical of a certain family of viruses (3–8). Several natural or synthetic agonists of TLRs are currently evaluated in preclinical research or clinical development as novel vaccine adjuvants (3–8).

Adjuvants based on insoluble salts of aluminum are currently the oldest and most used adjuvants in human vaccines and have been shown to be safe, well tolerated, and effective (1–6). For example, all the diphtheria–tetanus–pertussis vaccines commercially available, which are administered to hundreds millions of children each year and that are giving a definitive contribution to eliminating diphtheria, tetanus, and pertussis diseases, are adjuvanted using insoluble salts of aluminium (1–6). Recently an improved version of Alum, termed AS04, containing adsorbed monophosphoryl lipid A from *Salmonella minnesota* (which is a TLR4 agonist) has been licensed in a new vaccine against the human papilloma virus, to prevent cervical cancer in women (9).

Alum/TLR7 is a novel and improved Alum-based adjuvant containing a synthetic TLR7 agonist, with a benzonaphthyridine chemical scaffold, adsorbed to aluminum hydroxide (10–14).

Alum/TLR7 adjuvant is currently in phase I clinical development and preclinical data in the mouse model already demonstrated a significant superior capacity of this new adjuvant, compared to Alum, in eliciting an effective immune response against different pathogens (10–12, 14).

The adsorption on Alum of the specific benzonaphthyridine compound that identifies Alum/TLR7 significantly enhanced the humoral immune response against *Neisseria meningitidis* B after immunization with recombinant antigens from this bacterium adsorbed on Alum (10).

Using the *Bacillus anthracis* model, immunization with the bacterial toxin formulated with Alum/TLR7 increased the toxin-neutralizing antibody titers and, at the same time, the passive transfer of serum from these immune mice into naïve animals provided a significant increase in protection rate after challenge (10).

In a *Staphylococcus aureus* model, formulation with Alum/ TLR7 significantly enhanced the effectiveness of antibody and CD4 T cell responses induced by immunization of mice with four proteins as vaccine antigen candidates (11, 14).

The enhancement of the effectiveness of the humoral immune response after immunization with Alum/TLR7, compared to Alum, was also observed with a tetravalent glycoconjugate vaccine against *N. meningitidis* ACWY (12). Consistently, in *N. meningitidis* C model, the higher adjuvant potential of Alum/ TLR7 compared to Alum has been shown as dependent on the signaling activity of TLR7 (12).

Although the attachment of a TLR7 benzonaphthyridine compound to Alum is associated with a clear improvement of the humoral immune response to a vaccine, none of the previous studies evaluated the effect of this new adjuvant on the B cell response and particularly on the formation of the memory B cell compartment. Indeed, the generation of memory B cells is critical to mount an effective immunity (15–18) because memory B cells differentiated from the germinal center (GC) reaction within the B cell follicle of the lymphoid organs, express high affinity isotype-switched surface antibodies against the antigen (15–21). Consequently, antibody-secreting plasma cells (PCs) generated, through the GC reaction from memory B cells, produce effective antibodies with high affinity for the antigen (19, 22, 23). Thus, the generation of memory B cells actually confers the immunity to the microbial infections of the body (24, 25). Alum/TLR7, compared to Alum, is capable to generate a stronger and more efficient humoral immune response not only after the primary immunization (12, 14). In fact, this potent adjuvant effect of Alum/TLR7 is particularly evident after the boost using the combination of four glycoconjugate antigens from *N. meningitidis* ACWY strains (12). This observation drives to hypothesize that a significant higher expansion of the memory B cell compartment occurs during a primary immunization with Alum/TLR7, compared to Alum. Thus, we wondered if the enhancement in the humoral immune response observed after a secondary immunization with this new adjuvant could be associated with an increase of memory B cells after primary immunization. This finding would consequently demonstrate that the stronger adjuvant effect induced by Alum/ TLR7, compared to Alum, may be also correlated to an increase in the generation of memory B cells. In particular, we were interested in investigating the formation of the memory B cells within the lymphoid organs that are the principal locations where this critical and complex event takes place (20, 26). However, given the low number of endogenous antigen-specific memory B cells, can be challenging to perform this type of investigation. Therefore, to bypass this shortcoming, we set up a mouse adoptive transfer system in which, using transgenic naïve B cells, we increased the frequency of antigen-specific B cells within naïve mice and selectively followed the fate of these B cells after immunization with the cognate antigen.

# MATERIALS AND METHODS

### Mice

C57BL/6 female mice, 6/8 weeks old, were purchased from Charles River Laboratory. KL25 (27) transgenic female mice were bred at the San Raffaele Scientific Institute animal facility and 6/8 weeks old animals were used to isolate naïve B cells for the adoptive transfer. All mice were maintained under pathogen-free condition. All animal experiments were approved by local GSK Animal Welfare Body and performed in compliance with the European directive 2010/63/UE and the Italian law DL 26/14. The authorization codes for animal experimentation were AWB2012- 03, AWB2014-05, and AWB2015-01. Mice were sacrificed by cervical dislocation before the last blood sampling or before each lymphoid organ sampling.

# Preparation of Splenocytes

Spleens were smashed using a 70 µm Cell Strainer (Falcon, Becton Dickinson) and the splenocytes were collected using cold RPMI-1640 medium (Gibco, Life Technologies). Splenocytes were centrifuged at 300 *g* for 10 min at room temperature and the pelleted cells were treated with a red blood cell lysis buffer (BioLegend), according to the manufacturer's instructions. Obtained splenocytes were filtered with a 70 µm Cell Strainer (Falcon, Becton Dickinson).

# Adoptive Transfer

Naïve B cells were purified from splenocytes of KL25 transgenic mice through negative selection using a B cell isolation kit (Milteny Biotec), according to the manufacturer's instructions. Purity of B cell preparation was evaluated by flow cytometry after labeling the cells with anti-CD19 antibody and it was routinely around 98%. Isolated untouched naïve KL25 B cells were loaded with CFSE (ThermoFisher Scientific) at a final concentration of 5 µM, for 15 min at room temperature in 2 ml of PBS, protected from light. Then 5 or 10 volumes of PBS containing 5% of FCS was added to stop the loading reaction. CFSE-loaded cells were washed twice by centrifuging cells at 300 *g* for 10 min at room temperature. Washed cells were suspended in physiologic solution at a concentration of five million cells per 100 µl and intravenously injected (tail) into naïve C57BL/6 mice (five million KL25 B cells per mouse).

### Formulations

Envelope glycoprotein 1 of lymphocytic choriomeningitis virus (GP1-LCMV) (28) was used as model antigen and prepared as previously described (29). Formulations were prepared by adsorbing the antigen to Alum (concentration of Alum used: 2 mg/ml) or Alum/TLR7 (Alum: 2 mg/ml). Alum/TLR7 was prepared adsorbing a proprietary benzonaphthyridine compound TLR7 agonist to Alum. Each immunization dose contained 0.1 µg of GP1-LCMV. Alum/TLR7 contained 10 µg of the TLR7 agonist in each immunization dose. The volume dose was 20 µl. Antigen adsorption was performed incubating the antigen or the TLR7 agonist at 4°C for 2 h under gentle agitation. Before immunization, each individual formulation underwent the following quality control: (a) presence of bubbles or precipitates checked by visual inspection; (b) antigen identity, integrity, and Alum-adsorption checked by Western Blot analysis; and (c) TLR7 agonist adsorption checked by chromatography analysis.

### Immunizations

C57BL/6 mice were injected intramuscularly (calf muscle) in one leg with a dose of the following treatments: GP1-LCMV, GP1- LCMV adsorbed on Alum and GP1-LCMV adsorbed on Alum/ TLR7. For immunogenicity experiments, groups of 10 mice per each treatment were immunized twice, with a 4 weeks interval between the first and the second immunization. Mouse sera were collected the day before the first immunization (pre-immune sera) and 2 weeks after each immunization (post 1 and post 2 sera). For B cell response analysis, the day after the adoptive transfer of KL25 naïve B cells, groups of three adoptively transferred mice per each treatment were immunized (one time) intramuscularly (calf muscle) in one leg and the popliteal draining lymph nodes (LNs), contralateral popliteal non-draining LNs, and spleens were collected 3 days, 1 week, and 2 weeks after the immunization.

### Preparation of LN Cell Suspensions

Lymph nodes collected after the immunization were immediately processed by enzymatic digestion (pooling the LNs per each type of treatment a keeping separated draining and nondraining LNs) incubating them in RPMI-1640 medium (Gibco, Life Technologies) containing 250 µg/ml DNase I (Roche) and 500 µg/ml Liberase Research Grade (Roche) for 1 or 2 h (depending on the LN size) at 37°C and agitating by pipetting every 15 or 30 min. The obtained cell suspensions were centrifuged at 300 *g* for 10 min at room temperature and washed with RPMI-1640 medium (repeating the centrifugation). Washed cells were then filtered with a 70 µm Cell Strainer (Falcon, Becton Dickinson) and suspended in RPMI-1640 medium supplemented with 10% FCS (HyClone) and 1% PSG (Gibco, Life Technologies).

# Antibodies for Flow Cytometry

APC-Cy7-conjugated anti-CD45.1 (BioLegend), BV785 conjugated anti-CD45.2 (BioLegend), PE-Cy5-conjugated anti-CD38 (eBioscience), PE-CF594-conjugated anti-CD80 (BD Horizon), PE-Cy7-conjugated anti-CD73 (eBioscience), BV421 conjugated anti-IgM (BD Horizon), PE-conjugated anti-IgD (BD Pharmigen), Alexa Flour-647-conjugated anti-GL7 (BD Pharmigen), and APC-conjugated anti-CD19 (eBioscience). All antibodies were titrated against their respective isotype controls from the same manufacturer to determine the optimal dilution for cell labeling.

### Flow Cytometry

Cells (two or three million) of LN cell suspensions were labeled with the live/dead aqua cell stain kit (Molecular Probes; Invitrogen-Life Technologies) diluted 1:1,000 in 100 µl of PBS by incubation for 20 min at 4°C in dark condition and then washed with PBS by centrifugation at 300 *g* for 10 min at room temperature. After this step, the cells were incubated in the same conditions as above with Mouse BD Fc Block reagent (BD Pharmingen) diluted 1:1,000 in 100 µl of PBS, 1% BSA and washed again as previously described. Then, the LN cell suspensions were labeled, using the same procedures as described above, with an appropriate antibody mix, using antibodies diluted according to the titration (a mix of titrated isotype-matched antibodies was used as negative control). Finally, the cells were washed again as previously described and suspended in 150 µl of PBS for flow cytometry analysis in a FACS LSRII SOS1 instrument (Becton Dickinson). Data were analyzed with the FlowJo software.

### ELISA

96-well plates (maxisorp NUNC) were coated with 50 µl/well of purified goat anti Human IgG Fc fragment (Jackson) diluted (1:1,000) in carbonate buffer, pH 9.6 and incubated over night at 4°C. Then plates were washed three times with PBS, Tween 0.1% (wash buffer), and then blocked for 1 h at room temperature with 200 µl/well of PBS, Tween 0.01%, 5% fat-free milk as blocking buffer. After this incubation, plates were washed one time with wash buffer and 50 µl/well of GP1-LCMV-Fc supernatant were added. After 1 h of incubation at room temperature, the plates were washed five times with the wash buffer. Then mouse sera, serially diluted in blocking buffer, were added using a volume of 50 µl/well and plates were incubated for 1 h at room temperature and then washed five times with wash buffer. After this step, 50 µl of goat anti-mouse IgG H + L HRP (PerkinElmer), diluted 1:2,000 in blocking buffer, were added in each well and after 1 h of incubation at room temperature, the plates were washed five times with the wash buffer. Then 100 µl of TMB substrate (KPL) were added to each well and plates were incubated for 10 min, at room temperature, in the dark condition. To stop the reaction, sulfuric acid 1 M was added using a volume of 100 μl/well. Finally, absorbance at 450 nm was measured using a plate spectrophotometer (SpectraMax M2, Molecular Devices). Antibody titers were expressed as the reciprocal dilution corresponding to a cut-off at OD450 = 0.5.

### RESULTS

### Detection of Proliferated and Non-Proliferated Antigen-Specific B Cells

We took advantage of the use of KL25 transgenic mice whose B cells specifically recognize GP1-LCMV. This mouse strain has a C57BL/6 genetic background expressing the CD45.1 allele, whereas C57BL/6 mice possess the CD45.2 allele. In this way, naïve B cells from non-immunized KL25 mice can be easily tracked once injected into syngeneic C57BL/6 mice through the identification of the CD45 alleles, avoiding rejection at the same time. Thus, mice adoptively transferred with GP1-LCMV-specific naïve B cells were immunized with the cognate antigen GP1- LCMV (used as model antigen) in order to track the generation of the memory compartment.

Before to analyze the memory B cell formation by using the adoptive transfer mouse model described above, we asked whether, with our model antigen, Alum/TLR7 promotes a significantly higher antibody response compared to Alum. To address this question, we evaluated the endogenous antibody response to GP1-LCMV in C57BL/6 mice non-adoptively transferred. We immunized mice twice with GP1-LCMV alone or formulated with Alum or Alum/TLR7 and we measured the antigen-specific antibody response. We confirmed that, also using GP1-LCMV as antigen, Alum/TLR7 induces a significantly higher antigenspecific antibody titers compared to Alum, after both primary and secondary immunization (Figure S1 in Supplementary Material).

Having verified that Alum/TLR7 is a better adjuvant than Alum also using our model antigen, we moved to analyze the effect of this adjuvant on the memory B cell response after primary immunization, using the adoptive transfer mouse model. We intravenously transferred CFSE-loaded purified naïve B cells, specific for GP1-LCMV, into C57BL/6 syngeneic mice and the day after we immunized these mice intramuscularly in one leg with GP1-LCMV alone, adsorbed on Alum or adsorbed on Alum/ TLR7. Mice injected with the formulation buffer were used as negative control. Treated mice were sacrificed 3 days, 1 week, and 2 weeks after the injection, and the draining LNs, contralateral non-draining LNs, and spleens were collected. Lymphoid organs were enzymatically digested and the obtained cell suspensions were analyzed by flow cytometry to detect GP1-LCMV-specific B cells, according to the reported gating strategy (Figure S2 in Supplementary Material).

Based on the CFSE labeling, we identify two populations of GP1-LCMV specific B cells: CFSElow/negative proliferated cells and CFSEhigh non-proliferated cells (Figure S2 in Supplementary Material). Consistently with the described timing of B cell response (21, 22), by measuring the percentage of CFSElow/negative cells, proliferated B cells appear 1 week after the immunization and are still detectable after 2 weeks (**Figures 1B,C**).

Proliferated B cells are detectable only in the draining LNs and not in the non-draining LNs (**Figure 1**) or in the spleens (data not shown) and only after immunization with adjuvants (**Figure 1**), consistently with the poor immunogenicity observed when immunizing with the antigen alone (Figure S1 in Supplementary Material). We do not observe major differences in the detection of proliferated B cells between mice immunized with Alum or Alum/TLR7, based on the percentage of CFSElow/negative antigenspecific B cells (**Figure 1**).

## Proliferated Antigen-Specific B Cells Display an Activated/Isotype-Switched Phenotype

As expected and independently on the type of adjuvant used for the immunization, proliferated B cells display an activated phenotype, characterized by upregulation of the co-stimulatory molecule CD80 (**Figure 2A**), downregulation of the CD38 (**Figure 2A**), and above all downregulation of the IgD (**Figure 2B**). This switch toward an activated and isotype-switched B cell phenotype is more evident 2 weeks after

Figure 1 | Proliferated antigen-specific B cells are detected only within draining lymph nodes (LNs), 1 week and 2 weeks post immunization with adjuvanted antigen. Mice adoptively transferred with CFSE-labeled glycoprotein 1 of lymphocytic choriomeningitis virus (GP1-LCMV)-specific naïve B cells were immunized intramuscularly in one leg with the cognate antigen alone or formulated with Alum or Alum/toll-like receptor (TLR)7. Mice treated with formulation buffer alone were used as negative control. Immunized mice were sacrificed 3 days, 1 week and 2 weeks after the treatment. Draining LNs and contralateral non-draining LNs were collected to be analyzed by flow cytometry to identify proliferated (CFSElow/negative) and non-proliferated (CFSEpositive) antigen-specific B cells, according to the gating strategy reported in Figure S1 in Supplementary Material. The graphs show the percentage of CFSElow/negative proliferated antigen-specific B cells within draining (filled symbols) and non-draining (empty symbols) LNs 3 days (A), 1 week (B), and 2 weeks (C) after the immunization. Each symbol represents data from a single experiment. Results from five independent experiments are reported. The black orizonal lines represent the average percentage of proliferated cells in the five independent experiments, per each type of treatment. The gray line sets the threshold of proliferation based on the mice treated with formulation buffer alone in non-draining LN at each time point.

the immunization (**Figure 2B**), when the vast majority of proliferated B cells are IgD/IgM double negative, whereas 1 week after the immunization, roughly 50% of proliferated B cells still express IgM (**Figure 2C**). Indeed, the percentage of IgM/IgD expressing antigen-specific B cells is similar in mice immunized with Alum or Alum-TLR7 (**Figure 2C**). Consistently, non-proliferated B cells are CD38high, CD80negative, and IgD/IgM double positive (**Figures 2A,B**).

### Detection of Memory B Cells

To identify more selectively the memory B cells, we used the markers GL7 and CD73. We confirmed that non-proliferated B cells are double negative for GL7/CD73, whereas the activated/ isotype-switched proliferated B cells, independently on the adjuvant used for immunization, upregulate GL7 and differentially express CD73 over time (**Figure 3A**). In agreement with the current knowledge, 1 week after the immunization almost all proliferated B cells display a GC phenotype (being mostly GL7<sup>+</sup>/ CD73<sup>−</sup>/low), whereas 2 weeks after the immunization they move toward a mature memory B cell phenotype (becoming prevalently GL7<sup>+</sup>/CD73<sup>+</sup>) (**Figure 3A**). This finding is also consistent with the more pronounced IgM<sup>−</sup>/IgD<sup>−</sup> isotype-switched phenotype displayed by antigen-specific B cells 2 weeks after the immunization (**Figures 2B,C**). Thus, we discriminated memory B cells based on the expression of GL7 and CD73 molecules, considering GL7/CD73 double positive antigen-specific proliferated B cells as memory B cells.

### Expansion of Memory B Cell Compartment

In order to measure the memory B cell compartment, we counted the number of memory B cells within the draining LNs of Alum and Alum/TLR7 immunized mice and we found that the treatment with Alum/TLR7 enhances the number of memory B cells within the draining LNs, 2 weeks after the immunization (**Figure 3B**, right panel), whereas this was not observed 1 week after the immunization (**Figure 3B**, left panel). Interestingly, when immunizing with Alum, the number of memory B cells 2 weeks after the immunization is inferior (although not Figure 1 | Continued

Figure 2 | Phenotype of proliferated versus non-proliferated antigen-specific B cells. (A) Representative flow cytometry histograms for CD80 (upper panels) and CD38 (lower panels) expression in proliferated (red lines) versus non-proliferated (blue lines) antigen-specific B cells, 1 week and 2 weeks after the immunization. (B) Representative flow cytometry dot plots of IgM versus IgD expression by non-proliferated (upper panels, blue dot plots) versus proliferated (lower panels, red dot plots) antigen-specific B cells, 1 week and 2 weeks after the immunization. (C) Percentage of IgM+/IgD− cells and IgM−/IgD− cells within proliferated antigen-specific B cells 1 week and 2 weeks after immunization with glycoprotein 1 of lymphocytic choriomeningitis virus (GP1-LCMV) adjuvanted with Alum or Alum/toll-like receptor 7. Bar graphs plot the average results, with SD, of five independent experiments.

statistically significant) compared to the number of memory B cells 1 week after immunization. On the contrary, in mice immunized with Alum/TLR7 the number of the memory B cells is higher 2 weeks compared to 1 week after the immunization (**Figure 3B**). Therefore, while a decrease in the memory B cell number is observed in the passage from 1 week to 2 weeks after Alum immunization, at the same time an increase in the memory B cell number occurs after immunization with Alum/ TLR7 (**Figure 3B**). Consequently, immunization with Alum/ TLR7 leads to a larger memory compartment, compared to Alum, 2 weeks after immunization (**Figure 3B**). Ultimately, our data indicate that Alum/TLR7 promotes an enhancement of the expansion of the memory B cell compartment within draining LN by inducing a sustained proliferation of the activated antigenspecific B cells, which in turn differentiate into memory B cells, and/or a persistence of the memory B cells within the node.

## Intranodal Recruitment of Antigen-Specific B Cells

Evaluating the dimension of the whole LNs before processing them for flow cytometry analysis led us to observe that the draining LNs from Alum/TLR7 immunized mice appear larger than the other LNs (Figure S3 in Supplementary Material). Thus, we decided to count also the number of intranodal non-proliferated antigen-specific B cells and we found that only immunization with

Alum/TLR7 promotes a recruitment of naïve antigen-specific B cells into the draining LNs compared to the non-draining LNs (**Figure 4**). This enhancement of the intranodal antigen-specific naïve B cell number reflects the increase in the number of total lymphocytes in LNs of mice immunized with Alum/TLR7 (data not shown).

Figure 4 | Alum/toll-like receptor (TLR)7 promotes recruitment of non-proliferated naïve antigen-specific B cells within draining lymph nodes (LNs). Bar graphs reporting the number of naïve non-proliferated antigenspecific B cells within draining LNs (filled blue bars) and contralateral non-draining LNs (striped blue bars) from mice treated as indicated and collected 3 days (A), 1 week (B), and 2 weeks (C) after the immunization. Bar graphs plot the average results, with SD, of five independent experiments. Statistics: Mann–Whitney two-tailed test: \*\**P* < 0.01, \**P* < 0.05.

In conclusion, our data demonstrate that immunization with Alum/TLR7, a potentiated version of Alum adjuvant obtained by attaching a benzonaphthyridine compound TLR7 agonist to aluminiun hydroxyde, promotes, within the draining LN, both an increased expansion of memory B cells due to the intranodal sustained proliferation and/or accumulation of these cells and an increased recruitment of antigen-specific B cells.

### DISCUSSION

The development of the immunity against an infectious disease depends on the generation of B and T lymphocytes that specifically recognize epitopes of the antigens belonging to that particular pathogen responsible for the disease (24, 25). In this regard, it is definitely critical that B and T lymphocytes recognize certain specific antigens of the pathogen that play a key role in the mechanism of the infection and/or the disease determined by that specific pathogen (24, 25). Thus, the development of the immunity against a pathogen is entirely based on the specificity of the recognition of that particular pathogen (24, 25). To this end, the humoral immune response is an essential part of the immunity against both viral and bacterial infections because antibodies can block tissue colonization and cell entry, neutralize toxins or pathogen enzymes, activate complement, which kills bacterial cells, and facilitate phagocytosis of the pathogen (24, 25). These are all fundamental weapons that can neutralize the attack of a pathogen and lead to its elimination from the body (24, 25). However, despite the body's capacity of mounting an immune response against an infection, in most cases the pathogens are able to escape the surveillance of the immune system leading to the disease, while the immunity develops during convalescence from the illness (24, 25).

Vaccination aims to generate immunity without contracting the disease, which is absolutely critical for fatal or severely disabling diseases (1, 2, 24, 25, 30). Over the centuries, approaches to the development and production of vaccines have undergone drastic changes (30, 31). From the original and primitive "variolation," the introduction of modern technologies allowed the development of novel vaccines enhancing their efficacy, improving their safety, facilitating their manufacturing and licensure, and, above all, making it possible the immunization against vaccine-preventable diseases for which a vaccine was not previously available (30, 31). One of these key modern technologies is the adjuvant technology (1, 30, 31). Some modern vaccines are based on the identification of pathogen antigens that can be critical targets for the development of an effective immunity (mainly humoral immunity) Figure 4 | Continued to use these pathogen subunits to redirect the body's immune response against these few selected key target antigens (1, 3–6, 30, 31). In this way, an appropriate humoral immune response can be achieved to obtain a long-term and robust immunity, consequently enhancing the effectiveness of the immunization (1, 3–6, 30, 31). However, these target antigens can be poorly immunogenic if administered alone, thus vaccine adjuvants are particularly important component of a subunit vaccine (1, 3–6, 30, 31). In addition, the enhancement of the effectiveness of the immunity determined by adjuvants can provide a broader protective efficacy stimulating cross-protective antibodies that can recognize antigenic variants of the vaccine antigen expressed by different strains of that pathogen targeted by the vaccination (1, 3, 5, 32).

Vaccine adjuvant Alum/TLR7, currently in phase I clinical development, was shown to be a more effective adjuvant than Alum at enhancing a protective antibody response in preclinical animal models, using several target pathogens (10–12, 14). Particularly interesting is the fact that Alum/TLR7 promotes a significantly higher persistence of the effective antibody titers against *N. meningitidis* C until 8 months after the second immunization in mice (12). However, the generation of the humoral immunity is based also on the development of a memory compartment and not only on the production of antigen-specific antibodies (15, 17–20, 24, 25). The immunological memory, in fact, provides the immune system with the capacity to mount a faster, powerful, and effective immune response against a specific pathogen once the body is exposed again to that specific pathogen infection (15, 17–20, 24, 25). Moreover, a stronger effective antibody response against A, C, W, Y polysaccharide antigens of *N. meningitidis* has been observed after a second immunization with a ACWY tetravalent glycoconjugate vaccine formulated with Alum/TLR7, compared to formulation with Alum, suggesting that an increased memory response is induced by Alum/TLR7 (12). Despite this finding, the effect of this new adjuvant on the development of memory B cell compartment was not assessed in previous studies. In this light, our current study complements other investigations, which compared the adjuvant potential of Alum/TLR7 to that of Alum.

We demonstrated that the attachment of a synthetic benzonaphthyridine compound TLR7 agonist to Alum, not only leads to an enhancement of the antigen-specific antibody response but also significantly improves the capacity of this adjuvant to induce an expansion of the memory B cell compartment. Interestingly, this expansion of the memory B cells may be determined by both a sustained proliferation and/or a persistence of memory B cells within the draining LN. Indeed, when immunizing with Alum/TLR7, the number of antigen-specific memory B cells increases between 1 week and 2 weeks after the immunization. On the contrary, in case of immunization with Alum, the number of memory B cells 2 weeks after the immunization is lower than the number of memory B cells 1 week after the immunization. Surely, a portion of antigen-specific B cells entered in the GC differentiates into PCs that leave the LN, but this occurs when immunizing with both adjuvants (15, 16, 18–20, 22, 23). Thus, the decrease in the memory B cell number from 1 week to 2 weeks observed after immunization with Alum cannot be explained with the development of PCs from GC, because this event should occur also after immunization with Alum/TLR7, which moreover induces a higher antibody response compared to Alum. Thus, we can conclude that Alum/TLR7 promotes the observed intranodal expansion of the memory B cell compartment by a sustained proliferation and/or an accumulation of the memory B cells within the node. Our finding is particularly interesting, considering that the activation of memory B cells, that occurs during the secondary immunization of a vaccine course, is mostly based on the presentation of the intact antigen to the B cell receptor by follicular dendritic cells (DCs) (15–22, 26). For this reason, the increased number of memory B cells within the draining LN might facilitate the response to the re-exposure of the body to the same antigen injected in the same site, which is absolutely relevant for immunizations.

The expansion of the memory B cell compartment within the draining LN that we have found after immunization with Alum/ TLR7 is certainly due to the generation of activation or survival signals for antigen-engaged B cells during the GC reaction or signals that induce memory B cell retention within the LN (15–23). However, the immunological and molecular mechanisms underlying this phenomenon are difficult to hypothesize because the overall picture of memory B cell formation from the GC reaction, memory B cell circulation throughout the node and memory B cell lifespan, is very complex, still controversial and has yet to be elucidated in details (15–23). Several soluble or surface molecules might be involved, such as BAFF, Fas ligand, CD40 ligand, and IL-21, that can be expressed by different cell types, like DCs, follicular DCs, follicular helper T cells, or different macrophage subtypes (15–23). The TLR7 agonist compound may also play a critical role because TLR7 is expressed by B cells and the TLR engagement participates to the modulation of B cell responses (33, 34). Thus, the key signals that sustain the proliferation and/ or promote the intranodal accumulation of memory B cells after immunization with Alum/TLR7 might be delivered to B cells in two non-mutually exclusive ways: (1) directly by TLR7 stimulation on B cells and (2) indirectly by other cells that can, in turn, be activated either *via* TLR7 engagement (such as DCs) or *via* TLR7 stimulated DCs (such as follicular helper T cells) (3, 8, 15–23, 33–35). Among all possible different hypotheses, we consider particularly intriguing to evaluate whether DCs may participate directly to this induced expansion and/or persistence of memory B cells within the draining LN determined by immunization with Alum/TLR7. In fact, we would like to emphasize that immunization with Alum/TLR7 promotes recruitment of cells into the draining LN and many of them can be DCs migrated from the injection site. DCs can express and release the activating/survival signal for B cells BAFF (36, 37) and can be stimulated by TLR7 engagement (3, 8, 35). In addition, it has been discovered that the capture of influenza virus by DCs within the medullary region of the draining LN is necessary to produce an intranodal development of antibody-secreting cells (38), which is particularly intriguing for the hypothesis that DCs may have a similar role in the development of memory B cells.

We also observed that immunization with Alum/TLR7 induces a recruitment of antigen-specific B cells, into the draining LNs. Consequently, an additional mechanism to explain the persistence of the memory B cells within the draining LN as well as the result of the expansion of the intranodal memory B cell compartment could be the higher recruitment of the antigen-specific B cells into the LN. However, based on this interpretation, it is challenging the observation that 1 week after the immunization the number of memory B cells between mice immunized with Alum and Alum/TLR7 is comparable, despite Alum/TLR7 promotes a higher recruitment of antigenspecific B cells into the draining LN. Certainly, the increased recruitment of naïve B cells into the draining LN may feed the GC reaction and consequently sustain the generation of both PCs and memory B cells. However, whatever is the case, an important topic of future studies should be to explore the mechanisms leading to this induced cell recruitment into the draining LN and if recruited cells may participate to the delivery of signals to antigen-specific B cells that sustain their proliferation and/or promote their accumulation within the draining LN. Additionally, to investigate whether or not the expansion of the memory B cells on Alum/TLR7 treated mice may be entirely dependent on the TLR7 engagement on B cells, the same set of experiments presented in this study should be repeated transferring GP1-LCMV-specific B cells into TLR7 knockout mice.

Further studies are then needed to answer all these questions, which are very important to shed a light on the mode of action of Alum/TLR7 and obtain a deeper knowledge on how this adjuvant works.

Whatever are the molecular and cellular mechanisms underlying this phenomenon, we found, for the first time, that the attachment of a TLR7 agonist benzonaphthyridine synthetic compound to Alum significantly enhances not only the antibody response but also the expansion of the memory B cell compartment within the draining LN. Therefore, Alum/TLR7 adjuvant is able to sustain a more robust humoral immunity compared to Alum.

Alum adjuvant is the oldest and most used adjuvant for human immunization, particularly for pediatric vaccines (1, 2, 5, 6). The safety profile and the efficacy of Alum adjuvant are absolutely well documented by hundreds of millions vaccinations administered every year from decades (1, 2, 5, 6). However, the World Health Organization recommends the continuous progress of research and development aimed at the release of novel vaccines against vaccine-preventable diseases that are still a threat for the mankind or for some susceptible populations (1, 2). Generally, it refers to diseases for which effective vaccines either do not exist, such as malaria, AIDS, and bacterial diarrhea, or should be improved, such as tuberculosis and plague, or can be improved, such as pertussis (1, 2). Particularly, World Health Organization set a clear priority for the development of effective vaccines against malaria, tuberculosis, and AIDS (1, 2). In addition, many other vaccines for elderly or travelers can be developed as well as the continuous surveillance of meningococcal meningitis, particularly in the sub-Saharan belt, could envisage the development of new more effective meningococcal subunit vaccines (1, 2, 31).

Under this light, the release of a novel and more potent Alumbased adjuvant with a low reactogenic profile may be a strong hope for the development of new effective vaccines to combat vaccine-preventable diseases that are still serious concerns for public health globally and then to improve the health of the mankind to have a more equitable world.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the European directive 2010/63/UE and the Italian law DL 26/14. The protocol was approved by the local GSK Animal Welfare Body.

# AUTHOR CONTRIBUTIONS

HV: conducted the experiments, analyzed and interpreted the data, prepared the figures, critically revised the manuscript, approved final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work. BB: prepared formulations, critically revised the manuscript, approved the final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work. SS: bred KL25 mice, critically revised the manuscript, approved the final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work. MI: contributed to design of experiments and to interpretation of the data, critically revised the manuscript, approved the final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work. UD: contributed to design of experiments and interpretation of the data, critically revised the manuscript, approved the final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work. DP: conceived the work, designed research and experiments, analyzed and interpreted data, wrote the manuscript, approved the final version of the manuscript, and agreed to be accountable for the integrity and accuracy of the work.

# ACKNOWLEDGMENTS

We thank Marco Tortoli for mouse treatments, Silvia Mancianti for assistance to preparation of formulations, Simona Tavarini and Chiara Sammicheli for assistance to flow cytometry, and Pietro Di Lucia for preparation of purified GP1-LCMV:Fc antigen. This work was supported by (A) People Programme (Marie Sklodowska-Curie Actions) of the European Union's Seventh Framework Programme FP7/2007-2013 under the REA Grant agreement 317057 HOMIN-ITN and (B) European Commission through contract FP7-HEALTH-2011.1.4-4-280873 (ADITEC). This work was also supported in part by CTN01\_00177\_962865 grant (Medintech) from Ministero dell'Università e delle Ricerca (MIUR).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00641/ full#supplementary-material.

Figure S1 | Alum/TLR7 displays a superior ability in inducing antibody response against glycoprotein 1 of lymphocytic choriomeningitis virus (GP1-LCMV), compared to Alum. Mice (10 per treatment) were immunized twice, with 4 weeks interval between the first and the second immunization, using GP1-LCMV alone (buffer) or formulated with Alum or Alum/TLR7. Mouse sera were collected before the first immunization (pre-immune), 2 weeks after the first immunization (post 1) and 2 weeks after the second immunization (post 2). GP1-LCMV-specific IgG antibody titers were measured by ELISA. Results of two independent experiments (A,B) are reported. Values of antibody titers for each mouse in each immunization group are plotted as black dot. Statistics: Mann– Whitney two-tailed test, \*\*\*\**p* < 0.0001, \*\**p* < 0.01, \**p* < 0.05.

Figure S2 | Flow cytometry gating strategy to identify proliferated and non-proliferated antigen-specific B cells. Glycoprotein 1 of lymphocytic choriomeningitis virus (GP1-LCMV) specific B cells purified from KL25 transgenic mice were labeled with CFSE and transferred into recipient C57BL/6 syngeneic mice, containing a different CD45 allele. The day after, adoptively transferred mice were immunized in one leg with GP1-LCMV alone or

# REFERENCES


formulated with Alum or Alum/TLR7. Formulation buffer treated mice were used as negative control. After immunization, draining LNs, contralateral nondraining LNs, and spleens were collected, enzymatically digested, and analyzed by flow cytometry to identify antigen-specific B cells. Representative flow cytometry dot plots to show the gating strategy are reported. (A) Morphology to identify lymphocytes. (B) Live/dead staining to identify living lymphocytes. (C) Morphology to exclude doublets. (D) Antigen-specific B cells, identified as positive for CD45.1 (KL25 donor cells) and negative for CD45.2 (C57BL/6 recipient cells). (E) Identification of proliferated (CFSElow/negative) and non-proliferated (CFSEhigh) antigen-specific B cells by flow cytometry histogram.

Figure S3 | Popliteal draining lymph nodes (LNs) from mice immunized with Alum/TLR7 appear generally bigger than the others. Pictures of the popliteal draining LNs and popliteal contralateral non-draining LNs collected 3 days (A), 1 week (B), and 2 weeks (C) after the treatment, that compare their dimensions. Results of one experiment out of five independent experiments are shown.


**Conflict of Interest Statement:** This work was sponsored by Novartis Vaccines; in March 2015, the Novartis non-influenza Vaccine business was acquired by the GSK group of companies. The sponsor was involved in all stages of the study conduct and analysis. Alum/TLR7 adjuvant is property of GSK group of companies. At the time of the study, HV was a GSK contingent worker and HOMIN-ITN Ph.D. student, whereas SS and MI had a collaboration agreement with GSK Vaccines. BB, UD, and DP are employees of the GSK group of companies. BB and UD reports ownership of GSK shares.

*Copyright © 2018 Vo, Baudner, Sammicheli, Iannacone, D'Oro and Piccioli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Lorenza Tulli1 , Francesca Cattaneo2 , Juliette Vinot1 , Cosima T. Baldari <sup>2</sup> and Ugo D'Oro1 \**

*1GSK Vaccines, Siena, Italy, 2Department of Life Sciences, University of Siena, Siena, Italy*

### *Edited by:*

*Peter Andersen, State Serum Institute (SSI), Denmark*

### *Reviewed by:*

*Ali M. Harandi, University of Gothenburg, Sweden Carl G. Feng, University of Sydney, Australia*

> *\*Correspondence: Ugo D'Oro ugo.x.doro@gsk.com*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 06 February 2018 Published: 01 March 2018*

### *Citation:*

*Tulli L, Cattaneo F, Vinot J, Baldari CT and D'Oro U (2018) Src Family Kinases Regulate Interferon Regulatory Factor 1 K63 Ubiquitination following Activation by TLR7/8 Vaccine Adjuvant in Human Monocytes and B Cells. Front. Immunol. 9:330. doi: 10.3389/fimmu.2018.00330*

Toll-like receptors (TLRs) play a key role in the activation of innate immune cells, in which their engagement leads to production of cytokines and co-stimulatory molecules. TLRs signaling requires recruitment of toll/IL-1R (TIR) domain-containing adaptors, such as MyD88 and/or TRIF, and leads to activation of several transcription factors, such as NF-κB, the AP1 complex, and various members of the interferon regulatory factor (IRF) family, which in turn results in triggering of several cellular functions associated with these receptors. A role for Src family kinases (SFKs) in this signaling pathway has also been established. Our work and that of others have shown that this type of kinases is activated following engagement of several TLRs, and that this event is essential for the initiation of specific downstream cellular response. In particular, we have previously demonstrated that activation of SFKs is required for balanced production of pro-inflammatory cytokines by monocyte-derived dendritic cells after stimulation with R848, an agonist of human TLRs 7/8. We also showed that TLR7/8 triggering leads to an increase in interferon regulatory factor 1 (IRF-1) protein levels and that this effect is abolished by inhibition of SFKs, suggesting a critical role of these kinases in IRF-1 regulation. In this study, we first confirmed the key role of SFKs in TLR7/8 signaling for cytokine production and accumulation of IRF-1 protein in monocytes and in B lymphocytes, two other type of antigen-presenting cells. Then, we demonstrate that TLR7 triggering leads to an increase of K63-linked ubiquitination of IRF-1, which is prevented by SFKs inhibition, suggesting a key role of these kinases in posttranslational regulation of IRF-1 in the immune cells. In order to understand the mechanism that links SFKs activation to IRF-1 K63-linked ubiquitination, we examined SFKs and IRF-1 possible interactors and proved that activation of SFKs is necessary for their interaction with TNFR-associated factor 6 (TRAF6) and promotes the recruitment of both cIAP2 and IRF-1 by TRAF6. Collectively, our data demonstrate that TLR7/8 engagement leads to the formation of a complex that allows the interaction of cIAP2 and IRF-1 resulting in IRF-1 K63-linked ubiquitination, and that active SFKs are required for this process.

Keywords: toll-like receptor, innate immunity, interferon regulatory factor 1, ubiquitination, TNF receptor associated factor 6, Src family kinases

### INTRODUCTION

The innate immune system represents the first defense mechanism engaged by the host organism against pathogen infections and is also required to activate adaptive immune responses. A central role in coupling innate and adaptive immunity is played by toll-like receptors (TLRs) (1–4). Following recognition of highly conserved pathogen-associated molecular patterns (5, 6), TLRs activate a signaling cascade that in turn leads to the production of pro-inflammatory cytokines and co-stimulatory molecules, which are required for the response to pathogens (4, 7). Due to their critical role in the activation of the innate immune response TLRs are preferential targets of new vaccine adjuvants based on small molecules (8).

TLR7 and TLR8 belong to a subfamily of endosomal receptors that recognize single-stranded RNAs from viruses as well as endogenous nucleic acids released in the context of pathogenic events (9, 10). TLR7 has been shown to be a good target for a recently described vaccine adjuvant (11, 12). Engagement of these two TLRs by appropriate agonists triggers an intracellular signaling pathway that involves recruitment of the adaptor protein Myeloid differentiation primary response 88 (MyD88), which in turn binds interleukin 1 (IL-1) receptor-associated kinase family of protein kinases. Activation of these serine/threonine kinases leads, through TNFR-associated factor 6 (TRAF6), to activation of MAP kinases and nuclear translocation of the transcription factor NF-κB, which are required for induction of inflammatory cytokines. Members of the interferon regulatory factor (IRF) family of transcription factors are also induced by intracellular TLRs such as TLR7 and TLR8, resulting in triggering of other cellular functions associated with these receptors. TLR7 stimulation in plasmacytoid DC activates IRF-7 that in turn leads to IFNα production (13), while in conventional DC, TLR7 engagement is associated to interferon regulatory factor (IRF-1) induction, which in these cells controls cytokine gene expression (14–16). Src family kinases (SFKs) were also shown to participate in TLRs signaling (17–24). Activity of these kinases is finely regulated by a balance of phosphorylation and dephosphorylation events (25). Phosphorylation of their carboxyl-terminal tyrosine residue (Tyr-530 in human c-Src), mainly by C-terminal Src kinase, leads to an intramolecular interaction of this phosphorylated residue with the Src homology (SH) 2 domain, resulting in an enzymatically inactive closed conformation of SFKs. Dephosphorylation of the carboxyl-terminal tyrosine residue by several phosphatases or displacement of the intramolecular SH2 domain by another protein with a higher affinity SH2 domain opens up the kinase domain, which becomes catalytically active. This event allows intermolecular autophosphorylation of a tyrosine residue in the activation loop (Tyr-419 in human c-Src), which locks the catalytic pocket into an open conformation and is required for maximum activation of these kinases. PP2 is a small molecule inhibitor of SFKs, which binds specifically these kinases, preventing the auto-phosphorylation in the activation loop that is necessary for the full activation of SFKs (26). We previously demonstrated that pharmacological inhibition of SFKs with PP2 in human monocyte-derived dendritic cells (MoDCs) impaired TLR7/8-mediated release of several pro-inflammatory cytokines by interfering with the accumulation of IRF-1, thus suggesting an involvement of SFKs in IRF-1 regulation and TLR7/8 signaling (16).

Interferon regulatory factor 1 is the prototype member of a family of nine transcription factors (27, 28) and is expressed in a variety of cells in which plays a crucial role in promoting the expression of type I IFN genes following viral infection (29–31). Similar to other transcription factors, IRF-1 is tightly regulated at both transcriptional and posttranslational levels (32–35). In particular, IRF-1 can undergo K48-ubiquitination to be degraded *via* the proteasome pathway (36), or K63-ubiquitination through the recruitment of TRAF6 and cIAP2 to become activated following IL-1R stimulation (37).

Here, we show that a similar signaling pathway involving TLRs and SFKs controls IRF-1 expression and cytokine production in two other important classes of antigen-presenting cells, namely monocytes and B-lymphocytes, which are key target for vaccine adjuvants. Moreover, we provide evidence that SFKs control the TLR7/8-dependent release of pro-inflammatory cytokines by monocytes and B-lymphocytes by promoting K63-linked ubiquitination of IRF-1. Finally, we demonstrate a crucial role of SFKs in binding and activating the ubiquitin ligase TRAF6 and that its inhibition impairs the formation of a complex with cIAP2 and IRF-1 that is a crucial step for IRF-1 K63 linked ubiquitination.

### MATERIALS AND METHODS

### Cell Cultures

Human embryonic kidney cells stably expressing human TLR7 (hTLR7-HEK293 cells) were cultured in DMEM containing 4.5 g/ml glucose, supplemented with 10% heat inactivated FBS, 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM glutamine, 5 µg/ml puromycin, 5 µg/ml blasticidin, and 2 mM glutamine. Human monocytic leukemia cell line THP-1 were cultured in RPMI 1640 containing 2.5 g/l glucose, supplemented with 10% heat inactivated FBS, 10 mM HEPES, 10 mM Sodium Pyruvate, 100 U/ml penicillin, and 100 µg/ml streptomycin.

PBMCs were isolated from buffy coats of healthy donors using Ficoll gradient. Human primary B cells were purified by negative selection using the RosetteSep B-cell enrichment Cocktail (StemCell Technologies) followed by density gradient centrifugation on Lympholite (Cedarlane Laboratories) as previously described (38). Monocytes were isolated from purified human PBMC using anti-CD14 magnetic beads (Miltenyi Biotec). Informed consent was obtained according to the Declaration of Helsinki.

Epstein–Barr virus (EBV)-immortalized B cell line and human primary B cells were cultured in RPMI 1640 (Sigma-Aldrich, The Woodlands, TX, USA) supplemented with 7.5% bovine calf serum at 37°C in a humidified atmosphere with 5% CO2.

### Cell Treatment and Cytokine Detection

Cells were treated with 20 µM PP2 (Sigma-Aldrich) or DMSO for 30 min then stimulated with R848 (10 µM) (Invivogen) overnight. Supernatants were collected and the amount of inflammatory cytokines was measured using Mesoscale Assay Human-Pro-inflammatory 7-spot (Meso Scale Discovery), following the manufacturer's instructions.

### Cell Transfection

hTLR7-HEK293 cells were seeded in 100 mm diameter culture dishes (3 × 106 cells/dish) and after 24 h were transfected using Lipofectamine 2000 (Invitrogen) with an expression plasmid encoding hemagglutinin (HA)-tagged K48-only or K63-only ubiquitin (kindly provided by Dr. Jonathan Ashwell, NCI, NIH, Bethesda, MD, USA). Four days post-transfection cells were treated as described above. THP-1 cells were seeded in 100 mm diameter culture dishes (6 × 106 cells/dish) and immediately transfected by Lipofectamine 2000 (Invitrogen) with an expression plasmid encoding HA-tagged Ubiquitin. After 24 h, cells were treated as described above.

# Immunoprecipitation and Immunoblot Analysis

Cells were treated with 20 µM PP2 at concentration or DMSO for 30 min prior to incubation with R848 (10 µM) for 2 h, then lysed with a buffer containing 150 mM NaCl, 20 mM Tris–HCl, Triton X-100 1%, 1 µg/ml pepstatin A, 1 µg/ml leupeptin, 1 µg/ml aproteinin, 200 µg/ml sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and 1 mg/ml of *N*-ethylmaleimide, for 5 min on ice. Lysates were centrifuged for 20 min at 21,000 × *g* at 4°C. Protein concentration was measured using the Bradford assay (Sigma-Aldrich). Proteins were separated by 10% SDS-PAGE electrophoresis using the NuPage Gel System, according to the manufacturer's instructions and transferred to nitrocellulose membranes for Western blot analysis. After blocking with PBS containing 0.05% Tween 20 (PBST) and 5% BSA (Sigma-Aldrich), proteins were detected with specific antibodies.

For immunoblot and immunoprecipitation analysis, we used the following antibodies: anti-IRF-1 (cat# 8478 Cell Signaling for immunoblot and cat# sc-497 Santa Cruz Biotechnology for immunoprecipitation); anti-TRAF6 (cat# 8028 Cell Signaling); anti-Src (cat# 2108 Cell Signaling); anti-p-SFKs (Tyr416) (2101 Cell Signaling); anti-p-Tyrosine (cat# 8954 Cell Signaling); anti-HA (cat# 2367 Cell Signaling); anti-cIAP2 (cat# ab32059 Abcam), anti-β-actin (cat# ab8227 Abcam); and HRP-conjugated secondary antibodies (DAKO).

The same filters were stripped and re-probed with an antiactin antibody (cat#mab1501 Millipore) for loading control. Immunoblots were captured for densitometry using ImageJ software, and each signal was normalized to its respective protein signal on the same blot.

For immunoprecipitations, cell lysates were incubated overnight at 4°C with the primary antibody crosslinked with the BS3 reagent to Dynabeads® Protein G Immunoprecipitation Kit (Life Technologies). Crosslinking avoid the co-elution of IgGs with the targed antigen. Immunoprecipitated complexes were eluted and analyzed by Western blot. To analyze TRAF6 tyrosine phosphorylation, lysates were immunoprecipitated with anti-TRAF6 antibody. The immune complexes immobilized on Dynabeads were then resuspended in 2% SDS and heat-treated (90°C for 5 min) to disrupt the interaction between TRAF6 and SFKs. The protein mixture was diluted in PBS to reduce the SDS to a final concentration of 0.01%, and re-immunoprecipitated with the same anti-TRAF6 antibodies. The re-immunoprecipitated samples were eluted and analyzed by western blot. To control for efficiency of immunoprecipitation and protein loading and transfer, the blots were stripped and re-probed with the antibody used for immunoprecipitation. Immunoblots were captured for quantitative densitometry using ImageJ software, and each signal was normalized to its respective protein signal on the same blot.

# RNA Purification and Quantitative RT-PCR

Total RNA was extracted from THP-1, human B cells and HEK-293 using the RNeasy Plus Mini Kit (Qiagen) and retrotranscribed using the iScript™ cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Three independent reverse transcription reactions were performed on each sample. Real-time quantitative PCR (qRT-PCR) was performed in triplicate on each cDNA in 96-well optical PCR plates (Sarstedt, Nümbrecht, Germany) using SSo Fast EvaGreenR SuperMix (Bio-Rad) and a CFX96 Real-Time system (Bio-Rad). Results were processed and analyzed using CFX Manager Version 1.5 software (Bio-Rad). Transcript levels were normalized to HPRT1 used as housekeeping control for all the reactions. The primers sequences are as follow: HPRT1 forward 5'-AGATGGTCAAGGTCGAAG-3' and reverse 5'-GTATTCATTATAGTCAAGGGCATAT-3'; IRF-1 forward 5'-CGTGGGACATCAACAAGGA-3' and reverse 5'-GTG GAAGCATCCGGTACACT-3'.

### Statistical Analysis

Mean values, SDs and Student's *t-*test (unpaired) were calculated using Microsoft Excel Version 14.0.0 software (Microsoft) (Redmont, WA, USA). A *P* < 0.05 was considered statistically significant.

# RESULTS

## Activation of SFKs Is Required for Inflammatory Cytokine Production in both Monocyte-Like Cells and Primary B Cells

We have previously demonstrated that SFKs activity is required for the release of pro-inflammatory cytokines by human Mo-DCs stimulated with the imidazoquinoline compound R848 (Resiquimod), an agonist for human TLR7 and TLR8 (16). To understand whether this mechanism is shared by other immune cells, which are also targets for a TLR7/8 agonist vaccine adjuvant, we extended our analysis to monocytes and B-lymphocytes.

Human monocytic THP-1 cells and human primary monocytes and B cells were stimulated for 16 h with R848 in the absence or in the presence of the SFKs-specific inhibitor PP2 and cell supernatants were collected and analyzed using the multiplex Meso Scale Discovery immunoassay to measure the levels of 7 inflammatory cytokines in the culture supernatants. As shown in **Figure 1**, we observed a strong induction in the release of 4 of the 7 proinflammatory cytokines analyzed, specifically TNFα, IL-6, IL-8, and IL-1β in THP-1 cells (panel A) and in primary B cell (panel B), while in response to R848 only TNFα, IL-6 and IL-1β were released by human primary monocytes (panel C). However, in

Figure 1 | Src family kinases inhibition results in decrease of pro-inflammatory cytokine production in response to TLR7/8 stimulation in monocytes and B cells. THP-1 (A) and primary human B cells (B) were pretreated with PP2 (20 µM) or DMSO alone for 30 min and stimulated with R848 (10 µM) for 16 h. Monocytes (C,D) were pretreated with PP2 (20 µM) or DMSO alone for 30 min and stimulated with R848 (1 µM) for 16 h. Cytokine levels in the supernatants were quantified by Meso Scale Discovery immunoassay. Cells treated with vehicle only (DMSO) were used as control (Ctr). (A,B) Values represent the mean ± SD of cytokine concentration (picogram per milliliter) in the culture supernatants (*n* ≥ 3), \*\*\**P* < 0.001, \*\**P* < 0.01, \**P* < 0.05. Results are representative at least of three independent experiments. (C,B) The histograms show the mean (from duplicate samples) ± SD of cytokine concentration (pg/ml) in the culture supernatants from two different donors respectively in panels (C,D) and are representative of four experiments performed on monocytes from four different donors.

all cases, cytokine production was largely reversed when SFKs were inhibited by PP2 pretreatment.

Hence, SFKs activation is a critical event for cytokine production also in human monocytes and B cells.

### SFKs Play a Key Role in the Accumulation of IRF-1 in THP-1 and B Cells

Analysis of the signaling pathways triggered in MoDCs by R848 identified IRF-1 as the main effector responsible for the transcription of pro-inflammatory genes. Activation of this transcription factor was directly correlated with the activation of SFKs (16).

To evaluate whether IRF-1 was similarly involved in the release of pro-inflammatory cytokines in THP-1 and B cells, we analyzed its expression and activation in these cells after R848 stimulation. As shown in **Figures 2A,C**, stimulation of THP-1 cells or primary B cells with R848 for 2 h resulted in IRF-1 accumulation (upper western blot and left histogram in **Figures 2A,C**), which directly correlated with an increase in the auto-phosphorylation of SFKs on the activator Tyr residue (Y419 in c-Src) (middle western blot and right histogram in **Figures 2A,C**), that indicates activation of this type of kinases. These effects were all inhibited by pretreatment with PP2, confirming the importance of SFKs activation in the up-regulation of this transcription factor.

We next measured IRF-1 mRNA levels by qRT-PCR in cells treated as described above to assess whether the changes in IRF-1 expression were dependent on transcriptional regulation of the respective gene by SFKs. The level of *IRF1* mRNA was increased following stimulation with R848 in THP1 cells (**Figure 2B**) and primary B cells (**Figure 2D**). Interestingly, no decrease in the transcript was observed in either cell types when cells were pre-treated with PP2, suggesting that, as previously shown for MoDCs, in these cells SFKs regulate IRF-1 expression posttranscriptionally The analysis was also extended to EBV-immortalized primary human B lymphocytes (EBV-B). Interestingly, no changes in the mRNA levels were detected in these cells treated with R848, either in the presence or absence of PP2 (Figure S1A in Supplementary Material). Nevertheless, similar to THP-1 and primary B cells, we observed an accumulation of IRF-1 protein in R848-treated EBV-B cells that was blocked by PP2 (Figure S1B in Supplementary Material).

Collectively, these results support the hypothesis that posttranslational modifications, in addition to transcription activation, lead to IRF-1 accumulation as a consequence of SFKs activation.

### SFKs Activation Is Essential for K63-Linked Ubiquitination of IRF-1

To elucidate the mechanism responsible for IRF-1 downregulation in PP2 treated cells, we asked whether SFKs might prevent IRF-1 degradation by promoting posttranslational modifications in the protein. In fact, it has been reported that IRF-1 can undergo several posttranslational modifications, including sumoylation

Figure 2 | Src family kinases are required for interferon regulatory factor 1 (IRF-1) protein accumulation in response to TLR7/8 stimulation in monocytes and B cells. (A,C) Immunoblot analysis of IRF-1 and p-SFKs in THP-1 (A) and primary B cells (C) pretreated with PP2 (20 µM) or DMSO alone and stimulated with R848 (10 µM) for 2 h. After stripping, filters were re-probed with an antibody against actin as loading control. Results are representative at least of three independent experiments. The histograms on the right of the blot show the results of the densitometric analysis on three independent experiments. Each signal was normalized to its respective actin signal on the same blot and expressed as fold induction (mean ± SD) compared to untreated sample. (B,D) qRT-PCR analysis of *lRF-1* mRNA in THP-1 (B) and human B cells (D) pretreated with PP2 (20 µM) or DMSO alone and stimulated with R848 (10 µM) for 2 h. The relative abundance of the gene transcripts was determined on triplicate samples from at least three independent experiments using the ΔΔCt method and is expressed as the normalized fold expression (mean ± SD) compared to untreated sample. \*\**P* < 0.01, \**P* < 0.05.

and ubiquitination (35). K48-linked ubiquitination of IRF-1 has been shown to target the protein to the proteasome for degradation (36) whereas K63-linked ubiquitination is essential for IRF-1 activity (37).

To investigate whether SFKs modulate IRF-1 ubiquitination upon TLR7/8 engagement, we used HEK293 cells stably expressing human TLR7 (hTLR7-HEK293) as a specific cell model for studying the TLR7 signaling cascade. As observed in immune cells, stimulation of these cells with R848 led to an accumulation of IRF-1 that was reversed by PP2 treatment that also resulted in the inhibition of SFKs phosphorylation (**Figure 3A**). To analyze IRF-1 ubitiquination, hTLR7-HEK293 cells were transiently transfected with plasmids coding for HA-tagged K63-only or HA-tagged K48-only ubiquitin mutants and cell lysates were subjected to immunoprecipitation using an anti-IRF-1 antibody. As shown in **Figure 3B**, R848 stimulation resulted in an increase in the level of K63-linked ubiquitination of IRF-1, which was drastically reduced by pretreatment with PP2. Conversely, K48 linked ubiquitinated IRF-1 accumulated when cells were preincubated with PP2 (**Figure 3C**). It should be noted that, to normalize the levels of immunoprecipitated IRF-1 and thus avoiding that differences in the levels of ubiquitination could depend on the different levels of IRF-1 in the immunoprecipitates, a limiting amount of antibody was used to perform the immunoprecipitation. This resulted in similar levels of immunoprecipitated IRF-1 in all samples (lower blots in **Figures 3B,C**). Similar results were observed in THP-1 cells transiently transfected with HA-tagged K63-only or K48-only ubiquitin mutants (Figure S2 in Supplementary Material). To confirm the specificity of these results, we immunoprecipitated the lysates of hTLR7-HEK293 cells transiently transfected with HA-tagged K63-only or HAtagged K48-only ubiquitin mutants with an anti-HA antibody and then we processed the immunoprecipitates for IRF-1 immunoblotting (Figure S3 in Supplementary Material). Also in this case TLR7 engagement increased the levels of IRF-1 in the K63 ubiquitinated pool, which were inhibited by PP2 treatment, while Src kinase inhibition resulted in an expansion of the K48 ubiquitinated form. Taken together, these findings indicate that SFKs play a crucial role in the regulation of the balance between accumulation and degradation of IRF-1.

### SFKs Activation Regulates the Interaction of IRF-1 with cIAP2 and TRAF6

TNFR-associated factor 6 is an ubiquitin ligase that catalyzes the K63-linked ubiquitination of several signaling molecules (39–42). Of note, it has been previously reported that TRAF6 recruits the E3 ligase cIAP2 and IRF-1 in response to IL-1, forming a complex that allows K63-linked ubiquitination of IRF-1 by cIAP2 (37). Moreover, Liu A. et al. have demonstrated a functional association of SFKs and TRAF6 during TLR4 signaling (43). To establish whether TRAF6 participates in SFKsdependent IRF-1 regulation following TLR7/8 activation, lysates from hTLR7-HEK293 cells stimulated with R848 in the presence or absence of PP2 were subjected to co-immunoprecipitation experiments to assess protein–protein interactions. Analysis of TRAF6-specific immunoprecipitates showed that this protein

Figure 3 | Src family kinases activation is essential for interferon regulatory factor 1 (IRF-1) K63-linked ubiquitination. (A) Immunoblot analysis of IRF-1 and pSFKs in hTLR7-HEK293 cells pretreated or not with PP2 (20 µM) for 30 min and stimulated with R848 (10 µM) for 2 h. After stripping, membranes were blotted with anti-actin antibody as loading control. The histograms on the right of the blots show the results of the densitometric analysis on three independent experiments. Each signal was normalized to its respective actin signal on the same blot and expressed as fold induction (mean ± SD) compared to untreated sample. (B,C) hTLR7-HEK293 cells transiently transfected with plasmids coding for hemagglutinin (HA)-tagged K63-only (B) or K48-only ubiquitin mutants (C) were pretreated or not with PP2 (20 µM) for 30 min and stimulated for 2 h with R848 (10 µM). Total cell lysates were immunoprecipitated using an anti-IRF-1 antibody, separated by SDS-PAGE, and immunoblotted using anti-HA antibody for ubiquitin detection. Membranes were stripped and re-probed with an anti-IRF-1 antibody as control (bottom panels). The histograms on the right of the blots show the results of the densitometric analysis on three independent experiments. Each signal was normalized to the signal of the protein used for the pull-down and expressed as fold induction (mean ± SD) compared to untreated sample. \*\*\**P* < 0.001, \*\**P* < 0.01, \**P* < 0.05.

forms a complex with IRF-1, Src, and cIAP2 at steady state as well as following R848 stimulation. However, treatment with PP2 impaired the binding of TRAF6 with Src and cIAP2 (**Figure 4A**), thus showing that activation of SFKs is required for a stable interaction within this complex. On the other hand, when IRF-1

immunoblot. Histograms on the right side of the blots show the quantification of the relative amount of the pulled-down proteins, calculated on three independent experiments. Each signal was normalized to the signal of the protein used for the pull-down and expressed as fold induction (mean ± SD) compared to untreated sample. \*\*\**P* < 0.001, \*\**P* < 0.01, \**P* < 0.05.

immunoprecipitates were evaluated, only TRAF6 and cIAP2 were found in the complex, while no Src could be detected either before or after TLR7 triggering (**Figure 4B**). The absence of Src in this complex could depend on the fact that the kinase directly interacts with TRAF6 and that this interaction is not preserved in the anti-IRF-1 immunocomplex. Nevertheless, inhibition of SFKs by PP2 reduced the amount of associated cIAP2. Western blot of Src co-immunoprecipitated molecules confirmed that the kinase activity is required for TRAF6 binding (**Figure 4C**). Collectively, these results provide evidence that under basal conditions TRAF6 is associated with IRF-1, cIAP2, and SFKs in a SFKs activity-dependent manner.

# TLR7 Engagement Leads to Src K63-Linked Ubiquitination and TRAF6 Tyrosine Phosphorylation

Src family kinases have been proposed to promote TRAF6 phosphorylation that in turn can modify SFKs by adding K63 linked ubiquitin chains (18, 43). Therefore, we tested whether this mechanism might also occur following TLR7/8 activation. Total lysates from hTLR7-HEK293 cells transiently transfected with plasmids coding for HA-tagged K63-only were subjected to immunoprecipitation experiments using an anti-pSFKs antibody and then probed by immunoblot with an anti-HA antibody. As shown in **Figure 5A**, we found that R848 stimulation increases the level of Src K63-linked ubiquitination, a modification which is completely abrogated by inhibition of kinase activity by PP2.

To understand whether in turn, TRAF6 might be regulated by SFKs, we immunoprecipitated TRAF6 from hTLR7-HEK293 cells stimulated with R848 and pretreated or not with PP2. Immunoblot analysis with an anti-phospho-tyrosine antibody showed that TRAF6 was phosphorylated upon TLR7/8 stimulation in a kinase activity-dependent manner. Blotting this immunoprecipitate with an antibody against the activator Tyr residue (Y419) of Src confirmed that the active kinase is bound to TRAF6 in a kinase activity-dependent manner (**Figure 5B**). To confirm that the tyrosine phosphoprotein detected in TRAF6 specific immunoprecipitates is indeed TRAF6 and not the coimmunoprecipitated and co-migrating tyrosine phosphorylated

immunoprecipitated with anti-TRAF6 and immunoblotted with antibodies against pSFKs, Src, and pTyr. (C) TRAF6-specific immunoprecipitates were eluted with 2% SDS in order to disrupt the TRAF6-SFKs interaction and after dilution with PBS re-immunoprecipitated with anti-TRAF6 antibody. Results are representative of three independent experiments. Histograms on the right side of the blots show the quantification of the pulled-down proteins, calculated at least on two independent experiments. Each signal was normalized to the signal of the protein used for the pull-down and expressed as fold induction (mean ± SD) compared to untreated sample. \*\*\**P* < 0.001, \*\**P* < 0.01.

SFKs, the immunoprecipitated samples were treated with 2% SDS to disrupt the TRAF6/pSFKs interaction and re-immunoprecipitated using an anti-TRAF6 antibody to purify TRAF6 from all coimmunoprecipitated proteins. Probing these samples with an anti-phospho-tyrosine antibody clearly proved that TRAF6 is phosphorylated upon stimulation of TLR7/8, and this event is completely blocked by PP2 (**Figure 5C**) indicating that TRAF6 phosphorylation is mediated by SFKs. Membranes were stripped and re-probed with an anti-pSFKs antibody to verify the absence of pSFKs (data not shown).

Collectively, our results provide evidence that following TLR7/8 engagement the association between SFKs and TRAF6 leads to K63-linked ubiquitination of SFKs, which in turn is responsible for the phosphorylation of a tyrosine residue on TRAF6. We have shown that these modifications are crucial for the formation of a complex between IRF-1, TRAF6 and cIAP2 and lead to IRF-1 K63-linked ubiquitination. This posttranslational modification has a crucial role in IRF-1 stabilization, which is required for the nuclear translocation of the transcription factor and the subsequent activation of its target genes.

### DISCUSSION

In this study, we demonstrate that SFKs play an essential role in K63-linked ubiquitination of IRF-1 in response to TLR7/8 triggering. We had previously reported that activation of SFKs is required for the release of pro-inflammatory cytokines and the accumulation of IRF-1 during TLR7/8 signaling in human MoDCs (16). Here, we first show that this mechanism is also shared by other cells of the immune system, namely monocytes and B-lymphocytes, which are potential target of TLR7/8 agonists vaccine adjuvants. Moreover, we demonstrate that Src kinase associates with TRAF6 in a complex, which also contains IRF-1 and cIAP2, and that following TLR7 triggering Src undergoes K63-linked ubiquitination and phosphorylates TRAF6. These events lead to K63-linked ubiquitination of IRF-1 in the complex, which is responsible for the increased levels of this transcription factor elicited by TLR7 engagement.

Interferon regulatory factors (IRFs) play a crucial role in the connection between innate and adaptive immunity by modulating the activation of immune cells triggered by TLRs (28). Indeed, TLR signaling leads to the transcriptional activation of a variety of genes as a result of the increase in the levels of specific transcription factors (34, 44). To enhance the availability of a particular transcription factor, the cells can adopt two mechanisms: one is to increase the rate of transcription of the gene encoding that particular protein, the other is by promoting its stability and activation. The latter mechanism is mainly achieved by specific posttranslational modifications, which may prevent protein degradation and increase protein half-life, or they may promote or inhibit protein–protein interactions allowing nuclear translocation and binding to the regulatory regions of the target genes. This type of mechanisms plays a critical role in the regulation of IRF-1 (45) in addition to its control at the transcriptional level (46). Under basal conditions, the rate of IRF-1 turnover is high (half-life 20–40 min), such that its steadystate levels are maintained very low. Indeed, this transcription factor contains ubiquitination and degradation signals within its C-terminal region, which lead to its constitutive ubiquitination and degradation by the proteasome pathway (34–36). It is known that the IRF-1 levels rapidly increase in response to stimuli such as viral infection, DNA damage, or TLR stimulation (16, 47, 48) due not only to its transcriptional activation but also to changes in its posttranslational modification that greatly increase the protein stability.

To understand how SFKs modulate IRF-1 accumulation following TLRs engagement, we first measured the mRNA levels of IRF-1 in cells stimulated with the TLR7/8 agonist R848 and pretreated or not with the SKFs inhibitor, PP2. Because no differences in IRF-1 transcript levels were found, we hypothesized that SFKs may be involved in the posttranslational regulation of IRF-1, in particular, in the ubiquitination of this transcription factor. Ubiquitin has a central role in many cellular functions, such as signal transduction, receptor downregulation, protein– protein interaction, protein transport and degradation and gene transcription, which may be regulated by different types of ubiquination leading to different ubiquitin chains (49–51). The best characterized type of ubiquitin chain is formed by several ubiquitins linked through lysine 48 (K48) and is typically associated with proteasome-mediated protein degradation (49, 51). Moreover, several proteins may be K63-linked ubiquitinated. This modification is known to be important in several cellular processes and, in particular, in the signal transduction of various activation pathways (49, 51). The activation and degradation of several IRFs, such as IRF5, IRF7 (39, 52, 53), and also IRF-1 (36, 37), are regulated by both types of ubiquitination. Using hTLR7- HEK293 cells transiently transfected with plasmids encoding ubiquitin mutants that form only K63 or K48-linked chains, we found that activation of SFKs results in an increase in K63-linked ubiquitination of IRF-1, which is reversed by kinase inhibition following PP2 treatment. On the contrary, the levels of K48-linked ubiquitinated IRF-1 were found to be increased by PP2 treatment. These results suggest that, upon TLR7 engagement, IRF-1 preferentially undergoes K63-linked ubiquitination and that this modification rescues it from degradation. Hence, consistent with our current understanding of IRF-1 regulation, the protein levels of IRF-1 are controlled posttranslationally by K48-linked ubiquitination and proteasomal degradation (36). Once SFKs are activated by TLR7/8, they trigger an increase in K63-linked ubiquitination of IRF-1, resulting in the activation and accumulation of the transcription factor. Interestingly, several studies proved that both types of ubiquitination target the same residues in the DNA-binding domain of IRF-1. In particular Narayan et al*.* (54) suggested the possibility that K63-ubiquitination may counteract the constitutive K48-ubiquitination of IRF-1. We may, therefore, hypothesize that the two events compete, such that cells promote K48- or K63-linked ubiquitination depending on the external conditions.

It has been reported that IL-1-induced K63-linked ubiquitination of IRF-1 is catalyzed by the E3 ligase cIAP2, which in turn is activated through its ubiquitination by TRAF6. Moreover, these three proteins have been shown to interact with each other (43). TRAF6 is one of the major signal transducers of the TNF receptor family (42, 55) and has long been known to

Figure 6 | Schematic illustration of SFKs mechanism of action in interferon regulatory factor 1 (IRF-1) K63-Ubiquitination. (A) Activation of the TLR7/8 pathway promotes the interaction between SFKs and TNFR-associated factor 6 (TRAF6) allowing the recruitment of both cIAP2 and IRF-1 by TRAF6 with the subsequent K63-ubiquitination and accumulation of IRF-1. (B) Preincubation with PP2 impairs the capacity of SFKs to interact with TRAF6 and the formation of the TRAF6-IRF1-cIAP2 complex, inhibiting K63-ubiquitination of IRF-1, which is, however, K48-ubiquitinylated and it is consequently degraded *via* proteasome.

participate in TLR signaling, ubiquitinating itself and several proteins, including some IRFs (37, 39, 41). cIAP2, instead, not only regulates caspases and hence apoptosis, but has also been implicated in the modulation of inflammatory signaling (50, 56). We tested whether these proteins are also recruited to K63 linked ubiquitinated IRF-1 during TLR7/8 signaling and the role played by SFKs in regulating these interactions. We found that cIAP2 is able to bind TRAF6 and IRF-1, and that these interactions are affected by PP2, suggesting that the activation of SFKs is an upstream essential step in this signaling cascade. While TRAF6 also interacts with IRF-1, this event was found to be independent of SFKs activation. Finally, stimulation with TLR7/8 agonists leads to reciprocal modification of TRAF6 and SFKs in a SFKs activity-dependent manner. Reciprocal binding and modification between TRAF6 and SFKs have been extensively described for different signaling pathways in multiple cells of the immune system, although the hierarchical position between these two molecules remains still unclear. In particular, it was reported that the interaction between TRAF6 and c-Src requires an active SFK kinase domain and that the polyproline sequence of TRAF6

### REFERENCES


and the Src-homology 3 (SH3) domain of Src are required for the binding between these two proteins to take place (32, 46–49). Collectively, our results highlight a new mechanism, whereby TRAF6 acts as a linker between SFKs and cIAP2-dependent IRF-1 K63-linked ubiquitination.

In conclusion, our finding that SFKs are required for K63 linked ubiquitination of IRF-1 and pro-inflammatory cytokine release demonstrates that SFKs activation is a key step in TLR7/8 signaling. In particular, as depicted in **Figure 6**, the activation of SFKs appears necessary for their own interaction with TRAF6 and for the recruitment of both cIAP2 and IRF-1 by TRAF6. These events lead to the formation of a complex that allows the interaction of cIAP2 and IRF-1 resulting in IRF-1 K63-linked ubiquitination and stabilization.

## AUTHOR CONTRIBUTIONS

LT, FC, and JV performed and analyzed the experiments; CB and UD designed research studies; LT, FC, CB, and UD wrote the paper. All authors reviewed the results and revised and approved the final version of the manuscript.

# ACKNOWLEDGMENTS

We thank Dr. Jonathan Ashwell (NCI, NIH, Bethesda, MD, USA) for the plasmids encoding HA-tagged ubiquitins and Dr. Antonio Leonardi (University of Naples, Italy) for critical review of the manuscript. We also thank Marianna Taccone and Susanna Aprea for technical help in performing experiments with human primary monocytes.

### FUNDING

This work was partially supported by a grant from MIUR (Medintech project CTN01\_00177\_962865) and European Commission through contract FP7-HEALTH-2011.1.4-4-280873 (ADITEC). LT and FC were supported by a fellowship from Regione Toscana POR-FSE 2007-2013. JV was supported by the People Program (Marie Sklodowska-Curie Actions) of the European Union's Seventh Framework Programme FP7/2007- 2013 under the REA Grant agreement 317057 HOMIN-ITN.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00330/ full#supplementary-material.


by the ubiquitin ligase CHIP. *J Biol Chem* (2011) 286(1):607–19. doi:10.1074/ jbc.M110.153122


**Conflict of Interest Statement:** JV is a former employee of the GSK group of companies. UD is an employee of the GSK group of companies and reports ownership of restricted GSK shares. LT was a recipient of a scholarship from GSK Vaccines S.r.l on a grant from MIUR (Medintech training project CTN01\_00177\_962865). All other authors declare that they have no conflicts of interest with the contents of this article.

*Copyright © 2018 Tulli, Cattaneo, Vinot, Baldari and D'Oro. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# High Antigen Dose Is Detrimental to Post-Exposure Vaccine Protection against Tuberculosis

*Rolf Billeskov1 \*† , Thomas Lindenstrøm1†, Joshua Woodworth1 , Cristina Vilaplana2 , Pere-Joan Cardona2 , Joseph P. Cassidy3 , Rasmus Mortensen1 , Else Marie Agger1 and Peter Andersen1 \**

*Medicine, University College Dublin, Belfield, Dublin, Ireland*

*1Department of Infectious Disease Immunology, Statens Serum Institut, Copenhagen, Denmark, 2Unitat de Tuberculosi Experimental, Institut per a la Investigació en Ciències de la Salut Germans Trias I Pujol, CIBER Enfermedades Respiratorias, Universitat Autònoma de Barcelona, Badalona, Barcelona, Spain, 3Veterinary Sciences Centre, School of Veterinary* 

### *Edited by:*

*Jeffrey K. Actor, University of Texas Health Science Center at Houston, United States*

### *Reviewed by:*

*Buka Samten, University of Texas at Tyler, United States Juraj Ivanyi, King's College London, United Kingdom*

### *\*Correspondence:*

*Rolf Billeskov rolf\_bs@yahoo.com; Peter Andersen pa@ssi.dk*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 20 December 2017 Published: 15 January 2018*

### *Citation:*

*Billeskov R, Lindenstrøm T, Woodworth J, Vilaplana C, Cardona P-J, Cassidy JP, Mortensen R, Agger EM and Andersen P (2018) High Antigen Dose Is Detrimental to Post-Exposure Vaccine Protection against Tuberculosis. Front. Immunol. 8:1973. doi: 10.3389/fimmu.2017.01973*

*Mycobacterium tuberculosis* (Mtb), the etiologic agent of tuberculosis (TB), causes 1.8M deaths annually. The current vaccine, BCG, has failed to eradicate TB leaving 25% of the world's population with latent Mtb infection (LTBI), and 5–10% of these people will reactivate and develop active TB. An efficient therapeutic vaccine targeting LTBI could have an enormous impact on global TB incidence, and could be an important aid in fighting multidrug resistance, which is increasing globally. Here we show in a mouse model using the H56 (Ag85B-ESAT-6-Rv2660) TB vaccine candidate that postexposure, but not preventive, vaccine protection requires low vaccine antigen doses for optimal protection. Loss of protection from high dose post-exposure vaccination was not associated with a loss of overall vaccine response magnitude, but rather with greater differentiation and lower functional avidity of vaccine-specific CD4 T cells. High vaccine antigen dose also led to a decreased ability of vaccine-specific CD4 T cells to home into the Mtb-infected lung parenchyma, a recently discovered important feature of T cell protection in mice. These results underscore the importance of T cell quality rather than magnitude in TB-vaccine protection, and the significant role that antigen dosing plays in vaccine-mediated protection.

Keywords: tuberculosis, post-exposure vaccination, vaccine dose, T cell quality, functional avidity, CAF01, H56, adjuvant

### INTRODUCTION

One fourth of the world's population is estimated to harbor latent *Mycobacterium tuberculosis* (Mtb) infection (LTBI), of which 5–10% will eventually develop active and transmittable tuberculosis (TB) (1, 2). This constitutes an enormous reservoir of potential TB disease, and developing a vaccine that can prevent reactivation in LTBI individuals would greatly impact the global TB burden (3). Furthermore, efficient therapeutic vaccination would be an essential tool in combating MDR/ XDR-TB cases with low responsiveness to second-line antibiotics. However, it has proven very difficult to achieve vaccine protection in post-exposure or therapeutic animal TB models (4, 5), and little is known about the mechanisms of protection in the existing studies (6–8).

Evidence suggests a protective immune response against infection with Mtb is derived mainly from IFN-γ-producing Th1 cells that activate infected macrophages, since CD4-deficient, IFN-γ- and inducible nitric oxide synthase (iNOS)-KO mice are highly susceptible to Mtb infection compared to wild-type strains (9–12). However, the last 10–20 years of research has shown that TB immunity is not as straightforward as previously understood, with some studies even suggesting that classical Th1-derived cytokines are not necessary for protection (13, 14).

Recently, it was shown that optimal protective capacity of T cells against Mtb infection relies on the ability of T cells to home into the lung parenchyma to make close contact with granulomaresident Mtb-infected host cells (15–18). Notably, vaccination of naïve mice with H56/CAF01 in a preventive mouse model of TB induced high numbers of protective CD4 T cells with these homing attributes (16).

The goal of the current study was to examine in greater detail the mechanisms behind protection of H56 (Ag85B-ESAT-6- Rv2660c) formulated in CAF01 in a post-exposure model of TB, in which T cells have already been primed by Mtb prior to vaccination, resulting in a more terminally differentiated and less protective phenotype compared to H56/CAF01 vaccination of naïve animals (16). In preventive models, we have shown that the induction and retention of central memory (Tcm)-like T cells co-producing IL-2 and TNF are essential for long-term protection (19–22), and furthermore that vaccine antigen dose significantly affects T cell functional avidity (23), differentiation status, and their subsequent protection against TB (24). In line with this, recent human data from dose-escalating studies with H56 and related protein hybrids have shown that vaccination with higher doses of antigen results in a higher degree of T cell differentiation. Importantly, this phenomenon is accentuated in Quantiferon (QFT) positive individuals (24–28). In mice, the biological relevance of these findings is underscored by multiple groups observing inferior TB protection of terminally differentiated CD4 T cells characterized by high KLRG1 expression with a decreased ability to home into the lung parenchyma in adoptive transfer studies (15, 29, 30).

In the present paper, we use the standardized mouse model of pre- and post-exposure vaccination with H56/CAF01 to demonstrate that post-exposure protection against TB is very sensitive to vaccine antigen dose, in contrast to the preventive setting. We find that higher vaccine antigen doses drive terminal differentiation, decreased functional avidity, and resulted in a non-protective state of vaccine-promoted T cells with a decreased ability to home into the lung parenchyma. We conclude that T cells in the context of an Mtb primed immune response are highly susceptible to terminal differentiation by excess stimulation of high vaccine doses.

### MATERIALS AND METHODS

### Animal Handling

Studies were performed with 6- to 8-week-old ♀CB6F1 mice (♂C57BL/6x♀Balb/c; Envigo, Denmark). Mice were housed in appropriate animal facilities at Statens Serum Institut, and experiments conducted in accordance with regulations of the Danish Ministry of Justice and animal protection committees by Danish Animal Experiments Inspectorate Permit 2009/561-1655, 2012- 15-2934-00272, 2014-15-2934-01065 and in compliance with EU Directive 2010/63 and the U.S. Association for Laboratory Animal Care recommendations for the care and use of laboratory animals. Animals were rested 1 week after arrival and before commencement of experiments.

### Bacteria

*M. tuberculosis* Erdman were grown at 37°C on Middlebrook 7H11 (BD Pharmingen) agar or in suspension in Sauton medium (BD Pharmingen) as previously described (31).

### Antigens and Vaccines

Production of the H56 fusion protein has been described previously (6). For restimulation of cell cultures, 18-mer peptides covering the antigen-sequence with 10 a.a. overlaps (2 µg/ml per peptide) or recombinant proteins (Ag85B and H56; 5 µg/ml) proteins produced in *E. coli* were used.

### Experimental Infections and the Preventive and Post-Exposure TB Models

The preventive TB model has been previously described (6). Briefly, naïve mice were immunized three times with the indicated doses of H56 in CAF01, and challenged with a low dose of 10–50 CFU virulent Mtb Erdman with an inhalation exposure chamber (Biaera-AeroMP, capacity = 80 mice per aerosol round). Mice from different aerosol rounds were randomized to take potential aerosol round variation into consideration. Six weeks after challenge, mice were euthanized, and numeration of bacterial counts (CFUs) in the lungs were determined by serial threefold dilutions of whole-organ homogenates on 7H11 medium and presented as log10 means of bacterial counts. The post-exposure model was previously described (6, 7). Briefly, mice were challenged by the aerosol route with ~10–50 CFU of Mtb Erdman/mouse (Glas-Col/Biaera-AeroMP, capacity = 30/80 mice per aerosol round, respectively) and randomized if necessary as described above. Six weeks after the infection, mice were subjected to an antibiotic chemotherapy treatment (100 mg/l rifabutin, 100 mg/l isoniazid) in drinking water from week 6 to 12 of infection. Mice were immunized three times 3 weeks apart starting in the second last week of chemotherapy (weeks 10, 13, and 16 post challenge; Figure S1A in Supplementary Material). Bacterial counts after therapy was undetectable in the lungs (detection limit 10 CFU), and protection was assessed by bacterial enumeration in lungs 25 weeks after treatment (week 37 after infection).

### Immunizations

Mice were immunized three times at 2-week (preventive) or 3-week intervals (post-exposure) s.c. with a range of doses (0.005–50 µg) of recombinant H56 formulated in CAF01 as described (19).

### Cell Cultures and Immunological Readouts

Spleen and lung lymphocytes were isolated at different time points after either vaccination and infection and stimulated *in vitro* for cytokine release analysis by intracellular cytokine stain (ICS) or ELISA of culture supernatants as previously described (20). In the flow cytometric analysis of T cell magnitude and polyfunctionality, we showed data from seven experiments: five of the nine individual experiments shown in **Figure 1C** in which identical analyses of pulmonary immune responses were assessed by ICS and flow cytometry 1 week after last vaccination, and that contained both 0.05 and 50 µg H56 doses within the same experiments (Exp. 1, 4, 6, 7, 9 in Figure S1B in Supplementary Material); and two additional experiments in which only immune responses at week 17, but no long-term protection at week 37, was assessed, to include a total of *n* = 21 mice/group.

### Normalization of CFU to Compare Protection

As previously described (7), we assessed the delta log10 protection by subtracting individual log10 CFUs from the average log10 CFU of the control group:

Log protection = Average log10 CFU (unvaccinated control group) − individual log10 CFU.

Medians were used in **Figure 1C** since not all post-exposure experimental log10 CFUs were normally distributed. Similar results were obtained using means and medians.

### Histopathological Assessment

Lungs were removed aseptically post mortem at week 37 after aerosol infection with Mtb (25 weeks after end of antibiotic treatment, at time of bacterial load assessment). The right lung cranial lobe of each mouse was fixed by immersion in 10% neutral-buffered formalin and processed for histological examination. Sections were stained using hematoxylin and eosin and were evaluated without prior knowledge of treatment group. Lesions were quantified using computer-aided histomorphometry (Palm®robo software, version 1.2.3; Palm Microlaser Technologies AG Ltd., Bernried, Germany) by a pathologist with previous experience of murine models of TB infection (**Figure 2**). Immunolabelling of iNOS within lesions was carried out following dewaxing and rehydration of sections and an epitope retrieval step where tissue slides were microwaved for 20 min in tri-sodium citrate solution (pH 6.0). Subsequently, treatment of sections with normal goat serum blocking solution (Vectastain ABC kit, Vector Laboratories, Inc., Burlingame, CA, USA) was performed to block endogenous peroxidase. The primary antibody (rabbit polyclonal anti-mouse iNOS/NOSII, Upstate, Lake Placid, NY, USA) was then applied for 1 h at 1/1,000 dilution followed by sequential application of biotinylated goat anti-rabbit IgG (Vector Laboratories) and ABC solution for 30 min, respectively, at room temperature (Vectastain ABC kit, Vector Laboratories). Visualization of target cells followed application of diaminobenzidine-tetrahydrochloride solution (Sigma-Aldrich, Steinheim, Germany) and a hematoxylin counterstain. To evaluate potential vaccine-mediated immune pathology shortly after vaccinations, lungs were removed post mortem 2 weeks after first, and 1 week after second and third (=last) vaccination. The entire left lung lobe of each mouse was fixed by immersion in 10% neutral-buffered formalin and processed for histological examination. The scale of pulmonary inflammation was analyzed using image analysis software (NIS-Elements D 3.0n Nikon Instruments Europe BV, Amstelveen, Netherlands) to determine the percentage lung tissue affected by two TB-pathology specialists (**Figure 2D**). This experiment only evaluated early pathology and ended 1 week after third vaccination (week 17 post-infection), hence, no protection was measured at week 37 post-infection and the experiment is therefore not referenced in Figure S1 in Supplementary Material.

FIGURE 1 | High dose H56 in post-exposure vaccination is detrimental to protection against *Mycobacterium tuberculosis* (Mtb). Total lung bacterial burden (log10 CFU) for unvaccinated control (open triangle) and 5 µg H56 in CAF01 vaccinated (closed diamond) mice in preventive (A) and post-exposure (B) vaccination TB protection models. (A) CB6F1 mice were vaccinated (gray arrows) s.c. three times and rested 6 weeks before low dose aerosol Mtb challenge and subsequent lung CFU determination at weeks 3, 6, and 12 post-infection. Symbols represent mean ± SEM of *n* = 6–8 mice/group. \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 by two-way ANOVA with Tukey's posttest for multiple comparisons at each time point. (B) CB6F1 mice were aerosol-infected with Mtb and treated with antibiotics (gray box; Abx) prior to three s.c. H56 vaccinations (gray arrows, weeks 10, 13, and 16) and subsequent CFU determination 37 weeks post-infection. Symbols represent mean ± SEM, *n* = 4–6 mice/grp (weeks 0–12) and 15/group (week 37). \**p* < 0.05 Mann–Whitney test at week 37. (C) Meta-analysis of Δlog10 protection values (see Materials and Methods for calculation) for the indicated H56 dose using combined data from three independent preventive (open circles, 6 weeks post Mtb) and nine independent post-exposure (filled squares, 37 weeks post Mtb) vaccination experiments. Symbols, median ± 95% CI of *n* = 21–129 mice/group. \*\**p* < 0.01, \*\*\*\**p* < 0.0001 by Kruskall–Wallis non-parametric test with Dunn's test for multiple comparisons (vaccine groups vs. controls, separate analyses for preventive/post-exposure model).

## Adoptive Transfer Experiment and ESAT-64–17 Tetramer

C57BL/6 mice (*n* = 7/group) were vaccinated three times with a high (50 µg) or lower (5 µg) dose of H56 in CAF01 in a modified post-exposure model, in which mice were infected by the aerosol route with 10–50 CFU of virulent Mtb Erdman. Mice were then subjected to an antibiotic chemotherapy treatment (100 mg/l rifabutin, 100 mg/l isoniazid) in drinking water from week 14–22 of infection. Mice were immunized three times 2 weeks apart starting in the fourth week of chemotherapy (weeks 18, 20, and 22 post challenge; schematic overview in **Figure 5A**). Spleen and inguinal lymph nodes draining the SOI from post-exposure vaccinated donor mice were isolated 5 days after last vaccination, and CD4 T cells purified by negative CD4 T cell separation (Stemcell kit and magnet). The CD4 T cells were separated to a purity of 91–93%. Enriched CD4 T cells from the high dose animal were then stained with 10 µmol Cell Proliferation Dye 450 (CPD450), and cells from low dose immunized animals stained with 5 µmol CPD670. CD4 T cells pooled from seven high and low dose vaccinated donor mice, respectively, were then pooled in a 1:1 ratio and a total of 1–1.2 × 107 CD4 T cells adoptively transferred into each of seven syngeneic recipient mice infected 4 weeks before with Mtb Erdman (10–50 CFU by aerosol route). 18 –20 h after adoptive transfers, recipient mice were euthanized 3 min after an i.v. injection of anti-CD45.2-FITC. Lung lymphocytes of recipient i.v. stained mice were then subjected to an ESAT-64–17 I-Ab tetramer magnetic bead enrichment (NIH tetramer core facility, Bethesda, MD, USA). To this end, isolated lung lymphocytes were stained with PE-conjugated I-Ab:ESAT-64–17 tetramer at 37°C for 30 min followed by anti PE-bead (Miltenyi, Cologne, Germany) staining at 4°C for an additional 30 min. Tetramer binding cells were then positively enriched by MACS separation on LS columns. ESAT-6 tetramer enriched cells were subsequently surface stained and analyzed for abundance and localization of ESAT-6 specific cells from each donor origin. The CD45.2-i.v. stain allowed for discrimination of vascular and parenchymal cells; CD45.2-stained cells were derived from the lung vasculature, and CD45.2 negative cells derived from the lung parenchyma (protected from the intravascular stain).

### Statistical Methods

Statistical difference in protective efficacy and immunogenicity (total cytokine) of vaccines was evaluated using non-parametric ANOVA (Kruskall–Wallis) and Dunn's posttest comparing all groups to the controls unless otherwise stated in the figure legends. Normal distribution of bacterial counts was evaluated by D'Agostino and Pearson's omnibus normality test. Comparisons of boolean cytokine populations between vaccine groups were tested with a two-way ANOVA and Tukey's posttest comparing all groups against each other. Comparison of functional avidity log10(EC50)-values was carried out using one-way ANOVA and Tukey's posttest. IL-2 ratio between high and low vaccine dose was compared by Wilcoxon signed rank test. Differences in the proportion of donor cells homing to the infected lung parenchyma (i.v. −ve) from 50 µg and 5 µg post-exposure immunized mice after transfer into the same recipients were analyzed by a paired Student's *t*-test. A value of *p* < 0.05 was considered significant. Prism version 7 software (GraphPad) was used for analyses.

### RESULTS

### High Antigen Dose Abrogates Vaccine Protection against TB in the Post-Exposure Setting

In the past decade, our laboratory has focused on understanding and developing efficacious anti-TB vaccines for use in acute, late, and post-exposure stages of Mtb infection. Here we tested the effect of H56 antigen dose on protection in a preventive model, in which mice were vaccinated prior to challenge, as well as in a postexposure relapse model in which mice are vaccinated with partial antibiotic clearance of bacilli (see Figure S1A in Supplementary Material and Section "Materials and Methods" for model details).

A growing body of evidence from several groups, including our own, has shown the important effects of antigen dose on T cell function and preventive protection against TB (24, 28, 32). To study the effect of vaccine dose on both preventive and postexposure protective immunity, we retrospectively analyzed data from multiple independent murine experiments conducted in both of these models during the past years in which various H56 doses in CAF01 were included.

**Figures 1A,B** show typical results of the preventive and post-exposure TB vaccine models with significant protection using the H56 vaccine candidate at a standard dose of 5 µg, in line with previous studies (6, 7). The result of the analysis showed that a wide dose range (0.05–50 µg) of H56 resulted in significant protection levels after preventive vaccination (*p* < 0.0001), whereas a much narrower range of lower doses (0.05–0.5 µg) were protective in the post-exposure model (*p* < 0.01–0.0001; **Figure 1C**). Importantly, the highest dose (50 µg) completely abolished protection in the post-exposure model and was significantly inferior to post-exposure protection obtained with 0.05 µg H56 [*p* < 0.0001 (significance not shown in graph); **Figure 1C**]. A meta-analysis format visualizing preventive and post-exposure Δlog10-protection with 95% CI of high/low H56 vaccine dose for individual experiments clearly showed the abolished protection in the 50 µg post-exposure group, also at the single experiment level (Figure S1B in Supplementary Material).

Overall, we observed a clear difference between the vaccine dose optimum in the preventive and post-exposure settings, where the highest H56 dose completely abolished protection post-exposure despite giving significant preventive protection. Therefore, from here on, we chose to focus on the most protective low dose of 0.05 µg H56, and the non-protective high dose of 50 µg H56 to further elucidate the dose related difference in post-exposure protection in more detail.

### Only Low Dose Vaccination Reduce TB-Related Pathology in the Post-Exposure Model

In support of the observation that the high H56 dose of 50 µg was unable to restrict pulmonary mycobacterial growth, a histopathological assessment of the lungs at the time of CFU assessment (week 37 p.i.) was performed from one of the experiments included in **Figure 1C**. The pulmonary bacterial burden from this experiment was in line with the combined data from **Figure 1C** and showed that a low dose of 0.05 µg H56 given post-exposure was protective, whereas a high dose of 50 µg H56 was not protective (**Figure 2A**). The histopathological analysis showed that only low dose H56 vaccination resulted in significantly smaller lesions and less total area affected by TB inflammation compared to non-vaccinated controls at the necropsy time point week 37 p.i. (*p* < 0.05; **Figures 2B,C**). The reduced pathology in the low dose group was not associated with an overall increased macrophage activation as assessed by iNOS staining in lungs of these mice, since comparable levels was detected in the lungs of the three groups at this late time point (**Figure 2B**). It could be speculated that a high H56 antigen dose given post-exposure potentially could lead to direct immune pathology in the pulmonary lesions due to excess T cell activation. We therefore monitored pulmonary vaccine-related immune pathology shortly (1–2 weeks) after each post-exposure vaccination. At these early time points (week 12–17 p.i.), we found no evidence of increased pathology for any H56 dose compared to non-vaccinated mice suggesting that the difference in the long-term pathological outcome was not the consequence of any immediate exaggerated effect of excessive T cell stimulation post vaccination (**Figure 2D**). In conclusion, a low dose of 0.05 µg H56 protected mice from TB in the postexposure model by restricting mycobacterial growth and limiting pulmonary TB-related pathology, and this protection was completely absent after administering a high dose (50 µg) of the same H56 vaccine antigen.

### High Antigen Dose Given Post-Exposure Leads to a More Effector-Driven T Cell Phenotype

Given the significant difference in vaccine efficacy between high and low H56 doses in the post-exposure experiments, we next evaluated the effect of vaccine dose on the T cell response. Multiple experiments were performed with similar outcomes, and we combined the experiments in **Figure 1C** that contained identical vaccine responses analyzed by ICS from both 0.05 and 50 µg H56 in the same experiment. Combining the Δlog10-protection for these five experiments confirmed that 0.05 µg H56 protected significantly better than 50 µg as expected (not shown). We then assessed the magnitude and phenotype of the H56-response in the lungs of mice sacrificed one week after vaccinations (week 17 p.i.). We first observed that high and low dose vaccination induced similar percentages of CD4 T cells producing any of the measured cytokines (IFN-γ, TNF, IL-2, and/or IL-17A) as shown by ICS of H56-stimulated lung lymphocytes (**Figure 3A**). Hence, the lack of protection in the 50 µg H56 dose group was not related to reduced magnitude of the overall immune response. Second, we performed boolean gating of the three canonical Th1 cytokines, IFN-γ, TNF, and IL-2 in the same dataset as shown in **Figure 3A**. This showed that high dose vaccination resulted in a subtle, yet consistent, increase in terminally differentiated effector CD4 T cells producing IFN-γ alone (**Figure 3B**, red pies), whereas low dose H56 vaccination led to more IL-2 producing memory-like T cells also expressing TNF with or without IFN-γ (**Figure 3B**, green and blue pies, respectively, *p* < 0.013–0.05). Since more differentiated T helper cells lose the ability to secrete IL-2, we calculated and compared the ratio of vaccine specific T cells producing IL-2 from the ICS data shown in **Figure 3B**. After *in vitro* H56 stimulation, the ratio of cytokine positive CD4 T cells unable to secrete IL-2 (IL-2<sup>−</sup>) to the IL-2-producing (IL-2<sup>+</sup>) CD4 T cells was significantly higher in the high 50 µg H56 dose group compared to the low 0.05 µg H56 dose group (*p* < 0.05; **Figure 3C**), indicating a greater degree of T cell differentiation in the high dose group.

In conclusion, the high H56 dose of 50 µg led to a similar magnitude of vaccine specific T cells as the low dose; however, the high dose resulted in a tendency toward more differentiated T cells with a lower capacity to produce IL-2.

### High Antigen Dose Given Post-Exposure Leads to a Decrease in T Cell Functional Avidity

We recently published that low vaccine antigen doses in liposomal CAF adjuvants increased the antigen sensitivity, termed functional avidity, of CD4 T cells (32), and with the very low bacterial loads in the post-exposure model, the ability of T cells to respond to low antigen levels is highly relevant. We therefore compared the functional avidity of vaccine specific CD4 T cells 1 week after high and low dose H56 post-exposure vaccination. Splenocytes from low dose immunized mice responded substantially better to lower concentrations of *in vitro* antigen stimulation compared to the high dose group as reflected by IFN-γ secretion in culture supernatants (**Figure 4A**). Furthermore, the concentration of antigen required to reach 50% of the maximum response (EC50) was significantly higher for the high (50 µg) dose group compared to the low (0.05 µg) dose group (*p* = 0.014; **Figure 4B**). We next analyzed the functional avidity of H56-specific CD4 T cells in the lungs. In two separate experiments, we observed that low dose vaccination also resulted in H56-specific pulmonary T cells of higher functional avidity compared to high dose vaccination. Thus, *in vitro* stimulation of lung lymphocytes from low dose vaccinated animals resulted in IFN-γ secretion at lower antigen concentrations compared to high (50 µg) dose vaccination (**Figure 4C**). The antigen concentration needed for 50% maximal activation (EC50) of lung lymphocytes was greater after high compared to low dose vaccination (*p* = 0.054 in Exp#1, and *p* = 0.027 in Exp#2; **Figure 4D**). Moreover, low dose vaccination led to a

two-way ANOVA and Tukey's multiple comparisons (the relative values were normally distributed). (C) From the data shown in panel (B), we calculated the relative ability of vaccine specific CD4 T cells to produce IL-2. An IL-2 ratio (%IL-2−/%IL-2+) was calculated for any CD4 T cell producing IFN-γ, TNF, or IL-2 after H56 stimulation—a higher ratio indicates lower IL-2 production (higher differentiation) of H56-specific T cells. Statistical difference between high and low dose vaccine groups was assessed Wilcoxon's signed rank test. \**p* < 0.05. Data combined from seven experiments (see Materials and Methods for experiment inclusion).

splenocytes (A) or lung cells (C) from control (open circle) and H56-vaccinated (0.05 µg closed square; 50 µg open triangle) mice were cultured for 3 days with varying concentrations of H56, and IFN-γ release in the culture supernatants was measured by ELISA. (A) Data points represent mean ± SD of triplicate cultures of pooled cells from three mice/group (from Exp. 4 in Figure S1B in Supplementary Material). (B) EC50 was calculated using data from (A); bars, mean ± SD per vaccine group. This experiment was repeated twice. (C,D) IFN-γ release and calculated functional avidity (EC50) of lung cells from two independent experiments in which subsequent week 37 protection was not assessed. (C) Data points represent mean ± SD of triplicate cultures pooled from five mice/group. (D) Bars indicate mean ± SD EC50 calculated from (C). Statistical differences between functional avidity of vaccine groups were assessed by a one-way ANOVA and Tukey's posttest for multiple comparisons. \**p* < 0.05, \*\**p* < 0.01. No further identical experiments with lung T cell avidity were performed.

higher "per cell" IFN-γ production and stronger ESAT-6 tetramer binding, as assessed by flow cytometry in lungs 1 week after vaccination (data not shown), further supporting higher avidity of these vaccine-specific T cells.

In summary, low dose post-exposure vaccination led to protective CD4 T cells with greater functional avidity.

# High Dose Post-Exposure Vaccination Reduces Lung Parenchymal Homing Ability of Vaccine-Specific T Cells

Recent research has shown that a hallmark of protective CD4 T helper cells in murine TB is the ability to home into the lung parenchyma and interact with infected cells, and that this ability is tightly linked to the differentiation state of the T cells (15, 29, 30). Hence, more differentiated T cells are trapped in the lung vasculature and do not enter the parenchyma, where the TB lesions are located. Given that high dose post-exposure vaccination led to increased CD4 T cell differentiation compared to low dose vaccination, we speculated whether the high vaccine dose could potentially also impact the ability of vaccine-primed CD4 T cells to home from the circulation and into the lung parenchyma.

To address this, we co-adoptively transferred donor CD4 T cells purified from mice receiving either low or high postexposure vaccine doses into Mtb-infected syngeneic recipient mice (**Figure 5A**). We used ESAT-6:MHC-II tetramers combined with intravital i.v. staining to subsequently track the lung homing capacity of the transferred donor cells. CD4 T cells were isolated by magnetic enrichment (negative selection) from spleens and inguinal lymph nodes (draining the SOI) from mice postexposure vaccinated with either a high or a low H56 dose. For the high dose, we used 50 µg H56, while for the low dose we chose 5 µg H56, since we have observed more variation in the magnitude of vaccine-specific T cells using 0.05 µg, and observed similar T cell differentiation after 0.05 and 5 µg H56. After CD4 T cell enrichment, donor cells from high and low dose post-exposure vaccinations were differentially stained with cell-tracking dyes in order to distinguish donor cells after co-adoptive transfer into the same infected mouse (**Figure 5A**). Roughly 20 h after transfer, the ability of I-Ab:ESAT-64–17 specific donor cells to home into the lung parenchyma was analyzed by magnetic enrichment of ESAT-6 tetramer binding cells combined with an intravascular staining technique, where fluorescent anti-CD45.2 administered intravenously (i.v.) prior to euthanasia, allowed separation of cells from the vascular (CD45.2 i.v.<sup>+</sup>) and parenchymal (CD45.2 i.v.<sup>−</sup>) lung compartments. ESAT-6-specific donor cells from high and low dose post-exposure vaccinated animals could clearly be distinguished from each other, and from the double negative endogenous recipient lung cells (**Figure 5B**, left panel). Importantly, donor cells deriving from low dose vaccinated animals showed an improved ability to home into the lung parenchyma of infected recipients, as seen by the increased percentage of CD45.2 i.v. negative cells in lungs of recipient mice (**Figure 5B**, middle vs. right panel). Moreover, a significantly higher proportion of donor cells

T cells (singlets > lymphocytes > live CD4 T cells > ESAT-6 Tet + ve). Double-negative cells, endogenous recipient ESAT-6-specific CD4 T cells. CPD450+ cells, ESAT-6-specific CD4 T cells (high dose vaccinated donors). CPD670+ cells, ESAT-6-specific CD4 T cells (low dose vaccinated donors). The right plots show representative CD45 i.v. stain of donor cells from low dose (mid panel) and high dose (right panel) vaccinated donors. Percentages represent CD45 i.v. negative cells, i.e., relative proportion of parenchymal CD4 T cells from each donor population. (C) Data points represent percentage CD45 i.v. negative (parenchymal) lung CD4 T cells derived from low dose (filled squares) and high dose (filled circles) vaccinated donors gated as shown in (B). Lines represent matched pairs (donor cells transferred into the same recipient). Statistical analysis was performed by a paired Student's *t*-test. \**p* < 0.05. This experiment was performed once.

from low dose post-exposure vaccinated mice could be observed within the lung parenchyma (CD45.2 i.v. negative) compared to donor cells from high dose vaccinated mice after transfer into the same recipient mice (*p* = 0.026; **Figure 5C**), thus directly demonstrating an improved lung parenchymal homing capacity. In conclusion, the high dose of 50 µg H56 led to a decrease in the ability of vaccine-specific T cells to home into the lung parenchyma associated with highly differentiated cells, which could potentially be a contributing factor in the loss of protection observed from high H56 dose post-exposure vaccination.

### DISCUSSION

In this study, we observed that a wide range of high and low vaccine doses were protective in a preventive murine TB model, but similar high vaccine doses were detrimental to post-exposure vaccine protection. Loss of protection after high dose vaccination was associated with a more effector-driven phenotype and decreased functional avidity, which further correlated with a decreased parenchymal homing ability of the vaccine specific T cells.

TCR-stimulation strength has been linked to the type of response since early studies performed in the 1980s and 1990s (33–35), showing the strength of stimuli could regulate Th1/2 polarization, and later also to play a role in induction of follicular helper T cells (36, 37), regulatory T cells [reviewed in Ref. (38)], as well as memory induction (39). The importance of vaccine antigen dosing and subsequent protection has been observed in a number of infectious diseases and cancers in animals and humans, with the common conclusion that higher doses lead to increased immune responses and improved protection (40–44). However, high antigen concentrations can accelerate T cell differentiation (45), and our data clearly show the importance of carefully titrating vaccine antigen dose, not only for a specific disease, but also for different stages of that specific disease to obtain optimal protection. Importantly, while a broad range of H56 vaccine antigen doses were protective in prophylactic vaccination, only a narrow range of lower doses were protective in post-exposure vaccination. The wide protective range (saturating at 103 -fold increase from lowest protective dose) of H56/CAF01 given preventively is in contrast to our previous observations using a similar vaccine and adjuvant, H4 (Ag85B-TB10.4) in IC31, which had a narrower range of both protection and immunogenicity (0.05–1 µg). Importantly, in the H4/IC31 study, protection correlated closely with the magnitude of vaccine response that decreased dramatically at doses higher than 1 µg of H4. This underlines the influence of the adjuvant system as very similar vaccine molecules given in the CAF01 vs. IC31 result in different immune responses and optimal doses, with CAF01 having a broader plateau for maximum responses than IC31 that sharply decline at all doses above 1 µg (46). The different H56 performance pre/post-exposure is consistent with several reports showing that effective preventive TB-vaccine candidates did not protect when given therapeutically, and therapeutic vaccination even aggravated disease in some cases (4, 5). The lower optimal protective vaccine dose in the post-exposure setting could reflect strong Mtb-priming of T cells to vaccine antigens, in turn leading to greater sensitivity of those T cells to overstimulation after vaccination.

Overstimulation after high dose vaccination could be particularly important with vaccines containing ESAT-6, since recent work in humans showed that CD4 T cells recognizing ESAT-6 are more sensitive to exhaustion due to the high pulmonary expression of this antigen compared to less highly expressed antigens such as Ag85B (18). In line with this, the ESAT-6 antigen itself has been shown *in vitro* to hold immune-regulatory properties, both anti-inflammatory (reduction of macrophage IL-12 release and T cell activation) as well as proinflammatory (macrophage IL-6 production and lung epithelial IL-8 production) as well as impacting the Th1/Th17 balance (47–49). Thus, high concentrations of ESAT-6 in a vaccine could potentially increase these effects. However, as ESAT-6 in the H56 molecule is flanked by Ag85B and Rv2660c on either side, it is unknown whether the H56-contained ESAT-6 exhibits any of these biological functions.

Although our study does not pinpoint one particular T cell deficiency as responsible for the lack of protective effect, it is striking that the high dose vaccine response have an overall impaired T cell quality as evidenced by reduced functional avidity, increased terminal T cell differentiation, and impaired ability to home into the infectious site in the parenchyma. The lower ability to home from the vasculature into the infected lung parenchyma may be the sole consequence of the more differentiated state of the T cells as suggested by recent studies (16, 29, 30). However, it may also relate to the lower functional avidity of the T cells that render them less sensitive to minute concentration of antigens in the infected sites. Hence, we suggest that low dose vaccination given post-exposure is sufficient to drive a protective immune response, whereas higher doses negatively impacts T cell quality and protective capacity. These results are highly relevant for clinical vaccine studies involving QFT+ individuals and suggest that antigen doses must be carefully investigated in clinical trials targeting different populations.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations and regulations of the Danish Ministry of Justice and animal protection committees by Danish Animal Experiments Inspectorate Permit 2009/561-1655, 2012-15-2934-00272, 2014- 15-2934-01065, and in compliance with EU Directive 2010/63 and the U.S. Association for Laboratory Animal Care recommendations for the care and use of laboratory animals. Protocols were approved by the SSI ACUC headed by DVM Kristin Engelhart.

# AUTHOR CONTRIBUTIONS

RB, TL, JW, EA, and PA conceived and designed the studies. RB, TL, and JW performed murine TB experiments and analyzed the data. CV, P-JC, and JC performed histopathological analysis. RB, RM, and PA drafted the manuscript. RB, TL, JW, EA, PA, and RM finalized the manuscript.

# ACKNOWLEDGMENTS

We thank for the excellent technical assistance of Katja Bøgebjerg Carlsen, Janne Rabech, Linda Christensen, Allan Lykke Hansen, Sharmila Subratheepam, Merete Henriksen, animal caretakers, and veterinarians at the SSI, as well as Joe Brady and Brian Cloak (UCD). We acknowledge the NIH Tetramer Core Facility for provision of I-Ab :ESAT-64–17 and corresponding negative control tetramer I-Ab :hCLIP. This work was funded by SSI core funds and Danish Research Council grant 12-132230 (RB), European Commission through contract FP7-HEALTH-2011.1.4-4-280873 (ADITEC) and TBVAC H2020-PHC-2014-2015-643383, and by the National Institute of Allergy and Infectious Diseases of the National Institute of Health under Award Number R01AI134246.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01973/ full#supplementary-material.

### REFERENCES


by macrophages through activation of STAT3. *Sci Rep* (2017) 7:40984. doi:10.1038/srep40984


**Conflict of Interest Statement:** PA and EA are co-inventors of patents regarding the use of H56 [#WO0011214 (Molecular differences between species of the *M. tuberculosis* complex); PA and EA, #WO2006136162 (Tuberculosis vaccines comprising antigens expressed during the latent infection phase); PA] and CAF01 [#WO0069458 (Adjuvant combinations for immunization composition and vaccines); PA]. All rights have been assigned to the SSI, a state owned not-for-profit research organization, and the authors' co-inventorship did not influence design of studies or preparation of the manuscript. There are no further patents, products in development, or marketed products to declare.

*Copyright © 2018 Billeskov, Lindenstrøm, Woodworth, Vilaplana, Cardona, Cassidy, Mortensen, Agger and Andersen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Protective effect of Vaccine Promoted neutralizing antibodies against the intracellular Pathogen *Chlamydia trachomatis*

*Anja Weinreich Olsen, Emma Kathrine Lorenzen, Ida Rosenkrands, Frank Follmann and Peter Andersen\**

*Chlamydia Vaccine Research, Department of Infectious Disease Immunology, Statens Serum Institut, Copenhagen, Denmark*

There is an unmet need for a vaccine to control *Chlamydia trachomatis* (*C.t.*) infections. We have recently designed a multivalent heterologous immuno-repeat 1 (Hirep1) vaccine construct based on major outer membrane protein variable domain (VD) 4 regions from *C.t.* serovars (Svs) D–F. Hirep1 administered in the Cationic Adjuvant Formulation no. 1 (CAF01) promoted neutralizing antibodies in concert with CD4+ T cells and protected against genital infection. In the current study, we examined the protective role of the antibody (Ab) response in detail. Mice were vaccinated with either Hirep1 or a vaccine construct based on a homologous multivalent construct of extended VD4's from SvF (extVD4F \*4), adjuvanted in CAF01. Hirep1 and extVD4F \*4 induced similar levels of Ab and cell-mediated immune responses but differed in the fine specificity of the B cell epitopes targeted in the VD4 region. Hirep1 induced a strong response toward a neutralizing epitope (LNPTIAG) and the importance of this epitope for neutralization was demonstrated by competitive inhibition with the corresponding peptide. Immunization with extVD4F \*4 skewed the response to a non-neutralizing epitope slightly upstream in the sequence. Vaccination with Hirep1 as opposed to extVD4F \*4 induced significant protection against infection in mice both in short- and long-term vaccination experiments, signifying a key role for Hirep1 neutralizing antibodies during protection against *C.t.* Finally, we show that passive immunization of Rag1 knockout mice with Hirep1 antibodies completely prevented the establishment of infection in 48% of the mice, demonstrating an isolated role for neutralizing antibodies in controlling infection. Our data emphasize the role of antibodies in early protection against *C.t.* and support the inclusion of neutralizing targets in chlamydia vaccines.

Keywords: chlamydia, vaccine, neutralizing antibodies, protection, cell-mediated immunity

# INTRODUCTION

Worldwide, sexually transmitted infections (STIs) by *Chlamydia trachomatis* (*C.t.*) cause an annual estimated incidence of over 131 million cases (1). Despite effective antibiotics, the large proportion of asymptomatic infections (2) impedes the control of *C.t.* infections. In women, untreated infections can lead to severe long-term sequelae with pelvic inflammatory disease, chronic

### *Edited by:*

*Lee Mark Wetzler, Boston University School of Medicine, United States*

### *Reviewed by:*

*Paola Massari, Tufts University School of Medicine, United States Anita S. Iyer, Massachusetts General Hospital, United States*

> *\*Correspondence: Peter Andersen pa@ssi.dk*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 10 November 2017 Published: 11 December 2017*

### *Citation:*

*Olsen AW, Lorenzen EK, Rosenkrands I, Follmann F and Andersen P (2017) Protective Effect of Vaccine Promoted Neutralizing Antibodies against the Intracellular Pathogen Chlamydia trachomatis. Front. Immunol. 8:1652. doi: 10.3389/fimmu.2017.01652*

**199**

**Abbreviations:** *C.t*., *Chlamydia trachomatis*; Sv*,* serovar; ELISA, enzyme-linked immunosorbent assay; IFN-γ, interferon gamma; IFU, inclusion forming unit; IL-17, interleukin 17; TNF-α, tumor necrosis factor-alpha; IL-2, interleukin 2; SC, subcutaneous; IN, intranasal; FACS, flow cytometry; EB, elementary body.

abdominal pain, ectopic pregnancy, and infertility as the most severe complications (3–5). The annual direct medical costs for chlamydial infections in the US alone exceed \$500 million/year (6). Consequently, WHO has recently initiated a global roadmap targeting STIs. The long-term control strategy is to develop prophylactic vaccines (7).

Historically, the vast majority of vaccines work *via* the induction of antibodies (8). Antibody (Ab)-mediated neutralization can efficiently block pathogens from entering host cells or neutralize bacterial toxins. *C.t.* has a complex bi-phasic lifecycle and infects epithelial cells in a range of mucosal sites (9). The intracellular lifestyle of *C.t.* has resulted in a focus on cell-mediated immunity (CMI) for efficient recognition of infected cells and control through cytokines or cellular cytotoxicity (10–12). Optimally, a vaccine would completely prevent infection, but more likely, it will reduce the initial infectious load followed by CMI responses that will accelerate clearing of the remaining bacteria. A *C.t.* infection in itself may drive insufficient amounts of neutralizing antibodies to be protective. Instead, current thinking is that the infection also induces antibodies that have secondary protective functions as facilitators for both the adaptive and the innate immune responses (13, 14).

Our vaccine development strategy focused on the *C.t.* major outer membrane protein (MOMP). MOMP is the most prominent protein in the outer membrane and has been shown to function as a porin (15) and an adhesin (16). MOMP is the primary target for neutralizing antibodies during infection (16–18) and has four variable domains (VDs) protruding from the surface of *C.t.* (19). We recently described a recombinant engineered multivalent vaccine construct [heterologous immuno-repeat 1 (Hirep1)], based on extended variable domain 4 (extVD4) regions from the most prevalent serovars (Svs) D, E, and F. Formulated in the Cationic Adjuvant Formulation no. 1 (CAF01) adjuvant, this vaccine exposes key neutralizing epitopes in the VD4 domain together with conserved T cell epitopes (20). It can elicit broadly crossreactive antibodies toward multiple serotypes. Furthermore, adoptive transfer of sera from vaccinated mice into recipient wild-type mice mediated protection against a primary vaginal challenge. Interestingly, depletion experiments showed that this protection was (at least partly) dependent upon CD4<sup>+</sup> T cells suggesting that the mechanism behind was a synergy between neutralizing antibodies and CD4<sup>+</sup> T cells (20). It is therefore still unclear what role these antibodies can play on their own in the protective immune response against *C.t.*

In the present study, we extend our studies on the Hirep1/ CAF01 vaccine in the mouse model. We compare the immunogenicity and protective efficacy of Hirep1 with a similarly designed construct based on four repeats of extVD4 from SvF. We demonstrate that this construct has equal ability to induce CMI and Ab responses. However, despite a high degree of sequence similarity, it lacks the ability to induce neutralizing antibodies and protection. We also show that Hirep1-induced neutralizing antibodies can adoptively transfer protection into Rag1 knockout (KO) mice, which emphasizes the role of antibodies without the involvement of CMI responses in the early control of infection with *C.t*.

# MATERIALS AND METHODS

### Organisms

The *C.t.* serovar D (*C.t.* SvD) (UW-3/Cx, ATCC VR-885), SvE (BOUR, VR-348B), and SvF (IC-Cal-3, ATCC VR-346) were purchased from the ATCC and propagated in HeLa-229 cells. Six-well plates were centrifuged at 750 *g* for 1 h at RT. *C.t.* elementary bodies (EBs) were harvested, purified, and quantified as described previously (21) and stored at −80°C in a sucrosephosphate-glutamate (SPG) buffer. All procedures were done using Biosafety level 2 containments.

## Animals

Female B6C3F1 (C57BL/6J x C3H/HeN) mice, 6–8 weeks of age, were obtained from Harlan Laboratories. The mice were housed under standard environmental conditions and provided standard food and water *ad libitum*. Rag1-deficient (Rag1<tm1Mom>) mice (Rag1 KO) (22) were obtained from JAX Laboratories (JAX Stock #002216) and housed in high-barrier facilities at Statens Serum Institut. Animal experiments were conducted in accordance with regulations of the Danish Ministry of Justice and animal protection committees by Danish Animal Experiments Inspectorate Permit 2013-15-2934-00978 and in compliance with EU Directive 2010/63 and the US Association for Laboratory Animal Care recommendations for the care and use of laboratory animals.

# Fusion Protein Preparation and Peptide Synthesis

Recombinant proteins: Hirep1 and extended VD4's from SvF (extVD4F \*4) were produced as follows: based on the amino acid sequences (NCBI–GenBank) with an added N-terminal histidine tag, synthetic DNA constructs were codon-optimized for expression in *Escherichia coli* followed by insertion into the pJexpress 411 vector (DNA2.0). To avoid disulfide bridge formation during recombinant production, all cysteines were exchanged with serines. Purification was done essentially as described in Ref. (23). The 9-mer biotinylated pepset was produced by Mimotopes (United Kingdom) and the 20-mer peptides were produced by GeneCust (Luxembourg). For amino acid (aa) sequences see Ref. (20) and **Table 1**.

# Mouse Immunization and Infection

Mice were immunized with 5 μg/dose/route of recombinant immunorepeat (extVD4F \*4 or Hirep1) constructs. Mice received a total of three immunizations at 2-week intervals either subcutaneously (SC) at the base of the tail in a total volume of 200 µl or simultaneously with the intranasal (IN) route in a volume of 30 µl. The antigens were diluted in Tris-buffer (pH 7.4) and mixed by vortexing 1:1 with 100 µl (SC) or 15 µl (IN) CAF01 adjuvant consisting of 50 μg/dose of the glycolipid trehalose 6,6′-dibehenate (TDB) incorporated into 250 μg/dose of cationic liposomes composed of dimethyldioctadecyl-ammonium. The mice were rested for 6 weeks (short-term) or 72 weeks (long-term). Ten and 3 days before *C.t.* SvD or SvF challenge, the estrus cycle was synchronized by injection of 2.5 mg Medroxyprogesteronacetat

### Table 1 | Sequences of the VD4-based vaccine constructs.


*Overview and sequences of the extended VD4-based constructs Hirep1 and extVD4F \*4 used in the present study. Gray box: VD4 region, Blue box: conserved sequence within the VD4 region. In the SvF sequence, the cysteine has been replaced with a serine.*

(Depo-Provera; Pfizer) increasing mouse susceptibility to chlamydial infection by prolonging diestrus (24). The mice were challenged intravaginally (i.vag.) with 4 × 104 *C.t.* SvD (Rag1 KO mice), 1 × 106 inclusion forming unit (IFU) of *C.t.* SvF (B6C3F1 mice, short-term experiment), and 4 × 103 IFU of *C.t.* SvF (B6C3F1 mice, long-term experiment) in 10 µl SPG buffer.

### Measurement of Ab Levels in Plasma/ Serum and Swab Material

Blood was collected 10 days after last vaccination for quantification of vaccine-specific antibodies by enzyme-linked immunosorbent assay (ELISA). For isolation of serum, the tubes were centrifuged for 10 min at 10,000 *g*. To separate plasma, samples were centrifuged 10 min at 500 *g*. Maxisorp Plates (Nunc, Denmark) were coated with either recombinant antigens (1 µg/ml) or live *C.t.* SvD–F (5 µg/ml). The plasma or serum samples were diluted 1:20 and fivefold serially diluted before being added to coated maxisorp plates.

Enzyme-linked immunosorbent assay reactivity against the 9-mer overlapping biotinylated peptide panels spanning the extVD4 region of SvD, SvE, and SvF was investigated. Briefly, precoated ELISA plates were obtained from Mimotopes, blocked with skimmed-milk powder, washed and incubated with plasma prediluted 1:200. Swab material was collected at PID3 in 600 µl SPG buffer and added undiluted to coated plates. Antigen-specific total IgG was detected with isotype-specific HRP-conjugated rabbit antimouse (Zymed). The substrate was TMB-PLUS (Kem-En-TEC, Denmark). Endpoint titers were calculated as the highest dilution where OD450–620 exceeds the cutoff value. Cutoff values were calculated for each dilution step as mean OD450–620 of naive mice + 3 × SD(naive mice).

### Chlamydia-Specific Cellular Responses

Splenocytes (four individual mice/group) were stimulated for 1 h with 5 µg/ml of Hirep1 or extVD4F \*4 at 37°C/5% CO2 and subsequently incubated for 5 h at 37°C with 10 µg/ml brefeldin A (Sigma-Aldrich, USA) at 37°C. The intracellular cytokine staining procedure was done essentially as described in Ref. (25). The following antibodies were used for surface and intracellular staining: FITC anti-CD8a (53–6.7), APC-eF780-anti-CD4 (GK1.5), PE-anti-tumor necrosis factor-alpha (TNF-α), APCanti-interleukin (IL)-2, PE-Cy7-anti-interferon gamma (IFN-γ), and PerCP-Cy5.5-anti-IL17. All antibodies were purchased from BD Pharmingen or eBiosciences. Responses were analyzed using a FACSCanto flow cytometer (BD) and FlowJo v.10.2 (Tree Star Inc.). Blood lymphocytes (PBMC's) were purified on a density gradient. Cells were pooled from eight mice and cultured in duplicate or triplicate in round-bottom plates (Nunc, Denmark) containing 2 × 105 cells/well in a volume of 200 µl RPMI-1640 supplemented with 5 × 10<sup>−</sup><sup>5</sup> M 2-mercaptoethanol, 1 mM glutamine, 1% pyruvate, 1% penicillin–streptomycin, 1% HEPES, and 10% fetal calf serum (FCS) (Invitrogen, Denmark). Three pools of eight mice were analyzed per group. The cells were re-stimulated with overlapping 20-mer peptides (2 µg/ml) covering the extVD4 region from SvD, E, and F [for aa sequences see **Table 1** and Ref. (20)]. As a positive control for cell viability and as a negative control, cells were stimulated with Concanavalin A (5 µg/ml) and media, respectively. After 72 h of incubation at 37°C in 5% CO2, supernatants were harvested and stored at −20°C before use. The amounts of secreted IFN-γ were determined in supernatants by ELISA as previously described (26).

### Neutralization

Blood samples were collected 3 weeks post last vaccination and sera were isolated. The neutralization assay was performed essentially as described in Ref. (27). Briefly, HaK cells (ATCC) were grown to confluence in 96-well flat-bottom microtiter plates. The adherent cells were maintained in RPMI-1640 (Gibco BRL) with 5 × 10<sup>−</sup><sup>5</sup> M 2-mercaptoethanol, 1 mM glutamine, 1% pyruvate, 10 µg/mL gentamicin, and 5% heat-inactivated FBS at 37°C/5% CO2. The *C.t.* stocks were diluted and mixed 1:1 with a heat-inactivated, serial diluted serum pool (*n* = 16 mice/group). The suspension was inoculated onto HaK cells in duplicates and incubated for 24 h. In the competitive neutralization inhibition assay, Hirep1 sera were preincubated with 20 µg/ml of peptides prior to the addition of SvD. Inclusions were visualized by staining with polyclonal rabbit antirecombinant CT043 serum, followed by Alexa 488-conjugated goat antirabbit immunoglobulin (Life Technologies). Cell staining was done with Propidium Iodide (Invitrogen). The results were calculated as the percentage reduction in mean IFU relative to control sera. A 50% or greater reduction in IFU relative to the control was defined as neutralizing.

### Passive Transfer of Immune Serum to Rag1 KO Mice

Two experiments were done. Sera were isolated from B6C3F1 mice previously vaccinated three times SC with 5 µg Hirep1/

CAF01. In both experiments, mice vaccinated with CAF01 alone were included. The sera were heat-inactivated, sterile filtered and transferred by the intravenous and intraperitoneal routes to 10 (Exp.1) or 15 (Exp. 2) Rag1 KO mice using an equivalent of 2.7 donor:1 recipient. Rag1 KO mice receiving serum from control mice were used for comparison. Three days post serum transfer the mice were challenged with 4 × 104 IFU/mouse of *C.t.* SvD and swabbed, as described under "Vaginal *C.t.* load." The results are shown both as individual experiments (Figure S1 in Supplementary Material) and as a pool of the two experiments (**Figure 5**).

### Vaginal *C.t.* Load

Vaginal swabs were obtained at 3, 7, 10, 14, 17, or 21 days postinfection. Swabs were vortexed with glass-beads in 0.6 ml SPG buffer and stored at −80°C until analysis. The infectious load was assessed as described in Ref. (28). Inclusions were visualized by staining with polyclonal rabbit anti-MOMP serum made in our laboratory, followed by an Alexa 488-conjugated goat antirabbit immunoglobulin (Life Technologies, Denmark). Background staining was done with propidium iodide (Invitrogen, Denmark). IFUs were enumerated by fluorescence microscopy observing at least 20 individual fields of vision for each well. Culture-negative mice were assigned the lower cutoff of 4 IFU.

### Histopathology

Mice were anesthetized and euthanized by cervical dislocation and necropsied at PID21. The entire reproductive tract was removed en bloc and fixed in formalin. After fixation, the tissue was processed and embedded in paraffin according to standard procedures. Sections were cut to include the uterine horns, oviducts, and ovaries in the same section. The sections were mounted on Superfrost glass and stained with hematoxylin and eosin. Slides were scanned with ×20 objective and at 1,360 × 1,024 resolution by the 3D Histech Pannoramic Midi scanner (3D HISTECH Ltd., Budapest, Hungary) and evaluated with CaseViewer software (3D HISTECH Ltd., Budapest, Hungary) (Nordic Biosite). The genital tract tissue was evaluated for the degree of inflammatory infiltrates/presence of inflammatory cells according to the following semi-quantitative scoring system: (0) No cellular infiltration. (1) Infiltration at a single focus or scattered single cells. (2) Infiltration at two to four foci, small/moderate accumulations or confluent infiltration. (3) Infiltration at numerous (>4) foci/ aggregates or moderate confluent infiltration. (4) Severe/extreme confluent infiltration. The sections were blinded to the treatment group and assessed by a pathologist.

### Statistical Analysis

GraphPad Prism 7 was used for data handling, analysis, and graphic representation. The differences in log 10 IFU obtained in the efficacy studies were analyzed using Kruskal–Wallis test followed by Dunn's multiple comparison tests or the Mann– Whitney. Fisher's exact test compared numbers of cleared and non-cleared Rag1 KO mice. Ab titers and pathology scoring were analyzed by the Mann–Whitney test. *P* < 0.05 was considered significant.

### RESULTS

### Immunogenicity of extVD4F\*4 in Comparison with Hirep1

Despite a high degree of sequence similarities, we have previously observed that the VD4 regions from SvE and SvF (extVD4E and extVD4F , **Table 1**) generate immune responses with different capacity to neutralize *C.t.* (20). With the purpose of directly comparing the protective efficacy of two similarly designed constructs with different capacity to induce neutralizing antibodies, we designed an immunorepeat based on four repeats of the ext-VD4F region–extVD4F \*4 (**Table 1**) and compared it directly with Hirep1. B6C3F1 mice were vaccinated three times with Hirep1 or extVD4F \*4 adjuvanted with CAF01. Mice were immunized simultaneously *via* the IN and SC routes, a vaccination regime previously demonstrated to be effective in inducing combined systemic and mucosal immune responses (20, 29). After vaccination, antigen-specific Ab, and CMI responses were compared (**Figure 1**). Hirep1 and extVD4F \*4 both induced a high level of vaccine-specific IgG antibodies in plasma with endpoint titers of 312,500 and 62,500, respectively (**Figure 1A**). The CMI responses were measured by FACS analysis on total numbers of cytokine (IFN-γ, IL-17, TNF-α, IL-2) producing CD4+ T cells and demonstrated around 1% cytokine positive CD4<sup>+</sup> T cell induced by both constructs (**Figure 1B**). In both groups, the functionality of the CD4<sup>+</sup> T cell population was diverse and consisted preferentially of CD4<sup>+</sup> T cells producing IL-2 in various combinations (**Figure 1C**). This is in agreement with published results showing that CAF01 primes memory responses with capacity for IL-2 production (30).

### Ab Neutralizing Capacity and Bacterial Surface Recognition

Having demonstrated similar levels of immunogenicity at both the CMI and Ab levels, we proceeded to investigate and compare the functional effect of the antibodies, i.e., their capacity to recognize the bacterial surface and to induce neutralization. Hirep1-specific antibodies had a broad Sv specificity and strongly recognized the surface of both Svs D, E, and F, whereas serum from extVD4F \*4 vaccinated mice only demonstrated a very weak surface recognition (**Figure 2A**). The capacity to bind to the bacterial surface correlated strongly with the functional activity of the antibodies measured in a complement-independent *in vitro* neutralizing assay (**Figure 2B**). Hirep1 generated sera neutralized SvD–F with reciprocal 50% neutralizing titers ranging from 750 to 4,000, whereas the serum generated after extVD4F \*4 vaccination was unable to induce neutralization of any of the three Svs tested.

### Specificities of Ab and CMI Responses after Hirep1 and extVD4F\*4 Vaccination

Given the pronounced difference in the functional effect of the Ab response promoted by the two constructs, we next analyzed the fine specificity of the responses. Ab specificity was determined by the reactivity against sequential and overlapping nonapeptides covering the extended VD4 regions of SvD, E, and F (**Figure 3A**). Abs from Hirep1 vaccinated mice induced a broad immune

were included in the figure. Results are shown for one of the two independent experiments.

response with a very strong recognition of peptides spanning the highly conserved TTLNPTIAG region (blue box) with amino acids LNPTI as a critical binding motif. In addition, Hirep1 vaccinated mice recognized a SvD/E-specific sequence covering aa TAIFD and the SvF-specific sequence covering aa RIAQPR. In contrast, the Ab response generated by the SvF immunorepeat differed markedly with a very strong and focused response toward the SvF-specific sequence covering amino acids RIAQPR. This response did not include the highly conserved TTLNPTIAG region.

Epitope-specific T cell responses were determined by measuring the *in vitro* stimulatory properties of overlapping 20-mer peptides (10aa overlap) on PBMC's from vaccinated mice. After stimulation, IFN-γ release was measured by ELISA (**Figure 3A**, inserts). The SvF-specific P29 and P30 from SvD, E, and F were the dominant T cell epitope regions recognized after vaccination with both constructs. However, whereas P30 is the dominant T cell epitope in Hirep1 vaccinated mice, SvF P29 dominate the CMI response in extVD4F \*4 vaccinated mice—overlapping with the regions inducing the strongest Ab response. Having demonstrated that Hirep1 in contrast to extVD4F \*4-specific serum was able to neutralize *C.t.*, we continued by investigating the relative importance of the two major epitopes unique for Hirep1 as targets for neutralization; the TAIFD and the highly conserved 50% neutralization titer.

LNPTI regions. To assess their relative role in neutralization, we used a competition neutralization assay with Hirep1 serum (**Figure 3B**). A 17-mer peptide FDTTT**LNPTI**AGAGDVK covering the LNPTI region and an overlapping 20-mer peptide RIAQPKSA**TAIFD**TTTLNPT (SvD/E P29) covering the SvD/E-specific TAIFD region were pre-incubated with serum from Hirep1 immunized mice and the mixture was further incubated for 45 min. with a fixed concentration of SvD before being transferred to HAK cells. Pre-incubation with the 17-mer peptide covering the LNPTI sequence completely abrogated the neutralization capacity, whereas blocking the serum with the peptide spanning the SvD/E-specific region TAIFD had no effect on the neutralizing capacity of the serum.

### Short- and Long-term Protection against *C.t.* SvF

Based on the identical levels of T cell responses but the markedly different capacity to induce neutralizing antibodies, we continued by investigating how this would affect short- (6 weeks) and longterm (>1 year) protection against a SvF infection (**Figure 4**). Mice were vaccinated with either Hirep1/CAF01, extVD4F\*4/CAF01, or CAF01 alone and challenged with 1 × 106 IFU/mouse 6 weeks after vaccination and with 4 × 103 IFU/mouse following 1 year of resting. Viable *C.t.* shedding was measured by vaginal swabbing 3, 7, and 10 days post-inoculation (PID). Hirep1 vaccinated mice were significantly protected 6 weeks following vaccination and despite a 10-fold reduction in serum IgG levels (results not shown) they sustained protective immunity 1 year after vaccination. ExtVD4F \*4 vaccination failed to protect at both time-points postvaccination demonstrating a key role for neutralizing antibodies in protection as well as the limitations of a CMI response on its own.

### Passive Immunization of Rag1 KO Mice with Hirep1 Antibodies

As all these investigations pointed to a dominant role for neutralizing antibodies in the protection promoted by the Hirep1 vaccine, we continued by characterizing the role of Abs during a primary infection independent of T and B cells. We transferred Hirep1-specific serum to Rag1 KO mice producing no mature T or B cells. We performed two individual experiments. B6C3F1 mice were vaccinated SC three times with Hirep1/CAF01 or CAF01 alone. The sera were isolated from vaccinated mice, heat-inactivated, sterile-filtered and transferred to Rag1 KO mice (2.7 donor:1 recipient) 3 days before infection. At PID 0 the Rag1 KO were bled and levels of Hirep1-specific antibodies in the periphery (serum) were measured by ELISA (**Figure 5A**). The mice were infected with 4 × 104 IFU of *C.t.* SvD*.* At PID3, Hirep1-specific antibodies in the genital tract were analyzed by swabbing to determine the level of Ab transudation to the genital tract (**Figure 5B**). We detected Hirep1-specific antibodies in swab material from all recipient Rag1 KO mice. IFU levels were determined at PID3, 7, 10, 14, 17, and 21 and presented as a pool of

the two experiments (**Figure 5C**). IFU levels from the individual experiments are shown in Figure S1 in Supplementary Material. The Hirep1-specific serum transfer protected Rag1 KO mice very efficiently throughout the experiment. At the level of individual animals, 12 out of 25 (48%) mice in the Hirep1 Ab-treated group completely prevented establishment of infection (non-infected at PID3), compared to 4 out of 25 (16%) in the group that received control serum (*P* < 0.05 by Fisher's exact test), and this level was sustained throughout the infection period (**Figure 5D**). We finally did histopathology examinations of the genital tracts

(unpublished observations). Each dot represents the median number ± interquartile range of IFU recovered from vaginal swabs at days 3, 7, and 10 post-infection (*n* = 8). The Dunn's multiple comparison tests were used for comparison among groups. \**P* < 0.05 and \*\**P* < 0.01.

21 days post-infection. The genital tract tissues were evaluated for the degree of inflammatory infiltrates and presence of inflammatory cells according to a semi-quantitative scoring system from 0 to 4 (**Figure 5E** and see Materials and Methods). The majority of the control group had confluent diffuse infiltration with inflammatory cells and some animals also luminal exudate with inflammatory cells. The median score of the control group was 2 (**Figures 5E,F**, lower panel). The mice from the Hirep1 group that were cleared at PID3 showed no infiltration with a score of 0 (**Figures 5E,F**, top panel), whereas the mice that were not cleared at PID3 showed histopathology comparable to the control group. The median score in the Hirep1 group was 1 (**Figure 5E**). In general, no inflammation was detected in fallopian tubes or ovarian bursa and only mild and scattered inflammatory infiltrates were seen in the vagina.

### DISCUSSION

In the present study, we have dissected the role of a neutralizing Ab response against *C.t.* infection without the confounding influence of CMI responses. Designing experimental vaccines based on either recombinant MOMP, MOMP peptides, or MOMP DNA has been highly challenging and with limited success (31–33). Recently, we designed a recombinant engineered Hirep1 molecule [**Table 1** and (20)]. Hirep1 includes T and B cell epitopes from extended VD4 regions of the most prevalent SvD–F (extVD4s) including the species-specific neutralizing LNPTIAG epitope (34, 35).

Here, we describe a comparative evaluation of Hirep1 and a construct based on a homologous repeat of the extVD4F \*4. This molecule showed overall similar immunogenicity but lacked the ability to induce neutralizing antibodies. Despite the high degree of sequence similarities, the extVD4F \*4 generated immune serum recognized predominantly a SvF-specific region RIAQPR and had markedly different specificity against the conserved LNPTIAG region previously shown to contain neutralizing epitopes for SvF (34). In contrast, Hirep1 generated sera recognized a much broader epitope repertoire covering both the conserved and serotype-specific regions confirming previously published results (20). The functional Ab response, in terms of neutralization, was blocked with a peptide covering FDTTTLNPTIAGAGDVK demonstrating that although Hirep1 immunization induced antibodies that recognized different parts of the VD4 region only the LNPTI region was a target for

semi-quantitative scoring system from 0 to 4 (see Materials and Methods). Each dot represents the individual scoring. The median score ± interquartile range is indicated. The Mann–Whitney test was used for comparison among groups \**P* < 0.05. (F) Representative section of a genital tract with a score 0 referring to no cellular infiltration (top panels). Representative section of a genital tract with a score 2, referring to confluent cellular infiltration and furthermore luminal exudate with inflammatory cells (black arrow) (lower panels).

neutralizing antibodies. Thus, although LNPTI is conserved between extVD4F \*4 and Hirep1, subtle changes in aa surrounding this region have marked effect on the specificity of the Ab response—a challenge for vaccine design that the Hirep1 vaccine molecule addresses. Previous studies using MOMP have also shown high sensitivity of amino-acid changes on the neutralizing Ab response (36). Kari et al. immunized cynomolgus macaques with native MOMP purified from the clinical SvA isolate A2497, which generated highly neutralizing antibodies against the homologous strain but failed to neutralize SvA HAR-13, a strain differing by only four amino acids (37).

Besides broad Sv coverage, a novel *Chlamydia* vaccine will need to induce long-lived protection, at least covering the age group of 15–29 where the infection is most prominent. Here, we show that immunization with Hirep1 formulated in CAF01 sustained protective immunity for more than 1 year, confirming several previous observations of the strong and long-lived immunity induced by the adjuvant CAF01 in both animal models (30, 38) and human clinical trials (39). Following vaccination and challenge with *C.t.* SvF, only Hirep1 was able to induce both shortand long-term protection. Thus, although extVD4F \*4 induced a very similar CMI response, this was on its own not enough to significantly control infection, while the combined effect of neutralizing antibodies and CMI efficiently reduced the bacterial levels. The exact mechanism behind this interplay between CMI and neutralizing antibodies is unknown, but several studies have previously suggested mechanisms such as ADCC, Fc mediated enhanced phagocytosis and microbial killing (13, 14).

Despite a vast number of *in vitro* studies documenting a role of antibodies in both neutralization and complement activation (18, 40–44) the demonstration of a direct correlation between neutralizing antibodies and protection during a primary *C.t.* infection has been challenging. In general, studies in mice demonstrating a role for antibodies have predominantly demonstrated an important role during the secondary infection (in preconditioned tissue) (14, 45, 46) and no isolated role of antibodies has been suggested in those studies. In the current study, we demonstrate a protective efficacy of Hirep1-specific antibodies by transferring them to Rag1 KO mice producing no mature T and B cells (22). 48% (12/25) of mice receiving Hirep1 immune sera were uninfected day 3, and 10 of these mice had no sign of pathological changes, indicating that these mice completely controlled infection. These data strongly suggest a role for antibodies on their own in neutralizing the bacteria. The exact mechanism is not clear, but we believe that Hirep1-specific Ab molecules accumulate on the surface of *C.t.* and either directly inhibit them from binding and infecting the epithelial target cells or activate the complement system leading to direct lysis of the *C.t.* membrane. Possibly, antibodies could also slow down or even immobilize incoming bacteria *via* Ab-bacteria binding to mucin fibers that constitute Cervical-vaginal Mucus, a mechanism recently described for protection against HIV (47, 48).

The reason for the observed disagreement between our findings and previous work is most likely to be found in the level and functionality of antibodies. Previous studies have primarily investigated infection promoted antibodies (46), which in our hands have limited neutralizing capacity compared to Hirep1-specific antibodies (unpublished results). In support of the isolated capacity of antibodies to control infection are the observations by Cotter et al. demonstrating that MAbs delivered into serum and vaginal secretions of naive mice by using a backpack hybridoma system can reduce pathology (49). Similarly, studies from other animal models, have pointed to an isolated role of antibodies in protection against a primary *C.t.* infection (50).

In summary, we show that a vaccine inducing both neutralizing antibodies and CMI can significantly protect against infection in mice both short-and long-term post-vaccination.

### REFERENCES


Importantly, we provide evidence that antibodies on their own can prevent the establishment of *C.t.* infection in Rag1 KO mice. This emphasizes a previously unrecognized role of antibodies as a first line of defense against *C.t.* infection and supports the inclusion of neutralizing targets in the development of future Chlamydia vaccines.

### ETHICS STATEMENT

Animal experiments were conducted in accordance with regulations of the Danish Ministry of Justice and animal protection committees by Danish Animal Experiments Inspectorate Permit 2013-15-2934-00978 and in compliance with EU Directive 2010/63 and the US Association for Laboratory Animal Care recommendations for the care and use of laboratory animals.

### AUTHOR CONTRIBUTIONS

AO planned the study, performed the experiments, and made the laboratory analysis, statistics, interpreted data, and drafted the figures and manuscript. FF and PA planned the study, interpreted data, and revised figures and the manuscript. EL performed the histopathology examinations and revised figures and the manuscript. IR produced the recombinant constructs and revised figures and manuscript. All the authors approved the final manuscript.

### ACKNOWLEDGMENTS

We thank Sanne Kelly, Patricia Grénes, and Vita Skov for excellent technical assistance and Edel Boysen for proofreading the manuscript.

### FUNDING

This study was supported by the European Commission through the ADITEC consortium contract [FP7-HEALTH-2011.1.4-4-280873] and the Danish Council for Independent Research | Medical Sciences [Grant no.: DFF-4004-00424]. Part of the data has been presented as an oral presentation at the 8th Chlamydia Basic Research Society Meeting, April 07–10, 2017, Charlotte, NC, USA.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01652/ full#supplementary-material.


and IgG monoclonal antibodies in a murine model of *Chlamydia trachomatis* genital tract infection. *Infect Immun* (1995) 63:4704–14.

50. Rank RG, Batteiger BE. Protective role of serum antibody in immunity to chlamydial genital infection. *Infect Immun* (1989) 57:299–301.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. PA, AO, IR, and FF are coinventors on a patent application relating to *C.t.* vaccines. All rights have been assigned to Statens Serum Institut, a Danish not-for-profit governmental institute.

*Copyright © 2017 Olsen, Lorenzen, Rosenkrands, Follmann and Andersen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Seasonal Influenza Split Vaccines Confer Partial Cross-Protection against Heterologous Influenza Virus in Ferrets When Combined with the CAF01 Adjuvant

*Dennis Christensen1 \*‡ , Jan P. Christensen2‡, Karen S. Korsholm1 , Louise K. Isling3 , Karin Erneholm1†, Allan R. Thomsen2 and Peter Andersen1*

*1Department of Infectious Disease Immunology, Division of Vaccine, Statens Serum Institut, Copenhagen, Denmark, 2Department of Immunology and Microbiology, University of Copenhagen, Copenhagen, Denmark, 3Department of Quality Assurance/Quality Control, Section of Biological Service, Division of Vaccine, Statens Serum Institut, Copenhagen, Denmark*

Influenza epidemics occur annually, and estimated 5–10% of the adult population and 20–30% of children will become ill from influenza infection. Seasonal vaccines primarily work through the induction of neutralizing antibodies against the principal surface antigen hemagglutinin (HA). This important role of HA-specific antibodies explains why previous pandemics have emerged when new HAs have appeared in circulating human viruses. It has long been recognized that influenza virus-specific CD4(+) T cells are important in protection from infection through direct effector mechanisms or by providing help to B cells and CD8(+) T cells. However, the seasonal influenza vaccine is poor at inducing CD4(+) T-cell responses and needs to be combined with an adjuvant facilitating this response. In this study, we applied the ferret model to investigate the cross-protective efficacy of a heterologous trivalent influenza split-virion (TIV) vaccine adjuvanted with the CAF01 adjuvant, with proven ability to induce CD4(+) T-cell and antibody responses in mice, ferrets, pigs, primates, and humans. Our results indicate that CAF01-adjuvanted vaccine induces HA inhibition (HAI)-independent protection after heterologous challenge, manifested as reduced viral load and fever. On the other hand, we observe increased inflammation in the airways and more neutrophil and mononuclear cell infiltration in these ferrets when compared with optimally protected animals, i.e., ferrets receiving the same vaccine but a homologous challenge. This suggest that HAI-independent immunity induced by TIV + CAF01 can reduce viral shedding and systemic disease symptoms, but does not reduce local inflammation in the nasal cavity.

Keywords: influenza A virus, adjuvants, immunologic, heterologous immunity, ferrets, T cells

### INTRODUCTION

Annual influenza epidemics cause illness in up to 30% of the population dependent on age and general health status. Worldwide, these annual epidemics are estimated to result in 3–5 million cases of severe illness, and about 250,000–500,000 deaths, especially among children, elderly, and immune-deprived individuals (http://www.who.int/mediacentre/factsheets/fs211/en/).

### *Edited by:*

*Ken J. Ishii, National Institutes of Biomedical Innovation, Health and Nutrition, Japan*

### *Reviewed by:*

*Ali Ellebedy, Washington University in St. Louis, United States Rino Rappuoli, GlaxoSmithKline, Italy Michael Schotsaert, Icahn School of Medicine at Mount Sinai, United States*

> *\*Correspondence: Dennis Christensen den@ssi.dk*

### *†Present address:*

*Karin Erneholm, Timeline Bioresearch, Lund, Sweden*

*‡*

*These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 15 December 2017 Published: 08 January 2018*

### *Citation:*

*Christensen D, Christensen JP, Korsholm KS, Isling LK, Erneholm K, Thomsen AR and Andersen P (2018) Seasonal Influenza Split Vaccines Confer Partial Cross-Protection against Heterologous Influenza Virus in Ferrets When Combined with the CAF01 Adjuvant. Front. Immunol. 8:1928. doi: 10.3389/fimmu.2017.01928*

**211**

In addition to this, new strains occasionally appear, reflecting mutations and recent re-assortments, causing potentially life-threatening pandemics. In the past century, four such pandemics have occurred with the most lethal being the Spanish flu (1918–1920) killing between 50 and 100 million people, making it one of the most deadly disease outbreaks in history (1). For the past 20 years, there have been a number of highly pathogenic avian influenza strains circulating, which occasionally have crossed over to humans, in some cases with fatal outcome as recently reported for avian H10N8 (2), H7N9 (3), and H5N6 (4). Fortunately, human-to-human transmission of such highly lethal strains have not yet been reported, but it may only be a question of time before a human adapted variant will appear.

Seasonal flu vaccines primarily work through the induction of neutralizing antibodies against the principal surface antigen hemagglutinin (HA). HA attaches to host cell sialic acid receptors thus mediating infection of the cell. Neutralizing antibodies directed toward HA therefore reduce infection and they can even confer sterilizing immunity. This important role of HA-specific antibodies explains why previous pandemics have emerged when new HAs have appeared in viruses circulating in the human population. The latest pandemic in 2009 thus emerged when the circulating swine lineage of H1 HA, replaced the circulating human lineage (A/Brisbane/59/2007). The emerging H1N1 strain (A/California/04/2009) diverged from recent H1 variants with up to 50% in the residues corresponding to the antigenic sites of HA and were more closely related to the H1 HA strain causing the 1918 pandemic, with as little as 20% residue difference (5). To prevent or reduce severity of future influenza pandemics, it is therefore important to develop influenza vaccines, which do not rely so heavily on HA neutralization.

Cell-mediated immune responses against highly conserved epitopes in proteins such as the matrix proteins (M1 and M2), the nucleoprotein, and the polymerase basic proteins (PB1 and PB2) have been shown to mediate heterologous and heterosubtypic immunity against influenza A (6, 7). Cytotoxic CD8+ T cells are thought to be the primary mediators, but there is mounting evidence that also CD4+ T cells can facilitate of heterologous/ subtypic immunity although the underlying mechanisms are still not completely clear (8, 9). The main view is that CD4+ T cells provide help to CD8+T cells and B cells to promote viral clearance (8, 10). It has, however, been shown that both TH1 and TH17 cells transferred into naive mice can protect them against influenza, whereas TH2 cells do not play a role (11). This suggests that induction of TH1 and TH17 responses to the seasonal trivalent influenza split-virion (TIV) vaccine may increase cross-protection against heterologous/-heterosubtypic influenza strains and that the combination of TIV with the right adjuvant may facilitate this.

Several novel candidate adjuvants in clinical development have been assessed within the Advanced Immunization Technologies collaborative research program. Among these, the CAF01 adjuvant was recently shown to generate the strongest T-cell response in mice and that this T-cell response was biased toward a mixed TH1/TH17 response (12–15). This strong T-cell induction has been obtained in both mice, ferrets, pigs, non-human primates, calves, and humans (12, 16–21).

In this study, we applied the ferret model to investigate the cross-protective efficacy of T cells generated by a heterologous TIV vaccine adjuvanted with CAF01. The ferret model has for long been the preferred model for testing human influenza vaccines, the main reason being that ferrets share host range, clinical symptoms, and pathological changes in the respiratory tract with humans (22). On the other hand, the ferret model is difficult with respect to monitoring immune responses. The serum antibody responses to influenza infection and vaccination are much similar in humans and ferrets, and the methods are well established. However, the ferret model has only few tools for monitoring innate and adaptive cell-mediated immune responses. Vaccine efficacy determined in the ferret model therefore has to rely on monitoring of clinical symptoms and humoral responses. This complicates the evaluation of the cross protectiveness of influenza vaccines in situations where HA neutralization does not play a critical role. We therefore combine the classical clinical monitoring with immunohistological examination of the infection site to study the cross-protective efficacy of CAF01-adjuvanted TIV.

## RESULTS

Our primary goal was to evaluate whether CAF01 could enhance the heterologous protection by a TIV vaccine. Thus, we immunized four groups of ferrets with adjuvanted or non-adjuvanted suboptimal doses (1/10 human dose) of TIV vaccines containing split-virion proteins from Influenza B, H3N2, and H1N1 of either the A/Brisbane/59/2007 or the A/California/04/2009 strains (designated TIV2007 and TIV2009, respectively) (**Table 1**). Unimmunized ferrets were used as controls.

The ferrets were immunized twice 3 weeks apart. All groups including the unimmunized control group were challenged intranasally 4 weeks after last immunization, with 105 TCID50 of the A/Brisbane/59/2007 H1N1 influenza strain. Thus, two vaccination groups were challenged with a homologous virus and the other two groups were challenged with a heterologous virus (**Table 1**). We also tested the reverse combinations using A/ California/04/2009 as challenge strain. These data are not shown since the results were equivalent.

### Viral Load

The ability of the vaccines to inhibit influenza infection was investigated by swabbing the nasal cavity of the ferrets from 1 to 7 days after challenge and determining the viral load by quantitative RT-PCR (**Figure 1**). The ferrets immunized with the


Figure 1 | CAF01 enhances the protection of a split flu vaccine both against a homologous and a heterologous challenge. Ferrets immunized two times with either the seasonal trivalent influenza split-virion (TIV)2007 (A) or the seasonal TIV2009 (B) vaccine without adjuvant (▴) or with CAF01 adjuvant (▵). Unimmunized controls (⚫). Challenge with the A/Brisbane/59/2007 influenza strain. Viral load in nasopharyngeal swab samples was determined by quantitative real-time polymerase chain reaction. Shown are mean values + SD (*n* = 8). Statistical analysis: repeated-measures two-way ANOVA based on the data shown in each graph. Dunnet's test was used to compare unimmunized to the two other groups. Asterisks show statistically significant differences compared to the unimmunized group.

non-adjuvanted vaccines were not protected at the chosen suboptimal dose of homologous and heterologous TIV. By contrast, the CAF01-adjuvanted TIV vaccine significantly reduced the viral load at all time points after homologous challenge. In addition, the CAF01-adjuvanted vaccine reduced the viral load after heterologous challenge at both the two earliest and at the latest time points investigated, showing that adjuvanting the TIV with CAF01 increases the resistance to challenge with a heterologous influenza virus. The reduction in viral load was less pronounced after heterologous challenge.

### Humoral Immunity

Protection against influenza is commonly associated with antibody-mediated neutralization. The ability of the different vaccines to induce IgG antibodies against the dominant surface antigen, HA, from the challenge virus was therefore investigated (**Figure 2A**). Only the homologous vaccine adjuvanted with CAF01 induced significant TIV-specific IgG titers before challenge. There was a slight, but insignificant, increase also for the ferrets immunized with the heterologous vaccine adjuvanted with CAF01, suggesting that some cross-reactive antibodies were induced. The regional specificity of these antibodies was not investigated but the cross-recognition could be toward conserved parts of HA, e.g., the stem region. After challenge, the specific IgG antibody response was significantly increased above the levels in unvaccinated ferrets only in those groups receiving the CAF01-adjuvanted vaccines. In addition, analysis of hemagglutination inhibition (HI) titers, which represents the correlate of protection used for humans, showed that only the

Figure 2 | CAF01 enhances influenza-specific IgG antibody responses but not cross-reactive hemagglutination inhibition (HI) responses. Ferrets immunized two times with either the seasonal trivalent influenza split-virion (TIV)2007 or the seasonal TIV2009 vaccine with or without CAF01 adjuvant and challenged with the A/Brisbane/59/2007 influenza strain. Blood was drawn on 3 days before and 7 days after challenge. (A) IgG antibody titers against hemagglutinin (HA) and (B) HI titers against the challenge strain. Shown are mean values + SD [*n* = 8, except for (B): heterologous TIV + CAF01 where *n* = 7]. Statistical analysis: two-way ANOVA based on the data shown in each graph. Sidak's multiple comparison test was used to compare the vaccinated with the unimmunized group before (white) and after (black) infection. Star symbols show statistically significant differences compared to the unimmunized group ★*p* < 0.05, ★★★*p* < 0.001. Sidak's multiple comparison test was used to compare the intra-group difference before and after infection and the hash symbols show statistically significant differences compared to before infection #*p* < 0.05, ##*p* < 0.01, ###*p* < 0.001.

CAF01-adjuvanted homologous vaccine was able to induce HI titers to the challenge strain before challenge (**Figure 2B**). None of the other vaccines induced HI titers above the level of the unimmunized controls at any of the investigated time points but all groups had significantly higher HI titers after challenge. Thus, the CAF01-adjuvanted heterologous vaccine was unable to induce cross-reactive protective HA-specific antibodies detectable in either of these analyses, suggesting that the partial protection obtained with heterologous vaccination was not mediated by HA inhibition (HAI).

### Clinical Symptoms

We also examined if adjuvanting the vaccines could reduce clinical symptoms during influenza infection. Thus, the body temperature and body weight of the ferrets were monitored before and during challenge (**Figure 3**). Corresponding with the reduced viral load, only the CAF01-adjuvanted vaccines significantly protected against fever at day 2 after both homologous and heterologous challenge (**Figures 3A,B**). The CAF01-adjuvanted vaccine also protected the ferrets against weight loss after challenge with the homologous virus but not the heterologous vaccine (**Figures 3C,D**). On the contrary, the TIV + CAF01 immunized ferrets after heterologous challenge lost weight comparable to unimmunized and TIV immunized ferrets and did not gain weight again, as observed for the other groups within the 7 days the study lasted. This suggests that the TIV-CAF01 vaccine prevents systemic inflammation, whereas the weight loss could be explained by local inflammation in

Figure 3 | CAF01 adjuvantation reduces fever after both a homologous and a heterologous challenge. Ferrets immunized two times with either the seasonal trivalent influenza split-virion (TIV)2007 (A,C) or the seasonal TIV2009 (B,D) vaccine without adjuvant (▴) or with CAF01 adjuvant (▵). Unimmunized controls (⚫). Challenge with the A/Brisbane/59/2007 influenza strain. Changes in relative body temperature (A,B) were determined by rectal measurements, where an average of day −1 and 0 were set to 0 and the difference was determined as that on the day of measurement minus the starting point. Relative weight (C,D) were determined by measurements where an average of day −1 and 0 were set to 100% and the difference was determined as day of measurement divided by the starting point. Shown are mean values + SD (*n* = 8). Statistical analysis: repeatedmeasures two-way ANOVA based on the data shown in each graph. Dunnett's test was used to compare the unimmunized to the two other groups. Asterisks show statistically significant differences compared to the unimmunized group (*p* < 0.05).

the airways, causing failure to thrive, and reduced food and water intake.

### Gross Pathology after Challenge

Since antibodies could not readily explain the protective effect of the CAF01-adjuvanted heterologous vaccine, we assumed that it was T-cell mediated based on the known T-cell inducing properties of CAF01. However, T-cell responses are notoriously difficult to measure in ferrets due to the lack of specific reagents of sufficient quality, so we decided to investigate if the observed protection was associated with increased influx of lymphocytes to the site of infection and/or enhanced local pathology. Thus, in a follow-up experiment with all four vaccines, ferrets were euthanized and necropsied 4 days after challenge with A/ Brisbane/59/2007, as this time point is commonly associated with local pathology in ferret influenza models. Macroscopic lesions were not observed in the pharynx or in the lower respiratory tract in any of the groups.

### Histopathology after Challenge

As observed from the macroscopic evaluation, histological evaluation of the pharyngeal sections revealed no differences between the four vaccination groups (data not shown). However, in the nasal cavity, there was a significantly lower degree of inflammation in the homologous TIV + CAF01 group, compared to the other vaccination groups (**Figures 4** and **5**).

Generally, normal ciliated respiratory epithelium was present in the majority of each examined section from ferrets in the homologous TIV + CAF01 group (**Figures 4A,G**), while in the other groups the nasal mucosa was completely or almost completely devoid of ciliated epithelial cells as well as goblet cells and displayed large regions with hyperplastic epithelium infiltrated by moderate to large numbers of neutrophils (**Figures 4A,E,F,H**; described epithelium marked with arrowheads). Regions of variable size with flattened epithelium were also present in several ferrets.

In the lumen, mild-to-severe amounts of exudate composed of neutrophils, cellular debris, and mucus was present in the majority of ferrets from the homologous TIV, heterologous TIV, and heterologous TIV + CAF01 groups, while either absent or significantly reduced in ferrets from the homologous TIV + CAF01 group (**Figures 4B,E–H**; exudate marked with stars).

The ventral nasal concha thickness score was statistically significant higher in the homologous TIV as well as in the heterologous TIV and TIV + CAF01 groups compared to the homologous TIV + CAF01 group, while no difference was observed between the heterologous TIV and TIV + CAF01 groups (**Figure 4C**). The increased thickness of the ventral nasal concha was caused by an increased inflammatory cell infiltration of the lamina propria and a mild mucosal edema (**Figures 4D–H**).

The degree of inflammatory cell infiltration in the lamina propria was not statistically significantly different between the homologous TIV and TIV + CAF01 groups, whereas the infiltration of inflammatory cells was statistically significantly higher in the heterologous TIV + CAF01 group than in the heterologous TIV and homologous TIV + CAF01 groups (**Figure 4D**). In general, all ferrets that were not protected by

TIV ± CAF01 were comparable; however, the degree of cellular infiltration in lamina propria was statistically significant higher in heterologous TIV + CAF01 compared to both heterologous TIV and homologous TIV + CAF01. (E–H) H&E-stained histological sections of nasal cavities showing unaffected thin nasal concha covered by normal ciliated respiratory epithelium from ferret vaccinated with homologous TIV + CAF01 (G) and clearly affected nasal cavities from ferrets vaccinated with homologous TIV (E), heterologous TIV (F), and heterologous TIV + CAF01 (H) with exudate in lumen, thickened nasal concha covered by hyperplastic epithelium lacking cilia, presence of transmigrating neutrophils and with a mixed inflammatory cell infiltration in lamina propria (most prominent in group TIV2009 + CAF01). Stars show exudate, arrowheads show epithelium, and lamina propria is marked with double-headed arrows. Bar = 500 μm. Non-parametric one-way ANOVA based on the data shown in each graph. Kruskal–Wallis test by ranks were used to compare the groups. Statistical significant difference is shown by *p*-value.

Figure 5 | Composition of cellular infiltration to the site of infection is influenced by CAF01 adjuvantation. Ferrets immunized two times with either the homologous or heterologous trivalent influenza split-virion (TIV) vaccine with or without CAF01 adjuvant and euthanized 4 days after challenge with A/Brisbane/59/2007. (A) Score of number of infiltrating neutrophils' (gray bars) and lymphocytes (black bars) in lamina propria hematoxylin and eosin (H&E)-stained sections, *n* = 4 (homologous TIV), 6 (homologous + CAF01), and 8 (heterologous TIV ± CAF01). Bars show mean value + SD. (B–E) Representative histological sections with immunohistochemical detection of CD3 showing brown colored CD3(+) T cells in heterologous TIV (B,D) and heterologous TIV + CAF01 (C,E). Bars = 50 μm. (F) Score of number CD3(+) T cells and (G) plasma cells (H&E-stained section) in lamina propria. Dots represent individual values and lines median values. Statistical analysis: the Mann–Whitney two-tailed *U* test was applied to compare infiltration of cells between groups. Statistical significant difference is shown by *p*-value.

antibodies, i.e., those vaccinated with homologous TIV, heterologous TIV, and TIV + CAF01, experienced an increased influx of neutrophils. Furthermore, the cellular infiltration in the lamina propria was dominated by neutrophils in ferrets from the homologous and heterologous TIV groups (**Figure 5A**), while the cellular infiltration in the heterologous TIV + CAF01 group was dominated by substantial infiltration of mononuclear cells including lymphocytes (**Figure 5A**). Thus, vaccination with the CAF01-adjuvanted vaccine was associated with infiltration of mononuclear cells along with neutrophils, whereas the cell infiltration was dominated by neutrophils in those ferrets vaccinated with an unadjuvanted TIV vaccine.

Immunohistochemical (IHC) detection of CD3(+) T cells was performed on sections from the nasal cavities to investigate if heterologous vaccination was associated with a more pronounced influx of T cells compared to homologous immunization. A statistically significantly higher score of the number of infiltrating CD3(+) T cells in lamina propria was found in the heterologous TIV + CAF01 group as compared to both the homologous TIV + CAF01 and the TIV groups (**Figure 5F**). This is also illustrated in the representative slides of CD3 ICH stainings of histological sections of the nasal cavity from the two vaccine groups receiving heterologous challenge (**Figures 5B–E**). Finally, the number of plasma cells infiltrating into lamina propria after infection were scored, and no statistically significant increase was observed in the adjuvanted groups (**Figure 5G**).

### DISCUSSION

It is well established that T cells play an important role in the protection against influenza infection, especially when HA-specific neutralizing antibodies are not available. Several studies have thus shown that preexisting T-cell immunity is associated with lower virus shedding and less severe illness if antibody-mediated neutralization is not possible (23–25).

The efficacy of the seasonal influenza vaccines is very low when it comes to infections with influenza strains that are not closely matched for HA, because this vaccine predominantly induces an antibody response to this antigen and T-cell immunity is not obtained (26). In this study, we therefore combined the TIV vaccines with the CAF01 adjuvant which in addition to Ab responses is known to elicit a strong T-cell response in both mice and humans. Adjuvanting the TIV vaccines with CAF01 enhanced the protective efficacy, measured as reduced viral load and fever, against both homologous and heterologous challenge. Despite the absence of detectable cross-reactive HAI antibodies, protection after heterologous challenge was not associated with an increase in local pathology when comparing to unprotected ferrets. Adjuvanting the vaccines changed the cellular recruitment to the site of infection from being dominated by neutrophilic granulocytes to lymphocytes. Increased infiltration of lymphocytes with the adjuvanted vaccines suggests that infection recruits lymphocytes that have been induced by vaccination.

This suggest that CAF01 adjuvanted the TIV vaccine increases the CMI component of the immune system, which is also supported by the histological staining of CD3+ cells in the airways after intranasal challenge, showing increased infiltration of CD3+ T cells. This rapid infiltration of T cells can lead to containment of the viral infection and faster reduction of viral load in the lungs, either by direct killing of infected cells or by mediating recruitment of innate immune cells to eliminate the infection. This correlates with observations from previous studies showing that previous heterologous infection induces a certain extent of protective immunity to novel influenza virus subtypes (27–29). In addition, Laurie et al. show that multiple infections with different influenza A strains confer better protection than a single pre-infection, suggesting that immunity to more conserved B and T cell epitopes of the influenza virus play a role for this protection (27). This was supported by the observation that primary infection by influenza strains to which the ferrets were protected by vaccination conferred reduced protection in comparison to those that were not immune to the primary infection (27, 28). Also, Bodewes et al. saw a correlation between the induction of CD3+ CD8− T cells coming from primary H3N2 virus infection and heterosubtypic protection against H5N1 virus infection (28). This role of cross-reactive T cells to mediate partial protection against a HA antigenically novel influenza virus has also been shown in non-human primates, where animals receiving a primary influenza infection had T cells coming into the to the lungs as soon as 2 days post-heterologous secondary influenza infection, in comparison to 5–7 days in non-primed animals, and that the primed animals cleared the secondary infection faster (23).

Despite the observed reduced viral load and fever symptoms following the heterologous TIV + CAF01 vaccine, we still observed weight loss in these animals. This suggest that the heterologous TIV + CAF01 vaccine prevents fever, whereas the weight loss could be explained by an increased local inflammation in the nasal cavity, as an increased influx of neutrophils and mononuclear cells was found in this group, which most likely would lead to stuffy and runny noses in those ferrets. Ferrets are obligatory nose breathers, and a stuffy nose can lead to respiratory compromise due to the ineffectiveness of mouth breathing, which again may lead to reduced food and water intake. We therefore believe that the observed weight reduction in this case was a consequence of local inflammation in the nasal cavity caused by the infection, and enhanced by the vaccine-mediated recruitment of immune cells to the airways, which on the other hand reduced the fever.

Overall, this study indicate that the combination of TIV vaccines with the TH1/TH17-inducing adjuvant CAF01 leads to increased protection against homologous influenza infection but also heterologous influenza infection despite absence of neutralization *via* HAI. Although neutralizing Ab responses to other antigens than HA can help to increase protection with the CAF01-adjuvanted vaccine, the data suggest that CMI induced infiltration of mononuclear cells play a major role, when HAI is sidelined.

### MATERIALS AND METHODS

All animal experiments were approved by the Danish Animal Care and Ethics Committee and conducted in accordance with the Danish Animal Experimentation Act and the European Convention for the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes (permit no. 2012-15-2934-00503). Animals were housed in free-range area with up to 24 animals in each pen. The housing was controlled with temperature and humidity and animals were on a 12/12-h light cycle. Animals received cat food and water *ad libitum*. Animals were anesthetized with Zoletil for cats (90–120 µl dependent on their weight) when sampling and measurement of any form were performed and exsanguinated with 2-ml pentobarbital injection in the heart.

### Immunization Schedule

Outbred, 6–10 months old, female ferrets (*Mustela putorius* furo, *n* = 4–8) were randomly assigned to groups and vaccinated subcutaneous in the groin two times with 3 weeks in between, with a low dose (1/10 human dose) of either the 2007 or 2009 TIV vaccine (containing H3N2 and B, plus either A/Brisbane/59/07 or A/ California/07/09 virus antigens). The TIV vaccines were used alone or mixed with the CAF01 adjuvant. A control group of unvaccinated animals received 250 µl of phosphate-buffered saline (PBS) per injection instead of the vaccine. The ferrets were housed in a class II isolation facility at the Faculty of Health Sciences, University of Copenhagen, with free access to food and water. Before vaccinations, animals were confirmed to be seronegative for circulating and applied influenza A (H1N1 and H3N2) and influenza B viruses by hemagglutination inhibition assay (HAI) and ELISA.

Eight weeks after vaccination, all animals were inoculated intranasally (i.n.) with 105 TCID50 of A/Brisbane/59/2007 H1N1 produced in eggs. Thus, the challenge was homologous for one half of the vaccinated ferrets and heterologous for the other half.

Two repeat studies were performed of which one was terminated 4 days after challenge to sample tissue for histology. During challenge, nasal washes and blood samples were taken and temperatures measured daily from day 0 to day 7 post infection (p.i.). Nasal washes were performed using a pipette by application of 1 ml of PBS into the nostrils of each ferret. Nasal wash sample was collected and kept at −80°C. Blood samples were taken from the cranial vena cava. Temperature was taken by rectal measurement.

### Clinical Symptoms

Since ferrets have a large inter-individual set-point for "normal" temperature we measured the relative change in body temperature. The set point was determined as the average of the two measurements before infection in individual animals and the indicated changes were calculated as that on the day of measurement minus the set-point value.

Weight was measured during anesthesia. The starting point was determined as the average of the two measurements before infection in individual animals and this value was set to 100%. The indicated changes were calculated as that on the day of measurement divided by the starting value.

### IgG Antibodies

Detection of anti-HA antibodies in the sera from ferrets was performed by ELISA. In brief, 96-well MaxiSorp plates (Nunc, Denmark) were coated with 0.5 mg/ml of HA from either A/California/04/2009 in 15 mM Na2CO3, 35 mM NaHCO3 (pH 9.7), overnight at 4°C. The plates were blocked in PBS containing 2% (w/v) skimmed milk, and then the sera were added in twofold serial dilutions. On all plates an in-house positive reference serum pool was added in twofold dilutions. Specific antibodies were detected following incubation with rabbit anti-mouse IgG conjugated to horseradish peroxidase (HRP) (Zymed/Invitrogen) by addition of the TMB Plus Ready-to-use substrate (Kem-En-Tec, Taastrup, Denmark). The reaction was stopped at the optimal color development with 0.2 M H2SO4, and absorbance (OD) was read at 450 nm with wavelength correction at 620 nm to correct for optical imperfections including air bubbles in the plates. The antibody used for detection was HRP-conjugated anti IgG (Invitrogen, Frederick, MD, USA).

The relative concentration of HA-specific IgG was calculated based on the positive reference standard curve (top dilution set to 100 AU/ml). The corrected OD values were log–log transformed and the standard curve fitted manually to contain as many points as possible on a straight line (*r* > 0.99 or *n* ≥ 4). Only OD values for the dilutions of a sample within this linear interval of the reference were used to calculate the antibody titer of that sample.

Slope (*b*), intercept (*a*), and correlation (*r*) were calculated for the best fitted standard curve according to the method of least squares. According to the formula: *y* = *bx* + *a* (where *y* = OD and *x* = dilution), the relative dilution of each OD values for the samples can be calculated according to this formula: Anti-log[(((ODsample − ODbackground)) − Standard-intercept(*a*))/ Standard-slope(*b*)] = dilution.

The absolute concentration (titer) is then calculated by adjusting the calculated dilution with the plate dilution. This adjusted dilution is then divided into the concentration of the standard.

### Hemagglutination Inhibition

Blood was drawn from animals 3 days before and 7 days after infection with influenza and serum was prepared and stored at −20°C until use. HI titers were determined by endpoint titration. In 96-well round-bottom plates, serum was diluted twofold in 50 µl PBS and incubated with 50 µl 4 HA U influenza A/ Brisbane/59/2007 for 30 min. After incubation, 50 µl 0.5% chicken RBC was added and incubated for further 60 min before hemagglutination was scored as the well with the most diluted serum that can prevent hemagglutination. Controls included were serum control (50 µl serum + 50 µl PBS + 50 µl RBC), virus control (50 µl virus + 50 µl PBS + 50 µl RBC), and erythrocyte control (100 µl PBS + 50 µl RBC). All samples were analyzed in duplicate wells.

# Quantitative Real-time Polymerase Chain Reaction (qRT-PCR)

Viral loads in nasopharyngeal swab samples collected at days 1, 2, 3, 4, and 7 were determined using qRT-PCR. The influenza M gene was absolutely quantified while the housekeeping gene, murine glyceraldehyde-3-phosphate dehydrogenase (mGAPDH), was included in the amplifications to address inter-PCR variations. The reliability of this assay was previously validated through correlation with infective titers as determined by plaque assay.

The protocol has previously been described in Andersson et al. (30). In brief, the Brilliant II qRT-PCR, 1-step kit (Agilent Technologies) was used. The following was added to each well: 100 ng purified RNA, 12.5 µl master mix, 0.5 µl of each primer (10 µM), 0.5 µl mGAPDH probe (10 µM), 0.75 µl M probe Table 2 | Primers and probes used for quantitative real-time polymerase chain reaction.


*Primers and probes were synthesized at TAQ Copenhagen A/S.*

*FAM, carboxyfluorescein; BHQ-1, BlackHoleQuencher-1; HEX, hexachlorofluorescein; mGAPDH, murine glyceraldehyde-3-phosphate dehydrogenase.*

(10 µM), 0.375 µl (1:500) reference dye, and 1 µl RT/RNase block enzyme mixture diluted in Milli-Q water to a total volume of 25 µl. See **Table 2** for primers and probes used.

Samples were run on an Mx3000P Real-time QPCR instrument (Agilent Technologies) according to the following settings: 30 min/50°C, 10 min/95°C followed by 40 cycles of 30 s/95°C, 1 min/58°C, and 30 s/72°C. All samples were run in triplicates and standard curves in duplicates. The copy number was determined with an optical density read out. The obtained results were analyzed using the Stratagene MxPro software.

### Gross Pathology and Histopathology

Four days after challenge, all ferrets were euthanized and necropsied. The ferrets heads and a segment of the pharynx were fixed in 10% neutral-buffered formalin. Following decalcification of the heads by ethylenediaminetetraacetic acid (EDTA + NaOH), a segment of the nasal cavity including the nasal concha was obtained at the level of the upper canine tooth. The tissues were routinely processed, embedded in paraffin, and sectioned. From each ferret, one section of the pharynx and one section of nasal cavity were stained with hematoxylin and eosin (H&E). In addition, one section of the nasal cavity was used for IHC detection of CD3. All processing and staining of the section was done at Nordic Biosite. All sections were examined by light microscopy by a pathologist blinded to the groups. In H&E-stained section of the nasal cavities, the following were scored: thickness of ventral nasal concha (score 1–4; 1, normal; 2, slightly increased; 3, moderately increased; 4, severely increased), lack of normal ciliated respiratory epithelium (score 1–4; 1, lacking in ≥0–25%

### REFERENCES


of the section; 2, lacking in >25 to ≤50% of the section; 3, lacking in >50 to <90% of the section; 4, lacking in 90–100% of the section), the degree of inflammatory cell infiltration in lamina propria and the amount of exudate in the nasal cavity (score 0–4; 0, absent; 1, minimal; 2, mild; 3, moderate; 4, severe), the number of lymphocytes, neutrophils, and plasma cells (score 0–3; 0, absent; 1, few; 2 moderate; 3, many).

Immunohistochemical staining was performed by incubating tissue sections with monoclonal antibodies against CD3, followed by incubation with isotype-specific antibodies (DAKO). In IHC stained sections, the number of CD3(+) T-cells were scored (score 0–3; 0, absent; 1, few; 2, moderate; 3, many).

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of "Danish Animal Experimentation Act and the European Convention for the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes." The protocol was approved by the "Danish Animal Care and Ethics Committee" (permit no. 2012-15-2934-00503).

## AUTHOR CONTRIBUTIONS

DC, JC, KK, AT, and PA designed the study. DC, JC, KK, LI, and KE prepared laboratory work and analyzed the data. DC, KK, and PA interpreted the data and drafted the manuscript. All authors approved the final version.

### ACKNOWLEDGMENTS

We thank EC-FP7 project ADITEC (Grant agreement #280873), EC-FP7 project UNISEC (Grant agreement #602012), Innovative Medicines Initiative Joint Undertaking project BIOVACSAFE (Grant agreement #115308), and the Danish Strategic Research council project Centre for Nano Vaccines (Case #09-067052), and Independent Research Fund Denmark (Grant #4183- 00098B). Thanks to Rune Fledelius Jensen, Camilla Haumann Rasmussen, Allan Lykke Hansen, Tenna Vaishali Benthin, Deanna Bardenfleth, and Pernille Keller Andersen for excellent technical help. We also thank Max Per Kristiansen for help with antibody titer determination and Jorgen de Jonge for discussions on how to assess airway pathology in the ferret model.


**Conflict of Interest Statement:** DC, KK, and PA are employed by Statens Serum Institut, a nonprofit government research facility, and holds patents related to CAF01, all rights has been assigned to Statens Serum Institut. All other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer RR is currently coorganizing a Research Topic with one of the authors PA and confirms the absence of any other collaboration.

*Copyright © 2018 Christensen, Christensen, Korsholm, Isling, Erneholm, Thomsen and Andersen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# AS03- and MF59-Adjuvanted influenza vaccines in Children

*Amanda L. Wilkins1 \*, Dmitri Kazmin2 , Giorgio Napolitani3 , Elizabeth A. Clutterbuck <sup>4</sup> , Bali Pulendran2,5,6,7, Claire-Anne Siegrist <sup>8</sup> and Andrew J. Pollard4*

*<sup>1</sup> The Royal Children's Hospital Melbourne, Melbourne, VIC, Australia, 2Emory Vaccine Center, Emory University, Atlanta, GA, United States, 3Medical Research Council (MRC), Human Immunology Unit, University of Oxford, Oxford, United Kingdom, 4Oxford Vaccine Group, Department of Paediatrics, University of Oxford, The NIHR Oxford Biomedical Research Centre, Oxford, United Kingdom, 5Department of Pathology, Emory University School of Medicine, Atlanta, GA, United States, 6Department of Pathology, and Microbiology & Immunology, Stanford University, Stanford, CA, United States, 7 Institute for Immunology, Transplantation and Infection, Stanford University, Stanford, CA, United States, 8University of Geneva, Geneva, Switzerland*

Influenza is a major cause of respiratory disease leading to hospitalization in young children. However, seasonal trivalent influenza vaccines (TIVs) have been shown to be ineffective and poorly immunogenic in this population. The development of liveattenuated influenza vaccines and adjuvanted vaccines are important advances in the prevention of influenza in young children. The oil-in-water emulsions MF59 and adjuvant systems 03 (AS03) have been used as adjuvants in both seasonal adjuvanted trivalent influenza vaccines (ATIVs) and pandemic monovalent influenza vaccines. Compared with non-adjuvanted vaccine responses, these vaccines induce a more robust and persistent antibody response for both homologous and heterologous influenza strains in infants and young children. Evidence of a significant improvement in vaccine efficacy with these adjuvanted vaccines resulted in the use of the monovalent (A/H1N1) AS03-adjuvanted vaccine in children in the 2009 influenza pandemic and the licensure of the seasonal MF59 ATIV for children aged 6 months to 2 years in Canada. The mechanism of action of MF59 and AS03 remains unclear. Adjuvants such as MF59 induce proinflammatory cytokines and chemokines, including CXCL10, but independently of type-1 interferon. This proinflammatory response is associated with improved recruitment, activation and maturation of antigen presenting cells at the injection site. In young children MF59 ATIV produced more homogenous and robust transcriptional responses, more similar to adult-like patterns, than did TIV. Early gene signatures characteristic of the innate immune response, which correlated with antibody titers were also identified. Differences were detected when comparing child and adult responses including opposite trends in gene set enrichment at day 3 postvaccination and, unlike adult data, a lack of correlation between magnitude of plasmablast response at day 7 and antibody titers at day 28 in children. These insights show the utility of novel approaches in understanding new adjuvants and their importance for developing improved influenza vaccines for children.

Keywords: adjuvant, influenza vaccine, MF59, AS03, children

### INFLUENZA AND NON-ADJUVANTED VACCINES

Influenza causes significant morbidity and mortality worldwide and it is estimated that 20–30% of children become infected with influenza each year (1). Although influenza infection often results in a self-limiting illness, young children are at increased risk of secondary pneumonia, hospitalization, and death (2, 3). The global incidence of influenza-associated acute lower respiratory infections (ALRI)

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Raffael Nachbagauer, Icahn School of Medicine at Mount Sinai, United States Aldo Tagliabue, ALTA, Italy*

*\*Correspondence: Amanda L. Wilkins tolson.amanda@gmail.com*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 27 November 2017 Published: 13 December 2017*

### *Citation:*

*Wilkins AL, Kazmin D, Napolitani G, Clutterbuck EA, Pulendran B, Siegrist C-A and Pollard AJ (2017) AS03- and MF59-Adjuvanted Influenza Vaccines in Children. Front. Immunol. 8:1760. doi: 10.3389/fimmu.2017.01760*

**221**

in children less than 5 years old has been estimated at 20 million in 2008 (13% of all cases of pediatric ALRI) (4). In the same year, an estimated 28,000–111,500 deaths in children less than 5 years old were attributable to influenza-associated ALRI. The mortality burden is seen most in developing countries, where 99% of these deaths occurred. Moreover, influenza-related illness is responsible for substantial economic burden, contributing to an increasing number of outpatient appointments, missed school and antibiotic use in children (5, 6). Laboratory-confirmed influenza-related medical attendances in children less than 5 years of age have been reported as high as 27 emergency department visits per 1,000 children and 95 outpatient visits per 1,000 children (7).

Prevention of influenza is most effectively provided through vaccination and would ideally offer cross protection against drifted non-vaccine influenza virus strains. Children play an important role in transmission of influenza virus therefore the vaccination of this population is not only an important prevention strategy for direct protection but also indirect protection for the wider population (6, 8, 9). Licensed non-adjuvanted influenza vaccines for children include split or subunit inactivated influenza vaccines (IIV) and the live-attenuated influenza vaccine (LAIV). Young children are often naive to the influenza virus, have not previously been vaccinated, and are therefore unprimed. For this reason it is recommended that children receive two doses, 28 days apart, of an influenza vaccine in the first influenza season they receive immunization. Both IIV and LAIV have significant limitations for use in the pediatric age group. IIV is not licensed for use before the age of 6 months; it is poorly immunogenic in younger children, with an efficacy of 59% against confirmed influenza infection and 36% effectiveness against influenza-like illness (ILI) in children 6 months to 2 years old (10). Additionally, IIV provide poor cross-protection for mismatched influenza virus strains (10). LAIV has significantly better efficacy than IIV, with 55% fewer cases of confirmed influenza following LAIV compared with IIV (11). However, LAIV is not recommended for children less than 2 years of age due to increased rates of wheezing episodes postvaccination (11).

Currently, trivalent or quadrivalent influenza vaccination is recommended only for high-risk children in most countries—or for all children aged 6 months and older in certain developed countries including the US, UK, Australia, and Canada (12–15). In the UK the LAIV is funded through the routine immunization schedule for children aged 2–11 years (though roll out of the program to all these age groups is not yet complete) with evidence from surveillance data demonstrating direct protection in children against influenza infection and hospitalization (13, 16, 17). Conversely, LAIV has shown poor effectiveness in children in the US over the last three influenza seasons and was not recommended for the 2016–2017 season (18). Irrespective of these recommendations, uptake is suboptimal. Recent surveillance in the USA estimates only 26% of laboratory-confirmed influenza-associated pediatric deaths in children 6 months to 17 years having received an influenza vaccine prior to their illness (19). The common perception that influenza is a benign illness compared with other childhood infections, and the partial efficacy of influenza vaccines in young children limit its recommendations, its promotion and thus its uptake.

The limitations of IIV and LAIV in young children and the poor vaccination coverage result in one of the highest risk groups for influenza-related comorbidities receiving inadequate prevention and subsequent lack of herd protection for the remaining population. An approach to improving protection in children is the addition of adjuvants to the traditional IIV. Adjuvants are designed to enhance the immunological response to a vaccine and, when used for influenza vaccines, have afforded antigen dose sparing and improved cross-protection against non-vaccine influenza virus strains. A range of adjuvant formulations have been developed and there has been progress toward fully understanding the mechanisms involved their action in recent years. Historically there have been challenges involving the use of adjuvanted influenza vaccines in humans due to unacceptable adverse events (20–22). New and improved adjuvant systems have overcome this issue and there have been a number of approved adjuvanted influenza vaccines for children, including prepandemic, pandemic, and seasonal vaccines. This article provides an overview of the oil-in-water-adjuvanted influenza vaccines in children, MF59 and adjuvant systems 03 (AS03) (**Table 1**), highlighting their ability to provide improved protection for children against influenza.

### ADJUVANTED INFLUENZA VACCINES

### MF59-Adjuvanted Influenza Vaccines

MF59 is an oil-in-water emulsion composed of squalene and two surfactants, Tween 80 and Span 85. Squalene is a naturally occurring oil synthesized in the human liver and is a direct precursor to cholesterol (23). The Chiron Vaccines company developed MF59 and it was first licensed as part of a seasonal influenza vaccine for the elderly population in Italy in 1997. Over 100 million MF59 containing vaccines have been distributed in over 30 countries around the world. The MF59-adjuavanted inactive trivalent influenza vaccine (TIV) [Fluad®, MF59-adjuvanted trivalent influenza vaccine (ATIV), Novartis Vaccines] contains 15 μg of each influenza strain surface antigen and the MF59 adjuvant and is administered as a 0.5 ml dose. It is licensed for adults aged 65 years and over. Fluad Pediatric®, a 0.25 ml dose, has now been licensed for children aged 6 months to 2 years in Canada since 2015. Two MF59-adjuvanted monovalent A/H1N1 pandemic influenza vaccines (Focetria® and Celtura®, Novartis Vaccines) were licensed for children during the H1N1 influenza pandemic in 2009. Focetria® is an egg-based inactivated subunit vaccine and Celtura® a cell-culture-based inactivated subunit vaccine.

### AS03-Adjuvanted Influenza Vaccines

The AS03 adjuvant is an oil-in-water emulsion composed of squalene, polysorbate 80 and α-tocopherol (vitamin E). AS03 was first used in the prepandemic H5N1 vaccine Prepandrix (GlaxoSmithKline Biologicals s.a.) and was subsequently included in two influenza A(H1N1)pdm09 pandemic vaccines— Pandemrix®, GlaxoSmithKline Biologicals s.a., and Arepanrix®, GlaxoSmithKline Inc. Two AS03 formulations with differing amounts of tocopherol, AS03A (11.86 mg tocopherol) and AS03B (5.93 mg tocopherol), were used in the full dose and half dose


Table 1 | AS03- and MF59-adjuvanted vaccines for children.

*GSK, GlaxoSmithKline.*

vaccines, respectively. The 2009 A(H1N1) influenza pandemic was the first time the global deployment of a pandemic influenza vaccine had been undertaken. The benefit of using the AS03 adjuvant as part of a pandemic vaccine is its ability to induce high antibody titers with a reduced antigen dose (3.75 or 7.5 μg per strain compared with 15ug per strain in conventional TIV), making it possible to meet the global demand. Pandemrix® was accepted for fast track authorization and had been given to less than 200 children aged 3–9 years before it was licensed (24). Approximately 4.7 million doses of AS03-adjuvanted A(H1N1) vaccines have been administered to children since 2009 (25).

### IMMUNOLOGY

### Effect of Adjuvants on Systemic Antibody Responses to Influenza Hemagglutinin (HA)

A number of clinical trials have demonstrated that the seroprotection induced by the MF59-adjuvanted vaccines is superior to TIV, even in the very young (26–33) or the elderly (34). The threshold of protection was defined in immunized adults as HAI titers ≥40 (50% protection from reinfection) or a fourfold rise from baseline (35). However, this was proven to be insufficient to protect infants and young children and new protective thresholds were defined by Black et al. (36). Here, children aged 16–72 months received two doses of an MF59 ATIV (Fluad®) or a TIV vaccine (Influsplit®). Follow up for any influenza like illness was confirmed by RT-PCR. Immunogenicity and surveillance data collected allowed the investigators to model a protective HAI titer that would give 80% (≥330) and 90% (≥629) protection in this age group. In a subsequent, similar study of children aged 14–24 months, 100% of children achieved thresholds of both ≥330 and ≥629 after two doses of MF59 ATIV in response to A/H1N1 and A/ H3N2 vaccination in comparison with those receiving TIV (Imuvac®) where 8 and 47% of children achieved ≥330 to H1N1 and H3N2, respectively (29).

It seems that in all age groups, primed individuals respond more robustly to both TIV and ATIV vaccines (37–39).

### Adjuvants Induce Recruitment of Innate B Cells and IgM Production

Murine models have shown that influenza viruses cause inflammation of epithelial cells in the respiratory tract (40, 41). These innate inflammatory signals trigger local and systemic responses, resulting in a protective immune response against influenza virus (40, 41) and adaptive T and B cell responses (42).

The MF59 adjuvant has been shown, in mice and humans, to induce proinflammatory chemokines such as CXCL10 and cytokines [independently of type-1 interferon (IFN)] at the injection site, with recruitment of CD11b<sup>+</sup> blood cells (43–45). These chemokines and cytokines promote more efficient antigen uptake by, and differentiation of, monocytes, macrophages and granulocytes, and differentiation of monocytes into immature dendritic cells (DCs) (46). MF59 also primes for enhanced processing and presentation of antigen for broader recognition of epitopes (47, 48).

There is extensive evidence, mainly from murine studies, to show that influenza HA-specific IgM can mediate protection from initial infection and re-infection (42, 49–54). In mice, innate B1 cells have a clearly defined role in systemic and local protection through spontaneous, steady state secretion of natural IgM antibodies (49, 51, 52). Murine models of influenza infection noted that B1a cell secretion of viral–specific IgM is enhanced locally, but not systemically, following infection (52, 53). While the systemic response was mediated by conventional, B2 cell-derived IgM (49, 51, 52). In human infants (aged 14–24 months), systemic serum IgM, IgM-plasma cells (PCs) and IgM-memory B cell responses, specific to vaccine H3N2 and H1N1 components, were observed 1 month after two doses of TIV or ATIV, although no difference was observed between the groups (29).

The role of IgM and innate B cells in human responses is not clearly understood since HAI titers are a measure of total Ab function and not just IgG; however, the contribution of IgM in influenza virus neutralization assays has been demonstrated (55, 56), so it could be proposed that IgM also has a role in hemagglutination.

### Memory B Cells (BMEM), Antibody Secreting Cells (ASCs), and Cross-Reactive Antibodies Are Enhanced by Adjuvant

Influenza infection induces a mucosal (nasopharygeal lamina propria) B cell, antibody and cellular response which is maintained over time (37). In mice it was shown that BMEM localized to the lung could provide protection from reinfection, while the bone marrow resident long-lived plasma cell (LLPC), spontaneously secreted antibody to provide immediate protection (57).

There seems to be some separation of the mucosal response from the systemic response (peripheral blood and tonsils) induced by intramuscular (i/m) immunization (37, 58). However, detection of PCs or ASC in peripheral blood may be the best marker of recent infection or response to immunization in naive subjects (59).

While there is very little information in the literature on BMEM and ASC responses in infants and very young children there are some age group comparisons. The ASC responses in adults versus children (aged 2–3 years) immunized with TIV were similar between primed children and adults, but in unprimed children only the IgM-ASC response was equivalent to adults, while the IgG and IgA-ASC response was significantly lower (60). IgG-ASC responses have also been detected in other age groups following either TIV [in children aged 6 months to 4 years (39)] or TIV versus MF59 ATIV immunization [in children aged 14–26 months (29), and in adults (61)]. However, even with adjuvant use in younger children, the day 7 peak in frequency of IgG-ASC in children is less than in adults, which may be related to maturity of the immune system and previous priming (61). The peak in ASC at day 7 postimmunization is almost always referring to IgG response; however, similar peaks in IgA, and to a lesser extent, IgM are also described (37).

Induction of IgG-BMEM has been observed in both primed and unprimed adults, although the magnitude of the response was enhanced in the presence of MF59 (38). In children aged 14–24 months both TIV and ATIV vaccines induced a greater frequency of IgM-BMEM than IgG-BMEM. However, the functional, HAI, responses were more robust and long lived following MF59-ATIV than after TIV (29).

Even in the absence of adjuvant, Influenza HA induces polyclonal stimulation of B cells and production of IgM antibodies, some of which are cross-reactive with different flu strains (62–64). *In vitro* studies have revealed HA stalk-specific antibodies that show different binding patterns, which indicates multiple conserved epitopes (65).

Specificity of ASC and antibody in response to TIV is more strain specific, with little cross-reactivity in comparison with controlled infection (H3N2) where ASC were reactive with a number of different strains (66). Repeated exposure *via* TIV immunization also limited induction of cross-reactive stem antibodies while response to the immunodominant head structure increased (67), suggesting that primary responses (in younger cohorts) induce stem antibodies while the recall response, mediated by BMEM and LLPCs (in older cohorts), is to the head structure (68).

Thus, the presence of preexisting HA-specific-BMEM may reduce (or focus) the breadth of subsequent Ab and ASC specificity and presence of high titers of serum HA-specific antibodies corresponds with a poorer ASC response (61, 69–71). It was suggested that cross-strain responses could be improved if strains included in the seasonal vaccines varied more frequently (70). However, in the presence of MF59-adjuvanted vaccines, more robust BMEM responses to clade mismatched H5 viruses (38) and A/strain group mismatched viruses (H5N1 vs. H7N9) were achieved than with TIV alone (69).

The cross-reactive antibodies undergo affinity maturation following immunization (H1N1-pdm09), which correlates with increased expression of activation-induced cytidine deaminase (72). MF59 and AS03 have been shown to enhance the production of cross-reactive (38, 73) and strain-specific antibodies compared with non-adjuvanted versions of the same influenza A/strains (46, 61, 65, 74–83). A similar effect was also seen with rintatolimod (a TLR-3 adjuvant), given intranasally with LAIV against H5 and H7 strains (84). However, this approach was not as successful for B/strains of the virus (79).

The role of somatic hypermutation, during memory B cell development, in broadening the cross-specificity of preexiting memory was described by Fu et al. (85) who demonstrated acquisition of H5 specificity following a single mutation of an H1/H3-specific germline VH sequence (IGVH3-30, Mab 3I14) directed against the HA stem.

The A/strain-specific cross-reactive antibodies have been identified following immunization in adults, the elderly (47, 75), and in children (31, 86). Generation of these cross-reactive antibodies is one of the main aims of new influenza vaccine development in order to help protect against future, related, pandemic strains (38, 61, 64, 71).

## Role of T Follicular Helper cells (Tfh) and Enhancement by Adjuvant

Kopf et al. (53) suggested a primary role for B1 cell-derived IgM may be to enhance CD4<sup>+</sup> T cell priming at sites of infection. IgM-opsonized viral antigen may be captured by DCs that can prime T cell responses (53). IgM-Ag complexes may also flow back to draining lymph nodes (LNs), enhancing viral-specific CD4<sup>+</sup> T cell-B cell interactions and subsequent germinal center formation (53).

Thus the enhanced recruitment of antigen presenting cells induced by MF59 to sites where these CD4<sup>+</sup>T cell-conventional B cell interactions are occurring may partly explain the enhanced IgG memory responses achieved by ATIV vaccines and that the innate and adaptive mechanisms are required to achieve protective responses.

The role of CD4<sup>+</sup> T cells in supporting antibody responses against influenza HA has been accepted for many years (42). However, in recent years a subset of CD4<sup>+</sup> T cells, known as T follicular helper (Tfh) cells have been strongly implicated to be involved in robust, long-lived antibody responses to influenza infection (87, 88) and immunization with TIV (89) and ATIV (90).

Tfh cells differentiate under certain conditions at the T-B cell border of the lymphoid follicles and require proinflammatory conditions (91). Activated B cells secrete IL-6 which induces Bcl6 expression and enhances IL-21 secretion by CD4<sup>+</sup> T cells. IL-21 triggers differentiation of CD4<sup>+</sup> T cells into Tfh cells which secrete IL-21, maintaining their function.

There is some redundancy and only mice deficient in both IL-6 and IL-21 fail to make Tfh responses (91). IL-7 has also been implicated in Tfh development (87).

Immunization of adults and children with TIV induced robust Tfh responses by day 7 postimmunization, but only in the presence of immune memory (89). In naive children there was limited Tfh response—as was observed in infant mice (92). The frequency of Tfh cells also correlated with rise in antibody HAI titers (89), and immunization using LAIV induced circulatory (c)Tfh responses that strongly correlated with increased antibody avidity and expansion of HA-specific Tfh clones (93, 94). The cTfh population was identified in the peripheral blood prior to immunization and characterized as CXCR5<sup>+</sup>PD 1<sup>+</sup>ICOS<sup>+</sup>CD38<sup>+</sup>, with higher expression of CD27, CD25, CD28, CTLA4, PD1, Helios, and Ki67, but lower CD127 than total CD4<sup>+</sup> T cells (94).

While influenza infection and administration of non-adjuvanted influenza vaccines induced robust Tfh responses in adults, addition of MF59 as an adjuvant significantly enhanced the response (95) with expansion of HA-specific Tfh (CD4<sup>+</sup>ICOS<sup>+</sup>, CD4<sup>±</sup>ICOS<sup>+</sup>CXCR5<sup>+</sup>IL-21<sup>+</sup>) by day 7 postimmunization that highly correlated with HAI titers 1 and 6 months later (95).

Previous infection or repeated immunization led to competition for virus-specific CD4<sup>+</sup> T cells limiting naive Tfh expansion, but inducing expansion of preexisiting, clonal populations that subsequently resided in a memory population of ICOS-CD38 cTfh (94, 96).

Thus it could be proposed that cross-reactive, IgM antibodies, produced by innate B cells, trap antigen on DCs within the lymphoid follicles. This antigen activates B cells and is presented to CD4<sup>+</sup> T cells, enhancing Tfh responses and B cell help, leading to robust, highly avid IgG antibody production. In naive infants, lacking preexisting pools of memory Tfh, the response to influenza would be predominated by IgM, but priming of a Tfh population would occur following infection or immunization. Thus subsequent immunization induces protective IgG responses. In adults, primed by infections, Tfh memory pools exist enabling strong IgG responses. The importance of adjuvants, such of MF59, therefore, may be to induce cross-reactive cellular subpopulations with each dose, helping to avoid narrowing of HA-specificities present within the memory populations, increasing likelihood of protection against future, related pandemic strains of influenza virus. Accordingly, MF59 was described in mice as mediating its adjuvanticity on influenza HA by promoting Tfh and thus Germinal Center responses in adult and early life—but not to fail inducing Tfh cells and thus humoral responses in neonatal mice (92).

### Innate Responses and Transcriptomes

A protective role for IFN-related genes during influenza infection has been demonstrated in mice (97) and in humans (98). A study of mice knocked-out for the IFN-inducible transmembrane protein (IFITM) demonstrated the importance of this gene in protection from severe influenza infection with enhanced pathogenesis and overproduction of proinflammatory cytokines (97). A human minor *IFITM* allele (SNP rs12252-C) was also associated with hospitalization in pandemic A(H1N1)pdm09 influenza patients (97). The SNP rs12252-C allele was further investigated and associated with influenza infection severity in a study of Chinese patients infected with severe pandemic A(H1N1)pdm09 (98).

Although these findings suggest a role of type I IFN in limiting viral replication, these cytokines might also play a role in modulating adaptive immune responses capable of eliciting better protection against reinfection. Indeed mouse studies have shown that adjuvants triggering innate immune responses *via* activation of innate immune receptors such as TLR4 and TLR7 are superior in inducing protective immunity when compared with vaccines unable to do so (99).

Squalene-based adjuvants such as MF59 and AS03 are also capable of triggering innate immune responses *via* a yet unknown mechanism.

A study (100) comparing the immune response induced by vaccines containing alum, TLR7 agonists and MF59 found that MF59 is far superior to alum alone in its capacity to promote immune cell infiltration to the injection site in the muscle, resulting in antigen uptake by neutrophils, monocytes, and myeloid and plasmacytoid DCs and migration exclusively to the vaccine-draining LNs. This resulted in priming of higher numbers of antigen-specific CD4<sup>+</sup> T cells in the vaccine-draining LNs, increased T follicular helper cell differentiation and germinal center formation, and better antibody responses. Although this study failed to identify a type 1 IFN response in mice immunized with MF59, this adjuvant has been shown to induce increased IFN expression in infants (29). This innate response has been previously described in adults and is associated with stronger antibody responses (101). A similar observation was made in children aged 6 months to 14 years vaccinated with TIV or LAIV, where an association was found between upregulation of IFN genes at day 1 post-TIV and enhanced antibody responses, but only in children more than 5 years of age (102). In younger children (aged less than 5 years) IFN responses were not observed until day 7 post-LAIV (102).

A recent study (103) compared innate and adaptive immune responses in hepatitis B virus (HBV) naive individuals following receipt of a vaccine containing HBV surface antigen (HBsAg) adjuvanted with one of the following: AS01 [TLR4 ligand 3-*O*-desacyl-4′-monophosphoryl lipid A (MPL) and the purified saponin QS-21], AS03 (α-tocopherol and squalene in an oil-in-water emulsion), AS04 [MPL adsorbed on aluminum salt (AlPO4)], or Alum/Al(OH)3. Consistent with a role of innate responses in vaccine immunogenicity, the authors found that the adjuvanted vaccines capable of eliciting more pronounced antibody and CD4 T cell responses to HBsAg (AS01 and AS03), also induced an early mobilization of neutrophils and monocytes. Following vaccination with the AS01-adjuvanted vaccine, accumulation of cytokines, specifically IL-6, in serum was detectable as early as 3–6 h after vaccination. In addition, upregulation of IFN response genes was observed following the second dose of the AS01-adjuvanted vaccine but not following first or second dose of the other adjuvants. Notably, an increase in innate response and immunogenicity also correlated with more pronounced reactogenicity.

Recently, systems biological approaches have been used to define molecular signatures induced by vaccination in humans, and to understand their mechanisms of action. A systems-based approach was used to define the molecular signatures in response to vaccination with the live attenuated yellow fever vaccine (YF-17D), and to use such signatures to predict the immunogenicity of this vaccine (104). This offered proof of concept evidence of the utility of systems-based approaches in predicting vaccine immunity (104). In an independent study of the response to YF-17D, Sekaly and colleagues undertook a similar approach and obtained similar results (105). Subsequently, this approach has been extended to other vaccines such as the seasonal influenza vaccine (106–109), meningococcal vaccines (101) and shingles vaccines (110, 111), and in the infant population (29, 102). Importantly, recent studies have extended this approach to identifying signatures that predict vaccine-induced protection from disease (112, 113).

In addition, systems vaccinology approaches have been used to study responses to adjuvanted influenza vaccines. Both oil-in-water-based adjuvants, AS03 and MF59, induce specific transcriptional responses. These responses have been analyzed both in non-clinical mouse models and in clinical cohorts. In a non-clinical setting, upon injection of AS03 adjuvant, potent transcriptional responses have been observed both at the site of the injection and in the draining LNs as soon as 4 h postinjection (114). These changes affect a large number of chemotactic chemokines, believed to be involved in the recruitment of monocytes (CCL2, CCL3, CCL7), neutrophils (CXCL1, CXCL5, CXCL2, CSF3), eosinophils (CCL5), and DCs (CXCL9, CXCL10, CCL3, and CCL4). Of interest, while similar patterns of gene expression changes was observed in draining LNs, these changes tend to be more transient, with a peak at 4 h and diminishing signal at 24 h postinjection (114). It was also demonstrated that these responses were largely mediated by the α-tocopherol component of AS03, both in terms of the kinetics of the response, and the spectrum of chemokines being induced. *In vitro* studies identified monocytes and macrophages as the primary target cell type for α-tocopherol, and responsible for the production of chemokines in response to AS03 stimulation (114). In a clinical setting, a systems biology analysis of the effects of AS03 on responses to influenza vaccine has not yet been done in pediatric cohorts. However, in adults transcriptional responses in sorted cell populations were compared in cohorts receiving adjuvanted and non-adjuvanted H5N1 split-virion vaccine with high temporal resolution (115). This analysis demonstrated distinct gene expression patterns specific to distinct cell populations in peripheral blood, although different immune cell types responded at different time points. These responses were shown to correlate with cytokine production and antibody response (115). In another study, a large cohort of adult volunteers was followed longitudinally pre- and postadministration of A(H1N1) AS03 adjuvanted vaccine (Pandemrix®) (116). Early postvaccination the authors observed a transient decrease of expression of a large number of T cell-specific transcripts, accompanied by a strong upregulation of a large number of IFN response genes. Serum IFN gamma levels were accordingly elevated. Of special interest, age was a factor in gene expression patterns observed at day 1 postvaccination, with volunteers aged 35 or older demonstrating altered expression of several transcripts involved in early responses (116). While no pediatric subjects were included in this study, these results are relevant in light of the striking differences between responses to the same vaccine in infants and adults, discussed below.

Effects of MF59 on transcriptome have been extensively studies both in clinical setting (29) and mouse models (43, 44, 117), and reviewed by Olafsdottir et al. (118). MF59 induced strong localized transcriptional response, far exceeding in magnitude the response to CpG and Alum adjuvants (44), including the induction of a wide spectrum of cytokines and cytokine receptors, which included all (Alum), or nearly all (CpG), cytokines induced by other adjuvants. This potent transcriptional response was accompanied by a stronger recruitment of MHC class II and CD11b bearing cells to the site of injection (44). In a later study it was demonstrated that the observed effects on localized gene expression, cellular recruitment, antigen-specific humoral and T-cell responses, and antigen translocation to draining LN were all due to the combination of components of the adjuvant, as none of the individual components were able to elicit comparable responses (117). Further dissecting the functional transcriptional responses at the site of injection, Caproni et al. were able to demonstrate that the induction of proinflammatory genes, as well as genes relevant to transendothelial leukocyte migration correlated with the recruitment of CD11b<sup>+</sup> cells to the site of injection, and antibody and cellular responses (43). Of interest, in a mouse model, MF59 induced very weak IFN type I response, and IFN signaling through its cognate receptor was not required for mounting potent humoral response (43).

These results, however, were not recapitulated in a pediatric clinical study in which the effects of MF59 ATIV were compared with those of an unadjuvanted TIV (29). Indeed, in infants it was shown that MF59-adjuvanted TIV induces a strong and transient expression of a large number of IFN type I response genes, and that the induction of these genes at day 1 postboost vaccination tracked positively with HAI responses. Overall, MF59 ATIV induced a much stronger transcriptional response at day 1 postboost vaccination, although the magnitude of this response was much lower than in the adult cohorts investigated in a separate and independent study. These early responses were dominated by a large number of gene modules relevant to DC activation, antigen presentation, monocytes, IFN and antiviral response (**Figure 1A**). A notable feature of the responses to vaccination in infants is the high heterogeneity of responses. The unadjuvanted vaccine was able to induce gene expression patterns characteristic of innate immune responses in only a minority of subjects. In contrast, inclusion of MF59 adjuvant allowed the number of transcriptional responders to be pushed much higher, with only one subject still failing to mount an innate transcriptional response. Innate type transcriptional response early postvaccination is a correlate of immunogenicity in adults, and these correlates were recapitulated in the infants receiving MF59 ATIV. In contrast, the weak induction of an innate transcriptional response by unadjuvanted vaccines results in the lack of such correlates at day 1 postvaccination. In fact, it takes 7 days for the vaccines in an unadjuvanted arm to develop the spectrum of transcriptional correlates of immunogenicity similar to correlates that are evident in adjuvanted arm as early as day 1 postvaccination (**Figure 1B**). Even then, the correlations between the expression of innate immunity gene modules and HAI titers are weaker and encompass fewer modules than in the adjuvanted arm. Together, these data suggest that unadjuvanted vaccines induce weak and delayed innate transcriptional response, resulting in lower HAI titers, while the inclusion of MF59 adjuvant allows the development of a stronger and more uniform innate response, and a spectrum of transcriptional correlates of antibody responses resembling correlates observed in adults. An intriguing observation made in this study was an inverse relationship between the correlates of antibody responses observed at day 3 postvaccination in infants and those in adults. Further studies are currently underway to directly compare the effects of MF59 ATIV

Figure 1 | (A) Functional enrichment of transcriptional responses to adjuvanted trivalent influenza vaccine (ATIV) and trivalent influenza vaccine (TIV) vaccination. Enrichment scores generated by the GSEA analysis represented by color panel in right upper corner. Each row represents a module, which are grouped according to the high level notation group, as illustrated on the side bar. Positive enrichment indicates a combined upregulation of genes included in a module. (B) Correlates of immunogenicity (HAI titers) for the ATIV and TIV vaccines. GSEA was performed with genes ranked by correlation of expression to HAI titers. Distance from the center of the spider plot corresponds to the average enrichment score across all modules included in the high level annotation group. Only modules with enrichments significant by *p* < 0.05 are included; enrichments for insignificantly enriched modules are set to zero. Blue translucent zones indicate negative enrichment.

in adults and infants and to shed more light on the molecular mechanism of action of MF59-adjuvanted influenza vaccine in pediatric population.

### EFFICACY AND EFFECTIVENESS

### MF59

The efficacy of MF59 ATIV against RT-PCR-confirmed influenza infection was assessed in a randomized controlled phase III study in Germany and Finland, including 4,707 children aged 6–72 months across two consecutive influenza seasons. MF59 ATIV was compared with a conventional TIV and control (meningococcal C conjugate vaccine) (119). The absolute vaccine efficacy against all influenza virus strains for the MF59 ATIV was 86% (95% CI, 74–93) compared with 43% (95% CI, 15–61) for the non-adjuvanted TIV, with a relative efficacy of 75% (95% CI, 55–87) (119). These results are supported by the superior immunogenicity of MF59 ATIV seen in the same study. Given the limited use of TIV and LAIV in children less than 2 years of age, TIV secondary to immunogenicity and LAIV secondary to safety concerns, it is particularly important to highlight that the efficacy of MF59 ATIV in children aged 6–24 months against matched strains was 75% (95% CI, 20–92) compared with 2% for TIV. It should be noted that the overall efficacy in this study predominantly reflects protection against H3N2 (94 of the 110 culture-confirmed influenza cases were H3N2). Based on these promising study results and others, an application for marketing authorization with the European Medicines Agency (EMA) was submitted to the Committee for Medicinal Products for Human Use (CHMP). During the application process concerns were raised by the CHMP regarding flaws in good clinical practice (GCP) in the clinical trial described above. This application was withdrawn in 2012 following the initial assessment by the CHMP due to ongoing unresolved issues regarding the concerns with compliance with GCP (120).

Currently, there are no postlicensure data available to assess the effectiveness of MF59 ATIV in children. However, this information should be available in the future given the recent licensure of MF59 ATIV in Canada for children aged 6 months to 2 years.

There are limited data on the effectiveness of the monovalent A/H1N1 MF59 adjuvanted vaccine in children during the 2009 H1N1 pandemic. The MF59 adjuvanted pandemic vaccine Focetria® was one of several pandemic vaccines available and millions of doses were administered to children mainly in Europe and Latin America. The effectiveness against ILI and laboratory confirmed A(H1N1)pdm09 infection was estimated in the Netherlands; however, the children included were only those with an underlying medical condition indicating their need for vaccination. In the total cohort there was a crude vaccine effectiveness (VE) against ILI of 17.3% (95% CI, −8.5–36.9%). For children aged between 5 and 19 years the adjusted VE against ILI was 51% (95% CI, −50–84%) and VE of children 4 years or less was not able to estimate due to the small number of children included (121). One report from Spain did not show significant VE in children aged 1–17 years against medically attended ILI (VE: 12%, 95% CI, −142–68%) (122). Lastly, a recently published systematic review of 2009 pandemic influenza A(H1N1) vaccines did not find any studies fulfilling inclusion criteria that included children who had received an MF59-adjuvanted vaccine (123).

The monovalent, cell culture-derived inactivated subunit vaccine adjuvanted with MF59 (Celtura®) gained local regulatory approval in four countries in Europe and Latin America (Germany, Switzerland, Chile, and Peru). However, there have not been any published studies reporting the effectiveness of this vaccine in children.

Overall, despite some promising results in this published literature, further efficacy and effectiveness data are required for MF59 adjuvanted vaccines to strengthen the argument for licensure in young children.

### AS03-Adjuvanted Pandemic Vaccines

The VE and efficacy in children of the AS03-adjuvanted A(H1N1) vaccine has been assessed in many studies, including three systematic reviews (123–125). The recently published systematic review by Lansbury et al. (123) reported pooled VE against laboratory-confirmed influenza from four studies with a total of 932 children (126–129). Pooled VE was estimated at 88% (95% CI, 69–95%, *p* < 0.0001) for adjuvanted H1N1 vaccines compared with 45% (95% CI, −13–73%, *p* = 0.83) for nonadjuvanted vaccines. This difference was statistically significant and the result did not differ if the studies included were limited to only those which measured VE from 14 days after vaccination. Pooled VE against hospitalization due to laboratory-confirmed influenza A(H1N1) illness was estimated at 86% (95% CI, 67–94%, *p* < 0.00001) using results from two studies (130, 131). The majority of studies assessing effectiveness were done in Europe and Canada, as described below.

One multinational RCT reported a vaccine efficacy against RT-PCR confirmed influenza of 76.8% (95% CI, 18.5–93.4%) in children aged 6 months to less than 10 years of age receiving an AS03-adjuvanted vaccine (Arepanrix®, GSK) compared with non-adjuvanted vaccine during 2010–2011 influenza season (132).

### Europe

The EMA recommended the pandemic vaccine (Pandemrix® in most countries) should, in the first instance, be provided to "risk groups," which included children less than 2 years of age, followed by "target groups" (which included children of all ages over 6 months). Target groups were offered vaccination in a staggered fashion throughout Europe, with children often being included in the early stages to help reduce transmission and provide indirect protection. For example, in the United Kingdom children 6 months and older who were in a clinical risk group were eligible for a two 0.25 ml dose course of Pandemrix® vaccine initially. This was changed in December 2009 due to the increasing numbers of hospital admissions in children, and all healthy children aged 6 months to 5 years were eligible for one 0.25 ml dose (133).

Effectiveness of AS03-adjuvanted H1N1 vaccine (Pandemrix®) against RT-PCR positive A(H1N1) influenza infection was extensively evaluated in Stockholm County, Sweden. In Sweden at the time of the study two 0.25 ml doses were recommended for children aged 6 months and older, with a vaccination coverage of 52% for children aged 6 months to 2 years receiving at least one vaccine dose and 70% in children aged 3–18 years. The estimated VE for children aged 6 months to 12 years during the peak weeks of the 2009–2010 season was 89–92%, with most children having only received one dose of the vaccine (134). VE against hospitalization (used as a surrogate indicator for severe disease) in children aged 6 months to 17 years due to influenza in the same population, adjusted for comorbid conditions, was 91% (130). Örtqvist et al. followed this cohort in a long term effectiveness study in the subsequent influenza seasons (135). During the 2010–2011 influenza season the adjusted VE for those vaccinated with Pandemrix® in 2009 was estimated at 91.7%; however, during the 2012–2013 season, there was no evidence of ongoing protection from previous vaccination. There was almost no H1N1 virus circulating in the 2011–2012 season therefore VE was unable to be estimated. Very few children received the seasonal influenza vaccine in seasons following the 2009 pandemic, likely due parents' concerns regarding safety, therefore the long-term effectiveness reported here is thought to truly reflect the 2009 vaccination program rather than vaccination in subsequent seasons.

Similarly impressive results were observed in an English case control study which estimated a VE against laboratory-confirmed influenza in children aged less than 10 years of 77% (CI, 11–94) and 100% (95% CI, 80–100) in those aged 10–24 years (126). In another case control study across England and Scotland including over 2000 children aged less than 15 years no vaccine failures occurred, therefore the VE estimate in children less than 15 years of age was 100% (95% CI, 74–100) (136).

The comparison of VE between these studies is difficult due to varying methods used to estimate VE, the broad range of age groups, differing number of doses administered and a variety of approaches to collect confirmed influenza infection cases. Two methods to estimate the VE of the pandemic vaccine were compared in a German population, one a test-negative casecontrol method using virologic surveillance data and the other an innovative case-series methodology using nationally reported laboratory-confirmed influenza case data (128). In children less than 14 years of age the estimate of VE using both methods were similar, with the first estimating a VE of 79% (95% CI, 35–93, *p* = 0.007) and the second 87% (95% CI, 78–92, *p* < 0.001).

### Canada

In Canada, children aged 3–10 years were recommended to receive a single 0.25 ml dose of an AS03-adjuvanted A(H1N1) vaccine (Arepanrix®), and children 6 months to less than 3 years of age to receive two doses. In the Canadian province Manitoba, the VE against laboratory-confirmed influenza cases was estimated at 97% (95% CI, 72–100) in very young children (aged 6–35 months) compared with no protection provided by the seasonal TIV (127). Although statistically significant results for this same age group were not found in another communitybased study due to small sample size, an even higher VE against laboratory-confirmed influenza of 100% (95% CI, 79.5–100) was reported in children aged 3 years to less than 10 years (129).

Varying results have been reported regarding VE against pneumonia and hospitalization in Canada. In Quebec, the effectiveness of a single pediatric vaccine dose in preventing hospitalization due to influenza in children aged 6 months to 9 years was 85% (95% CI, 61–94) (131), whereas the VE against hospitalization was considerably less, 58% (30–75%), for children less than 5 years of age in another study (137).

# SAFETY

### MF59-Adjuvanted Influenza Vaccines

There are extensive data regarding the safety of MF59-adjuvanted influenza vaccines which have demonstrated an acceptable safety profile in young children (26, 27, 30, 32, 33, 119, 138–147). Most adverse reactions are mild-to-moderate and transient in nature and serious adverse reactions are rare. A systematic review and meta-analysis has provided an overview of safety for both seasonal and pandemic MF59-adjuvanted influenza vaccines in children (144). The analysis reported no increase in serious adverse events (SAEs) compared with control vaccines. The rate of SAEs in the adjuvanted group was 0.0–10.4% with a pooled relative risk of 0.74 (95% CI, 0.57–0.97) (144). The relative risk for the most common solicited adverse events including redness and pain at the injection site and systemic reactions such as fever, irritability and loss appetite were significantly higher for MF59-adjuvanted vaccines compared with control vaccines. The rates of solicited adverse reactions included 1.0–59.0% for pain (<1% for grade 3 pain) and 4.0–19.0% for fever (144). There were similar rates of unsolicited adverse event reporting between children who received adjuvanted compared with non-adjuvanted vaccines.

An integrated analysis evaluated the safety of MF59 adjuvanted vaccines (predominantly the seasonal trivalent or tetravalent vaccine) in children 6 months to 18 years of age (139). The analysis included five clinical trials—four trials with a seasonal MF59-adjuvanted vaccine and one trial with the prepandemic H5N1 vaccine. A total of 1,181 children received an MF59-adjuvanted vaccine compared with 545 children who received a non-adjuvanted vaccine. There was an increased incidence in solicited local and systemic reactions compared with non-adjuvanted vaccines; however, these were mostly mild and transient, resolving by day 3 postvaccination. Across all ages 55% experienced local reactions and 48% systemic reactions after the first dose in the MF59-adjuvanted groups, and 43% and 34%, respectively, in the non-adjuvanted group. These were slightly lower in both groups following the second vaccination. There was no difference in the rate of SAEs.

Following this analysis, safety data have been published in large immunogenicity and efficacy studies. The study by Vesikari et al. reported minimal difference in local reactions between adjuvanted and non-adjuvanted influenza vaccines aged 6–36 months, and fever was reported in 15.3 versus 13.3%, respectively (119). Similar results were demonstrated in a large phase III, randomized, multicenter study which included 3,125 children who received ATIV (143).

Studies focusing on the pandemic H1N1 and prepandemic H5N1 vaccines have also shown MF59 to have an acceptable safety profile in children (31, 148–156). Transient mild pain or tenderness and erythema were the most commonly reported local reactions and fatigue and myalgia the most common systemic reactions. Few, if any, children reported severe reactions including fever >40°C.

Theoretical concerns have been raised that MF59 vaccination may induce antibodies to squalene. Squalene is a naturally occurring product in the body and antibodies to the squalene component of the vaccine would therefore pose a risk of autoimmune disease in a vaccine recipient. Subsequent studies have demonstrated that vaccination with MF59 adjuvant did not induce antisqualene antibodies nor enhance preexisting antisqualene antibody levels (157).

Despite the association between the AS03-adjuvanted pandemic vaccine Pandemrix® and narcolepsy (described below), there has been no evidence to date of any increased risk of narcolepsy associated with MF59-adjuvanted vaccines in children or adults (158); however, postlicensure surveillance in children will be important to continue monitoring for this as the frequency reported for Pandemrix® was too low to detect in clinical trials.

### AS03-Adjuvanted Influenza Vaccines

Prior to the licensing of the pandemic AS03-adjuvanted influenza vaccines, evidence on the safety of the AS03 adjuvant was available from clinical trials, which demonstrated an acceptable reactogenicity profile (159–161). Following the rapid licensure of Pandemrix® in Europe and Arepanrix® in Canada, passive and active surveillance programs were initiated. During the 2009–2010 influenza season 31 million doses of Pandemrix® were distributed throughout Europe, and 12 million doses of Arepanrix® mainly in Canada and Latin America (162) with the collection of safety data *via* these national surveillance programs. With limited safety data available prior to its distribution, it was not until postlicensure surveillance revealed concerns regarding Pandemrix® in children.

Clinical trials have demonstrated acceptable rates of solicited local and systemic adverse events, albeit higher than nonadjuvanted vaccines, following vaccination with the AS03-adjuvanted pandemic vaccines (132, 163–170). A recent systematic review included four clinical trials enrolling children who received an AS03-adjuvanted influenza vaccine. There were significantly increased rates of local adverse reactions including pain and swelling, although mostly mild and transient, after receiving the AS03-adjuvanted vaccine compared with non-adjuvanted control vaccines (144). Pooled data from these studies showed local pain as the most frequent adverse event following AS03 adjuvanted vaccines in children, experienced by 31.7–84.6% of children, with rates of grade 3 pain between 4.3 and 12.4%. The rate of fever following vaccination was 11.0–23.8% and there was no significantly increased risk of developing an unsolicited adverse event (RR 1.0, 95% CI, 0.97–1.04) or convulsion (RR 1.14, 95% CI, 0.42–3.14) compared with non-adjuvanted vaccines. Moreover, there was no increased risk of SAEs, with 0.0–8.0% of children experiencing an SAE. There is evidence in children that the second vaccine dose, given 21–28 days after the first, results in higher rates of local and/or systemic reactions compared with the first dose, although this is not consistent across all studies (132, 163, 166). Immunocompromised children have experienced similar rates of adverse events compared with immunocompetent children (171, 172).

The EMA announced on August 27, 2010, that a safety review had been initiated following concerns raised in Sweden with a case series of six adolescents diagnosed with narcolepsy within 2 months of vaccination with Pandemrix® (173, 174). An investigation by the European Centre for Disease Prevention and Control (ECDC) and Vaccine Adverse Event Surveillance and Communication Consortium (VAESCO) was then undertaken in late 2010 (175). Following the initial report, formal studies assessing an association between Pandemrix® and narcolepsy have been undertaken in Finland, Sweden, France, Ireland, United Kingdom, and Norway which have confirmed an increased incidence of narcolepsy in young vaccine recipients (176–182). In Finnish children aged 4–19 years there was a rate ratio of 12.7 (95% CI, 6.1–30.8), with a vaccine-attributable risk of 1:16,000 (95% CI, 1:13,000–1:21,000) of developing narcolepsy following receipt of Pandemrix® (176). The EMA Eudravigilance database had received almost 1,400 reports of narcolepsy in Pandemrix® recipients by 2015 (183).

Narcolepsy is a rare sleep condition with onset often in adolescence which is characterized by excessive daytime sleepiness, episodes of unintended sleep and cataplexy. It is thought to be due to immune-mediated destruction of neurons which results in deficiency in hypocretin production in the hypothalamus, although no specific antibodies involved in this process have been identified. The majority of individuals with narcolepsy and cataplexy express the HLADQB1\*0602 allele, and infections including influenza A and *Streptococcus pyogenes* have been implicated in triggering narcolepsy in susceptible individuals (184).

The biological plausibility linking Pandemrix® and narcolepsy has been explored although the exact mechanism is yet to be identified. The AS03 adjuvant itself and specific components of AS03 (e.g., α-tocopherol) not present in other adjuvants were suggested to be responsible; however, the lack of association between other AS03-containing vaccines (e.g., Arepanrix®) and narcolepsy may refute this theory (162, 176, 185). Molecular mimicry has been proposed as a possible mechanism behind the association, with one study reporting a similarity between a peptide on the influenza nucleoprotein A and an extracellular domain of the hypocretin receptor 2 (186). This study was subsequently retracted due to inability to replicate results but, despite this, did not adequately explain why narcolepsy would not be associated with other H1N1 vaccines. Further studies investigating the presence of neuronal antibodies have not identified narcolepsy-specific antibodies in the sera or CSF of vaccinated children with narcolepsy (187).

Arepanrix®, the AS03-adjuvanted pandemic vaccine used in Canada, has not been associated with such a significant risk of narcolepsy, with only one extra case of narcolepsy per million doses received (183, 188). This is despite both Pandemrix® and Arepanrix® vaccines containing similar amounts of HA and AS03. The different method of production of the H1N1 antigen between the two vaccines, resulting in antigenic differences, has been suggested to result in enhanced levels of IgG-antibodies that may be implicated in the association of Pandemrix® with narcolepsy (189). However, in a separate study sera and CSF samples from 13 vaccinated patients with narcolepsy were compared with 44 vaccinated patients without narcolepsy, revealing no increase in narcolepsy-specific autoantibodies (187).

# CONCLUSION

Conventional influenza vaccines have suboptimal immunogenicity in young children and adjuvanted influenza vaccines offer a superior alternative. MF59 and AS03 have proven to be immunogenic in young children, provide cross protection against mismatched influenza virus strains and allow for antigen sparing which is important in the setting of pandemics where the global demand is high. Despite these positive results, the association between AS03 and narcolepsy has resulted in the future use of the current AS03 formulation in children limited. MF59 ATIV is efficacious in children leading to its licensure in Canada in children; however, further studies investigating the effectiveness of MF59 seasonal vaccines would potentially improve the likelihood of licensure in young children in other countries and more widespread use of this vaccine in children.

Looking to the future, the recent advancements in understanding the mechanism of adjuvants through elucidating the innate and adaptive immune response and relating these with gene expression profiles will allow both the improvement of current adjuvants and development of novel adjuvants. The progress with some newer adjuvanted influenza vaccines is promising. Adjuvants based on toll-like receptor agonists including TLR4 and TLR9 agonists used in pandemic and "universal" influenza vaccines, respectively, have shown excellent results in phase I trials. A variety of bacteria-derived adjuvants (e.g., flagellin, *Escherichia coli* heat-labile toxin patch, meningococcal outer membrane protein) which take advantage of the ability of bacterial components to activate the innate immune system have been incorporated in seasonal trivalent influenza vaccines, with some moving to phase III clinical trials. New technologies have allowed the development of these adjuvants among a myriad of others (e.g., liposomes, virus–like particles, saponins, viral vectors, and newer oil-in-water emulsions); however, many remain in the experimental phase in animals

# REFERENCES


and there is a lack of robust human data for the majority of these currently. Finally, a "universal" influenza vaccine which provides protection against all influenza virus strains, regardless of antigen drift or shift, with long-lasting protection remains an ultimate goal in the development of improved influenza vaccines. The combination of novel antigen formulations and adjuvants underlies many candidate vaccines currently in development. A number of these vaccines have entered clinical trials in recent years with the most advanced (recombinant M2e fused with flagellin, VAX102) reaching phase II trials. These vaccines face the challenge of providing an equivalent, if not better, immunogenic response than current seasonal influenza vaccines and ensuring an acceptable safety profile. Given it has been 20 years since the licensure of an MF59-containing vaccine and licensure of an influenza vaccine adjuvanted with MF59 for children has only occurred recently, it is likely to be some time before adjuvanted vaccines are widely used in the pediatric population.

The aim continues to be provision of the best possible protection of children from influenza while minimizing reactogenicity. That no vaccine against influenza is yet licensed for the most vulnerable, less than 6-month-old term or preterm-born infants is a challenge that may not remain unaddressed.

### AUTHOR CONTRIBUTIONS

AW, DK, GN, and EC wrote the initial drafts for the manuscript which was then revised and contributed to by BP, C-AS, and AP.

### FUNDING

This study was supported in part by European Commission FP7 Grant "Advanced Immunization Technologies (ADITEC)" and the National Institute for Health Research Oxford Biomedical Research Centre.


publications/healthy-living/canadian-immunization-guide-statement-seasonal-influenza-vaccine-2017-2018.html


children using MF59 adjuvant. *Pediatr Infect Dis J* (2009) 28(7):563–71. doi:10.1097/INF.0b013e31819d6394


components of the protective response to influenza virus infection. *J Exp Med* (2000) 192(2):271–80. doi:10.1084/jem.192.2.271


following pandemic 2009-H1N1 vaccination in humans. *PLoS Pathog* (2012) 8(9):e1002920. doi:10.1371/journal.ppat.1002920


polyfunctional immune responses. *J Exp Med* (2008) 205(13):3119–31. doi:10.1084/jem.20082292


ema.europa.eu/docs/en\_GB/document\_library/Press\_release/2010/08/ WC500096005.pdf


**Conflict of Interest Statement:** C-AS has received numerous educational or research grants, including from vaccine manufacturers, although none related to this work. AP has previously conducted clinical studies on behalf of Oxford University that were sponsored by vaccine manufacturers with grants from Okairos and Pfizer closing since January 2015. His department received unrestricted educational grants from Novartis/GSK/Astra Zeneca in 2015, Pfizer/GSK/Astra Zeneca in July 2016, and Gilead/MSD/GSK/Astra Zeneca in June 2017 to support a 3-day course on Infection and Immunity in Children. He is Chair of UK Dept. Health's (DH) Joint Committee on Vaccination and Immunization (JCVI) and chair of the scientific advisory group on vaccines for the European Medicines Agency and is a member of the WHO's SAGE. The views expressed in this manuscript do not necessarily represent the views of DH, JCVI or WHO. All other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Wilkins, Kazmin, Napolitani, Clutterbuck, Pulendran, Siegrist and Pollard. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# An Unexpected Major Role for Proteasome-Catalyzed Peptide Splicing in Generation of T Cell epitopes: is There Relevance for vaccine Development?

*Anouk C. M. Platteel 1†, Juliane Liepe2 , Willem van Eden1 , Michele Mishto3,4,5 and Alice J. A. M. Sijts1 \**

### *Edited by:*

*Peter Andersen, Statens Serum Institut, Denmark*

### *Reviewed by:*

*Simone Joosten, Leiden University Medical Center, Netherlands Juraj Ivanyi, King's College London, United Kingdom Peter Timmerman, Pepscan (Netherlands), Netherlands*

> *\*Correspondence: Alice J. A. M. Sijts e.j.a.m.sijts@uu.nl*

### *†Present address:*

*Anouk C. M. Platteel, Department of Medical Microbiology and Immunology, St. Antonius Hospital, Nieuwegein, Netherlands*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 13 September 2017 Accepted: 17 October 2017 Published: 03 November 2017*

### *Citation:*

*Platteel ACM, Liepe J, van Eden W, Mishto M and Sijts AJAM (2017) An Unexpected Major Role for Proteasome-Catalyzed Peptide Splicing in Generation of T Cell Epitopes: Is There Relevance for Vaccine Development? Front. Immunol. 8:1441. doi: 10.3389/fimmu.2017.01441*

*<sup>1</sup> Faculty of Veterinary Medicine, Department of Infectious Diseases and Immunology, Utrecht University, Utrecht, Netherlands, 2Max-Planck-Institute for Biophysical Chemistry, Göttingen, Germany, 3 Institut für Biochemie, Charité – Universitätsmedizin Berlin, Berlin, Germany, 4Berlin Institute of Health, Berlin, Germany, 5 Centre for Inflammation Biology and Cancer Immunology (CIBCI), Peter Gorer Department of Immunobiology, King's College London, London, United Kingdom*

Efficient and safe induction of CD8+ T cell responses is a desired characteristic of vaccines against intracellular pathogens. To achieve this, a new generation of safe vaccines is being developed accommodating single, dominant antigens of pathogens of interest. In particular, the selection of such antigens is challenging, since due to HLA polymorphism the ligand specificities and immunodominance hierarchies of pathogen-specific CD8+ T cell responses differ throughout the human population. A recently discovered mechanism of proteasome-mediated CD8+ T cell epitope generation, i.e., by proteasome-catalyzed peptide splicing (PCPS), expands the pool of peptides and antigens, presented by MHC class I HLA molecules. On the cell surface, one-third of the presented self-peptides are generated by PCPS, which coincides with one-fourth in terms of abundance. Spliced epitopes are targeted by CD8+ T cell responses during infection and, like non-spliced epitopes, can be identified within antigen sequences using a novel *in silico* strategy. The existence of spliced epitopes, by enlarging the pool of peptides available for presentation by different HLA variants, opens new opportunities for immunotherapies and vaccine design.

Keywords: antigen processing, CD8+ T cell, epitope, intracellular pathogens, proteasome, peptide splicing, vaccine

# INTRODUCTION

Infection with most viruses and bacteria elicits a protective immune response that eradicates the pathogen and provides immunity to subsequent infections. Vaccination approaches aim to trigger similarly protective immune responses, but usually fail to activate the broad range of immune effector mechanisms reacting to natural infection. For example, most vaccines are designed and/or tested for

**Abbreviations:** cTECs, cortical thymic epithelial cells; EBV, Epstein–Barr virus; ERAP, endoplasmic reticulum aminopeptidase; MHC-I, MHC class I; PCPS, proteasome-catalyzed peptide splicing; PlcB, phosphatidylcholine-preferring phospholipase-C (PC-PLC); TAP, transporter associated with antigen processing.

capacity to elicit humoral responses, but intracellular pathogens, once they have entered their target cells in infected hosts, are eliminated by T cells only. Currently, the design of safe vaccines that induce protective T cell immunity, capable of monitoring new infections, remains challenging. One of the reasons is that T cells of different individuals target different pathogen-derived epitopes. Although advanced prediction programs exist for the identification of such epitopes, a larger percentage of proteins seem not to contain any epitope candidates that may trigger a protective T cell response. In this mini review, we will focus on a recently discovered mechanism underlying T cell-mediated immune recognition. Proteasomes, the proteases that generate most peptide epitopes presented to CD8<sup>+</sup> T cells (1), not only degrade proteins but also paste non-contiguous sequences of a given antigen back together (2, 3). This mechanism is called proteasome-catalyzed peptide splicing (PCPS) (**Figure 1**). PCPS is a frequently occurring process and, importantly, expands the potential epitope pool significantly [for review, see Ref. (4)]. We here discuss the opportunities and implications of PCPS for pathogen-specific immune protection and vaccine design.

### SIGNALING INVADERS

CD8<sup>+</sup> T cells play an important role in immune protection to intracellular pathogens, including viruses and intracellular bacteria, and tumor growth. To signal infection or other intracellular aberrations, cells exploit the ubiquitin proteasome system, which safeguards the cellular proteome by degrading the majority of unfolded, immature, obsolete, and short-lived mature proteins located in the cytoplasm (1). Peptide fragments released from the proteasome may bind the transporter associated with antigen processing (TAP) for translocation into the ER. Here, they may undergo N-terminal trimming by endoplasmic reticulum aminopeptidases and be loaded into the antigen-presenting groove of MHC class I (MHC-I) molecules, if they contain an appropriate MHC-I-binding motif. Peptide loading stabilizes MHC-I molecules, which then traffic to the cell surface for display of the peptides to CD8<sup>+</sup> T cells.

The proteasome, by processing most MHC-I-presented antigens, shapes the antigenic peptide repertoire available for binding to MHC-I complexes. This is illustrated by the fact that—although the antigenic peptides monitored by CD8<sup>+</sup> T cells at the cell surface are influenced by the specificity of each step of the antigen presentation pathway—the two major factors selecting the MHC-I immunopeptidome are the affinity of the peptides for the cleft of the different MHC-I variants and proteasome cleavage specificity (10–13).

The proteasome can hydrolyze almost any peptide bond, but with a large range of efficiencies, resulting in huge differences in quantity between specific peptide products. One of the main cellular mechanisms to alter the peptide repertoire, produced by proteasomes, is by changing the cell's proteasome isoform content (14). The proteasome is a multi-catalytic enzyme complex, composed of a 20S core particle, responsible for proteolysis, and different regulatory complexes, including the 19S regulatory complex, which is responsible for substrate capture and unfolding in an ATP-dependent manner [for review, see Ref. (1, 14)]. The 20S catalytic core consists of four stacked rings of seven subunits each, with catalytic activity exerted by three β subunits—i.e., β1, β2, and β5—present in the inner two rings of this particle. Under stress conditions and cytokine exposure, these subunits can be replaced by their inducible homologs LMP2/iβ1, MECL-1/iβ2, and LMP7/iβ5, leading to the formation of immunoproteasomes. Depending on cell type and levels of constitutive or induced LMP2/iβ1, MECL-1/iβ2, and LMP7/iβ5 subunit expression, cells often possess "mixed proteasomes," which contain both standard and inducible catalytic β subunits (14). In addition, cortical thymic epithelial cells incorporate the thymus-specific β5t subunit in their proteasomes, probably to support the unique role of these cells in positive selection of CD8<sup>+</sup> T cells (14–16). The exchange of

between two splice reactants originating from two distinct proteins (*trans*-PCPS) (6, 7). Although the latter occurs *in vitro* (6–8), its occurrence in cells is disputable (7). We here show the example of the spliced peptide gp100mel35-39/35-39, which has been identified in *in vitro* digestions of synthetic substrate by purified proteasomes (9).

the catalytic β subunits largely affects the proteolytic dynamics of the proteasome (17) and thus, depending on cytokine milieu and expressed proteasome isoforms, different peptide products will predominate among the repertoire of peptides produced. These quantitative differences in the generation of specific peptides, by the different proteasome isoforms, can mark the immunogenicity of the individual peptides, i.e., quantitative differences in epitope generation can determine whether a specific T cell response is primed, and greatly affect the immunodominance hierarchy of CD8<sup>+</sup> T cells responding to infection, as demonstrated in mouse models (18–22).

## PCPS AND ITS POTENTIAL RELEVANCE IN CD8**+** T CELL RESPONSE DURING INFECTION

The active site of the catalytic proteasome β subunits is formed by a threonine (T) residue at position 1 of the mature form of these enzymes. This T1 catalyses the break of the peptide bond between two residues of a substrate, thereby leading to the formation of an acyl-enzyme intermediate between the active site T1 and the N-terminal portion of the substrate. The C-terminal peptide fragment is then released. For a long period, it was assumed that the N-terminal portion of the substrate would always be released by hydrolysis. However, more than a decade ago, it was discovered that not only an H2O molecule but also the NH2-terminus of an earlier produced peptide—i.e., the C-terminal splice reactant (**Figure 1**)—can compete with a H2O molecule to attack the acylenzyme intermediate. This then results in a peptide bond between the N-terminal portion of the substrate—i.e., the N-terminal splice reactant—and the C-terminal splice reactant. A new, spliced peptide is thereby formed and released by the proteasome (2, 6) (**Figure 1**). This transpeptidation reaction is likely to be the most frequent reaction type of PCPS and was demonstrated to contribute to the generation of a handful of tumor epitopes (4). Because of the small number of spliced epitopes described in over a decade, it was believed that PCPS was a rare event of no immunological relevance. However, recent findings demonstrating that around 30% of the MHC-I immunopeptidome variety of Epstein–Barr virus (EBV)-immortalized B cell lines and primary human fibroblasts is represented by spliced peptides (23) suggested the contrary. Although it still has to be proven that this phenomenon occurs in such great extent also in other cell types, spliced peptides are likely to fulfill an immune-relevant function during the immune response against infections.

First evidence supporting this hypothesis was obtained in a murine model of infection by the intracellular bacterium *Listeria monocytogenes*. We showed that CD8<sup>+</sup> T cells, specific for an immunodominant epitope derived from the listeriolysin O antigen, cross-recognized an overlapping peptide produced by proteasome-mediated splicing of the LLO sequence (**Figure 2A**) (24). These data illustrated that CD8+ T cells responding to infection could recognize both non-spliced and spliced variants of the same epitope, but did not strictly prove that PCPS increased the immune-relevant epitope pool. The latter was addressed in a second study, identifying two spliced epitopes derived from the *Listeria* phosphatidylcholine-preferring phospholipase-C (Plc)B protein (25). These spliced epitopes, i.e., PlcB189-191/163-167 and PlcB189-192/164-167 (**Figure 2B**), triggered a specific CD8<sup>+</sup> T cell response in *L. monocytogenes*-infected mice that could not be explained by any cross-reactivity toward non-spliced peptides of the same antigen or other antigens of the *L. monocytogenes*. Furthermore, mutation of the MHC-I anchor sites of these two spliced epitopes, within the PlcB antigen, abolished recognition of PlcB-transfectant cells by CD8<sup>+</sup> T cells of infected mice, suggesting that the two spliced epitopes dominate the response against the PlcB antigen during *L. monocytogenes* infection (25).

These sparse data prompt the question whether spliced epitopes are just an extra pool of antigenic peptides, largely overlapping with the non-spliced peptide pool, or whether they exert a broader immune-relevant function. The fact that two spliced epitopes drive the response of the CD8<sup>+</sup> T cells toward the PlcB antigen during *L. monocytogenes* infection (25) would suggest a broader function. For instance, it should be noted that various spliced epitope candidates that may be produced from *Listeria* PlcB largely exceeds that of non-spliced peptides, as calculated for both the mouse MHC-I H-2Kb molecule and the most frequent human MHC-I HLA-A and -B variants. In these *in silico* analyses, no non-spliced epitope candidates predicted to efficiently bind the predominant MHC-I variants were identified within the PlcB sequence, in opposite to hundreds of potential spliced epitope candidates (25). The potential immune relevance of spliced epitopes in enlarging the antigenic landscape of cells is further supported by the finding that one-third of human self-antigens is represented only as spliced peptides within the MHC-I immunopeptidomes of EBV-immortalized B cell lines and primary human fibroblasts (23). Moreover, with one-third of spliced peptides occupying one-fourth of the available MHC-I molecules, compared to two-thirds of non-spliced peptides occupying three-fourth of MHC-I molecules of the cell types investigated so far (23), the differences in quantity between spliced and non-spliced antigenic peptides may be rather small. In line with this assumption, Ebstein et al. (5) quantified the presence, at the cell surface, of gp100mel-derived spliced and non-spliced epitopes endogenously generated by proteasomes, using specific CD8<sup>+</sup> T cell clones. Despite substantial differences between epitopes, spliced and non-spliced epitopes were presented in comparable amounts.

Thus, spliced epitopes should be equally likely to trigger CD8<sup>+</sup> T cell responses as non-spliced epitopes. In addition, spliced epitopes that could be considered as variants of non-spliced epitopes, like LLO294/297-304, may either serve as alternative ligand for CD8<sup>+</sup> T cells primed by the non-spliced epitope or, perhaps, serve as antagonists, dampening the antigen-specific CD8<sup>+</sup> T cell response.

## TOOLS FOR THE PREDICTION OF SPLICED EPITOPES

Further studies into the immunological relevance of spliced peptides would benefit from *in silico* tools for predicting spliced epitope candidates within protein sequences of interest. However,

the combinatorial possibilities of spliced peptide generation are so many that without a preliminary *in silico* strategy, any analysis would be challenging. We recently developed and tested an *in silico* approach for spliced epitope identification, using the murine model of *Listeria* infection (25). As target antigens, we selected two small, secreted virulence factors, PlcA and PlcB, and calculated all possible spliced peptides originating from these proteins. Some restrictions were introduced, based on previous studies, i.e., we selected only spliced peptides that are 8 or 9 residues long, representing the typical length of peptides presented by mouse MHC class I molecules. Furthermore, we allowed: (i) 40 residues as maximal intervening sequence length; (ii) only *cis-*PCPS, which seems to be more common than *trans-*PCPS in living cells; and (iii) that the splice reactants can be spliced in either the same or reverse order as occurring in the antigen of interest (**Figure 1**). The resulting reduced spliced peptide database was then analyzed using the online available SMM prediction tool for MHC-I-peptide binding affinity, to select peptides with high binding affinity for mouse H-2Kb molecules. Setting a cutoff for binding affinity at 16 nM, 22 spliced epitope candidates were selected and tested *ex vivo* for recognition by splenic CD8<sup>+</sup> T cells of *L. monocytogenes*-infected mice. These experiments identified two spliced peptides, PlcB189-191/163-167 and PlcB189-192/164-167, as targets of the *Listeria*-specific CD8<sup>+</sup> T cell response. Selecting the final spliced epitope candidates *in silico*, based on their low predicted IC50, showed to be the right approach since PlcB189-191/163-167 and PlcB189-192/164-167 were the two peptides, ranked as the highest affinity binders (25). Thus, we observed a correlation between predicted IC50 (and H-2Kb —peptide complex stability) and priming of specific CD8<sup>+</sup> T cells during infection, consistent with earlier published works on non-spliced epitopes (11–13).

Despite the positive results of the above study, the applied MHC-I-peptide-binding affinity predictions could represent a problem, because the most efficient MHC-I-peptide-binding affinity prediction algorithms have been trained on non-spliced epitope databases. PCPS and cleavage preferences significantly differ in term of substrate sequence motifs (6). Thus, it is possible that the canonical sequence motif of the non-spliced antigenic peptides bound to a given MHC-I variant differ from that of the spliced antigenic peptides. This seemed to be the case for the MHC-I immunopeptidomes of the EBV-immortalized B cell lines and primary human fibroblasts (23). Probably, a great improvement will come from studying the sequence motifs preferred by PCPS. A first attempt in that direction has been done by Berkers et al. (8). Using small peptide libraries, composed of C- and N-terminal splice reactants carrying one of two required HLA-A\*02:01 anchor residues, and HLA-A\*02:01 molecules coupled with a chemosensitive ligand as read out, the authors defined which residues at P2′ and P1′, and P1 and P2 position most efficiently promote PCPS. Thereby, they confirmed that rather than being a random process, PCPS is finely regulated by the substrate sequence. This specificity could be explained by the binding of the splice reactants to PCPS-binding sites of the proteasome nearby its catalytic T1 (6). However, Berkers' approach was restricted to an epitope backbone sequence that was used as a reference and variations of that sequence were tested in terms of PCPS efficiency. It is known that not only residues in proximity but also 10 residues away may affect the cleavage-site usage by proteasomes (21, 26). Hence, the use of a fixed backbone sequence could bias the outcome of the analysis and partially restrict the results to that model sequence. Furthermore, the outcome of that pioneering work is specific for the binding motifs of HLA-A\*02:01 and could thus significantly differ for other HLA-I haplotypes.

We can reckon that the development of a PCPS prediction algorithm would benefit from a sequence unbiased approach, like provided by mass-spectrometry-based identification of spliced peptides in *in vitro* digestions of synthetic polypeptides (or proteins) carried out with purified proteasomes. This approach was the basis for the definition of preferred sequence motifs of proteasome-mediated peptide hydrolysis (20, 27, 28). Such a quantitative approach, in future, may define the proteasome preferences in terms of cleavage and PCPS, and thereby facilitate further research on the immunological relevance of PCPS.

# PCPS AND CROSS-REACTIVITY

Although spliced epitope-specific T cells most likely are selected in the thymus in a similar fashion as non-spliced epitope-specific T cells, PCPS offers so far unrecognized possibilities for crossreactivity of pathogen-specific CD8<sup>+</sup> T cell responses with self-antigens. Due to the limited research into PCPS-generated pathogen-derived epitopes, such auto-reactivity has not yet been shown for spliced epitope-specific CD8<sup>+</sup> T cells. However, in human T1D and mouse models of insulin-dependent diabetes mellitus, the autoreactive CD4<sup>+</sup> T cell response has been shown to target *trans*-spliced self-epitopes [(29–31), and reviewed in Ref. (4)]. In addition, most spliced CD8<sup>+</sup> T cell epitopes described up to date are derived from tumor-associated (self-)antigens, as discussed earlier. Thus, like for non-spliced epitopes, a spliced epitope-specific autoreactive T cell repertoire certainly exists. Given the tremendous number of possibilities for epitope splicing, there is a conceivable chance for a minor fraction of pathogen-induced spliced epitope-specific CD8+ T cells to display auto-reactivity.

## RELEVANCE OF PCPS FOR VACCINE DESIGN

Proteasome-catalyzed peptide splicing is not a random process of ligating peptide sequences. On the contrary, preliminary evidence suggests that PCPS prefers specific peptide sequence motifs and that those motifs are not those preferred for the canonical peptide hydrolysis reaction (6, 8). This implies that the spliced and non-spliced epitope pools can differ in term of characteristics and predisposition to bind specific HLA-I haplotypes. This can explain why a large portion of self-antigens seems to be represented at the cell surface only by spliced peptides (23). The few examples of MHC-I-spliced immunopeptidomes available so far do not allow us even to speculate if those antigens represented only by spliced peptides have either a peculiar intracellular localization, specific chemical characteristics, or turnover. Moreover, we do not know whether this large representation of spliced peptide in the MHC-I self immunopeptidome occurs also in the MHC-I viral immunopeptidome. However, our preliminary evidence that the mouse CD8<sup>+</sup> T cell response against a specific *L. monocytogenes* antigen (PlcB) is driven only through spliced epitopes suggests that PCPS may open novel opportunities for vaccine design.

To improve the efficacy of vaccination, especially against "difficult to treat" and rapidly mutating intracellular pathogens, vaccines need to be redesigned to elicit pathogen-specific, protective CD8<sup>+</sup> T cell responses. Currently, a new generation of live, safe, recombinant vaccine vectors that encode one or multiple small antigens, derived from the pathogen of interest, is evaluated for protective capacity (32–35). Nevertheless, the extensive polymorphism of the human HLA, resulting in differing epitope-specificities and immunodominance hierarchies of responding CD8+ T cells in different individuals, makes it difficult to predict which antigens will provide optimal protection within a larger percentage of the population. *In silico* analysis tools are frequently used to screen antigens, but may fail to predict relevant T cell epitopes that may induce a response in context of the majority of HLA-I haplotypes, as we illustrated for *L. monocytogenes-*derived PlcB (25). Including spliced peptides as potential epitopes could aid in this, by enlarging the number of potentially immune-relevant epitopes as well as the variety of pathogen' antigens that can be used to trigger a robust and persistent immune response. Therefore, we can speculate that further research into the immunological relevance of spliced epitopes during the immune response against infections, as well as optimization of *in silico* tools to accurately predict such epitopes within antigen sequences, can bring remarkable benefits for vaccine development.

# AUTHOR CONTRIBUTIONS

AP drafted the manuscript, prepared the figures, and gave final approval of the version to be published. JL, MM, and AS drafted and edited the manuscript, prepared the figures, and gave final approval of the version to be published. WE edited the manuscript and gave final approval of the version to be published.

# FUNDING

This work was financially supported by the European Union's Seventh Framework Program—Grant No. 280873 ADITEC to AS, and NC3Rs through a David Sainsbury Fellowship to JL (NC/K001949/1). MM was supported by Berlin Institute of Health (BIH, CRG1-TP1).

An antigenic peptide produced by peptide splicing in the proteasome. *Science* (2004) 304:587–90. doi:10.1126/science.1095522 3. Hanada K, Yewdell JW, Yang JC. Immune recognition of a human renal cancer antigen through post-translational protein splicing. *Nature* (2004) 427:252–6.

1. Sijts EJ, Kloetzel PM. The role of the proteasome in the generation of MHC class I ligands and immune responses. *Cell Mol Life Sci* (2011) 68:1491–502.

2. Vigneron N, Stroobant V, Chapiro J, Ooms A, Degiovanni G, Morel S, et al.


and epitope generation. *Eur J Immunol* (2014) 44:3508–21. doi:10.1002/ eji.201444902


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Platteel, Liepe, van Eden, Mishto and Sijts. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

REFERENCES

doi:10.1007/s00018-011-0657-y

# Molecular Signatures of Immunity and Immunogenicity in Infection and Vaccination

*Mariëlle C. Haks1 , Barbara Bottazzi2 , Valentina Cecchinato3 , Corinne De Gregorio3 , Giuseppe Del Giudice4 , Stefan H. E. Kaufmann5 , Antonio Lanzavecchia3 , David J. M. Lewis6 , Jeroen Maertzdorf <sup>5</sup> , Alberto Mantovani 2,7, Federica Sallusto3,8, Marina Sironi2 , Mariagrazia Uguccioni 3,7 and Tom H. M. Ottenhoff <sup>1</sup> \**

*1Department of Infectious Diseases, Leiden University Medical Center, Leiden, Netherlands, 2Humanitas Clinical and Research Center, Rozzano, Italy, 3 Institute for Research in Biomedicine, Università della Svizzera italiana, Bellinzona, Switzerland, 4GSK Vaccines, Siena, Italy, 5Department of Immunology, Max Planck Institute for Infection Biology, Berlin, Germany, 6University of Surrey, Guildford, United Kingdom, 7Humanitas University, Department of Biomedical Sciences, Pieve Emanuele-Milan, Italy, 8 Institute of Microbiology, ETH Zurich, Zurich, Switzerland*

### *Edited by:*

*Rashika El Ridi, Cairo University, Egypt*

### *Reviewed by:*

*Francesco Dieli, Università degli Studi di Palermo, Italy Katie Louise Flanagan, Monash University, Australia*

> *\*Correspondence: Tom H. M. Ottenhoff t.h.m.ottenhoff@lumc.nl*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 13 September 2017 Accepted: 31 October 2017 Published: 15 November 2017*

### *Citation:*

*Haks MC, Bottazzi B, Cecchinato V, De Gregorio C, Del Giudice G, Kaufmann SHE, Lanzavecchia A, Lewis DJM, Maertzdorf J, Mantovani A, Sallusto F, Sironi M, Uguccioni M and Ottenhoff THM (2017) Molecular Signatures of Immunity and Immunogenicity in Infection and Vaccination. Front. Immunol. 8:1563. doi: 10.3389/fimmu.2017.01563*

Vaccinology aims to understand what factors drive vaccine-induced immunity and protection. For many vaccines, however, the mechanisms underlying immunity and protection remain incompletely characterized at best, and except for neutralizing antibodies induced by viral vaccines, few correlates of protection exist. Recent omics and systems biology big data platforms have yielded valuable insights in these areas, particularly for viral vaccines, but in the case of more complex vaccines against bacterial infectious diseases, understanding is fragmented and limited. To fill this gap, the EC supported ADITEC project (http://www.aditecproject.eu/; http://stm.sciencemag.org/ content/4/128/128cm4.full) featured a work package on "Molecular signatures of immunity and immunogenicity," aimed to identify key molecular mechanisms of innate and adaptive immunity during effector and memory stages of immune responses following vaccination. Specifically, technologies were developed to assess the human immune response to vaccination and infection at the level of the transcriptomic and proteomic response, T-cell and B-cell memory formation, cellular trafficking, and key molecular pathways of innate immunity, with emphasis on underlying mechanisms of protective immunity. This work intersected with other efforts in the ADITEC project. This review summarizes the main achievements of the work package.

Keywords: vaccines, biomarkers, immunity and infections, immunity, assays

# INTRODUCTION

One of the key goals in vaccinology is to understand what factors drive protective immunity induced by vaccines. In many cases, the precise mechanisms underlying such immunity remain unknown, or at best incompletely characterized. Except for the case of neutralizing antibodies induced by viral vaccines, very few correlates of protection exist. The availability of omics technologies and advanced systems biology big data analytical platforms has enabled large scale analyses of vaccine-induced responses. These efforts revealed novel insights into immunity and correlates of protection, for example, in the human response to yellow fever and influenza vaccination (1, 2). However, for more complex vaccines such as those against bacterial infectious diseases, such understanding as yet is mostly lacking.

To fill this gap, EC FP7 supported ADITEC innovative project featured a work package on "Molecular signatures of immunity and immunogenicity." The working group aimed to identify key molecular mechanisms of innate and adaptive immunity during effector and memory stages of immune responses following vaccination. Specifically, technologies were developed to assess the human immune response to vaccination and infection at the level of the transcriptomic and proteomic response, T-cell and B-cell memory formation, cellular trafficking, and key molecular pathways of innate immunity, with emphasis on underlying mechanisms of protective immunity. This work was integrated with efforts in other sections of the ADITEC project, such as a work package on immunity and vaccine-induced responses in early life and aging; and work packages on the human response to adjuvants in clinical and translational vaccinology.

The following specific issues will be addressed in detail below:


targeting a conserved site in the hemagglutinin (HA) stem was found.


Each of these topics will be discussed in more detail, followed by a concluding section.

### A dcRT-MLPA ASSAY TO IDENTIFY NOVEL DIAGNOSTIC, PROGNOSTIC, AND PREDICTIVE BIOMARKERS IN RESPONSE TO VACCINATION, INFECTION, ACTIVE DISEASE, AND CURE

Dissecting innate, adaptive, and inflammatory immune signatures underlie the identification of biomarker signatures that can be used to diagnose infection or disease; to predict progression toward disease in the infected host; to monitor the efficacy of disease treatment; or to predict (in)adequate responsiveness to novel or existing vaccines. The latter signatures will greatly aid the rational development, testing and evaluation of novel vaccines. They may also allow the study of the impact of host factors such as age (elderly vs. children), coinfection [HIV, CMV, Epstein–Barr virus (EBV), etc.], comorbidity (type 2 diabetes, allergic, and inflammatory disorders) or iatrogenic immunosuppression (biologicals, other immunosuppressants) on responses to vaccines.

To investigate such human immune responses at the transcriptomic level, we have developed a dcRT-MLPA assay (3). dcRT-MLPA is a focused gene expression profiling platform that has proven suitable to identify, monitor, and validate multicomponent host diagnostic and prognostic/predictive biomarker signatures in many human cohort studies (4). The studies described in this article focus on the discovery of diagnostic, prognostic, and therapy-responsive biomarker signatures mainly in tuberculosis (TB). Genes were selected for inclusion in the dcRT-MLPA targeted gene expression profiling platform based on their ability to assess various compartments of the human immune response, with key roles in inducing and skewing immune reactivity and inflammation. This included genes representing adaptive immune responses (T/NK/B-cell, Treg, and Th1/Th2/Th9/Th17 and Th22 responses, cytotoxicity, and cell subset markers); innate immune responses (myeloid associated markers and scavenger receptors, pattern recognition receptors, and inflammasome components); inflammatory and IFN-signaling genes; and genes associated with apoptosis, cell growth/proliferation, transcriptional regulators/activators, mitochondrial stress, and inflammation.

### Diagnostic Biomarkers

Tuberculosis is a global health problem and the world's leading cause of death from infectious diseases [World Health (5)]. One-fourth of the worldwide population is latently infected with *Mycobacterium tuberculosis* (*Mtb*), the pathogen causing TB, while active disease develops in about 5–10% of infected individuals. Understanding the differences in immune responses between individuals able to control the infection versus those who develop active disease, preferably in a preclinical stage, would greatly facilitate vaccine development. However, since no correlates of protection have been uncovered yet and vaccination with BCG mainly contributes to the prevention and control of childhood but not adult TB, effective control of the TB pandemic still mainly depends on accurate diagnosis and early detection of active disease, allowing curative treatment to be initiated before TB transmission can occur. Because current immunological tools to diagnose TB infection, such as the tuberculin skin test and interferon-gamma release assays are not able to distinguish between latent TB infection (LTBI) and active TB disease, and have poor sensitivity in children and immune-compromised individuals, direct *ex vivo* whole blood RNA expression profiling was performed in genetically and geographically diverse adult, pediatric and HIV-infected cohorts to identify biomarkers that can classify clinical stages of TB independent of age, HIV status and genetic background.

Characterization of the human immune response to *Mtb* in cohorts from South Africa, Malawi, The Gambia, Ethiopia, India, and Paraguay identified biomarker signatures that were strongly associated with active TB and were strikingly distinct from that associated with latent infection or uninfected controls (3, 6–10). Although the compiled multicomponent signatures differed to a certain extent between study cohorts from different geographic origins, the most robust single classifier discriminating between active disease and latent infection or uninfected controls was *FCGR1A* [high-affinity IgG Fc receptor 1A (CD64)]. This marker is significantly higher expressed in individuals with active TB than in those with LTBI regardless of age (8, 10), HIV status (7, 9), or geogenetic differences (3, 6, 7). This demonstrates the power of this biomarker to classify the different clinical outcomes of exposure to *Mtb* and warrants further analysis of the role of FCGR1A in TB pathogenesis.

# Biomarkers of the Curative Response to Therapy

Multi-, extensive-, and total-drug resistant TB (MDR/XDR/ TDR-TB) continues to emerge and is primarily caused by the use of ineffective formulations of drugs (such as use of single drugs) and poor adherence by TB patients to the strict 6-month drug treatment regimen. Therefore, monitoring early kinetics in treatment responses is key to the effective delivery of anti-TB treatment (ATT) and to prevent *de novo* drug resistance.

Longitudinal follow up of TB cases during ATT showed that biomarkers that can discriminate active TB from LTBI and uninfected controls can also be used to monitor TB treatment responses independent of age, HIV status, or genetic background (3, 8; Gebremicael et al., forthcoming1 ). In immune-competent adults from The Gambia, gene expression profiles normalized over time and were similar to those observed in LTBIs and uninfected controls at the end of treatment (6 months) (3). However, changes in gene expression levels during ATT showed distinct kinetics. The expression levels of a proportion of genes normalized to the expression levels observed in LTBIs within 2 months (e.g., *BLR1*), whereas others reached control levels only after the full 6 months of treatment (e.g., *FCGR1A*). Importantly, in HIVcoinfected TB patients, transcriptomic profiles of ATT treatment responses (including expression of *FCGR1A*) were identified that were not affected by highly active antiretroviral therapy, and that normalized to levels observed in HIV+ latently TB infected and uninfected controls after completion of ATT (see text footnote 1). Moreover, in an Indian pediatric cohort, baseline levels of *BLR1* and *FCGR1A* displayed a capacity of >70% to predict the six month treatment outcome (8). Together, these data suggest that *FCGR1A* could possibly serve as part of a reliable and robust predictive signature of the treatment response, and as such help optimizing personalized medicine in TB while minimizing *de novo* drug resistance, independent of age, immune status and genetic background.

<sup>1</sup>Gebremicael GKD, Quinten E, Alemayehu Y, Gebreegziaxier A, Belay Y, et al. Host gene expression kinetics during treatment of tuberculosis in HIV-coinfected individuals is independent of HAART therapy (Forthcoming).

# Prognostic Biomarkers

Because most immunocompetent individuals maintain a lifelong latent infection with *Mtb* and active disease develops in only a minority of infected individuals, identification of correlates of TB progression would substantially contribute to combat the TB epidemic. In TB endemic areas, such biomarker signatures could identify individuals at risk of developing TB, allowing curative treatment to be initiated before TB transmission can occur.

Transcriptomic profiling of the immune responses of immune-competent adult progressors versus non-TB progressors in a well-characterized TB case-contact platform from The Gambia identified—among others—antiapoptotic gene *BCL2* as a marker that might predict the onset of active disease very early after infection (11). *BCL2* was significantly lower expressed in progressors compared with non-progressors, and dysregulation of *BCL2* occurred as early as 1 year before progression toward TB disease. Corroborating these findings at a cellular level, increased levels of apoptosis in effector T-cells were found to constitute a risk factor for TB disease progression (12).

In addition, in a retrospective case–control study in The Netherlands of a high risk population of HIV-infected drug users, *IL13* and *AIRE* were identified as markers correlating with progression to TB, months before clinical diagnosis (13). Furthermore, transcriptomic profiles of *IL13*-expressing TB cases were strongly enriched for type I IFN related signaling genes, suggesting that these genes represent processes that contribute to TB pathogenesis in HIV-infected individuals. Although the association between *IL-13-AIRE* and TB progression could not be validated in independent cohorts of HIVnegative South African and Gambian adult progressors and controls, activation of the interferon response, however, was validated as a signature predictive of progression toward active disease (14). Using RNA-Seq, a whole blood transcriptomic mRNA expression signature was identified in a large prospective cohort of LTBI adolescents based on the differential expression of 16 human genes (14). Interestingly, also *FCGR1A* (as well as *FCGR1B*) was again part of the predictive signature, in agreement with the importance of these and other IFN-regulated genes in TB pathogenesis.

Once these biomarker signatures have been validated in larger cohorts, they will provide an initial platform to improve diagnosis of TB, monitor TB therapeutic interventions and prospectively identify people at risk of developing TB. More importantly, these biomarker signatures may help unravel and improve our understanding of protective immunity to TB, which is a prerequisite for the development of an effective TB vaccine.

### TRANSCRIPTOME PROFILES IN INFANTS VACCINATED WITH RECOMBINANT BCG VPM1002

The current TB vaccine *Mycobacterium bovis* BCG prevents severe disseminated TB in children, but fails to protect against pulmonary TB. The recombinant BCG ΔureC:hly (VPM1002) was designed to improve safety and immunogenicity (15). This recombinant vaccine was modified to express the membraneperturbing listeriolysin (Hly) of *Listeria monocytogenes*, allowing egression of mycobacterial antigens into the cytosol for better antigen presentation. Simultaneously, its urease C deficiency prevents VPM1002 from neutralizing the acidic pH inside the phagosome, which is required for the biological activity of Hly (16).

*In vitro* and animal studies have indicated an improved cross-priming and increased apoptosis in VPM1002 infected macrophages, resulting in improved vaccine efficacy over parental BCG (15, 16). Superior protection by VPM1002 was later shown to relate to earlier recruitment of type 1 cytokine producing T-cells and a profound capacity to produce type 17 cytokine responses, which was not seen after BCG vaccination (17). Moreover, VPM1002 stimulates enhanced AIM2 inflammasome activation, enhancing autophagy and secretion of (IL)-1β and IL-18 (18). Experiments with immunodeficient mice revealed higher safety of VPM1002 as compared with parental BCG. Enhanced safety of VPM1002 is also illustrated by a lower incidence of abscess formation in immunized infants (19).

VPM1002 has now completed phase I clinical trials in healthy adults (19, 20) and a phase IIa trial in South African newborns. It is currently undergoing phase II trial assessment in HIV-exposed neonates (NCT01479972). Within the ADITEC project we have generated gene expression profiles from infants participating in the phase IIa trial. Two groups of 11 neonates each receiving either VPM1002 or parental BCG were included, and gene expression profiles were analyzed from samples taken at the time of immunization and at weeks 2, 6, 12, 18, and 26 postvaccination. Safety and immunogenicity data from this clinical trial revealed that safety parameters for VPM1002 and the parental BCG strain are comparable. Both vaccines induced IFNγ responses, while VPM1002 vaccination in addition resulted in an increased proportion of IL-17 producing CD8 T-cells (19).

The gene expression profiles that were generated showed a substantial variability in responses between individuals (**Figure 1A**). Following vaccination early after birth, dramatic changes in the gene expression profiles over time were observed in both groups. The most pronounced changes were associated with upregulation of immunoglobulin-related genes, which likely reflects the maturation of the immune system in infants. In a direct comparison between the groups vaccinated with parental BCG and VPM1002, we did not observe any significant differences in gene expression at any of the time points analyzed. Most likely, differences in gene expression induced by the two vaccines, if any, are too small to be detected and may be obscured by the pronounced changes over time. Second, this study was set up as a safety trial with a limited number of individuals. As a consequence, the gene expression analysis was statistically underpowered.

Therefore, we harnessed the power of gene modules, using a new bioinformatics tool called "tmod" (J. Weiner https://cran.rproject.org/web/packages/tmod/index.html) to calculate enrichments in genes associated with particular biological responses. Results from this analysis indicated subtle functional differences between the vaccination groups that could not be detected at the single gene level (**Figure 1B**). For example, the VPM1002

vaccinated group showed earlier and more pronounced changes in genes involved in T-cell activation. Similarly, a prolonged enrichment in myeloid/monocyte-associated genes could be observed in the VPM1002 group (**Figure 1B**). In a direct comparison between the groups, we did observe a weak enrichment suggesting a slightly stronger T-cell activation in the first weeks following VPM1002 immunization. Also, a more pronounced downregulation of *PAX3* target genes and pro-inflammatory cytokine and chemokine expression in the VPM1002 group, although none of these enrichments were statistically significant (**Figure 1B**).

### NOVEL TECHNOLOGIES FOR EPITOPE MAPPING FOR VACCINE DESIGN, AND EXPLORATION OF Tfh CELLS AS NOVEL BIOMARKERS

The analysis of the B-cell repertoire could not be considered complete if the fully sequenced IgG are not functionally characterized, quantified in the serum, and the epitopes recognized are not defined. All this provides essential information that can be used to guide vaccine design and optimization, eventually to predict vaccination efficacy and its duration. Considering that the majority of the functional B-cell epitopes are discontinuous non-linear epitopes having 3D-conformational structures, it is important to apply methodologies able to decipher these sometimes complex structures.

We used HDX coupled to MS to allow the mapping of conformational epitopes (21–27). The approach was also compared with the most sophisticated and available approaches such as protein chip (26, 27), phage display (21, 26, 27), X-ray (21), and cryo-EM (24, 25). Modern HDX–MS is more straightforward, rapid, and routine than in the past. As a result, the breadth of applications of the method, including epitope mapping, has expanded (28). HDX–MS relies upon the rapid exchange of backbone amide hydrogen for deuterium when a protein is diluted into a deuterated buffer. Each exchange event increases the protein mass of 1 Da that can be monitored by MS. Epitope mapping through HDX is based on differential rates of deuterium incorporation by the antigen when it is bound or not with a specific mAb (29). When an antigen–antibody complex is formed, the interface between the two components can occlude solvent, thereby reducing the exchange rate due to the action of solvent steric exclusion. Deuterium incorporation of the free antigen and of the antigen–mAb complex is compared and antigen sequences that present reduced exchange kinetics when the antigen is bound to mAb are highlighted as potential epitopes.

In a first phase, we probed the binding between mAb 12C1 and fHbp, using hybrid approaches including peptide arrays, X-ray crystallography, phage display, and HDX–MS, evidencing that HDX–MS, although not the highest resolution method, is the most effective in providing nearly complete information about the structure of epitopes (21). The approach has sufficient resolution to recognize fHbp overlapping epitopes with different functional properties (22). It was also successfully applied to map epitopes from NadA (23, 26) and NHBA (27), the two other protective epitopes of the vaccine against group B meningococcus, Bexsero. Hybrid approaches making use of HDX–MS and electron microscopy also evidenced the power of HDX–MS to map viral antigen epitopes of CMV gH–gL complex. In addition, to better characterizing the immune response to vaccination and to support antigen design, this method that speeds up the elucidation of recognized epitope will contribute to increase the number of available structures of antigens and antigen–antibody complexes, opening new possibilities for the development of novel tools that might reliably predict protective epitopes.

# INFLUENZA-SPECIFIC CIRCULATING T-CELL SUBSETS AS EARLY PREDICTORS OF SPECIFIC ANTIBODY RESPONSES

Development of vaccines is a long endeavor requiring long and expensive clinical trials to prove the efficacy of vaccines in the relevant target populations. The availability of biomarkers predicting protection or predicting protective immune responsiveness may accelerate the development of vaccines.

T follicular helper cells are a CD4+ T-cell subpopulation that is identifiable in lymph nodes and tonsils. Tfh cells are specialized in providing help to B-cells (30–32). The identification of a circulating counterpart of the Tfh subset in the blood (33, 34) would allow the measurement of these cells following vaccination, which may help defining markers predicting vaccine efficacy. Therefore, the question was to investigate whether IL-21+ CD4+ T-cells induced by specific vaccination were detectable in human blood, if and how vaccination modulated their frequency, and whether their expansion correlated with increased titers of functional antibodies. Indeed, following vaccination with MF59-adjuvanted avian H5N1 vaccine, H5N1-specific IL-21+ CD4+ T-cells were detectable in the blood, expanded after vaccination and accumulated in the CXCR5−ICOS1+ subset. The rise of vaccine-specific ICOS1+ IL-21+ CD4+ T-cells appeared to predict the postvaccination increase of functional antibodies in these vaccines. Finally, circulating CXCR5−ICOS1+ CD4+ T-cells contained increased numbers of T-cells able to help influenza-specific B-cell differentiation, such that they differentiated *in vitro* into antibody-secreting cells in a manner that was dependent on IL-21− and ICOS1 (35). More recent studies have confirmed these findings and have also shown that the early expansion of cells with a Tfh phenotype predicts the long term persistence of neutralizing antibodies against influenza virus (36). Thus, the expansion of antigen-specific ICOS1+ IL-21+ CD4+ T-cells in the circulation may represent an early predictor of a vaccine's ability to stimulate vaccine-specific immunity and a useful surrogate marker of a vaccine's immunogenicity in human beings.

### HETEROGENEITY OF MEMORY CD4**+** T-CELLS INDUCED BY PATHOGENS OR VACCINES

CD4+ T helper cells are crucial players in the adaptive immune response, contributing to protection against a wide range of pathogens through the functional regulation of other immune and non-immune cell types. CD4+ T helper cells are characterized by functional diversity that has evolved to provide the most appropriate type of response against different classes of pathogens in different tissues (37). Th1 cells produce IFNγ, express the transcription factor T-bet and, through the activation of macrophages, contribute to immunity against intracellular pathogens. Th2 cells produce IL-4, IL-5, and IL-13, express the transcription factor GATA-3 and, through the activation of eosinophils and mast cells, contribute to protection against helminth parasites. Th17 cells produce IL-17A, IL-17F, and IL-22, express RORγt and together with neutrophils, mediate protection against fungi and extracellular bacteria. Th1, Th2, and Th17 cells differ also in the expression of chemokine receptors that control their homing ability, a property that is coordinately obtained during the process of T-cell differentiation (38). CCR5 and CXCR3 are expressed by Th1 cells, CCR4, CCR3, and CRTh2 by Th2 cells, and CCR6 and CCR4 by Th17 cells. While the role of distinct subsets of CD4+ T-cells in protection against different types of pathogens has long been recognized, it has been unclear whether a given vaccine or pathogen will induce a single type of T-cells and whether a single naive T-cell challenged by a pathogen or by a vaccine may acquire multiple T-cell fates.

We combined antigenic stimulation, TCR deep sequencing and cloning of human Th1, Th2, and Th17 memory subsets, to study the distribution and TCR repertoire of pathogen- and vaccine-specific T-cells in immune donors (39). We found that memory T-cells induced by *Candida albicans* are present at high frequency in a CCR6+ compartment, which comprises Th17 and a subpopulation of Th1 cells co-expressing T-bet and RORγt (defined as Th1\*), and at low frequency in a CCR6− compartment, which comprises classical Th1 and Th2 cells (40) (**Figure 2**). We could then demonstrate, using next-generation TCR Vβ sequencing, that multiple different clonotypes were present in more than one subset and, in several cases even in all subsets, whereas by contrast other clonotypes were found only in one particular subset. We also studied tetanus toxoid (given with alum as adjuvant) vaccine-induced memory T-cells. Also in this case, TCR Vβ sequencing revealed a high level of clonotype sharing among Th1, Th2, Th1\*, and Th17 subsets, with multiple different clonotypes represented in three and even all four subsets. By contrast, *Mtb*-specific memory T-cells in healthy donors were highly enriched in the Th1\* subsets, with some specific T-cells present in the Th17 subset. In contrast to what was observed for *C. albicans*, however, only very few clonotypes were shared between *Mtb*-specific Th1\* and Th17. Taken together, our results indicate that human T-cells induced by pathogens or vaccines are functionally heterogeneous and comprise both clones polarized toward a single fate, as well as clones whose progeny has acquired

multiple fates. The highly significant intraclonal heterogeneity observed in this work further supports the one cell-multiple fates model of CD4 T-cell differentiation; furthermore these data reveal T-cell plasticity in the context of the human T-cell immune response. An important question that this study raises is whether the induction of functionally diverse T-cell subsets by for instance a vaccine is advantageous for the host. Clearly, defective induction of the correct type of Th cell response may increase susceptibility to infections, while the induction of a wider spectrum of different T-cell types, including effector and memory T-cells with different migratory abilities would enlarge the range of differentiated precursors that the host could recruit and expand whenever necessary.

## A NEW CLASS OF RARE INFLUENZA-NEUTRALIZING ANTIBODIES TARGET A CONSERVED SITE IN THE HA STEM

By immortalizing memory B-cells from donors upon influenza vaccination, a new type of rarely occurring influenzaneutralizing antibodies, targeting a conserved site in the HA stem was found.

High throughput cellular screens have been developed and used to isolate potently and broadly neutralizing antibodies against a plethora of pathogens. We used EBV in the presence of CpG to immortalize memory B-cells with high efficiency, and screened culture supernatants for the presence of the desired antibodies using multiple assays based on binding or viral neutralization. Using this approach, we studied the antibody response to influenza HA. The neutralizing antibody response is dominated by antibodies binding the most variable part of the HA, the globular head, which is undergoing antigenic drift continuously. As a consequence, most of anti-head antibodies, which are detected using the classical hemagglutination inhibition assay, neutralize only a few isolates within a given subtype. We and others have previously described heterosubtypic neutralizing antibodies that can bind and neutralize different influenza virus subtypes, comprising group 1 viruses and even group 1 and group 2 viruses (41). The broadly neutralizing antibodies target a conserved site in the HA stem and use a particular VH gene, VH1-69. These antibodies are highly mutated, and they are produced by some but not all individuals. This would suggest that a series of somatic mutations may be necessary for the development of such antibodies. To address the developmental pathway of anti-stem antibodies, we isolated, from a single donor, 197 VH1-69+ anti-stem antibodies. Using sequence information, we performed a genealogical analysis to reconstruct the developmental pathways of a number of the VH1-69 clones. We identified key elements which were required for affinity maturation. In all cases tested, the binding to HA was exclusively dependent on the mutated VH, while the light chain did not play a significant role since it could be substituted by irrelevant L chains. Strikingly, in most clones, affinity maturation was achieved through a single somatic mutation leading to a proline to alanine substitution at position 52a in HCDR2. This mutation was the first to occur in the B-cell clone and was sufficient to confer high-affinity binding to the selecting H1 antigen. This observation is consistent with rapid affinity maturation. Unexpectedly, however, further favorable mutations continued to accumulate, which further increased the breadth of antibody reactivity while making both the initial mutations functionally redundant. We also determined that phenylalanine 54, which was characteristic of all antibodies, was an essential binding residue. Interestingly, this position is polymorphic and individuals lacking a phenylalanine 54 allele did not produce VH1-69 antibodies to the HA stem. Finally, all VH1-69 anti-stem antibodies shared a 13-aminoacid-long HCDR3 with a tyrosine at position 98. This study defines VH1-69 polymorphism, HCDR3 sequence constraints and an individual somatic mutation as the three requirements involved in the generation of broadly neutralizing antibodies. It also reveals an hitherto unknown and unexpected redundancy in process of affinity maturation (42).

## MOLECULAR MECHANISMS REGULATING T-CELL TRAFFICKING AT MUCOSAL SITES IN HEALTH AND DISEASE: CHRONIC IMMUNE ACTIVATION DAMPENS LEUKOCYTE TRAFFICKING AND CALLS FOR NOVEL VACCINATION STRATEGIES

The homing of leukocytes in general, and antigen-specific T-cells in particular, to peripheral tissues, mucosal sites, and secondary lymphoid organs is controlled, among others, by the local production of chemokines, the expression of chemokine receptors on the cell surface, and an efficient cytoskeleton machinery (43, 44). The characterization of the surface expression of the different chemokine receptors on T-cells has guided the discovery of different T helper cell subsets and is still a precious tool for the characterization of novel functional subsets. Nevertheless, scanty information is available on the influence that pathological conditions can have on the activity of chemokine receptors, even if their expression on the T-cell surface remains unaltered compared with healthy individuals. In health and disease, the microenvironment can further control cell migration, by releasing factors that cooperate with chemokines for enhancing cell responses (45–49), or by producing natural chemokine antagonists that block chemokineinduced activities (49). In addition, to the proteins produced by the microenvironment, systemic chronic immune activation can dampen T-cell responses to chemokines (50). HIV-1 infection is a clear example for this phenomenon, since chronic immune activation due to microbial translocation has been associated with poor T-cell repopulation of the intestinal mucosa during ART (51, 52). CD4+ T-cells expressing the chemokine receptors CCR6+ and/or CXCR3+ can traffic to the intestine in response to CCL20 and to IFNγ-induced chemokines, to help maintain the integrity of the mucosal barrier (53). In HIV-1 infected patients, regardless of ART therapy, CD4+ T-cells inefficiently migrate after chemokine triggering, due to an unproductive polymerization of actin (50). Interestingly, an *in vivo* model of sustained toll-like receptor (TLR) 7 triggering with R848 recapitulated chronic immune activation, lymphoid system disruption, and poor response to chemokines as observed in HIV-1-mediated pathology (50, 54). It is indeed the persistent immune activation, and not the presence of the virus, that is causing the impaired response to chemokines. This deficiency is due to alterations in the cytoskeleton machinery that could be used as a marker to assess T-cell anergy caused by TLR7 triggering (55). Pharmacological intervention acting on the cytoskeleton can restore a proper response to chemokines both *in vitro* and *in vivo* (50) and represents novel therapeutic approaches, which are also amenable for combination with other therapies such as ART. It will be important to verify indeed whether such new immunotherapeutic interventions, aimed at dampening chronic immune activation during HIV-1 infection (56), result in efficient restoration of leukocyte migration. This is of particular interest since chronic activation of the immune system is not only present in HIV-1 infection, but is also a signature of other persistent infections as well as autoimmunity. Moreover, inflammation is a major component of senescence and aging-associated pathology (57–59). The study of frail cohorts of individuals and patients will identify those in need of receiving novel vaccine formulations that would improve the ability of leukocytes to migrate to the site of vaccination and later also during infection to properly mount an effective immune response.

# DISSECTING INNATE IMMUNITY AND DEVELOPING TOOLS FOR INNATE IMMUNE BIOPROFILING

Activation of an innate inflammatory response is a key step in the mechanisms of action of adjuvants. Gene expression profiling of the mouse muscle injected with different adjuvants revealed modulation of a cluster of innate immunity genes, including particularly: cytokines, cytokine receptors, chemokines, adhesion molecules, and proteins involved in host–pathogen interactions (60). In particular the long *PTX3*, a soluble pattern recognition molecule identified in the 1990s (61), was among the over 1,000 genes upregulated by adjuvant treatment. PTX3 is classified as a member of the pentraxins, a family of highly conserved molecules characterized by a unique multimeric structure (62, 63; Garlanda et al., forthcoming2 ). Based on the sequence of the composing promotors, two subfamilies can be distinguished: short and long pentraxins. C-reactive protein (CRP) and serum amyloid P component are prototypic short pentraxins, by contrast PTX3 is a prototypic long pentraxin subfamily member.

Despite their sequence homologies, CRP and PTX3 differ in molecular structure, gene organization and cellular sources. CRP is produced only by the liver in response to IL-6 while PTX3 is produced locally and more rapidly by myeloid and stromal cells in response to primary pro-inflammatory cytokines (IL-1β

and TNF-α) and ligation of TLRs. In addition, PTX3 is stored in neutrophils-specific granules, from which it is promptly released upon microbial recognition (63). From a functional point of view, PTX3 is endowed with multifunctional properties at the crossroad of inflammation, innate immunity, tissue remodeling, and female fertility. Of note, PTX3 plays an essential non-redundant role in resistance to selected microbes. The protein has antibody-like properties, binding selected pathogens (i.e., *A. fumigatus*, *Pseudomonas aeruginosa*, and more) and conserved microbial moieties (outer membrane protein A from *Klebsiella pneumoniae*; HA glycoprotein from influenza virus), and facilitating phagocytosis of recognized microbes in a manner dependent on Fcγ receptors and complement (64, 65). Moreover, PTX3 orchestrates complement activity (66) and, by regulating complement-dependent tumor promoting inflammation, can act as an extrinsic oncosuppressor gene in murine and human tumors (67). Finally, recent results obtained in different models of tissue damage highlighted a non-redundant role of PTX3 in remodeling and repair of tissue *via* its interaction with fibrin (68). This further supports the evidence that the recognition of matrix and microbial components is shared ancestral features of the humoral arm of the innate immune system.

In view of the upregulation of PTX3 by selected adjuvants we investigated the role of this protein in the antibody response using a well-known model of vaccination with OMV from *Nm*. We found that PTX3 binds *Nm* as well as OMV derived from *Nm*, exerting a protective role in a model of infection with *Nm* in the infant rat. *Ptx3*-deficient mice vaccinated with OMV without any adjuvant developed a lower antibody response compared with WT mice. In addition, co-injection of PTX3 enhanced the antibody response, especially in *ptx3*-deficient mice. Recognition of the antigen is essential for the effect exerted by PTX3, since immunization with an antigen not recognized by PTX3, such as ovalbumin, induced similar response in WT and *ptx3−/−* mice (69). This observation was confirmed by further investigations showing that *ptx3−/−* mice produced lower levels of IgM in response to administration of Pneumovax, a human vaccine containing capsular polysaccharide from multiple *Streptococcus pneumoniae* (*Sp*) serotypes (70). Similarly, *ptx3−/−* mice produced less antibodies following immunization with PR8 influenza A virus, which is also recognized by PTX3 (71), and in both cases PTX3 administration restored the response. Searching for an explanation of these effects, it has been found that PTX3 was released by B helper neutrophils (NBh) and bound to B-cells located in the marginal zone (MZ) in the spleen. NBh are an innate-like subset of antibody-producing cells positioned at the interface of the adaptive immune system and the circulation. The interaction between PTX3 and MZ-B-cells activated signals that were independent of FcγR but triggered class-switching from IgM to IgG. PTX3 promoted MZ-B-cell differentiation into extrafollicular plasmablasts and plasma cells, and enhanced IgM and IgG responses to the encapsulated bacterium *Sp*, or after immunization with capsular polysaccharides or bacterial carbohydrates. These results indicate that the humoral pattern recognition molecule PTX3, produced in response to pro-inflammatory cytokines or selected adjuvants, may amplify

<sup>2</sup>Garlanda C, Bottazzi B, Magrini E, Inforzato A, Mantovani A. PTX3, a key component of humoral innate immunity at the interface between defense and tissue remodeling. *Physiol Rev* (Forthcoming).

effective adaptive antibody responses that are induced by antigens recognized by PTX3, thus serving as an endogenous adjuvant (**Figure 3**).

The data outlined above provide a rationale to evaluate whether PTX3 could be a correlate of the shaping of the immune response induced by adjuvants in humans. In collaboration with D. J. Lewis and G. Del Giudice, PTX3 and CRP plasma levels were measured in a cohort of individuals injected with placebo or with licensed influenza vaccines adjuvanted or not. This study received ethical approval from London—Surrey Borders Research Ethics Committee (REC Ref: 13/LO/0044) and was registered on ClinicalTrials.gov before enrollment (NCT01771367). Preliminary unpublished results evidenced an increase in PTX3 plasma levels at early time points in individuals injected with the adjuvanted vaccine, confirming in humans the effects on PTX3 gene upregulation initially observed in mice (60). In addition, after immunization with an adjuvanted vaccine we observed an earlier induction of PTX3 (peak at 24 h postinjection) compared with CRP (peak at 48 h postinjection), likely due to the local expression of PTX3 versus the systemic production of CRP.

Several single nucleotide polymorphisms (SNPs) have been described in the human *PTX3* gene, mainly located in its non-coding regions with the only exception of one exonic SNP causing an amino acid variation in position 48 (Asp48Ala). *PTX3* haplotypes are associated with increased susceptibility to lung TB (72), *P. aeruginosa* infections in cystic fibrosis Caucasian patients (73), *A. fumigatus* infections in bone marrow transplanted patients (74), and urinary tract infections (75). Results in a large cohort of patients with fungal infections after solid organ transplantation (76) and in 2,609 bone marrow transplanted patients and their donor pairs (77) further reinforced the association of *PTX3* gene SNPs and susceptibility to fungal infection. The protection-associated haplotypes were also associated with higher PTX3 protein expression and circulating levels (74, 76), supporting an active role of PTX3 as non-redundant player involved in the innate defense against recognized pathogens.

In summary PTX3, a molecule of the innate immune system, produced in response to pro-inflammatory mediators, not only acts as an antibody-like molecule, recognizing pathogens and promoting their removal but also helps in antibody production by adaptive immunity, acting as endogenous adjuvant. This is schematically depicted in **Figure 3**.

# CONCLUDING REMARKS

Understanding the factors that are critical for the induction of protective immunity by vaccines is key for the rational development of novel vaccines and to optimize vaccine efficacy. Within the ADITEC work package "Molecular signatures of immunity and immunogenicity," the aim was to identify key molecular mechanisms of innate and adaptive immunity following vaccination and to uncover correlates of protection at the transcriptomic, proteomic and cellular level.

Transcriptomic profiling identified IFN-inducible gene *FCGR1A* as one of the most robust and consistent single biomarkers that could serve as part of reliable predictive signatures discriminating active TB disease from latent infection, identifying individuals at risk of developing TB, and monitoring TB treatment response kinetics. Such signatures are useful in stratifying risk groups for preventive treatment or clinical vaccine testing, as well as minimizing treatment duration and preventing possible *de novo* drug resistance, independent of age, immune status and genetic background. By contrast, when comparing transcriptome profiles between infants vaccinated with parental or recombinant BCG VPM1002 to unravel the mechanisms underlying superior protection against TB by VPM1002, differences at the single gene level could not be detected. Therefore, a new bioinformatics tool (tmod) was used to calculate enrichments in genes, uncovering more pronounced changes in T-cell activation and myeloid/monocyte-associated genes in the VPM1002 vaccinated group. Complementing the work on transcriptomics, HDX–MS was used to determine quantitative proteomic profiles and was successfully applied to define functionally active epitopes and antibodies in polyclonal sera following vaccination for various bacterial vaccines.

At the cellular level, early expansion of vaccine-specific CD4+ Tfh cells in peripheral blood was found to predict long term

# REFERENCES


persistence of neutralizing antibodies against influenza virus, suggesting this to be a possible biomarker of protective vaccine efficacy. In addition, it was observed that vaccine-induced human CD4+ T-cells are functionally heterogeneous and comprise not only clones polarized toward a single fate but also clones that fit the one cell-multiple fates model of CD4 T-cell differentiation. This suggests that inducing broader range of effector and memory T-cells with broad and different functional properties would benefit the host by expanding the number of options to respond to pathogens. Parameters affecting the ability of leukocytes to properly migrate to the site of vaccination and mount an effective immune response were also investigated. Interestingly, persistent immune activation caused impaired T-cell trafficking at mucosal sites, which was due to alterations in the cytoskeleton machinery. Pharmacological intervention acting on the cytoskeleton could restore a proper response to chemokines, calling for novel vaccination strategies in frail individuals. Finally, PTX3 was identified as a molecule produced by the innate immune system in response to pro-inflammatory mediators and was shown to be involved in the innate defense against recognized pathogens. PTX3 functions as an endogenous adjuvant, bridging the humoral arms of the innate and adaptive immune systems.

In summary, within the ADITEC work package "Molecular signatures of immunity and immunogenicity," dissecting the human innate and adaptive immune response to vaccination and infection identified not only novel biomarker signatures at the transcriptomic, proteomic, and cellular level that could serve as potential correlates of risk and of protection but also uncovered novel markers and mechanisms underlying protective immunity. Validating and further expanding on these findings will advance our understanding of protective immunity, which is a key step toward the development of effective novel vaccines.

# AUTHOR CONTRIBUTIONS

All the authors have contributed a section to the manuscript, based on their research in the EC FP7 ADITEC project.

# FUNDING

The authors gratefully acknowledge funding by the European Commission, EC FP7 ADITEC Grant Agreement No. 280873. The text represents the authors' views and does not necessarily represent a position of the Commission who will not be liable for the use made of such information.


ethnicity. *Clin Microbiol Infect* (2014) 20:O230–8. doi:10.1111/1469-0691. 12383


and pentamer reveal mutually exclusive cell entry complexes. *Proc Natl Acad Sci U S A* (2015) 112:1767–72. doi:10.1073/pnas.1424818112


[corrected]. LXXXIX. Update on the extended family of chemokine receptors and introducing a new nomenclature for atypical chemokine receptors. *Pharmacol Rev* (2014) 66:1–79. doi:10.1124/pr.113.007724


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Haks, Bottazzi, Cecchinato, De Gregorio, Del Giudice, Kaufmann, Lanzavecchia, Lewis, Maertzdorf, Mantovani, Sallusto, Sironi, Uguccioni and Ottenhoff. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Susan L. Baldwin1†, Fan-Chi Hsu1†, Neal Van Hoeven1 , Emily Gage1 , Brian Granger1 , Jeffrey A. Guderian1 , Sasha E. Larsen1 , Erica C. Lorenzo2 , Laura Haynes2 , Steven G. Reed1 and Rhea N. Coler1,3,4\**

*<sup>1</sup> Infectious Disease Research Institute, Seattle, WA, United States, 2Center on Aging, Department of Immunology, University of Connecticut School of Medicine, Farmington, CT, United States, 3Department of Global Health, University of Washington, Seattle, WA, United States, 4PAI Life Sciences, Seattle, WA, United States*

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Randy A. Albrecht, Icahn School of Medicine at Mount Sinai, United States Katie Louise Flanagan, Monash University, Australia*

### *\*Correspondence:*

*Rhea N. Coler rhea.coler@idri.org*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 12 September 2017 Accepted: 01 February 2018 Published: 19 February 2018*

### *Citation:*

*Baldwin SL, Hsu FC, Van Hoeven N, Gage E, Granger B, Guderian JA, Larsen SE, Lorenzo EC, Haynes L, Reed SG and Coler RN (2018) Improved Immune Responses in Young and Aged Mice with Adjuvanted Vaccines against H1N1 Influenza Infection. Front. Immunol. 9:295. doi: 10.3389/fimmu.2018.00295*

Elderly people are at high risk for influenza-related morbidity and mortality due to progressive immunosenescence. While toll-like receptor (TLR) agonist containing adjuvants, and other adjuvants, have been shown to enhance influenza vaccine-induced protective responses, the mechanisms underlying how these adjuvanted vaccines could benefit the elderly remain elusive. Here, we show that a split H1N1 influenza vaccine (sH1N1) combined with a TLR4 agonist, glucopyranosyl lipid adjuvant formulated in a stable oil-in-water emulsion (GLA-SE), boosts IgG2c:IgG1 ratios, enhances hemagglutination inhibition (HAI) titers, and increases protection in aged mice. We find that all adjuvanted sH1N1 vaccines tested were able to protect both young and aged mice from lethal A/ H1N1/California/4/2009 virus challenge after two immunizations compared to vaccine alone. We show that GLA-SE combined with sH1N1, however, also provides enhanced protection from morbidity in aged mice given one immunization (based on change in weight percentage). While the GLA-SE-adjuvanted sH1N1 vaccine promotes the generation of cytokine-producing T helper 1 cells, germinal center B cells, and long-lived bone marrow plasma cells in young mice, these responses were muted in aged mice. Differential *in vitro* responses, dependent on age, were also observed from mouse-derived bone marrow-derived dendritic cells and lung homogenates following stimulation with adjuvants, including GLA-SE. Besides enhanced HAI titers, additional protective factors elicited with sH1N1 + GLA-SE in young mice were observed, including (a) rapid reduction of viral titers in the lung, (b) prevention of excessive lung inflammation, and (c) homeostatic maintenance of alveolar macrophages (AMs) following H1N1 infection. Collectively, our results provide insight into mechanisms of adjuvant-mediated immune protection in the young and elderly.

Keywords: influenza, vaccine, adjuvant, elderly, T helper 1, H1N1

### INTRODUCTION

Influenza and influenza-related complications are leading causes of death in elderly populations. Older people exhibit increased morbidity and mortality in response to influenza infection due to uncontrolled respiratory inflammation, severe pneumonia, or multi-organ inflammation and failure (1). Although the annual influenza vaccine coverage rate has increased, the CDC estimates that

**256**

individuals older than 65 years of age who have received influenza vaccines in consecutive years are still at high risk of influenza infection (2).

The process of immunological aging, also called immunosenescence, is associated with a progressive loss of functional physiological integrity including collectively increasing DNA damage and genome instability, stem cell exhaustion, cellular senescence, and altered intercellular communication, among other processes [reviewed in Ref. (3)]. Immunosenescence causes the attenuation of both innate and adaptive immune systems including reduced levels and function of TLRs in macrophages and plasmacytoid dendritic cells (4, 5), decreased output of naïve B cell numbers (6), severe reduction of thymopoiesis (7), and insufficient immune synapse generation (8). Furthermore, aged individuals also demonstrate a significant reduction in the quantity, but not quality of antigen-specific antibody responses, and this reduction in antibody quantity is concurrent with a decrease in antigen-specific plasmablasts (9). Indeed, defective antigen presentation and reduced T cell and B cell repertoires in aged individuals results in poor cellular, humoral, and vaccineinduced responses.

As vaccination is the most effective method to prevent infection, ways to improve influenza vaccines for use in the aging human population are continually being explored. These strategies include adjuvant use, intradermal delivery, increased dosage, and altering antigen selection that typically make up seasonal vaccines [reviewed in Ref. (10)]. The first seasonal influenza vaccine containing an adjuvant (MF59), also known as Fluad™, was approved by the FDA in 2015 for use in people over the age of 65 years. Fluzone® High-Dose vaccine (an inactivated, split influenza virus vaccine containing 60 µg of HA for each component of the vaccine, rather than 15 µg) is also approved for use in people ≥65 years of age. Despite advancements in alternative vaccine options for the elderly population, they are still disproportionally affected and continue to exhibit severe morbidity and mortality each year from seasonal influenza infections. Thus, a need still remains for better strategies, including additional adjuvant options, for the influenza vaccine to overcome the challenges of immunosenescence in the elderly.

The disparity of morbidity and mortality burden in elderly individuals is partly due to the incomplete understanding of how adjuvanted vaccines may circumvent immunosenescence in elderly people. Mechanistic details regarding how different adjuvants work in the elderly may reveal ways in which adjuvant formulations can be tailored and adapted for optimal responses that provide enhanced protection for this susceptible population. Administration of adjuvants, including squalene-based oil-in-water emulsions, and those including a TLR4 agonist, have been leveraged in vaccines in order to enhance immune responses in multiple studies (11–18). These adjuvants include MF59 (as described above), and AS04 (comprised of monophosphoryl lipid A and Alum) included in FDA approved vaccines against hepatitis B virus and human papillomavirus (11). In preclinical studies, we have shown that glucopyranosyl lipid A (a synthetic TLR4 agonist) formulated in a stable oil-in-water emulsion (GLA-SE) expands immune responses to Fluzone® (12). GLA has also been included in several completed and ongoing human clinical studies for vaccines against *Leishmania* (NCT01751048), *Mycobacterium tuberculosis* (NCT02508376, NCT0246516, NCT01599897), HIV (NCT01966900, NCT01922284), Schistosomiasis (NCT03041766), malaria (NCT02647489, NCT01540474), and avian influenza (NCT01991561, NCT01147068). In humans, GLA-SE combined with rH5 was considered safe and improved antibody titers compared to the recombinant protein alone (15). Furthermore, no defect has been identified in humans following stimulation with GLA-SE on antigen-presenting cells (APCs) from aged compared to young APCs (16) and enhances T cell responses from older adults following stimulation of peripheral blood mononuclear cells (PBMCs) with live influenza virus (19). Therefore, our approach in this study was to employ the use of different adjuvants combined with an influenza vaccine to overcome the challenge of immunosenescence in aged animals.

In this investigation, one of our main findings reveal that aged mice have dampened cytokine responses to *in vitro* stimulation of DCs and lung homogenates with a T helper 1 (Th1) cytokineinducing agonist. In addition, we show that aged mice require two immunizations with adjuvanted sH1N1 vaccines for robust *in vivo* protection against influenza whereas young mice are protected after a single immunization. Finally, we demonstrate that immunization in young mice with sH1N1 + GLA-SE results in enhanced alveolar macrophage (AM) homeostasis within the lung, increased TLR7 expression within the AMs, and faster clearance of virus after H1N1 challenge compared to mice immunized with sH1N1 + SE or vaccine alone. These results emphasize the age-related differences following immunization, the ability to improve responses to influenza infection in the elderly with two immunizations, and a mechanism of enhanced vaccine protection against influenza with a TLR4 agonist adjuvant.

### MATERIALS AND METHODS

### Mouse Model

Young female C57BL/6 or CB6F1 mice were purchased from Charles River Laboratories (Wilmington, MA, USA) or the Jackson Laboratories (Bar Harbor, ME, USA) and were housed and maintained under specific pathogen free conditions at the Infectious Disease Research Institute. Experiments that included female C57BL/6 or CB6F1 mice aged 18–21 months were acquired with special request from the National Institute on Aging, from the aged rodent colony (National Institute of Health, Bethesda, MD, USA). Mice were housed in a biosafety level 2 (BSL2) environment for the entirety of these studies (including H1N1 challenge studies), and all procedures were performed in accordance with the regulations and guidelines of the IDRI animal care and use committee.

### Adjuvants and Immunization

Vaccines were formulated in saline, SE (2% final v/v oil concentration), an MF59-like adjuvant (2.5% final v/v oil concentration), or GLA-SE (5 µg of GLA in 2% SE). Mice were immunized intramuscularly (i.m.) one or two times 3 weeks apart. The split H1N1 influenza vaccine (sH1N1) (kindly provided by Novartis) and recombinant H1 (rH1) (A/California/4/2009; Protein Sciences Corp., Meriden, CT, USA) was used at 0.01 or 0.1 µg, respectively, for the immunizations. Serum was collected 3 weeks after prime or boost immunizations. Spleens and bone marrow were harvested for immunogenicity studies either 1 or 4 weeks following immunization(s) as described.

### Endpoint Antibody Titers

Sera were analyzed for H1-specific IgG1 and IgG2c endpoint antibody titers by antibody capture ELISA. Polysorp ELISA plates (Nunc-immuno polysorp 96 well plates, VWR) were coated with rH1 (A/H1N1/California/2009) (Protein Sciences Corp., Meriden, CT, USA) at a concentration of 1 µg/ml in 0.1 M bicarbonate coating buffer at 100 µl per well for 4 h at room temperature. Plates were then blocked with a 0.05% PBS-Tween solution plus 1% BSA and incubated overnight at 4°C, followed by five washes in 0.1% PBS-Tween and one PBS wash. Serially diluted mouse sera were added and plates incubated at room temperature, on a shaker, for 2 h. Plates were washed, dried, and secondary antibodies [IgG1-horseradish peroxidase (HRP) and IgG2c-HRP, Southern Biotechnologies, Birmingham, Al, USA] were added at a 1:2,000 dilution. Plates were incubated at room temperature for 1 h, washed, and eBioscience™ TMB solution (Thermo Fisher Scientific) was added to the plates. The enzymatic reaction was stopped with 50 µl per well of 1N H2SO4. Plates were then read on a VERSAmax microplate reader (Molecular Devices) at 450 nm with a reference filter set at 570 nm. Endpoint titers were determined as the last dilution to render a response of greater than 0.1 mean optical density using Prism software (GraphPad Software, La Jolla, CA, USA).

## Hemagglutination Inhibition (HAI) Antibody Responses

Hemagglutination inhibition assays were performed according to World Health Organization guidelines. Briefly, sera was treated with receptor destroying enzyme (from *Vibrio cholerae*, Denka-Seiken, Tokyo, Japan) overnight and heated at 56°C for 30 min to deactivate the enzyme. HAI antibodies were then tested against the vaccine strain (A/H1N1/California/4/2009) with 0.5% turkey red blood cells (Thermo Fisher Scientific). The HI titer was defined as the reciprocal of the highest dilution of sera, which completely inhibited the agglutination of the RBCs. All samples were run in duplicate, and pre-immune titers in all mice were ≤5.

### Weight and Survival Measurement

Ten mice/group were immunized i.m. either once or twice, 3 weeks apart. Three weeks after the prime or boost immunization, mice were infected *via* the intranasal route with 100LD50 A/H1N1/California/4/2009. Survival and clinical signs (weights) were assessed daily, over 14 consecutive days. Animals exhibiting significant weight loss or that were under duress (ruffled fur, weighted breathing, or hunched backs) were sacrificed. All procedures with A/H1N1/California/4/2009-infected mice were performed under BSL2 conditions per IACUC procedures and guidelines.

# *In Vitro* Stimulation of Bone Marrow Dendritic Cells (BMDCs) and Lung Homogenates

Bone marrow was harvested from the femurs of vaccinated mice, and single cell suspensions were prepared at a concentration of 2 × 105 cells/ml in complete RPMI-1640 with 20 ng/ml recombinant mouse GM-CSF (rmGM-CSF; PeproTech). On day 3, additional rmGM-CSF was supplied. BMDCs were harvested on day 6 and were seeded at a concentration of 6.6 × 105 cells per well. Cells were stimulated with a range of 2–20 µg/ml GLA-SE, 0.00008–0.8% SE, or 10 ng/ml Lipid A 506 (a TLR4 agonist as positive control, Peptide Institute Inc., Osaka, Japan) in the presence of 20 ng/ml rmGM-CSF. A sterile single cell suspension was prepared from young and aged mouse lung by passing tissue through a 100 µM cell strainer (Thermo Fisher Scientific). Erythrocytes were lysed using ACK lysis buffer (Gibco by Life Technologies), while remaining leukocytes were enumerated and plated in triplicate wells at a seeding density of 5 × 105 cells per well. Lung homogenates were stimulated with the same conditions as BMDCs described above. Following an 18-h stimulation, cell supernatants from BMDCs or lung homogenates were collected, and cytokine levels in the supernatants were determined by using custom Luminex-based multiplex immunoassay kits (Procarta Cytokine Assay Kit: Affymetrix, Santa Clara, CA, USA). Cell supernatants were incubated with polystyrene beads coated with antibodies corresponding to the different cytokines including: TNF-α, IL-10, IL12p40, and IL-6 and developed according to the manufacturer's instructions. Bead size and fluorescence were measured on a Luminex 200 and data were analyzed using the Masterplex QT software (Miraibio).

# Flow Cytometry

The antigen-specific T cell memory response generated by vaccination was determined by flow cytometry following incubation of splenocytes with 10 µg/ml rH1 (Protein Sciences Corp., Meriden, CT, USA). Cells were cultured at 1 × 106 cells per well in a 96-well plate (Corning Incorporated, Corning, NY, USA) in RPMI-1640 supplemented with 10% heat-inactivated FCS and 50,000 U penicillin/streptomycin (Invitrogen) for 18–20 h in the presence of GolgiStop (BD Bioscience). Each sample was incubated with Fc receptor blocking (clone2.4G2) before incubation on ice for 30 min with the following antibodies: anti-CD4 (clone RM4-5), anti-CD8α (clone 53-6.7), and anti-CD44 (clone IM7). Expression of selected cytokines was determined by incubation with anti-IFN-γ (clone XMG1.2), anti-IL-2 (clone JES6-5H4), anti-CD154 (clone MRI), anti-TNF-α (clone MP6-XT22), and anti-IL-17A (clone TC11-18H10.1). For Tfh cell analysis, singlecell suspension of inguinal lymph nodes (20) were prepared and stained with following antibodies: anti-CD4 (clone RM4-5), anti-CXCR5 (clone L138D7), anti-PD-1 (clone 29F.1A12), anti-B220 (clone RA3-6B2), anti-CD44 (clone IM7), and anti-CD8 (clone 53-5.8), F4/80 (clone BM8), CD11b (clone M1/70) for dump gate. For germinal center (GC) B cell analysis, single-cell suspension of spleen or inguinal LNs were prepared and stained with following antibodies: anti-B220 (clone RA3-6B2), anti-GL7 (clone G7), anti-CD95 (clone SA367H8), anti-IgD (clone 11-26c.2a), anti-IgG1 (clone RMG1-1). For lung immune cell analysis, lungs were chopped and digested in the presence of 70 µg/ml Liberase™ (Roche), 40 µg/ml DNase I (Roche), 10 mM Aminoguanidine (Sigma-Aldrich), and 5 mM KN-62 (Sigma-Aldrich). The homogenates were incubated for 30 min in a 37°C water bath. Single-cell suspensions were prepared by dispersing the tissues through a 70 µm nylon tissue strainer (BD Falcon). Cells were washed with 1× PBS and then stained with a fixable viability dye (Tonbo Biosciences) before being stained with following antibodies: anti-CD11b (clone M1/70), anti-CD11c (clone N418), anti-CD45 (clone 30-F11), anti-NK1.1 (clone PK136), anti-Ly6G (clone 1A8), anti-TLR2 (clone T2.5), anti-TLR3 (clone 11F8), anti-TLR4 (clone SA15-21), and anti-TLR7 (clone A94B10). Antibodies were purchased from BD Biosciences, eBioscience, BioLegend, or Tonbo Biosciences. Samples were analyzed with BD Fortessa or LSRII. Doublets and dead cells were excluded before analysis, and all the data were analyzed with FlowJo software (version 9; FlowJo, LLC).

# Enumeration of Long-Lived Antibody-Secreting Plasma Cells (ASPCs)

A bone marrow ELISPOT was used to determine the induction of vaccine-specific long-lived ASPCs following immunization with sH1N1 vaccine with and without adjuvants. ELISPOTs were performed as previously described (21) with minor revisions. The developed plates were counted by an ELISPOT plate reader (C.T.L. Serie3A Analyzer, Cellular Technology Ltd., Cleveland, OH, USA), and the data were analyzed using ImmunoSpot® software (Cellular Technology Ltd., Cleveland, OH, USA).

## Viral Load Measurement by Real-time Q-PCR

RNA from whole-lung samples was extracted with Trizol® reagent following the manufacturer's instructions (Thermo Fisher Scientific). The mixture of CHCl3 and Trizol was centrifuged at 11,500 *g* for 15 min at 4°C. After centrifugation, the aqueous layer was transferred and mixed with an equal volume of 70% RNase-free ethanol. The RNA extracts were purified with an RNA purification kit (Ambion®, Thermo Fisher Scientific). RNA was then reverse-transcribed with a SuperScript® IV first-strand synthesis system (Invitrogen®, Thermo Fisher Scientific) into cDNA. Expression of H1 was measured using customized TaqMan probes (Applied Biosystem, Thermo Fisher Scientific) including forward primer: 5′-ATTGCCGGTTTCATTGAAGG-3′; reverse primer: 5′-ATGGCATTCTGTGTGCTCTT-3′; probe: 5′-(FAM) ATGAGCAGGGGTCAGGATATGCAGCCGACC (TAMRA)-3′ to detect A/H1N1/California/4/2009 viral load (22). TaqMan probe for GAPDH was used as the internal control (Applied Biosystem, Thermo Fisher Scientific). Samples were analyzed using a Bio-Rad CFX384 real-time PCR detection system (Bio-Rad), and relative gene expression was calculated *via* the 2−ΔΔCT method.

# Statistical Analysis

Statistical analysis of antibody responses (endpoint antibody and HAI titers) and flow cytometry data was performed using one-way ANOVA with the Tukey multiple comparison test unless noted in the figure legend. Statistical analysis of cytokine production from BMDCs and lung homogenates represented in **Figure 4** were performed using two-way ANOVA and Sidak's posttest. A two-way ANOVA with the Tukey multiple comparison test was performed on data represented in **Figure 5** (with the exception of **Figure 5A**; which used one-way ANOVA and the Tukey multiple comparison test). Statistical analysis of survival curves was performed using the Log-Rank/Mantel–Cox test. All statistical analyses were performed using GraphPad Prism version 7 for Windows (GraphPad Software, La Jolla, CA, USA). *p-*Values of <0.05 were considered significant.

# RESULTS

### sH1N1 Vaccine Adjuvanted with GLA-SE Enhances IgG2c:IgG1 Ratios and High HAI Titers in Young CB6F1 Mice

We were particularly interested in determining whether adjuvants could help overcome the immunosenescence observed in elderly populations. The induction of antigen-specific IgG1 antibodies is dependent on Th2-biased immune responses, while class switching to IgG2c in mice is correlated with Th1-biased immune responses and has been well studied (20). We have previously reported that GLA-SE enhances antiviral protection through the induction of Th1-mediated immune responses (14) and, therefore, used this adjuvant formulation in this study. C57BL/6 mice were originally selected as the mouse strain for these studies; however, our data showed no evidence of HA (A/H1N1/California/4/2009) CD4 T cell epitopes in C57BL/6 mice, whereas CB6F1 mice have a confirmed CD4 T cell epitope (manuscript in preparation). These data provided the rationale for switching to the CB6F1 mouse strain for the remainder of the studies. As shown in Figure S1 in Supplementary Material, the recombinant rH1 vaccine combined with adjuvants MF59-like, SE, or GLA-SE in C57BL/6 mice induced both IgG1 and IgG2c responses post boost. An enhanced IgG2c:IgG1 bias was observed in C57BL/6 mice immunized with rH1 + GLA-SE (Figure S1B in Supplementary Material). While HAI titers in young C57BL/6 mice given adjuvanted rH1 were all >1:40 after two immunizations, much lower HAI titers were observed in aged C57BL/6 mice, where only one out of three of the aged animals in the MF59 and GLA-SE adjuvanted groups tested had post-vaccination HAI titers of >1:40 (Figure S1C in Supplementary Material).

We used the experimental design outlined in **Figure 1A** to assess the use of adjuvants in sH1N1 vaccine regimens and evaluate induced adaptive immune responses. Young (1 month old) or aged (18–21 months old) CB6F1 mice were immunized twice with the sH1N1 vaccine control (saline), or adjuvanted with MF59-like adjuvant, SE, or GLA-SE. Three weeks following the boost immunization, H1-specific IgG1 and IgG2c antibodies were evaluated (**Figure 1B**). We found that both young and aged mice receiving adjuvanted sH1N1 vaccines demonstrated measurable induction of IgG1 post boost over that of sH1N1 vaccine alone. However, the induction of IgG2c titers after two immunizations in aged mice were compromised even in adjuvanted groups, suggesting that the Th1-biased antibody response in aged mice,

aged CB6F1 mice. (A) Scheme of immunization procedure: CB6F1 mice were immunized i.m. either once or twice, 3 weeks apart, and antibody analysis was determined on sera collected 3 weeks following immunization. All mice were challenged with 100LD50 A/H1N1/California/4/2009 at day 54, and their physical condition was monitored over 14 days. Data for young mice were color-coded as blue; aged mice were color-coded as red for entire figure. (B) Sera from saline, sH1N1, sH1N1 + MF59-like, sH1N1 + SE, and sH1N1 + GLA-SE groups were analyzed for H1-specific IgG1 and IgG2c endpoint titers. Results are represented as the mean endpoint titer (log10) ± SEM. \*\* indicates *p-*value <0.01; \*\*\* indicates *p-*value <0.001; \*\*\*\* indicates *p-*value <0.0001. (C) Sera harvested from mice after a prime (day 20) or boost (day 42) immunization were analyzed for HAI titers. An HAI titer of five was assigned to responses below the assay detection limit. *p-*Values are denoted as follows: \* indicates <0.05; \*\* indicates <0.01; \*\*\* indicates < 0.001; \*\*\*\* indicates < 0.0001.

with the reduced dose of sH1N1 vaccine (shown to be effective in young mice), is not easily overcome by the action of adjuvants. Next, we tested whether adjuvanted sH1N1 vaccines could induce protective neutralizing antibodies in both young and aged mice. We observed that aged mice, which received MF59-like, SE, and GLA-SE adjuvanted vaccines could induce protective immune responses with two immunizations as compared to saline or sH1N1 vaccine alone, but not with one immunization (**Figure 1C**). These data suggest that adjuvants similar to MF59 like, SE, and GLA-SE could induce protective HAI titers in aged mice with a boost immunization; however, a boost with the TLR4 agonist adjuvant (GLA-SE) was unable to generate a Th1-biased environment capable of inducing IgG2c responses in aged mice with the sH1N1 vaccine.

# Enhanced Body Weights and Survival in Young and Aged CB6F1 Mice Following Two Immunizations with Adjuvanted sH1N1 Vaccines

CB6F1 mice were immunized once or twice with 0.01 µg of sH1N1 vaccine either alone or with MF59-like, SE, or GLA-SE adjuvants and then challenged with 100LD50 A/H1N1/California/4/2009 (**Figure 1A**). All mice that received a single immunization lost weight after viral challenge, yet, we found that young mice immunized with MF59-like, SE, and GLA-SE adjuvants could gain weight back by day 5 postinfection (**Figure 2A**, top left). Conversely, weight loss continued among aged CB6F1 mice until day 7, and only aged mice that received GLA-SE and SE adjuvants gained back weight over time (**Figure 2A**, top right). Both young and aged mice benefited from two immunizations with adjuvanted sH1N1 vaccines. Young mice given a boost with the adjuvanted vaccines quickly controlled the weight loss at day 3, and aged mice given adjuvanted vaccine only lost approximately 5% of starting weight compared to control animals (**Figure 2A**, bottom row). Aged mice given saline or sH1N1 vaccine alone lost approximately 20% of starting weight at day 6 (**Figure 2A**, bottom right). As shown in **Figure 2B**, survival in saline and sH1N1 young mice, and among all aged animals were severely compromised with one immunization. We observed that most of the young mice (with the exception of the saline group)

survived after receiving two immunizations (**Figure 2B**, bottom left). Survival among aged mice given sH1N1 combined with MF59-like, SE and GLA-SE adjuvants was significantly improved with two immunizations (**Figure 2B**, bottom right). Our data suggest that increasing the number of immunizations in aged mice from one to two immunizations could considerably improve clinical outcomes following the early stages of influenza infection. Similar to responses in young and aged mice immunized with adjuvanted sH1N1 vaccines, enhanced survival was observed in both young and aged C57BL/6 mice given two immunizations with adjuvanted rH1 vaccines (Figure S1D in Supplementary Material). This is particularly interesting based on the low HAI titers observed in aged mice following a boost immunization. The enhanced clinical outcome in aged mice with adjuvanted rH1 vaccine in the absence of HAI titers suggest that a compensatory immune response is contributing to the protective responses seen in aged mice.

# The GLA-SE-Based sH1N1 Vaccine Promotes the Generation of Cytokine-Producing T Helper Cells, Long-Lived Bone Marrow Plasma Cells (BMPCs) and GC B Cells in Young Mice

We next examined the cellular components that may contribute to protective immune responses in young and aged CB6F1 mice. Young and aged mice were immunized twice with adjuvanted sH1N1 vaccines using a slightly modified strategy, denoted in **Figure 3A**. The immunogenicity of T cell recall responses and GC B cell responses were determined 7 days following a boost immunization. Splenocytes from young and aged mice with two immunizations were stimulated with rH1 protein, and cytokine production from antigen-specific CD4<sup>+</sup> T cells were determined. We found that young mice that received both MF59-like or GLA-SE adjuvanted vaccines generated antigen-specific CD4<sup>+</sup> T cells that produce TNF-α, IL-2, and express CD154. However, only young mice given GLA-SE-adjuvanted vaccine induced IFNγ-producing CD4<sup>+</sup> T cells. From aged mice, we detected a minimal percentage of antigen-specific CD4+ T cells with increased nonspecific CD154 levels (representing activated T cells); however, some mice were able to produce IL-2 (**Figure 3B**). None of the aged mice induced TNF-α or IFN-γ-producing antigen-specific CD4<sup>+</sup> T cells following immunization. These data are consistent with our observation in **Figure 1A**, where aged mice failed to generate Th1-biased IgG2 antibody responses. Also, similar to previous observations (23), we found that aged mice have an increased percentage of non-specific follicular helper T cells (Tfh, CD4<sup>+</sup>CXCR5<sup>+</sup>PD-1<sup>+</sup>) as compared to young mice (**Figure 3C**). It has been shown that the quality and quantity of neutralizing antibodies is crucial for protective immune responses against viral infection (24). Thus, we examined the induction of GC B cells from spleen and inguinal LNs and long-lived BMPCs in both young and aged mice (**Figures 3D,E**). Interestingly, as shown in **Figure 3D**, within the spleen, young mice receiving two immunizations with sH1N1 alone had the highest number of GC B cells, whereas aged mice receiving sH1N1 combined with MF59-like adjuvant and GLA-SE had the highest number of GC B cells (although none of these responses reached statistical significance). The number of B220<sup>+</sup>CD95<sup>+</sup>GL7<sup>+</sup> GC B cells in the inguinal LN from young mice were enhanced after a boost immunization with all of the adjuvanted sH1N1 vaccines tested, although significant responses were observed only with sH1N1 + SE (**Figure 3D**). In aged mice, the MF59-like adjuvant induced the highest number of GC B cells in the LN (**Figure 3D**). We also found that young mice that received MF59-like or GLA-SE adjuvanted sH1N1 vaccines generated a significant number of antigen-specific BMPCs (**Figure 3E**). Aged mice that received MF59-like and GLA-SE adjuvanted vaccines showed a similar trend to that seen in young mice, although the responses were not statistically significant (**Figure 3E**).

### *In Vitro* Adjuvant Stimulation of BMDCs and Lung Cells from Aged Mice Exhibit Impaired Cytokine-Producing Ability Compared to Responses from Young Mice

Studies have shown that immune responses are reduced in aged animals due to decreased numbers of APCs and stromal cells (25), increased threshold required for TLR signaling (26), reduced B cell repertoire and humoral responses (7), and a phenomenon where immune organs eventually fill with fat or connective tissues (27). Following our observation that the induction of immune responses to sH1N1 vaccines were impaired in aged mice as compared to young mice, we next tested how adjuvants may affect early innate immune responses from both young and aged mice, characterized by *in vitro* cytokine production from BMDCs and lung homogenates. BMDCs, phenotypically and functionally similar to conventional DCs (28), are the surrogate to induction of systemic adaptive immune responses, whereas stimulated lung homogenates illustrate local immune responses. We tested BMDCs and lung homogenates from both young and aged CB6F1 mice in response to various doses of SE and GLA-SE and found that although BMDCs from aged mice are capable of producing innate inflammatory cytokines upon GLA-SE stimulation, aged cells produced significantly less cytokines compared to those from young animals (**Figure 4A**). Strikingly, we could scarcely detect inflammatory cytokines in the cell supernatants from GLA-SE-stimulated aged lung homogenates, especially for IL-12p40 and IFN-γ (**Figure 4B**). Our data suggest that although immunosenescence in aged mice affects innate immune responses both systemically (BMDCs) and locally (lung homogenates), the specific impairment of local immune responses in aged mice could lead to unfavorable and inefficient pathogen clearance at early stages following infection.

# GLA-SE Adjuvanted sH1N1 Vaccine Decreases Viral Load and Prevents Prolonged Lung Inflammation in Young CB6F1 Mice

Since we observed significant weight change and decrease in survival by day 5 after viral challenge in young and aged mice given either saline or sH1N1 vaccine alone (**Figure 2**), we hypothesized that mice immunized with adjuvanted vaccines would be

from aged mice were color-coded as red. \* indicates *p* value < 0.05; \*\* indicates *p* value < 0.01. Error bars indicate mean ± SEM. GC, germinal center; LN, lymph node; Tfh, follicular T helper cells; BMPC, bone marrow plasma cells. Statistical analysis was performed using one-way ANOVA and Tukey's multiple comparison test except 3E, which used one-way ANOVA and Bonferroni's test.

able to control local inflammation in the lung during the early infection phase, preventing subsequent mortality. In order to address this hypothesis, young CB6F1 mice were immunized once with saline, sH1N1, sH1N1 + SE, or sH1N1 + GLA-SE were then challenged with 100LD50 A/H1N1/California/4/2009. Following challenge, viral loads were determined at days 3 or 6 post challenge. Young CB6F1 mice that received GLA-SE adjuvant generated the highest functional neutralizing antibody titers following one immunization as compared to other groups (**Figure 5A**). Consistent with our previous data (**Figure 1C**), mice that received saline or a prime immunization with the sH1N1 vaccine did not generate significant neutralizing antibodies (**Figure 5A**); in addition, we found that the viral loads were 1,000-fold higher in saline and sH1N1 groups compared to responses in the sH1N1 + GLA-SE group at days 3 and 6 following influenza infection (**Figure 5B**). Interestingly, the sH1N1 + SE group had 100-fold higher viral load than the sH1N1 + GLA-SE group at day 3, but no differences in virus titer were observed at day 6, suggesting that sH1N1 adjuvanted with GLA-SE is superior to sH1N1 adjuvanted with SE, and animals immunized with sH1N1 + GLA-SE were able to clear viral infection earlier after influenza challenge (**Figure 5B**).

Next, we examined the inflammatory cytokines and chemokines in bronchoalveolar lavage fluid (BALF) from young mice at day 3 or 6 after viral challenge. To our surprise, we could not detect any inflammatory cytokines in BALF from mice that received either SE or GLA-SE adjuvants (**Figure 5C**). IL-6, TNF-α, IFN-γ, and IL-12p40 were detected in the BALF from saline and sH1N1 groups at day 3 and 6, indicating that a continuous inflammatory response occurs in the lungs of these mice. Notably, Th2-biased cytokine IL-4, IL-5, and chemokine Eotaxin (data not shown) in the BALF were detected at day 6 from mice that received the sH1N1 vaccine alone. These data corroborate our previous reports that immunization with protein only, or without proper Th1-driving adjuvants, leads to Th2-biased immune responses (29). It has also been reported that alveolar macrophages (AMs) are crucial for lung homeostasis and are the first line cells capable of detecting lung pathogens including bacteria, viruses, in addition to allergens and foreign particles (30). Thus, we examined whether AMs from young mice given a prime immunization were affected after exposure to A/ H1N1/California/4/2009 challenge. Interestingly, we found that unvaccinated, young CB6F1 mice (saline group) had dramatically decreased percentages of AMs at day 3 of infection as compared

\*\* indicates < 0.01; \*\*\* indicates < 0.001; \*\*\*\* indicates < 0.0001. BMDC, bone marrow dendritic cells.

to the other groups. At day 6, both saline and sH1N1 groups had a significantly decreased percentage of AMs as compared to adjuvanted groups, suggesting that AMs are potentially being eliminated by the viral infection (31) (**Figure 5D**). Similarly, a reduction in the percentage of AMs following influenza infection correlate with an increased viral load within the lung. As shown in Figure S2 in Supplementary Material, we found that in young mice given one immunization, only mice that received the sH1N1 + GLA-SE vaccine maintain AM homeostasis and prevent inflammatory cell infiltration (including eosinophils, neutrophils, inflammatory Mϕs, and NK cells) in the lung at both day 3 and 6 post-infection. After influenza infection, cells such as AMs that express TLR7 are able to detect single-strand RNA fragments released from viral particles. We found that expression of TLR7 on AMs were relatively consistent among all groups at day 3; however, the TLR7 levels were significantly decreased in

Figure 5 | GLA-SE adjuvanted sH1N1 vaccine decreases viral load and prevents prolonged lung inflammation in young CB6F1 mice. Young CB6F1 mice (age of 6–8 weeks old) were immunized once with saline, sH1N1, sH1N1 + SE, or sH1N1 + GLA-SE, and sera were harvested 3 weeks after prime. All mice were challenged with 100LD50 A/H1N1/California/4/2009 at day 28 and BALF and lung homogenates were harvested at day 3 and day 6 post-infection. (A) Hemagglutination inhibition (HAI) titer against A/H1N1/California/4/2009 was determined. (B) The relative expression of H1 was determined by real time Q-PCR. The expression of H1 of each sample was normalized to GAPDH (graph on log10 scale). (C) BALF from all infected mice were harvested at day 3 and day 6 postinfection, and the levels of IL-6, IFN-γ TNF-α, IL-12p70, IL-4, and IL-5 cytokines were determined by Luminex. (D) The percentage of AMs was determined by flow cytometry and were normalized to uninfected, naïve mice (as 100%). AMs in the lung homogenates were defined as CD11b−CD11c+Siglec-F+NK1.1− population. (E) The intracellular expression of TLR7 by AMs was determined by flow cytometry and was normalized to uninfected, naïve mice (as 100%). *p* Values are denoted as follows: \* indicates <0.05; \*\* indicates <0.01; \*\*\* indicates <0.001; \*\*\*\* indicates <0.0001. Results are represented as the mean ± SEM. BALF, bronchoalveolar lavage fluid; AM, alveolar macrophages.

saline, sH1N1 and sH1N1 + SE groups by day 6 as compared to sH1N1 + GLA-SE group (**Figure 5E**). Overall, our data suggest that the GLA-SE adjuvanted sH1N1 vaccine provides superior protection in young mice by clearing influenza virus early after infection, and through maintenance of AMs and TLR7 expression levels in the early infection phase.

### DISCUSSION

In this study, the influence of different adjuvants on the efficacy of two types of seasonal influenza vaccine: a rH1 protein vaccine, and a sH1N1 vaccine was evaluated. Adjuvants examined include two squalene-based oil-in-water emulsions [a MF59-like adjuvant and SE (a stable emulsion adjuvant)] and a synthetic TLR4 agonist formulated with SE (GLA-SE). In CB6F1 mice, we showed that both the MF59-like adjuvant and GLA-SE plus sH1N1 vaccine contribute to Th1-biased vaccine-specific IgG2 immune responses following a boost immunization, whereas significantly higher IgG2c titers were not induced in aged mice. HAI titers were, however, induced in both young and aged CB6F1 mice after two immunizations. In contrast, in C57BL/6 mice, we observed an increase in vaccine-specific IgG1 and IgG2c responses following a boost immunization with all of the adjuvanted rH1 vaccines tested. However, whereas HAI titers were above the 1:40 threshold with adjuvanted rH1 vaccines in young C57BL/6 mice [HAI titers above a titer of ≥1:40 in humans is associated with a 50% reduction in seasonal influenza infection and is considered a correlate of protection (32, 33)], HAI titers were not increased in aged mice. Following a boost immunization, aged C57BL/6 and CB6F1 mice induced significant levels of either vaccine-specific IgG2c or HAI titers, respectively, leading to protection against influenza challenge. This observation is of interest as cellular immunity is known to decline with age and compensatory mechanisms induced by adjuvants may aid protection in the elderly. Recently, the induction of IgG2 antibodies have been a focus of interest for protection against infection, including influenza infection. An M2e-specific IgG2c monoclonal antibody was reported to protect mice against influenza infection more effectively than an IgG1 monoclonal due to Fc activation properties (34).

The use of adjuvants as a strategy for enhancing the quality and magnitude of protective antibody responses to influenza vaccines is not a new concept. We have shown previously that the addition of adjuvants such as GLA-SE are able to increase dose-sparing properties and broaden HAI titers against drifted influenza viruses (12). In this manner, applying the GLA-SE adjuvant to influenza vaccine design addresses a noted and significant complication of reduced vaccine-specific antibody responses observed in aged individuals (9). GLA-SE combined with a recombinant H5N1 avian influenza vaccine (rH5) also increases the breadth of the antibody responses and provides protection against a heterosubtypic influenza virus, compared to the vaccine combined with SE (14).

One of the benefits of using GLA-SE as a vaccine adjuvant is the Th1-inducing innate responses that promote protective adaptive Th1-mediated cellular and humoral immunity (12, 29, 35). Characterization of this pure synthetic TLR4 agonist adjuvant in comparison to another TLR4 agonist, monophosphoryl lipid A (MPL), which is derived naturally from the bacterium *Salmonella minnesota*, has been shown following stimulation of DCs from both humans and mice (36). Both agonists were shown to increase DC activation and maturation and induce proinflammatory cytokines and chemokines (12). In elderly humans, the responses of monocyte-derived DCs stimulated with GLA-SE appear intact (16). Here, we add to this paradigm and demonstrate that BMDCs derived from aged mice respond to GLA-SE stimulation; however, cytokine responses are lower than observed following GLA-SE stimulation of young BMDCs.

Peripheral blood mononuclear cells from elderly patients stimulated with GLA-SE and a split virus vaccine have also been shown to increase the ratio of IFN-γ:IL-10, in addition to granzyme B, further supporting this as a potentially effective candidate adjuvant for use in the elderly (19). In our study, MF59-like adjuvant was also capable of promoting vaccine-specific CD4<sup>+</sup> T cell responses, GC B cells, and induction of long-lived BMPCs, and like the SE adjuvant, was able to induce HAI titers after a boost immunization leading to protection in mice against challenge with H1N1. The history, safety, and efficacy of the MF59 adjuvant has been recently reviewed (17, 18).

One of the most striking observations in our study was the differential cytokine responses to adjuvant in aged vs. young lung homogenates. The *in vitro* stimulatory response of GLA-SE on lung homogenates from young mice led to the induction of TNF-α, IL-12p40, IFN-γ, and IL-10, whereas only TNF-α and IL-10 were induced in samples from aged mice. This potential local impairment (within the lung) could have substantial consequences following acute infection with influenza virus and other pulmonary infections.

In order to determine the protective mechanism of action of sH1N1 adjuvanted with GLA-SE within the lungs of young mice, we further characterized the pulmonary responses following H1N1 infection, including assessment of viral load and inflammatory responses. We observed less virus in the lung at day 3 post challenge, in mice immunized once with sH1N1 combined with GLA-SE, and decreased levels of virus at day 6 post challenge in mice immunized with SE compared to saline or sH1N1 vaccine only. The magnitude of the neutralizing HAI titer was also highest in mice given the sH1N1 + GLA-SE. Furthermore, a drastic reduction in the inflammatory responses within the lungs of mice immunized with the sH1N1 + GLA-SE was observed compared to the Th2-biased inflammatory responses that were seen in the sH1N1 vaccine alone group. At day 6 of infection, the percentage of AMs was also significantly higher in the sH1N1 + GLA-SE group compared to both saline and sH1N1 immunized mice. Interestingly, TLR7 expression within AMs was also significantly higher in the sH1N1 + GLA-SE-treated mice compared to mice that were immunized with sH1N1 vaccine only and sH1N1 + SE. Recently, Wong et al. reported that AMs from aged mice have an intrinsically impaired function to limit lung damage during influenza viral lung infection and suggested that enhancing the function of AMs may improve outcomes in elderly individuals infected with respiratory viruses (31). It will be of great interest to determine if administration of Th1-inducing vaccines in aged animals can preserve the numbers and functions of AMs during acute phases of infection, and how this might reduce the morbidity and mortality in aged individuals. This significant hypothesis warrants further exploration. We have previously shown, in both mice and non-human primates, that Fluzone® (a trivalent inactivated influenza vaccine) adjuvanted with GLA-SE, increases the magnitude of HAI titers, induces Th1 cellular immune responses, and enhances cross-reactive antibody responses to drifted influenza strains (12). Our current work further addresses additional attributes of GLA-SE-adjuvanted influenza vaccines in the context of increased age.

Finally, the CDC estimates that during the 2015–2016 influenza season, the elderly (making up only 15% of the overall US population) accounted for half or more of hospitalizations associated with influenza and 64% of deaths associated with pneumonia and influenza (37). Our data suggest that the staggering burden of morbidity and mortality due to influenza infection in the elderly population may be alleviated by further detailed and mechanistic selection of proper adjuvants to be included in seasonal vaccines.

## ETHICS STATEMENT

These studies were carried out in accordance with the regulations and guidelines of the IDRI animal care and use committee (IACUC). The animal protocol was approved by the IACUC committee.

# REFERENCES


# AUTHOR CONTRIBUTIONS

SB and FCH wrote and prepared the manuscript; SB, FCH, EG, SL, and RC edited the manuscript; SB, NH, EG, RC, and FCH designed the experiments; NH, EG, and BG did HAI and viral infection; FCH and JG did Luminex assays; FCH did the quantitative PCR and flow analysis; EL and LH did *in vitro* stimulation of BMDCs and lung homogenates; and SB, FCH, and SL did the data analyses.

# ACKNOWLEDGMENTS

We thank Tara Evers, Sharvari Waghmare Joshi, Ghislain Ismail Nana, Tom Hudson, Elyse Beebe, Po-Wei Huang, and Valerie Reese for their technical assistance. We also thank Dr. Chris Fox and the formulation team at IDRI for the adjuvant formulations. In addition, we thank Dr. Winston Wicomb and the vivarium staff for their excellent care of the animals used in these studies. The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007–2011) under grant agreement 280873.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00295/ full#supplementary-material.


affects stability, antigen structure, and vaccine efficacy. *Influenza Other Respir Viruses* (2013) 7(5):815–26. doi:10.1111/irv.12031


**Conflict of Interest Statement:** SR is a founder of, and holds an equity interest in, Immune Design Corporation, a licensee of certain rights associated with GLA. All other authors have no financial conflicts of interest.

*Copyright © 2018 Baldwin, Hsu, Van Hoeven, Gage, Granger, Guderian, Larsen, Lorenzo, Haynes, Reed and Coler. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Efficacy Testing of H56 cDNA Tattoo Immunization against Tuberculosis in a Mouse Model

*Anouk C. M. Platteel1,2†, Natalie E. Nieuwenhuizen2†, Teresa Domaszewska2 , Stefanie Schürer2 , Ulrike Zedler2 , Volker Brinkmann3 , Alice J. A. M. Sijts1‡ and Stefan H. E. Kaufmann2 \*‡*

*1Department of Infectious Diseases and Immunology, Faculty of Veterinary Medicine, Utrecht University, Utrecht, Netherlands, 2Department of Immunology, Max Planck Institute for Infection Biology, Berlin, Germany, 3Microscopy Core Facility, Max Planck Institute for Infection Biology, Berlin, Germany*

Tuberculosis (TB), caused by *Mycobacterium tuberculosis* (*Mtb*), remains a global threat. The only approved vaccine against TB, *Mycobacterium bovis* bacillus Calmette–Guérin (BCG), provides insufficient protection and, being a live vaccine, can cause disseminated disease in immunocompromised individuals. Previously, we found that intradermal cDNA tattoo immunization with cDNA of tetanus toxoid fragment C domain 1 fused to cDNA of the fusion protein H56, comprising the *Mtb* antigens Ag85B, ESAT-6, and Rv2660c, induced antigen-specific CD8+ T cell responses *in vivo*. As cDNA tattoo immunization would be safer than a live vaccine in immunocompromised patients, we tested the protective efficacy of intradermal tattoo immunization against TB with H56 cDNA, as well as with H56\_E, a construct optimized for epitope processing in a mouse model. As *Mtb* antigens can be used in combination with BCG to boost immune responses, we also tested the protective efficacy of heterologous prime-boost, using dermal tattoo immunization with H56\_E cDNA to boost BCG immunization in mice. Dermal H56 and H56\_E cDNA immunization induced H56-specific CD4+ and CD8+ T cell responses and Ag85Bspecific IgG antibodies, but did not reduce bacterial loads, although immunization with H56\_E ameliorated lung pathology. Both subcutaneous and intradermal immunization with BCG resulted in broad cellular immune responses, with increased frequencies of CD4+ T effector memory cells, T follicular helper cells, and germinal center B cells, and resulted in reduced bacterial loads and lung pathology. Heterologous vaccination with BCG/H56\_E cDNA induced increased H56-specific CD4+ and CD8+ T cell cytokine responses compared to vaccination with BCG alone, and lung pathology was significantly decreased in BCG/H56\_E cDNA immunized mice compared to unvaccinated controls. However, bacterial loads were not decreased after heterologous vaccination compared to BCG alone. CD4+ T cells responding to Ag85B- and ESAT-6-derived epitopes were predominantly IFN-γ+TNF-α+ and TNF-α+IL-2+, respectively. In conclusion, despite inducing appreciable immune responses to Ag85B and ESAT-6, intradermal H56 cDNA tattoo immunization did not substantially enhance the protective effect of BCG under the conditions tested.

### Keywords: tuberculosis, vaccine, vaccination, DNA immunization, *Mycobacterium bovis* bacillus Calmette– Guérin, H56

### *Edited by:*

*Rino Rappuoli, GlaxoSmithKline, Italy*

### *Reviewed by:*

*Luciana Leite, Instituto Butantan, Brazil Juraj Ivanyi, King's College London, United Kingdom*

### *\*Correspondence:*

*Stefan H. E. Kaufmann kaufmann@mpiib-berlin.mpg.de*

*† These authors have contributed equally to this work.*

> *‡ Equal senior co-authors.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 12 September 2017 Accepted: 23 November 2017 Published: 11 December 2017*

### *Citation:*

*Platteel ACM, Nieuwenhuizen NE, Domaszewska T, Schürer S, Zedler U, Brinkmann V, Sijts AJAM and Kaufmann SHE (2017) Efficacy Testing of H56 cDNA Tattoo Immunization against Tuberculosis in a Mouse Model. Front. Immunol. 8:1744. doi: 10.3389/fimmu.2017.01744*

**Abbreviations:** BCG, *Mycobacterium bovis* bacillus Calmette–Guérin; GC, germinal center; *i.d.*, intradermal; *Mtb, Mycobacterium tuberculosis*; *s.c.*, subcutaneous; TB, tuberculosis; TCM, central memory T cells; TEM, effector memory T cells; TFH, follicular helper T cells; TTFC, tetanus toxin fragment C domain 1.

# INTRODUCTION

Tuberculosis (TB) remains a global health threat, with 10.4 million cases and 1.7 million deaths reported for 2016 (1, 2). It is estimated that approximately a quarter of the world's population has latent TB infection (LTBI) (3). Socioeconomic factors such as poor living conditions, stress and malnutrition, play a major role in susceptibility to developing TB disease, with the HIV pandemic also a major driver (4). Coinfection with *Mycobacterium tuberculosis* (*Mtb*) and HIV leads to accelerated deterioration of immunity, and TB is the most common cause of death in HIV<sup>+</sup> individuals. HIV contributes to the increased risk of TB by depleting CD4<sup>+</sup> T cells, affecting macrophage effector functions, tipping the Th1/Th1 balance and influencing granuloma formation (5). The risk of those with LTBI developing active disease is approximately 10% over a lifetime, but rises to 5–10% per year in those with HIV infection (6). HIV-exposed uninfected infants are another group at high risk of TB infection, as they have poorer T cell generation and IFN-γ production compared to HIV-unexposed infants (7).

An attenuated form of the causative agent of bovine TB, *Mycobacterium bovis* bacillus Calmette–Guérin (BCG), was introduced as a live vaccine in 1921 and is today the most used vaccine globally (8). BCG provides protection against TB meningitis and other forms of disseminated TB. In addition, it has contributed to a reduction in general childhood mortality by boosting non-specific immunity against common causes of childhood disease (9). However, BCG fails to protect adequately against the pulmonary form of TB (10). Protection also varies geographically; in the UK and Norway, BCG confers 50–60% protection that lasts up to 20 years (11), while the high incidence of TB in developing countries illustrates the need for an improved vaccination strategy. Furthermore, BCG can cause severe adverse effects in immunocompromised individuals, and it is not recommended for HIV-infected individuals, who are thus a target group for vaccination due to their increased risk of developing TB (12). As BCG is given at birth throughout the developing world, a new TB vaccine for children should take into account the fact that the majority of the population has already been BCG-vaccinated. New vaccination strategies aim to improve the efficacy and/or safety of BCG in several ways, including modifying it, combining it with booster vaccines, administering it with different adjuvants or altering the route of vaccination (13–19).

A crucial starting point in improving vaccine strategies is knowledge of immune responses that correlate with protection against TB. Although the precise mechanisms of *Mtb* control are unclear, T cell responses are known to be crucial (19). Successful attempts have been made to enhance the cellular response by modifying BCG; for example, a recombinant BCG strain was generated which expresses the *hly* gene encoding for listeriolysin O (LLO) from *Listeria monocytogenes,* in combination with deletion of urease C, resulting in a pH optimal for LLO activity (20–22). Vaccination of mice with BCG *ΔureC::hly* induced more CD4<sup>+</sup> central memory T cells (TCM) than BCG vaccination, which were protective against infection (21). In other studies, Ag85B-specific CD4<sup>+</sup> effector memory T cells (TEM) were shown to control infection in the lungs (23), and CD8<sup>+</sup> T cells also protect against *Mtb,* particularly during LTBI (24–28).

As most individuals in TB-endemic countries are vaccinated with BCG, another strategy for increasing protection against TB is to boost BCG prime immunization with subunit vaccines containing *Mtb* antigens, using non-live vaccines based on recombinant fusion proteins mixed with adjuvants or non-replicating attenuated viral vectors (13, 14, 16, 18, 19). Boosting with a second dose of BCG itself is currently not recommended by the World Health Organization, as it has not been found to improve protection significantly in earlier studies (29–32). Different *Mtb* antigens have been tested against TB, including Ag85A, Ag85B, TB10.4, HBHA, and others (19). Aagaard et al. (33) created the multistage fusion protein vaccine H56, which is composed of three antigens of *Mtb*: Ag85B, ESAT-6, and Rv2660c. Ag85B and ESAT-6 are both immunogenic proteins secreted early in infection (34, 35), with Ag85B found in both *Mtb* and BCG and ESAT-6 only in *Mtb*. Rv2660c was originally identified as a latency-associated gene of *Mtb* (36). Although it has not yet been detected at the mRNA or protein level (37), recombinant Rv2660c stimulated IFNγ responses in CD4+ T cells from individuals with LTBI and was also recognized by IgG antibodies (38) and increased protection was found after vaccination of mice with H56 compared to H1, a similar fusion protein that lacks only the Rv2660c gene (35). The Rv2660c (Mb2678c) gene was shown to be expressed at the transcript level by BCG *in vitro* (39). Vaccination with H56 or the antigens of H56 administered with adjuvant CAF01 (40, 41) or IC31 (41, 42) protected against *Mtb* infection in mice. Recently, we showed that intradermal (*i.d*.) tattoo administration of H56 cDNA fused to tetanus toxin fragment C domain 1 (TTFC) cDNA elicited vigorous antigenspecific CD4<sup>+</sup> T cell and CD8<sup>+</sup> T cell responses targeted to H56, without the need for adjuvants (43). The immunogenicity of H56 cDNA was enhanced by fusion to TTFC cDNA, and by substitution of the C-terminal epitope flanking residues, to optimize proteasome-mediated epitope generation. Unlike immunization with live bacterial strains such as BCG, cDNA immunization does not carry a risk of infection and it should, thus, be safer for immunocompromised individuals.

In the present study, we aimed to compare the efficacy of *i.d.* DNA tattoo immunization to that of subcutaneous (*s.c.*) BCG immunization *in vivo*, and to test *i.d*. H56 cDNA tattoo immunization as a booster for *s.c.* injection of BCG*.* We also tested *i.d.* immunization with BCG. Immunization with *i.d.* DNA tattoo alone did not protect against TB. Heterologous vaccination with BCG/DNA improved H56-specific T cell responses, increased antibody titers, and ameliorated lung pathology in our murine TB model. However, bacterial burdens were not reduced when compared to vaccination with BCG alone.

### MATERIALS AND METHODS

### Mice

Six- to eight-week old CB6F1 mice (male C57BL/6 × female BALB/c F1) were purchased from Charles River. All animal studies were ethically approved by the State Office for Health and Social Services, Berlin, Germany. We designed two protocols, one for single vaccination and one for prime-boost vaccination, each consisting of 10 mice per timepoint divided into two groups of five mice for manageability and performed in two separate experiments.

## DNA and Dermal DNA Tattoo Vaccination

The full-length H56 and H56\_E cDNA (33) was codon optimized, and inserted at the 3′end of TTFC cDNA (44, 45) into the pVAX1 vector (Invitrogen) as described previously (43). DNA tattoo immunization was performed with 15 µl cDNA (2 µg/µl) in Tris-EDTA buffer with a 9-needle bar mounted on a tattoo rotary device (Cheyenne) adjusted to 100 Hz at 1 mm depth for 1 min (46), and the mice were under isoflurane anesthesia.

### BCG Vaccination and *Mtb* Challenge

The *Mtb* H37Rv (American Type Culture Collection; catalog no. 27294) and BCG Danish 1331 (BCG SSI) (American Type Culture Collection; catalog no. 35733) were grown in Middlebrook 7H9 broth (BD) supplemented with albumin-dextrose-catalase enrichment (BD), 0.2% glycerol, and 0.05% Tween 80 or on Middlebrook 7H11 agar (BD) containing 10% (vol/vol) oleic acid-albumin-dextrose-catalase enrichment (BD) and 0.2% glycerol. BCG was grown to mid-log phase, washed with phosphatebuffered saline (PBS) and stored at −80°C in PBS/10% glycerol. BCG was washed in PBS before vaccination and administered at a dose of 106 CFU in 100 µl for *s.c.* immunization and 106 CFU in 25 µl for the *i.d.* tattoo. Aerosol challenge with *Mtb* was performed using an inoculum of 20–50 CFU.

# Single Immunization Model

The *s.c*. injection of BCG is known to lead to an approximately 1 log reduction in bacterial burden in mouse models (47, 48) and was included as the standard against which the efficacy of *i.d.* tattoo immunization with BCG or DNA were compared. The DNA vaccines included H56, a DNA vector containing the full-length codon-optimized H56 gene fused to TTFC cDNA, and H56\_E, in which six CD8<sup>+</sup> T cell epitope flanking sequences have been optimized to enhance proteasome-mediated processing (43), and were administered at day 0, 3, and 6. BCG was administered *s.c*. at day 0. One group of mice was also immunized by *i.d.* tattoo with BCG (day 0 only), in order to compare *s.c.* and *i.d.* administration of BCG. Immune responses were measured at day 21 and the efficacy of the methods was tested by subjecting the mice to an aerosol challenge with *Mtb* at day 60.

### Prime-Boost Model

Two groups of mice were *s.c.* immunized with BCG at day 0, and one of these was boosted by *i.d.* H56\_E DNA tattoo at day 40 for a heterologous immunization. Another group was primed and boosted *i.d.* with H56\_E cDNA tattoo. Unvaccinated mice were included as controls. Immune responses were measured 2 weeks after the booster vaccination. Mice were infected with *Mtb* at day 100 by aerosol challenge and bacterial burdens and lung pathology were assessed at day 190.

## Vaccine-Efficacy Studies

For measurements of the bacterial load, lungs and spleens were harvested and serial dilutions were performed in PBS-0.05% Tween 80 and plated on Middlebrook 7H11 agar. The percentage of inflammation per lung area was measured in a blinded manner on formaldehyde fixed tissue sections stained in Giemsa. Whole lung sections were scanned with a ZEISS Axioscan Z1 driven by ZEN and lung pathology analysis was performed using Volocity (Perkin Elmer). Lung images shown were chosen based on the image analysis, with the lung having the value closest to the average result per group shown here.

## Peptides

All epitopes, as determined previously (43), were synthesized using Fmoc solid phase chemistry. The sequence enumeration of the synthetic peptides referred to the vaccine H56 (33). CD4<sup>+</sup> T cell epitopes included H56242–262 (Ag85B derived) and H56288–307 (ESAT-6 derived). The CD8<sup>+</sup> T cell epitopes included H56354–363 (ESAT-6 derived) and five epitopes of Ag85B.

The eighty-five 15mer peptides used as pool to measure the total amount of H56-specific T cells (provided as a kind gift from Dr. Donatella Negri) spanned the entire amino acid sequence of H56, overlapping by 10 amino acid residues, and were synthesized by PRIMM Srl.

### Analysis of Specific CD8**+** and CD4**+** T Cell Responses Using Intracellular Cytokine Staining

For intracellular cytokine production, splenocytes were plated overnight with or without 1 µg/ml peptide or peptide pool containing 1 µg/ml of each peptide at 37°C. During the last 4 h, brefeldin (5 µg/ml, Sigma) was added to wells incubated with peptide or phorbol myristate acetate (50 ng/ml, Sigma) and ionomycin (250 ng/ml, Sigma) or to cells incubated with medium alone. Cells were then stained for the cell surface markers named below, fixed with 2% paraformaldehyde, permeabilized with saponin buffer (saponin, 1 g/l; CaCl2, 0.11 g/l; MgSO4, 0.125 g/l; NaN3, 0.5 g/l; bovine serum albumin (BSA) 1 g/l; 10 mM HEPES in PBS, pH 7.4), and stained for intracellular cytokines as described (43). The cell surface- and intracellular cytokine panel consisted of anti-TCRβ-A700 (Biolegend; clone H57-597), anti-CD4 Pacific Blue (Biolegend; clone GK1.5), anti-CD8 PerCP (Biolegend; clone 53-6.7), anti-IL-2 APC (Biolegend; clone JES6-5H4), anti-IFNγ PE-Cy7 (Biolegend; clone XMG1.2), anti-IL-17 PE (Biologend; clone TC11-18H10.1), and anti-TNF-α FITC (clone MP6-XT-22, grown in-house). Samples were acquired on an LSR II cytometer (BD Biosciences) with BD FACS Diva software and analyzed using FlowJo v10 (TreeStar).

# Analysis of Specific CD8**+** T Cell Responses by IFN-**γ** ELISpot

Multiscreen ELISPOT plates (Millipore) were coated with 2 µg/ ml anti-mouse IFN-γ (clone AN18, grown in-house) in PBS overnight at 4°C. Wells were washed and blocked with 5% BSA/ PBS. 5 × 105 erythrocyte depleted lymph node cells were plated with or without 1 µg/ml synthetic peptide overnight in Iscove's modified Dulbecco's medium (IMDM) with 10% fetal calf serum (FCS) Pen-Strep at 37°C. Plates were washed with PBS plus 0.01% Tween 20 (PBS-T), and IFN-γ was detected with biotinylated anti-IFN-γ (clone XMG1.2), followed by alkaline–phosphatase (AP)-conjugated streptavidin (homemade) in PBS-T supplemented with 2% BSA. The assay was developed with NBT/ BCIP substrate (Thermo Fisher Scientific) and analyzed using an Immunopost S6 Ultra-V Analyzer (Cellular Technology Limited).

### FIGURE 1 | Continued

Immunization with H56 or H56\_E cDNA induces Ag85B- and ESAT-6-specific T cell responses. Mice were immunized *s.c.* with bacillus Calmette–Guérin (BCG) and *i.d*. with BCG, H56, a DNA vector containing the H56 gene fused to TTFC or H56\_E, in which six CD8+ T cell epitope flanking residues in the H56 sequence were optimized to enhance proteasome-mediated processing. Intracellular cytokine staining and ELISpots were performed on splenocytes and inguinal lymph node cells, respectively, harvested at day 21 post vaccination and re-stimulated with different H56 peptides. The peptide pools consisted of eighty-five 15-mer peptides spanning the entire sequence of H56, as a measure of the total percentage of H56-antigen-specific cells. CD4+ T cell epitopes included H56242–262 (Ag85B derived) and H56288–307 (ESAT-6 derived). The CD8+ T cell epitopes included H56354–363 (ESAT-6 derived) and five epitopes of Ag85B. The results show pooled data from two independent experiments (*n* = 10 in total per group). For the unvaccinated group, cells from five mice were pooled into two samples in order to have enough cells. (A) Immunization scheme. (B) Gating strategy for FACS analysis. After gating on live, single cells, cells were gated on CD4+ (R1) or CD8+ (R2). Both R1 and R2 were gated on TCR-β+ cells (R3) and (R4) before gating single IFN-γ, IL-2, IL-17, and TNF-α cells. (C) Heat map showing mean frequency of spleen-derived CD4+ T cells secreting IFN-γ, IL-17, IL-2, or TNF-α after peptide stimulation, normalized to medium incubated cells. (D) Heat map showing mean frequency of spleen-derived CD8+ T cells secreting IFN-γ, IL-17, IL-2, or TNF-α after peptide stimulation, normalized to medium incubated cells. (E) Heat map showing mean number of IFN-γ producing cells per million lymph node cells after stimulation with peptide pools, CD4 epitopes or CD8 epitopes, normalized to medium incubated cells. The statistical differences of values shown in the heatmaps are shown in Tables S1 and S2 in Supplementary Material.

## Analysis of T and B Cell Subpopulations

Spleens and inguinal lymph nodes were collected and singlecell suspensions were generated in IMDM 10% FCS Pen-Strep. Cells were surface stained to quantify cell populations. T cell panel: anti-CD3 Alexa 700 (BioLegend; clone 17A2), anti-CD4 PE-Cy7 (BioLegend; clone RM4-5), anti-CD8 V500 (BD Horizon, clone 53-6.7), anti-CD62L APC (BD Pharmingen; clone MEL-14), anti-CD44 Pacific Blue (clone IM7, grown in-house), anti-CXCR5 PE (BioLegend; clone J252D4), and anti-CCR7 PerCP (BioLegend; clone 4B12). B cell panel: anti-B220 PE (BioLegend; clone RA3-6B2), anti-CD38 APC (BioLegend; clone 90), anti-Fas PE-Cy7 (BioLegend; clone Jo2), anti-GL7 FITC (BioLegend; clone GL7), anti-MHCII Pacific Blue (BioLegend; clone M5/114.15.2), anti-CD19 Alexa 700 (BioLegend; clone 6D5), and anti-CD138 Percp (BioLegend; clone 281-2). CD4<sup>+</sup> TCM were CD3<sup>+</sup>CD4<sup>+</sup>CD44highCD62high, CD4<sup>+</sup> TEM were CD3<sup>+</sup>CD4<sup>+</sup>CD44highCD62Llow, T follicular helper cells (TFH) were CD3<sup>+</sup>CD4<sup>+</sup>CXCR5<sup>+</sup>, germinal center (GC) B cells were B220<sup>+</sup>CD19<sup>+</sup>GL7<sup>+</sup>Fas<sup>+</sup>, and plasma cells B220<sup>+</sup>CD 19<sup>+</sup>CD138<sup>+</sup>CD38<sup>+</sup>.

### Specific Antibody Responses

Mycobacteria-, Ag85B- and ESAT-6-specific antibodies in serum were measured by indirect enzyme-linked immunosorbent assay using *Mtb* H37Rv lysate (BEI Resources), recombinant Ag85B (BEI Resources), or recombinant ESAT-6 (BEI Resources) to coat and AP-labeled anti-mouse IgG, IgG1, and IgG2c for detection (SouthernBiotech).

### Heat Maps

The scripts for **Figures 1** and **4** are available upon request. FACS data visualization was designed using R-package ggplot2 (v2.1.0; Springer-Verlag New York, NY, USA, 2009.) in R programming language [v3.2.3; R Development Core Team (2008)]. The mean values of percentages of CD4<sup>+</sup> or CD8<sup>+</sup> T cells producing cytokines after peptide stimulation were normalized to control samples stimulated with medium and visualized using heat maps. ELISpot data visualization was designed using R-package ggplot2 (v2.1.0; Springer-Verlag New York, NY, USA, 2009) in R programming language [v3.2.3; R Development Core Team (2008)]. The mean numbers of IFN-γ-producing cells per million cells after peptide stimulation were normalized to control samples stimulated with medium. For visualization purposes, the mean value for each experimental condition was transformed to square root of mean response signal and displayed as a heat map.

### Statistics

Data were tested for normality with Levene tests. Kruskal–Wallis with Dunn's multiple comparison test was used to analyze differences in bacterial burdens between groups. Differences in immunological parameters between groups were analyzed using a one-way ANOVA with Tukey's multiple comparisons test. *p* < 0.05 was considered significant. The differences in cytokine frequencies were assessed by creating linear regression models with treatment as the predictor and measured cytokine frequency as the dependent variable. The models were created for CD4<sup>+</sup> cells, CD8<sup>+</sup> cells, all peptide stimulations, and measured cytokines separately. Benjamini–Hochberg method was used to correct for multiple testing. *p*-values <0.05 were considered significant.

# RESULTS

## Immunization with H56 or H56\_E cDNA Induces H56-Specific T Cell Responses

Mice were immunized with BCG (*s.c*. or *i.d*.), with H56, a DNA vector containing the full-length codon-optimized H56 gene fused to TTFC cDNA, or with H56\_E, in which six CD8<sup>+</sup> T cell epitope flanking residues in the H56 sequence were optimized to enhance proteasome-mediated processing (43) (**Figure 1A**). Previously, we identified seven novel CD8<sup>+</sup> T cell epitopes in the H56 fusion protein (43). In order to compare Ag85B- and ESAT-6-specific T cell responses induced by the cDNA vaccines and BCG, we measured peptide-specific immune responses after the different immunization strategies, including responses to an H56 peptide pool. Splenocytes or lymph node cells from vaccinated mice were re-stimulated *ex vivo* with Ag85B- and ESAT-6-derived CD8<sup>+</sup> (43) and CD4<sup>+</sup> (49, 50) T cell epitopes and responses in the spleen were measured using intracellular cytokine staining and flow cytometry (**Figures 1B–D**) or in the draining lymph nodes by IFN-γ ELISpot (**Figure 1E**). The mean percentages of cytokine-producing CD4+ or CD8+ T cells after peptide stimulation were normalized to control samples stimulated with medium and visualized using heat maps (**Figures 1C,D**). Results of statistical analysis are shown in Table S1 in Supplementary Material. Incubation of splenocytes with a pool of overlapping 15-mers covering the entire H56 sequence induced increased IFN-γ production by CD4<sup>+</sup> T cells from H56 and H56\_E *i.d.* vaccinated mice compared to unvaccinated controls (**Figure 1C**). H56 cDNA immunized mice had increased IFN-γ production in response to both the Ag85B-derived peptide H56242–262 and the ESAT-6 derived peptide H56288–307. No differences in TNF-α, IL-17, or IL-2 secretion by spleen CD4<sup>+</sup> T cells were detected between the groups (**Figure 1C**). Differences in CD8<sup>+</sup> T cell IFN-γ production were not significant, although there was a trend toward increased recognition of the H56 protein (as measured by responses to the peptide pool) in the H56 or H56\_E cDNA *i.d.* immunized mice (**Figure 1D**). IFN-γ ELISpot analysis of total lymph node cells also demonstrated increased numbers of cells producing IFN-γ in the H56 cDNA *i.d.* vaccinated mice after stimulation with the H56 peptide pool compared to both unvaccinated and BCG *s.c.* vaccinated mice (**Figure 1E**; Table S2 in Supplementary Material). Overall, the H56 cDNA tattoo immunization increased H56-specific IFN-γ-producing T cell responses.

# Increased CD4**+** TEM, TFH, and GC B Cells following Vaccination with BCG

T cell memory populations and B cell responses after vaccination were measured by flow cytometry at day 21 in spleens and draining lymph nodes (**Figure 2A**). Both *s.c*. and *i.d*. BCG immunization resulted in increased percentages of CD4<sup>+</sup> TEM among spleen CD4<sup>+</sup> T cells compared to unvaccinated mice (**Figure 2B**), with CD4<sup>+</sup> TFH showing similar trends (**Figure 2C**). Frequencies of CD4<sup>+</sup> TCM were higher in the spleens of DNA *i.d.* vaccinated mice than in BCG *s.c.* vaccinated mice (**Figure 2D**). TEM responses differed in draining lymph nodes in that only H56\_E immunized mice had increased TEM (**Figure 2E**). A similar pattern was observed for TFH in lymph nodes compared to spleen, with all responses increased in all vaccinated groups compared to unvaccinated controls (**Figure 2F**). This pattern was also seen for TCM in draining lymph nodes (**Figure 2G**). There were no significant differences in frequencies of CD8<sup>+</sup> TEM or CD8<sup>+</sup> TCM between the groups (data not shown).

Humoral responses may contribute to protection against TB (51–53). We measured the frequencies of plasma cells and GC cells among B cells by flow cytometry (**Figure 2H**) in spleens (**Figures 2I,J**) and draining lymph nodes (**Figures 2K,L**). Frequencies of GC B cells were increased in spleens of both groups vaccinated with BCG compared to unvaccinated mice (**Figure 2I**), with a similar trend in plasma cells (**Figure 2J**). In the draining lymph nodes, GC B cells appeared increased in all immunized groups (**Figure 2K**) while plasma cells were similar between groups (**Figure 2L**). Antibody responses were very variable between animals, possibly because of the mixed background of CB6F1 mice, which are first-generation offspring of female BALB/c and male C57/BL6 mice. Although antibody responses varied between mice, the trend was directed toward increased levels of mycobacteria-specific IgG, including IgG1 and IgG2c, in mice vaccinated with BCG, which expresses multiple mycobacterial proteins (**Figure 3A**) and increased levels of Ag85Bspecific IgG, IgG1, and IgG2c in the H56 cDNA vaccinated group (**Figure 3B**). ESAT-6-specific antibodies were not detected (data not shown). Curiously, antibody levels were lower in mice immunized with H56\_E DNA than with H56 DNA, suggesting that B cell epitopes may have been influenced by the sequence modification.

### H56 and H56\_E cDNA Tattoo Immunization Does Not Reduce Bacterial Loads in Mice

To evaluate the protective efficacy of the different vaccines, bacterial loads were measured in lungs and spleens at day 90 after infection with *Mtb* and lung pathology was quantified as cell infiltration per lung area in a blinded manner. Mice vaccinated with BCG *s.c.* benefited from an approximately 1 log reduction in bacterial load in the lung compared to unvaccinated mice, whereas mice vaccinated with BCG *i.d*. did not (**Figure 3C**). Bacterial loads were not decreased after *i.d.* immunization with H56 or H56\_E cDNA. A similar trend was seen in the spleen (**Figure 3D**). Furthermore, mice vaccinated with BCG strains had strongly decreased lung inflammation compared to unvaccinated mice (**Figures 3E,F**). H56 cDNA immunization did not decrease lung pathology compared to unvaccinated mice; however, there was reduced lung pathology in H56\_E cDNA immunized mice (**Figures 3E,F**). Overall, cDNA tattoo immunization *i.d.* with H56 constructs was not as protective as BCG *s.c*. against TB in our model, but H56\_E immunization ameliorated pathology.

### Boosting BCG with *i.d.* H56 cDNA Tattoo Immunization Increases Ag85B and ESAT-6 Specific CD4**+** and CD8**+** T Cell Cytokine Responses

Although *i.d.* H56\_E DNA tattoo vaccination did not reduce *Mtb* burdens, it ameliorated lung pathology, and we hypothesized that it may increase the efficacy of the BCG vaccine in a prime-boost system by boosting responses to Ag85B and inducing responses to the *Mtb-*specific antigens ESAT-6 and Rv2660c found in H56. Therefore, we tested a heterologous prime-boost vaccination regimen (**Figure 4A**). In addition, we tested whether repeating *i.d.* H56\_E DNA tattoo immunization (homologous prime/ boost) could improve its protective efficacy. Spleen H56-specific CD4<sup>+</sup> IFN-γ responses were increased at day 14 after boosting of BCG with *i.d.* H56 and H56\_E DNA tattoo immunization as well as by homologous H56\_E cDNA immunization, compared to BCG-only immunized groups (**Figure 4B**; Table S3 in Supplementary Material). In addition, spleen H56-specific CD4<sup>+</sup> TNF-α responses were increased in the homologous H56\_E cDNA immunized group compared to the BCG *s.c.* immunized group, and in both groups boosted with H56\_E DNA compared to the unvaccinated controls. The BCG/H56\_E immunized group had increased frequencies of Ag85B-specific CD4<sup>+</sup>TNF-α+ cells. Homologous cDNA prime-boost induced increased percentages of H56-specific and ESAT-6-specific IFN-γ CD8<sup>+</sup> T cells compared to both BCG *s.c.* immunized and unvaccinated groups (**Figure 4C**; Table S3 in Supplementary Material). BCG *s.c.* immunized mice had increased Ag85B-specific CD8<sup>+</sup>IFN-γ+ responses compared to unvaccinated mice, which were further boosted by H56\_E cDNA compared to BCG alone. ELIspots on lymph node

FIGURE 2 | Differences in T- and B cell sub-populations following immunization with cDNA and BCG. Samples were collected at day 21 post vaccination. Splenocytes and lymph node cells were stained for different surface markers and analyzed by FACS. Pooled data of two experiments (*n* = 10 in total per group) are shown ±SEM. (A) Gating strategy for FACS analysis for T cell sub-populations. After gating on live, single cells, cells were gated on CD3+ (R1). R1 was gated on CD8+ (R2) or CD4+ (R3). CD4+ and CD8+ T cells were then analyzed for expression of CD44, CD62L, and CXCR-5 to identify CD4+ TCM, CD4+ TEM, and CD4+ TFH subsets. (B–D) Percentages of (B) TEM, (C) TFH, and (D) TCM, among spleen CD4+ T cells. (E–G) Percentages of (E) TEM (F), TFH, and (G) TCM, among lymph node CD4+ T cells. (H) Gating strategy for FACS analysis for B cell sub-populations. After gating on live, single cells, B cells were gated as B220+ and CD19+ (R1). Expression of GL7, Fas, CD138 and CD38 to identify plasma cells and germinal center (GC) B cells. (I) Percentages of GC B cells among spleen B cells. (J) Percentages of plasma cells among spleen B cells. (K) Percentage of GC B cells among lymph node B cells. (L) Percentage plasma cells among lymph node B cells. (A–L) Significant differences between groups compared to the unvaccinated group are marked as \* and significant differences between the groups with lines (ANOVA with Tukey's multiple comparison test; \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001).

Pooled data of two experiments with (*n* = 10 in total per group) are shown ±SEM. (A) *Mtb* lysate specific total IgG, IgG1, and IgG2c antibody titers as measured at 1:10 dilution. (B) Ag85B lysate specific total IgG, IgG1, and IgG2c antibody titers as measured at 1:10 dilution. (C) Bacterial loads in lungs and (D) spleen. (E) Quantification of percentage of infiltrated area per lung. (F) Giemsa staining of infected lungs. Representative histology slides per group are shown. (A,B,E) Significant differences between antibody titers of treatment groups compared to the unvaccinated group are marked as \* and significant differences between the groups with lines (ANOVA with Tukey's multiple comparison test; \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001). (C,D) Significant differences between the groups are shown with \* (Kruskal–Wallis with Dunn's multiple comparison test; \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001).

cells showed increased IFN-γ responses to Ag85B and ESAT-6 CD4 epitopes as well as the H56 peptide pool in mice that had received homologous H56\_E cDNA prime-boost (**Figure 3D**; Table S4 in Supplementary Material). Analysis of T cell subsets demonstrated no differences in frequencies of spleen TEM, TCM, or TFH after DNA boosting (**Figures 4E–G**), with a similar pattern in the draining lymph nodes (**Figures 4H–J**). Frequencies of CD8<sup>+</sup> TCM and CD8<sup>+</sup> TEM were not significantly different after different immunization strategies, similar to the single vaccination experiments (data not shown).

Frequencies of H56-specific multi-cytokine-producing CD4<sup>+</sup> T cells after different vaccination regimes were analyzed (**Figure 5**), since cytokine-producing capacity has been correlated to differentiation state (40, 54, 55). BCG induced much lower frequencies of H56-specific cytokine-producing cells overall. Strikingly, there were differences in cytokine production profiles between groups receiving BCG alone, heterologous prime-boost with BCG and H56\_E cDNA, or homologous H56\_E cDNA prime-boost. Specifically, in the heterologous prime-boosted group, CD4+ T cells responding to Ag85Bderived H56242–262 were primarily IFN-γ+TNF-α+, while in the H56\_E/H56\_E cDNA vaccinated groups they expressed a more mixed cytokine-producing profile. By contrast, CD4<sup>+</sup> T cells responding to ESAT-6-derived H56288–307 were TNF-α+IL-2<sup>+</sup> in

FIGURE 5 | Frequencies in double or triple cytokine positive CD4+ T cells is correlated to immunization strategy. Intracellular cytokine staining was performed on splenocytes harvested at day 14 after the booster vaccination and re-stimulated with different H56 peptides. The peptide pools consisted of eighty-five 15-mer peptides spanning the entire sequence of H56, as a measure of the total percentage of H56-antigen-specific cells. CD4+ T cell epitopes included H56242–262 (Ag85B derived) and H56288–307 (ESAT-6 derived). Analysis was done using FACS. The results show pooled data from two independent experiments (*n* = 10 in total). Within the CD4+ population, frequencies of peptide-specific single, double, or triple positive cells is shown in bar graphs. The pie graphs denote the proportion of each cytokine-producing subset of the responding cells.

the BCG/H56\_E group and had mixed profiles in the H56\_E/ H56\_E and the BCG-only vaccinated groups. Responses to the peptide pool were different again, with BCG prime/DNA boost mice showing mostly IFN-γ+ cells, while the H56\_E/H56\_E vaccinated group had a mixed profile.

No significant differences were found in lymph node GC and plasma B cells between groups (data not shown), and antibody levels were also similar (**Figures 6A,B**) although the antimycobacteria IgG1 titer was increased in mice receiving a booster immunization with H56\_E cDNA compared to the group vaccinated with BCG alone (**Figure 6A**). Titers of anti-mycobacteria antibodies in H56\_E cDNA/H56\_E cDNA prime-boost mice remained as low as in unvaccinated controls (**Figure 6A**), but Ag85B-specific IgG2c titers were raised (**Figure 6B**).

## Despite Inducing H56-Specific Immune Responses, a Booster Immunization with H56\_E cDNA Does Not Ameliorate Bacterial Loads Compared to BCG Vaccination in Mice

To determine the protective efficacy of the different prime-boost strategies, bacterial loads, and lung pathology were measured 90 days after *Mtb* challenge. BCG and BCG/H56\_E cDNA immunized groups had reduced lung bacterial burdens compared to the unvaccinated mice; however, there was no significant improvement in bacterial loads after boosting with H56\_E cDNA (**Figures 6C,D**). Boosting H56\_E cDNA prime with a homologous immunization did not increase efficacy, and mice were not protected over unvaccinated controls. Bacterial loads in spleens followed a similar pattern (**Figure 6D**), although vaccination was not as protective as in the lung. Quantification of infiltrating cells in the lungs by image analysis also revealed H56\_E homologous prime-boost mice to be the least protected (**Figures 6E,F**). Mice primed with BCG and boosted with H56\_E cDNA were the only group to show significant improvement in lung pathology compared to the unvaccinated group and tended to have less cell infiltration than the BCG-vaccinated mice too, suggesting that the H56-specific T cell responses associated with boosting with H56\_E cDNA did have a beneficial effect in ameliorating disease. Overall, protective efficacy relied on the presence of BCG in the vaccine regimen, suggesting that broader anti-mycobacterial responses may be required for optimal protection in addition to the generation of highly specific responses against particular antigens.

### DISCUSSION

Previously, we demonstrated that the immunogenicity of a DNA vector containing H56 cDNA could be enhanced by fusion of the H56 sequence to cDNA of TTFC and by altering epitope flanking residues to facilitate epitope processing (43). Here, we tested the efficacy of immunization with the optimized constructs against *Mtb* challenge. The results demonstrate that *i.d.* tattoo vaccination with an optimized H56\_E cDNA sequence fused to TTFC cDNA, either as a standalone vaccine or as a booster to BCG, did not reduce *Mtb* burdens in the lung compared to BCG alone, although it showed a tendency to improve lung pathology.

Bacillus Calmette–Guérin is the only licensed vaccine against TB, but it provides variable and inadequate protection against the disease (10). Therefore, recombinant BCG vaccines and other live-attenuated mycobacterial strains are being tested as replacement vaccines, while subunit vaccines containing new antigens, adjuvants, or viral vectors are being investigated for their ability to boost BCG (13–18, 40–42). Boosting of BCG prime vaccination with CAF01-adjuvanted *Mtb* fusion protein H56 (containing Ag85B, ESAT-6, and Rv2660 components) increased BCG-induced protection against TB in a murine model (33) and H56 administered in IC31 adjuvant as a booster following BCG prime vaccination was more protective than BCG alone in macaques (42). In order to test alternative methods of subunit vaccine delivery that could be used in prime-boost regimens without the need for adjuvants, we generated H56 encoding DNA constructs with H56 cDNA fused to TTFC cDNA (43). Several naked DNA vaccine candidates have been tested against TB in mice, with various degrees of immunogenicity and prophylactic efficacy (56), but the protective efficacy of tattoo immunization with constructs containing *Mtb* antigens has not been assessed previously. We tested whether *i.d.* H56 cDNA tattoo immunization was protective against TB in a mouse model, either as a standalone vaccine or as a heterologous boost to BCG prime. We found that *i.d.* tattoo immunization with H56 constructs did not protect against TB in mice, although it increased H56-specific T cell responses, spleen GC B cells, and antibody titers when given as a boost to BCG prime. Overall, our results showed that BCG immunization was more protective than the highly specific response to a few antigens induced by *i.d.* H56 cDNA immunization.

Understanding the underlying mechanisms of protective immunity to *Mtb* will assist in the rational design of effective vaccines (57). Our experiments comparing *i.d.* DNA immunization with H56 constructs and *s.c.* vaccination with BCG demonstrated that the two types of vaccines elicited distinct immune responses. BCG induced increased frequencies of CD4<sup>+</sup> TEM, TFH, GC B cells, and mycobacteria-specific antibodies, while H56 cDNA immunization induced H56-specific CD4<sup>+</sup>IFN-γ+ T cell responses and Ag85B-specific IgG antibodies. BCG shares numerous mycobacterial antigens with *Mtb*, including the immunodominant antigen Ag85B, while the DNA tattoo immunization elicits specific responses against H56, demonstrated by responses against ESAT-6 and Ag85B. Mice vaccinated with BCG accordingly had relatively high levels of mycobacteria-specific antibodies; while mice administered cDNA *i.d.* (particularly the unmodified H56 sequence) developed higher titers of Ag85Bspecific antibodies. Both BCG and the H56 cDNA constructs elicited Ag85B-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cells, with the peptide H56242–262 being a particularly strong inducer of IFN-γ responses by CD4<sup>+</sup> T cells.

Mice immunized with BCG *s.c.* were better protected against *Mtb* challenge than mice that received H56 cDNA *i.d.*, suggesting that broader anti-mycobacterial responses are an important component of vaccine-induced protection. Unlike viruses, which contain only a small number of antigens forming obvious vaccine

FIGURE 6 | Heterologous prime-boost with BCG/H56\_E cDNA decreases lung pathology but not bacterial loads. Serum was harvested at day 14 after the booster vaccination and lungs and spleens were collected at day 90 post *Mycobacterium tuberculosis* (*Mtb*) challenge were plated in serial dilution. Pooled data of two experiments with (*n* = 10 in total per group) are shown ±SEM. (A,B) Total IgG, IgG1, and IgG2c antibodies recognizing (A) *Mtb* lysate or (B) Ag85B were measured in serum of immunized mice at day 40 (total IgG: 1 in 100 dilution, IgG1 and IgG2c, 1 in 10 dilution). (C,D) Lungs and spleens collected at day 90 post *Mtb* challenge were plated in serial dilution. Pooled data of two experiments with (*n* = 10 in total per group) is shown ±SEM. (C) Bacterial loads in lungs and (D) spleen. (E) Quantification of percentage of infiltrated area per lung. (F) Giemsa staining of infected lungs. Representative histology slides per group are shown. (A–E) Significant differences between treatment groups and the unvaccinated group are marked as \* and significant differences between groups with lines. \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001. (A–E) ANOVA with Tukey's multiple comparison test; (C,D) Kruskal–Wallis with Dunn's multiple comparison test.

candidates, bacterial pathogens such as *Mtb* contain thousands of potential antigens, and no single antigen has been identified that is clearly associated with protective responses. This may be why it has been difficult for subunit vaccines to surpass live whole cell vaccines such as BCG in mouse models (10). Live BCG persists in mice for up to 16 months post vaccination and disseminates in the host, with T cell responses peaking around week 32 and waning thereafter (58, 59), whereas T cell responses against DNA vaccines could be measured only up to day 50 (43). Previously, we showed that BCG disseminates to the lungs after *s.c.* vaccination, where it persists for about 55 days (22). *I.d*. tattoo vaccination with BCG showed a trend toward decreased control of lung bacterial burdens compared to *s.c*. vaccination, however, lung pathology and spleen bacterial burdens were similar to that of *s.c.* vaccinated mice. BCG is most commonly administered by *i.d.* injection (not tattoo) in the clinic, and this route was found to be equivalent to percutaneous administration in both efficacy and safety over a 2-year follow-up period (60).

H56 and H56\_E cDNA were not protective as standalone vaccines; hence, we tested whether a homologous prime-boost regimen could improve the efficacy, using H56\_E cDNA. Homologous H56\_E cDNA prime-boost mice were also not protected compared to unvaccinated mice after *Mtb* challenge. These mice had the most vigorous H56-specific CD8<sup>+</sup> T cell responses, indicating that this response was not sufficiently effective in preventing murine TB, at least during active infection. Furthermore, mice in this group hardly produced any *Mtb*- and Ag85B-specific IgG, with only a few animals producing Ag85B-specific IgG2c. It is possible that alterations to the H56\_E sequence may have led to a change in B cell Ag85B epitopes or skewing of immune responses away from Ag85B antibody-producing conditions, since in the single-dose vaccination experiments, H56 cDNA also led to increased antibody titers compared to H56\_E cDNA.

Which immune responses are required for optimal protection against TB remains a key question in the attempt to generate a more effective vaccine. In our previous studies with BCG *ΔureC::hly* and BCG *ΔureC::hly ΔnuoG*, increased protection after vaccination against TB coincided with increased TEM, TCM, TFH, GC B cells, and mycobacteria-specific antibody titers as compared to BCG vaccination (22). CD4<sup>+</sup> TCM, by virtue of their capacity to generate new CD4<sup>+</sup> TEM, are considered important in long-term immunity to *Mtb*. Here, boosting BCG by *i.d.* immunization with H56\_E cDNA did not induce elevated CD4<sup>+</sup> TCM over BCG alone. The adjuvant CAF01 induces TCM responses (40) and homing of cytokine-producing cells to the lung parenchyma (61), which could explain why boosting with H56 protein in CAF01 adjuvant was more effective than boosting with the H56 cDNA tattoo immunization. In macaques vaccinated with BCG followed by an H56 booster in IC31 adjuvant, early recall responses to the vaccine antigens were associated with protection (42). A comparison between H56 cDNA and H56 protein/ adjuvant immunization could provide further insights into the type of immunity required for protection against TB. In addition, increased H56-specific CD8<sup>+</sup> T cell responses did not coincide with increased protective efficacy in our model. CD8<sup>+</sup> T cells play a critical role in vaccine-induced immunity to TB in macaques, whose CD8<sup>+</sup> T cells are more similar to humans (62). In mice, CD8<sup>+</sup> T cells appear to play a more important role in the chronic phase of TB than in the acute phase (28). Several studies have shown that antibodies and B cells may also play a role in immunity to TB (51–53). Interestingly, a previous study in guinea pigs found that boosting a recombinant BCG expressing ESAT-6 with intramuscular ESAT-6 cDNA injection reduced protection compared to BCG alone, despite inducing increased antigen-specific IFN-γ responses (63). Together with our study, this suggests that cDNA immunization as a whole might not induce the type of immune responses required for immunity to TB.

In summary, *i.d.* DNA tattoo vaccination using H56 and H56\_E constructs alone or in combination with BCG did not significantly improve upon protection induced by BCG vaccination alone in our model, despite inducing strong H56-specific CD4<sup>+</sup> T cell and CD8<sup>+</sup> T cell responses. Nevertheless, the H56-specific responses induced by H56\_E cDNA vaccination did ameliorate lung pathology, demonstrating the value of adding subunit vaccines containing defined *Mtb* antigens as a booster to BCG. Differences in T cell cytokines and antibody production following the different vaccination regimens as well as the limited number of antigens in H56 compared to BCG may explain the findings. Understanding why the types of immune responses generated by DNA tattoo vaccination does not suffice for protection against TB in contrast to responses generated by BCG or H56 administered with adjuvants such as CAF01 and IC31 will lead to optimized TB vaccination strategies.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the GV-SOLAS. The protocol was approved by the State Office for Health and Social Services, Berlin, Germany.

# AUTHOR CONTRIBUTIONS

AP and NN contributed equally to this work. AP, NN, AS, and SK conceptualized the study and wrote the manuscript. AP, NN, UZ, and SS performed experiments. AP, NN, TD, and VB performed data analysis. All authors contributed to the manuscript preparation. AS and SK are joint senior authors.

# ACKNOWLEDGMENTS

Support was by European Union's Seventh Framework Programme [FP7/2007-2013]—Grant No. 280873 ADITEC to AS and SK, Boehringer Ingelheim funds travel grant and EFIS travel grant to AP. We thank Dr. Donatella Negri for providing the 15-mers of H56, Dr. Alexis Vogelzang for technical support, Dr Gesa Rausch for assistance with animal ethics, and the staff of the animal facility.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01744/ full#supplementary-material.


### REFERENCES


Platteel et al. cDNA Tattoo Immunization against *Mtb*

Bacillus Calmette-Guérin ΔureC::hly vaccine's superior protection against tuberculosis. *J Infect Dis* (2014) 210:1928–37. doi:10.1093/infdis/jiu347


**Conflict of Interest Statement:** SK is co-inventor/patent holder of BCG *ΔureC::hly* (VPM1002). The other authors declare to have no conflicts of interest.

*Copyright © 2017 Platteel, Nieuwenhuizen, Domaszewska, Schürer, Zedler, Brinkmann, Sijts and Kaufmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Anne Marit de Groot 1†, Anouk C. M. Platteel1†, Nico Kuijt <sup>2</sup> , Peter J. S. van Kooten1 , Pieter Jan Vos2 , Alice J. A. M. Sijts1 \* and Koen van der Maaden2 \**

*1Department of Infectious Diseases and Immunology, Faculty of Veterinary Sciences, Utrecht University, Utrecht, Netherlands, 2MyLife Technologies, Leiden, Netherlands*

### *Edited by:*

*Peter Andersen, State Serum Institute (SSI), Denmark*

### *Reviewed by:*

*Ed C. Lavelle, Trinity College, Ireland Ji Wang, Harvard Medical School, United States*

### *\*Correspondence:*

*Alice J. A. M. Sijts e.j.a.m.sijts@uu.nl; Koen van der Maaden maaden@mylifetechnologies.nl*

*† These authors have contributed equally to this work.*

### *Specialty section:*

*This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology*

*Received: 06 September 2017 Accepted: 29 November 2017 Published: 13 December 2017*

### *Citation:*

*de Groot AM, Platteel ACM, Kuijt N, van Kooten PJS, Vos PJ, Sijts AJAM and van der Maaden K (2017) Nanoporous Microneedle Arrays Effectively Induce Antibody Responses against Diphtheria and Tetanus Toxoid. Front. Immunol. 8:1789. doi: 10.3389/fimmu.2017.01789*

The skin is immunologically very potent because of the high number of antigenpresenting cells in the dermis and epidermis, and is therefore considered to be very suitable for vaccination. However, the skin's physical barrier, the stratum corneum, prevents foreign substances, including vaccines, from entering the skin. Microneedles, which are needle-like structures with dimensions in the micrometer range, form a relatively new approach to circumvent the stratum corneum, allowing for minimally invasive and pain-free vaccination. In this study, we tested ceramic nanoporous microneedle arrays (npMNAs), representing a novel microneedle-based drug delivery technology, for their ability to deliver the subunit vaccines diphtheria toxoid (DT) and tetanus toxoid (TT) intradermally. First, the piercing ability of the ceramic (alumina) npMNAs, which contained over 100 microneedles per array, a length of 475 µm, and an average pore size of 80 nm, was evaluated in mouse skin. Then, the hydrodynamic diameters of DT and TT and the loading of DT, TT, and imiquimod into, and subsequent release from the npMNAs were assessed *in vitro*. It was shown that DT and TT were successfully loaded into the tips of the ceramic nanoporous microneedles, and by using near-infrared fluorescently labeled antigens, we found that DT and TT were released following piercing of the antigen-loaded npMNAs into *ex vivo* murine skin. Finally, the application of DTand TT-loaded npMNAs onto mouse skin *in vivo* led to the induction of antigen-specific antibodies, with titers similar to those obtained upon subcutaneous immunization with a similar dose. In conclusion, we show for the first time, the potential of npMNAs for intradermal (ID) immunization with subunit vaccines, which opens possibilities for future ID vaccination designs.

Keywords: nanoporous microneedles, intradermal vaccination, antigen release, humoral immune response, diphtheria, tetanus

**Abbreviations:** IMQ, imiquimod; MNA, microneedle array; npMNA, nanoporous microneedle array; IgG, immune globulin G; DT, diphtheria toxoid; TT, tetanus toxoid; TLR, toll-like receptor; DLS, dynamic light scatter; Lf, limits of flocculation; HPLC, high-performance liquid chromatography; TFA, trifluoroacetic acid; TMB, 3,3′,5,5′-tetramethylbenzidine; BSA, bovine serum albumin; PBS, phosphate-buffered saline; ACN, acetonitrile; HRP, horseradish peroxidase; OD, optical density; PBST, 0.01% Tween 20 in PBS.

# INTRODUCTION

The skin has great potential for vaccine delivery, because it is a large organ that is easy to reach. Delivery *via* the skin circumvents degradation challenges to biomacromolecules, as posed, for example, by the gastrointestinal delivery route (1, 2). The skin, with the stratum corneum as outer barrier, is designed to keep foreign materials including pathogens out of the body. Besides, the skin is immunologically very potent, with various professional antigen-presenting cells, such as dermal dendritic cells and Langerhans cells (3, 4), present in the dermis and epidermis, respectively. To circumvent the barrier function of the stratum corneum and reach antigen-presenting cells for vaccination purposes, microneedles can be used. Microneedles are needlelike structures with a length in the micrometer range and are a promising tool to deliver drugs and vaccines across the barrier. Furthermore, they represent a possible painless vaccination method (5), they present reduced contamination risks compared with traditional needles, they allow for injection by less trained personnel and even have potential for self-administration (6). However, microneedles need to be sufficiently long and strong enough to pierce the stratum corneum, but also preferably short enough to not reach the nociceptors. Various microneedles are under development, which are hollow-, solid-, dissolving-, or less known porous structured (6–10). For all types, multiple strategies have been investigated for the delivery of vaccine antigens into the skin, as reviewed by van der Maaden et al. (10).

Porous microneedles, which may be used as a single-unit-drug delivery system, can be prepared from pore-forming materials (11), from (nano)particles (12, 13), or by making solid microneedle material porous (14, 15). Porous microneedle arrays (MNAs) can be loaded with a drug, by loading the formulation into the pores of the MNAs. The drug is released when the microneedles are pierced into the skin *via* diffusion from the pores. To date, several materials have been used for the production of porous MNAs. Among these are biodegradable polymeric porous MNAs with a porosity of 75%, which, however, lack the strength to penetrate the skin (11). When using a brittle material, like silicon, the pores that are introduced in the material need to be sufficiently small to provide enough strength for skin piercing (14, 15). The use of porous silicon material, therefore, is limited to the delivery of low-molecular weight drugs (10). Using self-setting ceramics for production of porous MNAs increases MNA strength. However, drug loading into these MNAs requires circumstances that are unfavorable for formulating biomacromolecules, because it involves exothermic reactions or organic solvents (ethanol) (16).

In this study, microneedles composed of a biocompatible ceramic, alumina (Al2O3) (12), were tested for their suitability for intradermal (ID) vaccination. With an average pore size of 80 nm and an estimated porosity of 40%, these microneedles allow for encapsulation of large biomacromolecules before production (10, 13). In previous studies, it was shown that alumina nanoporous microneedle arrays (npMNA) can be successfully loaded with small molecules or nanoparticles with sizes up to 100 nm in solution or dispersion *via* absorption (*via* capillary forces), respectively, and to release these substances *in vitro* by diffusion. The npMNAs had sufficient strength to reproducibly pierce *ex vivo* human skin without breaking (10) and have shown to activate immune cells upon dermal application of peptide-loaded npMNAs in a murine model (17). However, characterization of ceramic alumina npMNAs loaded with larger, more relevant molecules, such as subunit vaccine antigens, had not been performed so far.

In this study, characterization and application of alumina npMNAs are described. Loading of npMNAs with diphtheria toxoid (DT) and tetanus toxoid (TT), antigen release in murine skin *ex vivo*, and *in vivo* immunogenicity of npMNA-delivered antigens were examined. We show that npMNA-mediated vaccine delivery elicits TT- and DT-specific antibody responses in mice, comparable to those induced by subcutaneous (SC) immunization with a similar dose. This is the first report showing the potential of porous microneedles in dermal immunization with subunit vaccines.

### MATERIALS AND METHODS

### Materials

Diphtheria toxoid and TT (for *in vitro* assays) were obtained from Staten Serum Institute (Copenhagen, Denmark) and imiquimod (IMQ) Vaccigrade was obtained from Invivogen. Trifluoroacetic acid (TFA), 3,3′,5,5′-tetramethylbenzidine (TMB) and bovine serum albumin (BSA), and 0.4% (w/v) were obtained from Sigma Aldrich. High-performance liquid chromatography (HPLC)-R grade acetonitrile (ACN) was from Biosolve and phosphate-buffered saline (PBS; pH 7.2, 1.5 mM KH2PO4, 2.7 mM Na2HPO4–7H2O, and 155 mM NaCl) was from Gibco (ThermoFisher Scientific). IRdye800cw carboxylic acid *N*-hydroxy succinimide ester (IRdye800cw-NHS) was purchased from Li-cor (Lincoln, NE, USA). Dexdomitor was purchased from Orion Corporation, Narketan ketamine from Vétoquinol, and Atipam was purchased from Dechra. Goat anti-mouse immune globulin G (IgG) total horseradish peroxidase (HRP) (GaM-IgG total HRP), goat anti-mouse IgG1 HRP (GaM-IgG1 HRP), and goat anti-mouse IgG2a HRP (GaM-IgG2a HRP) were obtained from Southern Biotech and microtiter plates 9610 used for ELISA were from Corning Costar.

### Preparation and Characterization of npMNA

Nanoporous microneedle arrays were produced by using a double replication technology as previously described (12). In brief, from an inverse silicon master a first positive PDMS mold was created, from which a second inverse PDMS mold was produced. Alumina npMNAs were fabricated at LouwersHanique B.V. from the second PDMS mold according to MyLife Technologies' proprietary manufacturing procedure (18), using a slurry that contains alumina nanoparticles with an average size of 300 nm and a plasticizer. After controlled drying, the resulting MNAs were removed from the PDMS mold and were sintered at 1,450°C. This results in removal of the plasticizer and the formation of nanoporous alumina material with an average pore size of 80 nm and a porosity of approximately 40% (10, 12). Microneedles used in this study had a length of 475 µm and a density of 150 microneedles/cm2 on a back plate of 0.7 cm2 (105 microneedles/ array; **Figures 1A,B**). The total volume in the nanopores of only the tips of the microneedles of a single MNA was calculated to be 0.25 µL. Bruker Nano Surface analysis was performed to characterize the geometry of the npMNAs.

### Preparation of npMNAs with Antigen-Loaded Microneedle Tips

To only load the tips of the microneedles, microneedles were pierced through a foil (Parafilm®) by using a UAFM-V1 electrical applicator (uPRAX Microsolutions) at a velocity of 65 cm/s. Next, a drop of 5 µL drug formulation was applied onto the foilpierced npMNAs to absorb a drug/vaccine formulation into the microneedle tips. After 5 s, the surplus drop of drug formulation and the foil were sequentially removed from the npMNAs. To confirm that only the microneedle tips can be loaded with a drug formulation, the tips of npMNAs were loaded with a 0.4% trypan blue solution as described earlier.

### Skin Penetration

To test the piercing ability, a npMNA was applied trice on the dorsal side of *ex vivo* murine ears (Balb/C), which were collected from surplus mice, by using 3D-printed uPRAX impact applicators (**Figure 1C**) having an average holding force of 4.08 ± 0.75 N (mean ± SD, *n* = 19). Subsequently, the three pierced *ex vivo* mouse ears were incubated with 50 µL of a 0.4% trypan blue solution at room temperature. After 30 min, the trypan blue was removed, and the ears were washed in 10 mL PBS. Finally, the blue dots (piercings) were counted, from which the penetration efficiency was calculated.

### Hydrodynamic Diameter and Size Distribution of Antigens

To determine whether the npMNAs are suitable devices to load and release subunit antigens DT and TT, the hydrodynamic diameter and size distribution of DT and TT were determined by using dynamic light scattering on a Zetasizer Nano (Malvern Instruments). For these measurements, DT and TT were at a concentration of 0.8 and 0.4 mg/mL, respectively.

### Release of IMQ and Antigen from Nanoporous Alumina *In Vitro*

Imiquimod has the potential to enhance the immunogenicity of antigens in the skin (19). To evaluate how IMQ is released from IMQ-loaded nanoporous material in the presence of DT and TT, npMNAs were loaded by applying a drop (on the microneedle side of a npMNA that were not pierced through a foil) of 5 µL PBS containing only 2.5 µg IMQ, or 2.5 µg IMQ and 2.5 Lf DT or TT. Such a drop is absorbed into an npMNA within seconds because of capillary forces. Next, the IMQ-loaded npMNAs were incubated in 2.5 mL release buffer (PBS) under shaking at 300 rpm, and samples of 75 µL were taken in duplicate at different time points (1, 5, 10, 30, 60, 120, and 240 min). The IMQ concentration in the release buffer was determined by using HPLC analysis on an Agilent 1100 series HPLC equipped with a UV detector using a Phenomenex Kinetex 150 mm × 4.6 mm 2.6 μm EVO C18 column. A linear gradient of 5% solvent A (ACN with 0.1% TFA) to 68% solvent B (milliQ with 0.1% TFA) from 0 to 12 min was detected at a wavelength of 242 nm at a retention time of 9.8 min.

To assess the release of antigen from antigen-loaded nanoporous material, npMNAs (that were not pierced through a foil) were loaded with 15 µL PBS that contained 5 Lf DT or TT as described earlier. Next, antigen-loaded npMNAs were incubated in 4 mL release buffer while shaking at 300 rpm. At different time points (1, 5, 10, 30, 60, 120, and 240 min), samples of 500 µL were taken in which the released amount of antigen was quantified by measuring the intrinsic fluorescence (emission wavelength of 348 nm) at an excitation wavelength of 280 nm and using standard concentrations of DT and TT ranging from 0.01 to 50 ng/mL on a Tecan Infinite M1000 plate reader. The release of antigen in the presence of IMQ was not investigated, because IMQ is fluorescent at similar wavelengths (excitation at 260 nm and emission at 340 nm) that are used to measure the intrinsic fluorescence of the antigens (20).

### Fluorescently Labeling of Antigens

To quantify the amount of DT and TT that is delivered from DTand TT-loaded npMNAs into skin, DT and TT were labeled with a near-infrared fluorescent dye (IRdye800cw-NHS). To this end, 1 mg/mL solutions of DT and TT in a 100 mM carbonate buffer pH 8.5 were prepared. Subsequently, 1 mL of each of these solutions was added to 500 µg of IRdye800cw-NHS. After 1 h shaking (300 rpm) at 37°C, the free dye was removed and the carbonate buffer was exchanged by PBS using a Zeba™ spin desalting column with a molecular weight cutoff (MWCO) of 7 kDa (Thermo Fisher Scientific). Next, IRdye800cw-labeled antigens were concentrated approximately 50 times by using 0.5 mL Amicon (Millipore) 10 kDa MWCO filters. Finally, the concentration of IRdye800cw-labeled DT and TT was determined by using a calibration curve of non-labeled antigens and measuring the intrinsic fluorescence (as described earlier).

### Quantification of Antigen in *Ex Vivo* Mouse Ears

To quantify the delivered amount of DT and TT into murine skin, npMNAs of which only the tips were loaded (using foil piercing) with fluorescently labeled antigens were prepared by using 5 µL of 12 Lf/μL (DT) and 6 Lf/μL (TT), as described earlier. The microneedles were applied by impact application and retained onto the skin by using a uPRAX 3D-printed applicator. After 30 min at room temperature, the antigen-loaded npMNAs were removed from the ears, and their fluorescence was compared with the fluorescence of standard solutions having known amounts of fluorescently labeled DT and TT, by using a IVIS® lumina II equipped with an ICG filter set. The intradermally delivered amounts of DT and TT were quantified by using Living Image® software (version 4.3.1).

### Preparation of Vaccine Formulations for Loading npMNAs for Immunization

Subunit vaccine formulations of DT and TT for loading the microneedle tips were prepared from antigen stock solutions (2.0 and 0.7 Lf/μL, respectively). Antigen stock solutions were concentrated (6–30×) by using 0.5 mL Amicon (Millipore) 10 kDa MWCO filters. Next, the concentration of the concentrates was determined by measuring the intrinsic fluorescence as described earlier. Finally, the antigen concentration was adjusted by diluting the concentrates in PBS to a concentration of 12 and 6 Lf/μL for DT and TT, respectively.

### Immunization of Mice

Seven-week-old Balb/C female mice (10 mice per group) obtained from Charles River (France) were immunized with 1.2 Lf (~0.50 μg) DT and 1.5 Lf (~0.77 μg) TT, or with 0.6 Lf DT and 0.75 Lf TT adjuvanted with 0.5 µg IMQ, on days 0, 21, and 42. The vaccine was administered *via* a SC injection of 100 µL in the neck using traditional hypodermic needles, or by dermal administration in the ear pinnae by using microneedles of which only the tips were loaded with vaccine formulation. Before each microneedle-based immunization, mice were anesthetized with 30 mg/kg ketamine and 0.1 mg/kg Dexdomitor by intraperitoneal injection. After the microneedles were removed, the anesthetic was antagonized with 0.4 mg/kg Atipam. On each ear, a DT- and TT-loaded npMNA was applied for 30 min by using a uPRAX applicator (**Figure 1C**). As a negative control mice were mock immunized *via* a PBS-loaded npMNA. Blood samples were collected from the tail vein 1 day before each immunization and serum was obtained by centrifugation; spleens were collected at day 47 (**Figure 1D**).

# DT- and TT-Specific IgG Total, IgG1, and IgG2a Titers

Diphtheria toxoid- and TT-specific antibody titers were determined using ELISA. ELISA plates were coated with 0.2 µg DT or 0.2 µg TT for 30 min and then blocked with 1% BSA in PBS for 15 min. Thereafter, 50 µL of serum sample at dilutions ranging from 1:25 until 1:200 (day 0), or from 1:200 until 1:25,000 (day 20, 41, and 47) were added, for 30 min. After extensive washing with 0.01% Tween 20 in PBS (PBST), wells were incubated for 30 min with GaM-IgG total HRP (1:5,000), GaM-IgG1 HRP (1:3,000), or GaM-IgG2a HRP (1:5,000). After extensive washing with PBST, antibody titers were quantified by adding 50 µL of stock TMB. Reactions were stopped after 60 s, with 100 µL of 1 M H2SO4, and the absorption was measured at a wavelength of 450 nm, with a reference wavelength of 650 nm, on a Microplate reader 550 (Bio-Rad). Titers of all animals at all time points for each isotype were measured in one experiment per antigen.

### Statistical Analysis

From 4 optical density (OD) values (at a wavelength of 450 nm) of diluted serum samples, the EC50 midpoint titers were determined using GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA, v6.05). Immunized mice showing OD values below the mean OD values + three times the SD measured for PBS treated mice were considered as non-responders and were given an arbitrary value of 0, equal to 10Log value of 1. Statistical differences between immunization groups were determined using a non-parametric one-way ANOVA Kruskal–Wallis test with a Dunn's test for multiple comparisons, and statistical significance was presented as follows: \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001, and ns, not significant. Ratio of IgG1:IgG2a was determined by dividing midpoint titers of individual isotypes, and if animals were considered as non-responders for one of the isotypes (see above), they were excluded from ratio analysis.

### RESULTS

### Characteristics of npMNA

Nanoporous microneedle arrays fabricated from alumina nanoparticles as previously reported (10), were characterized for geometry and dimensions *via* surface Bruker analysis, which showed that the ceramic microneedles had an average length of 475 µm and a needle shaft diameter of 275 µm (**Figure 2A**). For economic reasons, loading of only the tip of the microneedle with vaccine is an advantage, as residual vaccine quantities in the npMNAs will be strongly reduced. To investigate the possibility to only load the microneedle tips, npMNAs were pierced through

Figure 2 | (A) Bruker analysis was used for geometry and surface analysis and to measure the distance between the microneedle backplate and microneedle tip. The color is indicative for the size of the substrate-fillable microneedles. (B) Brightfield microscopy images of a nanoporous microneedle array (npMNA) of which only the microneedle tips are loaded with a trypan blue solution. (C) Representative image of a trypan blue piercing assay of *ex vivo* murine ears with an npMNA using the uPRAX applicator.

foil and tips were subsequently loaded with a trypan blue solution for visualization. Brightfield microscopy showed successful tip loading and no loading of the backplate of the npMNA (**Figure 2B**).

### Strength of npMNA by Skin Penetration

The ability of npMNAs to penetrate the skin is essential for ID antigen delivery. To determine whether the npMNAs are strong enough to penetrate the skin effectively and reproducibly, skin piercing was evaluated in *ex vivo* murine ear skin using a trypan blue assay (**Figure 2C**). Using the npMNAs resulted in an average piercing efficiency of 87 ± 17% (mean ± SD, *n* = 3). No visual breakage or reduction in microneedle strength or sharpness was observed. Together, these data show that the developed npMNA can be used to repeatedly penetrate the skin without breakage. To determine whether the npMNAs, having an average pore size of 80 nm, are suitable to be loaded with subunit vaccine antigens, the hydrodynamic diameter of DT and TT were determined by using dynamic light scatter. This revealed that DT and TT had a hydrodynamic diameter of 8.7 ± 2.8 nm (mean ± SD, *n* = 3) and 13.5 ± 5.6 nm (mean ± SD, *n* = 3), respectively, (**Figures 3A,B**). Therefore, the npMNAs should be suitable to be loaded with DT and TT into their nanopores.

### Antigen and Adjuvant Loading and Release *In Vitro*

After the npMNAs were loaded with either one of the antigens, the release of these antigens from the npMNAs in a release buffer was determined by measuring the intrinsic fluorescence of the antigens. After antigen-loaded npMNAs were incubated in release buffer for 30 min, 30% of both DT and TT were released from the npMNAs (**Figure 3C**). Besides, the release of IMQ from IMQ-loaded npMNAs was quantified after incubating them in a release buffer and using HPLC with UV detection. This revealed that approximately 50% of the npMNA-loaded IMQ was released after 30 min. Furthermore, it was observed that the co-delivery of IMQ and DT or TT did not result in a decreased release rate (**Figure 3D**). The release of IMQ reached a plateau at 60%, which indicates that IMQ partially adsorbs onto the npMNA. This was confirmed by incubating non-loaded npMNAs in an IMQ-containing buffer, having the same amount of IMQ as the IMQ-loaded npMNAs. The concentration of IMQ in the buffer decreased from 100 to 60% over time, showing that 0.2 µg IMQ was adsorbed onto the npMNA surface (data not shown). The effect of IMQ on the release of DT or TT could not be assessed due to interference with the fluorescence of IMQ (20). Together, these data showed that ceramic alumina npMNAs are suitable to be loaded with the subunit vaccine proteins DT and TT and the adjuvant IMQ, and that the antigens and adjuvant are released *in vitro*.

### Release of Antigen into *Ex Vivo* Skin

Next, the delivery of fluorescently labeled antigens from npMNAs into *ex vivo* murine skin was investigated. The antigen dose delivered into the skin was quantified after the application of the fluorescently labeled antigen-loaded npMNAs onto mouse ears. The delivery of the antigens into the ears was quantified by using infrared fluorescence imaging (**Figures 4A,B**) and was compared with a gradient of known amounts of fluorescently labeled antigens (**Figures 4C–E**). The delivery of DT was 0.61 ± 0.44 Lf (which corresponds with ~0.25 ± 0.18 μg) per MNA (**Figure 4A**), and the delivery of TT was 0.77 ± 0.23 Lf (~0.38 ± 0.11 μg) per MNA (**Figure 4B**).

Delivery efficiencies of DT and TT from the npMNAs were calculated by using the geometric values of the npMNAs. With an estimated total tip pore volume of 0.25 µL, the loaded amount of DT was calculated as 0.25 µL × 12 Lf/μL = 3 Lf/MNA. With a release of 0.61 Lf out of 3 Lf loaded, 20% delivery efficiency was achieved for DT, after 30 min of application onto the skin. For TT, an amount of 0.25 µL × 6 Lf/μL = 1.5 Lf/MNA was loaded and with a release of 0.77 Lf, the delivery efficiency was 0.77 Lf/1.5 Lf = 51%, after 30 min. For immunization studies, two arrays per mouse were used per antigen and this resulted in a delivery of 1.25 Lf (~0.5 μg) DT and a delivery of 1.53 Lf (~0.77 μg) TT. These values of delivered doses correspond with 26 and 31%, respectively, of the currently used human vaccination dose (5 Lf).

Figure 4 | Representative quantification image of the delivery of fluorescently labeled antigen into mouse ears. (A,B) An overlay of picture of the mouse ear and infrared fluorescence imaging. Two independent ear piercing experiments are shown for diphtheria toxoid (DT) (*n* = 2) (A) and tetanus toxoid (*n* = 3) (B). (C,D) Background fluorescence without and with mouse ear. (E) Gradient of solution containing 0.24, 0.6, and 1.2 Lf DT. Blue circles all indicate region of interest and have an equal diameter in all cases.

## Immune Response after Dermal Immunization

To determine whether npMNA-mediated ID delivery induces antigen-specific immunity, mice were immunized with both DT (1.2 Lf) and TT (1.5 Lf) using antigen-loaded npMNAs (ID administration) or using a needle and a syringe (SC injection). After each immunization, antibody titers in the serum were determined (**Figure 1D**). As expected, no DT- and TT-specific antibodies were detected 1 day before the first immunization (data not shown). At day 20 after immunization, approximately 50% of the immunized mice showed detectable IgG titers against DT, which were increased after the first boost measured at day 41. After the second boost (measured at day 47), all mice produced antibodies against DT (**Figures 5A–C**). IgG titers against TT were slightly higher compared with DT-specific titers (**Figures 5D–F**). When comparing ID administration with SC injection, no

statistical differences were found for DT-specific titers. IgG titers against TT were slightly higher upon SC delivery as compared with microneedle-mediated delivery (significant after the boost and second boost) (**Figures 5D–F**).

For both immunization routes, also DT/TT combinations adjuvanted with toll-like receptor (TLR) 7 agonist IMQ were tested to determine whether adjuvants may modify the quality of vaccination-induced antibody responses. Antigen quantities loaded in combination with IMQ were half the antigen dose loaded into npMNAs in the absence of this TLR7 agonist (0.6 Lf DT, 0.75 Lf TT, and 0.5 µg IMQ). Despite the lower antigen dose, similar antibody titers were detected in mice immunized with adjuvanted compared with unadjuvanted vaccine, with the exception of DT-specific responses measured at day 41 where 3/10 in the adjuvanted group compared with 1/10 mice in the unadjuvanted group had failed to respond (**Figure 5**). Overall, observing TT-specific antibody titers, the subcutaneously injected mice showed slightly higher titers than the intradermally immunized mice.

To determine whether IMQ skews the induced DT- and TT-specific response, relative quantities of the IgG isotypes IgG2a and IgG1, which serve as markers for T helper 1 and T helper 2 type lymphocytes, respectively, were determined after boost immunization. Relative quantities could not be calculated after prime immunization because of non-responders. Overall, ratios between DT and TT-specific IgG1:IgG2a, after first and second boost immunization, indicated that IgG1, and thus Th2 cell responses, prevailed (**Figure 6**; Figures S1B,C and S2B,C in Supplementary Material). Remarkably, addition of IMQ enhanced vaccine-induced DT-specific IgG1 responses (day 41) in SC but not in ID immunized mice. After the second booster immunization (day 47) these differences disappeared, and similar ratios between DT-specific IgG1:IgG2a were observed in all four immunization groups (**Figure 6**; Figures S1B,C and S2B,C in Supplementary Material). For TT specific isotypes, the kinetics were different. While after the first boost immunization, in all mouse groups, ratios between IgG1 and IgG2a were similar, both ID immunized mice receiving the adjuvanted vaccine and SC immunized mice receiving the unadjuvanted vaccine showed increased IgG1:IgG2a ratios after boost (**Figure 6**; Figures S1E,F and S2E,F in Supplementary Material). Taken together, although no major differences in IgG isotype ratios between groups were observed, these data demonstrate that followed vaccination regimen induces a predominantly Th2-skewed lymphocyte response.

### DISCUSSION

Nowadays, many microneedle technologies are under investigation for their potential future application in ID immunization, because they can be used to deliver drugs and vaccines in a minimally invasive and potentially pain-free manner into the skin. In the landscape of microneedle technologies, nanoporous microneedles are relatively new and pose an immunization method that enables the loading of drug formulations into the pores of the MNAs, which are released *via* diffusion upon piercing of the microneedles into the skin. In this first immunological study, we show that ceramic alumina npMNAs can be used for ID immunization with subunit vaccines, aimed to elicit humoral responses.

Nanoporous microneedle array strength is an important characteristic contributing to the efficiency of skin piercing but is closely related to the porosity of the material (15). High porosity can weaken the material resulting in breakage of the microneedle tips and thereby resulting in less efficient piercing of the skin. When designing npMNAs for the delivery of proteins or subunit vaccines, larger pores are necessary. The use of alumina (Al2O3) AKP30 particles for the production of npMNAs results in approximately 40% porosity, with an average pore size of 80 nm (13) and such npMNAs have sufficient strength to repeatedly penetrate the skin without breaking (10). Furthermore, it was previously shown that nanoporous microneedles can be loaded with small molecules and nanoparticles with sizes up to 100 nm (10). Alumina npMNAs, although with different microneedle geometry, but the same pore size distribution, have previously been loaded with small molecules (10) and short peptides (17). However, peptides are in general not ideal for prophylactic vaccination, because they contain only one minimal T or B cell epitope. Here, we show for the first time successful loading of npMNAs with the subunit vaccine proteins DT and TT, and the subsequent release of these antigens from npMNAs into skin, which potentially give rise to various epitopes.

Nanoporous microneedle arrays have an interconnected porous structured network throughout both the backplate reservoir and the microneedles (21), which allows for the loading of drug formulations in both the microneedles as well as the backplate reservoir. While loading relatively high amounts of drug formulations into the npMNAs (including backplate) can have advantages, for vaccines it might be disadvantageous. The diffusion of vaccines, or other biomacromolecules, from the backplate *via* the tips into the skin is a time consuming process and might therefore lead to a less efficient delivery efficiency of expensive molecules. In this study, we successfully developed a procedure to load only the microneedles tips of the npMNAs, post production of the npMNA, resulting in a relatively efficient use of vaccine formulations and allowing for limited application time on the skin. In addition, absorption of a formulation (by porous microneedles) could be favorable over adsorption onto (or coating of) the surface of, for instance, solid microneedles, because (1) the microneedle tip sharpness is retained and (2) potentially less excipients are required to retain the immunogenicity of a vaccine. Coating of solid MNAs generally requires a thick drug-containing layer to achieve the required amounts of drug/vaccine loading, and this could reduce the sharpness of the tips and thereby their skin piercing ability (22, 23). Furthermore, several excipients are required to adsorb the coating onto the microneedle surface and retain the immunogenicity of vaccines (24, 25). On the other hand, dissolving microneedles, wherein the drug/vaccine is embedded in the microneedle matrix require a more complex loading strategy in which the vaccine and excipients are added during the preparation phase. Hence, using dissolving microneedle technologies to only load the microneedle tips with a drug/vaccine is more challenging as compared with using porous microneedles. Therefore, npMNAs can be advantageous over coated and dissolving MNAs.

The release of DT, TT, and IMQ from drug-loaded npMNAs, was determined after incubating the drug-loaded npMNAs in release buffer *in vitro*. Indeed, around 30% of the vaccine antigens that were loaded into the npMNAs were released in buffer after 30 min. To establish if the vaccine subunit antigens were also released from the npMNAs after they penetrated the skin, and to quantify the amount of antigen delivered, *ex vivo* mouse ears were pierced with fluorescently labeled antigen-loaded npMNAs. The amount of delivered antigen was quantified, from which the release and delivery efficiencies were calculated. The ID delivery efficiency of DT was around 20%, and about 50% for TT. When the amounts of antigens that were released into the skin are correlated to the human vaccination dose (26, 27), the estimated diameter of the circular npMNA that needs to be used for human application should be 2.3 and 2.5 cm for DT and TT, respectively. This indicates a feasible size to deliver the corresponding vaccine doses.

Because our ceramic npMNAs have an average pore size of 80 nm, preferably low-molecular weight adjuvants are coformulated into the nanopores of the npMNAs. For example, alum is a potent (micrometer-sized) adjuvant for the induction of humoral immune responses, but it cannot fit into the nanopores of our npMNAs. Besides, alum causes granuloma formation and should therefore not be used as an ID adjuvant (28).

In this study, IMQ, which is a toll-like receptor 7 agonist with a molecular weight of 240 g/mol, was chosen as an adjuvant. IMQ is extensively researched for its adjuvanticity (19, 29, 30), and as TLR agonist holds promise for vaccination approaches, because it induces the release of pro-inflammatory cytokines (31, 32). The cytokine profile induced by IMQ specifically favors Th1 over Th2 type responses (33, 34), and thereby the induction of a cellular immune response. Furthermore, it has been shown that topical application, rather than ID injection, activated antigen-presenting cells in skin explants (35). IMQ 5% cream (Aldara) already has FDA approval for topical use for the treatment of warts, actinic keratinosis and superficial basal cell carcinoma. Together, this makes IMQ a potent and attractive adjuvant for ID immunization. In the adjuvanted vaccine formulation, half of the antigen dose was used to investigate whether the vaccine dose could be decreased using npMNAs with an adjuvant that fits into the nanopores of the npMNAs. Since no enhanced immunogenicity was observed using IMQ, we cannot make any statement about the adjuvanticity of IMQ using our npMNAs in combination with DT and TT. Furthermore, we found that, in most immunization groups, IMQ did not have a significant effect on ratios of DT- and TT-specific IgG1:IgG2 responses, i.e., IgG1 responses prevailed in all groups. Only in mice receiving a triple immunization using npMNAs, enhanced TT-specific IgG1:IgG2a ratios were observed in the IMQ-adjuvanted vaccine groups. Taken together, these findings suggest that npMNA-mediated ID immunization with IMQ-adjuvanted vaccine predominantly induces Th2, and not Th1 responses. When further exploring the immunological potential of ceramic npMNAs and selecting future adjuvants, one should consider the limited pore size of the npMNAs. Therefore, in future studies aimed to optimize microneedle-based intradermally delivered vaccines, we will focus on use of low-molecular weight adjuvants, such as cGAMP (36), for co-formulation into the nanopores of npNMAs.

Finally, in this study, it was shown that strong antibody responses were induced without using an adjuvant. The antibody

### REFERENCES


responses obtained in our study are in line with the ones described in literature (37, 38). For example, in previous studies doses of 0.3 µg [unadjuvanted DT (37)] and 0.1 µg [unadjuvanted DT and TT (38)] have been used for the induction of antibody responses (in mice and rats, respectively), which resulted in antibodyspecific log titers of 4–5. In our study, similar antigen-specific antibody titers were obtained (log titers of approximately 4) against DT and TT using a similar dose. Therefore, this study shows the potential of npMNAs for microneedle-based ID vaccination using subunit vaccines.

### CONCLUSION

Taken together, in this study we show that ceramic nanoporous microneedles are strong enough to repeatedly penetrate the skin and that they can be loaded with protein subunit vaccines such as DT and TT. After skin piercing with antigen-loaded npMNAs, the antigens are intradermally delivered, which resulted in an induction of antigen-specific antibody responses. In conclusion, we show for the first time the potential of npMNAs for ID immunization with subunit vaccines, which opens possibilities for future ID vaccination designs.

### ETHICS STATEMENT

Ethical approval was obtained from Animal Ethics Committee from Utrecht University, the Netherlands (DEC #2014.II.11.080).

### AUTHOR CONTRIBUTIONS

Substantial contributions to the conception was done by KM, PV, and AS and design of the work by KM, AS, and AP. Acquisition, analysis, and interpretation of data by AP, KM, AG, NK, and PK. Drafting the manuscript by AG and KM. All the authors revised the work and gave final approval of the version to be published.

### FUNDING

This work was financially supported by the European Union's Seventh Framework Program—Grant No. 280873 ADITEC acronym: NaPoVacs to AS and KM.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01789/ full#supplementary-material.


**Conflict of Interest Statement:** KM is co-owner of uPRAX Microsolutions. The other authors declare that they have no commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 de Groot, Platteel, Kuijt, van Kooten, Vos, Sijts and van der Maaden. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*