# ADVANCES AND CHALLENGES IN NANOMEDICINE

EDITED BY : Susan Hua and Sherry Y. Wu PUBLISHED IN : Frontiers in Pharmacology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-745-8 DOI 10.3389/978-2-88945-745-8

#### 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

# ADVANCES AND CHALLENGES IN NANOMEDICINE

Topic Editors: Susan Hua, University of Newcastle, Australia; Hunter Medical Research Institute, Australia Sherry Y. Wu, University of Queensland, Australia

Image: Volodymyr Horbovyy/Shutterstock.com

Nanotechnology is a multidisciplinary field that is revolutionizing the way we detect and treat damage to the human body. Nanomedicine applies nanotechnology to highly specific medical interventions for the prevention, diagnosis, and treatment of diseases. They are increasingly being used to overcome biological barriers in the body to improve the way we deliver compounds to specific tissues and organs. In particular, nanomedicines have been shown to be beneficial for stabilizing therapeutic compounds, overcoming obstacles to cellular and tissue uptake, and improving biodistribution of compounds to target sites in vivo. Nanomedicines have demonstrated significant therapeutic advantages for a multitude of biomedical applications, however the clinical translation of these nanotechnology platforms has not progressed as quickly as the plethora of positive results would have suggested. Understanding the advances in nanomedicine to date and the challenges that still need to be overcome, will allow future research to improve on existing platforms and to address the current translational and regulatory limitations.

This eBook "Advances and Challenges in Nanomedicine" has brought together experts in the fields of nanomedicine, nanotechnology, nanotoxicology, pharmaceutics, manufacturing, and translation to discuss the application of nanotechnology to drug delivery. This information is presented as original research, opinion, perspective, and review articles. The goal of this eBook is to generate collaborative discussion on the current status, general trends, challenges, strategies, and future direction of pharmaceutical nanotechnology, as well as highlight current and emerging nanoparticulate platforms with potential medical applications.

Citation: Hua, S., Wu, S. Y., eds. (2019). Advances and Challenges in Nanomedicine. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-745-8

# Table of Contents

*05 Editorial: Advances and Challenges in Nanomedicine* Susan Hua and Sherry Y. Wu *08 Advantages and Limitations of Current Techniques for Analyzing the Biodistribution of Nanoparticles* Lauren Arms, Doug W. Smith, Jamie Flynn, William Palmer, Antony Martin, Ameha Woldu and Susan Hua *25 Advantages and Limitations of Current Imaging Techniques for Characterizing Liposome Morphology* Annie-Louise Robson, Paul C. Dastoor, Jamie Flynn, William Palmer, Antony Martin, Doug W. Smith, Ameha Woldu and Susan Hua *33 Risk Assessment and Risk Minimization in Nanomedicine: A Need for Predictive, Alternative, and 3Rs Strategies* Lisa Accomasso, Caterina Cristallini and Claudia Giachino *40 Current Trends and Challenges in the Clinical Translation of Nanoparticulate Nanomedicines: Pathways for Translational Development and Commercialization* Susan Hua, Maria B. C. de Matos, Josbert M. Metselaar and Gert Storm *54 The Advances and Challenges in Utilizing Exosomes for Delivering Cancer Therapeutics* Mengliu Yang and Sherry Y. Wu *58 Biologically Targeted Magnetic Hyperthermia: Potential and Limitations* David Chang, May Lim, Jeroen A. C. M. Goos, Ruirui Qiao, Yun Yee Ng, Friederike M. Mansfeld, Michael Jackson, Thomas P. Davis and Maria Kavallaris *78 Nano-BCG: A Promising Delivery System for Treatment of Human Bladder Cancer*

Julieti Huch Buss, Karine Rech Begnini, Camila Bonemann Bender, Adriana R. Pohlmann, Silvia S. Guterres, Tiago Collares and Fabiana Kömmling Seixas

*87 Modulating the Tumor Microenvironment to Enhance Tumor Nanomedicine Delivery*

Bo Zhang, Yu Hu and Zhiqing Pang

*103 Targeting Accessories to the Crime: Nanoparticle Nucleic Acid Delivery to the Tumor Microenvironment*

Emily B. Harrison, Salma H. Azam and Chad V. Pecot


Shiyu Yan, Kewei Xu, Li Li, Wenyi Gu, Barbara E. Rolfe and Zhi P. Xu

# Editorial: Advances and Challenges in Nanomedicine

Susan Hua1,2 \* and Sherry Y. Wu<sup>3</sup>

*<sup>1</sup> Therapeutic Targeting Research Group, School of Biomedical Sciences and Pharmacy, University of Newcastle, Callaghan, NSW, Australia, <sup>2</sup> Hunter Medical Research Institute, New Lambton Heights, NSW, Australia, <sup>3</sup> School of Biomedical Sciences, University of Queensland, Brisbane, QLD, Australia*

Keywords: nanomedicine, nanotechnology, nanoparticles, targeted drug delivery, translation

**Editorial on the Research Topic**

#### **Advances and Challenges in Nanomedicine**

The use of nanotechnology in medicine has the potential to have a significant impact on human health by improving the diagnosis, prevention and treatment of diseases. Nanomedicines typically encapsulate therapeutic and/or imaging compounds in submicrometer-sized carrier materials. In the last several decades, the application of nanomedicine for clinical purposes has received significant attention from academia, researchers, government, funding agencies, and regulatory bodies (Allen and Cullis, 2004; Sercombe et al., 2015; Hare et al., 2017; Hua et al.). Nanomedicines are generally intended to increase the therapeutic index of compounds by allowing more efficient delivery to the target site to enhance therapeutic efficacy and/or by minimizing accumulation in healthy body sites to reduce toxicity. Nanoencapsulation can also protect therapeutics from degradation in biological environments and can provide solubilization (Talekar et al., 2015; Mishra et al., 2017; Shajari et al., 2017). This e-Book focuses on articles that discuss the advances and challenges in the nanomedicine field across a broad range of topics. A brief summary of each article is provided below.

Nanomedicines can have a combination of chemical, physical, and biological properties that influences their in vivo behavior.

	- In addition, Robson et al. discussed the advantages and limitations of available imaging techniques used to evaluate the morphology of liposomal formulations (Robson et al.). Validating and controlling the morphology of nanoparticles is important for clinical translation, however it is generally difficult to control and not well-characterized (Robson et al.).

There are a number of factors that can impose significant obstacles to the clinical translation of nanomedicines, irrespective of whether they are therapeutically beneficial or not.

• Accomasso et al. summarized the current state regarding the safety evaluation of nano-based therapeutics and discussed the importance of risk assessment and risk minimization in the development of nanomedicines (Accomasso et al.). With the rapid growth in the use of nanomaterials for medical applications, the most urgent need is developing and validating novel and practical approaches that are able to determine potential short-term and long-term health risks, including the extrapolation of acute in vitro results for the prediction of chronic in vivo effects (Accomasso et al.).

#### Edited and reviewed by:

*Dominique J. Dubois, Free University of Brussels, Belgium*

> \*Correspondence: *Susan Hua susan.hua@newcastle.edu.au*

#### Specialty section:

*This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology*

Received: *28 October 2018* Accepted: *13 November 2018* Published: *29 November 2018*

#### Citation:

*Hua S and Wu SY (2018) Editorial: Advances and Challenges in Nanomedicine. Front. Pharmacol. 9:1397. doi: 10.3389/fphar.2018.01397*

**5**

• More broadly, Hua et al. discussed the current trends and challenges in the clinical translation of nanoparticulate nanomedicines, as well as the potential pathways for translational development and commercialization (Hua et al.). Key issues related to the clinical development of nanoparticulate nanomedicines include biological challenges, biocompatibility and safety, large scale manufacturing, government regulations, intellectual property (IP), and overall cost-effectiveness in comparison to current therapies (Hua et al.).

Undoubtedly, the vast majority of nanomedicines in preclinical and clinical development as well as in clinical use are for targeting a wide variety of malignant tumors.


Modulation of the tumor microenvironment has recently emerged as an important strategy to improve the delivery of nanomedicines to tumors, given the importance of cancerassociated cells in tumor growth and metastasis.


development of various nanoparticle platforms to overcome nucleic acid delivery hurdles, several challenges still exist for effective tumor delivery (Harrison et al.). One such challenge has been the accumulation of nanoparticles in non-cancer cells within the tumor microenvironment, which has recently opened up novel therapeutic applications for nanoparticles (Harrison et al.).

The application of nanomedicine-based therapies for drug targeting to non-cancer conditions has also increased in recent years.


## SUMMARY

Overall, nanomedicine has the potential to revolutionize the way we detect and treat damage to the human body. Although nanomedicines have demonstrated significant therapeutic advantages for a multitude of medical applications, their translation has not progressed as rapidly as the plethora of positive preclinical results would have suggested (Luxenhofer et al., 2014; Sercombe et al., 2015; Hare et al., 2017; Hua et al.). The experimental development of nanomedicines is continually progressing at a fast pace, however significant challenges still exist in promoting these platforms into clinically feasible therapies. Therefore, continued translational success will require communication and collaboration between experts involved in all stages of pharmaceutical development of nanotechnologies, including pharmaceutical design and manufacturing, cellular interactions and toxicology, as well as preclinical and clinical evaluation (Hua et al.).

### AUTHOR CONTRIBUTIONS

SH drafted the manuscript. SH and SW critically reviewed the manuscript for important intellectual content.

### REFERENCES


### ACKNOWLEDGMENTS

We wish to thank The Pharmacy Research Trust of New South Wales, The Rebecca L. Cooper Medical Research Foundation, Gladys M. Brawn Fellowship, ausEE Research Foundation, and The University of Newcastle for providing financial support for our research.


**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 Hua and Wu. 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.

# Advantages and Limitations of Current Techniques for Analyzing the Biodistribution of Nanoparticles

Lauren Arms <sup>1</sup> , Doug W. Smith<sup>1</sup> , Jamie Flynn1,2, William Palmer 2,3, Antony Martin2,3 , Ameha Woldu1,2 and Susan Hua1,2 \*

*<sup>1</sup> School of Biomedical Sciences and Pharmacy, University of Newcastle, Callaghan, NSW, Australia, <sup>2</sup> Hunter Medical Research Institute, New Lambton Heights, NSW, Australia, <sup>3</sup> School of Environmental and Life Sciences, University of Newcastle, Callaghan, NSW, Australia*

#### Edited by:

*Xinhua Qu, Shanghai Ninth People's Hospital, Shanghai Jiao-Tong University School of Medicine, China*

#### Reviewed by:

*Christoph Eugen Hagemeyer, Monash University, Australia Bianca Christine Bernardo, Baker Heart and Diabetes Institute, Australia*

> \*Correspondence: *Susan Hua susan.hua@newcastle.edu.au*

#### Specialty section:

*This article was submitted to Translational Pharmacology, a section of the journal Frontiers in Pharmacology*

Received: *18 May 2018* Accepted: *03 July 2018* Published: *14 August 2018*

#### Citation:

*Arms L, Smith DW, Flynn J, Palmer W, Martin A, Woldu A and Hua S (2018) Advantages and Limitations of Current Techniques for Analyzing the Biodistribution of Nanoparticles. Front. Pharmacol. 9:802. doi: 10.3389/fphar.2018.00802* Nanomedicines are typically submicrometer-sized carrier materials (nanoparticles) encapsulating therapeutic and/or imaging compounds that are used for the prevention, diagnosis and treatment of diseases. They are increasingly being used to overcome biological barriers in the body to improve the way we deliver compounds to specific tissues and organs. Nanomedicine technology aims to improve the balance between the efficacy and the toxicity of therapeutic compounds. Nanoparticles, one of the key technologies of nanomedicine, can exhibit a combination of physical, chemical and biological characteristics that determine their *in vivo* behavior. A key component in the translational assessment of nanomedicines is determining the biodistribution of the nanoparticles following *in vivo* administration in animals and humans. There are a range of techniques available for evaluating nanoparticle biodistribution, including histology, electron microscopy, liquid scintillation counting (LSC), indirectly measuring drug concentrations, *in vivo* optical imaging, computed tomography (CT), magnetic resonance imaging (MRI), and nuclear medicine imaging. Each technique has its own advantages and limitations, as well as capabilities for assessing real-time, whole-organ and cellular accumulation. This review will address the principles and methodology of each technique and their advantages and limitations for evaluating *in vivo* biodistribution of nanoparticles.

Keywords: nanoparticles, nanomedicine, biodistribution, in vivo, imaging, techniques, advantages, limitations

### INTRODUCTION

Nanomedicine is the application of nanotechnology for the diagnosis, prevention and treatment of diseases. Nanomedicines are submicrometer-sized carrier materials (nanoparticles) designed to improve the biodistribution of encapsulated compounds by delivering them more effectively and more selectively to the pathological site (site-specific drug delivery) and/or by guiding them away from potentially endangered healthy tissues (site-avoidance drug delivery) (Lammers et al., 2012). This technology aims to improve the balance between the efficacy and the toxicity of therapeutic compounds (Lammers et al., 2012). Nanoparticles can exhibit a combination of physical (e.g., size, shape, lamellarity and homogeneity), chemical (e.g., composition, surface charge, surface coating and phase transition temperature), and biological (e.g., encapsulated compounds and conjugated surface ligands) characteristics that determine their in vivo behavior (Bharali and Mousa, 2010; Robson et al., 2018).

Despite the significant advances in drug delivery technologies and platforms in the last several decades, the clinical translation of nanomedicines has progressed incrementally (Sercombe et al., 2015; Hare et al., 2017). It has been suggested that effective nanomedicine development requires a diseasedriven approach, rather than the traditional formulation-driven approach where drug delivery system engineering has been the priority (Hare et al., 2017). This requires a strong understanding of the relationships between biology and technology, including the influence of disease pathophysiology on nanomedicine accumulation, distribution, retention and efficacy, and the correlation between delivery system properties and in vivo behavior in animals vs. humans (Hare et al., 2017).

A key component in the translational assessment of nanomedicines is determining the biodistribution of the nanoparticles following in vivo administration in animals and humans (Kunjachan et al., 2015). There is a range of techniques available for evaluating nanoparticle biodistribution, including histology, electron microscopy, liquid scintillation counting (LSC), indirectly measuring drug concentrations, in vivo optical imaging, computed tomography (CT), magnetic resonance imaging (MRI), and nuclear medicine imaging. Each technique has its own advantages and limitations, as well as capabilities for assessing real-time, whole-organ and cellular accumulation (**Figure 1**). This review will address the principles and methodology of each technique and their advantages and limitations for evaluating in vivo biodistribution of nanoparticles.

### HISTOLOGY

Microscopic visualization of nanoparticles in tissue sections is one of the common techniques used to assess biodistribution following in vivo administration in animals. This technique relies on histological processing of tissues to examine the association of nanoparticles with the cellular microenvironment under a microscope—typically light and fluorescence microscopy. In order to assess nanoparticle biodistribution, the organs or tissues of interest are harvested at set time points following in vivo administration in animals and undergo either conventional paraffin processing or cryostat processing of frozen or fixed tissues. The choice between the methods depends on the composition of the nanomedicine, as paraffin processing involves the samples being dehydrated, cleared (also called delipidation) and infiltrated. More specifically, water is removed from the specimen in successive stages using increasing concentrations of alcohol. A clearing agent, such as xylene, is used in the last phase to remove the alcohol and tissue lipids in the specimen, thereby allowing infiltration of an embedding agent (e.g., paraffin wax or cryogenic media) (Alturkistani et al., 2016). As this process involves the use of lipid soluble solvents, nanoparticles composed of materials that are easily degraded by these solvents (e.g., liposomes, solid lipid nanoparticles, and micelles) should only undergo cryostat processing of frozen sections.

For paraffin processing, tissues of interest are immediately fixed with chemicals (commonly formalin) to preserve structural integrity and prevent cell degradation (autolysis and putrefaction), prior to embedding and sectioning using a microtome. This process, however, can damage proteins in the tissue and can also denature them to a certain extent, which may include protein or peptide-based ligands conjugated to the surface of nanoparticles (Titford, 2009). On the other hand, frozen tissues may be fixed or unfixed and are sectioned using a microtome mounted in a refrigeration device known as a cryostat. Sections are then mounted on a glass slide and can be stained to enhance visualization of the nanoparticles and/or microscopic anatomy of cells and tissues. A variety of histological stains can be used to differentiate between biological structures, with hematoxylin and eosin (H&E) being the most common (Titford, 2009; Alturkistani et al., 2016). H&E staining provides excellent visualization of nuclei (stained purple) and cytoplasmic details (stained pink) within cells (Titford, 2009; Alturkistani et al., 2016). It is important to choose a stain to label biological structures that will not affect the nanoparticles themselves. Immunohistochemistry or co-labeling can also be used to visualize other aspects of the tissues, such as blood vessels, to allow appropriate orientation and evaluation of the cellular biodistribution of nanoparticles in tissue sections.

Conventional histopathology staining methods can be used to detect the biodistribution of certain types of nanoparticles. In particular, clusters of nanoparticles >200 nm in size can be visualized by light microscopy in tissue sections based on resolution limitations (Ostrowski et al., 2015; Robson et al., 2018). For example, the biodistribution of ultrasmall and small superparamagnetic iron oxide (USPIO and SPIO, respectively) nanoparticles that were injected intraperitoneally into C57BL/6 mice were studied histologically by measuring iron-positive areas (µm<sup>2</sup> ) in representative paraffin-embedded tissue sections of organs stained with Prussian blue (Tsuchiya et al., 2011; Pham et al., 2018). Similarly, cationic stains such as Alcian blue have been used to stain the negatively charged sulfate groups embedded within organic dendritic polyglycerol sulfate (dPGS) nanoparticles. Holzhausen et al were able to demonstrate specific localization of dPGS nanoparticles in hepatic Kupffer cells following intravenous injection in mice using Alcian blue in standard histopathological tissue sections (Holzhausen et al., 2013). Single-walled carbon nanotubes (SWCNT) labeled with colloidal gold have also been visualized as dark deposits on cryostat tissue sections using silver enhancement (Mercer et al., 2008). In addition, the biodistribution of a range of nanoparticles labeled with fluorescent dyes have been visualized in tissues sections using fluorescence microscopy, including silica nanoparticles (Cho et al., 2009) and polymeric micelles (Asem et al., 2016).

In terms of advantages, histology is a relatively cost-effective technique for assessing nanoparticle biodistribution and allows for the study of large tissue sections (**Table 1**). In comparison to other available techniques, histology can be used to study the specific accumulation and association of nanoparticles within a

cellular context. This technique also does not require exposure to ionizing radiation or contrast agents. However, histology is generally considered a qualitative method when assessing nanoparticle biodistribution and a number of limitations should be considered when approaching this technique. Light and fluorescence microscopy have generally low resolution compared to other microscopy techniques and are unable to image individual nanoparticles in the lower nanometer range, especially in tissues (Robson et al., 2018). In addition, a limited number of tissue sections (5–50µm thickness) are typically chosen to evaluate and approximate biodistribution in each organ, simply due to the sheer number of tissue sections that can be attained from each organ. This may affect the results as not all sections are examined; therefore, appropriate sampling methods should be utilized to provide a more reliable representation of nanoparticle biodistribution in the whole organ. Histology is also a time-consuming and laborious technique. Although cryostat sectioning may be faster to prepare than paraffinembedded tissue sections, the freezing process may negatively affect tissue structures and resolution, especially when using light microscopy. Furthermore, histology is susceptible to human error during slide preparation and analysis, and the identification of specific cell types can be difficult. In particular, the detection of organic nanoparticles in tissue sections often poses a particular challenge due to their closer similarities with biomolecules (Holzhausen et al., 2013). With regards to fluorescence imaging, the labeling of nanoparticles with fluorescent dyes may affect their physicochemical properties and subsequent in vivo behavior (Robson et al., 2018). Photobleaching of fluorescent dyes is another concern (Robson et al., 2018), especially when the fluorescent-labeled nanoparticles are likely to have some exposure to light during the study from in vivo administration to tissue harvesting and processing. This can result in a diminished fluorescent signal. Tissue autofluorescence is also a significant issue that needs to be addressed with appropriate control groups if using this technique. Autofluorescence occurs in most tissues and leads to a reduction in the signal detection sensitivity, which interferes with the accuracy of the results (Koo et al., 2006).

#### ELECTRON MICROSCOPY

Electron microscopy analysis of tissue samples can provide detailed information of the biodistribution of nanoparticles under very high magnification (Mayhew et al., 2009). This technique uses a beam of electrons focused onto the surface of the sample by various electromagnetic lenses. The electrons are scattered by the sample and are then refocused and magnified by a further series of electromagnetic lenses in the imaging column to produce a projected image (Mayhew et al., 2009; Robson et al., 2018). There are a number of different types of electron microscopes, with transmission electron microscopy (TEM), scanning electron microscopy (SEM) and variations of the two techniques having been utilized for this application. In comparison to TEM, in which the electron beam crosses the sample where it is then focused by the objective lens to form an image, SEM utilizes an electron beam that is scanned across or over a sample (rather than through a sample) and imaging is performed by mapping signal intensity synchronously with the scan to produce a magnified image of an object (Garcia-Negrete et al., 2015; Robson et al., 2018). Typically, tissue samples are fixed with chemicals (commonly formalin) and then undergo dehydration with serial alcohol and propylene oxide, prior to embedment in embedding resin (e.g., glycidether 100, EPON 812, Embed 812).

Electron microscopy has predominantly been used to determine the cellular association of nanoparticles in vitro (Schrand et al., 2010; Plascencia-Villa et al., 2012; Brown and Hondow, 2013; Goldstein et al., 2014), with only limited studies using this technique to evaluate nanoparticle biodistribution following in vivo administration (Muhlfeld et al., 2007; Mayhew et al., 2009; Jong et al., 2010; Kempen et al., 2013; Garcia-Negrete et al., 2015). For example, Jong et al. (2010) evaluated

#### TABLE 1 | Summary of current techniques for analyzing the biodistribution of nanoparticles.


*(Continued)*

#### TABLE 1 | Continued


• PET typically requires a cyclotron or generator

the biodistribution of gold nanoparticles (10 and 250 nm) 24 h post-intravenous injection in rats using TEM. Ultra-thin sections of 50–70 nm were prepared and stained by uranyl acetate and lead citrate. Results showed that 10 nm gold nanoparticles were present in the phagocytic cells of the reticuloendothelial system (RES), whereas 250 nm gold nanoparticles were unable to be detected in any of the organs investigated. This was likely due to the very low number of 250 nm particles that would be theoretically present in one TEM tissue section. In addition, several globular structures of approximately the expected size were found in liver cells and the endothelium of blood vessels in the brain; however, elemental analysis with energy dispersive X-ray (EDX) showed that these structures did not contain gold. This indicates that in vivo identification of nanoparticles cannot only depend on the detection of nanosized structures in cells.

More recent studies have used scanning transmission electron microscopy (STEM), which combines the principles of TEM and SEM (Kempen et al., 2013; Garcia-Negrete et al., 2015). STEM requires very thin samples (similar to TEM) and involves scanning a very finely focused beam of electrons across the sample in a raster pattern. This technique allows the use of other signals that cannot be spatially correlated in TEM (e.g., secondary electrons, scattered beam electrons, characteristic X-rays, and electron energy loss) and has improved spatial resolution compared to SEM (Kempen et al., 2013; Garcia-Negrete et al., 2015). For example, Kempen et al. (2013) used STEM to analyze the accumulation and distribution of polyethylene glycol coated Raman-active-silica-gold-nanoparticles (PEG-R-Si-Au-NPs) in the liver of intrarectally administered or tail-vein injected mice. Tissue samples were trimmed to <1 mm<sup>3</sup> and prepared and stained by osmium tetroxide and uranyl acetate. Sections (150 nm thick) were cut from the block face using an ultramicrotome and then placed on a copper grid. This approach utilizes the simultaneous bright and dark field imaging capabilities of STEM to readily identify PEG-R-Si-Au-NPs in mouse liver tissue. Results showed that nanoparticles injected intravenously accumulated in the liver while those administered intrarectally did not, indicating that they remain in the colon and do not pass through the colon wall into the systemic circulation.

Overall, the main advantage with electron microscopy is the high resolution, which allows visualization of the accumulation of nanoparticles in cells and the localization of nanoparticles in cellular organelles (Jong et al., 2010). Although this technique is generally considered a semi-quantitative method, a number of limitations should be considered when approaching this technique for evaluating nanoparticle biodistribution. Electron microscopy is a more expensive technique and is not capable of evaluating large tissue sections compared to standard histology (**Table 1**). For example, the analysis volume for TEM is generally low at 1–10 µm<sup>3</sup> for a single TEM session (Kempen et al., 2013). In addition, a limited number of ultra-thin tissue sections (50–150 nm thickness) are typically chosen to evaluate and approximate biodistribution in each organ, which may affect the results as not all sections are examined. Therefore, appropriate sampling methods and additional analytical methods should be utilized to provide a more reliable representation of nanoparticle biodistribution in the whole organ. Electron microscopy is also a time-consuming technique, with individual samples usually taking >3–4 h to analyze (Kempen et al., 2013). Relatively high numbers of nanoparticles need to be administered to enable the detection of nanoparticles in organs by electron microscopy, especially for larger nanoparticles (Jong et al., 2010; Kempen et al., 2013). An additional identification technique (e.g., EDX detection of the composing elements or a specific marker for the administered nanoparticles) may also be necessary for a positive identification of the nanomaterial in tissues and cells (Jong et al., 2010). Although electron microscopy can readily image soft matter samples, characterization of soft materials can be affected by the high-voltage electron beams (Garcia-Negrete et al., 2015). Therefore, artifacts need to be carefully protected against when an image is acquired, as burn-in spots can form on the image (Kempen et al., 2013). With regards to radiation exposure, low levels of X-rays can be produced from the backscattered electrons impinging on samples in electron microscopes. However, these units are well-shielded and any X-rays generated internally should not penetrate outside the unit. Finally, the sample preparation method will not be suitable for all nanoparticles. As this process involves the use of lipid soluble solvents, nanoparticles composed of materials that are easily degraded by these solvents (e.g., liposomes, solid lipid nanoparticles and micelles) should use other biodistribution techniques.

### LIQUID SCINTILLATION COUNTING (LSC)

Liquid scintillation counting (LSC) is a standard laboratory method to quantify the radioactivity of low energy radioisotopes—most commonly beta-emitting (β-emitting) and alpha-emitting (α-emitting) isotopes (Shigematsu et al., 1995; PerkinElmer, 2008). LSC analysis of samples requires a specific cocktail containing the aromatic organic solvent and scintillators (also referred to as "fluors") to absorb the radioisotopic energy and produce detectable light pulses, respectively. The basic principles of LSC rely on the energy released from a radioactive decay (emitting beta or alpha particles) to excite the aromatic solvent molecules. The energy of the solvent molecules is then transferred to the scintillator molecules to produce excited states of the electrons, which decay to the ground state and produce a light pulse that is characteristic for the scintillator. The emitted light is detected by the photomultiplier tube (PMT) of the liquid scintillation counter.

In vivo biodistribution of nanomedicines can be assessed by labeling nanoparticles with isotopic markers prior to administration in animals. For example, anti-ICAM-1 immunoliposomes and control liposomes were radiolabeled with [3H]-CHE and administered intravenously in rats with Complete Freund's Adjuvant-induced inflammation of the paw (Hua and Cabot, 2013). The use of [3H]-CHE is convenient for these studies because it is a stable, non-exchangeable, and non-degradable marker of liposomes, thus providing an estimate of the cumulative liposome dose in tissues (Hua and Cabot, 2013). Organs are then harvested and prepared for LSC. Depending on the sample type, the biological material can either be directly mixed to the cocktail with no or little pre-treatment, or a treatment/solubilization may be needed prior to scintillation cocktail addition (PerkinElmer, 2008). The latter is generally required when analyzing biological tissues and usually a portion of an organ is weighed and processed, due to the time taken for effective solubilization of larger tissues. Radioactivity is measured in terms of number of disintegrations per minute (DPM) of the isotope in each tissue sample. The amount of radioactivity can then be expressed as the number of becquerel (Bq) per gram of tissue using the following conversion: 1 Bq = 60 DPM. The becquerel is the SI derived unit of radioactivity.

LSC has the advantages of being a sensitive, specific, and quantitative technique for measuring nanomedicine biodistribution (**Table 1**). Removal of excess free isotopic markers that have not been incorporated into the nanoparticles is important prior to in vivo assessment. It should be noted that the quality and reproducibility of the data will depend on the choice of the cocktail as well as on the sample composition, volume, temperature, and counting device (PerkinElmer, 2008). This technique can be laborious, especially with the need to treat and solubilize the harvested tissues prior to LSC analysis. If a small portion of an organ is collected for LSC, this may not be an accurate reflection of whole organ biodistribution. Furthermore, LSC can only determine nanoparticle biodistribution at the tissue or organ level and does not provide any information regarding specific cellular association or accumulation of nanoparticles in tissues.

#### MEASUREMENT OF DRUG CONCENTRATION IN TISSUES

Nanoparticles loaded with therapeutic compounds can have their biodistribution evaluated by measuring drug concentration in tissues. This is an indirect approach and more specifically determines payload biodistribution. The assumption is that nanoparticles accumulate in specific tissues following in vivo administration, where they then release their cargo. It does not take into account possible premature drug release from the nanoparticles into the circulation and subsequent biodistribution of the free drugs themselves. This technique involves tissue samples being prepared for solubilization and extraction of the specific compound for further analysis. In order to achieve effective drug extraction from tissues, it is important to first determine the physicochemical properties of the compound and tissue matrix in the sample (Pavlovic´ et al., 2007).

In brief, biological tissues are broken down by methods such as grinding, blending, homogenization, sonication or sieving, as finer samples are more homogenous and easier to extract. Particulates are removed from the coarse biological material through methods such as centrifugation, filtration or solid-phase extraction. The supernatant is then collected and subjected to further extraction and purification. The extraction of drugs from biological tissues depends on its physicochemical properties, such as solubility, hydrophobicity/hydrophilicity, ionization, partition coefficient, and molecular weight. For example, solid-liquid extraction may be used, where a solvent is added to dissolve the analyte in the sample. The mixture is then filtered, decanted, or centrifuged to separate the solvent from the remaining sample. Following extraction, evaporation and reconstitution may be required before final analysis with high-performance liquid chromatography (HPLC) and/or mass spectrometry (MS) (Majors, 2013).

Measurement of drug concentration in tissues has been widely used for determining the biodistribution of nanomedicines. For example, Milane et al. (2011) assessed the biodistribution of epidermal growth factor receptor (EGFR)-targeted polymerblend nanoparticles loaded with the anti-cancer drugs, lonidamine and paclitaxel, in an orthotopic animal model of multi-drug resistant breast cancer. After euthanasia, the tumor mass, liver, lungs, kidneys, spleen, and heart were harvested and weighed. Tissue and plasma samples were then prepared using established methods for the extraction of lonidamine and paclitaxel in preparation for HPLC analysis. The data showed that both the non-targeted and the targeted nanoparticles were effective at increasing the tumor concentration of paclitaxel and lonidamine relative to free drug solution.

The main advantage of this technique is that it provides a quantitative measure of biodistribution that can be used to analyze whole or partial tissue samples (**Table 1**). This method does not involve exposure to ionizing radiation, incorporation of imaging molecules to nanoparticles, or the administration of contrast agents to enhance imaging outcomes. However, as mentioned earlier this indirect technique more specifically measures payload biodistribution. It is the compound encapsulated into or incorporated on the surface of the nanoparticles that is measured, which may provide unreliable results if the compound prematurely dissociates from the nanoparticles following in vivo administration. Furthermore, accurate measurement of drug concentration is highly dependent upon the quality of the tissue preparation and extraction procedure, which can be time-consuming and laborious. This technique is also unable to provide information on real-time biodistribution across time points in animals, but can be used as a secondary quantitative measure to support the biodistribution results attained from qualitative techniques.

### IN VIVO OPTICAL IMAGING

This technique refers to the use of equipment such as the In Vivo Imaging System (IVIS <sup>R</sup> ) and Kodak In-Vivo FX Imaging Station to visualize the biodistribution of nanoparticles in realtime in live animals or in harvested tissues and organs. These in vivo imaging systems are non-invasive and involve optical imaging technology to evaluate fluorescence or bioluminescence within the sample. Even though in vivo imaging systems typically possess these dual imaging capabilities, fluorescence imaging is used the most to evaluate the biodistribution of nanoparticles. Fluorescent imaging employs the ability of fluorophores, such as fluorescent proteins, dyes and conjugated polymers, to fluoresce after being excited with light of a particular wavelength (Janib et al., 2010; Coll, 2011; Priem et al., 2015). Fluorophores can be encapsulated within the nanoparticles (core or membrane) or conjugated to the nanoparticle surface. To optimize in vivo imaging sensitivity, fluorescent contrast agents should emit light in the red or near infrared (near-IR) wavelengths (∼600– 1,000 nm) (Coll, 2011; Liu Y. et al., 2012). This is particularly important for deep tissue samples to avoid coinciding with low photon absorption and autofluorescence in tissues, thereby enabling higher signal-to-background ratios (Vats et al., 2017). Once the sample is excited by a light source within the imaging chamber, fluorescence is emitted and captured on a chargecoupled device (CCD) camera that then converts this into electrical signals (Coll, 2011). A three-dimensional, tomographic image depicting the biodistribution of the fluorescent probe is then reconstructured.

In vivo imaging systems are commonly used to evaluate the biodistribution of nanoparticles, particularly in live animals across various time points to assess accumulation relative to disease progression. A variety of fluorescentlabeled nanoparticles have been imaged using this technique, including nanoporous silicon nanoparticles, carbon nanotubes, metal-based nanoparticles, polymer-based nanoparticles, and lipid-based nanoparticles (Connell et al., 2002; Zheng et al., 2003; Gao et al., 2010b; Goldberg et al., 2011; Milane et al., 2011; Tasciotti et al., 2011; Liu Y. et al., 2012; Zhang et al., 2012). For example, Milane et al. (2011) used this technique as a qualitative assessment of the biodistribution of EGFR-targeted polymer-blend nanoparticles in an orthotopic animal model of multi-drug resistant breast cancer. In this study, non-targeted and targeted nanoparticles loaded with DiR (near-IR) dye were administered via tail vein injection, and the biodistribution was visualized using a Kodak In-Vivo FX Imaging Station over 6 h. The results attained from in vivo optical imaging were found to be comparable with the quantitative data attained from HPLC analysis of drug distribution. Interestingly, some nanoparticles possess contrast that is inherently fluorescent such as quantum dot nanocrystals (Gao et al., 2010a,b; Liu Y. et al., 2012; Zhang et al., 2012; Zhao and Zeng, 2015). Quantum dots are semiconductor nanocrystals synthesized with a core-shell structure that enables imaging in the near infrared spectrum, thereby enhancing image sensitivity. They possess attractive optical qualities such as size-tunable fluorescence, photostability, high fluorescence quantum yields, and high resistance to photobleaching (Gao et al., 2010a). However, quantum dot preparations contain heavy metals such as cadmium, tellurium and selenium, which are potentially toxic to the body (Hardman, 2006; Kim et al., 2017).

Overall, in vivo optical imaging has the advantages of being direct, non-invasive and relatively simple to conduct (**Table 1**). It has fast image acquisition times and the procedure does not require exposure to ionizing radiation (Koo et al., 2006; Liu Y. et al., 2012). As imaging can be performed in realtime, biodistribution of nanoparticles can be assessed over many time points in the same group of animals—thus allowing a reduction in animal numbers. The images produced tend to have high sensitivity and enhanced temporal resolution (Liu Y. et al., 2012; Kim et al., 2017). This technique is generally considered a qualitative measure of biodistribution, as the intensity measured is not necessarily relative to the number of nanoparticles present in the tissues (Liu Y. et al., 2012). There are also a few limitations to this technique that should be considered. In vivo imaging systems can have limited tissue penetration (<1 cm) and is prone to attenuation with increased tissue depth (Koo et al., 2006; Kim et al., 2017). This is due to interference from light absorption and light scattering by tissue biomatter. This technique also has relatively low spatial resolution compared to CT and MRI (Massoud and Gambhir, 2003). In addition, labeling of nanoparticles with fluorophores may alter their physicochemical properties (e.g., surface charge, size, and surface functionalization) and in vivo behavior (Ann et al., 2013; Robson et al., 2018). Therefore, the choice of fluorophore and the method for labeling nanoparticles should be carefully considered. Another concern is that many fluorophores can undergo photobleaching (Robson et al., 2018), which affects their sensitivity to imaging. Tissue autofluorescence is a significant issue that can affect the interpretation of results, therefore fluorophores should have high signal-to-background ratios (Koo et al., 2006). Furthermore, in vivo imaging systems can only determine nanoparticle biodistribution at the tissue or organ level and do not provide any information regarding specific cellular association or accumulation of nanoparticles in tissues. It cannot visualize individual nanoparticles, but instead measures broader fluorescence intensity.

## COMPUTED TOMOGRAPHY (CT)

Computed tomography (CT) is a non-invasive, radiological imaging technique that uses X-rays to produce threedimensional, tomographic (cross-sectional) images of tissues. This technique is based on the variable absorption of X-rays by different tissues, which is a form of ionizing radiation with wavelengths of ∼0.01–10 nm (Kim et al., 2017). CT scanners typically consist of an X-ray tube, a detector unit, an image reconstruction system, collimators and filters. The X-ray tube is composed of a cathode and a tungsten-alloy anode housed within a vacuum. X-rays are generated within the tube by applying high voltage, which accelerates electrons from the heated cathode filament toward the anode. The accelerated electrons interact with electrons of the anode's tungsten nuclei and subsequently cause emission of X-rays. X-rays are then passed through the subject and are attenuated (absorbed or scattered), resulting in a loss of X-ray intensity (Lusic and Grinstaff, 2013; Liguori et al., 2015; Kim et al., 2017). Differential attenuation of X-rays across tissues according to their attenuation coefficient causes variation in radiation intensities and depicts information about tissue density and structure (Chatterjee et al., 2014; Liguori et al., 2015). This information is captured by detectors as a series of projections. Usually, the X-ray tube and detectors rotate synchronously on a circular axis around the subject with detectors positioned directly opposite, which enables a complete dataset of projections to be obtained over 360◦ . Computer algorithms are then applied to produce a three-dimensional reconstruction of the scanned object. Collimators and filters are used to limit unwarranted radiation and enhance the quality of the image (Liguori et al., 2015).

Contrast within the final image depends on the different densities and thickness of body structures. While different types of tissues can exhibit contrast, it can be particularly challenging to achieve high quality images and identify the interface between two different adjacent tissues (e.g., tumor in an organ) or to image soft tissues in contact with bodily fluids (Lusic and Grinstaff, 2013; Chatterjee et al., 2014). Therefore, contrast imaging agents are often used to increase CT sensitivity to enhance visualization and differentiation among different tissues. Contrast agents are usually elements having high atomic numbers and, therefore, higher number of electrons, which attenuate X-rays more efficiently by absorbing external X-rays. This results in decreased exposure on the X-ray detector (Lusic and Grinstaff, 2013). Contrast agents used clinically in patients undergoing CT are typically iodine- or barium-based compounds. Iodinated contrast agents are the main type of radiocontrast used for vascular imaging (e.g., vascular calcifications and hemorrhage), whereas barium sulfate is mainly used for imaging the gastrointestinal tract (Lusic and Grinstaff, 2013; Chatterjee et al., 2014; Kim et al., 2017).

CT has been utilized as a technique to allow in vivo imaging of the biodistribution of nanoparticles in real-time. Electron-dense elements are typically incorporated into the nanoparticles to enable visualization and differentiation of the nanoparticles in the tissues. Contrast agents that are more commonly incorporated into nanoparticles for CT analysis include iodine (Torchilin et al., 1999; Yordanov et al., 2002; Fu et al., 2006; Ho Kong et al., 2007; Elrod et al., 2009; de Vries et al., 2010; Hill et al., 2010; Hallouard et al., 2011), gold (Chie et al., 2010; Guo et al., 2010; Wang et al., 2011; Xiao et al., 2013), and bismuth (Rabin et al., 2006; Naha et al., 2014). However, various other elements such as gadolinium (Zhou et al., 2014), platinum (Chou et al., 2010), tantalum (Bonitatibus et al., 2010; Oh et al., 2011), tungsten (Jakhmola et al., 2014; Firouzi et al., 2017), and ytterbium (Pan et al., 2012; Jianhua et al., 2013) have also been used. Contrast agents for CT imaging can be loaded into the core of the nanoparticles, chemically grafted to the surface of nanoparticles, or inserted into the carrier membrane (e.g., lipid bilayer) (Cormode et al., 2014; Li et al., 2014). The in vivo biodistribution of numerous types of nanoparticles have been studied with CT, including nano-emulsions (de Vries et al., 2010; Hallouard et al., 2011), liposomes (Sachse et al., 1997; Leander et al., 2001; Elrod et al., 2009), micelles (Torchilin et al., 1999; Torchilin, 2002), lipoproteins (Cormode et al., 2008; Hill et al., 2010), polymercoated nanoparticles (Rabin et al., 2006; Muddineti et al., 2015; Firouzi et al., 2017), nanocapsules/nanospheres (Ashcroft et al., 2007; Ho Kong et al., 2007), nanotubes/nanorods (Ashcroft et al., 2007; Zhou et al., 2014), metal-based nanoparticles (Bonitatibus et al., 2010; Chou et al., 2010; Oh et al., 2011; Pan et al., 2012; Jianhua et al., 2013; Mieszawska et al., 2013; Cormode et al., 2014; Jakhmola et al., 2014; Naha et al., 2014; Kim et al., 2017), and dendrimers (Yordanov et al., 2002; Fu et al., 2006; Chie et al., 2010; Guo et al., 2010; Wang et al., 2011; Xiao et al., 2013).

CT has demonstrated to be an effective technique for producing reliable and high-resolution images for assessing the biodistribution of nanoparticles (**Table 1**). It has no tissue penetration limits and relatively quick image acquisition times (Massoud and Gambhir, 2003). This technique is generally considered a qualitative measure of biodistribution and can only determine nanoparticle biodistribution at the tissue or organ level. Furthermore, CT requires exposure to ionizing radiation and does not provide any information regarding specific cellular association of nanoparticles (Kim et al., 2017). Biodistribution of nanoparticles can be assessed in real-time and over many time points in the same group of animals, which reduces the number of animals required for longitudinal studies. However, CT alone can suffer from relatively poor visualization and differentiation among different types of tissues as mentioned above (Lusic and Grinstaff, 2013; Chatterjee et al., 2014). Hence, it often requires the administration of contrast imaging agents to increase CT sensitivity. This can pose a problem when nanoparticles labeled with contrast agents are used in conjunction with other contrast imaging agents to improve anatomical and tissue imaging. The detection limit of nanoparticle contrast agents is less sensitive compared to other modalities, such as nuclear imaging (Massoud and Gambhir, 2003; Kim et al., 2017). To overcome this issue, nanoparticles incorporating high concentrations of contrast agents are often required to improve imaging. Incorporation of contrast agents in nanoparticles may alter their physicochemical properties and in vivo behavior (Massoud and Gambhir, 2003; Kim et al., 2017).

### MAGNETIC RESONANCE IMAGING (MRI)

Magnetic resonance imaging (MRI) is a non-invasive imaging technique that produces three dimensional detailed anatomical images, without the use of ionizing radiation. MRI uses powerful magnets that produce a strong magnetic field that forces protons in the body to align with that field (Strijkers et al., 2007; Grover et al., 2015). Protons (hydrogen nuclei) are typically used in MRI imaging as they are particularly abundant in the water and fat of the body. Protons possess a positive charge and are constantly spinning around their own axes, which generates a magnetic field. The magnetic field for each proton is known as a magnetic moment and is a measure of an object's tendency to align with a magnetic field. Radiofrequency currents are pulsed through the patient to excite the protons to a higher energy state and spin them out of equilibrium, which creates strain against the pull of the magnetic field (Grover et al., 2015). When the radiofrequency field is turned off, the protons then realign with the magnetic field and the MRI sensors can detect the energy that is released in this process. In particular, MRI is able to produce high-resolution images by measuring the spin magnetization of polarized protons and their respective longitudinal (T1) and transverse (T2) relaxation rates in the body. It utilizes magnetic fields, electric field gradients and radio waves to produce three types of images: spin density weighted, T<sup>1</sup> weighted and T<sup>2</sup> weighted images (Strijkers et al., 2007; Grover et al., 2015). Field gradient coils are used to localize the MRI signal to particular tissues of interest. The signals are processed to extract frequency and phase data, and a mathematical algorithm is applied to construct an image. The

time it takes for the protons to realign with the magnetic field and the amount of energy released changes depending on the environment and the chemical nature of the molecules, which is used to differentiate between various types of tissues (Grover et al., 2015).

Contrast agents may be administered to a patient intravenously before or during the MRI procedure to increase the speed at which protons realign with the magnetic field, thereby shortening the T<sup>1</sup> and/or T<sup>2</sup> relaxation rates of protons located in their vicinity. Contrast agents that shorten T<sup>1</sup> (paramagnetic contrast agents) result in T<sup>1</sup> and T<sup>∗</sup> 1 hypersignal (brighter images), whereas those that shorten T<sup>2</sup> (superparamagnetic contrast agents) lead to a reduction in the T<sup>2</sup> and T<sup>∗</sup> 2 signal (darker images) (Strijkers et al., 2007; Kamaly and Miller, 2010). This improvement in image quality also enhances the differentiation between tissues. The effectiveness of contrast agents depends on its relaxivity, which is the proportionality constant of the measured rate of relaxation: 1/T<sup>1</sup> and 1/T<sup>2</sup> (Sun et al., 2008). Superparamagnetic iron oxide crystals (Fe3<sup>+</sup> or Fe2+) and paramagnetic lanthanide metals, such as gadolinium (Gd+<sup>3</sup> ), are the most widely used contrast agents for MRI imaging (Strijkers et al., 2007; Kamaly and Miller, 2010). Superparamagnetic iron oxide (SPIO, >50 nm in size) and ultrasmall superparamagnetic iron oxide (USPIO, <50 nm in size) are mainly used to shorten T2, leading to darker images in T<sup>2</sup> and T<sup>∗</sup> <sup>2</sup> weighted MRI (Jung and Jacobs, 1995; Strijkers et al., 2007; Kamaly and Miller, 2010). Conversely, paramagnetic gadolinium ions are used to shorten T1, resulting in brighter images in T<sup>1</sup> weighted MRI (Strijkers et al., 2007; Kamaly and Miller, 2010). The most clinically used MRI contrast agents are those that shorten T<sup>1</sup> relaxation rates, hence those that contain the element gadolinium are often preferred (Sun et al., 2008).

Gadolinium (Gd+<sup>3</sup> ) has seven unpaired outer shell electrons and a large magnetic moment, making it extremely useful for MRI imaging (Strijkers et al., 2007; Kamaly and Miller, 2010). Free gadolinium ions are highly toxic and, therefore, they are usually chelated with other ligands (e.g., diethylenetriamine pentaacetic acid, DTPA; tetraazacyclododecane tetraacetic acid, DOTA) to form complexes that are nontoxic and highly stable in the body during the period of administration (Wiegers et al., 1992; Rosen et al., 2011). For example, Park et al conjugated the peptide RGD to Gd-DOTA to obtain an MRI contrast agent with tumor targeting capability (Park et al., 2008). One of the more common gadolinium chelate used clinically and in drug delivery is gadoteridol, which is the chelate formed between Gd+<sup>3</sup> and 10-(2-hydroxy-propyl)- 1,4,7,10-tetraazacyclododecane-1,4,7-triacetate (Zhou and Lu, 2013). Gadolinium-based contrast agents can be incorporated into nanoparticles to enable real-time imaging of their in vivo biodistribution using MRI. A number of nanoparticle types have incorporated these contrast agents, including liposomes (Unger et al., 1989; Saito et al., 2005; Hossann et al., 2013; Smith et al., 2013; Skupin-Mrugalska et al., 2018), dendrimers (Margerum et al., 1997; Lee et al., 2005; Rongzuo et al., 2007), micelles (Parac-Vogt Tatjana et al., 2004; Kumar et al., 2010), polymeric-based nanoparticles (Liu et al., 2011), carbon-based nanotubes (Hartman et al., 2008; Richard et al., 2008), and mesoporous silica nanoparticles (Kobayashi et al., 2007; Kim et al., 2008; Taylor et al., 2008). For example, Saito et al. (2005) manufactured gadoteridol-loaded liposomes for real-time MRI evaluation of convection-enhanced delivery in the primate brain. Volume of distribution was analyzed for all delivery locations by histology and MRI, following administration in the corona radiata, putamen nucleus, and brain stem. The results showed that MRI of liposomal gadolinium was highly accurate at determining tissue distribution, as confirmed by comparison with histological results from concomitant administration of fluorescent liposomes. Gadolinium-based contrast agents can be incorporated within the core of nanoparticles, attached to the particle surface, or inserted into the carrier membrane (Unger et al., 1989; Hossann et al., 2013; Smith et al., 2013; Skupin-Mrugalska et al., 2018). It should be noted that encapsulation of gadolinium within the core can lead to lowered relaxivity, whereas surface attachment may be preferable to improve gadolinium's ability to interact with water (Tilcock et al., 1989; Kamaly and Miller, 2010). Relaxivity can be further improved by reducing the size of the nanoparticles (Tilcock et al., 1989; Kamaly and Miller, 2010).

MRI can also be used to evaluate the biodistribution of nanoparticles formulated with superparamagnetic iron oxide cores. Iron oxide crystals are mainly utilized to provide negative contrast in T<sup>2</sup> and T<sup>∗</sup> <sup>2</sup> weighted images. SPIO and USPIO nanoparticles are usually composed of a nano-sized magnetite (Fe3O4) or maghemite (γ-Fe2O3) core coated with a variety of materials to enhance stability, circulation time, biocompatibility and minimize toxicity (Peng et al., 2008). Bulk iron oxide is ferromagnetic, however when nano-sized, superparamagnetism is exhibited (Di Marco et al., 2007). The superparamagnetism of iron oxide nanoparticles is important for in vivo imaging. Polymers are the most widely used stabilizing materials and can be adsorbed into or anchored onto the iron oxide surface via hydrogen bonds, electrostatic forces or pseudo-covalent bonding (Estelrich et al., 2015). Examples include poly(ethylene glycol) (PEG), alginate, chitosan, dextran and its derivatives, starch, polyvinyl alcohol, albumin, poly(ethylene imine), organic siloxane, and sulphonated styrene-divinyl-benzene (Estelrich et al., 2015). In addition, SPIO can be used alone or incorporated into other nanostructures, such as magnetoliposomes (SPIOs are hybridized within a liposome carrier) (Martina et al., 2005; Plassat et al., 2007) and colloidal iron oxide nanoparticles (oleate-coated magnetite particles embedded in a hydrophobic matrix) (Senpan et al., 2009). Several formulations of iron oxide nanoparticles are already approved for clinical use (e.g., ferumoxides and ferucarbotran) for contrast-enhanced MRI of the liver (Reimer and Tombach, 1998). Their relatively large surface area also enables incorporation of biologically active substances to the surface of the nanoparticles. For example, Veiseh et al. (2009) developed a nanoprobe consisting of an iron oxide nanoparticle coated with biocompatible PEG– grafted chitosan copolymer, which allowed conjugation of a tumor-targeting agent, chlorotoxin, and a near-IR fluorophore. The results showed an ability for the nanoprobe to cross

the blood-brain barrier and specifically target brain tumors in a genetically engineered mouse model, as evidenced by in vivo MRI evaluation, in vivo optimal imaging and histology. The magnetism and subsequent MRI effectiveness of iron oxide nanoparticles is dependent upon their size, shape, morphology, structure, and homogeneity (Lin et al., 2012; Estelrich et al., 2015). Thus, variations in SPIO and USPIO nanoparticles can lead to different magnetic properties and thus alter their function in various applications. The coating and surface modifications can also influence in vivo stability and biodistribution of the nanoparticles. It should be noted that for conventional MRI, SPIO nanoparticles give negative contrast enhancement (dark signals) that are often confounded by the presence of artifacts due to hemorrhage, air, and partial-volume effects. To address these issues, many attempts have been made to generate positive contrast visualization methods in the last decade (Lin et al., 2012; Estelrich et al., 2015).

Overall, MRI has the advantage of producing high spatial resolution images (micrometers rather than several millimeters) compared to other techniques such as optical or radionuclide imaging (Massoud and Gambhir, 2003) (**Table 1**). It provides better soft tissue contrast than CT and can differentiate better between fat, water, muscle, and soft tissue (Massoud and Gambhir, 2003; Janib et al., 2010). MRI is not limited by tissue depth (unlimited penetration) and does not involve exposure to ionizing radiation. Furthermore, this technique allows non-invasive, three-dimensional, real-time imaging of the biodistribution of nanoparticles in vivo. However, MRI is more costly and has slow image acquisition and long post-processing times (Kim et al., 2017). As this technique can suffer from poor sensitivity, much higher amounts of contrast agent generally need to be administered (Massoud and Gambhir, 2003; Kim et al., 2017). MRI cannot be used in subjects with metallic implants/devices (Janib et al., 2010). In addition, incorporation of contrast agents in nanoparticles may alter their physicochemical properties and in vivo behavior (Massoud and Gambhir, 2003; Kim et al., 2017).

### NUCLEAR MEDICINE IMAGING

Single photon emission computed tomography (SPECT) and positron emission tomography (PET) scans are the two most common imaging modalities in nuclear medicine. They are both non-invasive techniques that produce three-dimensional images of the body by detecting gamma rays (γ-rays) that are emitted from radioactive substances that become localized and are taken up by specific tissues (Townsend, 2004; Ziegler, 2005; Pimlott and Sutherland, 2011; Van Audenhaege et al., 2015). Both techniques essentially involve administration of a radioactive tracer (radiotracer) into the subject that consists of a molecular probe with a radioactive isotope attached that is capable of emitting γ-rays. The choice of molecular probe is dependent on the tissue to be imaged and should ideally have high affinity and high selectivity for the target receptor or organ (Pimlott and Sutherland, 2011). As the isotope decays in the tissue, it emits gamma rays that are picked up by detectors (gamma scintillation camera system) placed around the subject. The scintillation crystals within the detectors then convert the γ-ray energy into lower-energy (near-optical) photons. This optical energy is converted into electrical signals by photomultiplier tubes and processed to obtain the location of the scintillation events in the crystal (Townsend, 2004; Ziegler, 2005; Peterson and Furenlid, 2011). The radionuclide is captured in a collection of projections, which are measured from numerous angles and linear positions in the subject. Image reconstruction techniques are applied to reconstruct these projections into a three-dimensional, tomographic image of the radiotracer's biodistribution and concentration within the tissue (Townsend, 2004; Ziegler, 2005; Peterson and Furenlid, 2011; Pimlott and Sutherland, 2011).

Although both PET and SPECT rely on the detection of gamma radiation, they differ in the type of radionuclides used. The radionuclides used in SPECT emit γ-rays by radioactive decay that is measured directly, whereas PET radionuclides emit positrons that annihilate with electrons up to a few millimeters away in the tissue to produce two gamma photons that are emitted in opposite directions (Massoud and Gambhir, 2003; Townsend, 2004; Ziegler, 2005). The γ-rays emitted in PET are captured in coincidence by opposing pairs of detectors aligned collinearly around the subject, which enable measurement of the radionuclide from multiple angles and planes (Townsend, 2004; Ziegler, 2005). Unlike PET, SPECT gamma cameras are rotated around the subject and a lead collimator is required to reconstruct the original location of the emitted γ-rays (Peterson and Furenlid, 2011; Van Audenhaege et al., 2015). PET positron emitters (e.g., <sup>15</sup>O, <sup>64</sup>Cu, <sup>13</sup>N, <sup>11</sup>C, and <sup>18</sup>F) emit higher energy γ-rays and possess shorter radioactive half-lives than SPECT radiotracers (Massoud and Gambhir, 2003; Townsend, 2004; Ziegler, 2005). The most common radioisotopes used for SPECT imaging include 99mTc, <sup>111</sup>In, and radioiodine (e.g., <sup>131</sup>I) (Hong et al., 2009; Pimlott and Sutherland, 2011).

Nanoparticles can be labeled with gamma-emitting radionuclides and positron emitters. These radiolabels can be attached to the nanoparticle surface, conjugated to the nanoparticle core, or encapsulated within a payload that is loaded into the nanoparticle. Radiolabeling is achieved through methods such as exogenous chelation of radiometals, direct proton/neutron bombardment, and chelator-free radiolabeling (Gibson et al., 2011; Liu T. et al., 2012 Sun et al., 2015; Lu et al., 2018; Yuan et al., 2018). Alternatively, radioactive precursors can be used to synthesize intrinsically radioactive nanoparticles (Zhao et al., 2014; Sun et al., 2015). The type of radionuclide and radiolabeling strategy requires careful consideration, as some nanoparticles may exhibit differing compatibility and imaging effectiveness across the various methods (Liu and Welch, 2012). PET can be used to image the biodistribution of a variety of nanoparticles, including quantum dots (Ducongé et al., 2008; Tu et al., 2011), iron oxide nanoparticles (Glaus et al., 2010; Yang et al., 2011), gold nanoparticles (Xie et al., 2010; Guerrero et al., 2012), liposomes (Oku et al., 2011; Petersen et al., 2011), solid lipid nanoparticles (Andreozzi et al., 2011), polymer-based nanoparticles (Fukukawa et al., 2008; Herth et al., 2009; Allmeroth et al., 2013), carbon-based nanoparticles (Liu et al., 2006; McDevitt et al., 2007), and micelles (Xiao et al., 2012). Similarly, SPECT imaging, often in combination with other imaging modalities, can also image the biodistribution of a similar range of nanoparticles, including dendrimers (Zhang et al., 2010a,b), micelles (Cheng et al., 2013; Hong et al., 2014), liposomes (Chang et al., 2010), carbon-based nanoparticles (Wu et al., 2009), iron-oxide nanoparticles (Madru et al., 2012), polymeric nanoparticles (Lu et al., 2011), gold nanoparticles (Morales-Avila et al., 2011; You et al., 2012), and silver nanoparticles (Chrastina and Schnitzer, 2010).

As PET and SPECT imaging rely purely on the detection of γ-rays, radiolabels must remain attached to the nanoparticles to accurately image their biodistribution. If disassociation occurs, imaging will not reflect true biodistribution, resulting in misleading and incorrect information (Liu and Welch, 2012). Therefore, it is important that the radiolabeling strategy, radionuclide type, and nanoparticle material are compatible, suited to the study purpose, and possess high in vivo stability (Liu and Welch, 2012; Sun et al., 2015). Although exogenous chelation of radionuclides is relatively easy, efficient and low cost, the resulting stability of radiolabels can be potentially problematic. Radionuclides may detach from chelators through transchelation or chelators may interact in vivo and subsequently disassociate from the nanoparticle (Bass et al., 2000; Boswell et al., 2004; Sun et al., 2015). The attachment of a chelator may also influence or damage the surface properties of nanoparticles, as high temperatures are required for chelation (Lu et al., 2018). This can adversely affect the conjugation capacity of targeting ligands (e.g., antibodies and PEG density), thereby resulting in impaired targeting, reduced circulation times and decreased imaging activity (Chang et al., 2008; Moghimi et al., 2012; Lu et al., 2018). Chelation issues can be avoided with direct bombardment radiolabeling, however this technique is limited by high costs, complexity of use, and the potential to damage the nanoparticles with ion-beam/neutron irradiation (Gibson et al., 2011). While intrinsic radioactive nanoparticles can exhibit high stability with limited radiolabel detachment, potential long-term toxicity and its limited applicability to only a few radioisotope-nanoparticle combinations present challenges for this technique (Liu T. et al., 2012; Chen et al., 2013; Goel et al., 2014). Furthermore, the chelator-free post-synthetic radiolabeling approach is fast, specific and can produce a high labeling yield, however is again limited to only a few nanoparticle and isotope combinations (Chen et al., 2013; Sun et al., 2015).

In comparison to other imaging modalities, PET and SPECT have the advantages of being able to image biochemical processes and are highly sensitive (nanomolar to picomolar level) (**Table 1**). Therefore, signals can be detected with very small amounts of labels which minimizes the disruption of cell function and surrounding tissue (Townsend, 2004; Ziegler, 2005; Pimlott and Sutherland, 2011; Van Audenhaege et al., 2015). These nuclear medicine imaging techniques are also quantitative and not restricted by tissue penetration limits (Koo et al., 2006; Janib et al., 2010; Kim et al., 2017). Several limitations should be considered for PET and SPECT imaging, including exposure to ionizing radiation, high costs, slow image acquisition times, and inability to be used for longitudinal studies due to radiolabel decay (Kim et al., 2017). Both imaging techniques also have low spatial resolution and provide a lack of anatomic information, hence they are often combined with other imaging modalities such as MRI or CT (Janib et al., 2010; Kim et al., 2017). When comparing between the two imaging techniques, SPECT has low photon detection efficiency and relatively low resolution due to the use of collimation, whereas PET is much more sensitive and provides more radiation event localization data owing to the detection of emissions "coincident" in time (Koo et al., 2006). The positron-emitting isotopes used in PET are also able to be substituted for naturally occurring atoms, thereby enhancing the ability to image molecular events (Massoud and Gambhir, 2003). However, SPECT can image multiple radionuclide probes simultaneously and is more widely available. SPECT scans are also significantly less expensive than PET scans, partly because their radionuclides are simple to prepare, more easily obtained, and typically possess longer half-lives than PET radionuclides (Massoud and Gambhir, 2003; Janib et al., 2010; Pimlott and Sutherland, 2011). In addition, PET typically requires a cyclotron or generator (Massoud and Gambhir, 2003).

## CONCLUSION

There is a range of techniques available for evaluating the biodistribution of nanoparticles in vivo. In general, the choice of technique depends on the: (i) physicochemical characteristics of the nanoparticle formulation; (ii) compatibility and stability of the nanoparticles with different labels and labeling methods; (iii) study duration (single or multiple time points); (iv) analysis type (quantitative or qualitative); (v) sample type (whole animal, whole organ/tissue, or tissue sections); and (vi) degree of detail required (organ/tissue accumulation or cellular association). Other aspects that should be considered include accessibility, costs, accuracy, image resolution, toxicity, complexity, and duration of the procedure. Each technique has its own advantages and limitations, as well as capabilities for assessing real-time, whole-organ, and cellular accumulation. The techniques which allow real-time and qualitative imaging of biodistribution in live animals are in vivo optical imaging, CT, MRI and nuclear medicine imaging (PET and SPECT). PET and SPECT are also able to provide quantitative data of uptake into specific organs or tissues, along with LSC and indirectly measuring drug concentration. Of the techniques available, only in vivo optimal imaging, CT and MRI are capable of imaging nanoparticle biodistribution across multiple time-points in longitudinal studies. In addition, histology and electron microscopy are the only techniques that can provide detailed information on the cellular association of nanoparticles following in vivo administration. Research on the use of other modalities for studying the biodistribution of nanoparticles in vivo are currently being explored, including ultrasound imaging of nanoparticles loaded with ultrasound contrast agents (e.g., insoluble gas perfluorocarbons or sulfur hexafluoride) (Janib et al., 2010; Shapiro et al., 2014; Kim et al., 2017). Nanoparticles with multifunctional theranostic capabilities, incorporating multimode contrast agents are also rapidly gaining popularity for biomedical applications.

### AUTHOR CONTRIBUTIONS

SH: conception of the work. LA and SH: drafting of the manuscript. SH: preparation of the figure and table. SH, DS, JF,

#### REFERENCES


WP, AM, and AW: reviewing the article critically for important intellectual content.

#### ACKNOWLEDGMENTS

The authors wish to thank The Pharmacy Research Trust of New South Wales, The Rebecca L. Cooper Medical Research Foundation, Gladys M. Brawn Fellowship, ausEE Research Foundation, and The University of Newcastle for providing financial support for our research.


using TEM sections through cells, tissues and organs. Ann. Anat. 191, 153–170. doi: 10.1016/j.aanat.2008.11.001


for the development of new drugs. Regul. Toxicol. Pharmacol. 22, 122–142. doi: 10.1006/rtph.1995.1078


Poly(amidoamine) (PAMAM)–folic acid conjugates. J. Med. Chem. 53, 3262–3272. doi: 10.1021/jm901910j


**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 Arms, Smith, Flynn, Palmer, Martin, Woldu and Hua. 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.

# Advantages and Limitations of Current Imaging Techniques for Characterizing Liposome Morphology

Annie-Louise Robson<sup>1</sup> , Paul C. Dastoor<sup>2</sup> , Jamie Flynn1,3, William Palmer3,4 , Antony Martin3,4, Doug W. Smith<sup>1</sup> , Ameha Woldu<sup>3</sup> and Susan Hua1,3 \*

<sup>1</sup> School of Biomedical Sciences and Pharmacy, University of Newcastle, Callaghan, NSW, Australia, <sup>2</sup> Centre for Organic Electronics, University of Newcastle, Callaghan, NSW, Australia, <sup>3</sup> Hunter Medical Research Institute, New Lambton Heights, NSW, Australia, <sup>4</sup> School of Environmental and Life Sciences, University of Newcastle, Callaghan, NSW, Australia

There are currently a number of imaging techniques available for evaluating the morphology of liposomes and other nanoparticles, with each having its own advantages and disadvantages that should be considered when interpreting data. Controlling and validating the morphology of nanoparticles is of key importance for the effective clinical translation of liposomal formulations. There are a number of physical characteristics of liposomes that determine their in vivo behavior, including size, surface characteristics, lamellarity, and homogeneity. Despite the great importance of the morphology of nanoparticles, it is generally not well-characterized and is difficult to control. Appropriate imaging techniques provide important details regarding the morphological characteristics of nanoparticles, and should be used in conjunction with other methods to assess physicochemical parameters. In this review, we will discuss the advantages and limitations of available imaging techniques used to evaluate liposomal formulations.

Keywords: liposomes, microscopy, imaging, nanoparticles, light microscopy, electron microscopy, atomic-force microscopy

### INTRODUCTION

Liposomes are a type of nanocarrier that have been widely investigated for drug-delivery purposes. They are composed of phospholipid bilayers which enclose a distinct aqueous space, thereby allowing encapsulation of both hydrophilic and hydrophobic compounds (Metselaar and Storm, 2005). Liposomes are able to stabilize therapeutic compounds and overcome barriers to cellular and tissue uptake (Ding et al., 2006; Hua and Wu, 2013). This allows them to improve targeting of compounds to sites of disease and consequently reduce accumulation in non-target organs (Bakker-Woudenberg et al., 1994; Mastrobattista et al., 1999; Hua, 2013; Hua et al., 2015; Sercombe et al., 2015; Zununi Vahed et al., 2017). There are four main types of liposomes based on their surface characteristics – conventional liposomes, PEGylated liposomes, ligand-targeted liposomes, and theranostic liposomes (**Figure 1**; Sercombe et al., 2015). Ligand-targeted liposomes provide the potential for site-specific delivery of drugs to certain tissues or organs that selectively express the targeted ligand (Willis and Forssen, 1998; Bendas, 2001; Sawant and Torchilin, 2012), whereas PEGylated liposomes confer steric hindrance to enhance the circulation half-life of the delivery system following systemic administration (Torchilin, 1994; Wang et al., 2015). Liposomes

#### Edited by:

Umesh Gupta, Central University of Rajasthan, India

#### Reviewed by:

Medha D. Joshi, Midwestern University, United States Domenico Criscuolo, Genovax S.r.l., Italy

#### \*Correspondence:

Susan Hua susan.hua@newcastle.edu.au

#### Specialty section:

This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology

Received: 06 November 2017 Accepted: 24 January 2018 Published: 06 February 2018

#### Citation:

Robson A-L, Dastoor PC, Flynn J, Palmer W, Martin A, Smith DW, Woldu A and Hua S (2018) Advantages and Limitations of Current Imaging Techniques for Characterizing Liposome Morphology. Front. Pharmacol. 9:80. doi: 10.3389/fphar.2018.00080

Frontiers in Pharmacology | www.frontiersin.org

incorporating a combination of the various delivery platforms can further improve the delivery of encapsulated compounds, depending on the route of administration and site of disease.

Following the manufacturing process, liposomes are characterized to ensure homogeneity across a number of parameters, including drug encapsulation, ligand-conjugation, lipid composition, surface charge, and morphological properties (e.g., size, shape, and number of lamellae) (Kuntsche et al., 2011). These characteristics are important as they can have a major impact on the behavior of liposomes both in vitro and in vivo (Sawant and Torchilin, 2012; Sercombe et al., 2015). Recognition and clearance of liposomes by the body's defenses, including the reticuloendothelial system (RES) and adsorption of opsonins with subsequent uptake by the mononuclear phagocytic system, are major contributors to the clearance and degradation of liposomes (Senior, 1987; Cullis et al., 1998; Ishida et al., 2001). Therefore, being able to determine the physicochemical properties of manufactured liposomes is important to optimize a formulation for further translational evaluation.

A major aspect in the physicochemical assessment of liposomes is visualizing the morphology of the nanoparticles using microscopy. There are a number of techniques available for imaging liposomes and other nanoparticles that can be broadly categorized into light, electron, or atomic-force microscopy (Bibi et al., 2011). Each technique has its own advantages and limitations, which should be considered when evaluating studies on nanoparticle-based drug delivery systems **(Table 1)**. This review will evaluate each imaging technique used to assess the morphological characteristics of liposomes.

### LIGHT MICROSCOPY

Light or optical microscopy refers to microscopes that utilize visible light and an arrangement of lenses to magnify a field of view (Murphy and Davidson, 2012c). Basic light microscopy itself is unable to provide comprehensive information about the lipid bilayer compared to the detail offered by other microscopy techniques. However, it can be used to rapidly obtain an image of vesicles using basic laboratory equipment (Bibi et al., 2011). This technique can be particularly useful when gathering general information on the size, shape, homogeneity, and degree of aggregation of a liposome sample (Nallamothu et al., 2006). Light microscopes have an ultimate resolution of ∼250 nm (governed by the smallest diffraction-limited spot size that can be achieved by the instrument) and, as such, are typically incapable of providing detailed information regarding the structures of small unilamellar vesicles (SUVs) and the lamellarity of vesicles (Bibi et al., 2011). Generally, light microscopy can only provide significant information on giant unilamellar vesicles (GUVs), which can range from single to hundreds of micrometers in diameter (Bagatolli, 2009). Incorporation of fluorescent probes, polarization techniques, and application of high-resolution confocal microscopy can provide more information about the 3D structure and lamellarity of the vesicles (Bagatolli, 2009).

### Polarization Microscopy

This type of light microscopy utilizes the unique ability of polarized light to interact with polarizable bonds of ordered molecules (Murphy and Davidson, 2012d). Enhanced light absorption occurs when molecular dipoles in the sample are aligned with the polarization vector of the incident light wave; resulting in phase differences between sampling light rays, which in turn produce interference-dependent changes in amplitude in the image plane (Murphy and Davidson, 2012d). Image contrast then arises not only from the effects of interference and diffraction, but also due to the presence of ordered molecular arrangements (Murphy and Davidson, 2012d). It can be used to study the form and dynamics of many ordered cellular structures, including lipid bilayers of plasma membranes (Bibi et al., 2011; Murphy and Davidson, 2012d). Images can be obtained in either monochrome or color. This technique provides an alternative method to visualize liposomes, particularly to confirm the formation of vesicles. It does not provide conclusive observations regarding the bilayer characteristics or lamellarity of the vesicles (Bibi et al., 2011). In addition, polarization microscopy is also limited by the size of the vesicles that can be visualized, with large vesicles in the micrometer range having the optimal clarity. This technique does not provide clear images of particles in the nanometer range.

#### Fluorescence Microscopy

Fluorescence or "epifluorescence" microscopy is a special form of light microscopy that exploits the ability of fluorochromes to emit light after being excited with light of a certain wavelength (Murphy and Davidson, 2012b). This technique is widely used in biophysics to provide essential information about the structure and dynamics of membrane components (Bouvrais et al., 2010). In these studies, fluorescent probes are specifically incorporated into the membrane under investigation, permitting visualization of the structure. A large range of fluorescent dyes with various chemical and photonic properties are commercially available, catering to a variety of research questions. For example, certain fluorescent dyes may prefer specific membrane environments exhibiting different arrangements and lateral packing (Bouvrais et al., 2010). As such, fluorescence microscopy allows for the potential to simultaneously apply multiple probes within a sample to provide information about the membrane structure itself. The fluorescent probes can be placed within the aqueous compartment as well as the lipid bilayer of liposomes (Bibi et al., 2011). This arrangement can be especially useful when viewing GUVs, where information can be obtained regarding the shape, size and fluidity of the lipid vesicles (Klymchenko et al., 2009; Bouvrais et al., 2010). Incorporation of probes, such as rhodaminelabeled lipids, directly into the lipid bilayer can also allow visualization of the lamellarity of liposomes (Bibi et al., 2011).

The disadvantage of fluorescent microscopy is that the addition of probes and dyes to a membrane system can potentially interfere with the properties of the liposomal delivery system (Bouvrais et al., 2010; Bibi et al., 2011; Murphy and Davidson, 2012b). However, the use of low dye concentrations

(≤1 mol%) has been shown to have minimal impact on the physical properties of the membrane (Bouvrais et al., 2010). It is also important to note that the choice of fluorescent dye is a critical step, as some dyes can induce large changes in the host membrane and/or cause experimental artifacts, resulting in inaccurate data interpretation (Bouvrais et al., 2010). In addition, photo-induced lipid peroxidation can also lead to domain formation even in simple dye systems. This process can result in the formation of large rafts and spontaneous facet formation (Bouvrais et al., 2010). Prolonged exposure to fluorescent light can also result in bleaching and loss of fluorescence intensity. Therefore, the choice of fluorescent dye and the development of new membrane probes are important considerations when using fluorescence microscopy (Klymchenko et al., 2009).

#### Confocal Microscopy

Confocal scanning microscopy has been an advancement in the area of fluorescence microscopy. Rather than illuminating the entire sample, an image is built by scanning one (or more) focused beams of light across the sample. Light returning from the illuminated sample passes through an aperture that rejects out-of-focus light from above and below the plane of interest; ensuring that only images from a small depth of field are obtained, greatly improving the out-of-plane resolution (Bibi et al., 2011; Murphy and Davidson, 2012a). Using this technique, a "z-stack" of images is collected, starting from the top of the vesicle followed by images taken in defined z-increments to the bottom of the sample, resulting in a composite 3D image of the sample (Bibi et al., 2011; Murphy and Davidson, 2012a). Confocal scanning microscopy has become a more attractive technique over epifluorescent light microscopes due to its superior image clarity. In the case of GUVs, this technique is capable of visualizing the internal structure of the lipid systems, which is often not possible with other microscopy methods (Ruozi et al., 2011; Mertins and Dimova, 2013). For example, separation of the aqueous and lipid bilayer phase can be clearly visualized in larger vesicles (Mertins and Dimova, 2013). However, confocal microscopy is still diffraction-limited and, therefore, unable to produce high-definition images of SUVs or oligolamellar liposomes (Ruozi et al., 2011).

## ELECTRON MICROSCOPY TECHNIQUES

Electron microscopy is a method for the visualization of vesicles under very high magnification (Henry, 2005). It is widely used in the characterization of lipid vesicles as the electron wavelength (and hence diffraction-limited resolution), is many orders of magnitude lower than that of optical microscopy, and therefore provides super-resolution for clear visualization of small liposomes (Bibi et al., 2011; Ruozi et al., 2011). This technique uses a beam of electrons focused onto the surface

#### TABLE 1 | Summary of current imaging techniques for characterizing liposome morphology.


(Continued)

#### TABLE 1 | Continued

fphar-09-00080 February 3, 2018 Time: 13:27 # 5


of the sample by various electromagnetic lenses. The electrons are then scattered by the sample, and are then refocused and magnified by a further series of electromagnetic lenses in the imaging column to produce a projected image (Henry, 2005). There are a number of different types of electron microscopes, each requiring a different sample preparation method.

#### Scanning Electron Microscopy (SEM)

Scanning electron microscopes (SEMs) utilize an electron beam that is scanned across or over a sample (rather than through a sample) to produce a magnified image of an object (Adler and Schiemann, 1985). Alder et al. first attempted to use SEM to characterize liposomes in 1984 (Adler and Schiemann, 1985). They showed that using the freeze-drying method to prepare the liposome samples for SEM resulted in a large proportion of visible lumps and crusted material (Adler and Schiemann, 1985). More recent studies have shown that the liposome structure itself may suffer perturbations due to the high-vacuum conditions and staining processes required for this preparation technique (Ruozi et al., 2011). SEM is now not commonly used for analyzing liposomes because it requires the sample to be dried or fixed prior to imaging (Ruozi et al., 2011). However, SEM can provide general information on the concentric structure of the different lipid layers, as well as give detail on the size and spherical morphology of a preparation (Nirale et al., 2009).

#### Transmission Electron Microscopy (TEM)

Transmission electron microscopy (TEM) is the most frequently used imaging method for the evaluation of the structure of nanoparticles (Henry, 2005; Kuntsche et al., 2011). It involves the preparation of a thin sample (<100 nm thick) that is placed in a vacuum chamber. The electron beam crosses the sample, where it is then focused by the objective lens to form an image. TEM can readily image soft-matter samples with a spatial resolution down to well below 1 nm in size (Henry, 2005; Kuntsche et al., 2011; Ruozi et al., 2011). This technique can be further categorized based on the sample-preparation method utilized – in particular, negative staining, freeze–fracture and cryogenic TEM. It can provide information on surface modifications of nanoparticles as it provides better contrast and contour of images than other microscopy techniques. For example, conjugation of transferrin to the surface of DSPC/cholesterol liposomes showed a particulate surface coating with negative-stain TEM, which was absent in the unconjugated liposome preparations (Anabousi et al., 2005). In a separate experiment, specific antibodies raised against human transferrin were added before the negative stain procedure to confirm the identity of the particulate surface coating (Anabousi et al., 2005). Therefore, TEM provides the possibility to achieve much-improved resolution when assessing the conjugation of ligands to the surface of liposomes.

#### Negative Staining Technique

In negative-stain TEM, sample preparation involves a small amount of hydrated sample being placed onto a grid. As the vesicles are in direct contact with the grid, their orientation and morphology may be affected, and hence this should be taken into account (Bibi et al., 2011). It is also necessary to embed the liposomes in a suitable electron-dense material (e.g., heavy metal salts like uranyl acetate or phosphotungstic acid) that provides high contrast, so vesicles can be viewed against a dark-stained background (Ruozi et al., 2011). The negativestaining technique is relatively fast and simple; however, it has been shown to cause changes to the original liposome structure and can lead to the creation of light and dark fringes that may be mistaken for lamellar structures (Bibi et al., 2011; Ruozi et al., 2011). In addition, placing the sample under vacuum can cause further dehydration of the sample, which can again cause changes in the structure of the vesicles (Bibi et al., 2011). Whilst the negative-stain TEM technique can provide much higher magnification for imaging nanoparticles, the damage to the liposome structure makes it difficult to accurately evaluate the morphological characteristics of the sample. Therefore, other imaging techniques are generally required to confirm results.

#### Freeze–Fracture Technique

The freeze–fracture technique does not require any drying process and can provide additional information about the internal structure of nanoparticles (Kuntsche et al., 2011). This method involves placing a sample on a TEM grid that is sandwiched between two copper or gold holders (Severs, 2007). The sample is vitrified via rapid freezing, typically with liquid propane or liquid nitrogen, before being fractured along areas of the sample with weak molecular interactions. This fracture surface can be further etched and shadowed with a thin platinum or carbon layer to provide a "negative" replica of the fracture sample plane (Severs, 2007; Kuntsche et al., 2011). The replica is then cleaned with an organic solvent to remove all residues

prior to visualization under a TEM microscope. As these replicas are so stable, they can be stored and viewed later (Kuntsche et al., 2011). The major advantage of this technique is that the replicas closely reflect the native state of the sample, and can provide detailed information on the 3D structure of the vesicles and bilayer organization (Bibi et al., 2011; Kuntsche et al., 2011). This information includes the multilamellar construction and bilayer packing of multilamellar vesicles. This technique can also assess aggregate size and may be particularly useful for examining the interaction of cationic liposomes with DNA (Bibi et al., 2011). However, artifacts may still occur in the sample during preparation due to insufficient freezing rate, re-deposition of solvent molecules and/or mechanical stress (Severs, 2007; Kuntsche et al., 2011). For example, a "rippling effect" can occur on the SUVs in the sample (Bibi et al., 2011). This is a common bilayer deformation that is due to a disorder in the transitions of the acyl chains prior to freezing. Incubating vesicles between the pre-transition and actual transition temperature can also cause ripples (Bibi et al., 2011). Interestingly, this can be used to provide information into the lipid phase transitions that occur with the varying nature of lipids used for liposomal preparations (Bibi et al., 2011).

#### Cryogenic TEM

Cryogenic TEM (cryo-TEM) is a variation of TEM where thin aqueous hydrated films which are vitrified in liquid ethane are used prior to imaging. This technique allows for the analysis of liposomes in their most native state, and is a valuable tool to determine the size, shape, internal structure, and lamellarity of liposomes (Weisman et al., 2004; Bibi et al., 2011; Kuntsche et al., 2011). The major advantage of rapidly freezing liposome samples is minimizing the formation of ice crystals and preserving proteins or other materials (Bibi et al., 2011; Kuntsche et al., 2011). This is beneficial where proteins or DNA have been encapsulated within the nanoparticles (Weisman et al., 2004; Kuntsche et al., 2011). Cryo-TEM is the most useful form of microscopy currently available to study liposomes, as it avoids issues with chemical fixation, dehydration, cutting and staining – all of which can affect the morphology of vesicles (Bibi et al., 2011). The resolution range is ∼5 to 500 nm, as defined by the thickness of the film (Almgren et al., 2000). However, several limitations should be noted with this technique, including the fact that only a 2D image is obtained from 3D objects (which generally also applies to most of the other microscopy techniques). To overcome this limitation, reconstruction of 3D shapes from a sufficiently large number of 2D images of randomly oriented non-spherical particles is possible (Orlova et al., 1999), as well as viewing the sample at different tilt angles to attain information about 3D shape (Van Antwerpen and Gilkey, 1994). Cryogenic electron tomography (cryo-ET) can also be used instead to attain 3D images (Le Bihan et al., 2009). Cryo-TEM also utilizes lower doses of electrons, which means that it often has a lower resolution compared to other methods (Bibi et al., 2011; Kuntsche et al., 2011). Artifacts are still possible due to the formation of a thin film of amorphous ice and the use of blotting on the sample applying shear forces during the film formation (Almgren et al., 2000; Bibi et al., 2011).

### Environmental Scanning Electron Microscopy (ESEM)

Environmental scanning electron microscopy (ESEM) is an imaging system that does not require the use of fixing, staining or freezing of vesicles, and can allow imaging of dynamic changes of wet systems without previous sample preparation (Muscariello et al., 2005; Ruozi et al., 2011). The main feature of ESEM is the presence of water vapor in the microscope chamber. The ability to maintain a water-containing atmosphere around the sample that may be partially or even fully hydrated is made possible by the use of a multiple-aperture, graduated vacuum system that allows the imaging chamber to be sustained at pressures up to 55 hPa (Bibi et al., 2011; Ruozi et al., 2011). The primary electron beam can generate secondary electrons that then encounter vapor molecules, leading to a cascade amplification of the signal before reaching the detector. Because of this, ESEM does not require sample preparation (Muscariello et al., 2005). This technique allows for variation in the sample environment through a series of pressure, temperature and gas compositions (Mohammed et al., 2004), which is useful when determining how environmental changes affect the vesicles. This is applicable to nanopharmaceutical formulation and stability studies (Bibi et al., 2011). ESEM has also been used to analyze drug loading into the bilayer of liposomes (Mohammed et al., 2004), as well as determining the size and shape of vesicles. A limitation of ESEM is that it cannot provide detailed information regarding the lamellarity and internal architecture of the nanoscale structures (Ruozi et al., 2011).

### ATOMIC FORCE MICROSCOPY (AFM)

Atomic force microscopy (AFM), also known as scanning-force microscopy (SFM), is a type of scanning probe microscope technique. It works by running a sharp tip attached to a cantilever and sensor over the surface of a sample and measuring the surface forces between the probe and the sample (Sitterberg et al., 2010). As the cantilever runs along the sample surface, it moves up and down due to the surface features and the cantilever deflects accordingly. This deflection is usually quantified using an optical sensor, with the laser beam being reflected on the back of the cantilever onto the light detector (Sitterberg et al., 2010). AFM does not need to operate in a vacuum and can operate in ambient air or under liquid; hence it is increasingly being used to image biological samples as well as nanoparticles (Liang et al., 2004a,b; Ruozi et al., 2005, 2009). AFM has outstanding resolution in the order of fractions of a nanometer and can provide a 3D image of liposomes along with details on morphology, size distribution, homogeneity, and stability (Liang et al., 2004a,b; Ruozi et al., 2005, 2007, 2009). Importantly, AFM can be used to characterize the surface modifications of liposomes and detect ligands (e.g., antibodies and polymers) conjugated at the liposomal membrane surface (Bendas et al., 1999; Moutardier et al., 2003;

Anabousi et al., 2005; Liang et al., 2005). For example, Bendas et al. used AFM to magnify the liposomal membrane border and were able to image trimeric structures, approximately 8–10 nm in diameter, which represented the coupling of IgG antibodies to the liposome surface (Bendas et al., 1999). The findings showed that the effectiveness of the technique was highly dependent on the conjugation method used, with antibodies conjugated directly to the liposomal surface being visible with AFM (restricted protein mobility), in comparison to antibodies attached to PEG chains. It was suggestive that the PEG chains caused high protein mobility and, therefore, were unable to be scanned (Bendas et al., 1999). Conversely, Anabousi et al showed that incorporation of PEGylated lipids into the liposomes induced a steric stabilization with liposomes maintaining a spherical shape (Anabousi et al., 2005). AFM images of the surface of unconjugated liposomes were smooth and no structures could be observed, whereas conjugation of transferrin to the surface of PEGylated liposomes were visualized as small globular structures (Anabousi et al., 2005). Similarly, Moutardier et al manufactured liposomes with polymeric cores (LSP) that consisted of drugs loaded into polymeric particles that formed the core of lipid vesicles (Moutardier et al., 2003). Images taken using AFM showed the presence of a polymer network on the exterior surface, which suggested that the collagen polymeric core radiated out and formed a surface layer on the LSP (Moutardier et al., 2003). In addition to surface structural details, AFM can provide information about the mechanical and chemical properties of a sample surface through force measurements (Ruozi et al., 2007). For example, Mao et al. used this technique to assess the elasticity and adhesive properties of liposomes (Mao et al., 2004). One limitation of AFM is the need for nanoparticles to be adsorbed onto support surfaces, such as mica or silicon wafers. The adsorption of liposomes onto a solid substrate has the potential to modify the size and shape of the vesicles, and cause their flattening. (Ruozi et al., 2007). In addition, the periodic contact of the probing tip can drag the liposomes as it moves across the vesicles in a sample (Jass et al., 2000). Despite this, AFM is still a useful tool in the evaluation of liposomes.

#### REFERENCES


### CONCLUSION

There is a wide range of imaging techniques available for evaluating the morphology of liposomes, with each having its own advantages and disadvantages. Light microscopy can provide general details regarding the size and shape of larger vesicles and the homogeneity of a sample in a relatively fast manner. Conversely, TEM is the most commonly used technique to examine the morphology of liposomes in much more detail; however, the potential for structural changes with each TEM sub-type due to staining and/or exposure to vacuum conditions need to be considered when interpreting the results. ESEM is most useful when determining liposomal changes in response to the environment, whereas AFM is emerging as a useful method in the morphological analysis of nanoparticles and provides maximum resolution of the liposomal surface. Both AFM and TEM are capable of imaging ligands conjugated to the surface of liposomes and provide complementary information on surface modifications. Overall, the choice of technique is dependent on what morphological characteristics and degree of detail are required. In addition, understanding the potential effects of the sample preparation method of each imaging technique is important in the selection process.

### AUTHOR CONTRIBUTIONS

Drafting of manuscript: A-LR and SH. Revising the article critically for important intellectual content: SH, PD, JF, WP, AM, DS and AW.

#### ACKNOWLEDGMENTS

The authors wish to thank The Pharmacy Research Trust of New South Wales, The Rebecca L. Cooper Medical Research Foundation, the Gladys M. Brawn Fellowship, ausEE Research Foundation, and The University of Newcastle for providing financial support for our research.



**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 Robson, Dastoor, Flynn, Palmer, Martin, Smith, Woldu and Hua. 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.

# Risk Assessment and Risk Minimization in Nanomedicine: A Need for Predictive, Alternative, and 3Rs Strategies

#### Lisa Accomasso<sup>1</sup> , Caterina Cristallini <sup>2</sup> and Claudia Giachino<sup>1</sup> \*

*<sup>1</sup> Department of Clinical and Biological Sciences, University of Turin, Turin, Italy, <sup>2</sup> CNR, Institute for Chemical and Physical Processes IPCF, Pisa, Italy*

The use of nanomaterials in medicine has grown very rapidly, leading to a concern about possible health risks. Surely, the application of nanotechnology in medicine has many significant potentialities as it can improve human health in at least three different ways: by contributing to early disease diagnosis, improved treatment outcomes and containment of health care costs. However, toxicology or safety assessment is an integral part of any new medical technology and the nanotechnologies are no exception. The principle aim of nanosafety studies in this frame is to enable safer design of nanomedicines. The most urgent need is finding and validating novel approaches able to extrapolate acute *in vitro* results for the prediction of chronic *in vivo* effects and to this purpose a few European initiatives have been launched. While a "safe-by-design" process may be considered as utopic, "safer-by-design" is probably a reachable goal in the field of nanomedicine.

#### Keywords: nanomaterial, nanomedicine, nanosafety, risk assessment, risk minimization

#### INTRODUCTION

Nanomedicine encloses many potential promises, ranging from optimized, targeted, and even personalized treatments with decreased toxicity, to very sensitive, and cheaper diagnostic approaches with contained costs, innovative functionalized biomaterials, and the prospect of cellular and tissue regeneration strategies (Wagner et al., 2006; Senjen, 2013).

Nanotechnology-based approaches have already been translated into highly accurate and sensitive diagnostic tests, most prominent of which is the early detection of neoplastic disease (Ferrari et al., 2009), targeted therapeutic interventions, following the statement that "targeted delivery will revolutionize disease treatment" (Duncan and Gaspar, 2011) and theranostic applications, having the possibility to combine in the same disease intervention both diagnosis (through imaging) and treatment (through thermal ablation) (Lammers et al., 2011).

Nanotechnological devices used in nanomedicine possess unique properties, not found in identical devices of smaller or larger dimensions, as they stem from their nanoscale dimension (Ferrari et al., 2009). However, a consensus has not been reached yet as for a scientific definition of "nanoparticle" (NP) (Satalkar et al., 2016) and as a consequence slightly different size limit definitions are applied in different fields of nanotechnology (Schütz et al., 2013).

In this review we first summarize the current state regarding safety evaluation of nano-based therapeutics and then we focus on the growing need for nanosafety studies for safer design of

#### Edited by:

*Susan Hua, University of Newcastle, Australia*

#### Reviewed by:

*Jonathan Shannahan, Purdue University, United States Kaisar Raza, Central University of Rajasthan, India*

\*Correspondence:

*Claudia Giachino claudia.giachino@unito.it*

#### Specialty section:

*This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology*

Received: *02 January 2018* Accepted: *28 February 2018* Published: *13 March 2018*

#### Citation:

*Accomasso L, Cristallini C and Giachino C (2018) Risk Assessment and Risk Minimization in Nanomedicine: A Need for Predictive, Alternative, and 3Rs Strategies. Front. Pharmacol. 9:228. doi: 10.3389/fphar.2018.00228* nanomedicines, including the employment of novel acute in vitro studies to extrapolate chronic effects that occur in vivo.

### RISK ASSESSMENT AND RISK MINIMIZATION IN NANOMEDICINE

The implementation of nanotechnology in medicine is a process that has occurred rapidly, suddenly moving from basic research and laboratory experimentation to clinical trials (Kola and Landis, 2004; Etheridge et al., 2013; Hafner et al., 2014). Many nanomedicine formulations have already entered the market. A recent study identified 247 nanomedicine applications and products, a very significant number, approved for or nearing in-human use (Etheridge et al., 2013) and it has been envisioned that the worldwide nanomedicine market may double by 2019 (BBCResearch, 2015). In the majority of cases, these artifacts are nanoformulations of current or novel drugs (58%) or nanobiomaterials (25%); however, nanotechnology has the potential to add innovative functionality to many pharmaceutical products and medical devices (Wagner et al., 2006). As for the regulatory aspects, it remains to be determined if nanomedicines fall into the category of medicinal products or medical devices. EU legislation makes clear distinction between the two, resulting in different regulatory approaches for risk assessment of medicinal products and medical devices. In the case of medicinal products, suitable clinical trials have to be performed prior to provision of a preliminary market authorisation for human use (Directive 2001/83/EC), while in the case of medical device market introduction requires lesser degrees of testing that can vary on the basis of the risk category the device falls into [Directive 2007/47/EC currently under revision: COM (2012) 542 final].

An elevated degree of unpredictability about prospective hazards and true advantages of NPs and nanomedicines, however, created remarkable obstacles along this translational pathway (Resnik and Tinkle, 2007). As such, nanotechnological approaches have opened up a few issues with respect to their proper risk assessment and risk minimization (Hogle, 2012), with particular emphasis on human and environmental toxicity (Allhoff, 2009; Ramachandran et al., 2012). Especially compelling in this respect are "first in human" (FIH) trials of nanotechnology medical applications, as they raise the highest degree of unpredictability in all clinical area (Kimmelman and John London, 2011). Especially important for FIH nanomedicine trials is the explicit description of the study purposes within the consent form. Today, the primary and sometimes exclusive aim of FIH trials is safety (King, 2012), even though innovative technologies like nanomedicines do not always fit the classical clinical trial phases that are followed when a new drug is developed. Thus, the information contained in the consent form should focus on these purposes and emphasize all aspects related to safety testing and risk identification (King, 2012).

Research in the field of nanomaterials (NMs) applied to medicine has continued to grow upon time, but it has been primarily focused around technological improvement, and not directed toward the definition of the potential risks of nanoproducts, thus nanosafety is an area that has remained poorly assessed. The fate of a nanomaterial upon entrance in the organism, whether it will be accumulated and become toxic, or rendered available at a biological level and transformed, or if and how it will interact with cells and macromolecules inside cells, are all crucial aspects that need to be understood. The conventional approaches listed in current test guidelines are not very likely to turn out as appropriate for the assessment of nanomedicine risks, rendering it urgent to develop NM-specific standards, guidelines, and tools. It is well-known, in fact, that often the bulk materials behave different in the nanometer regime and there is the need of generally accepted test methods for the characterization of nanomaterials. The methods currently existing can be not at all suitable to characterize nanomaterials for their specific properties. For example, when nanoparticles are dispersed in water, air or biological media they show a tendency to agglomerate and can even lose their nano-dimension. For this, it is important a careful and wide morphological, physico-chemical, and in vitro and in vivo biological characterization not only on the manufactured nanomaterials but also on nanomaterials after contact with relevant media. The evaluation of the effects of dispersion methods and of molecular interaction with biological components, cells and tissues on properties of nanomaterials represent a fundamental step for an effective control of nanomaterial risk. The degradation of the nanomaterials in the biological environment, the release of molecules or debris and the functionalization with organic substances could induce cytotoxic effects to be explored using methods already employed and validated in nanotoxicology but also improving the physicochemical characterization. QualityNano (www.qualitynano.eu; finished in 2015) has represented one of the first European initiatives along this line. It was an analytical research infrastructure addressing quality in NM safety assessment, through driving reliable and reproducible approaches to nanometrology and NM characterization before, during and after exposure to living systems. It included the development of standard operating protocols (SOPs) for analysis of the possible risks posed by NMs, focusing on assay reproducibility, use of appropriate positive and negative controls and controlling dose delivered to living systems (Senjen, 2013).

Assessment and management of risks, as well as risk communication are among the most challenging issues for nanomedicine clinical research (Resnik and Tinkle, 2007). Surely, our understanding and knowledge of different nanosized materials can be improved by single in vivo animal experiments and ex vivo laboratory testing, yet when a new nanomedicine product is to be tested in Phase I clinical trial they are not sufficient to resolve all of the uncertainty surrounding the first exposure of a human subject. What can be ethically accepted is that the risks potentially posed to human subjects by the new therapy are reasonable in relation to the new therapy potential benefits and, from a regulatory standpoint, that risks to the subject and society are minimized, wherever possible (Emanuel et al., 2000).

### PREDICTIVE TOXICOLOGICAL APPROACHES, ALTERNATIVE TEST STRATEGIES, AND 3RS APPROACHES

Extensive preclinical and clinical testing is needed prior to application of nanomedicine products in the three relevant areas of diagnosis, prevention, and treatment of disease, yet many aspects of NMs, including their toxicological, pharmacological, and immunological properties, have entered the scientific exploration only recently. Early safety studies are exceedingly needed to define whether the risk to benefit displayed by a specific nanomedicine is acceptable for the proposed use, thus determining if that nanotechnology will have the promise for further development in a clinical application (Duncan and Gaspar, 2011). Importantly, traditional approaches to toxicology, which are inherently descriptive in nature, will need to shift to predictive toxicology and this shift ascribes to both chemicals in general and NMs in particular (Oberdorster, 2010).

• Predictive toxicology is based on mechanism-based approaches relying on high-throughput screening (HTS) techniques. (i) The starting point is the generation of in vitro toxicity data resulting from the application of multiparametric, automated screening procedures. This in vitro phase of work may predict the possibility for disease or other pathological outcomes in vivo, based on the specific physicochemical properties of engineered NMs that are described (Nel et al., 2013b). (ii) The in vivo step will first of all validate the HTS techniques and will then improve them by establishing clear structure activity relationships. (iii) Then heat maps are developed on the basis of normalized data set, and self-organizing map features are exploited to organize all these information. (iv) Finally, appropriate combinations of both in vitro and in vivo approaches can be defined, with the final goal to establish hazard ranking and modeling. The landmark 2007 report from the US National Academy of Sciences, "Toxicity Testing in the Twenty-first Century: A Vision and a Strategy" (http://www.nap.edu/ catalog.php?record\_id=11970) is in agreement with this operation modality, clearly defining that a transition from qualitative laboratory testing and descriptive animal studies to mechanistic, quantitative testing funded on the employment of human cell types and high-throughput approaches will dramatically increase efficiency of toxicity evaluation (Nel et al., 2013b). Hazard assessment of large numbers of NMs can in this way be performed using pathways of toxicity (POTs), consisting in mechanism-based testing and representing the aim of the predictive toxicology approach (Nel, 2013). For the successful implementation of this methodology, careful selection of HTS techniques to be used in vitro as well as of POTs designed at the cellular level is needed, in order for them to reflect as many as possible pathogenic effects at the organism level. Although many methods and protocols were developed and validated for a predictive risk assessment of NMs, more work is needed regarding both physicochemical properties of NMs and their interaction with biological media. For example, the surface chemistry of the particles should be evaluated with particular attention considering that dissolution/dispersion and fate of NMs in biological media are affected by particle surface, surface charge, and radical formation potential.




points that depend on ad interim results and a critical aspect of the validation process becomes a correlation between in vitro and in silico with in vivo results. From an ethical and economical perspective it is not acceptable to test each NM in animals, thus a triage step based on an in vitro screening of these materials is necessary. Efforts in this direction are being made in the USA (Nel, 2013), but also in Europe collaborative initiatives are being created to increase open conveyance and sharing of results between different research groups. Overall, testing programme for regulatory purpose addressed a series of physico-chemical endpoints including methods and assays for a list of manufactured NMs, mostly inorganic carbon and metal oxide NMs (TiO2, SiO2, ZnO, CeO2). For different physico-chemical endpoints, several methods were used to evaluate: chemical composition (assay: EDX, ICP-OES ICP-MS, CHN elemental analysis), size [assay: DLS, TEM, SEM, AFM, (U)SAXS), WAXS], shape (TEM), coating (XPS, STEM-EDS, FTIR analysis of functional groups), surface area (VSSA, SAXS), Water solubility/dispersibility, cristallite size (XRD) (Rasmussen et al., 2018). In vitro studies addressed cytotoxicity, immunotoxicity, and genotoxicity testing using different cell culture models (i.e., blood, lung, placenta, brain, liver, gastrointestinal system) according to harmonized protocols.

The NanoTEST project (http://www.nanotest-fp7.eu/; finished in 2012) was one of the first examples. Many efforts were put in defining appropriate standard protocols, whose frequent lack represented an important problem experienced in testing NP potential hazards before clinical application (Juillerat-Jeanneret et al., 2015). A representative selection of commercial NMs currently or soon-to-be-applied in human medicine was investigated. To identify relevant short-term hazard models, the project used several standard toxicity assays for different markers such as cell viability, pro inflammatory response, oxidative stress, genotoxicity, immunotoxicity, cell uptake, and transport. Upon completion of the study, indications for full appraisal of NP toxicity included a few cytotoxicity measurements, a set of 2–3 representative cell types and five NP concentrations (Dusinska et al., 2015). NANoREG (www.nanoreg.eu; finished in 2017), FutureNanoNeeds (www.futurenanoneeds.eu; finished in 2017), and the ongoing NanoReg2 (www.nanoreg2.eu) are three other European projects aimed at defining a customary European strategy to the regulatory testing of fabricated NMs and at evolving an innovative frame to allow proper classification, better naming as well as hazard and environmental impact assessment of the future NMs before their extensive industrial employment.

### APPLICATION OF ALTERNATIVE TEST STRATEGIES TO RISK ASSESSMENT AND MINIMIZATION IN NANOMEDICINE

A number of different nano-specific, well-designed ATS are under development having the potentiality to provide answers to focused NM toxicity questions (Shatkin and Ong, 2016).

When the NM under study has an unknown toxicity, adoption of a Weight of Evidence (WoE) approach can be considered. Risk assessment is determined following careful hazard identification and prioritization taking into consideration and weighting all in vitro data available, both qualitative and quantitative, even in the absence of animal data. WoE mainly represents a methodological approach, where a collection of studies is analyzed based on expert opinions, systematic reviews or metaanalyses. When possible, quantitative WoE evaluations are also applied (Hristozov et al., 2014), for instance for prioritizing the riskiest occupational exposure scenarios that, in the case of NMs, can include processing methods, handling methods, length of time of exposure, protective equipment. They will be all considered and weighted accordingly, on the basis of quantitative data and expert judgement. Multiple techniques to test one end point should be applied to generate data with a high enough quality to allow for regulatory decisions to be taken based on WoE approaches.

Intelligent/Integrated Testing Strategies (ITS) are applied for the identification and prioritization of nanosafety research needs (Stone et al., 2014). Through this method, hypothesis-driven questions put to make risk decisions are answered by combining existing data, available analytical tools, experimental tests where the main goal is to avoid the need to test each developed NM. This strategy has been applied to accelerate the risk assessment process for materials of concern (Jaworska and Hoffmann, 2010) and benefits include reduced testing, consequently lowering the costs and limiting animal use; possibility to categorize NMs by (potential) mechanisms of action; applicability to a large assortment of testing strategies. Generally, ITS is based on of a stepped framework beginning with (i) an evaluation of existing data, that are organized using implements such as adverse outcome pathways (see below); (ii) measurement of chemical properties; (iii) biokinetic study of the NM; (iv) choice of suitable toxicity tests; and (v) employment of a WoE analysis that takes in consideration all the above results. Refinement steps can follow concerning both strategy and methods after which WoE is reevaluated (Oomen et al., 2014).

Conceptual frameworks such as adverse outcome pathways (AOP) are also been developed. Risk assessment is here performed through a sequential chain of events that are all causally linked and lead to an adverse outcome. Based on existing data, an initial molecular event is described and linked to a series of downstream key events acting at different biological levels (organism-cell-molecule) and eventually leading to the adverse outcome (Ankley et al., 2010).

All the above-mentioned approaches are needed in order to enable alignment of nanotoxicology with the 3Rs (Burden et al., 2017). The first step, establishment of a regulatory framework to enable implementation of alternative non-animal methods into risk assessment and acceptance, can benefit from WoE approaches to consider all available evidence from different non-animal methods. This will increase regulatory confidence in results from non-traditional methods, via guidelines and appropriate training, and will support risk assessors to understand the relevance and applicability of in vitro data for risk assessment and to adopt a rationale to deal with uncertainties and limitations inherent to experimental models (both in vitro and in vivo). The subsequent hazard prediction step will rely on ITS and adoption of a dual approach: hypothesis driven studies which test if a particular nanomaterial property impacts on toxicity, and studies which compare the toxicity of panels of nanomaterials. These parallel approaches will aim to accurately identify which NM properties confer toxicity and to establish a "reference data" for different endpoints for NMs which are deemed "representative" (dependent on the NM being studied) and the use of appropriate positive controls to relate the effects in vitro/in vivo. This involves ensuring that knowledge already in existence in other areas of toxicology is utilized to build knowledge within the discipline of nanotoxicology. After a validation step, AOPs frameworks can be exploited to adapt current standard in vitro approaches and to improve test item preparation, dosing, and understanding of toxicity mechanisms.

A rational design of the nanomaterials from the early phase of material selection, production method optimization, and product purification has to be considered of fundamental importance to prevent the safety issues of nanomaterial and increase their applicative potential. The concept of safer-bydesign emphasizes the importance of the contribution of more scientists such as engineers, chemists, physicians, and biologists to contrast the challenges of nanomaterials and satisfy the needs of the EU to regulate manufactured nanomaterials. The use of advanced analytical techniques (i.e., ICP, AFM, Chemical Imaging, biomarker detection) or their combination for the study of interactions between nanomaterials-relevant media in parallel with a better control of the preparation process will likely open up new scenarios in nanotoxicology testing (Dusinska et al., 2017; Oomen et al., 2018).

## CONCLUSIONS AND REMARKS

In nanomedicine, a proper risk evaluation in relation to health is unavoidable, in order to safeguard societal, ethical and regulatory acceptance, and public confidence. However, the individual testing approaches are limited and have turned out to be inadequate for nanotoxicology evaluations, thus risk assessment has needed to evolve to accommodate predictive toxicological analyses and ATS. The main short- to mediumterm objectives should include an improved comprehension of processes of interaction of NMs utilized in nanomedicine with organs, tissues, and cells and a clear strategy to tackle critical topics connecting to toxicity assessment specifically with respect to alternatives to tests on animals. Among the most compelling future objectives will be the need to test nanomedicines not only in healthy physiological environments, but also in disease environments that may alter biological responses and impact safety; in addition, to test nanoparticles not only individually but within complex mixtures, considering that nanoparticle incorporation into a variety of already utilized medical applications are likely to alter their risk profiles.

A safer-by-design concept has become increasingly important in risk assessment and minimization of nanomedicines, with the idea of integrating knowledge of NMs' potential adverse effects into the process of designing nanoproducts. This entails that nanomedicine safety is to be considered as an integrated route from the very first phases of research and innovation to the last phases of product validation, clearly different from the classical safety evaluation paradigm seeking to address potential concerns and to regulate NMs downstream, close to full product development, and market entrance.

### REFERENCES


### AUTHOR CONTRIBUTIONS

CC and CG contributed conception and design of the study; CG wrote the first draft of the manuscript; LA and CC wrote sections of the manuscript; LA contributed design of the table. All authors read and approved the submitted version.

### FUNDING

The support of M-ERANET (EU-funded network INCIPIT) and CRT Foundation (Grant 2016-079) is gratefully acknowledged.

### ACKNOWLEDGMENTS

We thank previous and current lab members for their valuable contributions.


**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 Accomasso, Cristallini and Giachino. 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.

# Current Trends and Challenges in the Clinical Translation of Nanoparticulate Nanomedicines: Pathways for Translational Development and Commercialization

#### Susan Hua1,2 \*, Maria B. C. de Matos <sup>3</sup> , Josbert M. Metselaar 4,5 and Gert Storm3,4,6

*<sup>1</sup> Therapeutic Targeting Research Group, School of Biomedical Sciences and Pharmacy, University of Newcastle, Callaghan, NSW, Australia, <sup>2</sup> Hunter Medical Research Institute, New Lambton Heights, Newcastle, NSW, Australia, <sup>3</sup> Department of Pharmaceutics, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht, Netherlands, <sup>4</sup> Department of Biomaterials Science and Technology, MIRA Institute for Biomedical Technology and Technical Medicine, University of Twente, Enschede, Netherlands, <sup>5</sup> Department of Experimental Molecular Imaging, RWTH University Clinic Aachen, Aachen, Germany, <sup>6</sup> Imaging Division, University Medical Centre Utrecht, Utrecht, Netherlands*

#### Edited by:

*Xinhua Qu, Shanghai Ninth People's Hospital, Shanghai Jiao-Tong University School of Medicine, China*

#### Reviewed by:

*Christoph Eugen Hagemeyer, Monash University, Australia Nazareno Paolocci, Johns Hopkins University, United States*

\*Correspondence: *Susan Hua Susan.Hua@newcastle.edu.au*

#### Specialty section:

*This article was submitted to Translational Pharmacology, a section of the journal Frontiers in Pharmacology*

Received: *31 May 2018* Accepted: *28 June 2018* Published: *17 July 2018*

#### Citation:

*Hua S, de Matos MBC, Metselaar JM and Storm G (2018) Current Trends and Challenges in the Clinical Translation of Nanoparticulate Nanomedicines: Pathways for Translational Development and Commercialization. Front. Pharmacol. 9:790. doi: 10.3389/fphar.2018.00790* The use of nanotechnology in medicine has the potential to have a major impact on human health for the prevention, diagnosis, and treatment of diseases. One particular aspect of the nanomedicine field which has received a great deal of attention is the design and development of nanoparticulate nanomedicines (NNMs) for drug delivery (i.e., drug-containing nanoparticles). NNMs are intended to deliver drugs via various mechanisms: solubilization, passive targeting, active targeting, and triggered release. The NNM approach aims to increase therapeutic efficacy, decrease the therapeutically effective dose, and/or reduce the risk of systemic side effects. In order to move a NNM from the bench to the bedside, several experimental challenges need to be addressed. This review will discuss the current trends and challenges in the clinical translation of NNMs as well as the potential pathways for translational development and commercialization. Key issues related to the clinical development of NNMs will be covered, including biological challenges, large-scale manufacturing, biocompatibility and safety, intellectual property (IP), government regulations, and overall cost-effectiveness in comparison to current therapies. These factors can impose significant hurdles limiting the appearance of NNMs on the market, irrelevant of whether they are therapeutically beneficial or not.

Keywords: nanomedicine, nanoparticles, drug delivery systems, clinical translation, challenges, commercialization, biological, regulations

### INTRODUCTION

Nanomedicine applies nanotechnology to highly specific medical interventions for the prevention, diagnosis, and treatment of diseases (Teli et al., 2010). In the last several decades, the application of nanotechnology for medical purposes has received significant attention from researchers, academia, funding agencies, government, and regulatory bodies (Allen and Cullis, 2004; Sercombe et al., 2015; Hare et al., 2017). One particular aspect of the nanomedicine field which has received a great deal of attention is the design and development of nanoparticulate nanomedicines (NNMs) for drug delivery (i.e., drug-containing nanoparticles), which are most often given by parenteral (particularly intravenous) administration. NNMs are intended to increase the therapeutic index of drugs (i.e., increase efficacy and/or reduce toxicity) by delivering them via various mechanisms: solubilization, passive targeting, active targeting, and triggered release (**Figure 1**). Nanoencapsulation gives the opportunity to protect fragile compounds that degrade easily in biological environments and to provide solubilization, i.e., to deliver compounds which have physicochemical properties that strongly limit their aqueous solubility and therefore systemic bioavailability (Talekar et al., 2015; Kim et al., 2016; Larsson et al., 2017; Mishra et al., 2017; Shajari et al., 2017). Targeted drug delivery and triggered release of NNMs have been shown to be beneficial for increasing the therapeutic index of compounds, by improving the in vivo fate of drug molecules such that more efficient delivery to the target site is achieved (to yield improved therapeutic effects) with less accumulation in many healthy body sites (to reduce toxicity). Also NNMs have been studied for their ability to stimulate target cell uptake and improve intracellular trafficking, processes sometimes required when they have localized in target tissues (Mastrobattista et al., 1999; Hua, 2013; Hua et al., 2015).

Although NNMs have demonstrated significant therapeutic advantages for a multitude of biomedical applications, their clinical translation has not progressed as rapidly as the plethora of positive preclinical results would have suggested (Luxenhofer et al., 2014). In order to move a NNM from the bench to the bedside, several experimental challenges need to be addressed. From a biological perspective, these include studies focused on understanding the in vivo fate and interactions of NNMs with the blood, tissue, cellular, and intracellular compartments in the host in healthy and diseased states (Nehoff et al., 2014; Sercombe et al., 2015; Hare et al., 2017). For NNMs to have clinical translation potential, the complexity in their design and development also needs to be minimized as much as possible to create systems that are able to be reproducibly prepared and characterized (Lammers, 2013; Barz et al., 2015). This review will address the current trends and challenges in the clinical translation of NNMs as well as the potential pathways for translational development and commercialization.

### CURRENT TRENDS IN THE CLINICAL TRANSLATION OF NANOMEDICINES

NNMs are often studied to improve drug targeting to specific sites of disease (i.e., site-specific drug delivery) and/or attenuate localization in healthy non-target tissues (i.e., site-avoidance drug delivery; Rizzo et al., 2013). The vast majority of NNMs in preclinical and clinical development as well in clinical use are for targeting a wide variety of cancers and tumors (Hare et al., 2017). The application of NNM-based therapies for drug targeting to non-cancer conditions has increased in recent years. In particular, NNMs have been developed to address the clinical challenge of effectively managing inflammatory diseases by exploiting the underlying biology of these conditions (Milane and Amiji, 2017). Non-cancerous inflammatory diseases that have been explored with NNM therapy include rheumatoid arthritis, inflammatory bowel disease, asthma, multiple sclerosis, diabetes, and neurodegenerative diseases (Milane and Amiji, 2017).

### Enhanced Permeability and Retention (EPR) Effect and Passive Accumulation of NNMs

The EPR effect refers to the preferential localization of NNMs in pathological tissues due to the enhanced permeability of the vasculature that supplies such tissues (e.g., tumors and inflammatory conditions). Deregulations in angiogenesis and/or the increased expression and activation of vascular permeability factors predominates at these sites, which can lead to fenestrations allowing passage of NNMs (Hashizume et al., 2000; Nehoff et al., 2014). In addition to the enhanced leakiness of tumorous and inflamed blood vessels, the EPR effect also relates to the observation that solid tumors tend to lack functional lymphatic drainage, which limits the removal of extravasated NNMs from the target site (Maeda et al., 2013; Danhier, 2016). These pathological properties allow NNMs to accumulate at pathological sites and is referred to as passive targeting. To achieve this, it is important that NNMs with drug cargo circulate long enough in the bloodstream (i.e., show prolonged circulation kinetics). This can be achieved by conjugating polyethylene glycol (PEG) to the surface of NNMs (**Figure 1**). Thus, the EPR effect is expected to increase the therapeutic efficacy of NNMs in comparison to small molecules, which often show inferior pharmacokinetic properties (Matsumura and Maeda, 1986; Hobbs et al., 1998; van der Meel et al., 2013). The EPR effect was first observed in 1986 (Matsumura and Maeda, 1986) and has since been exploited particularly for the development of NNMs for passive tumor targeting, leading to NNMs with adequate physicochemical properties and prolonged circulation half-life that accumulate in tumors over time (Maeda et al., 2013; Nakamura et al., 2015; Danhier, 2016). The EPR effect and thus extent of passive targeting is highly dependent on the tumor pathophysiology. Currently, it is recognized that EPR is a very heterogeneous phenomenon as it depends on the type of tumor and can vary significantly within the same tumor type (Lammers et al., 2012; Ojha et al., 2017). The degree of tumor vascularization and passive targeting of NNMs has been observed to be positively correlated (Theek et al., 2014). For example, Doxil <sup>R</sup> (pegylated liposomal doxorubicin) is the first FDA-approved NNM and has demonstrated superior efficacy in ovarian cancer and AIDS-related Kaposi's sarcoma compared to standard conventional therapies (Nichols and Bae, 2014). When doxorubicin is encapsulated within PEGylated liposomes, it delays and minimizes uptake and clearance by the reticuloendothelial system (RES), thereby prolonging circulation half-life. This allows the NNM to accumulate in the tumor tissue by exploiting the locally increased permeability of the tumor blood vessels, rather than in non-target healthy tissues which do not have such leaky vessels (Rahman et al., 2007). Furthermore, the use of pegylated liposomal doxorubicin avoids high plasma

therapeutic component, and may include a targeting element.

peak levels of free drug (Lyass et al., 2000) and significantly reduces the risk of cardiotoxicity by preventing doxorubicin release through the heart vasculature (Rahman et al., 2007).

#### NNMs and Active Targeting

Active targeting, also termed ligand-targeting or receptormediated targeting, involves the use of ligands (e.g., antibodies, peptides or sugar moieties) which are physically or chemically conjugated onto the surface of NNMs to facilitate localization to and/or uptake by target cells (van der Meel et al., 2013; Danhier, 2016; **Figure 1**). Ligand-targeted NNMs have enormous potential for site-specific delivery of therapeutic compounds to designated cell types in vivo, which selectively express or overexpress specific receptors (e.g., cellular receptors or cell adhesion molecules) at the site of disease (Willis and Forssen, 1998; Hua, 2013). For example, three sets of cellular targets are generally considered for active targeting in cancer—(i) targeting of cancer cells, which present overexpression of receptors for transferrin, folate, epidermal growth factor or glycoproteins; (ii) targeting of the tumor endothelium overexpressing vascular endothelial growth factors, integrins, vascular cell adhesion molecule-1 or matrix metalloproteinases; and (iii) targeting of stroma cells (e.g., macrophages, fibroblasts) that can acquire a tumor survivalpromoting phenotype in response to cytokines in the tumor microenvironment (Coimbra et al., 2010; Danhier et al., 2010; Kuijpers et al., 2010; Danhier, 2016). There is still much debate about whether ligand-targeted NNMs are capable of significantly enhancing NNM accumulation at target sites over non-targeted NNMs (passive-targeting), with conflicting results reported in the literature (Ferrari, 2005; Puri et al., 2009; Riehemann et al., 2009; van der Meel et al., 2013). Enhanced therapeutic effects have been demonstrated with ligand-targeted NNMs, despite showing no differences in accumulation in target tissues compared to non-targeted NNMs. For example, similar high levels of tumor tissue accumulation were achieved with both non-targeted liposomes and liposomes conjugated with HER2 monoclonal antibody fragments (7–8% injected dose/g tumor tissue) in HER2-overexpressing breast cancer xenografts models (Kirpotin et al., 1997, 2006). However, significantly superior therapeutic results was demonstrated with the doxorubicin-loaded anti-HER2 immunoliposomes in comparison to all other control groups, including recombinant anti-HER2 Mab trastuzumab, non-targeted liposomal doxorubicin, and free doxorubicin (Park et al., 2002). Differences in pharmacodynamics of the targeted NNM formulation in vivo was suggested as the reason for the improved anti-tumor effect, by enhancing intracellular drug delivery to HER2-overexpressing cancer cells (Kirpotin et al., 2006).

### NNMs for Triggered Release

A third targeting strategy based on stimuli-responsive NNMs, referred to as triggered drug release, is currently receiving much attention from academia and industry. This class of NNMs is designed with the goal of enhancing drug release in tumors by means of endogenous or exogenous stimuli. Endogenous stimuli-responsive NNMs exploit factors associated with the local environment at the site of disease (**Figure 1**). For example, low pH, presence of redox gradients or certain enzymes in the tumor microenvironment. Exogenous-responsive NNMs respond to external stimuli to trigger drug release, such as temperature, light, magnetic field or ultrasound. Of these strategies, the use of an external hyperthermic trigger to release therapeutic compounds from NNMs (e.g., thermosensitive liposomal doxorubicin, ThermoDox <sup>R</sup> ) appears to be the most promising to date (Needham et al., 2000). ThermoDox <sup>R</sup> was shown to be superior to its counterpart Doxil <sup>R</sup> in an in vivo model of non-resectable hepatocellular carcinoma (Torchilin, 2006; Sawant and Torchilin, 2012; Oude Blenke et al., 2013; Bertrand et al., 2014; Min et al., 2015; Jang et al., 2016; Shi et al., 2017). Thermosensitive liposomes are typically modified with temperature-sensitive lipids (e.g., distearoyl phosphocholine, DSPC) and/or polymers [e.g., poly(N-isopropylacrylamide)]. This composition allows the NNM to remain stable and retain their contents at physiologic temperatures, and undergo a phase change that makes them more permeable upon heating, thereby triggering the release of the cargo (Kono, 2001). The advantages of these NNMs can be further extended with the incorporation of imaging moieties (**Figure 1**) to enable monitoring of biodistribution, target accumulation and efficacy.

### NNMs Approved and in Clinical Trials

A number of NNM products are on the market with more in clinical development. The majority of NNMs in clinical development incorporate already approved drugs and are based on a variety of drug delivery platforms, including polymeric micelles, liposomes, dendrimers, and inorganic nanoparticles (Torchilin, 2006; Wagner et al., 2006; Sercombe et al., 2015). Despite the arsenal of nanoparticulate targeted systems currently under preclinical development or in clinical trials, it is indisputable that liposomes are dominant on the NNM market (**Table 1**) and were the first FDA-approved NNM (Caster et al., 2017; Shi et al., 2017). In fact, liposomes have all the necessary features to allow formulation of highly toxic and/or poorly soluble drugs, such as paclitaxel and amphotericin B (Min et al., 2015; Caster et al., 2017). Soon after their discovery in 1965 (Sessa and Weissmann, 1968; Deamer, 2010), liposomes were proposed as drug delivery vehicles for both small molecules as well as macromolecular drugs (Gregoriadis and Ryman, 1971; Gregoriadis et al., 1971). Years of research led to the development of the first FDA-approved NNM (Doxil <sup>R</sup> /Caelyx <sup>R</sup> ) as well as additional therapeutics (Allen and Cullis, 2013). Expectedly, many more NNMs are progressing to clinical investigation every year (**Table 2**), and again liposomal formulations represent the biggest share of the NNMs under clinical evaluation. The most frequently observed clinical benefit so far has been a reduction in toxicity with little evidence of improved efficacy. However, recently approved liposomal NNM, Vyxeos <sup>R</sup> (daunorubicin/cytarabine liposomal formulation), demonstrated improved survival and response rates, with tolerable toxicity in phase III clinical trials in older patients with therapy-related acute myeloid leukemia (t-AML) or AML with myelodysplasia-related changes (AML-MRC; Kim and Williams, 2018).

### CHALLENGES IN THE CLINICAL TRANSLATION OF NANOMEDICINES

The clinical translation of NNMs is an expensive and timeconsuming process. NNM technology is usually far more complex in comparison to conventional formulation technology containing free drug dispersed in a base (e.g., tablets, capsules and injections; Teli et al., 2010; Tinkle et al., 2014; Sainz et al., 2015). Key issues related to the clinical development of NNMs are listed in **Table 3**, and include biological challenges, large-scale manufacturing, biocompatibility and safety, intellectual property (IP), government regulations, and overall cost-effectiveness in comparison to current therapies (Allen and Cullis, 2004, 2013; Zhang et al., 2008; Sawant and Torchilin, 2012; Narang et al., 2013). These factors can impose significant hurdles limiting the appearance of NNMs on the market, irrelevant of whether they are therapeutically efficacious or not.

#### Biological Challenges

Traditionally, NNM development has been based on a formulation-driven approach, whereby novel delivery systems are firstly engineered and characterized from a physicochemical perspective. It is only when attempting to align the NNM with a pathological application that limitations in the clinical translation of the system have been identified. Understanding the relationship between biology and technology, including understanding the influence of disease pathophysiology on nanomedicine accumulation, distribution, retention and efficacy, as well as the biopharmaceutical correlation between delivery system properties and in vivo behavior in animals versus humans are important determinants for the successful translation of NNMs. Therefore, applying a disease-driven approach by designing and developing NNMs that are able to exploit pathophysiological changes in disease biology has been suggested to improve clinical translation (Hare et al., 2017).

From the outset in NNM development, it is essential to consider the relationship between disease pathophysiology and the heterogeneity of the disease in humans, and the importance of physicochemical characteristics of different NNMs to overcoming biological barriers to enable improved targeting to diseased tissue and/or reduced accumulation in non-target organs. Considerably less research effort has been dedicated to comprehensively understanding the correlations between NNM behavior and patient biology in specific clinical applications as well as disease heterogeneity in patients—which are likely


TABLE 1 | NNM formulations currently approved for marketing.

*(Ref: ema.europa.eu; drugs.com; fda.gov).*

the major reasons for the failure seen in the translation of promising NNMs in clinical trials (Hare et al., 2017). These biological challenges can be a significant deterrent for pharmaceutical industry investment into nanomedicines. In order to reduce investment risk for NNMs, the preclinical data sets need to comprehensively evaluate therapeutic efficacy, safety, biodistribution, and pharmacokinetics in appropriate animal models of the disease that are relevant to human disease. Evaluation of NNMs in multiple preclinical animal models that represent aspects of the clinical disease is preferred to achieve reproducibility of results for the specific disease and not for a specific animal model. In addition, animal models that reflect only a narrow spectrum of the clinical disease may provide useful data that can predict their suitability for treating a specific patient sub-group (Hare et al., 2017). Differences in the anatomy and/or physiology of the animal species compared to humans should be taken into account based on different routes of administration. Preclinical studies of NNMs should also be conducted under appropriate randomization and blinding to reduce bias, as well be evaluated against proper controls, including the gold standard treatment and not just free drug solution. These factors are currently lacking in many published studies, which makes it difficult to assess clinical applicability and translatability. Other considerations include designing preclinical studies to optimize NNM performance in vivo, dosing schedules, and treatment combinations based on the specific clinical disease, as well as understanding the influence of disease progression and severity on nanomedicine performance. This will determine whether specific patient sub-groups may respond more favorably to NNM-based treatment.

Interestingly, the majority of the NNM formulations in development and clinical trials are focused on cancer targeting, including more than 80% of the publications on nanomedicine in the last two decades alone (Park, 2017). Despite the large number of publications, the translation of the published studies to clinical applications has been disappointing. Cancer targeting of NNMs has generally been universally based on the EPR effect, despite the fact that EPR-mediated accumulation has only been reported for some tumor types (Maeda, 2015). Tumors, like other clinical diseases, can be highly heterogeneous and can show inter-patient and intra-patient variability as the disease progresses. Hence a one-size-fits-all approach when designing NNM-based treatment is unlikely to translate to clinically beneficial outcomes. The EPR effect has increasingly been exploited for NNM targeting in other non-cancer conditions, especially those involving an inflammatory component that causes leakiness of inflamed blood vessels (e.g., rheumatoid arthritis, atherosclerosis, and inflammatory bowel disease; Metselaar et al., 2003, 2004; Maiseyeu et al., 2009; Crielaard et al., 2012; Hua and Cabot, 2013; Hua et al., 2015; Milane and Amiji, 2017). It should be appreciated that not all diseases can be accessed with NNMs due to biological barriers and that the EPR effect is unlikely to be present in all clinical diseases. EPR is also not the only determinant of NNM efficacy. NNM activity is also influenced by the extent of cellular uptake and kinetics of drug release within target tissues (Hare et al., 2017).

Furthermore, the advantages of ligand-targeted NNMs in the clinical research phase have so far been negligible, despite the enhanced accumulation in target sites and therapeutic outcomes in a number of preclinical studies (Sercombe et al., 2015). Potential reasons for this discrepancy have previously been reviewed (Sawant and Torchilin, 2012; Allen and Cullis, 2013), and include factors such as target accessibility and expression, disease-dependent anatomical and physiological barriers, and formulation stability. In addition, the optimal targeting ligand density on the surface of each NNM has yet

#### TABLE 2 | NNM formulations in clinical trials.


*(Ref: clinicaltrials.gov).*

to be determined and will likely depend on characteristics of the molecular target (e.g., expression, location, internalization rate and immunogenicity; Puri et al., 2009; Hua and Wu, 2013; Kraft et al., 2014). Detailed analysis of the degree of NNM accumulation, cellular internalization, intracellular functionality and intracellular degradation will also be important considerations for clinical validation and translation (Puri et al., 2009). Through extensive experimentation, we are gaining a better understanding of the more appropriate clinical applications for ligand-targeted NNMs. Therefore, by taking a disease-driven approach to NNM development, it will be possible to build comprehensive preclinical data sets that best predict efficacy for patient sub-groups and support translatable clinical development.

#### Large-Scale Manufacturing

One of the important factors contributing to the slow pace in the clinical translation of NNMs is the structural and physicochemical complexity of the formulation itself. Platforms that require complex and/or laborious synthesis procedures generally have limited clinical translation potential, as they can be quite problematic to pharmaceutically manufacture on a large-scale (Teli et al., 2010; Tinkle et al., 2014; Barz et al., 2015; Sainz et al., 2015). Pharmaceutical manufacturing development is centered on quality and cost. Quality includes the manufacturing process and stability of the formulation, with NNM manufacturing being challenged by potential issues related to: (i) poor quality control; (ii) scalability complexities; (iii) incomplete purification from contaminants (e.g., by-products TABLE 3 | Considerations for the translational development of nanomedicines.

#### NANOPHARMACEUTICAL DESIGN

#### Key Considerations


#### PRECLINICAL EVALUATION

#### Key Considerations


#### Current Obstacles


#### CLINICAL EVALUATION FOR COMMERCIALIZATION

#### Key Considerations


#### Current Obstacles


and starting materials); (iv) high material and/or manufacturing costs; (v) low production yield; (vi) insufficient batch-to-batch reproducibility, consistency and storage stability of the final product (e.g., regarding size distribution, porosity, charge and mass); (vii) lack of infrastructure and/or in-house expertise; (viii) chemical instability or denaturation of the encapsulated compound during the manufacturing process; and (ix) scarcity of venture funds and pharmaceutical industry investment (Teli et al., 2010; Narang et al., 2013; Hafner et al., 2014; Tinkle et al., 2014).

An essential requirement for clinical translation is to have access to a preparation method that allows the production of large scalable quantities of NNMs, which is also consistently manufactured at the same high level of quality and batchto-batch reproducibility to set specifications (Grainger, 2013; Lammers, 2013; Barz et al., 2015). Suitable methods for the industrial scale production of several basic nanomedicine platforms, such as liposomes, have been successfully developed without the need for numerous manufacturing steps or the use of organic solvents (Jaafar-Maalej et al., 2012; Kraft et al., 2014). The challenges arise when the NNM system becomes more complex. For example, with the addition of surface modification with coatings and/or ligands, inclusion of multiple targeting components, or by encapsulating more than one therapeutic agent. Integration of multiple components into a single nanosized carrier requires multiple steps in the production process, which inevitably poses problems for largescale good manufacturing (cGMP) production, increases the cost of production, and makes the quality assurance and quality control (QA and QC) evaluation of such products more difficult (Teli et al., 2010; Svenson, 2012; Tinkle et al., 2014).

Characteristics of the manufactured NNM need to be welldefined and reproducibly generated to allow initiation of clinical translation. Chemistry, Manufacturing, and Controls (CMC) information is required for investigational new drugs (IND) at each phase of investigation to ensure proper identity, strength or potency, quality, and purity of the drug substance and drug product (FDA, 2003). The type of information submitted will depend on the phase of the investigation, the extent of the human study, the duration of the investigation, the nature and source of the drug substance, and the drug product dosage form (FDA, 2003). The characterization and validation of more complex NNMs can be particularly challenging due to the sheer number of parameters to address (e.g., size distribution, morphology, charge, purity, drug encapsulation efficiency, coating efficiency, and density of conjugated ligand/s; Teli et al., 2010). Batch-to-batch variation of NNMs can potentially lead to significant changes to their physicochemical properties (e.g., polarity and size), pharmacokinetic parameters (i.e., absorption, distribution, metabolism and excretion), and/or pharmacodynamic interactions (e.g., cellular interaction and activity; Teli et al., 2010; Tinkle et al., 2014; Barz et al., 2015). In addition, NNMs need to be stable after the manufacturing process, during long-term storage, and upon clinical administration (i.e., to avoid massive drug release or aggregation in the bloodstream en route to the site of action).

#### Biocompatibility and Safety

Detailed toxicology is essential for the clinical translation of NNMs to determine the overall safety for human use (Nystrom and Fadeel, 2012). Pharmaceutical regulatory authorities generally recommend that the sponsor carefully assess for any changes in the drug substance and drug product manufacturing process or drug product formulation at any phase of clinical development, in order to determine if the changes can directly or indirectly affect the safety of the product. CMC modifications throughout the IND process that can affect safety include: (i) changes in the synthetic pathway or reagents used to manufacture the drug substance, product or formulation; (ii) changes resulting in a different impurity profile; (iii) changes in the actual manufacturing method (e.g., chemical synthesis, fermentation, or derivation from a natural source); (iv) changes in the source material; (v) changes in the method of sterilization of the drug substance or drug product; (vi) changes in the route of administration; (vii) changes in the composition and/or dosage form of the drug product; (viii) changes in the drug product manufacturing process that can affect product quality; and (ix) changes in the drug product container closure system that can affect product quality (e.g., dose delivery; FDA, 2003). If any changes are identified, stringent procedures are in place to ensure appropriate comparison testing of the drug substance and/or drug product produced from the previous manufacturing process with the changed manufacturing process to evaluate product equivalency, quality, and safety (FDA, 2003). When analytical data demonstrate that the materials manufactured before and after are not comparable, sponsors should perform additional qualification and/or bridging studies to support the safety and bioavailability of the material to be used in the proposed trials (FDA, 2003).

Knowledge of the activity and toxicities of the free drug, the behavior of different NNM delivery systems and their interaction with biological components, and the influence of drug release rate on target and off-target concentrations of bioavailable drug allow the ability to predict potential side effects or toxicities in vivo (Hare et al., 2017). In particular, the rational design of NNMs from the early phase of material selection, production method optimization, and product purification is of fundamental importance to increase their clinical translation potential (Accomasso et al., 2018). Although the safety of some common materials such as phospholipids and biodegradable polymers have been studied previously (Storm et al., 1993), increasing the complexity of NNMs, such as the use of different synthetic compositions, coatings and ligands, can have a significant effect on the biocompatibility, biodistribution and toxicology profile of nanomedicines following in vivo administration (Allen and Cullis, 2004, 2013; Zhang et al., 2008; Sawant and Torchilin, 2012; Narang et al., 2013; Tinkle et al., 2014). For example, complement activation-related pseudoallergy (CARPA) is an acute adverse immune reaction caused by many NNMs (Szebeni, 2005; Sercombe et al., 2015; Szebeni and Storm, 2015; Jackman et al., 2016). The complement system is part of the innate immune response and is involved in a range of inflammatory and immunological processes (Moghimi and Hunter, 2001). CARPA is an immediate, non-IgE-mediated hypersensitivity reaction that can cause symptoms, including anaphylaxis, facial swelling, facial flushing, chills, headache, and cardiopulmonary distress (Szebeni, 2005). This adverse reaction is generally managed by slowing the infusion rate or ceasing therapy, as well as the use of standard allergy medications (e.g., antihistamines, corticosteroids and epinephrine; Sercombe et al., 2015; Szebeni and Storm, 2015). The development of immunogenic reactions to NNM-based therapies may lead to altered pharmacokinetics, loss of efficacy, and the rise of potentially serious toxicities (e.g., anaphylaxis; Szebeni and Moghimi, 2009; Szebeni and Storm, 2015).

There is a regulatory need for validated, sensitive and standardizable assays incorporating in vitro, ex vivo and in vivo protocols to appropriately assess the nanotoxicology of NNMs during the early stages of clinical development (Dobrovolskaia and McNeil, 2013; Jackman et al., 2016; Accomasso et al., 2018). Comprehensive in vitro or ex vivo assays for nanosafety testing are essential to screen for potential hazards prior to preclinical evaluation in animal models (Gaspar, 2007). For example, standardized in vitro protocols using different cell culture models (i.e., blood, liver, lung, brain, placenta, gastrointestinal system) to assess potential risk of cytotoxicity, immunotoxicity, and genotoxicity of NNMs (Accomasso et al., 2018). This is particularly important with the development of NNMs incorporating many new materials with the goal for use in the clinical setting. In order to do this effectively across the board, standardized reference materials would need to be established and the testing would also need to be relevant for the intended route of administration (Tinkle et al., 2014). Although current testing approaches are limited and insufficient for nanotoxicology evaluations for clinical translation, a number of techniques that are more specific for nanomedicines are under development. This includes alternative test strategies, high-throughput screening techniques, high-content screening, and computational modeling (Nel et al., 2013; Oomen et al., 2014; Dusinska et al., 2015; Accomasso et al., 2018). These techniques have the potential to analyze in a comparative way many NNMs simultaneously.

There is also a need to perform specialized toxicology studies in animal models to assess both short-term and longterm toxicity, as circulation half-lives and drug retention times are generally significantly increased with nanoencapsulation. A thorough understanding of the absorption, distribution, metabolism, and excretion of emerging nanomaterials in vivo is important to predict the toxicological responses to NNMs (Dobrovolskaia and McNeil, 2013; Tinkle et al., 2014). Adequate assessment protocols are needed to monitor various aspects of the NNM drug delivery process, including pharmacokinetics, biodistribution, target site accumulation, local distribution at the target site, localization in healthy tissues, kinetics of drug release, and therapeutic efficacy (Kunjachan et al., 2015). Incorporation of real-time imaging techniques have enabled better understanding of the interaction of NNMs with biological organs and tissues following in vivo administration (Gaspar, 2007; Nystrom and Fadeel, 2012; Dobrovolskaia and McNeil, 2013; Kunjachan et al., 2015).

In addition, biocompatibility, immunotoxicological, and inflammatory potential should be assessed, with functional outcomes correlated with mechanisms of tissue uptake and clearance (Gaspar, 2007). These parameters need to be wellinvestigated based on dose, dosage form and route of administration to establish safe limits prior to clinical trials (Gaspar, 2007; Nystrom and Fadeel, 2012). This is of particular importance for NNMs composed of materials that have never been used before in clinical applications. Even in the clinical trial phase, regulatory protocols should be in place to detect any toxicity caused not only by the encapsulated therapeutic compounds, but also novel mechanisms unique to nanotechnology (Gaspar, 2007; Nystrom and Fadeel, 2012). For example, short- and long-term effects of NNM accumulation in RES organs (esp. liver, kidneys, spleen, lungs, lymph nodes, and bone marrow; Senior, 1987; Szebeni and Barenholz, 2009; Szebeni and Moghimi, 2009), which are the main sites for NNM accumulation following systemic administration (Poste et al., 1976; Senior, 1987). The cells of the RES are also part of the innate immune system, which has raised concerns regarding whether macrophage saturation by NNMs can cause immunosuppression and increase the risk of infections (Sercombe et al., 2015; Liu et al., 2017). There have been no reports of clinically significant immunosuppression at therapeutic doses of non-cytotoxic NNMs, despite suggestions that excessive NNM deposition in macrophages may impair their phagocytic capacity or modulate other cellular functions (Szebeni and Barenholz, 2009; Szebeni and Moghimi, 2009). However, NNMs that contain cytotoxic compounds are capable of inducing macrophage destruction following uptake (Szebeni and Barenholz, 2009; Szebeni and Moghimi, 2009), with indirect signs that suggest the possibility of some immune suppression (Storm et al., 1998; Szebeni and Barenholz, 2009; Szebeni and Moghimi, 2009). For example, administration of Doxil <sup>R</sup> in mice was reported to interfere with the clearance of bacteria from the blood due to macrophage suppression (Storm et al., 1998; Szebeni and Barenholz, 2009). Addressing these issues are necessary to safeguard the application of emerging NNMs in the clinical setting.

#### Intellectual Property (IP)

Given the complexities of incorporating nanotechnology into biomedical and clinical applications, there needs to be more precise definitions on what constitutes novel IP of a nanomedicine (Satalkar et al., 2015). Nanomedicines are complex as they have a number of variable components, and bridge between the field of medicine and medical device (Paradise et al., 2009). Generally, the control of a NNM product requires an IP position on: (i) the encapsulated cargo; (ii) the carrier technology; and (iii) the characteristics of the drug and carrier together. Although this definition is straightforward, it does open up a number of problems with the issuing of patents to date (Bawa, 2007; Bawa et al., 2008). For example, NNMs that incorporate existing drugs with novel carrier technology, or those that incorporate existing drugs with existing carrier technology for a new biomedical or disease application. The IP situation becomes even more confusing with more complex drug delivery systems, such as those which incorporate commercially available targeting ligands (e.g., antibodies) or coatings (e.g., Eudragit <sup>R</sup> ) that are owned by other companies. IP strategies may likely involve multiple patents associated with any given technology and the need for cross-licensing arrangements (Murday et al., 2009). Therefore, new IP practices and protocols are required to simplify the pathway from invention to commercialization to reduce the time and expense required for negotiating collaboration and licensing agreements (Murday et al., 2009).

With the significant increase in the number of nanotechnology patent applications over the last few decades, other key issues that need to be addressed include patent review delays, patent thickets, and issuance of invalid patents (Bawa, 2005, 2007; Bawa et al., 2005). There needs to be a universal nano-nomenclature on identical or similar nanostructures or nanomaterials, and more refined search tools and commercial databases to avoid the issuing of multiple nanopatents on the same invention (Bawa et al., 2005; Bawa, 2007). Databases used by the Patent and Trademark Office (PTO) need to be able to search through nanotech-related prior art that resided in scientific publications world-wide, including earlier publications that preceded the emergence of online publication databases (Tinkle et al., 2014). Patent examiners also require expertise and training with respect to the emerging fields of nanotechnology and nanomedicine. The complexities with nanotechnology have led to the so called "patent thickets", which can lead to costly litigation and halt commercialization efforts (Tinkle et al., 2014). Therefore, improved clarity on IP and patenting surrounding nanotechnology in health and medicine is required, and will need to involve implementation of universal regulations and policies that are tailored toward this niche commercialization field.

### Government Regulations

Nanomedicines have significant potential to increase the growth of the pharmaceutical market and improve health benefits, however the current scientific and regulatory gap for nanomedicines is large and challenging. Commercialization of nanomedicines is highly dependent on a number of regulatory factors based on government policies in the area of manufacturing practice, quality control, safety, and patent protection (Gaspar, 2007; Tinkle et al., 2014; Sainz et al., 2015). The lack of clear regulatory and safety guidelines has affected the development of NNM products toward timely and effective clinical translation (Gaspar, 2007; Tinkle et al., 2014; Sainz et al., 2015). For example, polymers have been widely investigated as an effective platform for NNM strategies; however, their safety and efficacy is highly dependent on the polymer molecular weight, polydispersity, molecular structure, and conjugation chemistry (Gaspar and Duncan, 2009; Diab et al., 2012). Due to the increased number of novel polymeric materials and complex polymeric-based NNM formulations, there is an urgent need for an appropriate regulatory framework to assist in evaluation (Gaspar and Duncan, 2009). As each polymer-based NNM is different, it is important to consider each individually based on doses, administration routes, dosing frequency, and proposed clinical use. This would be the same for most other NNM platforms.

NNMs are currently regulated within the conventional framework governed by the key regulatory authority of each country (e.g., FDA, TGA, and EMA). Although NNMs have been on the market for nearly two decades, the first generation of NNM products passed regulatory approval by only having to meet general standards, applicable to medicinal compounds. These regulations are no longer appropriate to confirm the quality, safety, and efficacy of NNMs for clinical use (Gaspar, 2007; Tinkle et al., 2014; Sainz et al., 2015). Reasons for this are based on the complex structure of NNMs, their unclear interaction with cells and tissues within the human body, increased complexity of clinical use, and the multifunctional nature of some formulations (e.g., integration of therapeutics with imaging diagnostics; Gaspar, 2007; Tinkle et al., 2014; Sainz et al., 2015). Regulatory standards and protocols validated specifically for nanoparticles are needed that bridge both medicine and medical devices regulations. This should take into account a NNM's complexity, route of administration, pharmacokinetics, pharmacodynamics and safety profile, as well as provide information on the most appropriate clinical trial design and patient selection (Tinkle et al., 2014). There needs to be a fine balance to ensure the safety and quality of NNMs without over-regulation, which can negatively affect the progress of innovative products to the market, by escalating costs for achieving regulatory approval and/or consuming a significant portion of the life of a patent.

Development of global regulatory standards for NNMs should be established alongside key countries with invested interest. Although major steps have been taken in the last 5 years, a closer collaboration between regulatory agencies, academia, research and industry is needed (Gaspar, 2007; Murday et al., 2009; Hafner et al., 2014). This is of particular importance due to the limited availability of contract manufacturing organizations world-wide that specialize in producing NNM products in accordance with the requirements for good manufacturing practice (GMP; Hafner et al., 2014). It should be noted that this limited number of manufacturing organizations may be further divided based on their infrastructure capabilities of producing specific NNM platforms (e.g., liposomes, polymeric nanoparticles, dendrimers and drug-polymer conjugates). Therefore, NNMs produced in these manufacturing organizations will likely be marketed in multiple countries and thus should be governed under the same regulatory standards (Hafner et al., 2014). There will need to be complete evaluation and documentation of production processes for NNMs, incorporating appropriate industrial standards for both quality control and prevention of environmental issues (Gaspar, 2007). Manufactured NNMs will still need to meet general pharmaceutical standards such as purity, sterility, stability, manufacturing operations, and related industrial control standards (Gaspar, 2007). In addition, new analytical tools and standardized methods will need to be implemented to evaluate key physical characteristics of NNMs that can affect in vivo performance such as particle size and size distribution, surface chemistry, morphology, surface area, surface coating, hydrophilicity, porosity, and surface charge density (Gaspar, 2007; Tinkle et al., 2014; Sainz et al., 2015). These methods will vary for different nanomaterials and nanostructures. Thus, regulatory authorities should work together to develop the testing methods and appropriate standardized protocols for toxicity studies and regulatory requirements, which will be needed to ensure the efficacy and safety of current and emerging NNMs.

### PERSPECTIVES ON THE TRANSLATIONAL DEVELOPMENT OF NANOMEDICINES

From a therapeutic perspective, increasing drug accumulation at target tissues and minimizing systemic adverse effects are still the biggest design challenges to meet when developing new drug delivery systems. Even though promising NNMs may demonstrate significant efficacy in in vitro or ex vivo studies, it is important to evaluate the platforms in vivo using appropriate animal models of the disease. It is here where many of the current NNM platforms have shown limited specificity, accumulation and/or stability, therefore providing unsatisfactory results to warrant progression in the R&D process (Hua et al., 2015). Efficacy in an animal model also does not necessarily equate to efficacy in humans, as drug delivery within the human body is complex and can be highly variable, especially when associated with disease (Hare et al., 2017). Therefore, this concept of designing nanomedicines that act like a "magic bullet," which refers to the exclusive delivery of active compounds to specific organs, tissues or cells, is just not realistic when taking into account the pharmacokinetic and pharmacodynamic processes that occur following administration into the body (Barz et al., 2015). This term should refer to the development of realistic therapeutic platforms, in which therapeutic effects are maximized, doses are minimized, and complexity in dosage form design is reduced (Barz et al., 2015).

Complexity in dosage form design is a key factor in the ability for a NNM formulation to be translated to the clinic, irrelevant of its therapeutic efficacy. Simplification in formulation design is required to allow efficient and reproducible largescale manufacturing (Grainger, 2013; Lammers, 2013; Barz et al., 2015). Any added complexities to the basic NNM platform would need to show significantly improved benefits that is reliable and reproducible in animal models and patients, due to the added costs and complexity in the manufacturing process. For example, further studies are required to examine the benefits of ligand-targeted delivery systems over basic NNM platforms, in particular the reliability and consistency of the expression of the target across disease severity and in different patients (Hua et al., 2015; Sercombe et al., 2015; Hare et al., 2017). In addition, when translating findings from animal models to humans, we need to determine how to modify these formulations so that they are appropriate for human administration (Hua et al., 2015). In vivo studies are typically conducted in animal models of experimental diseases, especially in mice and rats, which can place limitations on the size and consistency of the dosage form that can be administered—for example, via oral, topical or intraperitoneal delivery (Hua et al., 2015; Sercombe et al., 2015). The practicability of designing dosage forms that are both acceptable to humans and efficacious should be further explored for clinical studies. Thus, there needs to be a balance between complexity, therapeutic efficacy, and clinical translation.

To transition NNMs to the clinic, attention should be given to nanosized carriers that are stable following in vivo administration, easily able to be up-scaled for manufacturing with high control over their physicochemical properties (e.g., size and polydispersity, morphology, drug encapsulation efficiency, and charge), as well as being composed of materials that are biocompatible, biodegradable, and non-toxic. As nanoparticles are able to enter cells and interfere with molecular pathways, synthetic polymers and lipids should be carefully evaluated for potential short-term and long-term toxicity for clinical application (Gaspar and Duncan, 2009). For example, potentially toxic in vitro and in vivo effects have been identified with the use of cationic polymers and lipids, including reduced number of mitoses, cell shrinking, detrimental effects on key cellular proteins (e.g., protein kinase C), and vacuolization of the cytoplasm (Lv et al., 2006).

### PATHWAY TO TRANSLATION AND COMMERCIALIZATION

The experimental development of NNMs is progressing at a fast pace, however significant challenges still exist in promoting these platforms into clinically feasible therapies (**Table 3**). The majority of NNMs in the clinic are for the treatment of cancer, predominantly by the parenteral route of administration. They are structurally based on simple nanomedicine platforms, in particular basic nanoparticles, surface charge-modified nanoparticles, and PEGylated nanoparticles (Hafner et al., 2014; Sainz et al., 2015). Although clinical applications of nanotechnology for non-cancer diseases are increasing based on promising experimental results, there are several barriers that have slowed progress in the preclinical and, especially, clinical stages of development. This includes issues surrounding complexity in manufacturing and characterization, lack of understanding of in vivo pharmacokinetics and pharmacodynamics, acute and chronic toxicity, and costeffectiveness (Gaspar, 2007; Teli et al., 2010; Hafner et al., 2014; Tinkle et al., 2014; Sainz et al., 2015). These challenges are even greater with increasing complexity of the NNM design.

The pace for the clinical translation of NNMs has been relatively slow as the development trajectory is very costly, complex and time-consuming, which has affected the attitudes of the pharmaceutical industry and capital investors. There has to be a clear positive benefit-to-risk ratio that will accompany the clinical implementation of products and procedures based on nanotechnology. In particular, the cost-benefit analysis may be a limitation to the clinical translation of some NNMs when compared to an approved counterpart or existing therapies. This analysis has to be clear before starting the development process. Emerging NNM products, which are more complex in structure and more expensive than conventional therapies, need to provide an overall reduction in health care costs and provide a worthwhile opportunity for the pharmaceutical industry to invest its R&D budgets (Hafner et al., 2014). This reduction in health care costs is likely to be obtained by increasing therapeutic efficacy, improving quality of life, reducing adverse effects or toxicities in non-target organs, and/or reducing the need for surgical or other high-risk interventions (Gandjour and Chernyak, 2011). Nanopharmaceuticals can offer the ability to extend the economic life of proprietary drugs and create additional revenue streams (Tinkle et al., 2014). In addition, market analysis, investment risk, potential profit margins, and value proposition of novel NNMs are important factors for the pharmaceutical industry and investors. Typically, pharmaceutical products that are developed to address larger disease populations with treatment expected in a primary or secondary care setting are preferred by the pharmaceutical industry. From a business perspective, the necessary infrastructure, understanding of NNMs, and skill set required for the commercial development of NNMs are not currently well represented at most pharmaceutical companies. These factors should be taken into account when assessing the overall cost-effectiveness of NNMs in comparison to existing therapies.

Nanomedicines generally face a number of regulatory approval hurdles. The control of materials in the nanosize range often presents greater scientific and technical challenges compared to conventional formulations (Gaspar, 2007; Teli et al., 2010; Hafner et al., 2014; Tinkle et al., 2014; Sainz et al., 2015). NNMs encompass a number of different types of nanomaterials and nanostructures, which make it even more challenging to establish appropriate regulatory protocols and tools to ensure standardized GMP manufacturing and characterization, safety and toxicology evaluation, and clinical trial design. These procedures are paramount to confirming therapeutic efficacy and safety prior to marketing approval for use in patients on a larger scale. Effective clinical translation will require an interdisciplinary approach to develop novel protocols, assays and infrastructure for the manufacturing and characterization of NNMs (Gaspar, 2007; Teli et al., 2010; Hafner et al., 2014; Tinkle et al., 2014; Sainz et al., 2015). This will need to involve experts from academia and industry with specialty in pharmaceutics, engineering, biology, medicine, and toxicology. Potential approaches to fast-track promising novel NNMs to clinical trials include the establishment or coordination of laboratories and centers that have expertise in (i) characterizing NNM platforms, (ii) conducting preclinical studies on NNMs for submission to regulatory agencies, (iii) scale up laboratory preparation of nanomaterials according to regulatory and industry standards for early clinical trials, and (iv) designing and conducting clinical trials of NNM platforms (Hafner et al., 2014).

## CONCLUSION

Overall, the use of nanotechnology in medicine has the potential to have a major impact on human health. It has been suggested to facilitate the development of personalized medicine for specific patient sub-groups, in which therapy is tailored by the patient's individual genetic and disease profile (Teli et al., 2010; Mura and Couvreur, 2012; Laroui et al., 2013). For example, diseasespecific characteristics such as capillary permeability (Calcagno et al., 2015), cellular receptor expression and molecular pathway activation could be analyzed and used to design personalized nanomedicines (Teli et al., 2010; Mura and Couvreur, 2012; Laroui et al., 2013). The physicochemical properties (e.g., size and structure) of the delivery system can also be modified according to the severity of the disease for optimal therapeutic benefits (Hua et al., 2015). This concept would significantly advance the way in which we treat patients. However, for this to occur, there are still a number of issues that need to be addressed as detailed in this review—from our basic understanding of the biology of specific diseases and the biological interaction of NNMs in patients, to commercialization hurdles related to manufacturing, costs, and regulatory standards. Finally, researchers need to consider minimizing the complexity of NNMs and take into account the final dosage form for human use, in order for a formulation to have the potential to be translated into a clinically applicable therapeutic. Reducing complexity to the minimum required for pathophysiological or medical need is paramount in nanoparticle design and synthesis to generate clinically translatable nanosized therapeutics.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### REFERENCES


#### ACKNOWLEDGMENTS

The authors wish to thank The Pharmacy Research Trust of New South Wales, The Rebecca L. Cooper Medical Research Foundation, Gladys M Brawn Fellowship, ausEE Research Foundation, and The University of Newcastle for providing financial support for our research.

nanomedicine: lessons from NanoTEST. Nanotoxicology 9 (Suppl. 1), 118–132. doi: 10.3109/17435390.2014.991431


**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 Hua, de Matos, Metselaar and Storm. 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.

# The Advances and Challenges in Utilizing Exosomes for Delivering Cancer Therapeutics

Mengliu Yang and Sherry Y. Wu\*

*School of Biomedical Sciences, The University of Queensland, Brisbane, QLD, Australia*

Keywords: exosomes, cancer therapeutics, delivery system, targeted delivery, gene therapy

Nanotechnology plays an important role in advancing treatment and diagnosis of a variety of human diseases. The use of nanocarriers often leads to better pharmacokinetic and safety profiles as well as enhanced bioavailability of the entrapped molecules. Several nanoparticle formulations have already been approved by Food and Drug Administration [Doxil (1995), onivyde (1996), Abraxane (2005)] or have advanced into clinical trials (Andre et al., 2002; Von Hoff et al., 2016; Subbiah et al., 2018). These particles are typically synthesized using lipids or polymers as these materials offer significant protection against degradation from serum nucleases and proteases. Targeting ligands can also be attached to the surface of these systems with ease to enable targeted delivery. However, the reliance of single targeting ligand may not be suitable for treatment of cancer as cancer cells could quickly adapt and change surface receptor expression profiles (Wu et al., 2014). To overcome this problem, recent research has focused on using naturally occurring exosomes to deliver therapeutic cargos. Exosomes, being natural transporters, offer significant advantage for cancer treatment as the surface of exosomes are decorated with numerous ligands that can be beneficial for preferential tumor targeting. Expression of selected ligands can also be enriched through molecular engineering (Alvarez-Erviti et al., 2011).

Certain types of exosomes have been demonstrated to have higher drug delivery efficiency when compared to commonly used nanocarriers (Kim et al., 2016), thus making them ideal candidates for delivering cancer therapeutics. Due to their favorable characteristics including superior targeting capability and safety profile, they are now being investigated as an emerging class of cancer therapeutics in several clinical trials with two trials already entering phase II testing (Besse et al., 2016) (NCT01854866). For instance, dendritic cell (DC)-derived exosomesloaded with tumor antigens have been used to vaccinate cancer patients with the goal of enhancing anti-tumor immune response (Escudier et al., 2005; Morse et al., 2005; Besse et al., 2016). While enhancing T cell response is yet to be achieved in these clinical studies, significant improvement in NK cell activity was observed (Morse et al., 2005; Besse et al., 2016). In addition to targeting immune cells, exosomes have also been used to target tumor cells directly (Kamerkar et al., 2017). The ideal delivery characteristic of exosomes is, in part, due to their surface protein expression profile (e.g., CD47), which allows evasion from phagocytosis by circulating monocytes (Kim et al., 2012; Kaur et al., 2014; Kamerkar et al., 2017). As the result, exosomal nanoparticles have increased circulatory half-life that is beneficial for tumor-targeting (Kim et al., 2016). Despite the promise, the development of exosomal delivery system is still in its infancy, with three major problems requiring further investigation: (1) isolation and purification of exosomes, (2) drug and antigen loading into exosomes, and (3) delivery of cargos to target cells. Here, we discuss recent progress in overcoming these challenges (**Figure 1**).

Lack of an efficient standardized isolation and purification method is a major challenge for bringing exosome technology into the clinic. It has been reported that exosomes can be isolated and purified by single or combinations of different methods, including immunoaffinity capture, size exclusion, polymeric precipitation, ultracentrifugation, microfluidics techniques, and

#### Edited by:

*Umesh Gupta, Central University of Rajasthan, India*

> Reviewed by: *Sandor Kerpel-Fronius, Semmelweis University, Hungary*

> > \*Correspondence: *Sherry Y. Wu sherry.wu@uq.edu.au*

#### Specialty section:

*This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology*

> Received: *06 April 2018* Accepted: *18 June 2018* Published: *16 July 2018*

#### Citation:

*Yang M and Wu SY (2018) The Advances and Challenges in Utilizing Exosomes for Delivering Cancer Therapeutics. Front. Pharmacol. 9:735. doi: 10.3389/fphar.2018.00735*

commercially available kits (Kim et al., 2016; Yim et al., 2016; Bellavia et al., 2017; Kamerkar et al., 2017). Combination of ultrafiltration and ultracentrifugation techniques has been used to generate clinical grade exosomes (Lamparski et al., 2002). An example of this preparation procedure includes concentrating exosomes by ultrafiltrating DC culture media or ascites fluid, followed by ultracentrifugation onto a sucrose/D2O density cushion to eliminate non-exosome proteins and to further reduce sample volume (Escudier et al., 2005; Morse et al., 2005; Dai et al., 2008; Besse et al., 2016). This isolation method resulted in generation of exosomes that can be safely administered into patients with minimal toxicity (Escudier et al., 2005). However, due to the complexity and labor intensiveness of this isolation strategy, new methods have been developed to further facilitate future clinical development of exosomal nanosystem. A recent study has described a unique approach to improve the efficiency of exosome isolation procedure. The researchers successfully isolated large number of transferrin receptorexpressing exosomes from reticulocytes, by incubating fresh serum with transferrin-coated superparamagnetic nanoparticles and separating exosomes by magnetic adhesion (Qi et al., 2016). As these exosomes exhibit superparamagnetic behavior with a strong response to an external magnetic field, they can be efficiently separated from the blood. This technology, combined with recently developed nanoscale flow cytometry (Morales-Kastresana et al., 2017), could facilitate the isolation of purified exosomes in large scale and be applied to other ligand of interest. Purification can be achieved through staining the exosomes using Carboxyfluorescein Succinimidyl Ester (CFSE) or other fluorophores followed by size exclusion chromatography and nanoFACS analysis. NanoFACS offers multi-parametric scattered light and fluorescence imaging of exosomes with high resolution and high sensitivity. Its use permits efficient assessment and enhancement of exosome purity.

In addition to developing reliable isolation and purification methods, researchers have recently developed new strategies to load cargos into exosomes. A decade ago, two independent research groups described exosomes as carriers of information and demonstrated their ability to transfer information from one cell type to another (Ratajczak et al., 2006; Valadi et al., 2007). This fundamental concept prompted many researchers to investigate the use of exosomes as a delivery system. Currently, three major types of drug loading strategies have been investigated: incubation (Qi et al., 2016; Bellavia et al., 2017), electroporation (Kamerkar et al., 2017), and sonication (Kim et al., 2016). The most widely used technique for generating cargo-containing exosomes for clinical testing is incubation. For instance, loading of antigens into exosome can be achieved through incubating antigens directly with conditioned DC-culturing media (the source of exosomes) or purified exosomes isolated from the culture media (Escudier et al., 2005; Morse et al., 2005; Besse et al., 2016). While this is a convenient method to load antigen or drug of interest into exosomes, it is hard to precisely control loading efficiency. A recent study has compared three methods of loading, namely incubation, electroporation, and sonication, using paclitaxel as a model molecule. It was shown that a loading efficiency of 29% could be achieved with the sonication approach while 1.5 and 5.3% were achieved for the incubation and electroporation methods, respectively (Kim et al., 2016). However, it must be noted that the sonication method resulted in slight particle aggregation. Thus, development of strategies to overcome the aggregation problem along with further improvement in loading efficiency are critical for future development of exosomal nanotechnology. In addition to traditional methods of drug loading, Yim and colleagues have recently reported a novel loading approach utilizing optically reversible protein-protein interaction (EXPLORs) technology (Yim et al., 2016). The researchers conjugated cargo proteins, mCherry, to photoreceptor cryptochrome 2 (CRY2) and induced their uptake into exosomes by overexpressing tetraspanin protein CD9 conjugated CRY-interacting basic-helix-loophelix1 (CIB1) in exosomes. The interaction between CRY2 and CIB1 was facilitated by blue light illumination. Application of this technology for nucleic acids loading along with strategies to enhance the stability of the resultant particles in vivo would be an exciting next set of challenges.

Another area of intense research is the development of better methods to enhance targeting ability of exosomal nanoparticles for cancer treatment. By utilizing normal human foreskin fibroblast-derived exosomes, Kamerkar and colleagues recently demonstrated the ability of exosomes to efficiently deliver KrasG12D siRNA to target undruggable oncogenic Kras in pancreatic tumor cells in vivo (Kamerkar et al., 2017). This resulted in diminished oncogenic KrasG12D expression, suppression of cancer cell proliferation, and an increase in overall survival in a mouse model of pancreatic cancer (Kamerkar et al., 2017). It was shown that fibroblast-derived exosomes display favorable protein expression profile on their surface which enabled efficient tumor targeting. These exosomal particles could then be taken up by tumor cells via Ras-induced micropinocytosis. In addition to the presence of naturally occurring ligands on exosomes surface, other researchers have also molecularly engineered exosome-producing cells to enrich the presence of particular ligand(s) on exosome surface in order to target a specific cancer type. For instance, Lamp2b-IL-3 expressing exosome was developed to target chronic myeloid leukemia (CML) cells preferentially as they overexpress IL-3 receptors (Bellavia et al., 2017). Utilizing this molecularly

#### REFERENCES


engineered system, Bellavia and colleagues successfully delivered BCR-ABL siRNA to CML cells, making them more sensitive to imatinib therapy in a CML mouse model. This technology could be applied for treatment of other IL-3 receptor overexpressing cancer types, such as lymphoma and acute myeloid leukemia. In addition to molecular methods to enhance targeting ability of exosomes to metastatic tumors, mechanical methods utilizing superparamagnetic nanoparticle-entrapped exosomes in combination with magnetic field at the tumor sites have also been developed to enhance tumor targeting. Using these superparamagnetic exosomes, Qi and colleagues successfully delivered doxorubicin to suppress tumor growth in a subcutaneous mouse model of liver cancer (Qi et al., 2016). This unique technique has provided a new approach to enhance the targeting ability of exosomes to localized tumors.

In summary, recent research has made significant progress in overcoming major barriers for using exosomes as a delivery system. Exosomes are ideal systems for delivering cancer therapeutics, owing to their size, surface expression profiles, low immunogenicity, low cytotoxicity, and long-term safety. Their use has opened a new promising avenue for cancer treatment. Scaling up the production of highly targetable therapeutic exosomes that can be used off-the-shelf which does not require generation from autologous source will be the next critical challenge to bring this promising delivery technology into the clinic.

#### AUTHOR CONTRIBUTIONS

SW initial conceptualization of the article. MY and SW wrote the manuscript.

#### ACKNOWLEDGMENTS

SW is supported by NIGMS (R35GM124708), Ovarian Cancer Research Fund Alliance (OCRFA), and Australian Ovarian Cancer Research Foundation.


exosomes derived from dendritic cells. J. Immunol. Methods 270, 211–226. doi: 10.1016/S0022-1759(02)00330-7


**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 Yang and Wu. 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.

# Biologically Targeted Magnetic Hyperthermia: Potential and Limitations

David Chang1,2,3, May Lim<sup>4</sup> , Jeroen A. C. M. Goos 5,6, Ruirui Qiao<sup>5</sup> , Yun Yee Ng<sup>4</sup> , Friederike M. Mansfeld1,3,5, Michael Jackson<sup>2</sup> , Thomas P. Davis 5,7 and Maria Kavallaris 1,3 \*

*<sup>1</sup> Children's Cancer Institute, Lowy Cancer Research Centre, University of New South Wales, Sydney, NSW, Australia, <sup>2</sup> Department of Radiation Oncology, Nelune Comprehensive Cancer Centre, Prince of Wales Hospital, Sydney, NSW, Australia, <sup>3</sup> ARC Centre of Excellence in Convergent Bio-Nano Science and Technology and Australian Centre for Nanomedicine, University of New South Wales, Sydney, NSW, Australia, <sup>4</sup> School of Chemical Engineering, University of New South Wales, Sydney, NSW, Australia, <sup>5</sup> ARC Centre of Excellence in Convergent Bio-Nano Science and Technology, Monash Institute of Pharmaceutical Sciences, Monash University, Melbourne, VIC, Australia, <sup>6</sup> Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States, <sup>7</sup> Department of Chemistry, University of Warwick, Coventry, United Kingdom*

#### Edited by:

*Susan Hua, University of Newcastle, Australia*

#### Reviewed by:

*James Hainfeld, Nanoprobes, United States Arkadiusz Jozefczak, Adam Mickiewicz University in Poznan, Poland*

\*Correspondence: *Maria Kavallaris m.kavallaris@ccia.unsw.edu.au*

#### Specialty section:

*This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology*

Received: *16 April 2018* Accepted: *10 July 2018* Published: *02 August 2018*

#### Citation:

*Chang D, Lim M, Goos JACM, Qiao R, Ng YY, Mansfeld FM, Jackson M, Davis TP and Kavallaris M (2018) Biologically Targeted Magnetic Hyperthermia: Potential and Limitations. Front. Pharmacol. 9:831. doi: 10.3389/fphar.2018.00831* Hyperthermia, the mild elevation of temperature to 40–43◦C, can induce cancer cell death and enhance the effects of radiotherapy and chemotherapy. However, achievement of its full potential as a clinically relevant treatment modality has been restricted by its inability to effectively and preferentially heat malignant cells. The limited spatial resolution may be circumvented by the intravenous administration of cancer-targeting magnetic nanoparticles that accumulate in the tumor, followed by the application of an alternating magnetic field to raise the temperature of the nanoparticles located in the tumor tissue. This targeted approach enables preferential heating of malignant cancer cells whilst sparing the surrounding normal tissue, potentially improving the effectiveness and safety of hyperthermia. Despite promising results in preclinical studies, there are numerous challenges that must be addressed before this technique can progress to the clinic. This review discusses these challenges and highlights the current understanding of targeted magnetic hyperthermia.

Keywords: magnetic hyperthermia, targeted therapy, iron oxide nanoparticles, cancer therapy, magnetic nanoparticles

### INTRODUCTION

Hyperthermia, a treatment aimed at raising the temperature of cancerous regions of the body to 40–43◦C, can induce cancer cell death by enhancing the cytotoxic effects of radiotherapy and chemotherapy (Wust et al., 2002). Extensive preclinical and clinical research into the application of hyperthermia has been conducted, with a number of randomized trials demonstrating that, when combined with radiotherapy, it has the potential to improve the outcomes of various cancers without significantly increasing toxicity (De Haas-Kock et al., 2009; Lutgens et al., 2010). Despite these promising results, hyperthermia is rarely incorporated into modern oncological management due to its ineffectiveness when applied as a single modality treatment and a lack of large phase 3 clinical trials combining hyperthermia with both standard chemotherapy and radiotherapy (van der Zee, 2002). Furthermore, a major disadvantage of conventional hyperthermia is that, in general, both malignant and non-malignant cells are equally sensitive to heating (Dewey et al., 1977; Roizin-Towle and Pirro, 1991). This is in contrast to chemotherapy or radiotherapy, which are generally more cytotoxic toward malignant cells. As a result, there has been significant interest in the concept of "biologically targeted magnetic hyperthermia," whereby targeted magnetic iron oxide nanoparticles (MIONs) are administered intravenously in order to heat tumors under an alternating magnetic field. In this review, we will discuss the current understanding of targeted magnetic hyperthermia and the limitations that must be overcome for further progression into clinical practice.

### HYPERTHERMIA AND CELL DEATH

Hyperthermia can cause cell death through a range of different mechanisms and there are no consistent differences in thermal sensitivities between malignant and non-malignant cells (Dewey et al., 1977; Roizin-Towle and Pirro, 1991). It has been shown in vitro that cell viability following hyperthermia treatment is heavily influenced by both the temperature and the duration of hyperthermia (**Figure 1**). Even half a degree rise in temperature can have a substantial impact on cell viability, highlighting the importance of effective and homogenous delivery of hyperthermia (Dewey et al., 1977). One of the possible mechanisms behind the reduction in cell viability is protein denaturation with subsequent activation and deactivation of several downstream pathways (van der Zee, 2002; Wust et al., 2002). Individual proteins have specific temperature thresholds for denaturation, with highly expressed proteins generally being more tolerant to heat (Leuenberger et al., 2017). Protein denaturation occurs from approximately 40◦C and higher temperatures will denature a greater proportion of proteins, which may explain why the rate of cell death rises with the temperature (Lepock, 2005b). At temperatures of 40–42◦C, only a small fraction of proteins is denatured, however, some of these can subsequently co-aggregate with native proteins, thereby significantly increasing the level of aggregation (Borrelli et al., 1996). It is this combination of heat-induced denaturation and subsequent co-aggregation that is thought to affect several downstream pathways including inactivation of protein synthesis, cell cycle progression and DNA repair (Dewey et al., 1977; Kampinga et al., 2004; Lepock, 2005a). Furthermore, possibly through a mechanism that is unrelated to protein denaturation, hyperthermia can have an adverse impact on the cytoskeleton, organelles, intracellular transport, and RNA processing (Richter et al., 2010). Another potential contributor to reduction in cell viability is heat-induced alterations in the plasma and subcellular organelle membranes, as well as membrane proteins (Richter et al., 2010; Mello et al., 2017).

Sufficient application of hyperthermia can result in cell death (**Figure 1**), but if cells survive several major classes of proteins will be activated leading to thermotolerance. These classes of proteins include: heat shock proteins that stabilize misfolded proteins, proteolytic enzymes that clear denatured/aggregated proteins, RNA-, and DNA-modifying proteins that repair damage, and others (Richter et al., 2010).

In addition to the responses to hyperthermia at a cellular level described above, hyperthermia may impart its effects via several additional, unique mechanisms on cell communities and these have been investigated in vivo. Tumors are generally associated with hypoxic and acidic environments due to poor vasculature, conditions in which cells are known to be more susceptible to hyperthermia (Gerweck et al., 1979; Eales et al., 2016). Elevated temperatures can lead to increased perfusion within the tumor, leading to greater chemotherapeutic drug delivery and higher oxygen concentrations, which in turn can sensitize tumors to radiotherapy (Song et al., 1996; Rau et al., 2000). Hyperthermia may enhance the immune response via several mechanisms, including increased migration of immune effector cells to the tumor, modulation of cell surface molecules and various pro-inflammatory cytokines, proliferation of effector cells, and increased immune cell cytotoxicity against malignant cells (Peer et al., 2010).

Despite the multitude of mechanisms by which hyperthermia can induce cell death, it is not efficient as a single agent treatment, mainly due to its poor specificity and the development of thermotolerance which may make subsequent hyperthermia treatments less effective. However, in combination with radiotherapy or chemotherapy, hyperthermia can lead to improved patient outcomes.

#### HYPERTHERMIA IN COMBINATION WITH RADIOTHERAPY AND CHEMOTHERAPY

In the clinic, hyperthermia can be applied to a local area, a specific region of the body or the entire body. In the past few decades, mild elevations of temperature have been achieved by various means including thermal chambers, hot water blankets, application of electromagnetic energy, perfusion of limb or body cavity with heated fluids, ultrasound and MIONs (van der Zee, 2002; Wust et al., 2002). In order to improve the efficacy, hyperthermia has often been evaluated as an adjunct treatment to enhance radiotherapy and cytotoxic chemotherapy. One way of expressing the enhancement of radiotherapy or chemotherapy is via the thermal enhancement ratio (TER), where TER is the ratio of the dose of radiation or drug alone that is required to achieve the end point to the dose of radiation or drug combined with heat to achieve the same end point (Overgaard, 1984). As an example, 60 min of hyperthermia at 42◦C, can result in a TER of nearly 2 for radiotherapy, making hyperthermia one of the most potent radiosensitizers (Overgaard, 1984). Hyperthermia is thought to enhance radiotherapy via protein denaturation and the subsequent inactivation of proteins involved in DNA repair. Inactivation of DNA repair proteins, particularly those involved in excision of clustered base damage, may prevent repair of the DNA damage induced by radiotherapy, leading to increased cell death (Kampinga and Dikomey, 2001). In vivo, hyperthermia can prime the tumor to radiotherapy via increased vascular perfusion and oxygenation of previously radioresistant, hypoxic areas (Song et al., 2005). Both preclinical and clinical evidence indicates that the TER is highest when hyperthermia is delivered simultaneously or in close temporal proximity to radiotherapy when protein denaturation and aggregation are likely to be at their greatest (van Leeuwen et al., 2017). Furthermore, the

TER increases with temperature and duration of hyperthermia (Overgaard, 1984).

Hyperthermia can synergistically enhance the efficacy of numerous chemotherapeutic agents including cisplatin, cyclophosphamide and bleomycin, whilst no significant enhancement for 5-fluorouracil, doxorubicin, and vincristine has been observed. For example, the application of 30 min of hyperthermia at 41.5◦C in vivo, can result in a TER of 1.48 for cisplatin and 2.28 for cyclophosphamide (Urano et al., 1999). Although the exact mechanism for chemosensitization is poorly understood, for alkylating or alkylating-like platinum agents like cyclophosphamide and cisplatin, their ability to interact with and encourage protein denaturation may be partly responsible (Lepock, 2005b). In vivo, hyperthermia can lead to chemosensitivity via increased tumor blood flow and increased vascular permeability resulting in increased accumulation of chemotherapeutic agent (Song et al., 2005).

There have been a number of randomized clinical trials on the impact of hyperthermia on various cancers in combination with radiotherapy or chemotherapy or both (**Tables 1**–**3**), with many other studies currently in progress (Valdagni et al., 1988; Berdov and Menteshashvili, 1990; Datta et al., 1990; Sharma et al., 1991; Sugimachi et al., 1994; Kitamura et al., 1995; Overgaard et al., 1996; Vernon et al., 1996; Sneed et al., 1998; Harima et al., 2001; van der Zee, 2002; Jones et al., 2005; Franckena et al., 2008; Verwaal et al., 2008; Huilgol et al., 2010; Issels et al., 2010; Colombo et al., 2011; Cihoric et al., 2015; Arends et al., 2016). The majority of studies demonstrated higher rates of local response with only mild to moderate toxicities. It is worth noting that there is some heterogeneity in the outcomes, which may be due to differences in heating protocols. An area of deficiency, and perhaps one of the reasons why hyperthermia is rarely used in the clinic is that delivering sufficient hyperthermia to the tumor, whilst sparing the surrounding normal tissue, is difficult.

### MAGNETIC HYPERTHERMIA

Despite the ability of hyperthermia to enhance radio- and chemotherapy treatments, toxicity due to the similar responses of malignant, and healthy tissues to hyperthermia remains a barrier to clinical application. A promising approach to overcoming this obstacle is magnetic hyperthermia, a form of hyperthermia that is currently undergoing clinical trials. It was first proposed by Gilchrist et al. (1957), who introduced the concept of injecting MIONs (20–100 nm), into lymphatic channels in order to heat residual cancer cells under an Alternating Magnetic Field (AMF) (Gilchrist et al., 1957). In 1993, Jordan et al. showed that delivering magnetic nanoparticles via direct injection into the tumor could result in much more effective and selective heating of tumors when compared to other heating techniques such as radiofrequency heating and ultrasound (Jordan et al., 1993). Furthermore, there is in vitro evidence that certain types of cancers including glioblastoma cells can take up magnetic nanoparticles more efficiently than non-malignant cells, although the exact mechanism is not well understood (Jordan et al., 1999). Since then, significant efforts have gone into the development of a clinical AMF system, resulting in the formation of a publicly listed company, MagForce AG based in Germany. The company has developed NanoTherm <sup>R</sup> aminosilane coated ferrofluid, NanoActivator <sup>R</sup> alternating magnetic field applicator, and NanoPlan <sup>R</sup> temperature simulation software.

In the past 2 decades, phase 1, and 2 clinical studies of intratumorally delivered magnetic nanoparticles and the subsequent application of AMF via the MagForce system have been successfully conducted for patients with glioblastoma and prostate cancers (Johannsen et al., 2005, 2007a,b; Maier-Hauff et al., 2011). Phase 1 clinical studies on patients with prostate cancer demonstrated the feasibility of the approach with no significant late treatment-related morbidity. The average temperatures achieved were in the hyperthermic

#### TABLE 1 | List of randomized clinical trials on hyperthermia combined with radiotherapy.


*(Continued)*

#### TABLE 1 | Continued


*(Continued)*

#### TABLE 1 | Continued


\**Meta-analysis of 5 randomized trials. The 5 trials were not published separately due to slow accrual.*

range (40–43◦C), as opposed to the thermoablative range (>50◦C). Although there were PSA declines following magnetic hyperthermia, responses in the monotherapy trial were of limited extent and duration and therefore, a phase 2 trial is now recruiting patients with intermediate risk prostate cancer and is evaluating magnetic hyperthermia in combination with low dose rate brachytherapy (Johannsen et al., 2010). Furthermore, MagForce has recently received an Investigational Device Exemption (IDE) for use in patients with intermediate prostate cancer undergoing active surveillance. Recruitment of patients with intermediate risk prostate cancer will commence after approval by ethics committees (Magforce, 2013, 2018). It is hoped that hyperthermia treatment in such patients can control the more aggressive component of the tumor and prevent or delay the need for radiotherapy or surgery. A phase 2 clinical trial involving 66 patients with recurrent glioblastoma, demonstrated a median overall survival of 13.4 months from the time of tumor recurrence (Maier-Hauff et al., 2011). Acute toxicities observed in this study included tachycardia (18.2%), headaches (13.6%), motor disturbances (21.2%), and convulsions (22.7%), which may be prevented with anti-epileptic drugs. In the magnetic hyperthermia study, however, no prolonged side effects were observed other than worsening motor disturbances, which may be related to disease progression rather than magnetic hyperthermia (Maier-Hauff et al., 2011). Following the phase 2 clinical trial, MagForce has been conducting a randomized, controlled trial (DRKS00005476) to determine the efficacy and safety of NanoTherm <sup>R</sup> monotherapy and NanoTherm <sup>R</sup> in combination with radiotherapy vs. radiotherapy alone in recurrent/progressive glioblastoma. The study is now closed and the final report of the data will be submitted to the official bodies this year (Magforce, 2013, 2018).

### MAGNETIC IRON OXIDE NANOPARTICLES FOR MAGNETIC HYPERTHERMIA

The most commonly used materials for magnetic hyperthermia are nanometre size (10–100nm) ferrite nanoparticles, in particular magnetite (Fe3O4) or maghemite (γ-Fe2O3). Fe3O<sup>4</sup> and γ-Fe2O<sup>3</sup> are commonly and collectively referred to as MIONs. The magnetic properties of MIONs arise from the presence of ions with different valency in their crystal structure. For instance, Fe3O<sup>4</sup> consists of two trivalent iron (III) ions and one divalent iron (II) ion. The unpaired ions result in parallel but oppositely aligned magnetic moments that do not cancel out and thus are subject to strong, spontaneous magnetization.

When exposed to an alternating magnetic field, MIONs produce heat via two main mechanisms: (1) hysteresis loss and (2) relaxational losses. Hysteresis losses occur in large MIONs which possess multiple magnetic domains. When such particles are subjected to an alternating magnetic field, the orientation of the magnetic moments will align continuously with the direction of the magnetic field as illustrated in **Figure 2**. This results in a difference in energy that is released in the form of heat (Kirschning et al., 2012). As MION size decreases, the number of magnetic domains will also decrease until a single magnetic domain remains at a threshold size of approximately 128 nm (Houlding and Rebrov, 2012). Below this size, MIONs are deemed superparamagnetic and in the presence of an AMF, heat is mainly produced by Néel relaxation and Brownian relaxation. Néel relaxation refers to rapid changes in the particle's magnetic moment when exposed to AMF (**Figure 2**). The rapid realignment is opposed by the particle's crystalline structure, resulting in heat generation. Brownian relaxation refers to the frictional heat generated from the physical rotation of particles within a supporting medium when the particles attempt to realign themselves with the changing magnetic field (**Figure 2**; Suto et al., 2009; Suriyanto et al., 2017). A more comprehensive discussion on the mechanism of heating is beyond the scope of this review and covered elsewhere (Ruta et al., 2015).

MIONs have the advantage of long term chemical stability and biocompatibility, and ease of surface modification and functionalisation when compared to other types of magnetic susceptible materials such as certain metals (e.g., iron, nickel or cobalt) or metal alloys (e.g., FePt, FeCo), (Dunn et al., 2014). Furthermore, MIONs can act as a contrast agent for computed tomography (CT) at high concentrations and magnetic resonance imaging (MRI) at lower concentrations, with several iron oxide nanoparticles previously approved by the FDA for these applications (Anselmo and Mitragotri, 2015). This is particularly useful since the concentration of the MIONs within the tumor can be estimated via CT and this can aid the estimation of hyperthermia dosimetry (Johannsen et al., 2007b). MIONs have also been shown to enhance the effects of radiotherapy even in the absence of AMF, potentially by increasing the generation

#### TABLE 2 | List of randomized clinical trials on hyperthermia combined with chemotherapy.


*(Continued)*

TABLE 2 | Continued


of reactive oxygen species (ROS) through the Fenton reaction (Huang et al., 2010; Klein et al., 2012; Khoei et al., 2014; Bouras et al., 2015). Finally, iron is an essential component of the human body and the average human adult naturally carries approximately 3.5–4 grams of iron. Consequently, unlike other inorganic nanoparticles, MIONs have been systemically delivered safely in large quantities in clinical settings (Hetzel et al., 2014). Furthermore, there is in vitro evidence that intracellular localized heating of ligand-decorated MIONS can lead to lysosomal damage of the target cells and induce cell death even in the absence of bulk heating (Creixell et al., 2011; Domenech et al., 2013).

#### MODE OF DELIVERY

MIONs can potentially be delivered to the tumor via intratumoral, intra-peritoneal, intra-arterial, intra-cavitary, and intravenous administration. Oral administration of MIONs is not feasible as most of the nanoparticles will be fecally excreted, owning to their large size (Chamorro et al., 2015). Intra-tumoral administration of MIONs efficiently localizes MIONs in the tumor and can result in effective heating of primary tumors such as prostate cancer. Intra-tumoral administration can result in very high concentrations of MIONs within the tumor and can remain localized in the tumor. When MIONs were directly injected to the prostate in men with localized prostate cancer, MIONs were still clearly visible on CT 6 weeks post injection, thereby allowing repeated magnetic hyperthermia treatments (Johannsen et al., 2005). In a separate post-mortem study of patients with glioblastoma who received MIONs, nanoparticles were restricted to the site of intra-tumoral injection, once again confirming a good retention profile (Van Landeghem et al., 2009). However, intra-tumoral delivery of MIONs is not practical for larger tumors with regional metastases and is more invasive than other techniques (**Figure 3**). Furthermore, poorly defined tumors like GBM may be better targeted by intravenously delivered MIONs which are less dependent on the operator for effective delivery, although penetrating the blood-brain barrier may be a challenge.

Intra-peritoneal mode of delivery is well suited to cancers that often spread to the peritoneal cavity such as ovarian, pancreatic and gastric cancers. Cancer targeting MIONs have been successfully delivered via the intra-peritoneal route and have demonstrated significant uptake by both primary and metastatic tumors in orthotopic mouse pancreatic cancer models (Gao et al., 2017). When compared to intravenous mode of delivery, the intra-peritoneal route achieved an intra-tumoral level that was 3-fold higher. The same system was able to carry chemotherapeutic drugs and significantly inhibited pancreatic tumors without systemic toxicity (Gao et al., 2017). Toraya-Brown et al. administered non-targeted MIONs intraperitoneally in an aggressive mouse metastatic ovarian cancer model and demonstrated significant accumulation of MIONs in the tumor (Toraya-Brown et al., 2013). They determined that the non-targeted MIONs were taken up by peritoneal phagocytes and delivered to tumors. When under an AMF, MIONs generated enough heat to induce cell death within tumors. A separate mouse study determined that up to 5 mg/kg of MIONs can be safely delivered intraperitoneally although at higher levels, signs of oxidative damage were detected within the hepatic and renal tissues (Ma et al., 2012). Furthermore, monocyte/macrophagelike cells with a propensity to migrate into tumors, can be loaded with MIONs externally and injected intraperitoneally, after which the cells will direct MIONs for magnetic hyperthermia (Basel et al., 2012).

#### TABLE 3 | List of randomized clinical trials on hyperthermia combined with radiotherapy and chemotherapy.


For bladder cancers, magnetic hyperthermia can be achieved by the direct injection of MIONs into the bladder cavity via a urinary catheter. The thick lining of the bladder will restrict the absorption of MIONs and once the treatment is completed, MIONs can be removed through the catheter, thereby minimizing systemic toxicity. The feasibility of this approach was demonstrated by Oliveira et al. in rat bladders, where temperatures of 42◦C were maintained in the bladder with minimal heating of surrounding tissues (Oliveira et al., 2013).

Intra-arterial administration of chemotherapeutic drugs has been successfully applied to liver cancers in the clinic. MIONs may be well suited to this task as they tend to accumulate in the liver via the reticuloendothelial system. With this in mind, the arterial delivery of iron oxide nanoparticles has been explored in preclinical models by several investigators (Lee et al., 2013, 2017; Kim et al., 2016). In rabbits, when MIONs were delivered with iodized oil and doxorubicin, there was an increased intra-tumoral accumulation of drugs and consequently, reduced numbers of viable tumor cells (Lee et al., 2013). For lung cancer, there have been early investigations into the potential formulation of aerosolized MIONs and their delivery via a combination of a nebulizer and a magnet (Dames et al., 2007; Tewes et al., 2014; Graczyk et al., 2015). However, it would be quite challenging to deliver sufficient quantities of MIONs for the application of magnetic hyperthermia using this approach.

Although the modes of administration mentioned so far are well suited to particular scenarios, intravenous administration is the most versatile method of delivery for the widest range of cancers. Not surprisingly, intravenous delivery is one of the most common routes of administration of chemotherapeutic drugs and in the past, FDA-approved MION MRI contrast agents have been delivered intravenously. (**Figure 3**). When MIONs are delivered in this manner, the accumulation of nanoparticles within the tumor depends in part on the enhanced permeability and retention (EPR) effect (Iyer et al., 2006). The EPR effect refers to the tendency of nanoparticles to preferentially accumulate within tumors due to their leaky vasculatures and poor lymphatic drainage. Once nanoparticles have reached the tumor, targeting ligands, such as small molecules, peptides or antibodies, bound to MIONs may lead to increased association and uptake of nanoparticles by malignant cells (DeNardo et al., 2007; Balivada et al., 2010). Their preferential accumulation within the malignant cells can lead to targeted heating of tumors and sparing of adjacent normal tissue under AMF (DeNardo et al., 2007; Balivada et al., 2010). Such approaches may potentially result in more homogeneous delivery of MIONs to the tumor and would be far less operator dependent when compared to other forms of targeting. In addition, the avoidance of the surgical morbidity associated with intra-tumoral injection of MIONs in the tumor may be attractive.

So far, preclinial in vivo studies (**Table 4**) have been performed in order to demonstrate the concept of biologically targeted magnetic hyperthermia (**Table 4**). Huang et al. intravenously injected extremely large quantities (1,700 mg Fe/kg) of untargeted MIONs into mice and achieved a subcutaneous tumor concentration of 1.9 mg Fe/kg of tumor. Despite the lack of targeting moiety, they were able to achieve a tumor to surrounding non-tumor concentration ratio of more than 16:1 via the enhanced permeability and retention effect alone. Application of AMF achieved significant tumor control when compared to either nanoparticles or AMF alone. In the same study, mice injected with even higher doses of nanoparticles (3,400 mg Fe/kg) survived more than 12 months without showing any clinical signs of toxicity (Huang and Hainfeld, 2013). Another in vivo study on intravenous administration of porphyrin coated MIONs demonstrated improved melanoma tumor control under AMF (Balivada et al., 2010). A third mouse study assessing the effectiveness of MIONs conjugated to ChL6, an antibody that targets tumor-associated antigen L6, demonstrated significant tumor accumulation and breast cancer tumor growth delays (DeNardo et al., 2007).

Despite the promising findings outlined above, preclinical studies often apply field strengths, frequencies or quantities of MIONs that are beyond what is clinically feasible and thus, further research is warranted in the areas of design, delivery, and the heating of nanoparticles, to achieve clinical translation in the future. In the following sections, areas requiring further research will be highlighted.

### FACTORS INFLUENCING THE EFFICACY OF BIOLOGICALLY TARGETED MAGNETIC HYPERTHERMIA

#### Field Strength and Frequency

Achieving and maintaining hyperthermia in the tumor is no easy task. Due to natural thermoregulatory processes, significant power must be delivered to elevate the temperature of a particular region of the body. The heating of MIONs is dependent on a variety of factors including the concentration of MIONs, frequency and the field strength. Currently, the only clinically available AMF system in the world, NanoActivator <sup>R</sup> (MagForce AG, Germany), can operate at a frequency of 100 kHz and is able to apply fields up to 18 kA/m (Jordan et al., 2001). Although higher frequencies are technically feasible, 100 kHz was chosen to minimize eddy currents and maximize the temperature differential between normal tissues and tumors containing magnetic nanoparticles (Jordan et al., 1993). Eddy currents are electrical currents that are induced within the conductor, in this case the human body, due to the changing magnetic field, as described by Faraday's law of induction. Excessive non-specific heating of normal tissues by eddy currents is the primary determinant of the maximum tolerable field strength and frequency.

At present, there is limited clinical data on the maximum tolerable field strength and frequency. In 1984, Atkinson et al. designed a single-turn induction coil for interstitial magnetic seed therapy and conducted experiments on thoraces of numerous volunteers. They found that field intensities up to 35.8 A/m at a frequency of 13.56 MHz can be tolerated for extended periods of time. Based on this study, the assumption was made that the product of field strength and frequency should not exceed 4.85 × 10<sup>8</sup> A/m·s (Atkinson et al., 1984). However, this is not an absolute limit and in certain scenarios, this limit may be exceeded (Dutz and Hergt, 2013; Obaidat et al., 2015). In phase 1 and 2 trials of the MagForce system, using lower frequencies of 100 kHz, patients with glioblastoma were able to tolerate up to 13.5 kA/m (1.35 × 10<sup>9</sup> A/m·s or a median value of 8.5 × 10<sup>8</sup> A/m·s) whilst patients with prostate cancer were only able to tolerate up to 5 kA/m (5 × 10<sup>8</sup> A/m·s) due to discomfort in the groin and/or perineal regions (Johannsen et al., 2007a; Maier-Hauff et al., 2007; Nieskoski and Trembly, 2014). This may have been due to boundary effects between tissues of different dielectric constants and conductivity, as well as narrowing of current path in the skin folds such as the groin, resulting in hot spots (Johannsen et al., 2007a). The higher tolerable field strength in patients with glioblastoma is likely to be due to the smaller radius of the head compared to the pelvis or thorax in other studies. Considering that higher field strengths and frequencies will translate to improved heating of tumors, further research is required into improving the tolerable limits of magnetic field strengths and frequencies via improved surface cooling of hotspots that develop in the body, such as the groin (Johannsen et al., 2007b). Furthermore, with shorter duration of treatment, it is possible that higher magnetic field strength or frequency may be achievable. Another possible limitation to the maximum field strength that can be applied clinically relates to the technical challenges of designing and manufacturing a much larger system than the smaller systems utilized in the preclinical studies (**Table 4**; Jordan et al., 2001). It is advisable that future preclinical studies on biologically targeted magnetic hyperthermia focus on the application of clinically relevant magnetic field strength and frequency of 18 kA/m and 100 kHz currently available on the MagForce system.

Assuming that MIONs have been delivered to the target, the temperature can be adjusted by the alteration of magnetic field strength or frequency. For example, the hyperthermia system from Magforce controls the temperature by adjusting the magnetic field strength. As the effect of hyperthermia is heavily influenced by the temperature reached and for how long this is maintained, it is extremely important to accurately monitor the temperature during therapy and this has been previously achieved with an invasive catheter or specialized software based on imaging (Mahmoudi et al., 2018). Future studies must ensure that hyperthermia is delivered sufficiently by close monitoring of the tumor temperature.

### DOSING AND TOXICITY OF MAGNETIC IRON OXIDE NANOPARTICLES

The rate of AMF-induced heating is highly dependent on the concentration of MIONs within the tumor. In clinical trials, up to 31.36 mg of Fe/cm<sup>3</sup> of tumor, in the form of MIONs, have been administered intra-tumorally in patients with glioblastoma (Maier-Hauff et al., 2011). Feraheme <sup>R</sup> (AMAG Pharmaceuticals, USA), an FDA approved iron oxide nanoparticle indicated for iron replacement, has been safely delivered intravenously in larger quantity than probably any other FDA approved inorganic nanoparticle so far and the recommended dose is 510 mg of Fe in the form of Feraheme <sup>R</sup> , followed by a second injection 3 to 8 days later. In the past, several patients have received two additional injections to a total dose of 2.02 g of Fe in the form of Feraheme <sup>R</sup> within a short period (Lu et al., 2010). In a hypothetical scenario, if 2.02 g of Fe in the form of MIONs, are intravenously administered to a patient with a 35 ml prostate tumor, and assuming that 1% of the dose would reach the tumor, this would result in only about 0.6 mg of Fe/cm<sup>3</sup> of tumor, far lower than what has been achieved with intra-tumoral administration. In addition, Feraheme contains approximately 3 nm iron oxide cores that are smaller than the MIONs that are typically associated with effective heating (Bullivant et al.,



\**Assuming 20 g average weight of mice.*

*EPR, enhanced permeability and retention; kA/m, kiloampere/metre; kHz, kilohertz; A/m*•*s, ampere/meter*•*second.*

2013). For example, the nanoparticles used by MagForce contain a 12 nm iron oxide core surrounded by aminosilanes and larger crystal cores are likely to be associated with different toxicity profiles. In mice, Huang et al. was able to deliver much higher concentrations of MIONs (5.1 g Fe/kg) and determined an MTD<sup>50</sup> value of 4.7 g Fe/kg, more than 100 times that delivered per kg in the Feraheme study (Huang and Hainfeld, 2013).

As the interaction of MIONs with their biological environment, and therefore their toxicity, varies with morphology, size, and surface modifications such as the addition of biocompatible coatings and targeting moieties, as well as the route of administration, each formulation needs to be tested thoroughly in vitro and in vivo. MIONs can mediate toxicity through several mechanisms that all have to be taken into account when evaluating their safety. Most intracellular toxicity is caused by generation of reactive oxygen species whereas in vivo disturbances of blood clotting, iron homeostasis and macrophage function, as well as organ toxicities, are additional considerations (Ilinskaya and Dobrovolskaia, 2013; Wu et al., 2014; Wei et al., 2016; Shah and Dobrovolskaia, 2018). A more detailed discussion of MION toxicity can be found in specialized review articles (Reddy et al., 2012; Liu et al., 2013; Arami et al., 2015).

To achieve sufficient heating via intravenous delivery of MIONs, further research is necessary to assess the tolerability of larger quantities of MIONs with bigger cores which are more suited to magnetic hyperthermia, and this will have to be finely balanced with size requirements for efficient intra-tumoral accumulation of nanoparticles.

#### HEATING EFFICIENCY OF MAGNETIC IRON OXIDE NANOPARTICLES

In order to minimize the quantity of iron oxide nanoparticles necessary for adequate magnetic hyperthermia, the development of nanoparticles with higher heating efficiency is desirable. The most common parameter for quantifying the heat generated via magnetic induction of MIONs is the Specific Absorption Rate (SAR). The experimental measurement of SAR is relatively simple. It typically involves suspending a known amount of MIONs in a liquid of known heat capacity. The test sample is exposed to an AMF of a specific strength and frequency, and the change in temperature is measured continuously over a period of time. The temperature measurement is carried out with fiber optic temperature probes to avoid electromagnetic interference with the measurement. The SAR is then calculated from the following equation (Kallumadil et al., 2009; Huang et al., 2012):

$$SAR = \frac{C}{m\_{np}} \left. \left( \frac{dT}{dt} \right) \right|\_{t=0}$$

where C is heat capacity of the fluid per unit mass of fluid, mnp is the mass of magnetic phase suspended in the fluid and dT/dt refers to the initial slope of temperature rise T, as a function of time, t.

It is important to note that SAR is a system-dependent parameter, that is, its value depends on the strength (H) and frequency (f) of the applied magnetic field. Therefore, direct comparison between measurements that are made using different field strength and frequency is not possible. A better parameter for this purpose is the Intrinsic Loss Power (ILP) which is mathematically described by the equation below (Kallumadil et al., 2009):

$$ILP = \frac{SAR}{H^2 f} = \frac{C}{H^2 f \ m\_{np}} \left. \left( \frac{dT}{dt} \right) \right|\_{t=0}$$

The ILP parameter is introduced under several key assumptions: (1) Test samples are single domain nanoparticles that heat up mainly via rotational relaxation; (2) Magnetic induction systems are of low frequencies at approximately 105–10<sup>6</sup> Hz; (3) Applied field strength is under the saturation field of the MIONs; (4) For the case of polydisperse MIONs in solution, the crystallite polydispersity index (PDI) has to be greater than 0.1 (Rosensweig, 2002; Kallumadil et al., 2009). If these assumptions are not satisfied, the derived ILP values may not be valid. It is important to note that the published ILPs are only a guide, and the absolute values may not always be reliable due to the variability in the methods used to measure them and given the heating rates are very sensitive to factors such as polydispersity (Gonzales-Weimuller et al., 2009; Wildeboer et al., 2014). Different types of MIONs have highly variable heating properties. Kallumadil et al. found significant variations in the ILP between various commercially available MIONs, ranging from 0.15 to 3.12 nHm<sup>2</sup> /kg. Heating rates can be influenced by several factors such as the ferrous iron content, size, hydrodynamic diameter, shape, number of cores, method of synthesis, and introduction of other metals such as Mn and Zn (Kallumadil et al., 2009; Blanco-Andujar et al., 2015; Hauser et al., 2015; Phong et al., 2017).

Due to the large number of variables, it is difficult to determine precisely how individual factors can impact the heating performance. In addition, the viscosity of the solvent and concentration of MIONs can further dictate the heating properties (Salas et al., 2014). Despite this, there are studies that do provide general insights to the relationship between the various characteristics and the heating properties. Several investigators have shown that in general, larger MIONs are more efficient at generating heat than smaller MIONs. (Gonzales-Weimuller et al., 2009; Lartigue et al., 2011; de La Presa et al., 2012; Jeun et al., 2012). For example, Lartigue et al. produced MIONs ranging from 4 to 35nm and coated them with rhamnose, a type of sugar. When heated under 168 kHz and 21 kA/m, the SAR was 0 W/g of Fe for 4 nm MIONs, 32 W/g of Fe for 10 nm MIONs, 61 W/g of Fe for 16 nm MIONs, and 76 W/g of Fe for 35 nm MIONs (Lartigue et al., 2011).

The shape of the nanoparticle can have a significant influence on the heating performance. Song et al. produced and compared the heating performance of quasi-cubical and spherical Fe3O<sup>4</sup> nanoparticles under 100 kHz and 30 kA/m. Under equal concentration of Fe, the SAR for quasi-cubical nanoparticles were far superior (Song et al., 2012). Another study by Nemati et al. compared deformed cube (octopods) shaped MIONs with spherical nanoparticles of similar volume and demonstrated superior heating performance of the octopods (Nemati et al., 2016). Liu et al. produced ring shaped MIONs (nanorings) and compared the heating performance with a commercial MION called Resovist across a range of magnetic field strengths. Although the difference cannot be entirely attributed to the shape alone due to the differences in size, nanorings demonstrated superior heating performance, especially under the higher ranges of magnetic field strength (Liu et al., 2015). Consequently, magnetic hyperthermia via nanorings resulted in superior tumor control in vivo (Liu et al., 2015). Despite the superior heating rates of some of the oddly shaped MIONS, it is important to be aware that the shape can also influence the rate of uptake and toxicity (Hinde et al., 2017). These factors must be considered when designing nanoparticles for clinical applications.

The surface coating can have a significant impact on the heating performance of MIONs. Complete coating of MIONs with a low heat conductor such as SiO<sup>2</sup> shell can prevent the outflow of heat and reduce the heating efficiency (Gonzalez-Fernandez et al., 2009; Rivas et al., 2012). Furthermore, the thickness of the coating can also impact the heating efficiency. Liu et al. coated MIONs with polyethylene glycol (PEG) polymer of various length ranging from 2,000 to 20,000 Da and found that MIONs coated with shorter polymers generally heat better, possibly due to increased Brownian loss, improved thermal conductivity and dispersibility (Liu et al., 2012). One exception to this was the 31 nm MION which heated better when coated with longer PEG polymers. This was ascribed to potential agglomeration of the 31 nm MIONs with the shorter PEG, highlighting a delicate balance between stability and heating performance. The coating can also influence the pharmacokinetics of MIONs in the body which is an important consideration when developing MIONs for hyperthermia (Arami et al., 2015). Doping MIONs with Mg or Zn is another strategy that has resulted in nanoparticles with superior heating profiles, resulting in better tumor control in vivo (Jang et al., 2009).

Interestingly, one of the highest ILPs (23.41 nHm<sup>2</sup> /kg) to have been reported in the past was on bacterially derived MIONs, which have a mean core diameter of approximately 30 nm (Hergt et al., 2005). Bacterial magnetosome-like cubic nanoparticles were later produced by Martinez-Boubeta et al. and demonstrated superior heating efficiency compared to spheroidal MIONs of similar size (Martinez-Boubeta et al., 2013). Le Fevre et al. have evaluated the effectiveness of magnetic hyperthermia via intra-tumorally delivered magnetosomes and achieved superior tumor control compared to chemically synthesized MIONs (Le Fèvre et al., 2017). Recently, Sangnier et al. demonstrated that magnetosomes can be tagged with tumor targeting peptide, arginine-glycine-aspartic acid (RGD), then administered intravenously in mice models for targeted delivery to tumors (Plan Sangnier et al., 2018). They applied photothermal therapy rather than magnetic hyperthermia as it was thought to be more effective. However, such approaches are likely to be limited for deep seated tumors in humans and thus, further work is required to evaluate its application for magnetic hyperthermia. Many other types of nanoparticles have been produced in the past for magnetic hyperthermia and more details can be found in other specialized review articles (Blanco-Andujar et al., 2017; Hedayatnasab et al., 2017). Higher heating efficiency would be highly desirable as it would reduce the quantity of nanoparticles, field strength and frequency required to induce significant heating.

#### TARGETING OF MIONS

Intravenously administered nanoparticles preferentially accumulate within tumors owing to their leaky vasculature and poor drainage. This EPR effect is well documented and was recently demonstrated in human tumors (Clark et al., 2016). In addition, structural and surface modification of MIONs can further increase tumor accumulation and up to ∼15.5%ID/g have been reported in the past (Xu et al., 2016).

Targeting of cancer cells with antibodies or other ligands can further improve the accumulation of nanoparticles within the tumor. MIONs conjugated to antibodies have been previously delivered to several tumor specific antigens including L6, HER-2 and PSMA for medical imaging and magnetic hyperthermia (DeNardo et al., 2007; Zhang et al., 2011; Tse et al., 2015). As mentioned earlier, one of the best examples is a study by DeNardo et al. in which MIONs conjugated to ChL6, an antibody that targets tumor-associated antigen L6, demonstrated significant tumor accumulation and breast cancer tumor growth delays under an AMF (DeNardo et al., 2007). Despite the potential for enhanced delivery, targeting can be associated with significant challenges in terms of the chemistry of conjugation and stability of ligand or antibody bound to nanoparticles. For example, MLN2704, a prostate specific antigen directed immunoconjugate for delivering chemotherapeutics to prostate cancer was associated with significant toxicity and limited activity due to deconjugation of the targeting antibody once in circulation (Milowsky et al., 2016). In a clinical trial of CALAA-01, a ligand bound nanoparticle siRNA delivery system, 21% of patients discontinued the study due to an adverse event and it was proposed that ligand instability was responsible for the undesirable toxicity (Zuckerman and Davis, 2015). Some of these limitations can be overcome by the application of bispecific antibodies that can spontaneously bind to both the poly ethylene glycol (PEG) coated nanoparticles and cancer specific antigens such as prostate specific membrane antigen (PSMA) or epidermal growth factor receptor (EGFR). Bispecific antibodies are composed of 2 separate single-chain fragment (scFv) and are smaller than whole antibodies. It can be stored in the freezer separate to the nanoparticles, thereby overcoming the stability issue. When administered with any PEGylated nanoparticles prior to or at the time of delivery, bispecific antibodies will spontaneously associate itself with PEGylated nanoparticles. Within the tumor, bispecific antibodies will bind to cancer specific antigens and keep the nanoparticle in close proximity to the target cancer cells, thereby, enhancing tumor accumulation (Howard et al., 2016).

To overcome the limitations of antibodies, MIONs can alternatively be conjugated to cancer specific peptides, glycosaminoglycans or aptamers. In order to target ovarian cancer, Taratula et al. synthesized MIONs conjugated to an ovarian cancer targeting Luteinizing Hormone-Release Hormone (LHRH) peptide. In vitro, LHRH peptide coating improved the ability of MIONs to associate with ovarian cancer cells and resulted in a significant reduction in cell viability under an alternating magnetic field (Taratula et al., 2013). For reduced immunogenicity, MIONs can be coated with hyaluronic acid, a biocompatible material that is naturally found in our body. Hyaluronic acid can target cancer cells via CD44 receptor, a commonly found cell surface marker in epithelial tumors and its potential role in magnetic hyperthermia has been demonstrated in vitro (Thomas et al., 2015). Nair et al. produced glioma targeting aptamers, composed of oligonucleotides, for conjugation with dextran coated iron oxide nanoparticles. Using the targeted MIONs, they were able to induce preferential damage to glioma cells via mechanical oscillation induced by a rotating magnetic field (Nair et al., 2010). It is possible that such nanoparticles may be applied for magnetic hyperthermia in the future.

For further enhancement of hyperthermia, MIONs can be directed toward intracellular organelles of cancer cells via conjugation of organelle targeting peptides. Peng et al. administered transferrin and nuclear targeting TAT peptide conjugated MIONs to mice and applied photothermal hyperthermia (Peng et al., 2017). When compared to transferrin conjugated MIONs, nuclear targeting MIONs demonstrated significant improvement in tumor control (Peng et al., 2017). Additional studies are required to confirm that such intracellular targeting strategies may be applicable for magnetic hyperthermia. Despite these exciting approaches to targeting, there is multitude of factors that can influence its effectiveness and a detailed evaluation can be found in specialized reviews (Rosenblum et al., 2018).

Another novel approach to improved tumor targeting is to suppress the reticuloendothelial system with drugs prior to the delivery of MIONs. For example, Abdollah et al. demonstrated that the suppression of Kupffer cells in the liver with dextran sulfate can significantly increase the circulating half-life of non-targeted MIONs by inhibiting the liver uptake (Abdollah et al., 2014). It is uncertain whether dextran sulfate suppression can also be applied in combination with ligand- or antibodyconjugated MIONs to prevent liver uptake and further research is warranted in this area.

Overall, several strategies are being evaluated in order to effectively target nanoparticles to the tumor whilst sparing normal tissue. Improved targeting will ultimately be the key to delivering sufficient quantities of MIONs for selective heating of tumors.

#### MAGNETICALLY TARGETED MIONS

Due to their magnetic properties, MIONs can be directed toward the tumor via a magnetic field. This can be applied in combination with targeted MIONs for effective magnetic hyperthermia. There are several notable examples of this approach. For gene therapy, MIONs have been used to direct intravenously administered silencing RNAs toward gastric tumors in mouse models under a magnetic field (Namiki et al., 2009). In a separate study, Garcia-Jimeno et al. were able to direct magnetoliposomes, with the aid of a magnetic field, toward the target and away from the liver and the spleen of mice (García-Jimeno et al., 2012).

For maximum uptake and retention in the tumor, it is important for MIONs or other nanoparticles to extravasate and reach the cancer cells. This can be achieved by disrupting the endothelial barrier with an external magnetic field. Qui et al. injected MIONs into mouse tail vein and used an external magnetic field to direct the particles into the lateral tail vein. Histological examination revealed that MIONs accumulated in the endothelial tissue. When a fluorophore was injected systemically, fluorescence signal was higher in the tail of the mice subjected to the magnetic field and MIONs, due to a disruption of endothelial lining (Qiu et al., 2017). Combining these approaches may potentially improve the therapeutic efficacy of magnetic hyperthermia in the future.

#### OTHER METHODS TO IMPROVE THE IMPACT OF TARGETED MAGNETIC HEATING

Other novel methods of improving the effectiveness of magnetic hyperthermia have been explored. Espinosa et al. applied near-infrared laser irradiation (808 nm) during magnetic hyperthermia in vivo and demonstrated 2–5 fold improvements in heating when compared to magnetic hyperthermia alone although such approaches would be limited to surface tumors owing to the poor tissue penetration of laser irradiation (Espinosa et al., 2016).

In the past, there have been attempts to biologically enhance the effectiveness of magnetic hyperthermia with hyperthermia enhancing drugs such as the heat shock protein (HSP) 90 inhibitor Geldanomycin. When cells are heated, HSP 90 plays a key role in stabilizing proteins, thus, limiting the downstream effects of protein denaturation. Therefore, the inhibition of HSP 90 can lead to improved effectiveness of hyperthermia and reduce thermotolerance. For example, Ito et al. delivered Geldanomycin, and applied magnetic hyperthermia in a mouse melanoma model, which resulted in significant improvement in tumor control when compared to magnetic hyperthermia alone (Ito et al., 2009). This approach is particularly promising as HSP 90 inhibitors can independently enhance the effectiveness of radiotherapy, even in the absence of hyperthermia (Schilling et al., 2015).

There are many other hyperthermia enhancers that have been reported in the past but the majority of these agents have not been evaluated in combination with magnetic hyperthermia (Marchal et al., 1986). Protease inhibitors are another class of potent hyperthermia enhancers that have been evaluated in vitro. It is thought that the enhancement is achieved by inhibiting the clearance of denatured proteins within the cells (Zhu et al., 1995).

Another novel strategy is to combine magnetic hyperthermia with thermally sensitive liposomes. This can be achieved by creating a liposome with magnetic iron oxide cores embedded within. When an AMF is applied, the magnetic nanoparticles will trigger the release of the lipososomal contents. As AMF can be applied to a specific region of the body, this could result in targeted drug release and improved therapeutic effectiveness. For example, Yang et al. produced a CD90 targeted magnetoliposome encapsulating 17-AAG, a HSP 90 inhibitor. The magnetoliposome was able to simultaneously heat liver cancer stem-like cells and trigger the release of 17-AAG, thereby improving the effectiveness of magnetic hyperthermia (Yang et al., 2015).

## CONCLUSION

Hyperthermia can lead to cell death via modulation of various cellular processes and is an effective treatment that can enhance the outcomes of radiotherapy and chemotherapy. One of the disadvantages is the lack of specificity toward malignant cells compared to healthy tissue. Systemic administration of targeted MIONs has the potential to improve the specificity of hyperthermia and improve its efficacy. However, several limitations must be resolved before this technology can progress to clinic. Future preclinical studies should focus on designing MIONs that can target and heat tumors more effectively. Furthermore, various hyperthermia enhancers should be evaluated in combination with magnetic hyperthermia, with the ultimate objective of achieving clinical feasibility.

### AUTHOR CONTRIBUTIONS

DC drafted the manuscript, drew the figures and constructed the tables. DC, ML, JG, RQ, YN, FM, MJ, TD, and MK discussed the outline and critically reviewed the paper, the content, and the figures used.

#### ACKNOWLEDGMENTS

Special thanks to Helen Forgham for designing **Figure 3.** Authors are also grateful for support from the Children's Cancer Institute, which is affiliated with the University of New South Wales (UNSW Sydney) and Sydney Children's Hospital Network. DC is supported by an Australian Government Research Training Program Scholarship, and a RANZCR Genito-Urinary Fellowship Grant (supported by Tolmar Australia Pty Ltd). MK is supported by an NHMRC Principal Research Fellowship (APP1119152) and is funded by a Cancer Council New South Wales Program Grant (PG16-01). MK and TD are both funded by an Australian Research Council Centre of Excellence in Convergent Bio-Nano Science and Technology (CE140100036), and NHMRC Program Grant (APP1091261). TD is supported by an Australian Laureate Fellowship from the Australian Research Council.

#### REFERENCES


Arami, H., Khandhar, A., Liggitt, D., and Krishnan, K. M. (2015). In vivo delivery, pharmacokinetics, biodistribution and toxicity of iron oxide nanoparticles. Chem. Soc. Rev. 44, 8576–8607. doi: 10.1039/C5CS00541H

Arends, T. J., Nativ, O., Maffezzini, M., De Cobelli, O., Canepa, G., Verweij, F., et al. (2016). Results of a randomised controlled trial comparing intravesical chemohyperthermia with mitomycin C versus bacillus calmette-guerin for adjuvant treatment of patients with intermediate- and high-risk non-muscle-invasive bladder cancer. Eur. Urol. 69, 1046–1052. doi: 10.1016/j.eururo.2016.01.006


tumor targeting and multimodality imaging. Nanoscale 8, 12683–12692. doi: 10.1039/C5NR09193D


Zuckerman, J. E., and Davis, M. E. (2015). Clinical experiences with systemically administered siRNA-based therapeutics in cancer. Nat. Rev. Drug Discov. 14, 843–856. doi: 10.1038/nrd4685

**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 Chang, Lim, Goos, Qiao, Ng, Mansfeld, Jackson, Davis and Kavallaris. 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.

# Nano-BCG: A Promising Delivery System for Treatment of Human Bladder Cancer

Julieti Huch Buss <sup>1</sup> , Karine Rech Begnini <sup>1</sup> , Camila Bonemann Bender <sup>1</sup> , Adriana R. Pohlmann<sup>2</sup> , Silvia S. Guterres <sup>3</sup> , Tiago Collares <sup>1</sup> and Fabiana Kömmling Seixas <sup>1</sup> \*

<sup>1</sup> Laboratory of Cancer Biotechnology, Biotechnology Graduate Program, Technology Development Center, Federal University of Pelotas, Pelotas, Rio Grande do Sul, Brazil, <sup>2</sup> Pharmaceutical Sciences, Federal University of Rio Grande do Sul, Porto Alegre, RS, Brazil; Institute of Chemistry, Federal University of Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, Brazil, <sup>3</sup> Pharmaceutical Sciences, Federal University of Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, Brazil

Mycobacterium bovis bacillus Calmette–Guerin (BCG) remains at the forefront of immunotherapy for treating bladder cancer patients. However, the incidence of recurrence and progression to invasive cancer is commonly observed. There are no established effective intravesical therapies available for patients, whose tumors recur following BCG treatment, representing an important unmet clinical need. In addition, there are very limited options for patients who do not respond to or tolerate chemotherapy due to toxicities, resulting in poor overall treatment outcomes. Within this context, nanotechnology is an emergent and promising tool for: (1) controlling drug release for extended time frames, (2) combination therapies due to the ability to encapsulate multiple drugs simultaneously, (3) reducing systemic side effects, (4) increasing bioavailability, (5) and increasing the viability of various routes of administration. Moreover, bladder cancer is often characterized by high mutation rates and over expression of tumor antigens on the tumor cell surface. Therapeutic targeting of these biomolecules may be improved by nanotechnology strategies. In this mini-review, we discuss how nanotechnology can help overcome current obstacles in bladder cancer treatment, and how nanotechnology can facilitate combination chemotherapeutic and BCG immunotherapies for the treatment of non-muscle invasive urothelial bladder cancer.

Keywords: bladder cancer, nanotechnology, BCG, Nano-BCG, monoclonal antibody, EGFR

### INTRODUCTION

Bladder cancer (BC) is the second most common malignancy of the urinary tract, the fourth most common cancer in men with a yearly incidence rate of 330, 380 cases, and the 11th most common among women with a yearly incidence rate of 99,413. Worldwide, carcinomas of the bladder represent the ninth most common cause of cancer, with 430,000 patients diagnosed with BC annually (Jemal et al., 1999; Ferlay et al., 2013). The incidence of BC also increases with advancing age, as 90% of new diagnoses are made in people over the age of 55 (average age of 73 years at diagnosis) (American Cancer Society, 2017).

Most (75%) BC cases are non-muscle-invasive bladder cancers (NMIBC) at diagnosis with the other 25% representing muscle invasive bladder cancers (MIBC) or metastatic cancers (Moch et al., 2016). Urothelial carcinomas can be categorized as low grade or high grade according to their

#### Edited by:

Susan Hua, University of Newcastle, Australia

#### Reviewed by:

Grannum R. Sant, Tufts University School of Medicine, United States Kaisar Raza, Central University of Rajasthan, India

> \*Correspondence: Fabiana Kömmling Seixas seixas.fk@gmail.com

#### Specialty section:

This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology

Received: 21 October 2017 Accepted: 21 December 2017 Published: 12 January 2018

#### Citation:

Buss JH, Begnini KR, Bender CB, Pohlmann AR, Guterres SS, Collares T and Seixas FK (2018) Nano-BCG: A Promising Delivery System for Treatment of Human Bladder Cancer. Front. Pharmacol. 8:977. doi: 10.3389/fphar.2017.00977

architectural and cytological atypia and include papillary urothelial neoplasm or low malignant potential (Cheng et al., 2012). Pathological assessment is the gold standing for tumor classification. Ta (non-invasive papillary) and Tis [carcinoma in situ (CIS)] are tumors that are restricted to the mucosa, while T1 and T2 are tumors that invade the lamina propria and the muscularis propria, respectively (Sobin and Gospodarowicz, 2009).

Initial BC treatments involve transurethral resection (TURBT) to facilitate removal of the visible tumor (Hall et al., 2007; Sylvester, 2008; Babjuk et al., 2013). Further therapy is dependent on pathologic stage and grade of the tumor and often mediated through intravesical instillation. Although the response rate to therapy in patients with NMIBC is high (∼80%), 50–90% of NMIBC patients suffer from recurrence within 5 years, with muscle invasion found in up to 20% of recurrent patients (Rübben et al., 1988; Lamm and Allaway, 2000; Hussain et al., 2009). This review focuses on currently available BC therapies and describes nanotechnology tools to enhance therapeutic effects and overcome side effects, emphasizing its use to improve BCG immunotherapy.

## BLADDER CANCER TREATMENTS

Following TURBT, a single intravesical chemotherapy treatment is recommended for patients with low to intermediate risk NMIBC, with mitomycin, epirubicin, and gemcitabine representing common drugs of choice (Kamat et al., 2016). It has been shown that the relative risk for tumor recurrence is reduced by 50% if the chemotherapy instillation is given with 24 h after TURBT (Kaasinen et al., 2002).

Intravesical immunotherapy with Mycobacterium bovis Bacillus Calmette-Guérin (BCG) is the treatment of choice for patients with high-risk NMIBC. BCG immunotherapy is the gold standard treatment for NMIBC due to its ability to reduce recurrence and progression to MIBC (Ahn et al., 2014). A meta-analysis with individual patient data comparing BCG immunotherapy with intravesical mitomycin chemotherapy has shown BCG to be superior in terms of reducing recurrence and delaying disease progression; however, no significant differences in progression or overall survival were observed (Malmström et al., 2009). Conventional BCG treatment consists of a percutaneous BCG vaccine administered 2–6 weeks after TURBT followed by 6 weekly courses of intravesical BCG administration (Morales et al., 1976; Kresowik, 2009; Kamat et al., 2015). However, specific BCG substrain preferences, schedules, and dosages differ across geographic regions due the wide range of BCG substrains licensed for human tuberculosis vaccination and BCG immunotherapy (Gan et al., 2013). Based on multiple meta-analyses, it is recommended to continue BCG therapy for 1 to 3 years if tolerated by the patient to decrease recurrence and progression of NMIBC (Shelley et al., 2001; Sylvester et al., 2002; Böhle and Bock, 2004; Hall et al., 2007; Gontero et al., 2010; Babjuk et al., 2013).

Despite these guidelines, it is estimated that 20% of patients with high-risk NMIBC treated with BCG will progress to muscle invasion or suffer from NMIBC recurrence within 5 years (Rübben et al., 1988). MIBC is a major clinical issue due to its aggressiveness and high 5 year mortality rate. To maximize survival rates, radical cystectomy (RC) represents the best treatment option. RC consists of removal of the bladder, prostate, seminal vesicles, proximal vas deferens, and proximal urethra in men, and bladder, uterus, ovaries, fallopian tubes, urethra, and part of vagina in women (Arcangeli et al., 2015). Many patients cannot tolerate the morbidity of RC and instead opt for continued local therapy as an effort to spare their bladder (Ahn et al., 2014). Some studies have shown progress with the use of Mytomicin C (van der Heijden et al., 2004; Halachmi et al., 2011), Gemcitabine (Skinner et al., 2013), Valrubicin (Steinberg et al., 2000), Docetaxel (Barlow et al., 2013), Nab-Paclitaxel (McKiernan et al., 2011), mycobacterial cell wall extract (Morales et al., 2009), EGFR (Rebouissou et al., 2014) and a combination therapy of Gemcitabine and Mytomicin C for the treatment of BC (Lightfoot et al., 2011). More recently, trimodal treatment with simultaneous delivery of chemotherapy and radiotherapy has emerged as an effective bladder sparing treatment with similar survival rates compared to RC (Arcangeli et al., 2015). However, none of these local therapies have shown to be more effective for BC control than early RC (Ahn et al., 2014).

### CHEMOTHERAPY OPTIMIZATION USING NANOTECHNOLOGY FOR BLADDER CANCER THERAPY

Nanotechnology consists in the study and application of materials on the nanometer scale (Ebbesen and Jensen, 2006) and application of nanotechnology in the medical field is referred to as nanomedicine (Sweeney, 2015). Nanotechnology have proven to be a powerful tool for the development of new chemotherapies or immunotherapies for BC. The development of new drug delivery systems has been growing and is expected to continue to increase over the next few years (Brito et al., 2017).

In this context, several studies have utilized nanoparticles (NPs) to increase the therapeutic effectiveness and reduce adverse effects of chemotherapy by targeting chemotherapeutic agents to a specific tissue and increasing its bioavailability (Yurgel et al., 2014; Kang et al., 2017; Yao et al., 2017; Zhu et al., 2017). Polysaccharide-based NPs loaded with Mitomycin C and surrounded by the bioadhesive polymer chitosan mixed with polylactic acid or with poly(ε-caprolactone) have been utilized in an attempt to optimize BC treatments. This NP promoted favorable drug loading and release profiles along with improved anticancer efficacy and cellular interactions (Bilensoy et al., 2009). Erdogar et al. (2012) has also demonstrated that bioadhesive and cationic NPs loaded with Mitomycin C are able to increased exposure of the bladder to the drug resulting in a drug reservoir at the action site, which might improve local treatment (Erdogar et al., 2012). In addition, cationic core-shell nanoparticles loaded with Mitomycin C have also improved antitumor efficacy in tumor-induced rat models (Erdogar et al., 2014).

Magnetic NPs (MNPs) also show promise for delivery of quimiotherapic agents to the target tissue (Stapf et al., 2017). To limit doxorubicin's (Dox) cytotoxic effects on healthy cells, MNPs (iron oxide) were conjugated with Dox to ensure efficient delivery to cancer sites, resulting in increased BC sensitivity compared to Dox alone (Nowicka et al., 2013). In addition, monoclonal antibodies (mABs) bound to MNPs can increase the ability of MNPs to target BC cells and enable thermotherapy to cope with BC recurrence (Rezaei et al., 2017).

NPs loaded with molecules with high urothelium mucoadhesivity is another approach used to optimize the delivery of molecules to the target tissue. To increase Dox specificity for BC cells, thiol-functionalized NPs loaded with Dox were synthesized and induced cytotoxicity against UMUC3 cancer cells (Zhang et al., 2014). Besides the possibility of conjugation with different molecules, studies have also reported the importance of developing methods to modify the surface of mesoporous silica NPs to enhance the antineoplasic effects of Dox on BC (Wei et al., 2017). An enhanced therapeutic effect against UMUC3 cells was also demonstrated using Dox and peptide-modified cisplatin synergistically loaded onto positively charged mucoadhesive chitosan–polymethacrylic acid nanocapsules (Lu et al., 2016).

In addition, platinum agents can be loaded onto a variety of polymeric, lipid, and inorganic nanocarriers, including liposomes, NPs, and nanotubes to increase their antitumoral effects (Browning et al., 2017). The use of cisplatin nanocarriers is associated with reduced toxicity and adverse events (Sudha et al., 2017); however, novel strategies are required to increase drug uptake and release at the target site. In this regard, cisplatin NPs were evaluated in a preclinical study against NMIBC and cisplatin-loaded biocompatible poly(L-aspartic acid sodium salt) (PAA) NPs demonstrated potential for improved intravesical treatment of NMIBC while reducing local and systemic side effect (Kates et al., 2017).

Nanotechnology tools have also been used in clinical trials. Albumin-bound-Paclitaxel NPs showed minimal toxicity and systemic absorption when used to treat NMIBC during the first human intravesical phase I trial (McKiernan et al., 2011). In addition, phase II trials have demonstrated minimal toxicity of intravesical nab-paclitaxel in NMIBC patients with a response rate of 35.7% (Mckiernan et al., 2014). The formulation of albumin-bound-paclitaxel NPs has also been used for the first time to treat unresectable metastatic urethral cancer. Follwoing therapy, a 70% reduction in the size of the tumor was observed in addition to 19 months of progression free survival (Abaza and Alemany, 2014).

### RNAI OPTIMIZATION OF NANOTECHNOLOGY BLADDER CANCER THERAPY

The use of interference RNA (RNAi) combined with nanotechnology is another promising approach for BC treatment. RNAi technology can be used to inhibit tumor growth through messenger RNA inhibition of several activated oncogenes (Xin et al., 2017). In human BC, some upregulated genes associated to the development of resistance to chemotherapy have been inhibited using RNAi technology through knockdown of the target gene (Pan et al., 2016; Wang et al., 2017). Within this context, RNAi technology is a highly effective approach to combat chemoresistance and improve advanced BC outcomes.

Although RNAi technology could be used to overcome multidrug resistance and restore cells sensitivity, there are several challenges associated with RNAi delivery to diseased sites for gene therapy (Melamed et al., 2017). NPs appear to be a promising tool to help overcome existing biological barriers to RNAi delivery. Through this approach, studies have reported upregulation of BC specific genes, which can be effective targets of NP-siRNA therapeutic approaches (Seth et al., 2011; Müller et al., 2016).

Small dsRNAs, known as small activating RNA (saRNA), produce the opposite effect of RNAi by inducing gene expression (Li et al., 2006; Chen et al., 2008). The therapeutic potential of dsRNA (P21-322) coupled with nanotechnology has been demonstrated using a 20-fluoro-modified derivative loaded into lipid NPs in an orthotopic model of BC. Antitumoral activity and induction of p21 expression was confirmed in vitro and in vivo (Kang et al., 2012), suggesting induction of specific genes can provide an alternative route to BC treatment.

#### FUNCTIONALIZATION OF NANOPARTICLES TO OPTIMIZE BLADDER CANCER THERAPY

The functionalization of nanoparticles with monoclonal antibodies is a promising strategy for targeted delivery to and uptake by cells overexpressing the antigens specific for these antibodies (Eloy et al., 2017). Development of molecules that exhibit affinity for targets expressed on tumor cells represents an emerging therapeutic approach (Diesendruck and Benhar, 2017). For example, the receptor tyrosine kinase EGFR exhibits altered expression in several types of solid tumors and its overexpression in UC is directly correlated with advanced tumor stages (Kassouf et al., 2008). The incidence of EGFR positivity in epithelial tumors varies by tumor type. BC is often characterized by a high mutation rate and high EGFR expression in approximately 50% of cases (Colquhoun and Mellon, 2002). Thanks to the expression of EGRF on BC cells, an anti-EGFR monoclonal antibody conjugated to gold nanorods was able to effectively bind EGFR-expressing BC cells and reduce the systemic exposure and clearance of nanoparticles from the body (Yang et al., 2015). These results indicate the use of monoclonal antibodies represents another possible approach for delivery of molecules to specific tissues.

### NANO-BCG: OPTIMIZING BCG IMMUNOTHERAPY USING NANOTECHNOLOGY TO TREAT BLADDER CANCER

As previously mentioned, BCG is considered the standard treatment for NMIBC. However, BCG immunotherapy


 cancer.


TABLE 1 |

Continued

is associated with frequent induction of adverse effects in patients (Poletajew et al., 2017) leading researchers to investigate novel alternatives to increase their effectiveness (Begnini et al., 2013). Delivery systems and nanotechnological approaches are interesting tools to improve currently available BCG therapies and prolong exposure of the bladder tissue. The main advances of nanotechnology tools for improvement of BCG immunotherapies against BC are listed in **Table 1**.

Cationic Chitosan (CS) NPs encapsulating BCG developed with the purpose of increasing the antitumor efficacy of BCG following intravesical administration have demonstrated significant advantages for the treatment of BC (Erdogar et al., 2014). Application of magnetic thermosensitive hydrogel developed with chitosan has been shown to be effective in increasing exposure of the bladder to BCG in addition to potentializing its immunological response (Zhang et al., 2013).

With the goal of overcoming side effects resulting from the administration of viable BCG, BCG cell wall skeleton (BCG-CWS) used as an immunomodulator in cancer patients (Uenishi et al., 2007, 2009; Hayashi et al., 2009) was shown to effectively control cell proliferation in UC, representing an efficient and safe alternative to BC immunotherapy (Kato et al., 2010). Nakamura et al. (2014a) was the first to develop a nanoencapsulated BCG-CWS (CWS-NP) using a liposome evaporated emulsified lipid (LEEL) method, resulting in a strong antitumor effect against MBT-2 BC cells as well as in vivo induced tumors (Nakamura et al., 2014a). CWS-NP was also shown to produce significant antitumor effects through internalization of BCG-CWS in BC cells, which contributes to the initiation of antitumor immunological activity (Nakamura et al., 2014b).

Another nanotechnological approach applied to BCG therapy consists of the use of modified nanoparticles incorporating BCG cell wall (BCG-CW) or skeleton. Utilization of octaarginine-modified liposomes incorporating BCG-CW (R8-liposome-BCG-CW) results in increased immunotherapeutic potential of BCG-CW for NMIBC through cellular internalization resulting in growth inhibition in vivo (Joraku et al., 2008). R8-liposome-BCG-CWS has also be used to investigate the suppressive effects of liposomes using a rat N-butyl-N-(4-hydroxybutyl) nitrosamine (BBN) induced BC model. This approach demonstrated that R8-liposome-BCG-CWS displays inhibitory effects against CD in vivo (Miyazaki et al., 2011b). Using this same approach, other studies have demonstrated that R8-liposome-BCG-CWS treatment results in induction of surface specific ligands (NKG2D) in BC cells, making them more susceptible to lymphokine-activated killing (LAK), indicating these cells are affected by R8-liposome BCG-CWS administration (Miyazaki et al., 2011a). These results demonstrate the efficiency of nanotechnology applications for optimization and development of novel BCG immunotherapy approaches for BC.

### PERSPECTIVE

Considering the promising results demonstrated by utilizing nanotechnology to develop new BC therapies, including success application of those approaches to BCG therapy, we believe that nanotechnology will provide significant advances for improving BC treatment. In this review, we described the main advances and applications of nanotechnology tools for development of novel treatments against BC, providing evidence that nanotechnology has contributed greatly to this effort by controlling drug release for longer periods, enabling the encapsulation of multiple drugs simultaneously, decreasing side effects, and increasing bioavailability. Nevertheless, we believe that combination therapies show more promise for effective treatment of this complex disease than individual approaches (**Figure 1**).

Although nanotechnology has resulted in significant progress for BCG immunotherapy, including increasing immunotherapeutic effects, prolonging exposure of the bladder tissue, and reducing adverse side effects, we believe this approach can be further enhanced through intravesical BCG delivery using NPs functionalized with antibody molecules against highly expressed receptors on the surface of BC cells such as EGFR to further target drug delivery and avoid systemic exposure and clearance of NPs from the body. Therefore, considering the emerging and motivating results using these approaches to treat BC, we believe that BCG delivery using NPs functionalized with monoclonal antibodies, in particular anti-EGFR, will provide a significantly improved delivery system for treatment of human BC.

### AUTHOR CONTRIBUTIONS

JB drafted the manuscript. CB wrote the introduction section, KB wrote the bladder cancer treatments section, and all authors, mainly AP, SG, TC, FS, and KB were responsible for conception, manuscript review, and critical intellectual input.

#### REFERENCES


non-muscle-invasive bladder cancer. Clin. Cancer Res. 23, 6592–6601. doi: 10.1158/1078-0432.CCR-17-1082


the LEEL method in a mouse bladder cancer model. J. Control. Release 196, 161–167. doi: 10.1016/j.jconrel.2014.10.007


**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 Buss, Begnini, Bender, Pohlmann, Guterres, Collares and Seixas. 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.

# Modulating the Tumor Microenvironment to Enhance Tumor Nanomedicine Delivery

Bo Zhang1, 2, Yu Hu<sup>2</sup> \* and Zhiqing Pang<sup>1</sup> \*

*<sup>1</sup> School of Pharmacy, Fudan University, Key Laboratory of Smart Drug Delivery, Ministry of Education, Shanghai, China, 2 Institute of Hematology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China*

Nanomedicines including liposomes, micelles, and nanoparticles based on the enhanced permeability and retention (EPR) effect have become the mainstream for tumor treatment owing to their superiority over conventional anticancer agents. Advanced design of nanomedicine including active targeting nanomedicine, tumor-responsive nanomedicine, and optimization of physicochemical properties to enable highly effective delivery of nanomedicine to tumors has further improved their therapeutic benefits. However, these strategies still could not conquer the delivery barriers of a tumor microenvironment such as heterogeneous blood flow, dense extracellular matrix, abundant stroma cells, and high interstitial fluid pressure, which severely impaired vascular transport of nanomedicines, hindered their effective extravasation, and impeded their interstitial transport to realize uniform distribution inside tumors. Therefore, modulation of tumor microenvironment has now emerged as an important strategy to improve nanomedicine delivery to tumors. Here, we review the existing strategies and approaches for tumor microenvironment modulation to improve tumor perfusion for helping more nanomedicines to reach the tumor site, to facilitate nanomedicine extravasation for enhancing transvascular transport, and to improve interstitial transport for optimizing the distribution of nanomedicines. These strategies may provide an avenue for the development of new combination chemotherapeutic regimens and reassessment of previously suboptimal agents.

Keywords: tumor microenvironment, nanomedicine, tumor nanomedicine delivery, interstitial fluid pressure, tumor perfusion, extracellular matrix

#### INTRODUCTION

In recent times, nanomedicine delivery to tumors has attracted extensive attention in the field of tumor treatment (Allen and Cullis, 2004; Peer et al., 2007). The advantage of nanomedicines over free drugs is based on the enhanced permeability and retention (EPR) effect (Fang et al., 2011; Maeda, 2012). The fundamental characteristics of EPR physiology are highly permeable tumor vessels allowing the enhanced permeability (EP) of large particles including proteins, macromolecules, liposomes, micelles, and other particles large enough to avoid renal clearance, into the tumor interstitium combined with impaired lymphatic drainage limiting clearance and causing enhanced retention (ER) of those extravasated particles. Both features result from the rapid growth of a tumor and collapse of the existing blood

#### Edited by:

*Susan Hua, University of Newcastle, Australia*

#### Reviewed by:

*Saraswati Sukumar, School of Medicine, Johns Hopkins University, United States Jai Prakash, University of Twente, Netherlands*

\*Correspondence:

*Yu Hu dr\_huyu@126.com Zhiqing Pang zqpang@fudan.edu.cn*

#### Specialty section:

*This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology*

Received: *03 November 2017* Accepted: *15 December 2017* Published: *22 December 2017*

#### Citation:

*Zhang B, Hu Y and Pang Z (2017) Modulating the Tumor Microenvironment to Enhance Tumor Nanomedicine Delivery. Front. Pharmacol. 8:952. doi: 10.3389/fphar.2017.00952*

**87**

and lymph vessels in the limited interstitial space (Leu et al., 2000; Dreher et al., 2006). With EPR effect as the main principle for passive targeting strategy, nanomedicine delivery to tumors has achieved success to varying degrees. However, the clinical benefits of the three EPR-based Food and Drug Administration (FAD)-approved nanomedicines including pegylated liposomal doxorubicin (Doxil/Caelyx), liposomal daunorubicin (DaunoXome), and nanoparticle albumin-bound paclitaxel (Abraxane) for the treatment of solid tumors were demonstrated to be only modest (O'Brien et al., 2004; Gradishar et al., 2005; Jain and Stylianopoulos, 2010), posing considerable challenges for the clinical translation of new nanomedicines. Accumulating evidence revealed that EPRdependent drug delivery was always compromised by the tumor microenvironment including irregular vascular distribution, elevated tumor interstitial fluid pressure (IFP), poor blood flow, rich extracellular matrix (ECM) and abundant tumor stroma cells (Nichols and Bae, 2014). Delivery barriers posed by the tumor microenvironment are the main reasons responsible for the modest survival benefits of FDA-approved nanomedicines (Jain and Stylianopoulos, 2010).

The tumor microenvironment, as an important component of tumor tissues, functions as the soil for the seeds i.e., tumor cells to proliferate, differentiate, and promote tumor growth (Zhang et al., 2015, 2016a). The components of tumor microenvironment include the extracellular matrix (ECM) and different kinds of stromal cells such as tumor-associated fibroblasts (TAF), tumorassociated macrophages, and pericytes (Hanahan and Weinberg, 2011). As a pathologic condition, the tumor microenvironment is remarkably abnormal: tumor blood flow is low, perfusion is uneven in tumors, the tumor vessel permeability is highly heterogeneous, interstitial fluid pressure (IFP) is elevated, and a large number of active stromal cells and ECM are often dense and stiff.

Systemically administrated nanomedicines need to undergo a three-step process in solid tumors to achieve their therapeutic effect: vascular transportation to different areas of the tumor, trans-vascular transport across the vessel wall, and interstitial transport to reach the tumor cells (Wong et al., 2011). Delivery of nanomedicines differs markedly between tumors and normal tissues owing to structural differences. The abnormality in organization and structure of the tumor vasculature leads to heterogeneous blood flow, which directly influences the vascular transport of nanomedicines (Jain, 1994, 2005). Additionally, the vascular hyper-permeability and lack of functional lymphatic vessels inside tumors results in elevated IFP (Boucher et al., 1990), which not only compresses tumor vessels to aggravate the heterogeneous blood flow, but also reduces convective transport of nanomedicines (Jain, 1987). Besides, compression from proliferating tumor cells, stromal cells, and the ECM could also compress tumor vessels (Stylianopoulos et al., 2012). Furthermore, the dense ECM hinders interstitial diffusion of nanomedicines (Jain, 1987). When 90-nm liposomes, approximately the size of liposomal doxorubicin, an approved nanomedicine, was intravenously administered in tumor-bearing mice, these particles leaked out of tumor vessels but did not move far away from the vessel wall (Yuan et al., 1994a). Even directly intra-tumor injected 150-nm particles did not move far from the injection site (McKee et al., 2006). Altogether, the complex tumor microenvironment could negatively affect vascular transport, trans-vascular transport and interstitial transport of nanomedicine, and compromise nanomedicine delivery for tumor treatment.

To improve the therapeutic benefits of nanomedicine, different strategies including active targeting nanomedicine (Zhang et al., 2014a,b), tumor-responsive nanomedicine (Zhu et al., 2012; Huang et al., 2013), as well as optimization of the physiochemical parameters of nanomedicine such as size (Tong et al., 2013; Tang et al., 2014), charge (Han et al., 2015), and shape (Chauhan et al., 2011) have been developed. However, these methods rely on the advanced development of the nanomedicine itself, which could not conquer the abovementioned delivery barriers of the tumor microenvironment (Chauhan and Jain, 2013). Accordingly, a modification of the tumor microenvironment was recognized as an important tool to improve tumor nanomedicine delivery (Jain and Stylianopoulos, 2010; Miao et al., 2015). In this review, in terms of the three processes including vascular transport, trans-vascular transport and interstitial transport that nanomedicines need to experience before reaching tumor cells and achieving therapeutic benefit, we tried to summarize different strategies of modulating tumor microenvironment to improve tumor nanomedicine delivery from the corresponding three aspects including improving tumor perfusion, facilitating nanomedicine extravasation, and enhancing interstitial transport of nanomedicine.

#### NANOMEDICINE TRANSPORT BARRIERS FROM TUMOR MICROENVIRONMENT

The main transport barriers of a tumor microenvironment include abnormal tumor vasculature, elevated IFP, dense ECM, and stromal cells (**Figure 1**). These components varied with respect to different tumor types. To better understand and address the complexities of a tumor microenvironment, we used two representative models—highly permeable tumors and highly desmoplastic tumors (Stylianopoulos and Jain, 2013). Highly permeable tumors such as gliomas and melanomas are always rich in vessels and a small amount of pericytes, TAFs, and ECM, wherein tumor cells are in the vicinity of tumor vessels. On the other hand, highly desmoplastic tumors such as pancreatic cancers (Cabral et al., 2011), bladder cancers (Zhang et al., 2014), and some breast cancers (Stylianopoulos and Jain, 2013) are always hypovascular with numerous TAFs, a dense ECM, and high coverage rate of pericytes on the endothelium, such that the tumor cells are isolated into nests by TAF complexed with ECM and are a certain distance from the tumor vessels (Feig et al., 2012; Zhang et al., 2016c). Highly permeable and highly desmoplastic tumors were also referred to as tumors with tumorvessels architecture and tumors with stroma-vessels architecture, respectively, in some reports (Smith et al., 2013; Miao et al., 2016).

#### Abnormal Tumor Vasculature Networks

In normal tissues, an exquisite counterbalance is achieved between the proangiogenic molecules such as VEGF and endogenous antiangiogenic molecules such as sVEGFR1 and

cells to exert their therapeutic effects.

thrombospondins (Carmeliet and Jain, 2000; Jain, 2003). In tumor tissues, however, the proangiogenic effect is abnormally upregulated and the pathological angiogenesis occurs in a disorganized manner. Compared to normal vessels, tumor vessels are highly irregular and chaotic in structure with wide endothelial gaps and a heterogeneous basement membrane (Carmeliet and Jain, 2011). The tortuous and leaky nature of tumor vessels contributes to compromised tumor blood flow, as the high tortuosity of tumor vessels results in an elevated geometric resistance that retards the blood flow (Sevick and Jain, 1989a). Furthermore, this sharp drop of blood flow due to the geometric resistance exerts a marked influence on the viscous resistance of blood in tumor vessels (Sevick and Jain, 1989b). Additionally, the leaky nature of tumor vessels allows the permeability of fluid into the interstitium. Fluid loss would increase the hematocrit of tumor blood to elevate its viscosity and further impede tumor blood flow (Sevick and Jain, 1989b; Netti et al., 1996; Sun et al., 2007). This combined effect of tortuosity of and leakage from tumor vessels compromises blood flow in brain tumors to one to three orders of magnitude, slower than that in the pial vessels surrounding normal tissues (Yuan et al., 1994b). Inefficient blood flow in tumors always leads to poor delivery of systemically administered drugs. Therefore, the tumor vascular network poses a major barrier in vascular transportation of nanomedicine, specifically highly permeable tumors with a variety of hyper permeable vessels.

### Elevated IFP

IFP is a type of stress exerted by fluids and is uniformly elevated throughout the tumor bulk in many tumors. Fluid flow is involved in three processes including flow along the tumor vessels, through the tumor interstitium, and drainage of excessive fluid by the lymphatic vessels (Jain et al., 2014).

Abnormalities in the tumor microenvironment concerning these three processes lead to elevated IFP. The leaky tumor vessels allow the extravasation of excessive fluid and plasma macromolecules into the tumor interstitium. In normal tissues, excessive fluid could be drained by an effective lymphatic network to maintain a balanced tissue interstitial pressure. However, lymphatic drainage in tumors do not function properly and hence, IFP is elevated in tumor tissues. Apart from abnormality in tumor blood vessels and lymphatic vessels, abnormal hydraulic conductivity is also a regulator of IFP, especially in highly desmoplastic tumors. The hydraulic conductivity depends on the volume fraction, surface charge, chemical composition, and organization of fibers in the tumor interstitial space (Levick, 1987; Stylianopoulos et al., 2008). Tumors abundant in collagen could display an order of magnitude lower in hydraulic conductivity than those with low collagen content (Netti et al., 2000). The negatively charged glycosaminoglycans could increase flow resistance because of their ability to trap water (Levick, 1987). Therefore, depletion of glycosaminoglycans with matrix metalloproteinases-1 and -8 increases the hydraulic conductivity and thus, the interstitial fluid velocity (Mok et al., 2007). Furthermore, a high density of tumor cells and stromal cells might reduce the interstitial space available for fluid flow, thereby increasing fluid resistance and IFP. In normal tissues, IFP is in the range of 0–3 mm Hg. However, experimental and human solid tumors exhibit high IFP, typically ranging from 5 to 40 mmHg, which may reach 75–130 mmHg in highly desmoplastic pancreatic tumors (Milosevic et al., 2004; Provenzano et al., 2012). Therefore, in many tumors, elevated IFP and reduced microvascular pressure (MVP) hinder nanomedicine delivery by the following mechanisms. First, as IFP and MVP also impose fluid stresses on vessel walls, elevated IFP could compress tumor vessels to cause blood stasis and/or vessel collapse to reduce the vascular transport of nanomedicine. Another consequence of high IFP is that interstitial fluid might escape from the tumor periphery into surrounding normal tissue, carrying not only nanomedicine to reduce the trans-vascular transport of nanomedicine but also growth factors or tumor cells to drive tumor metastasis or drug resistance (Chary and Jain, 1989; Netti et al., 1996). Finally, IFP might force the nanomedicine to extravasate by passive diffusion, instead of convention, a much faster transport process, which compromises the interstitial transport of nanomedicine especially larger nanomedicines. To conclude, elevated IFP inside tumors hinders vascular, trans-vascular, and interstitial transport of nanomedicines.

#### ECM

The ECM is the non-cellular component widely present within all tissues and organs. ECM is mainly composed of two types of macromolecules: proteoglycans (PGs) and fibrous proteins (Järveläinen et al., 2009; Schaefer and Schaefer, 2010). PGs such as hyaluronan fill the majority of the extracellular interstitium of the tissue in the form of a hydrated gel. The fibrous ECM proteins include collagens, elastins, fibronectins, and laminins (Dequidt et al., 2007). ECM is a highly dynamic structure being constantly remodeled, either enzymatically or non-enzymatically, and its final components are controlled by a myriad of post-translational modifications. ECM is tissue-specific and the component varies greatly among different tissues including cancerous ones (Frantz et al., 2010).

Under normal conditions, the unique composition and structure of the ECM functions as a growth regulator. ECM and ECM-associated enzymes and growth factors regulate cell proliferation and differentiation, maintaining cell survival and dynamic homeostasis (Li et al., 2010). However, ECM is commonly deregulated and becomes disorganized in diseases such as cancer. Fibrosis due to excessive ECM production or limited ECM turnover occurs in many types of cancers. Especially in highly desmoplastic tumor such as pancreatic cancer and some breast cancers (Stylianopoulos and Jain, 2013), a dense ECM composed of collagen, hyaluronan, and fibronectin is always found (Feig et al., 2012; Zhang et al., 2016c). In contrast, tumors with abundant vessels always harbor a low level of ECM. Signaling pathways involved in ECM production included transforming growth factor-beta (TGF-β), Hedgehog (Hh) signaling, and platelet-derived growth factor (PDGF). ECM turnover is subjected to enzyme-mediated remolding including heparanase, cysteine proteases, 6-O-sulfatases, urokinase, and many matrix metalloproteinases (MMPs) (Egeblad et al., 2010; Lu et al., 2012). In highly vascularized and permeable tumors such as glioma and melanoma, ECM is always scarce.

The dense ECM in the tumor interstitium not only compressed tumor vasculature and reduced vascular transport of nanomedicine but also isolates tumor cells into nests within a certain distance from collapsed vessels and resists the free penetration and homogeneous distribution of nanomedicine in three main ways (Bailey et al., 2008; Miao et al., 2015). First, limited interstitial volume plus high stromal fraction and large matrix molecules result in a dense network (Padera et al., 2004), effectively reducing blood flow and limiting convection of nanomedicine. Second, the fibrillar structure, mesh size, and collagen thickness directly limit the diffusion of nanomedicine. The diffusion capacity is inversely related to the size of nanomedicine. Matrix mesh size ranges between 20 and 40 nm in solid tumors. Particles larger than the mesh size are completely prevented from diffusing through the ECM, those near the mesh size can be hindered to a certain extent, and only small particles can penetrate almost freely (Nichols and Bae, 2012). Third, the tortuous nature of the interstitial space poses an additional barrier for drugs of all size, because it elongates the diffusion path of the nanomedicine from blood vessels to tumor cells (Chauhan et al., 2009). The resistance of nanomedicine delivery from ECM mainly occurs in highly desmoplastic tumors. For highly permeable tumors with tumorvessels architecture, the ECM is not as abundant, dense, and stiff as that in highly desmoplastic tumors, the and nanomedicine can penetrate throughout the tumor tissues much more easily after its extravasation from tumor vessels (Cabral et al., 2011). Unfortunately, the ECM is much more denser and thicker in human tumors than in mouse models (Miao and Huang, 2015). In conclusion,rich ECM in tumors resists vascular and interstitial transport of nanomedicine.

#### Stromal Cells

Stromal cells include TAF, tumor-associated macrophages (TAM), and pericytes. The origin of TAF is still debatable. TAF probably originated from resident tissue fibroblasts, bone marrow-derived mesenchymal stem cells, hematopoietic stem cells, epithelial cells (epithelial-mesenchymal transition; EMT), and endothelial cells (endothelial-mesenchymal transition; EndMT) (Shiga et al., 2015). It was now widely accepted that TAF significantly contributes to cancer progression (Brennen et al., 2012). TAF is abundant in highly desmoplastic tumors and produces large amounts of ECM to isolate tumor cells into a nest. TAF has been regarded as the major component of tumor stroma and contributes to the binding-site barrier for interstitial transport of nanomedicine (Miao et al., 2016). Large numbers of TAF associated with dense ECM also compress tumor vessels to compromise the vascular transport of nanomedicine. Besides, reports have shown that uptake of the anisamide ligand-modified nanomedicine by TAF was 7-fold higher than that of the other cells because of the different expression level of the sigma receptor between TAF and other cells (Miao et al., 2016).

TAM is the major cancer-related inflammatory cell primarily converted from monocytes that are closely associated with the prognosis of many cancer types. Other inflammatory cells include granulocytes, dendritic cells, and myeloid derived suppressor cells, which are also important constituents of the tumor microenvironment (Mocellin et al., 2001; Hu et al., 2016). When polarized toward the anti-inflammatory state by the tumor microenvironment, TAM promotes immune evasion and angiogenesis, thereby driving tumor growth (Cieslewicz et al., 2013). The off-target effect of nanomedicines is inevitable because of the phagocytic properties of inflammatory cells. Roode et al. showed that the association between TAM and NP were 4-fold greater than that of cancer cells despite TAM constituting only 1% of all cells in tumors (Roode et al., 2016). The off-target uptake of nanomedicine by stromal cells including TAF and TAM could certainly reduce the uptake of nanomedicine by tumor cells and therefore the therapeutic benefits.

Pericytes are another important type of stromal cells located mainly in the perivascular space, which also affect nanomedicine delivery. Neither leaky, immature vessels with little coverage nor over-mature vessels with high pericyte coverage are favorable for nanomedicine delivery. Excessively leaky vessels in highly vascularized tumors affect nanomedicine delivery mainly by compromising blood perfusion and thereby the vascular transport of nanomedicine (Stylianopoulos and Jain, 2013). In contrast, high pericyte coverage is always found in highly desmoplastic tumors, which reduce the endothelial gap and limit the trans-vascular transport of nanomedicine, especially for larger nanomedicine (Cabral et al., 2011).

### STRATEGIES TO MODULATE TUMOR MICROENVIRONMENT

In accordance with the three processes including vascular, transvascular and interstitial transport that nanomedicines need to experience before reaching tumor cells, strategies of modulating tumor microenvironment to improve nanomedicine delivery for tumor treatment can be divided into three categories: improving tumor perfusion, facilitating nanomedicine extravasation, and enhancing interstitial transport of nanomedicine (**Table 1**).

#### Improving Tumor Perfusion

#### Tumor Vasculature Normalization

The newly formed tumor vessels are always tortuous and leaky allowing the extravasation of nanomedicine but simultaneously increasing IFP, which prevents adequate and homogeneous blood flow and vascular transport of nanomedicine. To improve nanomedicine delivery for tumor treatment, normalization of vessels has emerged as an effective approach. Vessel normalization transforms the abnormal phenotype of tumor vessels into a phenotype that closely resembles that of fully functional normal vessels by repairing the basement membrane and increasing coverage rate of pericytes, and ultimately decreasing vessel leakiness. Optimizing the structure of tumor vessels could reduce the extravasation of excessive fluid and lower IFP, and then restore tumor blood flow, thereby improving vascular transport of nanomedicine. Many proangiogenic molecules including VEGF, fibroblast growth factor (FGF), and PDGF are overexpressed in tumors and involved in angiogenesis, which cause chaotic structural development in these newly formed tumor vessels (Goel et al., 2011). Therefore, strategies to block these proangiogenic signaling molecules were designed to repair tumor vessels. For example, VEGF inhibitors Bevacizumab, the FDA-approved antiangiogenic monoclonal antibody (mAb), capable of reverting abnormal structure of tumor vessels toward a more normal phenotype have been applied in the treatment of metastatic colorectal cancer (Salgaller, 2003; Ellis, 2005; **Table 1**), which were of high potential to improve nanomedicine delivery for tumor treatment. Moreover, some angiogenic signaling pathways such as mTOR signaling (Guo et al., 2014), Notch 1 signaling (Maes et al., 2014), and D2 receptors-angiopoietin 1 signaling (Chakroborty et al., 2011) involved in vessel normalization have also been modulated to improve nanomedicine delivery (**Table 1**). In our previous research, it was also shown imatinib mesylate (IMA) could normalize the tumor vessels of A549 tumors by inhibiting platelet-derived growth factor signaling pathway (Zhang et al., 2016d). Interestedly, IMA treatment could significantly reduce the accumulation of nanoparticles (NPs) around 110 nm but enhanced the accumulation of micelles around 23 nm. Furthermore, IMA treatment limited the distribution of NPs inside tumors but increased that of micelles with a more homogeneous pattern (**Figure 2**). Finally, the anti-tumor efficacy study displayed that IMA pretreatment could significantly increase the therapeutic effects of paclitaxelloaded micelles. As tumor vessel normalization minimized endothelial gap, it could prevent tumor cells shedding into tumor vessels, and reduce the possible tumor metastasis to a certain degree.

To utilize vessel normalization strategy to improve nanomedicine delivery for tumor treatment, four concerns should be kept in mind. First, the strategy can only improve the delivery of small molecular weight drugs or relatively smaller nanomedicines ranging from 20 to 40 nm, but reduces the delivery of large nanomedicines around 100 nm as it decreases the endothelial gap of tumor vessels (Chauhan et al., 2012; Zhang et al., 2016d, 2017b). Second, the normalization is transient and the followed nanomedicine should be applied in the normalization window. Third, a judicious dose of vascular normalizer is highly recommended to prevent excessive pruning of tumor vessels, which might impair vascular efficiency and thus the delivery of concurrent therapy (Huang et al., 2012; Jain, 2013). Fourth, as vasculatures are always severely compressed in highly desmoplastic tumors and are refractory to vasculature normalizers (Smith et al., 2013), this strategy might only be used in tumors that are highly permeable but not highly desmoplastic, or at least combined with other strategies capable of reopening compressed vessels.

#### Tumor Vessel Dilation

Vasoconstrictive endothelin-1 (ET1) and its receptor ETA via which ET-1 mediates vasoconstriction are both abundant in tumor tissues for maintaining the contractile tone of tumor vessels. The expression level of ET1 and ETA in tumor vessels was 13-fold and 5-folder higher than that of size-matched normal vessels, respectively (Sonveaux et al., 2004). BQ123, a selective antagonist against ETA could inhibit ET1-ETA signaling, induce tumor vessel dilation, and trigger a tumorspecific increase in blood flow. The blood flow improvement induced by BQ123 improved the delivery of free drugs to tumors despite an increase in IFP (Martinive et al., 2006). In addition, it was also demonstrated that BQ123 could increase the delivery of photothermal nanomedicine around 100 nm for effective photothermal therapy of tumors (Wang et al., 2017). Some inflammation factors such as bradykinin capable


of dilating vessels could also directly increase tumor perfusion. In our previous study, it was shown captopril, a widely used hypotensor in clinics, could dilate tumor blood vessels by increasing bradykinin expression and even increase tumor vessel permeability to enhance nanomedicine delivery for tumor therapy (Zhang et al., 2017a).

#### Facilitating Nanomedicine Extravasation Inflammatory Mediators for Enhanced Permeability

Inflammatory mediators such as TNFα (Seki et al., 2011), prostaglandin analogs (Tanaka et al., 2003), VEGF (Monsky et al., 1999), and nitric oxide (NO) donors (Seki et al., 2009), capable of enhancing vascular permeability, have been utilized to increase nanomedicine accumulation in tumors up to 2– 6-fold higher than that of the control group (**Table 1**). Apart from vascular permeability enhancement, vasodilatation and blood-flow improvement by usage of inflammatory mediators were also involved in improving nanomedicine delivery for tumors. However, a series of effects of inflammatory mediators mentioned above could also lead to elevated IFP against nanomedicine delivery. Thus, accumulation of nanomedicine in tumors comprehensively depends on these factors. As inflammation might potentially promote cancer development (Atsumi et al., 2014), local application (Seki et al., 2011) or

targeted delivery of inflammatory mediators to the tumor site should be adopted.

#### Pericyte Depletion

In highly desmoplastic tumor, the coverage rate of pericytes on endothelium was about 70%, much higher than highly permeable tumors, which significantly limit the transvascular movement of nanomedicine into tumor interstium. Therefore, strategies by using low dose of a TGF-β inhibitor, LY364947 was developed to reduce the pericyte coverage of endothelium and increase size gaps between endothelium to increase therapeutic benefits of gemcitabine-loaded liposomes for pancreatic cancer (Cabral et al., 2011; Meng et al., 2013) and Doxil for diffuse-type gastric cancer (Kano et al., 2007).

#### Platelet Depletion

It is well known that platelets contribute a lot to hemostasis. Apart from its role in thrombus formation, platelets are highly involved in tumor progression and metastasis. In addition, it could also support tumor vascular homeostasis and protect the integrity of tumor vessels (Kisucka et al., 2006; Ho-Tin-Noé et al., 2009). Studies showed that platelets reduction induced bleeding in the tumor site and increased leakiness of tumor vasculature. Therefore, platelets reduction in thrombocytopenic mice increased efficacy of chemotherapy for breast cancer (Demers et al., 2011). To avoid potential bleeding in normal organs caused by low platelet counts, a recent study by Li et al. designed a tumor microenvironment-responsive nanoparticle capable of delivering antiplatelet antibody R300 to selectively deplete platelet in tumor tissues, therefore augmenting vascular permeability and improving nanomedicine delivery for tumors (Li et al., 2017). Platelet depletion represented as a promising approach to augment transvascular delivery of nanomedicine to tumors.

#### Physical Stimulus

Radiation can improve tumor-targeted delivery of nanomedicine (Davies Cde et al., 2004; Giustini et al., 2012). Some possible mechanisms are as follows: First, radiation could upregulate the level of vascular endothelial growth factor (VEGF) by activating hypoxia inducible factor 1 (HIF1) (Moeller et al., 2004; Stapleton et al., 2016) or via multiple mitogen-activated protein kinase dependent pathways (Park et al., 2001) to increase the permeability of tumor vessels (Kobayashi et al., 2004). Results showed that the tumor vessels' permeability of magnetic resonance imaging-contrast agent with the molecular weight above 200 kDa was increased by 32.8% after irradiation (10 Gy). In addition, radiation can rapidly kill the sensitive tumor cells. The reduced cell density helped to alleviate compression stress from tumor cells, reopen collapsed vessels, and therefore increase tumor blood flow (Nagano et al., 2008; Khawar et al., 2015). The effect of radiation on tumors is complex and dose-, time- , and tumor-type dependent (Garcia-Barros et al., 2003; Fuks and Kolesnick, 2005; Kioi et al., 2010). Milosevic's recent review provides further evidence of the same (Stapleton et al., 2016).

Kong et al. pioneered the use of mild hyperthermia (HT) for nanomedicine extravasation into tumor tissues by improved vascular permeability (Kong et al., 2000, 2001). Studies further demonstrated that mild HT could also help improve tumor perfusion and reduce IFP (Sen et al., 2011; Winslow et al., 2015), probably via creation of vascular fenestrations and perturbation of the vascular endothelium (Kirui et al., 2015), and thus allow deep nanomedicine penetration throughout the tumors, rather than perivascular accumulation (Li et al., 2013). However, there was no direct evidence to demonstrate the pore size change of tumor vessels after HT treatment. The extravasation depth and intensity in tumor interstitium was considered to vary greatly among different types of tumors, which depends on the pattern of the endothelial lining and the intrinsic property of the surrounding tumor microenvironment (Eberhard et al., 2000), such as structure of the interstitial matrix. Apart from tumors with a certain vascular component, highly desmoplastic tumors could also respond well to mild HT treatment (Kirui et al., 2014, 2015). The temperature is a crucial element in the heating method, wherein results showed that 41–43◦C was appropriate, because a very high temperature might damage the endothelial lining of tumor vessels and induce coagulation response. Thrombin formation could choke the vessels and compromise nanomedicine delivery. Alternatively, insufficient temperature might exert a very minimal effect on the tumor vessel endothelium to increase the endothelial gap (von Maltzahn et al., 2011; Li et al., 2013).

Ultrasound has been used to improve nanomedicine delivery for tumors by both mechanical and HT effect (Goins et al., 2016). For mechanical effects, many reports showed that gasfilled bubbles could be used to transiently produce pores in blood vessels (Durymanov et al., 2015) or cell membranes (sonoporation) (Yoon et al., 2014; Ma et al., 2016), through which nanomedicines of different types can effectively extravasate tumor vessels or enter into tumor cells (Thakkar et al., 2013), therefore achieving improved delivery of nanomedicine (Rapoport et al., 2013, 2015). Besides, ultrasound also produces heat at an acoustic intensity and in a time-dependent manner. Recently, Frazier used magnetic resonance imaging-guided, highintensity focused ultrasound (HIFU) to produce a spatially uniform 43◦C heating pattern in a xenograft tumor model and improved the accumulation of Evans blue dye in heated tumors to nearly 2-fold higher than in unheated tumors (Frazier et al., 2016).

#### Improving Interstitial Transport of Nanomedicine ECM Disruption Strategies

Dense ECM always resists free penetration of nanomedicine throughout tumor tissues to reach tumor cells. Therefore, modification of ECM has been extensively explored to improve the delivery and distribution of nanomedicine in tumor tissues. The ECM modification strategy includes direct ECM disruption

and reduction of ECM synthesis by inhibiting TAF activity. Several studies have shown that different kinds of enzymes directly degrade the components of ECM such as collagen and hyaluronic acid and can improve the delivery of nanomedicine (**Table 1**). For instance, collagenase-coated nanomedicine could penetrate deeper into the core of in vitro tumor spheroids than control ones (Goodman et al., 2007; Cui et al., 2013). Besides, enzymatic digestion of collagen and decorin facilitates >10-fold increase in the diffusion of macromolecular dextran into tumor tissue, supporting matrix degradation as a useful tool to improve macromolecule distribution (Magzoub et al., 2008). In another study, intravenous injection of collagenase-1 into xenograft-bearing mice models increased the accumulation and gene expression of lipoplex in tumors by 1.5- and 2 fold, respectively, further confirming collagen digestion to be a useful strategy to improve nanomedicine delivery (Kato et al., 2012). Hyaluronan, or hyaluronic acid, a large linear glycosaminoglycan, composed of repeating N-acetyl glucosamine and glucuronic acid units, was also a crucial component of ECM (Provenzano et al., 2012), which was found abundant in non-small cell lung cancer (NSCLC), prostate, pancreatic, and breast cancers. Hyaluronidase was shown to induce a 4 fold increase in the distribution of liposomal doxorubicin in a human osteosarcoma xenograft model (Eikenes et al., 2005). A phase 1b trial of docetaxel combining PEGPH20 in metastatic

refractory NSCLC has been completed (NCT02346370) with results pending (Wong et al., 2017). In addition, a nanomedicine combining PEGylated hyaluronidase (PEGPH20) to improve the efficiency of chemotherapeutics for hyaluronan-high pancreatic cancer is currently in phase 3 clinical trial (NCT02715804) (Provenzano et al., 2012; Wong et al., 2017). Lysyl oxidase (LOX) is a key element in the crosslinking of collagen and increasing the stiffness of collagen fibers (Egeblad et al., 2010; Kanapathipillai et al., 2012). LOX-activity inhibition has proven successful in preventing ECM remodeling and stiffening (Levental et al., 2009; Barry-Hamilton et al., 2010), which may overcome the deregulated ECM barrier for nanomedicine delivery (Khawar et al., 2015).

It was noteworthy that ECM disruption was seldom used to increase nanomedicine delivery in highly vascularized tumors, which might be the relative lack of ECM in these tumors. However, our group found that fibrin, a kind of ECM component was rich in tumors harboring rich tumor vessels (Dvorak, 1986). The reason might be due to leakage of coagulation factors from circulation to tumor tissues and the high express level of tissue factor on tumor cells, both of which together contribute to local coagulation response in tumor tissues (Dvorak et al., 1985; Liu et al., 2011). As the end product of coagulation response, fibrin is mostly covalently cross-linked in tumor interstitium as an important component of tumor ECM (Dvorak, 1986; Pilch et al., 2006) and is mainly located in the vicinity tumor vessels (Nakahara et al., 2006), a distinct distribution pattern totally different from that of other components of matrix such as collagen and hyaluronic acid, which are always extensively distributed throughout tumor tissues. The special distribution pattern of fibrin was demonstrated to compress tumor vessels nearby, which reduce blood flow and compromise nanomedicine delivery for tumors. Moreover, as fibrin is always covalently cross-linked near tumor vessels, the penetration of nanomedicines in the tumor interstitium could also be hindered. Treatment with rtPA, a clinically widely used drug, at a dose of 25 mg/kg for 2 weeks, could safely and successfully deplete fibrin deposition, reopen compressed tumor vessels, reduce erythrocytes retention in tumor vessels, improve tumor blood flow, and further enhance the accumulation and penetration of nanoparticles (**Figure 3**; Zhang et al., 2016b).

However, systematic treatment with ECM disruptors such as collagenase, hyaluronidase, and rtPA may cause damage to healthy tissues, and site-specific action might be safer for clinical transformation. Therefore, tumor-specific degradation of ECM was achieved by coating nanomedicines with specific ECM enzymes (Zhou et al., 2016) or PEGlated ECM enzymes (Hingorani et al., 2016). Zhou et al. showed that hyaluronidase modified on the surface of nanoparticles was more effective than free hyaluronidase to help facilitate nanoparticle diffusion and achieved better therapeutic benefits (Zhou et al., 2016). Reports also showed that the dose of hyaluronidase seemed critical, because high doses of hyaluronidase may collapse water-swelling cage structures of hyaluronan rendering the ECM more viscous and less permeable, thereby reducing the diffusion coefficient of nanomedicine (Eikenes et al., 2010). In another research, Bromelain, a crude enzymatic complex purified from pineapple stems that belongs to the peptidase papain family, was decorated to mesoporous silica nanoparticles (Br-MSN), which showed an increased ability to digest and diffuse in tumor ECM in vitro and in vivo (Parodi et al., 2014).

TAF was mainly responsible for ECM production. Resting fibroblasts were transformed to TAF by cancer-derived growth factors such as TGF-β, Hedgehog moiety (Olive et al., 2009; Stylianopoulos et al., 2012), and PDGF (Olson and Hanahan, 2009). This trans-differentiation process of TAF is always characterized by the encoding of ECM-associated components such as collagens, hyaluronan, fibronectin, and MMPs (Cirri and Chiarugi, 2011). Therefore, ECM deregulation can be realized by blocking the growth factors involved in signaling for TAF stimulation. Taking TGF-β-associated signaling as an example, antibodies or other agents capable of blocking TGFβ signaling have proven to inhibit collagen synthesis and enhance nanomedicine delivery in xenograft models. The TGFβ neutralizing antibody ID11 improved tumor tissue delivery of Doxil and obtained better control of mammary carcinoma in xenograft models (Liu et al., 2012). Recently, our group used cyclopamine, a naturally occurring steroidal alkaloid, to inhibit the Hedgehog signaling pathway which contributes a lot to ECM production in pancreatic carcinoma by acting on the Smoothened (SMO) receptor (Heretsch et al., 2010). Compared with the control group, cyclopamine treatment successfully disrupt ECM in in pancreatic cancer, increased functional vessels about 2 folds at a dose of 50 mg/kg for 3 weeks, and significantly improved the accumulation (by a 2.6-fold) and penetration of nanoparticles in tumor tissues (Zhang et al., 2016c).

ECM modulation suits nanomedicine regardless of their size, because ECM modification could reopen compressed vasculatures to improve tumor blood flow and decrease the hindrance of nanomedicine penetrating ECM to reach tumor cells (Chauhan and Jain, 2013). Generally speaking, this strategy most benefits the delivery of larger nanomedicines (Cabral et al., 2011; Jacobetz et al., 2012), as they are more hindered by the ECM (Pluen et al., 2001).

#### Stromal Cell Reprogramming or Depletion Strategies

Apart from dense ECM, desmoplastic tumors always harbor a high density of stromal cells, among which TAF has been regarded as the major component of tumoral stroma as a potential therapeutic target for nanomedicine delivery. TAF depletion can improve the interstitial transport and distribution of nanomedicine by optimization of the tumor interstitium. Quercetin nanoparticles, capable of suppressing Wnt16 expression could reduce the number of TAF and improve nanomedicine delivery to bladder carcinoma (Hu et al., 2017). Inspired by the close association between cyclooxygenase-2 (COX-2) and tumor-associated angiogenesis, as well as tumor matrix formation, our group explored the tumor microenvironment modulation effect of celecoxib, a special COX-2 inhibitor widely used in clinics. Very interestingly, oral celecoxib treatment at a dose of 200 mg/kg for 2 weeks could successfully normalized the tumor microenvironment, including tumor-associated fibroblast depletion, fibronectin bundle disruption, tumor vessel normalization, and tumor

perfusion improvement. Furthermore, it also significantly enhanced the in vivo accumulation and deep penetration of 22-nm micelles rather than 100-nm nanoparticles in tumor tissues and improved the therapeutic efficacy of paclitaxel-loaded micelles in tumor xenograft-bearing mouse models (Zhang et al., 2017b).

Although TAF depletion could undoubtedly modify the tumor microenvironment to improve nanomedicine delivery, recent studies have also indicated that direct TAF depletion might drive tumor metastasis and progression (Özdemir et al., 2014), suggesting a paradoxical effect of TAF depletion. One explanation for this paradoxical effect is that the TAF-depleting strategy runs the risk of eliminating the key element needed for tissue homeostasis (Miao et al., 2015). To avoid this paradox, an alternative approach is to transform activated TAF to a dormant form. Losartan is a antihypertensive agent with antifibrosis properties. Research has shown that losartan treatment reprogramed TAF and reduced the collagen and hyaluronan content in desmoplastic models of human breast and pancreatic tumors in mice and improved the distribution and therapeutic effects of systematic administered Doxil (Diop-Frimpong et al., 2011; Chauhan et al., 2013). The antifibrotic effect of losartan was associated with reduced number of activated TAFs and therefore decreased expression of downstream profibrotic factors, such as connective tissue growth factor (CTGF), TGF-β1, and ET-1 (Chauhan et al., 2013), which led to an ongoing clinical trial of losartan combined with chemotherapy in pancreatic tumors (NCT01821729). Vitamin D receptor (VDR) and Wnt-β-Catenin signaling pathway was upregulated in pancreatic stellate cell (PSC), a form of TAF in PDA (Omary et al., 2007). Other studies showed that the VDR ligand and all-trans retinoic acid (ATRA) can act through VDR or Wnt-β-Catenin signaling pathway to reprogram PSC to the quiescent state to reduce the fibrotic content in tumor interstitium (Froeling et al., 2011; Sherman et al., 2014; Chronopoulos et al., 2016), which is promising for the second-wave nanomedicine therapy.

TAMs are prominent components and critical modulators of the tumor microenvironment and contribute to tumor development, invasion, and metastases. Evidences have shown TAMs protect tumor cells from chemotherapy and suppress the immune response of cytotoxic T cells (Jinushi et al., 2011), highlighting the essential of targeting TAMs for cancer treatment. TAM depletion could decrease off-target uptake of nanomedicine by TAMs and thus increase drug delivery to tumor cells. However, the effect of selective TAM depletion on nanomedicine delivery for tumor was seldom reported, and TAM depletion might be a new strategy for tumor microenvironment modulation to enhance tumor nanomedicine delivery.

### FUTURE PERSPECTIVES

Nanomedicine drug delivery system has attracted extensive attention in the field of tumor treatment. The complex tumor microenvironment including structural abnormalities in tumor vessels, dense ECM structure, and high density of stromal cells as well as physicochemical environment such as elevated IFP pose barriers and compromise the delivery of nanomedicines**.** In this review, we summarized these barriers and provided strategies to overcome them for improved nanomedicine delivery.

However, some aspects deserve special attention. First, tumors could be generally divided into two types such as those with abundant permeable but uncompressed vessels and tumors with dense ECM and large amount of TAF. More precise classification of tumor microenvironment-like tumor cells might be urgently needed for precise tumor nanomedicine delivery. We need to adopt different strategies according to the characteristics of the tumor microenvironment. For example, tumor vessel normalization is more effective for tumors with abundant, highly permeable but not compressed vessels, but not so appropriate for highly desmoplastic tumors. ECM disruption strategy demonstrates promising prospects to enhance nanomedicine delivery for tumors with abundant ECM even in clinical trials. As for tumors with a certain amount of both vessels and ECM, strategies capable of modulating vessels and ECM should be combined to obtain an optimal effect. In addition, it is important to study new animal models capable of quantitative analysis of parameters involved in the tumor microenvironment, such as IFP to quantify the negative contribution of tumor microenvironment parameters to tumor nanomedicine delivery and help develop corresponding strategies. Furthermore, the tumor microenvironment is too complex and one strategy might have multiple modulation effects on the tumor microenvironment. With respect to ECM disruption, it not only disrupt tumor ECM to optimize interstitial transport of nanomedicines, but also alleviate compression for tumor vessels to improve tumor perfusion to bring more nanomedicines to the tumor site. Another good illustration is the tumor vessel normalization strategy. Tumor vessel normalization could repair tumor vessels structure, which could improve tumor perfusion and reduce IFP to increase the delivery of small nanomedicine around 20–40 nm. However, it inversely compromised the delivery of larger nanomedicine around 100 nm because of the size reduction of endothelial gaps.

Therefore, nanomedicines with suitable qualities should be combined with approaches modulating the tumor microenvironment to overcome nanomedicine transport barriers that the advanced design of nanomedicines cannot conquer. With these in mind, we believe nanomedicines of the future could be far more effective than those available at present.

### AUTHOR CONTRIBUTIONS

ZP and YH conceived the principal idea and revised the manuscripts. BZ and ZP co-wrote the manuscript.

### ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (81773283, 81472757, 81361140344, and 81600175).

### REFERENCES


carriers for intra-tumoral chemotherapy. Adv. Healthc. Mater. 2, 1236–1245. doi: 10.1002/adhm.201200467


carcinomas transplanted in rat and mouse cranial windows. Cancer Res. 54, 4564–4568.


**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 Zhang, Hu and Pang. 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.

# Targeting Accessories to the Crime: Nanoparticle Nucleic Acid Delivery to the Tumor Microenvironment

Emily B. Harrison<sup>1</sup> , Salma H. Azam<sup>2</sup> and Chad V. Pecot3,4,5 \*

<sup>1</sup> Center for Nanotechnology in Drug Delivery, UNC Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>2</sup> Curriculum in Genetics and Molecular Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>3</sup> UNC Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>4</sup> Division of Hematology/Oncology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>5</sup> Department of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States

Nucleic acid delivery for cancer holds extraordinary promise. Increasing expression of tumor suppressor genes or inhibition of oncogenes in cancer cells has important therapeutic potential. However, several barriers impair progress in cancer gene delivery. These include effective delivery to cancer cells and relevant intracellular compartments. Although viral gene delivery can be effective, it has the disadvantages of being immunostimulatory, potentially mutagenic and lacking temporal control. Various nanoparticle (NP) platforms have been developed to overcome nucleic acid delivery hurdles, but several challenges still exist. One such challenge has been the accumulation of NPs in non-cancer cells within the tumor microenvironment (TME) as well as the circulation. While uptake by these cancer-associated cells is considered to be an off-target effect in some contexts, several strategies have now emerged to utilize NP-mediated gene delivery to intentionally alter the TME. For example, the similarity of NPs in shape and size to pathogens promotes uptake by antigen presenting cells, which can be used to increase immune stimulation and promote tumor killing by T-lymphocytes. In the era of immunotherapy, boosting the ability of the immune system to eliminate cancer cells has proven to be an exciting new area in cancer nanotechnology. Given the importance of cancer-associated cells in tumor growth and metastasis, targeting these cells in the TME opens up new therapeutic applications for NPs. This review will cover evidence for non-cancer cell accumulation of NPs in animal models and patients, summarize characteristics that promote NP delivery to different cell types, and describe several therapeutic strategies for gene modification within the TME.

Keywords: gene delivery, cancer, nanomaterials, immuno-oncology, cell targeting

#### INTRODUCTION

Over the last several decades, impressive advancements in nucleic acid delivery have brought these technologies to the clinic. Three gene therapies have been approved by the U.S. Food and Drug Administration (FDA), all of which rely on viral delivery systems. Non-viral systems are less immunogenic than viral systems, which may be required in cases where repeat administration is necessary, and they are non-mutagenic. While stable viral integration of genes may be a beneficial

#### Edited by:

Sherry Y. Wu, The University of Queensland, Australia

#### Reviewed by:

Domenico Criscuolo, Genovax S.r.l., Italy Chang Hsin-I, National Chiayi University, Taiwan

> \*Correspondence: Chad V. Pecot pecot@email.unc.edu

#### Specialty section:

This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology

Received: 24 November 2017 Accepted: 16 March 2018 Published: 04 April 2018

#### Citation:

Harrison EB, Azam SH and Pecot CV (2018) Targeting Accessories to the Crime: Nanoparticle Nucleic Acid Delivery to the Tumor Microenvironment. Front. Pharmacol. 9:307. doi: 10.3389/fphar.2018.00307

treatment for genetic disorders, a more transient regulation of gene expression may be preferred in other contexts. Non-viral nucleic acid delivery has not yet reached FDA approval, but several nanoparticles (NP)-based therapeutics are currently in clinical trials. For a thorough review of non-viral nucleic acid therapies in clinical development, we suggest (Yin et al., 2014). As these delivery platforms reach regulatory approval in the United States and elsewhere, they will pave the way for nucleic acid therapeutics in cancer and other diseases.

In parallel with advancements in nucleic acid delivery, the development of immunotherapies has revolutionized cancer treatment. Although historically cancer therapies have focused on directly killing cancer cells through chemotherapy and radiation, the success of immune checkpoint inhibitors and chimeric antigen receptor (CAR) T-cells has demonstrated that turning the immune tumor microenvironment (TME) against cancers can have strong therapeutic effects (Hoos, 2016). However, these immune-oncology drugs are only effective for subsets of patients, suggesting that additional factors are at play. An immune suppressive TME is one critical factor that can hamper T-cell invasion and anti-tumor effects. Taken together, harnessing NP-based nucleic acid delivery to the TME could transform a pro-tumoral and immuno-suppressive TME into a toxic environment for cancer cells. Here we review pre-clinical studies that demonstrate the feasibility of nucleic acid delivery to the TME for cancer therapy.

### NANOPARTICLES FOR NUCLEIC ACID DELIVERY

Naked nucleic acids display unfavorable biodistribution and pharmacokinetics: once injected into the blood stream, RNA and DNA can be quickly degraded by nucleases, phagocytosed by immune cells in the blood, or excreted through the kidneys. Therefore, to be effective, nucleic acids require delivery vehicles (Yin et al., 2014). All nucleic acids share a similar chemical structure: repeated nucleotides each composed of a five-carbon sugar linked to a nitrogenous base and connected by a phosphate backbone. While nucleic acids vary in size and contain either ribose or deoxyribose (in RNA and DNA, respectively), they are all negatively charged and hydrophilic. These properties allow them to be efficiently encapsulated into NPs. NPs are a diverse group of biomaterials that form structures in the nanometer scale. These include particles made of gold, silica, polymers, lipids, and others. While there are exceptions, lipids and polymers are the most common materials used for delivery of nucleic acids. For example, most commercially available transfection reagents use cationic lipids for effective intracellular delivery of DNA and RNAs such as mRNA, microRNA (miRNA) and short interfering RNAs (siRNAs) in vitro. In large part, in vivo nucleic acid delivery relies on similar principles but faces additional barriers such as stability in the circulation and delivery to target cells.

#### Lipids

Lipid systems for in vitro gene delivery were first developed in the 1980s and were primarily composed of amphiphilic cationic lipids (Felgner et al., 1987). These molecules contain a polar head group, linker, and fatty acid chains that selfassemble into micellar, lamellar, or hexagonal structures in water: examples include N-[1-(2,3-dioleyloxy)propyl]-N,N,Ntrimethylammonium chloride (DOTMA) and 1,2-Dioleoyl-3-trimethylammonium-propane (DOTAP). Incorporation of cholesterol and neutral lipids such as 1,2-Dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) can also increase stability and transfection efficiency. Permanently charged lipids result in toxicity, therefore ionizable systems have been developed. Ionizable lipids are positively charged in mildly acidic conditions where they can complex with nucleic acids; however, they remain uncharged at neutral pH which avoids systemic toxicity (Rietwyk and Peer, 2017).

#### Polymers

Polymers can also be used to encapsulate nucleic acids for in vivo delivery. Generally, polymers can be divided into two groups: natural or synthetic. Biologically occurring molecules such as peptides, oligosaccharides, and even nucleic acids themselves are natural polymers. Synthetic polymers are chemically produced, such as poly(lactic-co-glycolic acid) (PLGA). Polymers can occur as a single repeating unit (homopolymers) or multiple unit types (copolymers). Additionally, polymers with discrete segments consisting of different repeating units, called blockcopolymers can be made with a variety of useful properties. The cationic polymers poly-L-lysine (PLL) and polyethylenimine (PEI) were the earliest polymers used for condensing DNA. PEI has superior transfection efficiency and has been developed for in vivo and clinical delivery of nucleic acids (Yin et al., 2014). Combining PLGA, which is safe, biodegradable, and forms stable NPs, with PEI into mixed polymer NPs allows for effective gene delivery in vivo (Bivas-Benita et al., 2004). The neutral polymer polyethylene glycol (PEG) has also been used to reduce electrostatic interactions of both polymer and lipid-based NPs in vivo (Storm et al., 1995). Another nucleic acid delivery agent gaining popularity is chitosan, a natural co-polymer that can act as a biodegradable gene delivery agent (Wang et al., 2011). There are a wide variety of materials used in nanoformulations for nucleic acid delivery. The combination of these diverse materials and formulation techniques results in NPs with an array of sizes, charges, and surface properties. The properties of NPs used for nucleic acid delivery to the TME in preclinical studies are summarized in **Table 1**.

### NP Delivery to Tumors

In the case of solid tumors, delivery to cancer cells is a formidable hurdle (Pecot et al., 2011), but evidence of tolerability and intracellular delivery has been demonstrated in phase I clinical trials for both lipid and polymeric NPs (Zuckerman and Davis, 2015). While leaky and inefficient vasculature can allow accumulation of NPs in the tumor (Prabhakar et al., 2013), solid tumors also have stroma that contains a vast milieu of noncancerous constituents that include fibroblasts, tumor-associated macrophages (TAMs), endothelial cells and extracellular matrix components (ECM) that additionally impair access to the tumor parenchyma. In the case of NPs loaded with small molecule drugs,


TABLE 1 | Characteristics of nanoparticles used to target the tumor microenvironment in pre-clinical models.

PEG, polyethylene glycol; PLL, poly-L-lysine; PEI, Polyethylenimine; TPP, tripolyphosphate; PLGA, poly(lactic-co-glycolic acid); PBAE, poly(β-aminoester); na, not applicable; nr, not reported.

delivery to TAMs in the TME may be beneficial for local and sustained release of drug (Miller et al., 2015). This is not the case for nucleic acid delivery, where not only is delivery to appropriate cells required, but the payload must also reach key intracellular compartments. While strategies to allow better penetrance of NPs through the tumor stroma are being explored, caution is warranted. Disruption of tumor stroma may remove important elements of nutrition and growth factors, but it can also promote resistance (Miao et al., 2015). One alternative strategy is to target tumor-associated cells within the TME for cancer therapy. Commonly found in the tumor periphery, these cells are the first to encounter NPs as they leave the circulation. Many cell types within the TME also express unique cell surface markers, which can be utilized for targeted delivery. Given the influence of TME cells on all of the hallmarks of cancer, this is an enticing direction to pursue (Hanahan and Coussens, 2012). TME cell types, their role in cancer biology, and surface markers commonly used to target them are summarized in **Figure 1**.

### TYPES OF NUCLEIC ACIDS FOR NP DELIVERY

#### DNA

Delivery of exogenous DNA offers a great therapeutic opportunity for cancer. One approach is overexpression of genes that can suppress or kill tumor cells. These genes can be human in origin, viral proteins, bacterial toxins, or proteins designed for desired functions. One such example is NP delivery of the gene encoding anti-tumoral viral protein E1A, which was successfully delivered and expressed in humans (Chang et al., 1997; Yoo et al., 2001; Ueno et al., 2002). Many barriers to gene delivery are universal to all oligonucleotide species: stability in circulation, cell uptake, and endosomal escape. DNA must additionally be delivered to the nuclear compartment to permit access to transcriptional machinery. It was shown over 30 years ago that DNA microinjected into the nucleus, but not the cytosol, produces a gene product (Capecchi, 1980). Improved delivery systems and alterations in DNA sequences can enhance nuclear delivery. For example, import into the nucleus can be aided by inclusion of a nuclear localization signal (NLS) in the plasmid DNA (pDNA) sequence (Brandén et al., 1999; Zanta et al., 1999). Also, strong viral or eukaryotic promoter sequences can be added to activate transcription (Capecchi, 1980). Plasmids can be further tailored by using tissue-specific promoters to reduce potential off-target effects (Gorski et al., 1986).

Important safety concerns for DNA delivery are immune stimulation and risk of insertional mutagenesis. The risk of insertional mutagenesis is much higher for viral delivery systems, but cannot be excluded for episomal pDNA (Glover et al., 2005; Baum et al., 2006). Like other oligonucleotides, DNA can stimulate immune responses that should be considered when designing therapeutics. Immune responses to DNA are caused by unmethylated CpG motifs which stimulate B-cell proliferation and cytokine release through TLR9 (Krieg et al., 1995; Klinman et al., 1996; Hemmi et al., 2000). The robust immunostimmulatory effect of CpG DNA makes it a potentially useful vaccine adjuvant (Gurunathan et al., 2000). Alternatively, the immune effects of CpG can also be removed by altering the DNA sequence to replace CpG with CpG-S motifs (Krieg et al., 1998). Thorough reviews of DNA delivery technology platforms have been published elsewhere (Nishikawa and Huang, 2001; Yin et al., 2014).

Delivery of DNA encoding a gene of interest is a powerful tool for gene therapy and important advances toward this goal have been made. While no therapies have been approved by the FDA thus far, several non-viral DNA strategies are being evaluated clinically, including in clinical trials for cancer therapy. In a phase 1 clinical trial, the tumor suppressor gene TUSC2 was delivered to lung cancer patients using DOTAP-cholesterol liposomes, resulting in transgene expression and activation of apoptotic pathways (Lu et al., 2012). This therapy is now in phase 1/2 trials. Other gene delivery trails include IL-12 gene delivery in PEG–PEI–cholesterol NPs to enhance immune response (Kendrick et al., 2008; Anwer et al., 2010) and co-delivery of two tumor suppressor genes somatostatin receptor subtype 2 (SSTR2) and deoxycytidine kinase::uridylmonophosphate kinase (DCK::UMK) complexed to PEI (Buscail et al., 2015). Delivery of the cytotoxic diphtheria toxin A gene as a "suicide" gene under cancer specific promoters in PEI NPs is also being evaluated in ovarian, pancreatic, and bladder cancer and has demonstrated good safety profiles and anti-tumor efficacy (Sidi et al., 2008; Smaldone and Davies, 2010; Hanna et al., 2012; Gofrit et al., 2014). If these non-viral DNA therapeutics show substantial efficacy in clinical trials, they will pave the way for non-viral DNA in cancer therapy and other diseases. After 3 decades of technology development, therapeutic DNA delivery in humans is becoming a reality.

#### mRNA

The goal of mRNA delivery is the same as DNA delivery, to deliver a therapeutic gene that will be translated into protein within target cells. In contrast to DNA, mRNA needs to reach the cytosol and be recognized by ribosomes. Transfection efficiency with mRNA is higher than for DNA, especially in non-dividing cells (Yamamoto et al., 2009). Also, mRNA delivery does not pose the risk of insertional mutagenesis. Although stability of naked mRNA is poor, chemical modifications and protection from serum endonucleases in NP delivery vehicles can increase mRNA stability. As with DNA, exogenous mRNA can also stimulate an immune response through TLR3, TLR7, TLR8, and retinoic acid receptor responder protein 3 (RARRES3 or RIG-I) (Yin et al., 2014). Chemical modifications can reduce recognition of mRNA by the immune system (Karikó et al., 2005).

The structure of mRNA is critical for recognition by the eukaryotic translation machinery. At the core of the mRNA structure is an open reading frame (ORF) that is translated into protein. Flanking the ORF are two untranslated regions (UTRs) at the 3<sup>0</sup> and 5<sup>0</sup> ends that allow for regulation of translation. Finally, a 5<sup>0</sup> methyl cap and a 3<sup>0</sup> poly adenosine tail book-end the mRNA and are necessary for efficient translation (Gallie, 1991). Increasing the poly(A) tail length also improves stability (Holtkamp et al., 2006). Commercial kits are available to synthesize mRNA with all necessary structural components from

plasmid DNA, though optimization for target cell types can also help improve translation efficiency (Yamamoto et al., 2009).

There are over a dozen clinical trials using mRNA for vaccines, adjuvants, or to express antigens either in dendritic cells ex vivo or by direct injection. However, gene replacement therapy is still in pre-clinical development (Kaczmarek et al., 2017).

#### MicroRNAs (miRNAs)

Discovered in 1993, miRNAs are a class of non-coding, regulatory RNAs that have critical roles in nearly all biological processes, including cancer. miRNAs can serve as both oncogenes and tumor suppressors (Farazi et al., 2013). Primary miRNA transcripts have characteristic hairpin structures that are recognized and processed by RNase III enzyme Drosha, which produces a stem loop precursor miRNA (pre-miRNA) of ∼70 nucleotides (Lee et al., 2003). Final cleavage by Dicer results in a mature dsRNA (Hutvágner et al., 2001). The mature ∼22 nucleotide miRNA associates with the RNA-induced silencing complex (RISC), and one or both strands of the duplex guide RISC to complementary sequences within target mRNA. Strand selection by RISC is likely based on stability. The two strands are named 5p and 3p corresponding to the 5<sup>0</sup> and 3<sup>0</sup> ends of the miRNA precursor hairpin, respectively. Target sequences complimentary to the 2–7 nucleotide "seed" region of the miRNA sequence are frequently found in the 3<sup>0</sup> UTR of mRNAs, but can also be within coding or intronic regions. Binding of miRNA to target mRNA results in degradation or destabilization of the mRNA and can also cause translational repression (Ha and Kim, 2014). Generally speaking, each miRNA can target hundreds of unique mRNAs, and thus can regulate transcriptome-wide changes. In this way, miRNAs are critical regulators of cell identity and state (Kosik, 2010). Additionally, miRNAs are essential for immune cell development and immune activation (Xiao and Rajewsky, 2009) as well as cross-talk between cancer cells and the TME (Chou et al., 2013).

To replace a downregulated miRNA, synthetic doublestranded RNAs (dsRNAs) carrying the same sequence as the endogenous mature or precursor miRNA can be used. These miRNA "mimics" are smaller and more stable than mRNA, allowing ease of encapsulation in several types of NPs. Chemical modifications to miRNAs can be made in the same way as mRNAs or siRNAs to increase stability and reduce inflammatory response. Nucleic acid based inhibitors of miRNAs include locked-nucleic acids (LNAs), antagomirs, anti-miRs, and miRsponges have been characterized in more detail elsewhere (Ling et al., 2013). Therapeutic strategies modulating miRNA function are already in clinical trials, and a comprehensive review of miRNA therapeutics is provided elsewhere (Rupaimoole and Slack, 2017). Inhibiting miR-122, a critical player for hepatitis infection is being investigated in multiple clinical trials, and

inhibition of miR103/107 is in clinical trials for alcoholic fatty liver disease. Both drugs utilize unencapsulated anti-miRs. While these therapeutics target the liver, a relatively accessible organ target for nucleic acid delivery, trials are also in progress for cancer therapy. Delivery of miR-16 with EGFR-targeted EnGeneIC Delivery Vehicle nanocells completed phase 1 clinical trials in mesothelioma with an acceptable safety profile and signs of efficacy (van Zandwijk et al., 2017). In contrast, phase 1 trials for miR-34 mimics for multiple solid tumors were recently terminated due to severe immune-related and marrow suppressive adverse events. A 110 nm liposomal carrier composed of ionizable lipids was used in these trials. The precise source of the inflammatory reactions—be it due to the carrier, miRNA mimic, or synergy between the two—is not known (Beg et al., 2017; Rupaimoole and Slack, 2017). It is clear that going forward, extensive pre-clinical evaluation of immune stimulation by miRNA-directed therapies must to be an important consideration.

#### siRNA

siRNAs are ∼21 nt dsRNAs that interact in the cytoplasm with the RISC complex to degrade target mRNAs. In structure, siRNAs are identical to miRNAs and can be modified and encapsulated in the same way. Other than the fact that miRNA mimics are based on endogenous genes while siRNAs are synthetically designed, the primary difference between miRNAs and siRNAs is the sequence specificity. Instead of a miRNA seed region of 6–8 nt with complementarity to the 3-UTR of target mRNAs, by design siRNAs usually have 100% complementarity to target mRNAs. As such, siRNAs have potent activity on a single target. In contrast, miRNAs have the potential to target hundreds of mRNAs, but generally this inhibition is to a lesser extent. The targets of siRNAs are degraded by the endonuclease activity of Argonaut 2 associated with the RISC, while miRNAs usually cause deadenylation or translational repression of target mRNAs by association with other Argonaut complexes devoid of nuclease activity. Therefore, the pharmacodynamics of miRNA mimics and siRNAs have important differences. However, the pharmacokinetics and biodistribution of these two oligonucleotides is the same, since their chemical structure is identical. Another consideration is stimulation of immune responses by exogenous RNA. For instance, siRNAs within liposomal or polymeric NPs can induce an inflammatory cytokine response that is greater than either component alone. The response is mediated by TLRs and is sequence dependent, with GU-rich sequences inducing the most potent cytokine response (Judge et al., 2005). Modifications of RNA, such as incorporation of 2<sup>0</sup> -O-methyl nucleosides, can reduce inflammatory response without decreasing gene silencing (Judge et al., 2006). Inflammatory cytokine responses have been observed in siRNA and miRNA clinical trials, and therefore evaluating immunostimulation should be a concern when translating these therapeutics to the clinic. Currently, dozens of clinical trials of siRNA-based therapies have been completed or are ongoing, for in depth review see Wittrup and Lieberman (2015) and Zuckerman and Davis (2015).

### Other Non-coding RNAs

Several classes of non-coding RNAs have been identified including piwi-interacting RNAs, endogenous siRNAs, long-noncoding RNAs, and circular RNAs. The function of these species and their role in disease biology are being actively investigated (Esteller, 2011). As the role of non-coding RNAs in cancer biology continues to unfold, therapeutic approaches to modulate them will be of increasing interest (Gutschner and Diederichs, 2012). Therapeutic delivery of non-coding RNAs will face similar challenges to other types of RNA, such as stability and uptake into appropriate cells and intracellular compartments.

### Genome Editing

The targeted manipulation of genomic DNA in living cells is possible through the use of engineered nucleases, such as mega nucleases, zinc finger nucleases (ZFN), transcription activatorlike effector-based nucleases (TALEN), and the clustered regularly interspaced short palindromic repeats (CRISPR)-Cas system (Cox et al., 2015). The enthusiasm surrounding genome editing technology is reminiscent of that generated by the discovery of RNAi (Fire et al., 1998). However, similar hurdles impede therapeutic translation. To enact genome editing in vivo nucleases must be present within target cells. While delivery of protein may be possible, most approaches utilize delivery of mRNA or DNA encoding nucleases. Using viral vectors, such as AAV, to deliver nucleases raises additional concerns. Integration of nuclease DNA leads to constitutive expression of nucleases capable of genome editing and increases the chances of off-target mutations. For this reason, transient expression – as is achieved with NP systems-may be preferable.

CRISPR-Cas9 is a highly specific gene editing tool that is rapidly becoming a standard lab technique. In its simplest embodiment, it requires the presence of a Cas9 protein and single guide RNA (sgRNA) to be present in the same cell (Liu C. et al., 2017). The first use of CRISPR-Cas9 in clinical trials has already taken place. In this trial the T-cell exhaustion receptor PD-1 was genetically removed ex vivo by CRISPR-Cas9 technology and cells were then delivered to the patient (Cyranoski, 2016). Similarly, CRISPR-Cas9 is being used to remove the CCR5 receptor from hematopoietic stem cells as a therapeutic strategy for HIV. Since HIV enters cells through CCR5, removing this receptor is expected to reduce the ability of HIV to infect transplanted cells (Li et al., 2015). Application of this technique to other cell-based therapies could soon follow. It will certainly be more challenging to edit cells in vivo but several groups are working toward this important goal, including developing strategies for cancer treatment (Sánchez-Rivera and Jacks, 2015).

### TARGETING CELLS IN THE TME

#### Mononuclear Cells Macrophages

Tumor-associated macrophages (TAMs) are often pro-tumoral by promoting angiogenesis, invasion, metastasis, and immune tolerance (Noy and Pollard, 2014). Given these and other important roles in tumor progression, there is a growing

interest in targeting TAMs as a cancer therapeutic. Macrophages are often characterized by the balance of pro- and antiinflammatory characteristics as "M1" or "M2" macrophages, respectively. While in most cases this terminology is an oversimplification, it is a useful reference to describe macrophage subsets. M1 macrophages are characterized by high levels of major histocompatibility complex (MHC) class II molecules, pro-inflammatory cytokines and inducible nitric oxide synthase 2 (Nos2). Conversely, M2 macrophages express low levels of these markers and instead express high levels of arginase-1 and scavenger receptors like the mannose receptor (MR) (Martinez et al., 2009). Initial in vitro studies suggested macrophages could kill tumor cells, however, in the TME secreted factors shift TAMs to a pro-tumoral M2 phenotype (Sica et al., 2006). Therefore, reprograming TAMs to an M1-like state could reverse the protumoral effects. We review here characteristics of NPs that promote macrophage uptake and detail studies utilizing NP gene delivery to target TAMs for cancer therapy.

Most NP platforms are in the size range of viruses (20–250 nm) and bacteria (0.2–10 µm), and consequently they are readily taken up by phagocytic cells of the immune system, such as macrophages and dendritic cells. In fact, uptake by the mononuclear phagocytic system (MPS), also called the reticuloendothelial system (RES), has been considered a critical obstacle to NP drug delivery. One example of this is the first FDA approved NP drug Doxil, a nanoliposome formulation of doxorubicin. Doxil was coated with a hydrophilic neutral polymer, PEG, to reduce its recognition by the MPS, creating a "stealth" effect (Working et al., 1994). Despite these modifications, clearance by macrophages still has a major influence on Doxil pharmacokinetics (La-Beck et al., 2012). The proclivity of macrophages for taking up NPs can even be utilized to image macrophages in vivo (Weissleder et al., 2014). Interestingly, M2 macrophages take up both 300 nm PEG hydrogel NPs and 30 nm quantum dots at higher rates than M1 macrophages due to increased levels of scavenger receptors such as MR, suggesting that TAMs may be especially sensitive to NP delivery. This phenomenon does not extend to microparticles, as 6 µm PEG hydrogels are not affected by macrophage polarization (Jones et al., 2013). Additionally, single cell pharmacokinetics of NPs within tumors has revealed that TAMs can serve as reservoirs for NPs, releasing small molecule drugs (Miller et al., 2015). The fate of oligonucleotide payloads in macrophages is less certain, although there is some evidence that macrophages may also transfer genes to cells at sites of inflammation (Haney et al., 2013; Mahajan et al., 2016). Because much effort has gone into reducing phagocyte clearance of NPs, there is a wealth of studies detailing the characteristics that reduce and consequently those that enhance uptake into this population. Detailed review of factors influencing macrophage NP uptake is covered elsewhere (Gustafson et al., 2015).

Macrophage NP uptake can occur through micropinocytosis, phagocytosis, and receptor-mediated endocytosis (Gustafson et al., 2015). Factors influencing uptake include charge, size, and surface chemistry. A positive surface charge facilitates uptake by many types of cells having a negative membrane potential. In contrast, greater net charge in either the positive or negative direction increases uptake of NPs by macrophages. The uptake of chitosan NPs with charges ranging from −40 to +35 was examined for murine macrophages. In these cells, increasing charge in both the negative and positive direction increased macrophage uptake, while non-phagocytic cells more efficiently took up positively charged particles. NP size is also a key determinant: large NPs 300–500 nm in size are taken up more efficiently than 150 nm particles by murine macrophages (He et al., 2010). In addition, uptake of particles by macrophages is highly dependent on serum protein adsorption. PEGylation can decrease, but not eliminate, protein adsorption and macrophage uptake (Xie et al., 2007; Walkey et al., 2012). These factors are important for non-targeted or passive uptake; however, further cell-type specificity can be achieved with targeting moieties.

Ligands or antibodies to cell surface receptors can be used to decorate the NP surface and enhance macrophage uptake. Receptors that mediate macrophage NP uptake include folate receptor (FR), MR, cluster of differentiation 163 (CD163), Legumanin, galactose-type C-type lectins, and cluster of differentiation 11b (CD11b) (Binnemars-Postma et al., 2017). Mannose is one of the most common macrophage targeting ligands, but MR (also known as CD206) is also present on other phagocytes, such as DCs (McKenzie et al., 2007), which could result in off-target effects. However, MR is upregulated in M2-like TAMs with decreased MHC II expression (Movahedi et al., 2010). Additionally, CCR2 knockout mice that have fewer TAMs have reduced tumor uptake of MR-targeted nanobodies, indicating that MR binding is through CCR2 derived cells, including macrophages (Movahedi et al., 2012). Mannosylated polymeric micelles are able to deliver siRNA and mediate TAM gene silencing in vitro and in vivo (Shann et al., 2013; Ortega et al., 2015). The galactose-type C lectin receptor has also been targeted for macrophage nucleic acid delivery by attaching its ligand, galactose, to the surface of NPs (Huang et al., 2012; Liu L. et al., 2017). One sophisticated approach utilized a dual targeting moiety: an apolipoprotein A1 mimetic (α-peptide) served as a ligand for SR-1B and was linked to a M2 macrophage binding protein (M2pep) to deliver NP-encapsulated siRNAs (Qian et al., 2017). Macrophages also express receptors capable of recognizing a variety of pathogen-associated molecular patterns (PAMPs) and incorporation of PAMPs into NP design can facilitate NP uptake by TAMs. For example, nanocomplexes incorporating glucan, a PAMP associated with fungi, were shown to target TAMs (Zhang et al., 2015a,b). In summary, TAMs act as natural sinks for NPs. Further targeting with receptor specific ligands or antibodies can facilitate uptake, but complete discrimination between macrophages and other mononuclear cells has not been clearly demonstrated.

A small number of studies have shown efficacy of nanoparticle nucleic acid delivery to target and reprogram TAMs for cancer therapy. In a melanoma mouse model, delivering anti-CSF-1R siRNA targeted to TAMs reduced tumor growth by 87% and prolonged survival. Non-targeted particles also inhibited tumor growth, but not as dramatically. This therapeutic effect corresponded with decreased immunosuppressive cytokines IL-10 and TGF-β, and increased immunostimmulatory cytokines IL-12 and IFN-γ as well as increased the function of CD8+

T-cells (Qian et al., 2017). Similarly, pro-inflammatory miR-155 was delivered in redox and pH sensitive NPs targeted with galactose moieties to the macrophage galactose-specific C-type lectin receptor. Galactose targeting increased in vitro miR-155 uptake in TAMs, but not B16-F10 cells. Delivery of miR-155 NPs increased IL-12 and MHCII positive cells, as well as decreased M2 markers. Increased numbers of activated T-cells and NK cells were observed, and anti-tumoral effects were elicited (Liu L. et al., 2017). Intratumoral injection of modified glucan nanocomplexes carrying siRNA has also been shown to effectively inhibit gene expression in macrophages. Delivery of siRNA to migration inhibitory factor (MIF) in glucan NPs reduced both released and intracellular MIF in TAMs and in cancer cells. This resulted in reduced M2 markers and increased inflammatory cytokines TNF-α and IL-2. This treatment also increased CD4+ and CD8+ cells in the tumor and promoted anti-tumor immunity (Zhang et al., 2015a,b). These reports support that oligonucleotide delivery can be used to reprogram TAMs from an M1 to M2 phenotype to promote anti-tumor effects.

#### Inflammatory Monocytes

Inflammatory monocytes (IMs) can give rise to TAMs and other myeloid suppressor cells which promote angiogenesis and subsequent metastasis. Recruitment of IMs relies on the chemokine CCL2 (Qian et al., 2011). Blocking this axis with receptors against CCL2 or its cognate receptor CCR2 has been the subject of clinical trials, but pharmacological inhibition of this axis has proved challenging in part due to rebound effects (Lim et al., 2016). An alternative strategy used screening approaches to identify both optimal lipids and siRNA sequences for inhibition of CCR2 in monocytes, no targeting ligands were used. In a lymphoma model, inhibition of CCR2 in monocytes reduced tumor size and number of TAMs. This therapy also inhibited expression of VEGF and reduced microvessel density in the tumors (Leuschner et al., 2011). Whether NP-based targeting of the CCL2-CCR2 axis can evade withdrawal effects seen with antibody targeting of the CCL2-CCR2 axis remains to be seen. However, the ability to silence genes in monocytes has demonstrated clear therapeutic potential.

#### Dendritic Cells

As part of the innate immune response to pathogens, dendritic cells recognize foreign materials through pattern recognition receptors (PRRs) or compliment binding leading to phagocytosis. Inside dendritic cell lysosomes, processing of pathogenic proteins results in the generation of peptide fragments that are presented on MHC receptors to be recognized by members of the adaptive immune system. Cross presentation of antigens from DCs to CD8<sup>+</sup> T-cells is required for anti-tumor immunity. As such, DCs are the primary targets of cancer vaccines and their actions are required for effective cytotoxic T-cell response in checkpoint blockade inhibitor therapies. NPs from 20 nm to 3 µm are readily taken up by dendritic cells, presumably due to the size similarity to viral and bacterial pathogens. The size of NPs and their ability to present multivalent antigens clearly points to vaccine applications (Bachmann and Jennings, 2010). For these reasons an increasing number of NP-based vaccines with or without additional immune agonists are being designed for cancer therapy (Mizrahy et al., 2017). Here we will consider those that additionally incorporate nucleic acid delivery. A smaller number of studies have utilized gene delivery to modulate the activation of dendritic cells and subsequent cross-presentation to T-cells.

Surface coating can also increase DC NP uptake. For example, natural coatings such as mannosylation or glycosylation increase DC uptake through interactions with MR (Jiang et al., 2015; De Coen et al., 2016; Wang et al., 2016). However, as previously described, MRs are also expressed on macrophages which may compete with DCs for NP uptake (Stahl and Ezekowitz, 1998). Alternatively, antibody-based targeting has also been reported. One example is a clinical trial using targeting antibodies against MR to deliver a peptide antigen to APCs. This therapy induced humoral and T-cell responses in melanoma patients (Morse et al., 2011). Reports indicate that antibodies against DEC-205 can also enhance DC uptake and increase downstream immune activation relative to non-targeted NPs (Raghuwanshi et al., 2012; Walters et al., 2015). Using antibodies against DEC-205 fused with a tumor antigen induced humoral and cellular immunity in patients with advanced malignancies (Dhodapkar et al., 2014). While MR and DEC-205 are commonly used for NP or vaccine targeting, other targets have also been tested (Sehgal et al., 2014a), including targets that enhance uptake in subsets of DCs (Sehgal et al., 2014b). However, in some cases the material composition of the NP carrier may be more important than targeting ligands. For instance, lipid based "nanogels" were more readily internalized than PLGA NPs (Look et al., 2014). Also, linear PEI nanocomplexes were more effective than anti-CD11c antibody-targeted zwitterionic liposomes at siRNA delivery to DCs (Cubillos-Ruiz et al., 2009).

Several reports both in vitro and in vivo have shown that delivering siRNAs in addition to antigenic peptides and adjuvants to DCs can further enhance anti-tumor immune responses. Heo et al. (2014) used polymeric micelles to deliver a tumor antigen and siRNA for the immunosuppressive Suppressor of Cytokine Signaling 1 (SOCS1) to dendritic cells in vitro. Delivery of SOCS1 siRNA increased secretion of pro-inflammatory cytokines by cultured DCs and activation of T-cells by cross presentation (Heo et al., 2014). Two additional studies by Heo et al. (2014) examined in vivo efficacy of NP siRNA to DCs. In one study, the investigators formulated multifunctional polymeric NPs carrying tumor model antigen OVA, dendritic cell activator imiquimod (R837), and siRNAs for STAT3. The immune activation induced by R837 is inhibited by STAT3, so the authors hypothesized that this combination would produce a more robust activation of DCs. PLGA (R837/STAT3 siRNA) NPs were taken up efficiently by DCs, elicited cytokine response, antigen crosspresentation, and trafficking of DCs to draining lymph nodes when injected in vivo. Furthermore, incorporation of STAT3 siRNA significantly increased anti-tumor immunity (Heo and Lim, 2014). In a different approach, tumor bearing mice were first treated with hyaluronic acid (HA) and paclitaxel (PTX) complexes to induce immunogenic cell death. This treatment was followed by administration of NPs containing CpG adjuvant and IL-10 siRNAs. IL-10 is an immunosuppressive cytokine and

its inhibition further enhanced the immune response. These multifunctional NPs trafficked to draining LNs and promoted antitumor immunity in vivo (Heo et al., 2015).

In ovarian cancer tumor-associated DCs (tDCs) have a particularly tolerogenic role (Huarte et al., 2008). By inhibiting tolerogenic pathways in DCs with siRNA, therapeutic benefits were observed in ovarian cancer models. In addition to siRNA, miRNA can also be delivered to tDCs to induce anti-tumor immunity. For example, delivery of siRNAs against members of the ER stress pathway, XBP1 and IRE1, which inhibit cross presentation in tDC through lipid accumulation. NP delivery of XBP1 or IRE1 siRNA reduced metastasis and increased survival in ovarian cancer models. Importantly, this phenomenon was ablated in Rag2 deficient mice suggesting that immune and not direct cancer targets were responsible (Cubillos-Ruiz et al., 2015). miR-155 is considered to be an oncogenic miRNA, however, it is also necessary for cross-presentation of DCs. Using PEI NPs, delivery of miR-155 mimics produced potent anti-tumor effects with about 33% of mice showing no disease progression 80 days after controls had succumbed to disease. This anti-tumor effect was accompanied by transcriptome wide changes in tDCs, highlighting the utility of miRNAs for reprograming the TME (Cubillos-Ruiz et al., 2012). Interestingly, the authors also found that a bulged dsRNA that required processing by RNase enzyme DICER was most effective at gene silencing. Another aspect of the studies by Cubillos-Ruiz et al. (2009, 2012, 2015) was an immunestimulatory effect of PEI NPs containing even non-targeting RNA through TLR pathways. Overall, RNA delivery to tDCs has been shown to be an effective therapeutic strategy in mouse models of ovarian cancer.

Historically, vaccines have relied on peptide antigens, but an alternative vaccine strategy is delivery of DNA or mRNAs. In brief, DNA or mRNA encoding an antigen are injected, the genetic material is taken up by cells at the injection site, and then translated into protein. Proteins encoded in the DNA or mRNA can be expressed in myocytes or keratinocytes at the injection site and are subsequently recognized by APCs or directly taken up by DCs followed by internal processing and presentation. DNA vaccines are currently used in veterinary medicine, but have thus far not been successfully translated to humans (Rice et al., 2008). Attempts at DNA vaccines in humans have relied on non-specific targeting of injected DNA. Increased gene delivery through electroporation and NP delivery systems has been reported, but generally do not utilize cell-specific targeting. In one report, plasmid DNA for the nucleocapsid of severe acute respiratory syndrome coronavirus (SARS-CoV) was delivered in chitosan NPs targeted with anti-DEC-205 antibody. NP DNA delivery successfully stimulated IgG and IgA antibodies against SARS-CoV nucleocapsid, in contrast to naked DNA, which produced no detectable antibody response. In addition, DC targeting with anti-DEC-205 antibody significantly increased serum IgG against SARS-CoV nucleocapsid (Raghuwanshi et al., 2012). This approach could be translated to cancer immunotherapy as well, but consideration of particle size may be critical to induce a cytotoxic T-cell (CTL) response, given that CD8<sup>+</sup> DCs are necessary to induce a CTL response and are restricted to the lymph node. Therefore, particles must drain to the lymph node, which requires a particle size of 20–200, with 40 nm being ideal (Bachmann and Jennings, 2010).

Overall, DCs are an exciting TME target for NP nucleic acid delivery; they are intimately involved in the anti-tumor response and are required for the actions of checkpoint blockade therapies. Serendipitously, their phagocytic abilities and PRRs also make them easy targets for NP delivery. These qualities have generated increased interest in NP-based vaccines, which will likely lead to several clinical trials. As multifunctional NPs are designed to deliver antigens and adjuvants to DCs, gene delivery strategies should also be considered.

#### Cancer-Associated Fibroblasts

Within the tumor stroma, cancer-associated fibroblasts (CAFs) modulate tumor growth and metastasis by secreting growth factors, chemokines, and extracellular components (Kalluri and Zeisberg, 2006). In many tumors, especially desmoplastic tumors with a dense stroma, CAFs often lie between blood vessels and cancer cells. This makes CAFs an impediment to cancer-directed NP delivery (Miao et al., 2016). Cisplatin NPs with or without targeting are largely taken up by CAFs in desmoplastic pancreatic tumors (Miao et al., 2017b). Damage to fibroblasts initially reduces their supportive role and promotes tumor regression. However, chronic exposure induces expression and release of soluble factors such as Wnt16 and resistance to chemotherapy. Co-delivering Wnt16 siRNA along with cisplatin NPs can prevent resistance through this pathway (Miao et al., 2015). Since this finding, several studies have now shown that plasmid DNA can be delivered to and expressed in CAFs using lipid-based NPs. In one study, delivery of a gene that produced a soluble TNFα-related apoptosis inducing ligand (sTRAIL) to CAFs caused apoptosis in the tumor parenchyma, and ultimately tumor regression (Miao et al., 2017b). Similarly, several studies have shown that delivery of pDNA encoding "traps" can be successfully delivered to CAFs in vivo for cancer therapy. Traps are fusion proteins designed to be secreted and ultimately bound to soluble factors in the TME, such as chemokines and cytokines. By inhibiting these factors, metastasis and immunosuppression have been shown to be reduced, ultimately improving survival in animal models. In one report, a CXCL12 trap in combination with a PD-L1 trap promoted T-cell infiltration and reduced liver metastasis of pancreatic cancer more than either therapy alone (Miao et al., 2017a). Combination of CXCL12 and PD-L1 traps also decreased immune suppressive lymphoid structures and enhanced anticancer vaccine efficacy (Goodwin et al., 2017). These studies together suggest that CAFs can be used as cellular factories for production of proteins that inhibit the immunosuppressive TME. This work demonstrates the possibility that replacement or inhibition of endogenous genes in CAFs may be a feasible therapeutic strategy.

#### T-cells

T-cells are important mediators of anti-tumor immunity and the targets of immune-oncology drugs such as checkpoint blockade inhibitors. The receptors PD-1 and CTLA4 expressed on T-cells promote exhaustion and thus immune evasion by cancer cells. Antibodies blocking these receptors and their

ligands have proven to be effective stimulators of anti-tumor immunity and have quickly become a staple of anti-cancer therapy (Hoos, 2016). An alternative T-cell based approach to promote immune recognition of cancer cells is autologous, genetically engineered T-cells. These cells are engineered to express CARs specific to cancer epitopes using viral transduction. CAR-T cells have recently been approved by the FDA for acute lymphoblastic leukemia (ALL) and clinical trials are ongoing in many other cancer types (Landoni and Savoldo, 2017). Whether by removing checkpoint blockade or through genetic modification, T-cells have been demonstrated to be a successful target for cancer immunotherapy. New approaches to further harness the power of T-cells are being developed in many areas, including nanomedicine.

#### CAR-T-cells

Expression of CARs in primary T-cells relies on viral transduction and integration of DNA into the genome in vitro. Manufacturing genetically engineered cells for autologous transplantation is an intensive process with relatively low yields. T-cells are resistant to many forms of gene delivery and standard transfection protocols are not effective. Current gene delivery methods to T-cells rely on viruses or electroporation (Freeley and Long, 2013). Viral methods can be mutagenic and electroporation of cell membranes can lead to irreversible cell damage and low yields. The incorporation of efficient and transient gene expression with NP platforms to produce engineered T-cells holds promise for improved immunotherapies. Photoporation based on NPs is one such strategy. In this approach, transient permeabilization is achieved by adding gold NPs to CD8+ T-cells followed by short laser pulses, creating a photothermal effect. This strategy had lower cytotoxicity than nucleofection with comparable siRNA-mediated gene knockdown (Wayteck et al., 2017).

One potential improvement to CAR T-cell therapy is increasing the specificity of T-cells by means of removing noncancer specific TCRs. Toward this aim, Moffett et al. report NP delivery of mRNAs to T-cells using anti-CD3 and anti-CD8 antibody targeting. Delivery of megaTAL nuclease mediated elimination of the T-cell receptor alpha constant region (TRAC), effectively removed the ability of T-cells to produce their own TCRs and resulted in the specific expression of the CAR (Moffett et al., 2017). This approach may foreseeably reduce off-target immune responses, but was not tested in vivo. In another approach, Moffett et al. (2017) increased the proportion of central memory T-cells, a critical cell population in establishing an effective immune response. Enrichment of the central memory T-cell phenotype was achieved by introducing Foxo13A encoding mRNA into a CD3 targeted NP platform. Treatment of T-cells with these Foxo13A-encoding NPs increased the activity of CARmodified T-cells in a mouse model of B-cell lymphoma (Moffett et al., 2017).

While these reports are intriguing, given the prevalence of viral methods in autologous T-cell therapy, it is questionable whether NP-based T-cell gene delivery will be clinically translatable. Despite greater than 200 clinical trials for CAR-T cells, none currently use NP-based methods.

#### T-cells in Vivo

Nanoparticle-mediated nucleic acid delivery to T-cells in vivo has also been demonstrated. These NP systems rely on antibodies to surface proteins expressed on T-cells. In one instance, b7 integrin targeting antibody was used to deliver lipid-based NPs containing siRNAs to leukocytes. Systemic delivery of only 2.5 mg/kg mediated gene knockdown (Peer et al., 2008). Delivery of siRNAs to CCR5, a critical receptor for HIV entry, with lymphocyte function-associated antigen 1 (LFA-1) targeted particles decreased susceptibility of humanized mice to HIV infection (Kim et al., 2010). In these two studies, the subsets of leukocytes targeted were not described and are likely heterogeneous, considering the targeted receptors are present on many leukocytes. Alternatively, using anti-CD4 antibody decorated lipid NPs can specifically deliver siRNA to T-cells in vivo. Ramishetti et al. (2015) found that internalization, not endosomal escape, may be the limiting factor for T-cell gene delivery. Intriguingly, CD4 subsets with high or low CD4 expression had different rates of internalization and subsequent gene silencing (Ramishetti et al., 2015). Further research into the T-cell internalization pathways and characterization of internalization after binding to other T-cell specific receptors is warranted. Collectively, in vivo delivery of oligonucleotides to T-cells with NPs is achievable, but the potential therapeutic benefit for cancer is yet to be determined.

#### Blood Vessels

Angiogenesis refers to the growth of new blood vessels from pre-existing vascular networks. Healthy vasculature is quiescent due to a controlled balance between pro- (e.g., VEGF and FGF) and anti- (e.g., angiostatin and thrombospondin) angiogenic factors that regulate endothelial cell proliferation and migration (Jain, 2003). As tumors outgrow their local oxygen supply, they hijack this regulation and permanently shift the balance to a pathologic, pro-angiogenic state during the "angiogenic switch" (Folkman, 1971; Hanahan and Folkman, 1996). This produces chaotic and dysfunctional vasculature. While normal blood vessels consist of a continuous monolayer of tightly adhered ECs, closely associated mural cells that promote vessel stability, and a continuous basement membrane; tumor vessels have loosely associated ECs with large gaps between them, poor mural cell recruitment, and an irregular and discontinuous basement membrane (Baluk et al., 2005). This reduced vessel wall integrity promotes leakiness and cancer cell intravasation. Thus, directly targeting tumor vessels to either inhibit their growth or promote their normalization is believed to have the potential to inhibit tumor growth and aggression, as well as metastasis (Folkman, 1971; Carmeliet and Jain, 2011). Interestingly, it is the "leaky" nature of tumor blood vessels that both makes it challenging to deliver drugs such as chemotherapy to the tumor core, but also greatly facilitates delivery of NPs to cancer cells due to the "enhanced permeability and retention effect" (Prabhakar et al., 2013).

Oligonucleotide delivery to tumor endothelium has been achieved with multiple NP platforms. Generally, successful delivery of NPs to vasculature is confirmed by visualizing colocalization of fluorescently labeled nucleic acids packaged in

NPs with an endothelial stain such as the cell surface marker CD31. Chitosan NPs have been demonstrated to co-localize to both tumor and endothelial cells in vivo and effectively deliver siRNAs to both cell types (Lu et al., 2010). In an orthotopic model of ovarian carcinoma, treatment with chitosan NPs carrying siRNAs targeting human EZH2 (expressed in the transplanted cancer cells) or murine EZH2 (expressed in the endogenous murine vasculature) inhibited tumor growth. However, the NPs carrying murine targeting siRNA had more potent effects on inhibiting disease burden, suggesting chitosanmediated targeting of tumor vasculature had more potent therapeutic effects than targeting cancer cells directly (Lu et al., 2010). Second-generation NPs rely on incorporation of ligands to target endothelial cell-specific surface proteins. For example, ligands to integrin αVβ3, such as the peptide RGD, can be used to facilitate NP uptake into neo-vasculature. Studies have shown that NPs containing the chemotherapeutic drug doxorubicin can be directed specifically to tumor vasculature using this ligand, causing loss of tumor blood vessels and decreased metastasis (Murphy et al., 2008). Similarly, delivery of an anti-miR to inhibit the pro-angiogenic miR-132 with these same NPs in an orthotopic xenograft mouse model of human breast cancer yielded therapeutic effects on inhibiting tumor vasculature and decreasing tumor burden (Anand et al., 2010). miRNAs have also been delivered using RGD-labeled chitosan NPs. Delivery of miR-200 family members using this approach reduced angiogenesis by direct and indirect mechanisms and resulted in reduced disease burden in ovarian cancer models (Pecot et al., 2013). RGD-chitosan mediated delivery of siRNA targeting PLXDC1, a growth-promoting gene, has been shown to effectively silence target gene expression in endothelial cells, with subsequent effects on promoting endothelial apoptosis and inhibiting tumor growth (Hee-Dong et al., 2010). The αVβ<sup>3</sup> integrin also facilitates uptake of viral genomic material and therefore may be an effective route for NP based gene delivery (Stewart and Nemerow, 2007). In one report, delivery of mutant Raf-1 gene with αVβ3 targeted cationic lipid NPs caused apoptosis of vessels and surrounding tumor tissues (Hood et al., 2002). Another receptor that can mediate uptake into the vascular endothelium is CD31, a classical marker of blood vessels. While αVβ3 is thought to be expressed specifically by tumor neovasculature (as well as some cancer cell types), CD31 is expressed on all endothelium (both blood, and to a lesser extent, lymphatic). Using CD31 ligands to deliver siRNAs resulted in specific decrease of target genes in vascular endothelium. By delivering siRNA to CD31 itself, tumor growth and metastasis were inhibited in a prostate cancer model (Santel et al., 2006). An alternative approach to ligand-based targeting is chemically modified dendrimers that can specifically target the endothelium (Khan et al., 2015). 7C1

#### REFERENCES

Anand, S., Majeti, B. K., Acevedo, L. M., Murphy, E. A., Mukthavaram, R., Scheppke, L., et al. (2010). MicroRNA-132-mediated loss of p120RasGAP activates the endothelium to facilitate pathological angiogenesis. Nat. Med. 16, 909–914. doi: 10.1038/nm.2186

NPs are another type of NP that have been reported to localize faithfully and specifically to the endothelium in multiple models of aberrant vascular function, including tumor angiogenesis. These NPs are able to elicit at least 50% knockdown of target endothelial gene expression, and simultaneously deliver siRNAs targeting multiple genes in the endothelium (Dahlman et al., 2014).

### SUMMARY

In an era where clinical trials in nucleic acid delivery have become a reality, we can expand our scope to consider new and exciting gene and cell targets for cancer therapy. NP uptake by cells within the TME has traditionally been considered a delivery obstacle for NP-based systems, however, turning TME cells into targets could lead to new therapeutic strategies. Biology has taught us that non-transformed cells can act as accessories to cancer growth and spread, but that strategies to reprogram cells in the TME could result in revolutionary therapies. The studies highlighted in this review demonstrate NP-based nucleic acid delivery strategies for reprograming the TME. In effect, turning the TME from a permissive space for cancer growth to a hostile one. This strategy is synergistic with current immunotherapy and antiangiogenic approaches and could feasibly extend the efficacy of these paradigm-shifting treatments.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

CP was supported in part by an NCI grant (R01-CA215075-01), a Mentored Research Scholar Grants in Applied and Clinical Research (MRSG-14-222-01-RMC) from the American Cancer Society, the Jimmy V Foundation Scholar Award, the UCRF Innovation Awards, the Stuart Scott V Foundation/Lung Cancer Initiative Awards for Clinical Research, the University Cancer Research Fund, the Lung Cancer Research Foundation, the Free to Breathe Metastasis Research Award, and the Susan G. Komen Career Catalyst Award. SA was supported in part by a grant from the National Institute of General Medical Sciences under award 5T32 GM007092. EH was supported in part by a grant from the National Cancer Institute of the National Institutes of Health under award number T32CA196589.

Anwer, K., Barnes, M., Fewell, J., Lewis, D., and Alvarez, R. (2010). Phase-I clinical trial of IL-12 plasmid/lipopolymer complexes for the treatment of recurrent ovarian cancer. Gene Ther. 17, 360–369. doi: 10.1038/gt.2009.159

Bachmann, M. F., and Jennings, G. T. (2010). Vaccine delivery: a matter of size, geometry, kinetics and molecular patterns. Nat. Rev. Immunol. 10, 787–796. doi: 10.1038/nri2868



interleukin 6, interleukin 12, and interferon gamma. Proc. Natl. Acad. Sci. U.S.A. 93, 2879–2883. doi: 10.1073/pnas.93.7.2879



with recurrent malignant pleural mesothelioma: a first-in-man, phase 1, openlabel, dose-escalation study. Lancet Oncol. 18, 1386–1396. doi: 10.1016/S1470- 2045(17)30621-6


**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 Harrison, Azam and Pecot. 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.

# The Potential of Nanotechnology in Medically Assisted Reproduction

Mariana H. Remião<sup>1</sup> , Natalia V. Segatto<sup>1</sup> , Adriana Pohlmann<sup>2</sup> , Silvia S. Guterres<sup>2</sup> , Fabiana K. Seixas<sup>1</sup> and Tiago Collares<sup>1</sup> \*

<sup>1</sup> Biotechnology Graduate Program, Molecular and Cellular Oncology Research Group, Laboratory of Cancer Biotechnology, Technology Development Center, Federal University of Pelotas, Pelotas, Brazil, <sup>2</sup> Post-graduation Program in Pharmaceutical Sciences, Federal University of Rio Grande do Sul, Porto Alegre, Brazil

Reproductive medicine is a field of science which searches for new alternatives not only to help couples achieve pregnancy and preserve fertility, but also to diagnose and treat diseases which can impair the normal operation of the reproductive tract. Assisted reproductive technology (ART) is a set of methodologies applied to cases related to infertility. Despite being highly practiced worldwide, ART presents some challenges, which still require special attention. Nanotechnology, as a tool for reproductive medicine, has been considered to help overcome some of those impairments. Over recent years, nanotechnology approaches applied to reproductive medicine have provided strategies to improve diagnosis and increase specificity and sensitivity. For in vitro embryo production, studies in non-human models have been used to deliver molecules to gametes and embryos. The exploration of nanotechnology for ART would bring great advances. In this way, experiments in non-human models to test the development and safety of new protocols using nanomaterials are very important for informing potential future employment in humans. This paper presents recent developments in nanotechnology regarding impairments still faced by ART: ovary stimulation, multiple pregnancy, and genetic disorders. New perspectives for further use of nanotechnology in reproductive medicine studies are also discussed.

Keywords: assisted reproductive technologies, nanotechnology, nanobiotechnology, multiple pregnancy, in vitro maturation, gene therapy, embryology

### REPRODUCTIVE MEDICINE AND NANOTECHNOLOGY

Infertility and subfertility defined as the difficulty to conceive are conditions affecting people worldwide. The World Health Organization considers infertile couples those who fail to achieve a clinical pregnancy after, at least, 1 year of regular unprotected sexual intercourse (Zegers-Hochschild et al., 2017). Couples who experience these difficulties can turn to reproductive medicine technologies to help solve the problem. One of the most revolutionary treatments in this area is assisted reproductive technology (ART) comprising of in vitro embryo production (IVEP).

Regarding reproductive medicine, nanotechnology can be very useful in the development of non-invasive detection, diagnosis, and minimally invasive treatment of infertility-related disorders (oncological or non-oncological) (Barkalina et al., 2014a). To improve diagnostics, nanotechnology is applied mainly to the development and improvement of nanobiosensors and imaging techniques. Nanobiosensors are devices capable of identifying antigens, proteins, nucleic acids, and reactive oxygen and nitrogen species with quickness and sensitivity (Shi et al., 2007; Zhu et al., 2015).

#### Edited by:

Susan Hua, University of Newcastle, Australia

#### Reviewed by:

Domenico Criscuolo, Genovax S.r.l., Italy Christine Wyns, Université catholique de Louvain, Belgium

\*Correspondence:

Tiago Collares tiago.collares@ufpel.edu.br; collares.t@gmail.com

#### Specialty section:

This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology

Received: 03 November 2017 Accepted: 26 December 2017 Published: 11 January 2018

#### Citation:

Remião MH, Segatto NV, Pohlmann A, Guterres SS, Seixas FK and Collares T (2018) The Potential of Nanotechnology in Medically Assisted Reproduction. Front. Pharmacol. 8:994. doi: 10.3389/fphar.2017.00994

These technologies are underlying the development of interesting 'lab-on-a-chip' tools. Besides advantages of nanotechnology, this tool requires small volumes of analyte and reagents (Craighead, 2006; Hill and Li, 2017). The functionalization of zinc oxide nanoroads – gold nanoparticles (Gasparotto et al., 2017), iron oxide nanoparticles (Pal et al., 2015), and silicacoated gold nanoparticles with cadmium selenide quantum dots (Johari-Ahar et al., 2015) with anti-CA125 antibodies represent successful strategies to develop higher sensitivity tools for ovarian cancer detection. In addition, anti-HE4 antibody attached to silver nanoparticles was also used to develop a fast, specific, and stable ovarian cancer detection system (Yuan et al., 2012).

Biosensors using anti-PSA antibodies to detect PSA antigen represent one of the most used strategies for detection of prostate cancer. Gold nanoparticles functionalized with anti-PSA antibodies have been used in a bio-barcode assay showing ultrasensitivity (Thaxton et al., 2009) and in silicon nanowire field-effect transistors providing real-time prostate cancer detection (Presnova et al., 2017). Gold nanoparticles and anti-PSA antibodies supported in graphene oxide (Pal and Khan, 2017) or bound in cuprous oxide@ceric dioxide core–shell nanocomposites (Li et al., 2017) were also used to develop novel, accurate, and sensitive electrochemical immunosensors.

Diagnostic imaging has been improved by metallic and nanostructured particles, as these nanomaterials have great benefits compared to contrast agents. Iron oxide is one of the main contrast agents used for magnetic resonance imaging (MRI), and when nanostructured, it can be functionalized for additional benefits. Iron oxide nanoparticles can be formed in poly(vinyl alcohol), rendering them degradable over time (Bannerman et al., 2017) or, as showed in tumor xenograft animal models, can be associated to diatoms to improve tumor retention when a magnetic field is applied (Todd et al., 2014). In addition, iron oxide nanoparticles can be directed to a tumor site. For example, the functionalization of superparamagnetic iron oxide nanoparticles (SPIONs) with anti-prostate-specific membrane antigen (PSMA) increased the detection limit and the sensitivity of MRI in prostate tumor cell culture (Sillerud, 2016). In addition to iron oxide, other nanomaterials already tested in cell culture and/or animal models can be used as contrast agents to enhance imaging diagnostics, including gold nanoparticles (Indrasekara et al., 2013; Cole et al., 2015), carbon nanotubes (Liu et al., 2007; Vittorio et al., 2011), liposomes (Martina et al., 2005; Mukundan et al., 2006), dendrimers (Miyake et al., 2015), and quantum dots (Guo et al., 2014; Yao et al., 2016).

In the treatment of oncological diseases of the reproductive system, recent drug delivery and cell-target strategies have been developed. For example, one of the main anticancer drugs used, doxorubicin, has been associated to nanoformulations to increase its efficacy. These include mesoporous silica nanoparticles (Guo et al., 2017) and lipid-coated mesoporous iron oxide– based magnetic nanoassemblies (Pradhan et al., 2016) tested in human cell culture and xenograft mouse models, respectively. PEGylated liposomes have also been tested for cervical and ovarian cancer using human cells (Sriraman et al., 2016). Magnetic nanoparticles (Hua et al., 2017) have been used to treat cervical cancer in human cell cultures and xenograft mice. Other strategies include the delivering of siRNA in cationic dendritic starch (Engelberth et al., 2017), layer-bylayer engineering of upconversion nanoparticles (Lin et al., 2017), and mesoporous silica nanoparticles (Roberts et al., 2017) resulting in improved cell death in human ovarian cancer cells.

For non-oncological diseases of the reproductive system, some alternatives were tested in human cell culture. To treat uterine leiomyoma, strategies included the use of magnetic nanoparticles complexed to adenovirus (Shalaby et al., 2016) and nanoparticles loaded with 2-methoxyestradiol (Ali et al., 2013). In animal models, carbosilane dendrimer (Chonco et al., 2012) and nanoparticles-in-film (Cunha-Reis et al., 2016) were tested for the treatment of HIV infections. Another condition that could impair fertility is endometriosis, and the strategies already generated using nanomaterials are listed in **Table 1**.

In the field of fertility preservation, nanotechnology was shown to improve the potential of cryopreserved human immature testicular tissue to restore fertility. Dextran–chitosan nanoparticles loaded with vascular endothelial growth factor (VEGF) were tested for tissue engraftment after cryopreservation of the tissue in mice, resulting in higher vascular density and spermatogonia recovery in transplanted tissues (Poels et al., 2016). For female gametes, an interesting strategy for swine oocyte cryopreservation was developed. The addition of low concentrations of hydroxy apatite nanoparticles (less than 0.5%) in cryoprotectant agents increased the developmental rate of vitrified/devitrified germinal vesicles oocytes (Li et al., 2016). These are a few of the different contributions that nanotechnology has been giving to medically assisted reproduction.

### POTENTIAL CONTRIBUTIONS OF NANOTECHNOLOGY TO ASSISTED REPRODUCTIVE TECHNOLOGY

Although ART is successfully applied as a clinical treatment worldwide, some challenges remain. Because of this, strategies developed in animal models are highly important for identifying new alternatives to overcome these problems. When it comes to ART, embryo development in mammalian models is highly similar to humans (Niemann and Wrenzycki, 2000; Barkalina et al., 2016). Lagomorph, murine, swine, bovine, and non-human primates are the main species used to study IVEP techniques to be applied to humans.

Similarly, the implementation of nanotechnology, which has already been developed for non-human animals, could be applied to assisted reproduction in humans (Langbeen et al., 2015; Barkalina et al., 2016). As mentioned previously, this technology has already been tested and used in sectors adjacent to reproductive medicine. Therefore, the main


challenges of ART nowadays are how nanotechnology can intervene in order to boost the techniques already used today.

#### OVARIAN STIMULATION AND IN VITRO MATURATION

To perform ART procedures, ovarian stimulation is routinelly required in order to obtain a higher number of oocytes and increase the chances of embryo production to enable the selection of the best quality embryos for transfer (Fauser et al., 2005). Despite the increased number of oocytes that can be obtained using this procedure, some impairment has been observed. In addition to the high costs and the modest success rates, there are also potential health risks for the patients such as ovarian hyperstimulation syndrome in case of hyperresponse to ovarian stimulation (Huang et al., 2010).

In vitro maturation (IVM) is one of the most promising strategies for overcoming problems related to ovarian stimulation. Oocyte maturation consists of modification of

genomic structures, organelle restructurations, and molecular production to allow the gamete to receive spermatozoa for fertilization (Fulka et al., 1998; Mao et al., 2014). Using the IVM technique, immature oocytes are collected from ovaries of non-stimulated patients, followed by selection and exposure to IVM medium consisting of a base medium for cell culture supplemented with hormones, including FSH, LH, and estradiol. However, despite its clinical utility and successfull application in farm animals (Goto et al., 1988; Hwu et al., 1998), IVM of human oocytes remains an experimental approach not widely accepted in fertility clinics worldwide (Chang et al., 2014; Tannus et al., 2017). This is likely due to the lower pregnancy and live birth rates using in vitro compared to in vivo maturated oocytes, likely due to inadequacies of the culture media (Combelles et al., 2002; Ortega-Hrepich et al., 2013).

It is well established that embryo quality is dependent on oocyte quality (Lonergan et al., 2003; Ferris et al., 2016). In addition, correct and complete oocyte maturation is essential to efficient embryo production. Regarding IVM, the process can be disrupted by excess production of ROS, which is one of the major causes of oocyte depletion (Tamura et al., 2008; Karuputhula et al., 2013). For IVM, the addition of antioxidants is helpful, but these molecules may not exert their function with high efficiency due to their instability in in vitro enviroment, making utilization of nanomaterials an interesting strategy for molecule protection (Lucas et al., 2015; Komninou et al., 2016; Remião et al., 2016; Duarah et al., 2017; Manconi et al., 2017). One study from our group has show increased cleavage and blastocyst production rates, decreased ROS levels, and decreased the number of apoptotic cells/blastocyst when bovine oocytes were supplemented with nanoencapsulated melatonin in a IVM medium (Remião et al., 2016).

In another study, tretinoin was nanoencapsulated in lipidcore nanocapsules (LNC) and supplementation with the minor tested concentration (0.25 µM) in IVM medium was benneficial for bovine oocytes, resulting in higher cleavage and blastocyst rates, decreased P66Shc protein levels (the 66-kDa isoform of the growth-factor adapter Shc), and decreased ROS production. These benefits were not observed using the same concentration of non-encapsulated tretinoin (Lucas et al., 2015). Therefore, this represents a potential strategy for increasing the effectivness of human IVM and IVEP.

#### MULTIPLE PREGNANCY

Multiple pregnancies are a current problem in ART. The incidence of multiple pregnancies is related to pre-term birth, birth of babies with low weight and other complications, and risks to mothers and babies (Fauser et al., 2005; Vulliemoz et al., 2012). The high incidence of multiple pregnancies when using ART is related to the fact that sometimes more than one embryo is transferred into the female reproductive tract (Friedman et al., 2011; Mersereau et al., 2017).

In order to overcome the multiple pregnancy problems in ART, one alternative is the transfer of single embryos performed at a higher frequency (Mancuso et al., 2016). The methodologies assisting this condition are IVM, in vitro blastocyst culture, and embryo cryopreservation, techniques that have been highly studied in small and large animals and have been utilized commercially for many years (Sinclair, 2008).

Preimplantational genetic screening (PGS) and preimplantational genetic diagnosis (PGD) can also be useful to avoid multiple pregnancies, by discarding embryos with genetic disorders. To perform PGS and PGD, embryos are biopsied and evaluated using techniques such as karyotyping, fluorescent in situ hybridization (FISH), quantitative polymerase chain reaction (qPCR), array comparative genomic hybridization (aCGH), and next generation sequencing (NGS) (Chen et al., 2017).

Nanotechnology can help researches improve the application of PGS and PGD. Although highly employed, the current detection methods could be more sensitive and specific, more affordable and accessible to patients, faster, and easier to use to facilitate use in human reproduction clinics. Gold, silver, carbon, and magnetic nanomaterials are the main materials used to develop new methods of genetic diagnostics (Zhu et al., 2015). Nanotechnology combined with colorimetric (Stoeva et al., 2006) and electrochemical (Ozsoz et al., 2003) methods for nucleic acid analysis and detection has brought more sensitivity, lower cost, and increased simplicity and portability to diagnostics. This and other strategies recently developed for DNA analysis can be applied in the future to simplify PGD and PGS diagnostic procedures.

Another strategy for embryo selection is the culture of human embryos until day 5/6, when they reach the blastocyst stage. It has been previously shown that blastocyst transfer (day 5/6) presents better results than cleavage embryos (day 2/3) (Abuzeid et al., 2014; Yin et al., 2017). However, some clinics transfer embryos at the cleavage stage because most embryos fail to reach day 5/6 due to difficulties in mimicking the complexities of the in vivo environment (Alper et al., 2001; Tsirigotis, 1998). In vitro culture and manipulation of gametes and embryos stimulates production of exogenous ROS and leads to oxidative stress, reducing embryo quality (Agarwal et al., 2006; Truong et al., 2016). To overcome the challenge of embryo culture leading up to the blastocyst stage, research groups have looked for alternative approaches to improve in vitro embryo culture, including the addition of antioxidant molecules to the medium.

Studies on IVEP in animal models indicate antioxidant supplementation in medium is beneficial for blastocyst production. Antioxidants presenting beneficial effects in animal model in vitro embryo cultures include L-carnitine (Abdelrazik et al., 2009), hyaluronan (Romek et al., 2017), resveratrol (Salzano et al., 2014), and melatonin (Wang et al., 2013, 2014). However, in the case of bovine IVM, nanotechnology provides interesting alternatives for protecting of these molecules in in vitro environments (Lucas et al., 2015; Komninou et al., 2016; Remião et al., 2016). A recent publication confirmed this approach may represent a relevant alternative: supplementation of IVC medium with melatonin-loaded LNC increased embryo quality and blastocyst hatching in a bovine model (Komninou et al., 2016). This strategy is beneficial since the nanocapsules are biodegradable and do not result in toxicity when exposed to bovine oocytes (Lucas et al., 2017) or administered intradermally in rats (Bulcão et al., 2014).

### GENETIC DISORDERS

fphar-08-00994 January 9, 2018 Time: 17:50 # 5

The development and improvement of genome editing technology in the last few years has introduced gene therapy as a pre-emptive solution for correction of genetic anomalies. Monogenic diseases may be easily corrected using gene therapy, as they are caused by a single defective gene (Ma et al., 2017). Some monogenic diseases have already been targeted by gene therapy techniques, including lipoprotein lipase deficiency (Gaudet et al., 2016), hemophilia B (Nathwani et al., 2017), β-hemoglobinopathies (Negre et al., 2016), Wiskott– Aldrich syndrome (Aiuti et al., 2013; Morris et al., 2017), and inherited retinal degenerations (Gupta and Huckefeldt, 2017), although these diseases have not been treated in embryos.

Two recent reports have already shown the possibility of gene editing human embryos to correct genetic disorders. The studies used the CRISPR-Cas9 method to fix the human β-globin gene (Liang et al., 2015) and heterozygous MYBPC3 mutation (Ma et al., 2017), mutations responsible for β-thalassemia and hypertrophic cardiomyopathy, respectively. Although these studies have raised ethical concerns and the technology is still experimental, without proven efficacy and safety, both publications bring an important alternative to reproductive medicine through the treatment of diseases that until now were considered incurable (Ishii, 2017).

Nanotechnology development has resulted in some interesting non-viral strategies for molecule delivery in cells (Barkalina et al., 2015) contributing to the optimization of gene editing. One of them is the study of Sun et al. (2015) that delivered the Cas9 protein and a guide RNA through a DNA nanoclew to human osteosarcoma tumors in mice. Diverse studies have shown efficient gene delivery in mammalian cells via nanomaterials (Guan and Rosenecker, 2017; Riley and Vermerris, 2017). To produce genetically modified embryos, nanomaterials can be used to increase the efficiency of gene transfer via sperm mediated gene transfer. Silica nanoparticles (Barkalina et al., 2014b), magnetic iron nanoparticles (Kim et al., 2010), halloysite clay nanotubes (Campos et al., 2011), and poly(vinyl alcohol)-coated iron oxide nanoparticles (Makhluf et al., 2008) have already shown promise for delivery of nucleic acids and/or proteins to bovine spermatozoa.

Single-cell embryos can also be directly modified using nanomaterials. Das et al. (2016) hypothesized that if singlecell stage zona-free bubaline embryos are transfected with commercial transfecting agents and developed until the blastocyst stage (Selokar et al., 2015), nanomaterials could also be used to introduce genes into embryos at this stage and condition. This could be an alternative to not only viral vectors, but also other expensive methods such as pronuclear microinjection (Das et al., 2016). However, more studies are needed before introducing this technology into practice due to the possible toxic effects.

### CHALLENGES FOR THE USE OF NANOTECHNOLOGY IN REPRODUCTIVE SCIENCE

Nanotechnology has already and can continue to provide advantages for reproductive medicine. Despite the great solutions it can offer, some challenges still faces the use of this technology in medicine. As it is an emerging science, few studies have been performed to validate all the possibilities for treatments or diagnostics. One of the main questions that still need to be addressed regarding the use of nanotechnology is the toxicity it could cause. Despite new diagnostic methodologies being closer to being applied commercially, the ways in which nanomaterials are administered to organisms, embryos, or gametes need to be further studied.

Some nanomaterials are toxic to organism, mainly when exposure occurs during pregnancy and embryo development. For example, when pregnant mice are exposed to titanium dioxide nanomaterials, these materials can cross the placental barrier and cause anatomical defects in the fetuses (Melnik et al., 2013; Naserzadeh et al., 2017). In addition, silver nanoparticles decrease oestrogen plasma levels, increasing the number of resorbed fetuses (Campagnolo et al., 2017) and affecting embryonic growth (Austin et al., 2016). Carbon nanotubes also decrease the number of live fetuses per dam (Fujitani et al., 2015), the number of blood vessels on placenta, and increase the number of abortions (Qi et al., 2014).

Because of this, the utilization of these nanomaterials for reproductive proposes should be done carefully. One alternative is to search for additional nanomaterials that do not present toxicity, with biodegradable structures as the LNC and the dextran/chitosan nanoparticles, representing the most promising nanostructures for use in health applications.

### FINAL CONSIDERATIONS

Despite the recent advances in assisted reproductive technologies, some challenges remain, mainly related to pregnancy rates, multiple births, and genetic disorders. To overcome these problems, new alternatives must be identified. Nanotechnology represents a valuable tool that must be explored further to help researchers identify solutions for reproductive medicine. Nanomaterials can bring specificity, practice, and sensibility to next-generation diagnostic and treatment modalities.

It is expected that, as in other areas of medicine, the employment of nanotechnology could be helpful and beneficial to patients. In addition, researchers must be encouraged to develop more in vitro and in vivo tests using animal models to test safety and efficiency of these new methodologies. In addition, human clinical reproductive trials may also help accelerate commercial availability of new alternatives for ART.

### AUTHOR CONTRIBUTIONS

fphar-08-00994 January 9, 2018 Time: 17:50 # 6

MR, NS, AP, SG, FS, and TC had an equal participation in writing and approving the present manuscript.

#### REFERENCES


#### FUNDING

This study is part of the National Institute of Science and Technology in Pharmaceutical Nanotechnology: a transdisciplinary approach INCT-NANOFARMA, which is supported by (FAPESP, Brazil) grant no. 2014/50928-2, by (CNPq, Brazil) grant no. 465687/2014-8, CNPq (grant no. 311567/2016-9), and by PRONEX/FAPERGS-CNPq grant no. 165510000467-6.

and affect the placenta and the foetus. Nanotoxicology 11, 687–698. doi: 10.1080/ 17435390.2017.1343875



nanocapsules with distinct supramolecular structures: polymeric (NC) and lipid-core nanocapsules (LNC) on bovine embryo culture model. PLOS ONE 11:e0157561. doi: 10.1371/journal.pone.0157561




biosensor based on localized surface plasmon resonance. Int. J. Nanomedicine 7, 2921–2928. doi: 10.2147/IJN.S32641


**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 Remião, Segatto, Pohlmann, Guterres, Seixas and Collares. 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.

# Liposome-Encapsulated Baicalein Suppressed Lipogenesis and Extracellular Matrix Formation in Hs68 Human Dermal Fibroblasts

#### Chien-Liang Fang<sup>1</sup>† , Yiwei Wang<sup>2</sup>† , Kevin H.-Y. Tsai<sup>3</sup>† and Hsin-I Chang<sup>3</sup> \*

#### Edited by:

Sherry Y. Wu, The University of Queensland, Australia

#### Reviewed by:

Amirali Popat, The University of Queensland, Australia Cristian Rodriguez-Aguayo, University of Texas MD Anderson Cancer Center, United States

\*Correspondence:

Hsin-I Chang hchang@mail.ncyu.edu.tw

†These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Pharmaceutical Medicine and Outcomes Research, a section of the journal Frontiers in Pharmacology

Received: 09 November 2017 Accepted: 13 February 2018 Published: 06 March 2018

#### Citation:

Fang C-L, Wang Y, Tsai KH-Y and Chang H-I (2018) Liposome-Encapsulated Baicalein Suppressed Lipogenesis and Extracellular Matrix Formation in Hs68 Human Dermal Fibroblasts. Front. Pharmacol. 9:155. doi: 10.3389/fphar.2018.00155 <sup>1</sup> Division of Plastic and Reconstructive Surgery, Department of Surgery, Ditmanson Medical Foundation Chia-Yi Christian Hospital, Chiayi City, Taiwan, <sup>2</sup> Burns Research Group, ANZAC Research Institute, Concord Hospital, University of Sydney, Concord, NSW, Australia, <sup>3</sup> Department of Biochemical Science and Technology, National Chiayi University, Chiayi City, Taiwan

The dermis of human skin contains large numbers of fibroblasts that are responsible for the production of the extracellular matrix (ECM) that supporting skin integrity, elasticity and wound healing. Previously, an in vivo study demonstrated that dermal fibroblasts siting in the lower dermis are capable to convert into skin adipose layer and hence fibroblast lipogenesis may vary the structure and elasticity of dermis. In the present study, Hs68 human dermal fibroblasts were utilized as an in vitro model to study the lipogenesis via using adipogenic differentiation medium (ADM). Baicalein, isolated from Scutellaria baicalensis, is one of the flavonoids to inhibit adipocyte differentiation due to high antioxidant activity in vitro. In order to develop a suitable formulation for baicalein (a poorly water-soluble drug), soybean phosphatidylcholine (SPC) was used to prepare baicalein-loaded liposomes to enhance drug bioavailability. Our results demonstrated that liposome-encapsulated baicalein protected cell viability and increased cellular uptake efficiency of Hs68 fibroblasts. Lipid accumulation, triglyceride synthesis and gene expressions of lipogenesis enzymes (FABP4 and LPL) were significantly increased in ADM-stimulated Hs68 fibroblasts but subsequently suppressed by liposome-encapsulated baicalein. In addition, ADM-induced TNF-α expression and related inflammatory factors was down-regulated by liposome-encapsulated baicalein. Through ADM-induced lipogenesis, the protein expression of elastin, type I and type III collagens increased remarkably, whereas liposome-encapsulated baicalein can down-regulate ADM-induced ECM protein synthesis. Taken together, we found that liposome-encapsulated baicalein can inhibit ADM-induced lipid accumulation and ECM formation in Hs68 fibroblasts through the suppression of lipogenesis enzymes and inflammatory responses. Liposome-encapsulated baicalein may have the potential to improve wound healing and restore skin structure after skin injury.

Keywords: baicalein, liposomes, Hs68 human dermal fibroblast, adipogenic differentiation medium, lipogenesis

## INTRODUCTION

fphar-09-00155 March 2, 2018 Time: 15:48 # 2

Dermis in skin is the layer lies between epidermis and subcutaneous adipose tissue. Dermis is composed of fibrous and elastic tissue, thereby providing strength and flexibility to skin. Dermis is the thickest layer of the skin, but that can be varied based on body mass index (BMI), gender and locations. Recently, many researches revealed the relationship between adipose tissue in correlation with dermal fibroblasts. Ezure and Amano (2010) and Hew et al. (2016) demonstrated that the subcutaneous adipose layer of mice was remarkably thickened after being fed with high-fat diet, while the dermal layer was thinned (Ezure and Amano, 2010; Hew et al., 2016). Hence, such increase of adipose tissue may reduce the proliferation of dermal fibroblasts and elasticity of skin (Ezure and Amano, 2010). Moreover, adipogenesisdefect animal model indicated that intradermal adipocytes could mediate fibroblast recruitment during skin wound healing (Schmidt and Horsley, 2013). Interestingly, Wojciechowicz et al. (2013) confirmed that dermal fibroblasts siting in the lower dermis are capable to convert into skin adipose layer independently without influence from subcutaneous adipose tissue. Furthermore, the thickness of the lower dermis increased was found concomitant with the extension and downgrowth of hair follicles. Therefore, a key objective of studying the lipogenesis in Hs68 fibroblasts is to determine if and how lipogenesis can modulate inflammatory responses and change extracellular matrix (ECM) structure and composition in human dermal fibroblasts.

Dermal fibroblasts are the major cell population in dermis which are responsible for ECM production and wound healing. In the study reported by Rakar et al. (2012) primary human dermal fibroblasts were found to differentiate toward adipocytes, osteoblasts and chondrocytes using different induction media. Hence, relatively undifferentiated fibroblasts can express a particular phenotype depend on physiological stimuli and microenvironmental factors to which they are exposed (Koumas et al., 2003). This discovery supports the concept of fibroblast plasticity and proposes that fibroblasts can be transformed into adipocytes. In previous report suing human neonatal and adult lung tissues, alveolar interstitial fibroblasts were detected with the presence of lipid and the expression of adipocyte differentiation-related protein (ADRP), a protein necessary for lipid uptake, leading to their classification as lipid-containing fibroblasts or lipofibroblasts (Rehan et al., 2006). Lung specimens harvesting from patients with idiopathic pulmonary fibrosis (IPF) were found to have decreased lipofibroblast marker expression compared with non-IPF control samples (Bhattacharya, 2016), resulting in a hypothesis that conversion of fibroblasts into lipid-containing cells or lipofibroblasts may be able to prevent tissue fibrosis, abnormal wound healing, and hypertrophic scars. However, in skin wound healing, the role of lipogenesis in dermal fibroblasts were under investigated and the mechanisms has yet to be defined. Our previous study confirmed that adipogenic differentiation medium (ADM) can stimulate the differentiation of Hs68 human dermal fibroblasts to adipocytelike cells through the lipid accumulation and mRNA expressions of PPAR-γ, LPL and FABP4 (Fang et al., 2016). In the present study, we aim to further investigate the impact of lipogenesis on the inflammation responses and ECM formation in Hs68 fibroblasts.

Baicalein (5,6,7-trihydroxy-2-phenyl-4H-1-benzopyran-4 one), one of the remarkable flavonoids, is isolated from the root of Scutellaria baicalensis. Baicalein possesses a variety of biological activities, including high antioxidant, antiinflammatory, anti-proliferative, anti-apoptotic and anti-tumor activities (Gao et al., 2013). Baicalein has been demonstrated to decrease skin thickness and to suppress the expression levels of matrix metalloproteinase (MMP)-9, and vascular endothelial growth factor (VEGF) in ultraviolet (UV) B irradiated skin of mice models (Kimura and Sumiyoshi, 2011). Moreover, baicalein is known to inhibit radiation induced expression of nuclear transcription factor nuclear factor kappa B (NF-κB) and Cyclooxygenase-2 (COX-2) in human keratinocytes (Kimura and Sumiyoshi, 2011). Seo et al. (2014) found that baicalein can inhibit lipid accumulation and adipocyte differentiation by suppressing adipogenic factors such as PPARγ and C/EBPα through m-TOR signaling pathway in 3T3-L1 fibroblasts. However, using baicalein in therapeutic application are limited due to its low water solubility and poor oral bioavailability (Huang et al., 2014; de Oliveira et al., 2015). Recently, Rajkumari et al. (2017) have used baicalein as a reducing and capping agent in the synthesis of gold nanoparticles to inhibit Pseudomonas aeruginosa PAO1 biofilm formation. Li X. et al. (2017) also synthesized amine-modified mesoporous silica nanoparticles to encapsulate baicalein to exhibit anti-inflammatory effect on primary human gingival epithelial cells. Interestingly, baicalein was found to reduce the cytotoxicity of ZnO nanoparticles in Caco-2 cells (Li Y. et al., 2017). Therefore, liposomal nanoencapsulation of baicalein was introduced and investigated aiming to improve therapeutic efficacy through the increase of drug solubility and cell absorption efficiency (Moulaoui et al., 2015)

Composite phospholipid liposomes, similar to the lipid bilayer of cell membrane, have been formulated in nano-scaled sizes to increase the in vivo bioavailability for hydrophobic drugs (Kalepu and Nekkanti, 2015; Ong et al., 2016). Liposomes can be modified with various lipids to enhance drug loading efficiency and release characteristics (Mohammed et al., 2004) with reduced cytotoxicity, better biocompatibility and stability (Zeng et al., 2016). Various techniques, such as Bangham method, the detergent depletion method, the ether/ethanol injection method, the reverse phase evaporation and the emulsion method are previously reported to formulate drugloaded liposomes with high entrapment efficiency (EE), narrow polydiversity index (PDI) and long term stability (Galovic´ Rengel et al., 2002; Bergstrand et al., 2003). Based on low aqueous solubility and high cytotoxicity of baicalein on Hs 68 human dermal fibroblasts, phosphatidylcholine (PC) -based liposomes were used to encapsulate baicalein. Thereafter, we investigated the inhibitory effects of liposome-encapsulated baicalein on the ADM-induced adipogenesis, inflammatory responses and ECM synthesis in human dermal fibroblasts, Hs68.

### MATERIALS AND METHODS

fphar-09-00155 March 2, 2018 Time: 15:48 # 3

### Chemicals and Cell Culture

Baicalein was purchased from Sigma-Aldrich, USA. Phospholipon 90G (phosphatidylcholine 90%) was acquired from American Lecithin Company, Germany. All cell culture materials including, Dulbecco's modified eagle's medium (DMEM), fetal bovine serum (FBS), L-glutamine, adipocyte differentiation medium (ADM) were obtained from Gibco (Grand Island, NY, United States). Oil red o staining kit was purchased from Lifeline Cell Technology (Frederick, MD, United States). Adipogenesis assay kit was purchased from Sigma–Aldrich (St. Louis, MO, United States). All reagents and solvents are for research use only. Human foreskin fibroblasts (Hs68) were obtained from ATCC (Manassas, VA, United States) and Murine macrophages RAW 264.7 were purchased from Bioresource collection and Research Center, Food Industry Research and Development Institute, Taiwan. Both cell lines were cultured in DMEM supplemented with 10% v/v FBS, 100 units/ml penicillin and 100 µg/ml of streptomycine under steady state condition at 37◦C with 5% CO<sup>2</sup> in a humidified incubator.

### Liposomal Formulation

Baicalein-loaded liposomes were generated using a thin-film hydration and size reduction procedure as previously described (Yeh et al., 2015). Briefly, 100 mg of phospholipids were dissolved in 8 ml of chloroform, series amounts (20–60 µg) of baicalein were dissolved in 2 ml of ethanol and mixed together in a round-bottom flask. The organic solvents were evaporated by using rotary evaporator (Eyela, N-1000, Japan) at 45◦C prior to vacuum dry to form a dry lipid film. The lipid film was rehydrated by addition of 2 ml PBS (Phosphate buffered saline). Liposomes were then resized and uniformed through extruding polycarbonate membranes with series decreased pore sizes from 400 nm, 200 nm to 100 nm (Avanti Mini Extruderm, Alabaster, AL, United States). Empty liposomes were prepared by the same process with drug-free methanol.

#### Particle Characterization

The particle stability of liposomes was identified by storing and measuring the particle sizes over 2 weeks. The particle sizes were measured by using a Dynamic Light Scattering Instrument (HORIBA, LB-550, Japan). The solution of baicalein-loaded liposomes was diluted in approximate 30 times with doubledistilled water to make sure that the light scattering intensity was in the instrument's detectable range. The Polydispersity index (PDI) and zeta potential of liposomes were determined by Dynamic Light Scattering Analyzer (Malvern, Malvern Nano-Zs, England).

### Entrapment Efficiency

Loading efficiency of baicalein in liposomes was assessed using a high-speed centrifugation once the liposomes were formulated. Baicalein-loaded liposomes were spun down at 80,000 rpm by Beckman ultra-high speed centrifuge for 30 min. Next, the supernatants were carefully discarded and pellets were subsequently redissolved in the equal volume of ethanol. The concentration of baicalein in liposomes were then measured using an ELISA reader (Tecan, infinite M200) at wavelength of 277 nm. The encapsulation efficiency of baicalein in liposomes was calculated from the standard curve. The entrapment efficiency (EE) was obtained by the following formulation:

EE% = the amount of baicalein in liposomes initial amount of baicalein for drug loading×100%

### Cell Uptake of Baicalein-Loaded Liposomes

Cell uptake of baicalein-loaded liposomes in Hs68 human foreskin fibroblasts was examined using the fluorescent microscopy. DiI solution (1,1<sup>0</sup> -Dioctadecyl-3,3,3<sup>0</sup> ,30 - Tetramethylindocarbocyanine Perchlorate, 10 mg/ml, 1 µl) was added into liposomal solution to form DiI-loaded liposomes. Hs68 fibroblasts were cultured in 6 cm dishes at a density of 3 × 10<sup>5</sup> cells/ dish overnight at 37◦C with 5% CO2. After 1 day culture, cell culture medium was removed and replaced with DiI-loaded liposomes at various time intervals (0–24 h) prior to washing with PBS and fixation with 0.075% (v/v) formaldehyde solution for 30 min. In the control group, Hs68 fibroblasts were treated with DiI solution for 24 h. After fixation, cells were secondary washed with PBS and stained nuclei using DAPI solution (10 µg/ml) for 10 min. Finally, Hs68 fibroblasts were rinsed and mounted with PBS and photographed by microscope (Nikon TI-E) and CCD camera system (SPOT RT3). Fluorescent photographs were quantitative analyzed using Image J software (NIH, United State). Cell uptake efficiency was determined via measuring DiI fluorescence intensity.

### Cell Viability

Human foreskin fibroblasts, Hs68 and murine macrophages, RAW264.7 were seeded at a density of 10<sup>4</sup> cells/well and 5 × 10<sup>4</sup> cells/well in 96 well plates individually for cell viability analysis. After incubation overnight, cells were then treated with pure baicalein, empty or baicalein-loaded liposomes in concentrations of 10 or 20 µg/ml respectively for 24 h. After baicalein treatment, cell culture medium were replaced with 200 µl of 100 µg/ml MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) reagent and incubated for 4 h before measuring absorbance at 570 nm using ELISA reader (Tecan, infinite M200). Relative cell viability was demonstrated by percentage compared with control.

### Nitrite Assay

RAW264.7 macrophages were seeded at a density of 5 × 10<sup>5</sup> cells/well in 24 well plates, following by treatment with pure baicalein, empty or baicalein-loaded liposomes, in concentrations of 10 and 20 µg/ml respectively for 24 h. After removing of supernatants, no-phenol red medium (200 µl) were added for 8 h of incubation. The release of nitric oxide from inflamed macrophages was measured by determining nitrite concentration. Nitrite-contained medium (100 µl) were mixed with 100 µl of griess reagents in 96 well plates. Absorbance was measured at wavelength of 550 nm using ELISA reader (Tecan, infinite M200) after 15 min of shaking in dark. The reference values of nitric oxide were shown as the mean percentage of absorbance and standard deviation in comparison with lipopolysaccharide (LPS)-treated cells from two independent experiments.

## Oil Red O Staining

fphar-09-00155 March 2, 2018 Time: 15:48 # 4

Hs68 fibroblasts were seeded at a density of 10<sup>5</sup> cells/ dish in 6 cm dishes. After 24 h incubation, the culture medium was replenished with 20% (v/v) ADM and cultured for 14 days to induce lipogenesis, while pure baicalein, empty or baicaleinloaded liposomes were added at day 7. At the end of 14 days culture, cell culture medium was removed and rinsed with PBS twice. For histochemical examination, intracellular accumulation of lipid in Hs68 fibroblasts was performed using the Oilred O staining kit (Lifeline Cell Technology, Carlsbad, CA, United States). Concisely, cells were fixed with 4% (v/v) paraformaldehyde fixative solution for 30 min, rinsed with PBS and incubated with 100% 1, 2-propanediol dehydration solution for 10 min at room temperature. Following by fixation and dehydration, oil red O stain solution was added and incubated at 37◦C for 30 min prior to imaging of cell morphology under microscope (Nikon TI-E) and CCD camera system (SPOT RT3). Finally, 85% (v/v) 1,2-propanediol stain differential solution was added for 1 min to differentiate stain and the absorbance was measured by ELISA reader (Tecan, infinite M200) at the wavelength of 520 nm.

### Triglyceride Assay

Quantitative analysis of triglyceride content in ADM-induced Hs68 fibroblasts was conducted using an adipogenesis assay kit (Sigma–Aldrich) according to the manufacturer's instructions. Hs68 fibroblasts were cultured and adipogenesis was induced as described above. Free drug, empty liposomes or baicalein-loaded liposomes were added to cells respectively at day 7. After 21 days of incubation, 100 µl of lipid extraction buffer was added and incubated for 30 min at 90◦C. Cell medium became cloudy and then cooled down at room temperature. Cell medium was shaken for 1 min for homogenization followed by transfering 5–50 µl of lipid to another 96 well plate which were filled with adipogenesis assay buffer up to a total volume of 50 µl. Lipase solution (2 µl) was added and incubated for 10 min at room temperature for the degradation of triglyceride, following by adding master reaction mix for reaction of 30 min before measuring the absorbance at 570 nm using the ELISA reader (Tecan, infinite M200). Triglyceride content in Hs68 fibroblasts was calculated from the triglyceride (triolein)-equivalent standard curve.

#### Fluorescent Antibody Technique

Hs68 fibroblasts were seeded at a density of 10<sup>5</sup> cells/dish and treated with ADM for 14 days in the presence or absence of liposomal samples. After 14 days of adipogenic differentiation, medium was removed and rinsed with PBS. Cells were fixed with 10% formaldehyde at 4◦C for 30 min and washed with PBS twice. 0.1% Nonidet P-40 in PBS was added and incubated at room temperature for 10 min. After discarding Nonidet P-40 solution, cells were washed with PBS twice and blocked with BSA solution (2% w/v in PBS) for 30 min. Thereafter, the primary antibodies of anti-collagen I IgG produced in rabbit (1:500 in PBS, ab34710, Abcam, Cambridge, United Kingdom), anti-collagen III IgG produced in mouse (1:1000 in PBS, ab6310, Abcam, Cambridge, United Kingdom), and anti-elastin IgG produced in rabbit (1:200 in PBS, ab21610, Abcam, Cambridge, United Kingdom) were added separately and kept at 4◦C for overnight. After washing with PBS twice, the secondary antibody of H&L Dylight 594 anti-mouse IgG produced in goat (1:50 in PBS, ab96881 Abcam, Cambridge, United Kingdom) conjugated with reddish fluorescein (FITC) and green fluorescent antirabbit IgG produced in goat (1:200 in PBS, 111-095-003, Jackson ImmunoResearch Laboratories, WestGrove, PA, United States) were added in turn for 1 h of incubation. Finally, cells were washed with PBS twice and DAPI (10 µg/ml) was used to stain the nucleus of cells. The cell morphology was photographed by fluorescent microscopy equipped with CCD camera system (SPOT RT3). For quantitative analysis of elastin, type I and III collagen in hs68 fibroblasts, we used Image J software (NIH) to measure the fluorescence intensity in 10X images obtained with equal acquisition parameters.

### Collagen Assay

Collagen content in Hs68 fibroblasts was measured by Sircol collagen assay kit (Biocolor, United Kingdom) according to manufacturer's instructions. Hs68 fibroblasts were seeded at a density of 10<sup>5</sup> cells/ well in 6 well plates. Hs68 fibroblasts were incubated with 20% ADM for 7 days at 37◦C in a 5% CO<sup>2</sup> atmosphere and liposomal samples were added at day 3. At day 7, the medium was removed and washed with PBS. Pepsin solution (10 mg/ml in 0.5 M cold acetic acid) was added to each well for collagen isolation. After the supernatants were transferred to other eppendorf tubes, 1 ml Alkali Reagent was added and mixed by inverting contents for 30 min, then the samples were centrifuged at 12,000 × g for 10 min at 4◦C. Afterward, the supernatant was removed and 750 µl ice-cold Acid-Salt Wash Reagent was added into the collagen-dye pellet to eliminate unbound dye from the surface of the pellet. After centrifuged at 12,000 × g for 10 min, the supernatants were removed and 250 µl Alkali Reagent was added to each sample to dissolve the dye. Finally, the absorbance was measure by an ELISA reader at a wavelength of 555 nm.

#### Elastin Assay

Elastin content in Hs68 fibroblasts was measured by Fastin elastin assay kit (Biocolor, United Kingdom) according to manufacturer's instructions. Firstly, Hs68 fibroblasts were seeded at a density of 10<sup>5</sup> cells/ well in 6 well plates. Hs68 fibroblasts were incubated with 20% ADM for 7 days at 37◦C in a 5% CO<sup>2</sup> atmosphere, and test samples were treated to cells at day 3. At day 7, the medium was removed and washed with PBS. Then, 1 ml of trypsin solution was added into each well to detach Hs68 fibroblasts, and cell lysing solution was transferred to a 1.5 ml micro centrifuge tube. In order to convert cell bound elastin to water soluble α-elastin, 300 µl of cell lysing solution was mixed with 100 µl of 1.0 M oxalic acid to make the final concentration of 0.25 M and heated at 100◦C for an

hour. Afterward, an equal volume of elastin precipitating reagent was added and vortexed gently for 15 min to make sure the complete precipitation. After centrifugation at 10,000 g and 4◦C for 10 min, the supernatant was discarded and replaced with 1ml of Dye Reagent. The samples were shaken evenly for 90 min and centrifuged at 10,000 g for another 10 min. The unbound dye was drained out and 250 µl of Dye Dissociation Reagent was added to each sample to dissolve the dye. Finally, the absorbance was measure by an ELISA reader at a wavelength of 513 nm.

#### Quantitative Real-Time PCR

Hs68 fibroblasts were cultured in 6 cm dishes at different cell densities for the studies of TNF-α induction and ADM stimulation. After treatment, the total RNA in cells were extracted using Trizol reagent (Protech Technology, Taiwan) in reference to the manufacturer's instructions. Messenger RNA was subsequently reverse transcribed to complementary DNA following the method of TProfessional basic (Biometra, Germany). The obtained cDNA was quantified to 20 ng and the measurement was conducted in StepOnePlusTM Real-Time PCR system with FastStart DNA Mater-PLUS SYBR Green I (Applied Biosystem, United States). The primary sequences were shown in **Table 1**. The efficiency of DNA amplification was performed using the mean cycle threshold (Ct) method, which represent the number of cycles at which the fluorescence surpasses the threshold. 1Ct value in each group was calculated from Ct values of different genes subtracting the Ct value of GAPDH. The resulting relative mRNA expression was showed as fold change (2−11Ct) relative to the expression values in control cells.

#### Statistical Analysis

Statistical analysis was completed by using commercial software Minitab 16. Data are presented as means ± standard deviations. All p-values were calculated by two-tails t-test and paired t-test. P < 0.05 was considered the significant difference.

### RESULTS

#### Characterization of Liposomal Formulations

Physical property is an important parameter in formulation of liposomes as it could affect drug delivery or interaction between drugs and cells. In the present study, physical characteristics of baicalei-loaded liposomes, including particle size, entrapment efficiency, polydispersity index (PDI) and zeta potential, were assessed. **Table 2** illustrates that the particle sizes of empty liposomes, 10 µg/ml baicalein-loaded liposomes (10 µg/ml BC-Lip) and 20 µg/ml baicalein-loaded liposomes (20 µg/ml BC-Lip) are in the range of 135–171 nm. Liposomes encapsulated with baicalein were found with reduced particle size in a dose-dependent manner. Moreover, an increase of baicalein concentration in liposomes led to a decrease in PDI from 0.546 to 0.462. PDI is an important parameter which is used to describe variation of particle size in a population of particles and hence baicalein-loaded liposomes have a relatively narrow size distribution. Moreover, baicalein-loaded liposomes present a decline in entrapment efficiency from 33.65 to 25.40% when drug concentration increase from 30 to 80 µg/ml. The surface charge of baicalein-loaded liposomes was assessed through zeta potential measurement. Since baicalein-loaded liposomes are mainly composed with phosphatidylcholine, an amphoteric phospholipid, these liposomes should exhibit a neutral surface and hence the zeta potential was around zero. Similar values of zeta potential were obtained among these liposomal formulations, suggesting that the presence of baicalein did not alter the electrophoretic mobility of liposomes. The


TABLE 2 | Physical parameters of the liposomal formulations after extrusion.


stability of liposomal formulations was further determined through particle size (**Figure 1**). Although particle sizes of empty and baicalein-loaded liposomes were slightly different at day 1, all the liposomes displayed similar stability profiles over a 2-week period at 4◦C in the fridge. The stability of liposomes in vivo can also be affected by interactions between lipoproteins and few other proteins in the blood. In our previous serum stability studies of liposomes, no significant change in the particle size distribution for all the types of liposomal formulations was noted (Yeh et al., 2015).

### Effect of Liposomal Formulations on the Cell Viability, Delivery Efficiency and NO Production

The in vitro cytotoxic of free baicalein, empty and baicaleinloaded liposomes on Hs68 fibroblasts and RAW264.7 macrophages were examined using the MTT assay. As shown in **Figure 2**, free baicalein at the concentration of 20 µg/ml was found to reduce cell viability sharply to 14%. In comparison, liposome-encapsulated baicalein at the same concentration showed highly remained cell viability (around 100%), indicating that liposomal formulation was capable to prevent the cytotoxicity of baicalein with significantly enhanced cell viability. However, both free baicalein and liposomeencapsulated baicalein at the concentration of 30 µg/ml showed low cell viability (less than 15%).

Cell uptake of baicalein-loaded liposomes in Hs68 fibroblasts was analyzed based on release of DiI in to cells after cell uptake of DiI-labeled liposomes (Yeh et al., 2015). In here, DAPI stained cell nucleus apparent in blue fluorescence. As shown in **Figure 3A**, Dil-loaded liposomes were gradually taken by cells showing more red fluorescence signals over 24 h. Quantitative analysis of DiI red fluorescence signals demonstrated the fluorescent intensity in cells which were treated with DiI-loaded liposomes was four times higher compared to control cells which were treated with Dil only over 24 h (**Figure 3B**). Hence, liposomal formulation was helpful to improve cell viability and cellular uptake efficiency of baicalein in Hs68 fibroblasts.

To confirm the anti-inflammatory activity of liposomeencapsulated baicalein, cell viability and nitrite production of RAW264.7 were measured. Liposome-encapsulated baicalein showed higher cell viability compared to free drug in treating RAW264.7 macrophages. Baicalein and liposome-encapsulated baicalein dose-dependently reduced cell viability of RAW264.7 macrophages (**Figure 4A**). Based on this result, concentrations with cell viability over 80% were selected for subsequent NO inhibition experiment. The nitrite accumulation in the cells was significantly increased after LPS stimulation. Therefore, cells were simultaneously treated with LPS following by free baicalein, empty or baicalein-loaded liposomes, respectively in order to determine anti-inflammatory activity of baicalein (**Figure 4B**). LPS-induced nitrite production in RAW264.7 macrophages was significantly inhibited through the treatment of free baicalein, empty or baicalein-loaded liposomes. Baicalein inhibited the NO production in a dose-dependent manner, particularly at 20 µg/ml of baicalein (greater than 80% inhibition). liposome-encapsulated baicalein also exhibited the suppressive effect on LPS-induced nitrite production (in approximate 65–75% inhibition), but not in a dose-dependent manner. The result of MTT assay confirmed that baicalein or liposome-encapsulated baicalein had direct antiinflammatory effects which was not correlated with cell damage (cell viability > 80%).

### The Effect of Liposome-Encapsulated Baicalein on Lipid Accumulation and Triglyceride Synthesis in ADM-Induced Hs68 Fibroblasts

Lipogenesis effects of baicalein or liposome-encapsulated baicalein were examined respectively on Hs68 fibroblasts

FIGURE 3 | Cellular uptake of DiI-loaded liposomes in Hs68 fibroblasts. (A) Hs68 fibroblasts were incubated with DiI and DiI-loaded liposomes at different time intervals (1–24 h). The fluorescent image was photographed under fluorescent microscopy paired with CCD system. (B) The fluorescent intensity of DiI-loaded liposomes inside the cells was quantified using Image J. The data are expressed in relative index compared with control. The results are presented as the means ± standard deviation (×100 magnification, scale bar = 200 nm, n = 4).

within ADM induced lipogenesis for 14 days. Oil Red-O was utilized to measure intracellular lipid accumulation. Under light microscopy, lipid droplets were apparent via treatments of ADM in the presence or absence of empty liposomes (**Figure 5A**). Moreover, cells co-treated with ADM and empty liposomes had a significantly increase in intracellular lipid content compared to the cells treated with ADM only. In contrast, baicalein (10 µg/ ml) showed a significant inhibitory role against ADM-induced lipid formation, while baicaleinloaded liposomes showed comparable inhibitory effects on lipid formation but not dose dependently (**Figure 5B**). These results were further confirmed by quantitative analysis of intracellular triglyceride contents. Suppressive effect on triglyceride accumulation was observed in Hs68 fibroblasts (**Figure 5C**). As expected, ADM induced triglyceride synthesis in Hs68 fibroblasts was significantly suppressed by free baicalein (10 µg/ml), while after liposome-encapsulation, baicalein-loaded liposomes were found to reduce ADM induced triglyceride synthesis, suggesting that a similar but mild effects of baicalein after liposome-encapsulation on the lipid formation and triglyceride synthesis in Hs68 fibroblasts.

### The Effect of Liposome-Encapsulated Baicalein on Gene Expressions of Lipogenesis Enzymes in ADM-Induced Hs68 Fibroblasts

To understand the molecular mechanism of baicalein's effects with or not encapsulated in liposomes on lipid accumulation in Hs68 cells, we examined the expression of lipogenesis enzymes, lipoprotein lipase (LPL) and fatty acid binding protein 4 (FABP4). A previous study demonstrated that the gene expressions of LPL and FABP4 in human adipose-derived stromal cells were significantly increased during adipogenic differentiation (D7–D21) and hence they could be served as potential adipogenic differentiation markers (Ambele et al., 2016). Cells treated with ADM, in presence or absence of empty liposomes, resulting in greater gene expression of LPL (2 fold) and FABP4 (sixfold) comparing to control and hence empty liposomes had no additive effect on lipogenesis (**Figure 6**). The increase of these lipogenic enzyme genes indicates that Hs68 fibroblasts can be induced to lipogenesis by ADM treatment no matter the addition of liposomes. In contrast to ADM-induced lipogenesis in Hs68 fibroblasts, free baicalein treatment demonstrated 70 and 48% decrease in LPL and FABP4 expression levels, respectively. Cells co-treated with ADM and liposome-encapsulated baicalein also exhibited suppressive effects on LPL (62% decrease) and FABP4 (49% decrease) mRNA expressions. Such expression profiles of LPL and FABP4 are consistent with lipid and triglyceride levels. Taken together, both free baicalein and liposome-encapsulated baicalein inhibited lipid formation and triglyceride synthesis in Hs68 fibroblasts through suppression of lipogenic enzyme genes.

### The Effect of Liposome-Encapsulated Baicalein on Gene Expression of Inflammation-Related Factor After ADM Treatment

As obesity is known to induce mild inflammatory responses, while inflammation plays an important role in insulin resistance, diabetes and other diseases. Hotamisligil reported that the mRNA expression of tumor necrosis factor alpha (TNF-α) mRNA expression in subcutaneous adipose tissue were about 2.5-fold higher in obese women compared to women with normal BMI. Importantly, reduction of body weight (17% w/w) were found to correlate with 22% decrease of TNF-α expression, and lead to indirectly improvement of insulin sensitivity (Hotamisligil et al., 1995). Thus, TNF-α expression was used in the present study as an indicator to determine the release of adipokines in the adipose tissue (Cawthorn and Sethi, 2008). To clarify whether ADM-induced adipogenesis can stimulate inflammation in Hs68 fibroblasts, we measured the mRNA levels of TNF-α. In here, we found that ADM treatment resulted in 1.5-fold increase in the gene expression of TNF-α, and which can be reduced approximate 75 and 80% by baicalein and liposome-encapsulated baicalein respectively (**Figure 7**). In addition, ADM-induced

TNF-α expression can be suppressed by liposome-encapsulated baicalein in a dose-dependent manner. A previous report using fresh peritoneal murine macrophages harvested from C57RL/6 mice or ANA-1 macrophage line derived from the bone marrow of C57RL/6 mice demonstrated that LPL can induce TNF-α gene expression and protein secretion (Renier et al., 1994). Moreover, Xu et al. (2015) also indicated that FABP4 could modulate TNF-α secretion in FABP4/aP2 knockout macrophages, suggesting ADM treatment may induce TNF-α expression in Hs68 fibroblasts through the up-regulation of lipogenic enzyme gene expressions (LPL and FABP4).

ADM-induced adipogenesis caused inflammatory responses in Hs68 fibroblasts through the up-regulation of TNFα gene expression. We further treated Hs68 fibroblasts treated with 20 ng/ml of TNF-α following by measure of the mRNA expressions of inflammatory cytokines, such as cyclooxygenase-2 (COX-2), interleukin-6 (IL-6) and IL-8. As shown in **Figures 8A–C**, incubation of Hs68 fibroblasts with TNF-α increased the gene expressions of COX-2, IL-6 and IL-8 (about 2, 2 and 3.5-fold) and the addition of baicalein, empty and baicalein-loaded liposomes individually downregulated TNF-α-induced inflammatory gene expressions. Since TNF-α was reported to stimulate ECM degradation by inducing the expression of matrix metalloproteases (MMP)-1 and MMP-3 in dermal fibroblasts (Shindo et al., 2014), we then examined the effect of baicalein, empty and baicalein-loaded liposomes on gene expressions of MMP-1 and MMP-3. Incubation of Hs68 fibroblasts with TNF-α significantly increased gene expressions of MMP-1 and MMP-3 compared to non-treated control cells, whereas baicalein and liposome-encapsulated baicalein significantly reduced TNFα-induced MMP-1 and MMP-3 mRNA levels (**Figure 8D**). It is interesting to note that empty liposomes have similar

inhibitory effect on TNF-α-induced MMP-1 and MMP-3 mRNA expressions but not in statistical significance. Through the up-regulation of TNF-α gene expression, ADM-induced

relative to cells treated with ADM only and \$p < 0.05 relative to cells

co-treated with ADM and empty liposomes.

mean ± standard deviation (n = 3). #P < 0.05 relative to control and <sup>∗</sup>p < 0.05 relative to cells treated with TNF-α only.

lipogenesis may cause inflammation and ECM degradation in Hs68 fibroblasts. However, baicalein, empty and baicaleinloaded liposomes show their potential to suppress inflammatory responses and MMP expressions via down-regulation of TNF-α pathway.

### The Effect of Liposome-Encapsulated Baicalein on ECM Synthesis in ADM-Induced Hs68 Fibroblasts

Dermal fibroblasts have major responsibility to produce ECM for maintaining skin homeostasis and for orchestrating skin tissue regeneration. Since ECM plays a critical role in regulation of skin cell morphogenesis, activity and function, ADM, baicalein and liposome-encapsulated baicalein was further investigated on ECM synthesis. In the presence of ADM and empty liposomes, Hs68 fibroblasts were found enlarged and altered in cell shape. In addition, cells co-treated with ADM and empty (or baicalein-loaded) liposomes produced an elaborate type I collagen matrix (**Figure 9A**), whereas control cells or cells cotreated with ADM and baicalein showed no cross-linked network of collagen microfibers. Quantitative analysis of the fluorescent signals revealed that cells treated by ADM or empty liposomes exhibited higher expression levels of elastin, type I and type III collagens compared to control and cells co-treated with ADM and baicalein (or liposome-encapsulated baicalein) (**Figures 9B–D**).

These results indicated that baicalein (or liposome-encapsulated baicalein) can largely reduce the stimulatory effect of ADM and liposomal formulation on ECM synthesis.

To further confirm the effect of baicalein and liposomeencapsulated baicalein, collagen and elastin protein level in Hs68 fibroblasts were analyzed using ELISA assay. Results are in agreement with immunohistochemical staining showing a similar declining trend in collagen and elastin production (**Figure 10**). Taken together, ADM and empty liposomes could induce ECM expression level in Hs68 fibroblasts but baicalein or liposomeencapsulated baicalein may have the reverse effects.

### DISCUSSION

Previous studies revealed that baicalein is capable of attenuating ROS generation and exhibit high anti-oxidant activity (Chang et al., 2011). We also found that baicalein displays strong DPPH radical scavenging action, particularly at the concentration of 20 µg/ml (around 90%, **Supplementary Figure S1**). In a study reported by Kimura and Sumiyoshi (2011) baicalein could inhibit UV-B-induced MMP-9 and VEGF expression level through suppression of COX-2 but expression of NFκB and hence baicalein is believed to have a remarkable anti-inflammation activity. In the present study, our findings demonstrate that baicalein is capable to decrease LPS-induced nitrite production in RAW264.7 macrophages and TNF-α stimulated COX-2, IL-6 and IL-8 expressions in Hs68 fibroblasts. These results agree with previous studies, showing that baicalein possesses excellent anti-inflammatory effect (Kimura and Sumiyoshi, 2011; Chen et al., 2014). However, due to its low solubility in aqueous solutions and low bioavailability in vivo, medical application of baicalein are very limited, particularly for skin administration. Few recent studies suggested that the therapeutic efficiency of natural compounds could be improved by liposomal nanoencapsulation because of its high compatibility and easy incorporation efficiency (Cadena et al., 2013; Caddeo et al., 2014). Encapsulated drugs in nanoparticles

\$p < 0.05 relative to cells co-treated with ADM and empty liposomes.

are known to reduce drugs leakage, prolong the residence time on skin and facilitate the internalization of drugs into cells (Moulaoui et al., 2015). Tsai et al. (2012) found that encapsulated baicalein with nanostructured lipid carrier (NLC) system can be used in brain therapy. In the present study, we used soybean phosphatidylcholine (SPC) for baicalein-loaded liposomal formulations to enhance drug bioavailability. Notably, particle size and PDI were decreased with baicalein–loaded liposomes, and no significant differences in zeta potential was measured between empty liposomes and baicalein-loaded liposomes. The reduction of particle size may be due to stronger drug interactions via hydrogen bonding and hence there is no effect on zeta potential. After 14 days stored at 4◦C, only less 20% variation in particle sizes was noted for empty or baicalein-loaded liposomes. Moreover, our previous studies of liposome' stability in serum demonstrated no significant change in the particle size distribution for all the liposomal formulations (Yeh et al., 2015). Hence, liposomal formulations are in high uniformity and homogeneity based on the low PDI and great stability within 14 days. In comparison with free drug, liposome-encapsulated baicalein (20 µg/ml, equal to 74 µM) obtained high cell viability in Hs68 fibroblasts due to the reduction of drug leakage and the anti-inflammatory activity of liposome-encapsulated baicalein is consistent with free drug.

Rakar et al. (2012) showed that primary human dermal fibroblasts were able to differentiate toward adipocytes, osteoblasts and chondrocytes using different induction media. Additionally, Takeda et al. (2017) also observed dramatic morphological change of human dermal fibroblasts over differentiation in which cells morphology convert from characteristic elongated fibroblasts into round adipocyte-like cells. In here, ADM-induced lipogenesis enlarged the cell size of Hs68 fibroblasts with oval and round cells rather than spindle shape. Such morphological changes of Hs68 human dermal fibroblasts during lipogenesis is consistent with the "morphotypes" described in previous studies (Rakar et al., 2012; Takeda et al., 2017). Conversely, Lu et al. (2006) previously demonstrated that baicalein at concentrations of 160–640 µM inhibited the proliferation of porcine preadipocytes with over expression of adipogenesis related genes: PPARγ2 and fatty acid synthase (FAS) but the gene expression was suppressed in lower concentrations of baicalein at 40–320 µM. Similarly, the inhibitory role of baicalein on lipid accumulation in 3T3-L1 fibroblasts and zebrafish was published by Seo et al. (2014). Baicalein inhibited triglyceride accumulation during adipogenesis and significantly down-regulated the gene expression of lipogenesis enzyme, FABP (Cha et al., 2006). In our study, as expected, baicalein or liposome-encapsulated baicalein remained cells with fibroblast-like morphology rather than to adipocytes, and reduced lipid droplet formation in ADM-induced Hs68 fibroblasts. Furthermore, Baicalein and liposome-encapsulated baicalein showed significant inhibitory effect on ADM-induced lipid accumulation and triglyceride synthesis through the suppression of adipogenesis marker, FABP4 and LPL. Thus, we believed that baicalein and liposomeencapsulated baicalein could suppress ADM-induced lipogenesis in human dermal fibroblasts.

As lipogenesis is known to induce inflammatory responses via the increase of TNF-α expression (Hotamisligil et al., 1995; Cawthorn and Sethi, 2008), ADM-induced lipogenesis markedly increased TNF-α expression in Hs68 fibroblasts through the upregulation of LPL and FABP4 expression, and which was further inhibited by the addition of baicalein, empty and baicalein-loaded liposomes (Renier et al., 1994; Xu et al., 2015). Of note, baicalein, empty liposomes and baicalein-loaded liposomes demonstrated prominent suppression in COX-2, IL-1, IL-6, and IL-8 gene

expressions in TNF-α induced Hs68 fibroblasts. These findings are in agreement with previous studies (Hsieh et al., 2007; Kimura and Sumiyoshi, 2011; Luo et al., 2017), showing baicalein can consistently suppress gene and protein expressions of Cox-2, IL-1, IL-6 and IL-8. Results herein further confirm that liposomes didn't effect on inhibitory role of baicalein on gene expressions of inflammatory responses. Moreover, our results showed that pure baicalein and baicalein-loaded liposomes suppressed gene expression of MMP-1 and MMP-3, while similar results were reported before by Chen et al. (2015) through measuring the gene and protein expressions of MMP-1, 3, and 13 in baicalein treated human OA chondrocytes. They found baicalein can dosedependently reduce gene and protein expressions of MMP1, 3 and 13 in IL-1β-induced human OA chondrocytes (Chen et al., 2015), while other researchers also demonstrated that baicalein can down-regulate gene and protein expressions of MMP-1 in H2O2-treated human HaCaT keratinocytes (Kim et al., 2012). According to our data, baicalein and baicalein-loaded liposomes have similar inhibitory effect on gene expressions of MMP-1 and MMP-3 and that is consistent with previous findings (Kim et al., 2012; Chen et al., 2015). Besides, baicalein inhibited the gene and protein expressions of MMP-2 and MMP-9 and meanwhile promoted the expressions of tissue inhibitors of metalloproteinase (TIMP)-1 and TIMP-2 in hepatocellular carcinoma MHCC97H cells and mouse melanoma B16F10 cells (Chen et al., 2013; Choi et al., 2017), suggesting that baicalein can suppress inflammatory responses through the down-regulation of TNF-α expression, resulting in the inhibition of MMP expressions in Hs68 fibroblasts.

Since ECM is playing a key role in skin regeneration, we therefore examined protein expressions of elastin, type I and type III collagens in Hs68 fibroblasts. Our results indicated that ADM stimulation up-regulated elastin, type I and type III collagens in Hs68 fibroblasts. Of note, liposomes slightly increased type I and type III collagens but not elastin as compared with ADM-treated cells. Similar to our findings, human subcutaneous adiposederived cells were found to have presence of COL1A1 gene (Mariman and Wang, 2010). In a DNA microarray analysis of subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) in Wistar rats, ECM-related genes such as type I, III, and V collagen were expressed higher in SAT than VAT (Mori et al., 2014). Moreover, collagen type I protein was highly expressed and formed a fibrous structure in rat SAT as well as dermis but not in rat VAT. In comparison with undifferentiated cells, a decrease of type I, type III and type V collagens was observed in 3T3-L1 cells during early phase of adipogenic differentiation (Wang et al., 2004; Mori et al., 2014). Enlarged adipocytes reduced 3T3-L1 fibroblast proliferation and gene expression of collagen type I and elastin and increased gene expression of MMP 13 (Ezure and Amano, 2011). Therefore, Hs68 fibroblast may present similar ECM expressions (such as type I and type III collagens) to human subcutaneous adipose-derived cells, rat SAT and VAT but not mouse 3T3-L1 adipocytes. Besides, baicalein and liposome-encapsulated baicalein suppressed the

protein expressions of elastin, type I and type III collagen in ADM-induced Hs68 fibroblasts which may due to the inhibition of lipogenesis. Therefore, we elucidated that lipogenesis in Hs68 fibroblasts can increase ECM formulation, particularly in the protein expression of elastin, type I and III collagens. It is worth noting that baicalein had inhibitory effect on ECM formation in lipid-containing fibroblasts.

In summary, we propose the following network: Physiological stimuli (ADM treatment) influenced the expressions of lipogenic enzyme genes (LPL and FABP4); simultaneously, lipogenesis enzymes control the lipid accumulation and inflammatory responses, which are the key factors for ECM remodeling in Hs68 fibroblasts (**Figure 11**). Besides, Liposomeencapsulated baicalein provide enhanced cell viability and cellular uptake efficiency of Hs68 fibroblasts together with down-regulation of ADM-induced lipid accumulation and ECM formation in Hs68 fibroblasts through suppression of lipogenesis enzymes and inflammatory responses. In conclusion, we suggested that liposome-encapsulated baicalein can provide an opportunity as medical or cosmetic products to prevent lipogenesis and maintain ECM structure in skin.

### REFERENCES


### AUTHOR CONTRIBUTIONS

C-LF designed and performed the experiments and analyzed the data. KT performed the experiments. YW wrote the manuscript and contributed to data analysis. H-IC supervised the project, contributed to data analysis, and wrote the manuscript.

### FUNDING

This work was supported by a research grant from Ditmanson Medical Foundation Chia-Yi Christian Hospital Research Program (R104-20).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar. 2018.00155/full#supplementary-material

FIGURE S1 | The anti-oxidant activity of baicalein measured by DPPH scavenging assay.



osteogenic phenotype shifts. Differentiation 84, 305–313. doi: 10.1016/j.diff. 2012.08.003


**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 reviewer AP and handling Editor declared their shared affiliation.

Copyright © 2018 Fang, Wang, Tsai and Chang. 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.

## The Pathways for Layered Double Hydroxide Nanoparticles to Enhance Antigen (Cross)-Presentation on Immune Cells as Adjuvants for Protein Vaccines

#### Shiyu Yan<sup>1</sup> , Kewei Xu<sup>2</sup> , Li Li<sup>1</sup> \*, Wenyi Gu<sup>1</sup> , Barbara E. Rolfe<sup>1</sup> and Zhi P. Xu<sup>1</sup> \*

<sup>1</sup> Australian Institute for Bioengineering and Nanotechnology, The University of Queensland, Brisbane, QLD, Australia, <sup>2</sup> School of Medicine, The University of Queensland, Brisbane, QLD, Australia

Nanoparticles (NPs) are intensively investigated as adjuvants in new generation vaccines, while how these NPs promote the immune responses has not been well understood. In this research, we have tried to elucidate the possible pathways for layered double hydroxide (LDH) NPs to provoke immune responses. As previously reported, LDH NPs efficiently deliver antigens to antigen presenting cells (APCs). In this research, we have found that these internalized LDH NPs are not released by these APCs within 8 h. We have for the first time found that macrophage cells exchange the internalized LDH NPs with other surrounding ones, which may promote immune responses in an additional way. Moreover, the internalized LDH-antigen NPs significantly facilitate the maturation of immature DCs and enhance cross-presentation of epitope/MHC class I complexes on the DC surface. This research would help understand the NP adjuvant mechanism and further assist the design of new specific NPs as more efficient nano-adjuvants.

Keywords: nanoparticle adjuvant, layered double hydroxides, antigen presentation cells, dendritic cell maturation, cross-presentation, immune responses

#### INTRODUCTION

Various nanomaterials have been widely examined to deliver immunogens and immune stimulants as adjuvants in vaccine development (Fan and Moon, 2015; Zhang et al., 2015). For example, polystyrene delivering ovalbumin (OVA) induces systemic immune responses in sheep (Scheerlinck et al., 2006). Poly (lactic-co-glycolic acid) (PLGA) co-delivering tyrosinase-related protein 2 (TRP-2) and toll-like receptor ligand 4 (TLR4) induces cell-mediated immunity (Hamdy et al., 2008). Mesoporous silica (MS) as the antigen adjuvant also shows good adjuvant activities in HIV (Cheng et al., 2012) and porcine circovirus (Guo et al., 2012) vaccines. Calcium phosphate (Cap) adjuvanted herpes simplex vaccine shows systemic immune response in mice (He et al., 2002) and layered double hydroxide (LDH) delivering DNA vaccine shows high efficiency in transfection, and promotes immunity (Li et al., 2011; Wang et al., 2014; Williams et al., 2014). In particular, we have previously reported that LDH co-delivering OVA/TLR9 ligand CpG and Intimin β (IB) promotes potent humoral and cell-mediated immunities (Yan et al., 2014, 2018; Chen et al., 2016). However, how nanomaterials as adjuvants stimulate strong immune responses has not been well understood.

#### Edited by:

Susan Hua, University of Newcastle, Australia

#### Reviewed by:

Robert Braidwood Sim, University of Oxford, United Kingdom Neil Martin O'Brien-Simpson, The University of Melbourne, Australia

#### \*Correspondence:

Li Li l.li2@uq.edu.au Zhi P. Xu gordonxu@uq.edu.au

#### Specialty section:

This article was submitted to Translational Pharmacology, a section of the journal Frontiers in Pharmacology

Received: 07 May 2018 Accepted: 03 September 2018 Published: 20 September 2018

#### Citation:

Yan S, Xu K, Li L, Gu W, Rolfe BE and Xu ZP (2018) The Pathways for Layered Double Hydroxide Nanoparticles to Enhance Antigen (Cross)-Presentation on Immune Cells as Adjuvants for Protein Vaccines. Front. Pharmacol. 9:1060. doi: 10.3389/fphar.2018.01060

The most critical adjuvant processes include the assisted cellular uptake of antigen and subsequent antigen presentation or cross-presentation by antigen-presenting cells (APCs) (Toes et al., 1996). After subcutaneous administration of vaccine formulations, APCs are recruited to take up the nanomaterialantigen particles, and then circulate to the local regional nodes. During this period, the nanomaterial-antigen particles are processed within the APCs to present the epitope and prime naïve lymphocytes (Slingluff, 2011). Therefore, the understanding of APCs' cellular uptake and APCs' antigen (cross)-presentation pathways via nanoparticles is very important in adjuvant design and development.

It has been confirmed that LDH NPs facilitate negatively charged antigens (such as BSA) to attach onto and enter the cell (Gu et al., 2015; Chen et al., 2016). As reported by Li et al. (2010) 60–65% bone marrow dendritic cells (BMDCs) took up LDH nanoparticles within 3 h. LDH is a family of anionic clay minerals, with the general formula of [M2<sup>+</sup> 1−x M3<sup>+</sup> <sup>x</sup> (OH)2] <sup>x</sup>+[An<sup>−</sup> <sup>x</sup>/n·yH2O]x−, where M2<sup>+</sup> is a divalent cation, M3<sup>+</sup> a trivalent cation, and An<sup>−</sup> an anion (Braterman et al., 2004). LDH has positively charged hydroxide basal layers where the trivalent cations substitute for the divalent cations, which are balanced by the hydrated anions intercalated in the interlayer space. MgAl-LDH NPs possess low toxicity and good biocompatibility, high loading of proteins and proteomic vaccines and a high capability to facilitate the cellular uptake of payloads (Xu et al., 2006c, 2008a), which may explain why LDH NPs can act as effective adjuvants to stimulate strong immune responses in vaccine development (Yan et al., 2014). However, there is no report regarding (1) whether LDH-antigen complexes facilitate maturation of APCs; (2) whether APCs that take up NPs exchange these NPs with other APCs; and (3) how antigen is cross-presented by APCs through LDH-antigen complexes.

In this study, we reinvestigated the antigen cellular uptake of LDH-dye NPs by murine macrophage cells and bone-marrow dendritic cells, and examined enhancement of presentation and cross-presentation of the model antigen OVA delivered by LDH NPs. We also employed the mimicking of antigen presentation via MHC class I pathway using LDH NPs to prime T cell activation in B3Z CD8+ T hybridoma system. Our results demonstrate the possible pathways to explain how LDH-delivered antigen significantly improves the dendritic cells maturation and enhances the antigen cross-presentation on DCs' surface.

### MATERIALS AND METHODS

### Preparation of LDH, LDH-FITC, and LDH-Congo Red (LDH-CR) NPs

Mg2Al(OH)6Cl·mH2O LDH NPs were prepared as described in previous work (Xu et al., 2006a,b). In brief, 10 mL of mixed salt solution containing MgCl<sup>2</sup> 6H2O, (0.30 M) (Chem-Supply, 99.0–101.0%), with AlCl<sup>3</sup> 6H2O (0.10 M) (Scharlau, 95–101%) was poured into 40 mL of NaOH (Sodium hydroxide pellets; Ajax Finechem) solution (0.15 M) under vigorous stirring. After 10 min stirring, LDH slurry was collected and washed twice with deionized water by centrifugation (SIGMA4 <sup>R</sup> -16K Centrifuge) at 4700 rpm for 10 min. Then the slurry was manually dispersed in 40 mL of deionized water and transferred into an stainless steel autoclave with a Teflon lining (Parr Acid Digestion Vessels) for heating at 100◦C for 6 h, giving rise to a homogeneously dispersed MgAl-LDH suspension.

To make LDH-FITC NPs, ¼ of manually dispersed LDH slurry was mixed with 0.5 mL of 0.025 M FITC2<sup>−</sup> (fluorescein isothiocyanate; Sigma-Aldrich)" and shaken for 1 h, followed by separation and washing via centrifugation. The slurry was then manually dispersed in 10 mL of deionized water, which was similarly treated at 100◦C for 6 h, yielding a well dispersed LDH-FITC NP suspension.

LDH-Congo red (LDH-CR) NPs were prepared similarly. Congo-red (0.0125 M; Sigma-Aldrich) was pre-mixed with 40 mL NaOH (0.15 M) solution before adding 10 mL of mixed salt solution containing MgCl<sup>2</sup> (0.30 M) and AlCl<sup>3</sup> (0.10 M). The resultant suspension was separated and the collected slurry washed twice. Finally, the slurry was dispersed in water and treated in an autoclave at 100◦C for 14 h, yielding an LDH-CR NP suspension.

The particle size distribution of these LDH NP suspensions was measured with a dynamic light scattering (DLS) instrument (Nanosizer Nano ZS, MALVERN Instruments) to estimate the average hydrodynamic particle size and check the dispersion state.

## Cell Culture

RAW 264.7 macrophage cells (ATCC) were grown on 93 mm × 21 mm Petri dish in complete RPMI 1640 medium (Life Technologies Corporation, Australia) supplemented with 10% fetal bovine serum and adjusted to contain 100 µg/mL streptomycin and 100 units/mL penicillin, all from Invitrogen. Cell subcultures were made by scraping or mechanical isolation.

DC2.4 cells (kindly provided by A/Prof Mingnan Chen, University of Utah, United States) were grown in complete RPMI 1640 medium supplemented with 10% fetal bovine serum and adjusted to contain 1% L-glutamine, streptomycin and penicillin, all from Invitrogen.

Bone marrow dendritic cells were generated according a previous publication (Lutz et al., 1999). All animal studies were performed with adherence to the guidelines of the Animal Ethics Committee of The University of Queensland. Femurs and tibias were obtained from 6 to 8 weeks C57BL/6 female mice. Bone marrow was mashed into the single cell suspension via a 70 µm cell strainer. On the first day, 2 × 10<sup>6</sup> BM leukocytes were seeded in each Petri dish in 10 mL complete RPMI medium with 10% fetal bovine serum and 0.05 mM of 2-mercaptoethanol and adjusted to contain 1% L-glutamine, streptomycin, and penicillin. In addition, 200 ng recombinant murine granulocyte macrophage colony stimulating factor (rmGM-CSF; Sigma-Aldrich) was added as supplement. At day 3, another 200 ng rmGM-CSF in 10 mL medium was added into the dish. At days 6 and 8, half of cell supernatant was collected, and after centrifugation, cells were resuspended into 10 mL fresh medium with 200 ng rmGM-CSF, and then added into original Petri dish.

#### Cellular Uptake of LDH NPs

fphar-09-01060 September 18, 2018 Time: 19:4 # 3

After subculture of RAW 264.7 macrophage cells, 35 mm × 10 mm Nunclon cell culture dishes were used for cell growth at the density of 5 × 10<sup>5</sup> cells/mL for overnight. Then, 5 or 25 µg/mL LDH-FITC NP suspension was added into these dishes for cellular uptake. Controls were added with the same volume of PBS. Cells were cultured at 37◦C in a 5% CO<sup>2</sup> incubator and then collected at the time point of 0.5, 1, 2, 4, or 8 h. The collected cells were washed, and fixed in 4% Paraformaldehyde (PFA; Sigma-Aldrich) solution for FACS analysis (FCM, BD AccuriTM C6, BD Biosciences, San Jose, CA, United States).

Similarly, freshly obtained BMDCs were cultured in 6-well plates at the density of 1 × 10<sup>6</sup> DC/well in 1.5 mL medium containing 50 µg/mL LDH-FITC NPs. After incubation for 0.5, 1, 2, 4, or 8 h at 37◦C, BMDCs were collected for FACS analysis to determine the uptake kinetics. For dose-dependent uptake assay, BMDCs were cultured in 1.5 mL medium containing 10, 20, 50, 100, and 200 µg/mL of LDH-FITC NPs for 2 h at 37◦C, and then collected for uptake quantification using FACS. Similarly, the cellular uptake kinetics of LDH-CR NPs were also examined.

#### Cellular Exocytosis of Internalized LDH NPs

To analyze the release of internalized LDH-FITC NPs by the cells, 25 µg/ml LDH-FITC NP suspension was added into the dishes for uptake by macrophages for 2 h. Then, the cells were collected and washed with PBS to remove free LDH NPs, followed by further culture in fresh medium. The cells were finally collected and washed with PBS at the time point of 0, 0.5, 1, 2, 4, or 8 h post incubation, and fixed in 4% PFA solution for FACS analysis.

### Exchange of Internalized LDH NPs Between Macrophages

Intercellular exchange of LDH NPs between macrophage cells was examined by co-culturing two individually labeled macrophage populations. In brief, 25 µg/mL of LDH-FITC and LDH-CR NPs were separately added into RAW 264.7 macrophage cell suspensions for cellular uptake for 2 h, yielding two cell populations (each labeled with a specific dye), i.e., M8LDH−FITC and M8LDH−CR. Two cell populations were then mixed at the equal cell number and then co-cultured for 4 h in fresh medium. The co-cultured cells were collected and fixed for FACS analysis, and cell images were taken using a Carl Zeiss LSM 510 confocal laser-scanning microscope (CLSM, Carl Zeiss MicroImaging GmbH, Germany). For comparison, two cell populations, i.e., M8LDH−FITC and M8LDH−CR, were also cultured for 4 h separately in fresh medium and analyzed.

#### BMDC Maturation Induced by LDH-OVA NPs

At day 7 or 8, BMDCs were cultured in ultra-low attachment plates and pulsed with OVA (albumin from hen egg white, lyophilized powder, ≥98%, Grade VI; Sigma-Aldrich) or the equal amount of OVA in complex with LDH in RPMI 1640 medium (without GM-CSF) for 16 h. Cells were then harvested and washed, and stained with Alexa Fluor <sup>R</sup> 488 anti-mouse I-A/I-E Antibody (Clone 2G9; BioLegend) to determine dendritic cell maturation. Here LDH NPs with OVA were made by mixing them at the mass ratio of 2:1, at the concentration of 200 and 100 µg/mL, respectively.

### SIINFEKL-Antigen Presentation in DC2.4 Cells Promoted by LDH NPs

After cell internalization, OVA antigen would be enzymatically degraded into the functional epitopes. OVA H-2Kb-restricted CTL epitope (OVA257−264, SIINFEKL) would interact with MHC class I complexes, leading to the presentation of the MHC class I-functional epitope (like SIINFEKL) complex on the surface of DCs. To do this assay, DC 2.4 cells were cultured in 96 well cell culture plates, and pulsed with OVA in complex with LDH in RPMI 1640 medium (without GM-CSF) for 16 h. Then cells were harvested and washed, and stained with APC or PE anti-mouse H-2Kb of MHC class I bound to SIINFEKL antibody (Clone 25-D1.16; BioLegend) to determine the degree of epitope presentation (SIINFEKL/H-2Kb complexes) on DC2.4 cell surface.

#### B3Z CD8+ T Hybridoma Cell Activation

T cell priming can also indicate SIINFEKL epitope presented on the murine Kb MHC class I molecules (Karttunen et al., 1992). B3Z cell (kindly provided by A/Prof Mingnan Chen, University of Utah, United States), a CD8+ T-cell hybridoma, induces β-galactosidase (β-gal) production through T cell receptor interaction with SIINFEKL/H-2Kb complexes. To do this assay, DC 2.4 cells were cultured in 96 well plates at a density of 1 × 10<sup>5</sup> cells/mL, and were pulsed with OVA or LDH-OVA at a designed concentration for 16 h. After washing with PBS, the same number of B3Z cells were added to DC 2.4 cells. After 24 h co-culture, cells were washed and incubated with lysis buffer and chlorophenol red β-galactoside for 4 h. After stopping the reaction by EDTA and glycine, the OD value of the buffer was measured at 570 nm with that at 635 nm as the reference, in order to measure the activation degree of B3Z cells, which also reflects the cross-presentation amount of SIINFEKL/H-2K<sup>b</sup> complexes on DC 2.4 cells.

#### Statistical Analysis

Data presented as mean ± standard error of the mean (SEM) were analyzed by one-way analysis of variance (ANOVA), followed by multiple comparisons using Tukey's test within GraphPad Prism software. A p-value <0.05 was considered statistically significant. <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; and ∗∗∗∗p < 0.0001.

### RESULTS

### Physicochemical Features of LDH-FITC and LDH-CR NPs

Both LDH-FITC and LDH-CR NPs were well dispersed in aqueous suspensions, showing a moderate particle

size distribution (**Figures 1A–C**). The equivalent mean hydrodynamic diameter for LDH-FITC and LDH-CR was 106 and 250 nm with the polydispersity index (PDI) of 0.132 and 0.255, respectively. Most LDH-FITC NPs were distributed within a range of 40–220 nm, while LDH-CR NPs were in 60–800 nm. The larger LDH-CR NPs may result from the longer heating time in the autoclave and the slight aggregation due to the higher CR loading. The estimated FITC was 10% of the anion exchange capacity and CR was ∼20%. The higher CR loading may also facilitate the LDH-CR crystallite growth at a relatively quicker rate than the lower FITC loading (**Figures 1A,B**; Xu and Braterman, 2003). In addition, FTIR spectra and XRD patterns confirm the layered structure of LDH-FITC and LDH-CR (**Supplementary Figure S1**), with Cl<sup>−</sup> as the most abundant anion in the LDH interlayer.

Interestingly, when LDH-FITC and LDH-CR NP suspensions were mixed with culture medium separately, the average hydrodynamic particle size was increased by about 2 times (**Supplementary Figure S2**), suggesting slight aggregation caused by serum proteins through the bridging effect, as reported previously in our group (Gu et al., 2015). This slight aggregation does not severely affect the cellular uptake by immune cells, as presented shortly.

#### Immune Cell's Uptake Kinetics

The uptake kinetics of LDH-FITC NPs by immune cells (macrophages and DCs) was quantified by measuring the fluorescence intensity of each cell using the flow cytometry. As shown in **Figure 2** for macrophage uptake, the mean fluorescence intensity (MFI) was increased with the incubation time from 0.5 to 8 h at the LDH-FITC concentration of 5 and 25 µg/ml, respectively, indicating the cellular uptake is timedependent. Interestingly, at both LDH-FITC doses, MFI increase was relatively quicker in the first 4 h than in the subsequent 4 h, as previously observed for the uptake of many other cells (Xu et al., 2008b; Oh et al., 2009; Wong et al., 2010).

Relatively, the uptake amount (MFI) at the low dose of LDH-FITC NPs (5 µg/ml) is much smaller than that at the higher dose (25 µg/ml) at all incubation time points, reflecting the cellular uptake is dose-dependent. In particular, FITC-positive cells reached 85–95% just after incubation for 1–2 h at the higher dose, i.e., almost all cells took up an enough amount of

LDH-FITC in 1–2 h (**Supplementary Figure S3**) to distinguish themselves from un-treated cells. This thus indicates that the uptake of LDH-FITC NPs by macrophage cells is very rapid, and in consistence with our previous findings for other cells (Xu et al., 2008b; Musumeci et al., 2010). Similarly, LDH-CR NPs were also quickly taken up by macrophage cells (**Supplementary Figure S4**; Oh et al., 2009). The quick cellular uptake of LDH NPs can be largely attributed to the quick endosomal escape of LDH NPs during endocytosis, as reported previously (Ladewig et al., 2010; Gu et al., 2011).

As further shown in **Supplementary Figure S5**, the freshly obtained BMDCs took up LDH-FITC NPs also quickly, in a doseand time-dependent way, as reported previously for BMDCs (Li et al., 2010) and other mammalian cells (Xu et al., 2008b; Oh et al., 2009).

### No Exocytosis of Internalized LDH NPs by Macrophage Cells

Our results indicate that there were ∼90% FITC-positive macrophage cells after culture for 2 h at the LDH-FITC dose of 25 µg/ml (**Supplementary Figure S3**), as also shown as the point at 0 h in **Supplementary Figure S6**. Thus, these cells took up an essential amount of LDH-FITC NPs. After replacement of LDH-FITC containing medium with fresh medium, these LDH-FITC-positive cells were further cultured to examine whether they release the internalized LDH-FITC NPs.

As shown in **Figure 3**, the relative MFI was reduced from 100 to 72% (28% reduction) after 8 h incubation in fresh culture medium. In particular, the relative MFI decreased from 100 to 86% (14% reduction) in the first 1 h, much more quickly than in the subsequent 7 h (14%). There are a few possible factors that contribute to the reduction of LDH-FITC NPs in each cell. The first factor is cell division. The total cell number may increase by ∼30% through division after 8 h incubation supposing that the cycle time of RAW 264.7 cells is 15–20 h. We believe that the cell division would largely explain the MFI reduction during this 8 h.

The second factor is the release of the fluorescent tag (FITC) from the LDH interlayer, which may be degraded by the cell or diffuse out of the cell. If FITC release takes place in later endosome, the free FITC may also be quenched in the low pH environment. FITC release and quench may be responsible for the quick MFI reduction in the first 1 h just after cells were treated. The third factor is cellular exocytosis (release), which seems to contribute little to the MFI reduction.

Based on this test and analysis, we may conclude that macrophage cells do not obviously exocytosize the internalized LDH NPs, but keep them within the cells and passage to the next generation. As reported previously, iron oxide nanoparticles (IONPs) internalized by cells are retained within the cells and passaged to the cells in subsequent 3–4 generations (Gu et al., 2005).

### LDH NP Intercellular Exchange Between Macrophage Cells

As shown in **Figures 4A,B**, and **Supplementary Table S1**, the fluorescence intensity of LDH-FITC and LDH-CR NPtreated cells (M8LDH−FITC and M8LDH−CR) was increased from 7,300 (M8control) to 176,000 (M8LDH−FITC) and from 2,400 (M8control) to 50,000 (M8LDH−CR) after 2-h uptake and 4-h post-incubation (**Supplementary Table S1**), with ∼80% cells being fluorescence positive (**Table 1**). When M8LDH−FITC and M8LDH−CR cells were mixed in the equal number and the fluorescence intensity was quickly measured. The cytometry profile was their simple combination (**Figures 4A–C**), i.e., half of their individual positive cell percentage, i.e., 39% for each population [**Table 1**, M8LDH−FITC + M8LDH−CR (0 h) and **Figure 4C**], without obvious change for the intensity of these two cell populations (**Supplementary Table S1**).

After 4 h incubation of the mixed cells, two populations moved into the cross to close each other (**Figure 4D** and **Supplementary Figure S7**). For example, the FITC intensity of two cell populations [M8LDH−FITC + M8LDH−CR (4 h)] was 26,800/64,200, in sharp contrast to 12,400/173,000


[M8LDH−FITC + M8LDH−CR (0 h)] (**Supplementary Table S1**). Similarly, the CR intensity was 4,500/18,700 vs. 3,400/49,700, respectively (**Supplementary Table S1**). Very remarkably, there were 5.1% cells being both FITC-positive and CRpositive (**Figure 4D**, indicated with the red circle), while the percentage of only FITC-positive and only CR-positive cells was significantly reduced to 14.7 and 9.5%, respectively [**Table 1**, M8LDH−FITC + M8LDH−CR (4 h)]. The histograms indicate that M8LDH−FITC + M8LDH−CR (4 h) cells obviously shift in both FITC and Congo red channels (**Supplementary Figure S7**), and confirmed that some cells in M8LDH−FITC + M8LDH−CR (4 h) population contained both LDH-FITC and LDH-CR NPs, clearly showing that macrophage cells exchange the LDH NPs with each other. The nanoparticle exchange may occur via the possible mechanisms for the transfer of antigens between APCs, such as synapse (Mittelbrunn and Sanchez-Madrid, 2012), via tunneling nanotubes (TNT) (Domhan et al., 2011), or through gap junctions (Yewdell and Dolan, 2011).

This exchange has been also captured in the fluorescence image. As shown in **Figure 5D**, the arrow indicates that the macrophage cell has both LDH-FITC and LDH-CR NPs inside. These data thus reveal that the macrophage cells exchange the

internalized NPs with their neighbors, which appeared to occur even in suspension when the two populations of cells were just mixed and contacted within a minute, as there were 0.7% cells being already both FITC-positive and CR-positive [**Table 1**, M8LDH−FITC + M8LDH−CR (0 h)].

### BMDC Maturation Promoted by LDH-OVA

High expression of MHC class II complexes on the DC surface, i.e., DC maturation, is very critical for generation of a high level of antigen-specific antibody (Kukutsch et al., 2000). To demonstrate the maturation effect of LDH-OVA vaccine formulation, BMDCs cultured at day 7 or 8 were collected as the target DCs. LDH-OVA was formulated at the LDH:OVA mass ratio of 2:1, and BMDCs were then exposed to culture medium containing this LDH-OVA formulation for 16 h. The I-A/I-E antibody was used to distinguish two kinds of DC subpopulations, i.e., MHC II high and MHC II low (**Supplementary Figure S8**), which are representative for mature and immature DCs according to the previous study (Kukutsch et al., 2000).

Our data show that there was a significant increase in terms of mature DCs when DCs were stimulated by the LDH-OVA formulation. As shown in **Figure 6A**, the MFI of MHC II high DC population treated with LDH-OVA was significantly higher than that of the blank control and OVA only-activated DC group. Consistently, the mature DC was up to 54.2% when LDH-OVA was used to stimulate DCs, significantly higher than the control group (35.4%) and OVA-stimulated group (48.0%) (**Figure 6B**). Thus LDH NPs significantly promote the maturation of DCs, as reported previously (Li et al., 2010).

FIGURE 6 | Layered double hydroxide-ovalbumin (LDH-OVA) induces BMDC maturation. The MFI (A) and percentage (B) of MHC class II high population. The data shown are representative for three independent experiments. Data are the mean ± SEM. <sup>∗</sup>p < 0.05; ∗∗p < 0.01 (one-way ANOVA, with a Tukey's multiple comparison test).

## Enhanced Antigen Cross-Presentation Promoted by LDH NPs

As previously presented, blank LDH NPs are readily taken up by macrophage cells and DCs, which can be used to carry the target antigens and facilitate their cellular uptake, such as BSA (Chen et al., 2016) and OVA. After internalization, OVA antigen is probably dissociated from LDH-OVA particles either in later endosome or cytoplasm, and then enzymatically degraded into the functional epitope. This epitope interacts with MHC class I complexes, leading to the presentation of the MHC class I-functional epitope (like SIINFEKL) complexes on the surface of DCs.

In this research, 25-D1.16 antibody was used to specifically bind with the complex (SIINFEKL/H-2K<sup>b</sup> ) to confirm and quantify the antigen cross-presentation through the MHC class I pathway, which is necessary for inducing the formation of antitumor CTL CD8+ T cells (Burgdorf et al., 2007). As shown in **Figure 7A**, LDH-OVA vaccine significantly enhanced the presentation of SIINFEKL/MHC I complexes on the surface

of DC 2.4 in terms of the MFI, with up to 6.3% of DC 2.4 presenting complexes, in sharp contrast with nearly no antigen presented in DC 2.4 cells treated with OVA only (0.45%) and control medium (0.62%) (**Figure 7B**). The higher antigencomplex presentation on the DC surface may be largely attributed to the promoted cellular uptake and moreover, the enhanced subsequent processes, such as enzymatic degradation of OVA to epitope with the help of LDH NPs, as well as formation of MHC I-epitope complexes through the cytosolic pathway with endoplasmic reticulum (ER) or phagosomal loading (Joffre et al., 2012). Moreover, as shown in **Supplementary Figure S9**, LDH-SIINFEKL vaccine resulted in high SIINFEKL presentation by DCs which is comparable with the presentation using free SIINFEKL as the positive control. Free SIINFEKL peptide is well known to be readily loaded onto MHC class I after exogenous loading/incubation with DC's or APCs (Cho et al., 2016).

## Enhanced T Cell Priming by LDH-OVA-Treated DCs

T cell priming can also indicate whether the antigen is crosspresented on the DC surface in the form of MHC class I-epitope complexes. Thus, B3Z cell, a CD8+ T-cell hybridoma, was chosen

to specifically recognize SIINFEKL epitope presented on the murine Kb MHC class I molecules (Karttunen et al., 1992) and determine the T cell priming extent. The priming of the SIINFEKL epitope to B3Z cells induces β-galactosidase (β-gal) synthesis by B3Z cells. The induced β-gal amount thus quantifies the cross-presentation of SIINFEKL/H-2K<sup>b</sup> complexes on DC cells and the degree of T cell activation. In this research, an equal number of B3Z and DCs (LDH-OVA stimulated) were co-cultured for 24 h, and the OD value, i.e., the β-gal amount produced by B3Z, was measured in a plate reader. As shown in **Figure 8**, the OD value of lyzed B3Z cells that were cocultured with LDH-OVA pulsed DCs was significantly higher than that of other two control groups, indicating that B3Z cells were significantly activated and further confirming that the OVA epitope was successfully cross-presented on DC 2.4 cells via LDH NP adjuvants.

#### DISCUSSION

Layered double hydroxide nanomaterials are reported to significantly promote the immune responses in mice model and show a high promise as effective nano-adjuvants (Li et al., 2011; Wang et al., 2014; Williams et al., 2014). In particular, our groups have demonstrated that LDH NPs are able to induce both highlevel antibody and cellular immune responses for antibacterial and anti-tumor treatment (Yan et al., 2014; Chen et al., 2016). Apparently, the adjuvanticity of LDH NPs is related to the particle size and composition, the mass ratio of LDH: antigen, and the dose injected, which have been reported (Williams et al., 2014; Yan et al., 2014; Chen et al., 2016). In principle, the activity is largely determined by the effects of LDH NPs on the biological processes of immune cells, including long-term stimulation (depot effect), cellular uptake, APC maturation, antigen processes within APCs and the antigen presentation on the APC surface, as well as the activation of target T and B cells, which have been well investigated in the current research, together with our previous work (Yan et al., 2014; Chen et al., 2016).

When LDH-antigen NPs are injected subcutaneously, these NPs form a loosely aggregated lump, which then causes a socalled depot effect, i.e., long-term stimulation. For example, the lump of LDH NP-adjuvanted vaccine was found to last for ∼1 month beneath the skin, giving a higher and sustained level of specific antibody (Chen et al., 2016, 2018). Thus, the depot effect is beneficial to the long-term immune responses (Mckee et al., 2007; Henriksen-Lacey et al., 2010). Moreover, the lump recruits many inflammatory cells (Aimanianda et al., 2009; Chen et al., 2018), thus LDH-antigen NPs on the lump surface can be readily taken up by or facilitate the delivery of antigen (such as OVA) to these immune cells. The current research has further confirmed that LDH NPs are readily taken up by macrophage cells (**Figure 2**) and BMDCs (**Supplementary Figure S5**), which is also supported by previous reports (Li et al., 2010; Wang et al., 2014). As revealed elsewhere, this facilitation to cellular uptake results from the quick endosome escape (Choy et al., 2004; Xu et al., 2008b). As weakly alkaline LDH NPs are partly dissolved in the slightly acidic endosome, so the ion concentration increases and the enhanced osmotic pressure inside the endosome leads to water influx and bursts the endosome, releasing the LDH NPs into the cytoplasm (Xu et al., 2008b; Gu et al., 2011). Therefore, antigen is mostly associated with LDH NPs after endosome escape and then possibly processed to load with MHC I molecules through the cross-presentation pathway. This process is very much different from the case using polymeric nanoparticles to adjuvant antigens through lysosomal pathway (Lai et al., 2007, 2008; Fernando et al., 2010).

The most potent APCs are immunologically competent dendritic cells (DCs), while their ability to regulate immunity is dependent on their maturation (Banchereau and Steinman, 1998). After the LDH-antigen vaccine is injected subcutaneously, immature DCs are recruited to the site of inflammation in peripheral tissues, and take up LDH-antigen NPs (Waeckerle-Men et al., 2004; Chen et al., 2016, 2018). In this research, we found that LDH NPs assist OVA to mature DCs by promoting significantly more MHC II complexes on the DCs' surface (**Figure 6**), in coordination with the activation signals received from the surrounding cytokines and chemokines, costimulatory molecules and proteases (Li et al., 2010; Williams et al., 2014). On the other hand, the exogenous OVA antigens are quickly delivered to cytoplasm by LDH NPs, and processed into epitopes for complexing with MHC I molecules (cross-presentation), which is also benefited from LDH's quick endosome escape

with antigen processing (TAP), epitopes are transferred into endoplasmic reticulum (ER), where they are loaded with MHC class I molecules (Step E). Afterward, these complexes are transferred to exocytic vesicle by Golgi apparatus (Step F). Finally, MHC class I complexes present these epitopes on the surface of antigen presentation cells (Step G). (Right) Generally exogenous antigens are intracellular internalization in endocytic vesicles, such as acidic intracellular compartments. Antigens are degraded into epitopes (Step H). Then endocytic vesicles fuse with lysosome, where epitopes are loaded with MHC class II molecules (Step J). Note that MHC class II molecules are made inside the ER and transferred to lysosome by Golgi apparatus (Step I). Finally, these epitopes are presented with MHC class II complexes on the cell surface (Step L) by exocytic vesicles (Step K).

(as schematically shown in **Figure 9**). Meanwhile, LDH-antigen NPs could also attract the proteasome and enzymatic proteases. These proteases on the LDH surface may more efficiently process the adjacent antigens into antigenic epitope. Moreover, shortchain epitope is more easily released from the LDH surface, which may quickly form more epitope-loaded MHC class II and I complexes and their subsequent (cross)-presentation. This postulated mechanism may be supported by the enhanced MHC class II high population and more SIINFEKL/MHC I complexes in LDH-OVA group (**Figures 6**, **7**), respectively. Therefore, the mature DCs have significantly enhanced presentation of antigenloaded MHC class I complexes on the cell surface upon the stimulation of the LDH-antigen vaccine.

A more interesting issue is that immune cells (such as APCs) that take up LDH-antigen NPs may exchange these NPs with surrounding immature cells (including macrophages and DCs) at the site of injection, during the circulation in blood/lymph systems, and in the lymphoid nodes. As demonstrated in this research, macrophage cells exchanged their internalized LDH NPs with each other during in vitro culture (**Figures 4**, **5**). In such a way, DCs that take up LDH-antigen NPs at the injection/inflammatory site may transfer these LDH NPs to the surrounding immature DCs, thus "infect" and activate these immature DCs to mature (epitope/MHC class II presentation) and present epitope/MHC class I complexes (cross-presentation). As reported recently, cellular communication by exchanging materials may occur via synapse (Mittelbrunn and Sanchez-Madrid, 2012), tunneling nanotubes (TNT) (Domhan et al., 2011), or gap junctions (Yewdell and Dolan, 2011). Subsequently, this exchange induces a high level of specific antibody by stimulating B cells and activates more potent cytotoxic T cells for cell-mediated immune response, as reported in our previous paper (Yan et al., 2014) and this research (**Figure 8**), leading to the remarkable improvement in the immune responses.

#### CONCLUSION

In summary, we report that APCs (such as macrophages and DCs) can take up LDH NPs efficiently, and more

### REFERENCES


significantly macrophages exchange the internalized LDH NPs with surrounding ones. We also report that the internalized LDH-antigen NPs can significantly facilitate the maturation of immature DCs and enhance the antigen cross-presentation of MHC I complexes on the DC surface. The high adjuvanticity of LDH NPs may be attributed to specific properties of LDH materials, such as the weak alkalinity for endosome escape and capability of co-adsorbing enzymes on the surface for enzymatic degradation. These findings may provide some guidelines for design new adjuvants for next generation vaccines.

### AUTHOR CONTRIBUTIONS

ZX and LL designed the current experiments in consultation with WG and BR. SY conducted most of the experiments and collected and analyzed the data. KX assisted in the experiments and data collection. SY and ZX wrote the manuscript. LL, KX, WG, and BR contributed to the revisions of the manuscript.

## FUNDING

This work was financially supported by Australian Research Council (ARC) DP grant (DP170104643) and AIBN Group Leader PhD Living Allowance Scholarship.

### ACKNOWLEDGMENTS

The authors acknowledged the facilities and the assistance of Queensland Node of the Australian National Fabrication Facility (ANFF-Q), the University of Queensland.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar. 2018.01060/full#supplementary-material


Tunneling nano-tube like structures in primary human renal epithelial cells. PLoS One 6:e21283. doi: 10.1371/journal.pone.0021283


fphar-09-01060 September 18, 2018 Time: 19:4 # 11

anti-tumour vaccine formulations. Dalton Trans. 47, 2956–2964. doi: 10.1039/ c7dt03725b


**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 Yan, Xu, Li, Gu, Rolfe and Xu. 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.