# THE PHYSIOLOGY OF INFLAMMATION – THE FINAL COMMON PATHWAY TO DISEASE

EDITED BY : Alexandrina Ferreira Mendes, Maria Teresa Cruz and Oreste Gualillo PUBLISHED IN : Frontiers in Physiology

#### 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 978-2-88945-729-8 DOI 10.3389/978-2-88945-729-8

### 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

# THE PHYSIOLOGY OF INFLAMMATION – THE FINAL COMMON PATHWAY TO DISEASE

Topic Editors:

Alexandrina Ferreira Mendes, Faculty of Pharmacy and Center for Neuroscience and Cell Biology, University of Coimbra, Portugal Maria Teresa Cruz, Faculty of Pharmacy and Center for Neuroscience and Cell Biology, University of Coimbra, Portugal Oreste Gualillo, the NEIRID Group (Neuroendocrine Interactions in Rheumatology and Inflammatory Diseases), Servizo Galego de Saude and Instituto de Investigación Sanitaria de Santiago de Compostela (IDIS), Santiago University Clinical Hospital, Spain

Chronic diseases are increasingly recognized as involving low grade inflammation, that is, a self-perpetuating tissue response to stress caused by exogenous or endogenous triggers, that progressively evokes danger-associated molecular pattern release, ultimately driving tissue damage and loss of function. This response is frequently unapparent clinically, thus the designation "low grade". This eBook comprises nineteen reviews and original articles that provide the most updated knowledge on the causes and roles of this inflammatory response in a variety of diseases and conditions. The editorial that precedes these articles not only summarizes each one, but provides a broader interpretation of the role of inflammation in health and a variety of disease conditions, the underlying mechanisms and the targets more promising for therapy. Finally, it also highlights the most relevant and emerging research topics that are already shaping future directions for the development of more fine-tuned and innovative therapies.

Citation: Mendes, A. F., Cruz, M. T., Gualillo, O., eds. (2019). The Physiology of Inflammation – The Final Common Pathway to Disease. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-729-8

# Table of Contents

*06 Editorial: The Physiology of Inflammation—The Final Common Pathway to Disease*

Alexandrina Ferreira Mendes, Maria Teresa Cruz and Oreste Gualillo

# CHAPTER 1

# INFLAMMATION, INTESTINAL MICROBIOTA AND DISEASE


Rosaria Meli, Claudio Pirozzi and Alessandra Pelagalli

*40 L-Threonine Supplementation During Colitis Onset Delays Disease Recovery*

Joana Gaifem, Luís G. Gonçalves, Ricardo J. Dinis-Oliveira, Cristina Cunha, Agostinho Carvalho, Egídio Torrado, Fernando Rodrigues, Margarida Saraiva, António G. Castro and Ricardo Silvestre

*47 Could Sodium Chloride be an Environmental Trigger for Immune-Mediated Diseases? An Overview of the Experimental and Clinical Evidence* Eric Toussirot, Matthieu Béreau, Charline Vauchy and Philippe Saas

# CHAPTER 2

# ROLE OF THE INTERPLAY BETWEEN METABOLISM AND INFLAMMATION IN HEALTH AND DISEASE

*57 Role of the Inflammation-Autophagy-Senescence Integrative Network in Osteoarthritis*

Claire Vinatier, Eduardo Domínguez, Jerome Guicheux and Beatriz Caramés

*82 Obesity, Metabolic Syndrome, and Musculoskeletal Disease: Common Inflammatory Pathways Suggest a Central Role for Loss of Muscle Integrity*

Kelsey H. Collins, Walter Herzog, Graham Z. MacDonald, Raylene A. Reimer, Jaqueline L. Rios, Ian C. Smith, Ronald F. Zernicke and David A. Hart

*107 Human Milk and Donkey Milk, Compared to Cow Milk, Reduce Inflammatory Mediators and Modulate Glucose and Lipid Metabolism, Acting on Mitochondrial Function and Oleylethanolamide Levels in Rat Skeletal Muscle*

Giovanna Trinchese, Gina Cavaliere, Chiara De Filippo, Serena Aceto, Marina Prisco, Jong Tai Chun, Eduardo Penna, Rossella Negri, Laura Muredda, Andrea Demurtas, Sebastiano Banni, Roberto Berni-Canani, Giuseppina Mattace Raso, Antonio Calignano, Rosaria Meli, Luigi Greco, Marianna Crispino and Maria P. Mollica

# *122 Obesity, Fat Mass and Immune System: Role for Leptin*

Vera Francisco, Jesús Pino, Victor Campos-Cabaleiro, Clara Ruiz-Fernández, Antonio Mera, Miguel A. Gonzalez-Gay, Rodolfo Gómez and Oreste Gualillo

# *142 Relaxin-2 in Cardiometabolic Diseases: Mechanisms of Action and Future Perspectives*

Sandra Feijóo-Bandín, Alana Aragón-Herrera, Diego Rodríguez-Penas, Manuel Portolés, Esther Roselló-Lletí, Miguel Rivera, José R. González-Juanatey and Francisca Lago

# CHAPTER 3

# INFLAMMATION AND CANCER


Sheng Li, Xiaoping Liu, Tongzu Liu, Xiangyu Meng, Xiaohong Yin, Cheng Fang, Di Huang, Yue Cao, Hong Weng, Xiantao Zeng and Xinghuan Wang

# CHAPTER 4

# INFLAMMATION, ENDOTHELIAL DYSFUNCTION AND MACROPHAGES IN DEGENERATIVE DISEASES

*185 The Evaluation of Flow-Mediated Vasodilation in the Brachial Artery Correlates With Endothelial Dysfunction Evaluated by Nitric Oxide Synthase Metabolites in Marfan Syndrome Patients*

Oscar Lomelí, Israel Pérez-Torres, Ricardo Márquez, Sergio Críales, Ana M. Mejía, Claudia Chiney, Enrique Hernández-Lemus and Maria E. Soto

*195 Lipid and Non-lipid Factors Affecting Macrophage Dysfunction and Inflammation in Atherosclerosis*

Mark S. Gibson, Neuza Domingues and Otilia V. Vieira

*213 Macrophage Depletion Lowered Blood Pressure and Attenuated Hypertensive Renal Injury and Fibrosis*

Lei Huang, Aimei Wang, Yun Hao, Weihong Li, Chang Liu, Zhihang Yang, Feng Zheng and Ming-Sheng Zhou

# CHAPTER 5

# NEW TARGETS FOR CHRONIC DISEASES: THE ROLE OF TOLL-LIKE RECEPTORS AND THE THERAPEUTIC POTENTIAL OF MESENCHYMAL STEM CELLS

*224 Paracrine Anti-inflammatory Effects of Adipose Tissue-Derived Mesenchymal Stem Cells in Human Monocytes*

Maria I. Guillén, Julia Platas, María D. Pérez del Caz, Vicente Mirabet and Maria J. Alcaraz


# Editorial: The Physiology of Inflammation—The Final Common Pathway to Disease

#### Alexandrina Ferreira Mendes 1,2 \*, Maria Teresa Cruz 1,2 and Oreste Gualillo3,4

<sup>1</sup> Center for Neuroscience and Cell Biology, University of Coimbra, Coimbra, Portugal, <sup>2</sup> Faculty of Pharmacy, University of Coimbra, Coimbra, Portugal, <sup>3</sup> The NEIRID Group (Neuroendocrine Interactions in Rheumatology and Inflammatory Diseases), Servizo Galego de Saude, Santiago de Compostela, Spain, <sup>4</sup> Instituto de Investigación Sanitaria de Santiago de Compostela (IDIS), Santiago University Clinical Hospital, Santiago de Compostela, Spain

Keywords: aging, low grade inflammation, chronic diseases, nutrition, microbiota

**Editorial on the Research Topic**

#### **The Physiology of Inflammation—The Final Common Pathway to Disease**

The 19 original and review articles included in the research topic "Physiology of Inflammation- the common pathway to disease" collectively convey an updated perspective on a variety of physiologic and pathologic conditions that trigger inflammation and how this response develops and progresses affecting tissue functions, either contributing to protect the organism or, instead, being a damaging process that disrupts homeostasis and drives most chronic diseases. While acute inflammation is usually a self-limited process, chronic inflammation, excluding that associated with autoimmune diseases, is characterized by a low-grade persistent inflammatory response that frequently is not clinically evident, but can be detected by the presence of increased levels of cytokines, chemokines, prostaglandins, nitric oxide, proteases, and other inflammatory mediators both at the tissue and plasma levels (Minihane et al., 2015). Accumulating evidence associates low grade inflammation with aging and the development and progression of most chronic diseases, from metabolic disturbances, like diabetes mellitus, obesity and the metabolic syndrome, to neurodegenerative, musculoskeletal, renal, cardiovascular diseases, and even behavioral diseases, among others (Scarpellini and Tack, 2012; Zhu et al., 2014; Nefla et al., 2016; Mihai et al., 2018; Speer et al., 2018). This low grade inflammation not only contributes to the morbidity and mortality associated with chronic diseases, as it can also impact health status in apparently healthy people. In this regard, a recent epidemiologic study involving more than 20,000 individuals found a significant increased risk in overall mortality in individuals with high low grade inflammation scores relative to those with lower scores, independently of possible confounders, including presence of chronic diseases and a number of health-related behaviors (Bonaccio et al., 2016). Moreover, low grade inflammation has also been implicated as critical determinant of chronic fatigue, either associated with cancer or constituting the chronic fatigue syndrome (Lacourt et al., 2018).

Therefore, understanding the role and mechanisms of low grade inflammation both as a contributor to those chronic diseases as well as to the overall health status is essential to provide clues for the development of innovative more efficient therapeutic and preventive strategies. The papers included in this topic contribute to these goals by providing clues for a better understanding of those roles and mechanisms, also highlighting specific issues pertaining to distinct conditions that are relevant not only to understand the underlying mechanisms, as for identification of specific targets and development of more finely-tuned therapies that address specific components and drivers of each condition. The papers included in this research topic identify relevant questions that need to be answered to achieve that goal and propose directions for future research, as well as potential solutions for, at least, some of those conditions.

#### Edited by:

Geoffrey A. Head, Baker Heart and Diabetes Institute, Australia

> \*Correspondence: Alexandrina Ferreira Mendes afmendes@ff.uc.pt

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 25 October 2018 Accepted: 19 November 2018 Published: 04 December 2018

#### Citation:

Mendes AF, Cruz MT and Gualillo O (2018) Editorial: The Physiology of Inflammation—The Final Common Pathway to Disease. Front. Physiol. 9:1741. doi: 10.3389/fphys.2018.01741

One of the most recent advances in understanding the mechanisms that drive inflammation in chronic diseases is the recognition of the role of the intestinal microbiota. While much remains to be elucidated, five papers in this research topic address this subject from different perspectives. Cristiano et al. highlight the role of the intestinal microbiota and factors that modulate it, both during development and after birth, in shaping brain neuronal circuits critical for the regulation of behavior through life. The mechanisms involved are beginning to be unveiled highlighting new avenues for prevention and therapy of behavioral disorders, namely autism spectrum disorders. Adding to this, Cardoso and Empadinhas suggest that Parkinson's disease can be elicited by toxic metabolic products produced by an altered gut microbiota that, by causing mitochondrial damage, activate the neuronal innate immune system triggering inflammation which, in turn, causes neurodegeneration and, thus, can also contribute to other neurodegenerative diseases.

The articles by Meli et al; Gaifem et al., and Toussirot et al. point out aquaporins and the dietary consumption of threonine and sodium chloride as relevant modulators of intestinal inflammation whose pharmacological or nutritional modulation can impact not only intestinal diseases, like inflammatory bowel disease, but also systemic ones, including rheumatoid arthritis, and multiple sclerosis. Interestingly, such interventions affect the innate and adaptive immune systems, thus likely interfering not only with their functions in general, but also with their ability to respond to normal and altered microbiota, which together can impact on the pathogenesis of several chronic diseases.

Another emerging area of research focus on the relationships between metabolism and inflammation. In this issue, five papers address those relationships, four of them focusing on the contribution of the metabolic syndrome and other metabolic imbalances to the activation of inflammatory pathways that drive the development and progression of many musculoskeletal diseases, from osteoarthritis and sarcopenia to autoimmune diseases, like rheumatoid arthritis. The interplay between inflammation and metabolic imbalances and its impact on important homeostatic mechanisms, like the intrinsic circadian rhythm, autophagy and cell senescence, in cells of the musculoskeletal system is presented by Vinatier et al. as a critical determinant of musculoskeletal diseases and of osteoarthritis in particular. Importantly, this paper also highlights requirements of pre-clinical models to improve their validity and translational value. Collins et al. in turn, discuss the central role of skeletal muscle damage by obesity-induced inflammation in disturbing the whole musculoskeletal system and driving its associated disorders, proposing new directions for clinical risk assessment and management of patients with metabolic syndrome, as well as identifying research opportunities. Trinchese et al. evaluated the anti-inflammatory and metabolic effects of milk from different species, concluding that human and donkey milk, but not cow milk, have significant anti-inflammatory effects and improve lipid metabolism both in the liver and skeletal muscle, raising important questions as to the nutritional effects of dairy cow products. Further adding to this subject, Francisco et al. focused on the role of leptin in linking the immune system and the adipose tissue, especially in obesity, to facilitate the development of autoimmune and other inflammatory rheumatic diseases, highlighting potential therapeutic targets. Feijóo-Bandín et al. focused on a similar subject, but addressing the role of another hormone, relaxin, in modulating inflammation, and metabolism at the cardiovascular level, highlighting its potential value as a therapeutic target for cardiovascular diseases.

Still focusing on the musculoskeletal system, Pérez-Baos et al. review the mechanisms involved in systemic inflammationassociated sarcopenia, critically describing the experimental models currently in use, their triggers and mediators, with special emphasis on the role and potential as a therapeutic target of the Janus Kinase/Signal Transducer and Activator of Transcription (JAK/STAT) signaling pathway. In a correlated area, Voltarelli et al. show that subcutaneous implantation of a melanoma tumor in mice induces the production of inflammatory mediators that cause a catabolic state driving muscle loss and decreased locomotor activity. These findings contribute to explain cancerassociated cachexia, thus highlighting new therapeutic targets with relevant clinical implications. Also pertaining to cancer, the paper by Li et al. and collaborators identifies a set of genes whose lower expression in bladder cancer patients correlates with poor survival, suggesting their potential use as prognostic biomarkers. Interestingly, some of those genes are regulated by inflammatory pathways, suggesting that modulation of inflammation in bladder cancer patients may improve overall survival. Nonetheless, further studies are required to confirm this hypothesis.

Inflammation increasingly emerges as an important component of many other diseases classically envisaged as degenerative or metabolic. In this issue, three such examples are provided by the papers by Lomelí et al; Gibson et al., and Huang et al. The first shows that some inflammatory mediators, namely VCAM-1 and nitric oxide metabolites, are correlated with the degree of endothelial dysfunction found in patients with Marfan Syndrome, a connective tissue disease most frequently associated with mutations in the fibrillin-1 gene. Interestingly, that correlation seems to precede structural changes, being pointed out as a potential marker for earlier diagnosis and therapeutic intervention. Also considering the role and causes of inflammation in atherogenesis and endothelial dysfunction, although in a different context, Gibson et al. provide an up to date review on the role of individual lipids and lipid metabolites in modulating macrophage phenotype and consequently their inflammatory status. Importantly, this review identifies causes of variability in different studies on the role of lipids on macrophage functions and phenotype directly impacting endothelial dysfunction and proposes new experimental approaches to minimize this variability and improve the translational value of resulting knowledge. Finally, Huang et al. provide evidence for a crucial role of macrophages in the development of hypertensive renal injury and fibrosis which in turn further aggravates hypertension. Renal macrophages thus emerge as potential targets for the development of novel therapies to reduce hypertensive renal injury, which may also contribute to control hypertension itself.

Starting from a different point of view, but likewise aiming at identifying strategies to reduce inflammation, Guillén et al. show that conditioned medium from human adipose tissuederived mesenchymal stem cells (ASC) has anti-inflammatory properties reducing the production of inflammatory mediators and promoting a non-inflammatory phenotype of human macrophages. Further studies to determine whether ASC from obese patients retain these anti-inflammatory properties are needed, in as much as studies to demonstrate which component(s) of the ASC secretome are endowed with these properties and their potential therapeutic efficacy in inflammation-associated diseases.

Approaching another critical component of inflammation, the article by Alonso-Pérez et al. focuses on the role of Toll-Like Receptors (TLR), particularly TLR4, in modulating anabolic and catabolic processes, many of which are activated by inflammatory mediators, like TLR agonists, in osteoblasts, osteocytes, and mesenchymal stem cells. Finally, the paper by Garcia-Rodriguez et al. further explores this subject by reviewing current knowledge on the role of TLRs on the chronic low-grade inflammation and osteogenic responses associated with calcific aortic valve disease. In this regard, the similarities between modulation of bone and valve interstitial cell functions by endogenous TLR agonists are striking, suggesting that inhibition of inflammatory responses through modulation of these receptors may have multiple effects and therapeutic applications.

With this brief summary, we expect to give the readers a broad perspective on the subjects covered by this research topic highlighting the role of inflammation in health and a variety of disease conditions, the underlying mechanisms and the targets more promising for therapy. Moreover, we also attempted to summarize the most relevant and innovative ideas that may help shape future research. Nonetheless, many other relevant and emerging aspects could have been focused. Among those, the interplay between the circadian rhythm and inflammation and the mutual regulation of clock and inflammatory/antiinflammatory genes, both in peripheral organs and tissues and in the hypothalamic suprachiasmatic nucleus, the central circadian clock regulator, are critically involved in disease initiation and progression (Geiger et al., 2015; Man et al., 2016) and can

# REFERENCES


have profound effects in the efficacy and safety of therapeutic interventions for many chronic diseases. Elucidating those interactions and their consequences is thus an emerging and exciting area of research with a foreseeable huge impact on disease management.

We hope the readers will enjoy reading these papers as much as we enjoyed editing them and we sincerely thank all contributors, authors and reviewers, for their dynamic participation and commitment that made possible this outstanding research topic.

# AUTHOR CONTRIBUTIONS

AM drafted the manuscript. AM, TC, and OG commented on the manuscript and approved the final version.

# FUNDING

AM is beneficiary of the projects Healthy Aging 2020: CENTRO-01-0145-FEDER-000012 and POCI-01-0145-FEDER-028424 supported by Programa Operacional Competitividade e Internacionalização (COMPETE 2020) through FEDER and by the Portuguese Foundation for Science and Technology. MC is beneficiary of the projects HealthyAging2020: CENTRO-01- 0145-FEDER-000012 and CENTRO-01-0145-FEDER-029369. OG is Staff Personnel of Xunta de Galicia (Servizo Galego de Saude, SERGAS) through a research-staff stabilization contract (ISCIII/SERGAS). OG is members of RETICS Programme, RD16/0012/0014 (RIER: Red de Investigación en Inflamación y Enfermedades Reumáticas) via Instituto de Salud Carlos III (ISCIII) and ERDF (FEDER). The research work of OG is funded by Instituto de Salud Carlos III and ERDF (FEDER) (PI17/00409). OG is a beneficiary of the project funded by Research Executive Agency of the European Union in the framework of MSCA-RISE Action of the H2020 Programme, Project 734899 – Olive-Net.


**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 Mendes, Cruz and Gualillo. 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.

# Interplay Between Peripheral and Central Inflammation in Autism Spectrum Disorders: Possible Nutritional and Therapeutic Strategies

Claudia Cristiano<sup>1</sup> , Adriano Lama<sup>1</sup> , Francesca Lembo<sup>1</sup> , Maria P. Mollica<sup>2</sup> \*, Antonio Calignano<sup>1</sup> and Giuseppina Mattace Raso<sup>1</sup> \*

<sup>1</sup> Department of Pharmacy, School of Medicine, University of Naples Federico II, Naples, Italy, <sup>2</sup> Department of Biology, University of Naples Federico II, Naples, Italy

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

Anne Regnier-Vigouroux, Johannes Gutenberg-Universität Mainz, Germany Kwang-Mook Jung, University of California, Irvine, United States

#### \*Correspondence:

Maria P. Mollica mariapia.mollica@unina.it Giuseppina Mattace Raso mattace@unina.it

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 15 January 2018 Accepted: 20 February 2018 Published: 07 March 2018

#### Citation:

Cristiano C, Lama A, Lembo F, Mollica MP, Calignano A and Mattace Raso G (2018) Interplay Between Peripheral and Central Inflammation in Autism Spectrum Disorders: Possible Nutritional and Therapeutic Strategies. Front. Physiol. 9:184. doi: 10.3389/fphys.2018.00184 Pre- and post-natal factors can affect brain development and function, impacting health outcomes with particular relevance to neurodevelopmental diseases, such as autism spectrum disorders (ASDs). Maternal obesity and its associated complications have been related to the increased risk of ASDs in offspring. Indeed, animals exposed to maternal obesity or high fat diets are prone to social communication impairment and repetitive behavior, the hallmarks of autism. During development, fatty acids and sugars, as well as satiety hormones, like insulin and leptin, and inflammatory factors related to obesity-induced low grade inflammation, could play a role in the impairment of neuroendocrine system and brain neuronal circuits regulating behavior in offspring. On the other side, post-natal factors, such as mode of delivery, stress, diet, or antibiotic treatment are associated to a modification of gut microbiota composition, perturbing microbiota-gut-brain axis. Indeed, the interplay between the gastrointestinal tract and the central nervous system not only occurs through neural, hormonal, and immune pathways, but also through microbe-derived metabolic products. The modification of unhealthy perinatal and postnatal environment, manipulation of gut microbiota, nutritional, and dietary interventions could represent possible strategies in preventing or limiting ASDs, through targeting inflammatory process and gut microbiota.

Keywords: ASDs, inflammation, obesity, gut-brain axis, therapeutics

# INTRODUCTION

Over the past decades, the frequency of autism diagnoses has been steadily climbing and has increased the interest of the scientific community (Fombonne, 2009; Boyle et al., 2011; Elsabbagh et al., 2012). Together with Asperger syndrome, Pervasive Developmental Disorder-Not Otherwise Specified and childhood disintegrative disorder, autism is one of the so-called autism spectrum disorders (ASDs) according to the Diagnostic and Statistical Manual of Mental Disorders version 5. Generally, symptoms start at the age of 8 months, become reliably diagnosed around 2 or 3 years and then continue during the entire life (Rapin and Tuchman, 2008). ASD is defined by impairments in social communication, repetitive behaviors, and restricted interests. These variety of symptoms and their different severity, indicate the involvement not only of genetic factors (Abrahams and Geschwind, 2010) in its etiology, but also that of environmental factors (Fujiwara et al., 2016), which have different effects across individuals.

Growing evidences suggest that individuals with ASDs present prominent activation of microglia and astrocytes and severe chronic inflammation (Vargas et al., 2005), both in the periphery and in multiple brain areas, characterized by increased levels of tumor necrosis factor (TNF)α (Al-Ayadhi, 2005; Chez et al., 2007), interleukin (IL)-1 (Al-Ayadhi, 2005), IL-6 (Wei et al., 2011), and interferon-γ (El-Ansary and Al-Ayadhi, 2012).

Exposure to detrimental factors during early brain development can impact its structure and function and increases long-term susceptibility to neurodevelopmental disorders (Bale et al., 2010). Obesity during pregnancy has been suggested to induce fetal brain inflammation (Graf et al., 2016), consistently nutritional composition during pregnancy is another impacting factor on brain functions: for example, polyunsaturated fatty acids (PUFAs) deprivation during pregnancy reduced learning and memory in the progeny, that restored its functions when diet was properly supplemented (Lozada et al., 2017).

In addition, recent studies show that ASD is associated with gastrointestinal (GI) ailments and changes in microbiota composition (van De Sande et al., 2014; Rosenfeld, 2015). Thus, several strategies are proposed to reduce food-related effects in ASD, as gluten- and casein-free diet, elimination of complex carbohydrates and their replacement with monosaccharides, and micronutrients intake, all interventions convergent to correct neurogenesis and neuro-network development. In addition, the use of probiotics, prebiotics, and post-biotics to ameliorate the dysbiosis-associated GI and ASD symptoms has been also proposed.

Now, it is noteworthy to better analyze the impact of all these factors on the behavior and physiology of future generations, also thank to the support of the wide range of animal models resembling human pathophysiology of ASDs (Chadman, 2017) and whether efficacious intervention strategies can be optimized. Evidence associating ASDs to these different conditions will be better analyzed in the following sections.

# MATERNAL OBESITY AND ASDs

Different studies have recently investigated critical factors during maternal pregnancy that may influence and cause the ASD development in the offspring. Among plausible maternal stressors, Zerbo et al. (2013) underline the impact of maternal fever; other studies evidence that viral and bacterial infections together with maternal gut microbiome alteration are associated with the development of ASD in offspring, through the activation of the maternal immune system (Lee et al., 2015; Choi et al., 2016; Mahic et al., 2017).

Moreover, the impact of obesity during pregnancy on cognitive function and development of CNS disorders in offspring has recently been investigated (Veena et al., 2016; Edlow, 2017). In particular, several findings indicate a link between maternal obesity and the increased risk of developing ASD in children (Leonard et al., 2006; Hamlyn et al., 2013).

Epidemiologic studies provide evidence that obesity and its complications are increasingly rising in western countries (Flegal et al., 2012; Ng et al., 2014), and with it, the rates of obesity during pregnancy (Kim et al., 2007; Heslehurst et al., 2010; Gregor et al., 2016).

In particular, obesity is often a comorbidity for some ailments, as diabetes, cancer, and cardiovascular diseases (Guh et al., 2009). It is associated with pre-natal complications, such as gestational diabetes (Solomon et al., 1997), pre-eclampsia (Bodnar et al., 2005), hypertension (Magriples et al., 2013), and placental dysfunction (Hastie and Lappas, 2014), significantly impacting mother health and embryonic and fetal growth. Moreover, maternal obesity could also induce post- partum complications (Huda et al., 2010; Siega-Riz and Gray, 2013; Moussa et al., 2016) and even long-lasting detrimental effects on offspring metabolism, organ and brain development, increasing the risk of neurodevelopmental disorders (Misiak et al., 2012). However, studies investigating this issue have shown divergent and unclear results.

In details, Dodds et al. (2011) found that mothers with a prepregnancy weight exceeding 90 kg had an increased possibility of develop autistic children. Accordingly, Krakowiak et al. (2012) demonstrated a clear relationship between obesity in pregnancy and the occurrence of ASD in their offspring. Similar findings were shown by Bilder et al. (2013) and Reynolds et al. (2014), which also linked maternal obesity to a delay in language skills. More recently, this correlation was also found in offspring of obese women gaining excessive weight during pregnancy (Li et al., 2016).

Other studies have shown a weak direct, but strong indirect relationship between maternal obesity and ASD in the offspring, particularly when maternal obesity was associated with a very low birth weight of new-borns which had two-fold risk to be autistic compared to normal weight new-borns (Pinto-Martin et al., 2011; Abel et al., 2013; Moss and Chugani, 2014). Moreover, ASD is also developed in children born prematurely (Limperopoulos et al., 2008; Abel et al., 2013; Moss and Chugani, 2014). These findings suggest that maternal obesity can raise the incidence of many negative elements, which in turn enhance the risk of developing ASD.

Notably, children born from mothers who were underweight appear to be at greater risk as well of developing ASDs (Getz et al., 2016). Last year, in a study on Danish population, Andersen et al. (2017), reported that, the risk of ASD in the offspring was observed both in underweight [body max index (BMI) < 18.5] and obese (30 ≤ BMI < 35) mothers compared to normal weight (18.5 ≤ BMI < 25) ones. Thus, both conditions may be associated with an increased risk for ASD.

However, although these studies find a clear association, Rivera et al. (2015), underline that they suffer of methodological limitations (Krakowiak et al., 2012; Hinkle et al., 2013; Moss and Chugani, 2014), and deficit in statistical analysis (Stein et al., 2006). Accordingly, Gardner et al. (2015) did not report in their study a clear relationship between obesity in pregnancy and the occurrence of ASD. Considering the importance of these findings, it is critical to also examine the mechanisms by which maternal obesity induced changes in offspring behavior and brain (Van Lieshout et al., 2011; Van Lieshout, 2013).

Several findings from human and animal studies indicate that the increase in inflammatory cytokines, and the dysregulation of nutrients (fatty acids, glucose) and hormones (leptin and insulin) occurring in obese mothers are possible events involved in the impaired development of offspring (Ramsay et al., 2002; Challier et al., 2008; Madan et al., 2009). Indeed, crossing the placenta, these factors impact the development of neuronal circuits of the offspring, impairing behavioral phenotype and increasing the risk to develop ASD. Moreover, obesity results in an increased level of many inflammatory markers, as C reactive protein, IL- 6, IL-1β, and TNF-α (Das, 2001). This relationship has been observed in pregnant obese women, leading to endothelial (Stewart et al., 2007), placental dysfunction (Leung and Bryant, 2000; Nordahl et al., 2008), and impaired fetal development contributing to the risk of ASD in children (Meyer et al., 2011).

A recent review from Nuttall (2017) evaluates maternal obesity, along with infection, and toxicant exposure, as an environmental risk factor for the occurrence of ASD resulting from a maternal inflammatory response.

In several animal studies, pups born from obese mothers revealed alterations in brain-derived neurotrophic factor, a reduction in neuronal progenitor cell proliferation, atypical synaptic stability, and a reduction in dendrite length and branching (Tozuka et al., 2009, 2010; Yu et al., 2014; Hatanaka et al., 2017), all factors generally revealed in mouse models of ASD. To date, little is still known about the possible epigenetic mechanisms responsible of the development of cognitive functions in the offspring and most of the work has focused on DNA methylation in rodent models.

Considering the multiple mechanisms and factors behind the maternal obese status possibly involved in the development of ASD in offspring, a direct association between obese mothers and child ASD diagnosis cannot be established.

# GUT-BRAIN AXIS

The scaffold of the gut-brain axis is represented by the central nervous system (CNS), the autonomic nervous system (both sympathetic and parasympathetic branches), the neuroendocrine and neuroimmune systems, the enteric nervous system (ENS), and the gut microbiota. All these components intercommunicate via the immune system, the vagus nerve, and other host microbe interactions, and influence each other, constituting a complex network. The main neuroendocrine pathway is the hypothalamic-pituitary-adrenal (HPA) axis, activated in response to various physical and psychological stressors. A crucial player in this communication between peripheral signals and the CNS is the gut microbiota (GM), therefore, this interplay was re-named microbiota-gut-brain axis, viewed as a bidirectional communication system between gut microbes and CNS (Cryan and Dinan, 2012). During development, GM together with immune system, co-evolves with brain neural circuits required for social and emotional cognition, indeed CNS neurotransmission (serotonergic system) is disturbed by an alteration of GM (Clarke et al., 2013).

In mice without gut microbiota colonization (germ-free mice), a mild stress lead to a higher release of corticosterone and adrenocorticotrophic hormone (ACTH) compared to mice with common and no pathogen bacteria (specific pathogen free mice). Moreover, the administration of Bifidobacterium infantis partially reversed the stress response in a time-dependent manner, showing the critical role of intestinal colonization in normal development of the HPA axis (Sudo et al., 2004).

GM capability to influence the brain activity is based on the production of neuroendocrine hormones and neuroactive compounds that play a pivotal role in shaping cognitive networks underlying social cognition, emotion, and behavior (Dinan et al., 2015). Neurochemicals produced by GM can influence the host via two ways: they can either be taken up from the gut into the portal circulation and absorbed crossing the blood-brain barrier (BBB) to modulate cerebral function or they can directly interact with receptors expressed by the ENS, influencing the brain function through ENS-CNS connection. The host GM modulates several features of basic neurodevelopmental processes: integrity of BBB (Braniste et al., 2014); neurogenesis (hippocampus and amygdala) (Ogbonnaya et al., 2015); maturation and activity of microglia (Matcovitch-Natan et al., 2016); myelination in prefrontal cortex (Hoban et al., 2016); synthesis and expression of neurotrophins (Desbonnet et al., 2015), neurotransmitters (O'mahony et al., 2015), and their receptors (BDNF, NR2B, synaptophysin, PSD-95).

# Gut Microbiota in Early Life

GM can be influenced and modulated in a dynamic and rapid manner by several factors, including gestational age, mode of delivery, feeding, and antibiotics/probiotics exposure. Indeed, GM in new-born is characterized by low diversity and relative prevalence of Proteobacteria and Actinobacteria phyla parallel to brain development. Then the diversity increases and changes with a prevalence of Firmicutes and Bacteroidetes (Eckburg et al., 2005). Indeed, the first key factor influencing GM is the gestational age, in fact pre-term babies had a large interindividual variation in GM diversity compared to full-term ones (Barrett et al., 2013). Indeed, pre-term babies are often delivered by caesarian section, that is another factor contributing to GM composition. Natural delivery leads to a new-born microbial composition resembling mother's vaginal canal, while those infants delivered by C-section have instead a GM akin to that of mother's skin (Dominguez-Bello et al., 2010; Del Chierico et al., 2015; Hill et al., 2017). Beyond from the mode of delivery, pre-term children are more frequently fed due to a functionally immature gut, leading to the dominance of pathogenic bacteria and a lower diversity, accompanied also by reduced production of bacteria-derived metabolites, such as short chain fatty acids (SCFAs) (Arboleya et al., 2012). However, in this case, breast feeding would increase pre-term infant microbiota consistently with the utilization of human milk oligosaccharides and substrates (i.e., Bifidobacterium species). Indeed, in this phase the GM composition could remain stable or shift, due to environmental events, such as fever, antibiotic therapy, and formula feeding (Costello et al., 2012).

Whether and how all these homeostatic modifications in GM in early life can impact brain development and maturation is an open question whose understanding would enable the development of nutritional and therapeutic interventions for the prevention of later neurodevelopmental illness.

# Gut Microbiota and ASDs

Beyond the social and psychological pattern, GI symptoms are a common comorbidity in ASD (Adams et al., 2011; Williams et al., 2011; Chaidez et al., 2014; Li et al., 2017), even if the underlying mechanisms are still unclear.

Germ-free rats and mice revealed, compared to specific pathogen free animals, a reduced sociability, an increased anxiety-like behavior and peculiar brain gene expression profiles, alterations of neurophysiology, strongly reinforcing the idea that GM may affect mammalian brain development and subsequent adult behavior (Desbonnet et al., 2014; Stilling et al., 2015). In particular, an increasing body of evidence shows that the microbiota-gut-brain axis could participate in ASD. The use of antibiotics, possibly affecting GM homeostasis, during pregnancy is a potential risk factor for infantile ASD (Atladottir et al., 2010). Improvement in behavioral and GI symptoms of ASD patients was observed following different treatments able to manipulate GM (i.e., antibiotics and probiotics) (Sandler et al., 2000; Parracho et al., 2005). All these findings suggest that GM modification may contribute to improve ASD symptoms and that microbiota assortment could represent the boundary between environmental and genetic risk factors that convey in ASD.

# Gut Microbiota Dysbiosis in ASD Patients

Several human studies identified dysbiosis in ASD patients with respect to neurotypical control subjects. ASD patients exhibited decreased Bacteroidetes/Firmicutes ratio (Tomova et al., 2015). High-throughput sequencing analysis showed a significant reduction in relative abundance of phylum Bacteroidetes in autistic subjects, mostly classified as severe ASDs (Strati et al., 2017). On the contrary, Son et al. (2015) did not find any alteration in GM composition in a study comparing fecal microbiota in ASD children and neurotypical siblings by qPCR. Specific bacterial genera have been reported to be differentially abundant in ASD individuals compared to controls: Clostridium, Sutterella, and Desulfovibrio species, Bacteroides vulgatus, Collinsella, Corynebacterium, Dorea, and Lactobacillus were found increased in ASD patients (Parracho et al., 2005; Finegold et al., 2010; Williams et al., 2012; De Angelis et al., 2013; Strati et al., 2017), while Bifidobacterium, Prevotella, Akkermansia muciniphila, Alistipes, Biophila, Dialister, Parabacteroides, and Veillonella were frequently found reduced in the intestinal microbiota of ASD children compared to control subjects (Finegold et al., 2010; Wang et al., 2011; Kang et al., 2013; Strati et al., 2017). Despite the vast majority of researches indicate a distinctive GM composition in ASD, it is not possible, up to now, to define a gut microbial signature for ASD, because of small scale pre-clinical studies and conflicting results often reported. Homogeneity for enrolment criteria and technologies used for GM definition would greatly help to share microbiome data to find out association between microbial communities and different pathological states.

# Relationship Among Gut Microbiota, Immune Dysregulation, GI, and Behavioral Alterations in ASDs

A bulk of knowledge pinpoints a role for GM in immune dysregulation and inflammation in ASD (Depino, 2013). Autistic patients show alterations in circulating and brain cytokines and in inflammatory factors along with abnormal responsiveness to stimulation of peripheral blood monocytes and macrophages, suggesting a prominent role of inflammation in the onset and/or development of the disease (Grigorenko et al., 2008; Careaga and Ashwood, 2012; Hsiao, 2013). Moreover, deficiencies in the integrity of both gut epithelium barrier and BBB have also been reported in ASD individuals (de Magistris et al., 2010; Fiorentino et al., 2016). Thus, one possibility is that bacterial products, metabolites, and antigens able to translocate through the altered GI barrier, induce peripheral immunoreaction; peripheral cytokines can activate the vagal system which in turn regulates CNS activity (Yarandi et al., 2016). It is also possible for cytokines and bacterial compounds to reach the blood stream, cross the BBB and directly signal to CNS (Ashwood et al., 2011), impacting on neuronal plasticity and consequently on mood and behavior. Experiments that clearly show a correlation among microbial dysbiosis, GI alterations, immunity, and behavior have been reported for animal models of ASD. In a mouse model of environmental risk factors, such as in utero valproic acid (VPA) exposure, gut inflammation, altered microbiota, and ASD-like behavioral abnormalities in male offspring were reported (de Theije et al., 2014). In mouse models of maternal immune activation (MIA) the offspring displays principal features of ASD, decreased permeability of GI tract, altered serum metabolome and increased levels of IL-6 in the colon. Notably, probiotic administration (Bacteroides fragilis) proved to be a key treatment to ameliorate alterations in commensal microbiota, to restore intestinal permeability and cytokine production and to improve behavioral abnormalities (Hsiao, 2013; Hsiao et al., 2013). Beyond behavioral tests, in a recent study, Coretti et al. (2017) also explored GM profile, intestinal function and immunological features of adult male and female BTBR mice, which represent the main animal model of ASD. These authors identified some key genera, such as Bacteroides, Parabacteroides, Sutterella, Dehalobacterium, and Oscillospira, in GM profile of BTBR mice. In particular, female BTBR showed a strong link among an autistic outcome, an increased expression of TNF-α, of Parabacteroides and Sutterella, together with a decrease of Dehalobacterium, Oscillospira, and unclassified member of TM7, a subgroup of Gram-positive still uncharacterized, also known as Candidatus Saccharibacteria. On the other hand, the autistic profile of male BTBR mice was accompanied by an increased presence of unclassified members of Helicobacteriaceae and low expression of anti-inflammatory IL-10. Furthermore, in these mice, low levels of Dehalobacterium, Ruminococcus, and Desulfovibrio were associated with increased gut permeability (Coretti et al., 2017). Altogether these studies indicate a strong relationship among GM composition, gut integrity, immunological state, and behavior.

# Influence of Microbial Metabolites on Brain Functions in ASDs

Due to the capability to cross BBB, the influence of bacteriaproduced metabolites, such as short chain fatty acids (SCFAs), free amino acids (FAAs), and phenol compounds (4-ethylphenylsulfate, 4EPS), in neuropsychological disorders has been deeply clarified (Morris et al., 2017). Several studies showed that the administration or high concentrations of different SCFAs modulated the pathological features of ASD in opposite ways. Indeed, high levels of propionic acid (PPA) have been associated to the onset of ASD. Thomas et al. (2012) demonstrated that intracerebroventricular administration of PPA increased the stereotyped and repetitive behaviors in rats, influencing the metabolism of key neurotransmitters, such as dopamine and norepinephrine and epinephrine. Furthermore, autistic children fed with products containing PPA or which are metabolized in this SCFA, showed an increase of autistic behaviors and GI symptoms. The deprivation of these foods or reduction of PPA-producing bacteria induced by antibiotics ameliorated their clinical condition (Mellon et al., 2000). On the other hand, sodium butyrate, another SCFA, showed opposite effects of PPA. In two mice models of ASD, the administration of sodium butyrate was able to improve autistic outline, reducing repetitive movements and increasing sociability interaction (Takuma et al., 2014; Kratsman et al., 2016).

# DIET AND BEHAVIOR

Most children with ASD displaying GI disorders, display malabsorption (Goodwin et al., 1971), maldigestion (Cade et al., 2000), microorganism overgrowth (fungi, bacteria, and virus) (Finegold et al., 2010), and altered intestinal integrity (de Magistris et al., 2010), causing symptoms, such as diarrhea, constipation, bloating, belching, and visibly undigested foods (Klukowski et al., 2015). Indeed, impairment in carbohydrate metabolism could underline some of the GI ailments occurring in some ASD patients, even if their role in the neurological and behavioral traits is unclear.

The impact of nutrition-related factors in the etiology and symptoms of ASD encourages to combine the conventional therapy based on behavioral and pharmacological approach together with an appropriate diet in order to improve gut health and alleviate the disease severity (van De Sande et al., 2014). Therefore, adopting individual diets tailored to disease symptoms, would prevent the onset of GI dyscomfort, taking into account the dietary reference values for food energy and nutrients and food preferences of the patient.

Nutritionists also require constant patient management in ASD mostly when they are obese, overweight or wasted, to correct an inappropriate diet. Anyway, the dietary therapy needs to be accompanied by other strategies to better handle ASD.

Available online at: Reducing intake of certain food products (i.e., gluten-free or casein-free diets) is associated with less incidence of numerous GI disorders. On the other hand, several studies show that nutritional deficiencies of autistic patients are filled with the supplementation of vitamins and minerals, fatty acids ω-3, probiotics, in combination with pharmacological and psychological interventions, even if supplementation interventions show contrasting, but promising results.

# Gluten- and Milk Protein-Free Diets

A typical strategy to decrease food related effects in ASD is a gluten-free and casein-free (GFCF) diet.

GFCF diet consists in the elimination of food containing gluten, and products containing gluten trace amounts. This diet also eliminates casein, a protein present in cow milk and dairy products. The absence in this diet of milk and dairy products leads, however, to calcium, phosphate, and vitamin D deficiency. Therefore, nutrition specialists usually recommend soy or rice milk as substitutes for cheese (Kawicka and Regulska-Ilow, 2013).

The interest in GFCF diet is related to the opioid excess theory of ASD. This theory suggested that autistic children present abnormal metabolism of those two proteins (gluten and casein) possibly due to peptidase deficiencies. This alteration may result in excessive opioid activity in the CNS, altering its function. In particular, gluten- or casein-derived peptides are suspected to be involved in ASD, resembling opioid-like molecules (Piwowarczyk et al., 2017), able to interact with opioid receptors expressed both in the CNS and GI tract or with opioid metabolizing enzymes.

Data to assess the effects of GFCF diet are limited since nutritional strategies and outcome considered varied among studies as did standard diet and monitoring of compliance to GFCF regimen. Many reports showed a reduction of several behavioral symptoms in ASD patients (Knivsberg et al., 2002; Patel and Curtis, 2007), but re-introduction of gluten or casein containing products leads to autistic symptom recurrence (Piwowarczyk et al., 2017).

# The Specific Carbohydrate Diet

The specific carbohydrate diet (SCD), introduced by Gotschall (2004), represents a strategy to reduce symptoms of malabsorption able to limit growth of pathogenic gut microbiota in ASD patient.

This diet largely recommends monosaccharides, whose origin is fruit, vegetables, and honey, whereas it removes complex carbohydrates. The formulation of this diet is based on the evidence that autistic patients have an abnormality in carbohydrate digestion and adsorption, causing residual food accumulation that represents a breeding ground for pathogenic intestinal flora (Williams et al., 2011). The components of SCD are meat, vegetables (onions, spinach, peppers, cauliflower, and cabbage), eggs, natural cheeses, nuts (walnuts, brazil nuts, and almond), fresh fruits, soaked lentils, and beans (Gotschall, 2004). Autistic children treated with SCD report an improvement in attitude, increased skills, and responsiveness (Gotschall, 2004).

# Low Oxalate Diet

The presence of certain substances, such as oxalate, in GI tract can be strongly linked to an impairment of CNS structure and functionality (Levy et al., 2007). Furthermore, Konstantynowicz et al. (2012) demonstrated that ASD patients exhibited high levels of oxalates in plasma and urine, underlining the possible correlation between this molecule and the disease. The nutriments containing oxalate are numerous beetroots, spinach, cocoa, figs, black tea, lemon zest, black grapes, green apples, kiwis, tangerines, strawberries, oats, wheat, berries, millet, cashew nuts, hazelnuts, peanuts, almonds, and blueberries (Marcason, 2006). Given its presence in several foods, the daily intake of oxalate in a typical western diet exceeds the acceptable daily intake (ADI) (1000 mg/day instead of 250 mg/day). Therefore, ASD patients should reduce the intake of these foods to 40–50 mg/day (Jaeger and Robertson, 2004) but, at the same time, they should add dietary supplements containing arginine, vitamins A and E, glucosamine, taurine, glutathione, magnesium, thiamine, citrate, CoA, magnesium, zinc, and calcium.

# Micronutrients

Micronutrients, essential for neurogenesis and neuro-network development (Curtis and Patel, 2008), have been reported to be reduced in serum, hair, or tissues from ASD children, as shown for magnesium (Strambi et al., 2006), zinc, selenium, vitamin A, vitamin D, vitamin B complex, vitamin E, and carnitine (Filipek et al., 2004). This deficit is related to several factors: autistic children often experience significant eating difficulties, specific food selectivity (Williams et al., 2000; Arnold et al., 2003; Herndon et al., 2009), poor digestion (Williams et al., 2011), gut inflammation (Jyonouchi et al., 2001; Bauer et al., 2007), and reduced levels of vitamin-producing microbiota in the intestine (Guarner and Malagelada, 2003; Wang et al., 2011). Therefore, multivitamin and mineral oral supplement are widely recommended interventions for ASD (Golnik and Ireland, 2009), improving ASD symptoms (Adams and Holloway, 2004; Mousain-Bosc et al., 2006).

It has also been reported that during pregnancy a deficiency of folic acid (Schmidt et al., 2012; Suren et al., 2013) and vitamin D (Cannell, 2008; Grant and Soles, 2009), increase the risk for offspring developing ASD.

# Polyunsaturated Fatty Acids

PUFA, namely arachidonic acid (AA, 20:4 ω-6), eicosapentaenoic acid (EPA, 20:5 ω-3), and docosahexaenoic acid (DHA, 22:6 ω-3), are crucial for brain development and cognitive and memory functions (Das, 2013).

Maternal intake of PUFAs improved memory in the progeny, suggesting that, supplementation of PUFAs needs to start during pregnancy and continue after delivery until brain development is complete in adolescence. In agreement, breastfeeding improves brain growth and development and memory, since breast milk is richer in AA and DHA compared with formula (O'connor et al., 2001; Auestad et al., 2003; Alessandri et al., 2004). These results suggest that availability of adequate amounts of various PUFAs (both ω-6 and ω-3 fatty acids), especially AA and DHA to the fetus and new-born, is recommended (Das, 2003a,b; Das and Fams, 2003). In fact, women with higher intake of PUFAs before and during pregnancy had a reduced risk of having a child with ASD than those with lower PUFA intake (Fujiwara et al., 2016).

Moreover, PUFAs inhibited the production of neurotoxic TNF-α, IL-1, and IL-6, and increased nitric oxide synthesis, inhibiting neuronal apoptosis and facilitating memory improvement and consolidation (Das, 2003a,b, 2010, 2011; Das and Fams, 2003).

Several studies showed that both plasma and red blood cell phospholipid fatty acid composition is altered in autistic patients. In particular, the levels of ω-3 fatty acids, including DHA, were reduced in the erythrocytes or plasma of ASD patients compared to controls (Vancassel et al., 2001; Bell et al., 2004; Wiest et al., 2009; Al-Farsi et al., 2013; Brigandi et al., 2015; Yui et al., 2016), without significant reduction in the total (ω-6) PUFAs, with a significant increase in the (ω-6)/(ω-3) ratio (Vancassel et al., 2001), although other studies did not support these claims (Bu et al., 2006; Bell et al., 2010).

Indeed, it was reported that the supplementation of EPA/DHA (Amminger et al., 2007) or addition of high amounts of AA (Yui et al., 2012) produced a significant beneficial effect in ASD.

Intriguingly, metabolism of PUFAs, involving several cofactors such as antioxidants, minerals, trace elements, and various vitamins seems to be compromised (Das, 1985). AA, EPA, and DHA are precursors to anti-inflammatory bioactive lipids, such as lipoxins, resolvins, and protectins, involved in wound healing, and neuroprotection from various endogenous and exogenous insults (Das, 2013). It has been suggested that in some autistic patients, the metabolism of PUFAs is deficient or abnormal therefore anti-inflammatory lipids are lower, in spite of an increase of proinflammatory cytokines, oxidative stress, an alteration of various neurotransmitters (dopamine, serotonin, catecholamines, and BDNF), leading to ASD onset and progression, despite PUFA administration.

Interestingly, PUFAs are also able to induce an increase in BDNF levels in the brain (Wu et al., 2004; Bousquet et al., 2009), where it regulates neuronal development and plasticity. In ASDs, abnormalities in PUFAs and BDNF are shown, thus suggesting an interplay between these endogenous molecules impacting brain growth and development and cognitive function.

# MANIPULATION OF GUT MICROBIOTA

# Probiotics and Prebiotics

Probiotics, commonly recommended in GI disorders, are living microorganisms considered beneficial for human health, generally belonging to Gram positive taxa, (i.e., Lactobacillus and Bifidobacterium genus). On the basis of microbiota involvement in ASD etiology, probiotics have been considered as a therapeutic tool able to impact brain development, and behavior. Therefore, the rationale of the use of probiotics, has been to re-establish the healthy equilibrium of GM altered in ASDs. One of the first evidence showing the impact of GM on ASDs has been demonstrated by Sandler et al. (2000). Eight weeks treatment of autistic children with poorly absorbed oral vancomycin was able to induce a regression of autistic behavioral traits, an effect that was suddenly blunted after vancomycin withdrawal. This study performed on 11 children showed the requirement of a long-term antibiotic therapy not feasible for ASD patients. On the other hand, it also underlined the role of Gram-positive bacteria belonging to vancomycin spectrum in ASD etiology and/or development. The possible underlying mechanism of this antibiotic was supposed to be the decline of neurotoxinproducing pathogens, such as Clostridium, whose surviving spores to therapy were responsible, after germination, of the recurrence of ASD symptoms. In fact, Linday (2001) suggested as adjunctive treatment to ASD therapy the administration of Saccharomyces boulardii for its efficacy in Clostridium difficile colitis.

In a double-blind, placebo-controlled, crossover-designed probiotic feeding study, autistic children ranging between 3 and 16 years of age were divided into two groups, one receiving the vehicle (placebo) for 3 weeks and Lactobacillus (L.) plantarum WCFS1 for the following 3 weeks, and the other receiving first the probiotic and then the placebo (Parracho et al., 2010). Oral administration of L. plantarum was efficacious in autistic children compared to placebo group, not only modulating GM composition and GI symptoms, but also improving autistic behavior. Actually, probiotics have been suggested in autistic patients where GI comorbidities are also shown (Golnik and Ireland, 2009), but are not safe in immunocompromised children (Geraghty et al., 2010).

In a prospective study on preterm children treated with L. reuteri or L. rhamnosus in order to prevent GI colonization by Candida species, the authors reported the reduction of Candida in stool from treated infants with both probiotics, as well as GI symptoms by L. rhamnosus administration; notably, all treated newborns showed lower incidence of poor neurological outcomes compared to untreated control group, evaluated up to 12 months by the Hammersmith Infant Neurological examination (Romeo et al., 2011). The role of Candida GI colonization in neurological status of newborns was suggested by the elevated urine level of a metabolite of several Candida species, D-arabinitol, in autistic patients (Shaw et al., 2000). Indeed, the oral treatment of autistic children with Lactobacillus acidophilus for 2 months reduced D-arabinitol in urine, indicating this approach useful for improving autistic behavior (Kaluzna-Czaplinska and Blaszczyk, 2012). However, while the level of concentration ability and carrying out orders were improved in autistic children, the response to emotions was not varied.

The influence of GM and its metabolites on ASD has been demonstrated by Hsiao et al. (2013) through maternal immune activation (MIA) mouse model. Actually, the dysbiosis accompanying this model induced a modification in the production of bacteria-derived metabolites that, spilled into the bloodstream of the offspring, underlined the neurological anomalies. In fact, the administration of the Bacteroides fragilis reduced MIA-induced increase in 4-ethylphenylsulphate in the offspring and was able to restore gut integrity, to modulate GM and to reduce the impairment in socialization and communication, the stereotypies and the sensorimotor dysfunction.

Some improvement in GI symptoms as well as a reduction in the severity of autistic traits were observed in children with ASD after 6 months of supplementation with Delpro (West et al., 2013), that was a probiotic mixture containing L. acidophilus, L. casei, L. delbrueckii, Bifidobacterium longum, and Bifidobacterium bifidum combined with the immunomodulator Del-Immune V.

GM impact on host neurology has been shown by the correction of the impairment of social behavior of the offspring from high fat diet-fed mothers by co-housing these animals with healthy pups delivered from mothers on standard diet. Moreover, the treatment with L. reuteri, which was shown reduced in the offspring, selectively improves social behavior by promoting oxytocin-mediated functions (Buffington et al., 2016).

The efficacy of probiotics on reducing inflammation in autistic children having GI comorbidities was also shown by Russo (2015), even if the type of probiotics administered was not reported. Russo showed that various probiotic-based therapies induced a reduction in myeoloperoxidase and copper level compared to untreated patients (Russo, 2015), suggesting GM modulation and reduced GI inflammation as main mechanisms underlining those effects.

A recent study performed in children to evaluate the impact of probiotic on CNS functions was conceived administering L. rhamnosus GG or placebo in the first 6 months of life of infants and evaluating the onset of neuropsychiatric disorders in the following 13 years (Pärtty et al., 2015). The results indicate a reduction in the later onset of attention-deficit/hyperactivity disorder or Asperger syndrome by the probiotic, showing a link between the early GM and development of these disorders, although no single constant microbiota composition component or change was detected. The study by Tomova et al. (2015) showed an imbalance in GM in autistic children, accompanied by GI dysfunction whose severity correlated with that of autistic symptoms. Unfortunately, apart from the restoration of GM in these patients by a probiotic mix of three strains of Lactobacillus, two of Bifidobacterium, and one of Streptocossus, the possible reversal of the autistic behavior was not analyzed. Conversely, the case study by Grossi et al. (2016) strengthened the efficacy of probiotic therapy in ASD core symptoms. Twelve years old children received VSL#3 mixture of 10 probiotics for 4 months, then the patients were monitored for subsequent 4 months. Apart from the reduction of abdominal symptoms, no improvement was shown for the restricted, repetitive behaviors, but for the social affect, indicating probiotics as possible therapeutics for behavioral abnormalities associated with ASD.

A very recent prospective, open-label study on 30 ASD children from 5 to 9 years old, showed that 3 months of treatment with L. acidophilus, L. rhamnosus, and Bifidobacterium longum, modulated GM, increasing Lactobacilli and Bifidobacteria levels in the stool, and improved both GI and autistic symptoms (Shaaban et al., 2017).

In conclusion, many of these studies on probiotic efficacy in ASDs had several limitations, among which the sample size, the patient enrollment, and the criteria for diagnosis; even the design was not always planned in order to assess significant evaluation of clinical outcomes or show side effects. A clinical study on 100 autistic preschoolers is currently in progress in order to gather more information on the effects of a probiotic treatment both on GI function and neurophysiological pattern (Santocchi et al., 2016).

Prebiotics are carbohydrates, including inulin and oligosaccharides, and some food ingredients indigestible by the host, able to induce modifications in the composition and activity of GM. A study by Schmidt et al. (2015) showed a reduction of salivary cortisol secretion in healthy volunteers by fructo-oligosaccharides and Bimuno galacto-oligosaccharides treatment, indicating a suppression of neuroendocrine stress response and an increase of attention span, both events impacting ASD patients. Therefore, even if no study has analyzed strictly the effects of prebiotics alone in ASD animal models or patients, several findings strengthen the possible improvement by these treatments, able to correct pathological features common in ASDs.

It has been demonstrated, indeed, that prebiotics can directly or indirectly affect signaling molecules. At CNS level, for instance fructo- and galacto-oligosaccharides are able to restore BDNF at hippocampal level (Savignac et al., 2013), probably through the involvement of gut hormones. Moreover, prebiotics do not have the survival problem in the GI tract that probiotics do.

# Fecal Microbiota Transplantation

Fecal microbiota transplantation (FMT) is a procedure through which patients receive the fecal microbiota from healthy donors. Generally, this practice is beneficial in ailments where dysbiosis plays a pivotal role, since it aims to restore GM homeostasis. Obviously, GI disorders, such as inflammatory bowel diseases and irritable bowel syndrome, are the most common ailments suggested for this intervention, even if also other disorders, like autoimmune disease and obesity has been tested with beneficial effects (Aroniadis and Brandt, 2013). Clinically controlled studies are needed to assess not only the efficacy, but also the possible limitations of this intervention, i.e., the choice of donors, the dosage and duration of this treatment, the delivery system, and the side effects. Moreover, before FMT, patients need to be pretreated with antibiotics and bowel-cleansing regimens (Krajmalnik-Brown et al., 2015; Toh and Allen-Vercoe, 2015). The mechanisms underlying FMT positive effects are not completely clear: indeed, they could be related to viable microbiota or viruses, and/or to substances contained in the

# REFERENCES


feces, such as bile acids, vitamins, short chain fatty acids. Considering the emerging role of GI dysbiosis in ASD, there is a growing attention on FMT approach in ASD. FMT in autistic children has been described by Aroniadis and Brandt (2013). Therefore, the efficacy of FMT in autistic children has been recently studied in an open label trial (Kang et al., 2017). Even if an improvement of GI symptoms and autistic behaviors was reported, this study showed lacking points, since it was not placebo controlled, blinded, or randomized.

# CONCLUSIONS

Here, we have briefly reported how unhealthy perinatal and post-natal environment could be associated to later changes in behaviors, establishing the crucial importance of the bidirectional communications between peripheral signals and brain. Maternal obesity, early life exposure to over-or under-nutrition and dysfunction of the microbiota-gut-brain axis are possibly associated with the onset and progression of ASDs. Therefore, the interplay between nutrition and microbial composition in ASDs patients is worthy of further investigation. Gut homeostasis, as well as neurobehavioral patterns, depend upon microbes for certain key nutrients, in terms of diversity and amount. Therefore, the alterations of GM would lead to an unbalance of metabolite profile, immune deregulation and activation of pathways impacting neurological function. Up to date no treatment is fully successful in treating ASD, however, a combination of complementary interventions and techniques could relieve at least in part behavioral symptoms. Adjuvant therapies for ASDs include tailored diets, and gut microbiota manipulation (i.e., the use of probiotics and prebiotics), in order to correct dysbiosis and restore a healthy conversation between gut and brain. This would partially lessen both gastrointestinal and neurobehavioral symptoms.

# AUTHOR CONTRIBUTIONS

CC, AL, FL, MPM, AC, and GM wrote the paper. CC, MPM, and GM supervised the review editing. GM decided the overall structure of the review, coordinating the working group.


red blood cells. Int. J. Mol. Sci. 16, 10061–10076. doi: 10.3390/ijms1605 10061


development and disabilities at kindergarten. Int. J. Obes. 37, 1344–1351. doi: 10.1038/ijo.2013.128


receiving the Delpro R probiotic and immunomodulator formulation. J. Prob. Health 1:102. doi: 10.4172/2329-8901.1000102


**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 Cristiano, Lama, Lembo, Mollica, Calignano and Mattace Raso. 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 Microbiome-Mitochondria Dance in Prodromal Parkinson's Disease

#### Sandra M. Cardoso1,2 \* and Nuno Empadinhas<sup>1</sup>

<sup>1</sup> Center for Neuroscience and Cell Biology, University of Coimbra, Coimbra, Portugal, <sup>2</sup> Institute of Cellular and Molecular Biology, Faculty of Medicine, University of Coimbra, Coimbra, Portugal

The brain is an immunologically active organ where neurons and glia cells orchestrate complex innate immune responses against infections and injuries. Neuronal responses involve Toll-like or Nod-like receptors and the secretion of antimicrobial peptides and cytokines. The endosymbiotic theory for the evolutionary origin of mitochondria from primitive bacteria, suggests that they may have also retained the capacity to activate neuronal innate immunity. In fact, it was shown that mitochondrial damage-associated molecular patterns could signal and activate innate immunity and inflammation. Moreover, the mitochondrial cascade hypothesis for sporadic Parkinson's disease (PD) argues that altered mitochondrial metabolism and function can drive neurodegeneration. Additionally, a neuroinflammatory signature with increased levels of pro-inflammatory mediators in PD affected brain areas was recently detected. Herein, we propose that a cascade of events initiating in a dysbiotic gut microbiome drive the production of toxins or antibiotics that target and damage mitochondria. This in turn activates neuronal innate immunity and triggers sterile inflammation phenomena that culminate in the neurodegenerative processes observed in the enteric and in the central nervous systems and that ultimately lead to Parkinson's disease.

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

Carsten Culmsee, Philipps University of Marburg, Germany Abdu Adem, United Arab Emirates University, United Arab Emirates

#### \*Correspondence:

Sandra M. Cardoso cardoso.sandra.m@gmail.com

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 24 January 2018 Accepted: 16 April 2018 Published: 09 May 2018

#### Citation:

Cardoso SM and Empadinhas N (2018) The Microbiome-Mitochondria Dance in Prodromal Parkinson's Disease. Front. Physiol. 9:471. doi: 10.3389/fphys.2018.00471 Keywords: mitochondria, bacteria, microbiome, neuronal innate immunity, Parkinson's disease

# PARKINSON'S DISEASE OVERVIEW

Parkinson's disease (PD), the most frequent neurodegenerative movement disorder, is characterized by severe loss of midbrain dopaminergic neurons in the SNpc and by the presence of intra-cytoplasmic inclusions of aggregated SNCA, known as LBs (Obeso et al., 2010). Sporadic PD (sPD) is a multifactorial disorder that evolves over decades without any motor complications. PD has a long prodromal phase during which several other symptoms develop, namely related to olfactory impairment, sleep disturbances, and depression (Reichmann et al., 2009). Another common underlying symptom described for the prodromal phase in PD patients is GI dysfunction

**Abbreviations:** AMP, antimicrobial peptides; CNS, central nervous system; DAMP, damage-associated molecular pattern; ENS, enteric nervous system; FPR, formyl peptide receptors; GI, gastrointestinal; IL, interleukin; LBs, Lewy bodies; LPS, lipopolysaccharide; mtDNA, mitochondrial DNA; mtROS, mitochondrial produced reactive oxygen species; NLR, NODlike receptors; Nlrp3, NLR family pyrin domain containing 3; NOD, nucleotide-binding oligomerization domain-like; PAMP, pathogen-associated molecular pattern; PD, Parkinson's disease; PET, positron emission tomography; PRR, pattern recognition receptor; SNCA, alpha-synuclein; SNpc, substantia nigra pars compact; TIR, Toll/interleukin-1 receptor; TLRs, Toll-like receptors; TNFα, tumor necrosis factor α.

that also includes dysphagia, gastroparesis, and severe constipation (Pfeiffer, 2003). These symptoms correlate with Braak staging whereas SNCA-immunopositive Lewy neurites and LBs target specific induction sites: initially in the dorsal motor nucleus of the glossopharyngeal and vagal nerves and in the anterior olfactory nucleus (Braak et al., 2003a).

Because LBs are also detected in the ENS of earliest and asymptomatic stage patients (Natale et al., 2011), the GI tract was proposed as an early target of PD pathology. Despite this, several evidences resulting from brain autopsy, animal models and cellular studies show that PD neurodegeneration involves multiple cellular processes, including mitochondrial dysfunction, oxidative stress, proteasomal and autophagic impairments and neuroinflammation (Olanow, 2007).

# MITOCHONDRIAL INVOLVEMENT IN PARKINSON'S DISEASE ETIOLOGY

Mitochondria host biochemical reactions essential for normal cell functioning, namely energy production and maintenance of redox homeostasis (Lobet et al., 2015). Mitochondria were associated to sPD pathology when deficits in mitochondrial NADH dehydrogenase (complex I) activity were identified in the SNpc of post-mortem PD patient's brains and in their platelets (Cardoso, 2011). Our group demonstrated that dysfunctional mitochondria from PD patients' trigger several pathogenic features observed in PD subject brains, such as the generation of protein aggregates (LBs "like") (Esteves et al., 2009), microtubule disassembly, disruption of intracellular trafficking (Esteves et al., 2010) and accumulation of autophagosomes and autophagic substrates (Arduíno et al., 2012).

The complex I inhibitors 1-methyl-4-phenyl-1,2,3,6 tetrahydropyridine and rotenone are widely used as in vitro and in vivo models of PD given that they recapitulate the main features of the disease (Meredith and Rademacher, 2011; Xiong et al., 2012). Mitochondrial dysfunction in PD tissues and models is also characterized by a decrease in mitochondrial membrane potential (Mann et al., 1992; Esteves et al., 2008), and by an increase in mitochondrial pool fragmentation and cristae disruption (Baloyannis et al., 2006; Santos et al., 2015). Accordingly, at a functional level, brain bioenergetics is compromised in PD where PET scans show decreased glucose utilization in PD individuals in the occipital cortex compared to control individuals (Schapira, 2008).

# RELEVANCE OF A BACTERIAL ORIGIN OF MITOCHONDRIA

After exposure to a new pathogen, our innate immune system protects us from infection. Innate immune responses are not specific to a particular pathogen and depend on the recognition of several conserved features of pathogens (Ward and Rosenthal, 2014). The innate immune response relies on PRRs to identify PAMPs, many of which are normal components of bacterial cells (Pallen, 2011).

Mitochondria share a common ancestor with Alphaproteobacteria and so proposed to be derived from ancestral bacterial endosymbiosis. The evidence supports a common origin for mitochondria and bacteria related to the Rickettsiales that have extremely reduced genomes and have obligate intracellular lifestyles (Fitzpatrick et al., 2006). mtDNA shares features with the genome of Rickettsia prowazekii but the similarities between bacteria and mitochondria extend beyond the abundance in the distinctive lipid cardiolipin in the inner membrane, to the numerous small molecule transport systems and to an electron transport chain that pumps protons across the inner mitochondrial membrane with the resulting proton motive force driving ATP synthesis via the F<sup>1</sup> ATP synthase. Additionally, both the matrix of mitochondria and the cytosol of bacteria contain DNA, tRNA, ribosomes, and numerous soluble enzymes; both reproduce by binary fission and bear a N-formylmethionine start residue in their proteins. Remarkably, some bacterial PAMPs persist in mitochondria, such as formyl peptides that activate FPRs and unmethylated CpG dinucleotides, which activate TLR. Therefore the innate immune system does indeed recognize mitochondrial bacterial motifs, also called DAMPs. Upon mitochondrial release of DAMPs a sterile inflammation is activated that mimics the response to infection (Pallen, 2011).

# MITOCHONDRIA DAMPs TRIGGERS INNATE IMMUNITY

The innate immune response can also be triggered by tissue damage independently of infection, a process also referred to as sterile inflammation, during which damaged cells release endogenous messengers known as DAMPs that are able to activate TLRs (Wilkins et al., 2017). At least 13 mammalian TLR isoforms are known, and each is capable of recognizing certain types of PAMPs or DAMPs (Takeuchi and Akira, 2010). Since mitochondria are an important source of DAMPs, the release of these mitochondrial DAMPs upon injury activates the innate immune system (Taanman, 1999). mtDNA is similar to bacterial DNA containing CpG motifs, which activate the TLR9 (Taanman, 1999; Zhang et al., 2010). Moreover, mitochondrial protein synthesis is initiated with the residue N-formylmethionine, similar to bacterial protein synthesis (Rabiet et al., 2007). The resulting bacterial N-formylated peptides are known to act as PAMPs by binding and activating G protein-coupled FPRs (Gurung et al., 2015), while the mitochondrial N-formylated peptides act as DAMPs through activation of the FPR1 (Taanman, 1999). Several studies have now described a crucial role for mitochondria in the regulation and activation of NLR specifically the Nlrp3 inflammasome (Latz et al., 2013). The inflammasomes are intracellular molecular platforms activated upon cellular infection or sterile stressors, which activate the pro-inflammatory cytokines interleukin-1β (IL-1β) and IL-18, to trigger cell death (reviewed in Zhou et al., 2011). A variety of insults, resulting from cellular infection or stress, can promote mitochondrial dysfunction and activate the Nlrp3 inflammasome (Schroder and Tschopp, 2010; Latz

et al., 2013). While initial studies showed that mitochondrial dysfunction and mtROS production are necessary for Nlrp3 inflammasome activation (Nakahira et al., 2011; Shimada et al., 2012), further evidence has shown that mtDNA translo cation to the cytosol plays an active role in this process, where it can directly bind to and activate the Nlrp3 inflammasome (de Andrade Rosa et al., 2006). In addition, the mitochondrial lipid cardiolipin is also essential for Nlrp3 inflammasome activation, by directly binding to Nlrp3, downstream of mitochondrial dysfunction (Iyer et al., 2013). Altogether, by sensing mitochondrial DAMPs, the Nlrp3 inflammasome plays a critical role in integrating mitochondrial dysfunction in a pro-inflammatory signaling response, thus explaining the association of mitochondrial damage with inflammatory diseases.

Despite the great number of studies describing mitochondria as a source of DAMPs, the potential for mitochondrial DAMPs to trigger, or exacerbate, inflammation in the brain is now being explored. In recent studies, this potential was tested by treating different brain cell types with mitochondrial components, and measuring markers of inflammation. Neuronal and microglial cell lines exposed to mitochondrial lysates displayed increased markers of inflammation, with mtDNA being identified as the candidate DAMP responsible for the inflammatory changes (Wilkins et al., 2015). Recently, it was observed that stereotactic injection of mitochondrial lysates or purified mtDNA into animal hippocampi induced pro-inflammatory modifications (Wilkins et al., 2016), such as increased levels of hippocampal TNFα mRNA, glial fibrillary acidic protein and NFκB phosphorylation in the cortex.

# NEURONAL INNATE IMMUNITY ACTIVATION IN PARKINSON'S DISEASE

Innate immunity reacts to different insults that may challenge the integrity of the CNS. This process is initiated by receptors of the TLR family that are activated by PAMPs or DAMPs. In the brain, this response is considered to be mediated by microglial cells, the major antigen-presenting cells in the CNS. Nevertheless, neurons also express critical TIR domain-containing adaptors that transduce signals of TLR, namely TLR1, TLR2, TLR3, TLR4, TLR7, and TLR9, and regulate the expression of various cytokines (Liu et al., 2014). Indeed, TLR3 and 7, localized in the neuronal endosomal compartment, play a role in neurite outgrowth. In neurons, TLR9 mainly found in the ER (Shintani et al., 2014), reduces the calcium transfer to the mitochondria promoting autophagy and cell survival (de Bernard and Rizzuto, 2014; Shintani et al., 2014). Others also found that TLR4 signaling regulates axonal growth, neuronal plasticity and even adult neurogenesis (Okun et al., 2011). Moreover, in vitro activation of neuronal TLR4 by LPS induces a strong expression of neuronal chemokines. These data revealed that neuronal TLR4 activation may play a central role in the onset of innate immunity during CNS infection or harm (Leow-Dyke et al., 2012). It is assumed that the cytokines produced by neurons may be just enough to recruit and activate local microglia without causing global brain inflammation. So it is perceived that also neuronal cells are able to mount an innate immune response. In fact, CNS neurons can be crucial sensors of infection since they respond to LPS by producing pro-inflammatory chemokines that in turn lead to activation of endothelial cells (Leow-Dyke et al., 2012). Interestingly, also ENS neurons respond to LPS and produce TNF-α (Coquenlorge et al., 2014). Regardless of PD being characterized by a slow and progressive degeneration of dopaminergic neurons in the SNpc, the cause of this neuronal loss is still poorly understood. Most relevant is the possibility that genetically determined age-dependent decline in mitochondrial function of the PD-typical pathologic cascade, gut bacteria or even their metabolites targeting the mitochondria, could activate innate immunity in dopaminergic neurons, due to the exposure of DAMPs, and in this way contribute to low-grade inflammation.

It was shown in PD cellular and animal models that mitochondrial network is highly fragmented. Mitochondrial fission is a prerequisite for the selective targeting of dysfunctional mitochondria for degradation by the lysosome in a process called mitophagy (Santos et al., 2015; Esteves et al., 2018). Nevertheless, it was recently proven that mitochondrial fission leads to the exposure of the inner membrane phospholipid, cardiolipin, which serves an important defensive function for the elimination of damaged mitochondria (Chu et al., 2013). Since cardiolipin is only found in mitochondrial and bacterial membranes it is considered a mitochondrial-derived DAMP that is detected by the Nlrp3 (He et al., 2016). NLR and TLR activation trigger the production of pro-inflammatory cytokines and AMPs (Lampron et al., 2013). Recently, it was also demonstrated that PD-associated SNCA proteins might be involved in the innate immunity response (Stolzenberg et al., 2017). It was proven that SNCA production mobilizes immune defenses against pathogens and the levels of mRNA of inflammatory cytokines in colonic biopsies from PD patients correlates with disease duration (Devos et al., 2013). Moreover, it was described that SNCA inserts in the mitochondrial membrane leading to mitochondrial dysfunction and also potentiate its fragmentation (Shen et al., 2014). These results seem to point to a positive feedback loop whereas mitochondrial dysfunction increases cardiolipin exposure, which in turn activates neuronal innate immunity. The question still remains on what is the role of SNCA under these conditions, one possibility being its involvement in the innate immunity pathway culminating in the potentiation of mitochondrial dysfunction.

# MITOCHONDRIA: TARGETS OF THE HUMAN MICROBIOME

Mitochondria play a key in the regulation of many immune functions through metabolic control, calcium homeostasis and ROS production, thus assisting host defenses against pathogens (Lobet et al., 2015). However, many bacteria have evolved several different types of effectors and mechanisms that target the mitochondria precisely as a strategy to circumvent mitochondriadependent surveillance. In light of the endosymbiotic theory it

can be argued that these mechanisms may have evolved in the ancestor bacteria as a strategy to gain competitive advantage in overpopulated environments. The mammalian gut harbors highly complex microbial communities with constantly balanced microbe–microbe and microbe–host interactions involving cooperative and competitive mechanisms to maximize the available resources and a shared co-existence (Sana et al., 2017). Some microbial species take advantage of nutrients produced by others in the community, while others target and kill their competitors by releasing toxic metabolites or by secreting effectors. We propose that gut dysbiosis increasingly associated to PD may, in the complex and competitive gut ecosystem, promote unrestrained and chronic production of microbial toxins that also target the mitochondria of ENS and CNS neurons (**Figure 1**). Indeed, PD biopsy studies confirmed the presence of LBs in organs innervated by the vagus nerve (Cersosimo et al., 2013), indicating this as the obvious route for disease progression between the gut and the brain. This led to the hypothesis that an exogenous toxin or pathogen can trigger the disease and spread via retrograde axonal vagus transport from the ENS to the CNS, and that the GI symptoms in the vast majority of PD patients are pre-motor manifestations of the disease (Braak et al., 2003a,b; Hawkes et al., 2007). Indeed, full truncal vagotomy significantly decreased the risk for subsequent PD, which suggests that the vagus nerve is indeed a key player in PD pathogenesis, again corroborating the involvement of an enteric pathogen or toxin in disease progression (Svensson et al., 2015). Recently, the development of SNCA pathology in genetically susceptible mice was shown to require the presence of gut microbiota, as evidenced by the limited pathophysiology observed in germ-free and antibiotictreated susceptible mice, although the effect of antibiotics on the mitochondria was not evaluated. Remarkably, administration of certain microbial metabolites to genetically modified germ free mice reproduced major features of the disease, comparable to PD induced in mice with a complex microbiota (Sampson et al., 2016). Although PD gut microbiota signatures have recently begun to emerge, their functional interpretation still remains largely elusive (Hill-Burns et al., 2017; Scheperjans, 2018).

Many toxins produced by eventual gut microbes can damage mitochondria. Some Clostridium difficile strains secrete toxins that inhibit the mitochondrial ATP-sensitive potassium channels, drive mitochondrial membrane hyper-polarization, apoptosis and disruption of the gut epithelial barrier (Matarrese et al., 2007). Vibrio cholera secretes a toxin (VopE) that inhibits mitochondrial network reorganization (Suzuki et al., 2014). Although these toxins have not been identified in commensal

FIGURE 1 | Schematic diagram indicating that neuronal mitochondria are primary gut bacteria targets. A dysbiotic gut harbors an inflammatory microbiota that could potentiate the production of microbial toxins. Either bacteria or bacterial toxins could activate innate immunity in the ENS and CNS through the vagus nerve, the gut-brain axis. Neuronal innate immunity is triggered by bacterial PAMPs or due to mitochondrial DAMPs. Mitochondrial damage may occur through the action of AMPs produced by the neuron as an arm of innate immunity activation or by the action of antibiotics produced by bacteria. PAMPs and mitochondria DAMPs activate the NLRs and TLRs leading to neuronal production of cytokines. These pro-inflammatory cytokines are released and activate low-grade inflammation through microglia. This chronic inflammation impacts neurons exacerbating AMPs production and mitochondrial damage.

microbiota, gene clusters for their biosynthesis may be present and drive the synthesis of related toxic compounds that may impact mitochondria. Hypotheses for the etiology of PD pathology based on chronic exposure to environmental toxins have been proposed as the underlying cause for PD onset (Tanner et al., 2011). In theory, the gut microbiota might also represent the "environmental" source of toxins to which the host would be exposed. In addition to the above examples of toxins produced by pathogenic gut microbiota, low molecular weight antibiotics of different classes are also known to induce mitochondrial dysfunction and oxidative damage with pathological consequences (Kalghatgi et al., 2013). Some classes of antibiotics target the bacterial protein synthesis machinery and inadvertently also the mitochondrial ribosomes (mitoribosomes) with potentially severe side effects in the host (Hobbie et al., 2008; Qian and Guan, 2009). The mitochondrial protein synthesis apparatus is similar to that of bacteria as a result of a shared origin and later endosymbiosis. Consequently, mitochondrial ribosomes are frequently unintended off-targets of antibiotics such as the aminoglycosides directed to bacterial ribosomes (Hobbie et al., 2008). Actually, sensitivity to a given antibiotic is likely a multifactorial trait but the genetic makeup of sensitive individuals, including the observed higher mutation rates in mtDNA accumulated as a consequence of aging, may also be a major contributing factor (Qian and Guan, 2009; Pacheu-Grau et al., 2013). Considering that most of the known antibiotics in use since the 1940's are of microbial origin and more prominently produced by members of the phyla Actinobacteria and Firmicutes as well as by some Fungi, is it reasonable to anticipate that some

# REFERENCES


members of the gut microbiota may possess the genetic resources to synthesize other antibiotics or antimicrobials including toxins that may target their distantly related counterparts, the mitochondria. Indeed, unknown antimicrobials remain hidden in the largely unexplored human microbiome (Donia et al., 2014). Although their identity and effects in mitochondria with possible damage eventually leading to activation of innate immunity have not been addressed, the enormous biosynthetic potential for metabolites impacting microbes clearly indicates that we haven't seen but the tip of the mediators regulating the complex microbial interactions within us, and which might foster mitochondrial damage and neurodegenerative processes.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

Work in our laboratories was supported by Fundação para a Ciência e a Tecnologia (FCT) and by EU-FEDER funding through the Operational Competitiveness Program—COMPETE grant UID/NEU/04539/2013, by the European Regional Development Fund, Centro 2020 Regional Operational Program (CENTRO-01-0145-FEDER-000012-HealthyAging2020), and by Prémio Santa Casa Neurociências Mantero Belard MB-40-2016.

response by activation of the enteric nervous system. J. Neuroinflammation 11, 202. doi: 10.1186/s12974-014-0202-7




Zhou, R., Yazdi, A. S., Menu, P., and Tschopp, J. (2011). A role for mitochondria in NLRP3 inflammasome activation. Nature 469, 221–225. doi: 10.1038/ nature09663

**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 Cardoso and Empadinhas. 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.

# New Perspectives on the Potential Role of Aquaporins (AQPs) in the Physiology of Inflammation

Rosaria Meli <sup>1</sup> , Claudio Pirozzi <sup>1</sup> and Alessandra Pelagalli 2,3 \*

<sup>1</sup> Department of Pharmacy, University of Naples Federico II, Naples, Italy, <sup>2</sup> Department of Advanced Biomedical Sciences, University of Naples Federico II, Naples, Italy, <sup>3</sup> Institute of Biostructure and Bioimaging, National Research Council (CNR), Naples, Italy

Aquaporins (AQPs) are emerging, in the last few decades, as critical proteins regulating water fluid homeostasis in cells involved in inflammation. AQPs represent a family of ubiquitous membrane channels that regulate osmotically water flux in various tissues and sometimes the transport of small solutes, including glycerol. Extensive data indicate that AQPs, working as water channel proteins, regulate not only cell migration, but also common events essential for inflammatory response. The involvement of AQPs in several inflammatory processes, as demonstrated by their dysregulation both in human and animal diseases, identifies their new role in protection and response to different noxious stimuli, including bacterial infection. This contribution could represent a new key to clarify the dilemma of host-pathogen communications, and opens up new scenarios regarding the investigation of the modulation of specific AQPs, as target for new pharmacological therapies. This review provides updated information on the underlying mechanisms of AQPs in the regulation of inflammatory responses in mammals and discusses the broad spectrum of options that can be tailored for different diseases and their pharmacological treatment.

### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

### Reviewed by:

Fulvio D'Acquisto, Queen Mary University of London, United Kingdom Stefania Marzocco, Università degli Studi di Salerno, Italy

#### \*Correspondence:

Alessandra Pelagalli alpelaga@unina.it

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 14 December 2017 Accepted: 31 January 2018 Published: 16 February 2018

#### Citation:

Meli R, Pirozzi C and Pelagalli A (2018) New Perspectives on the Potential Role of Aquaporins (AQPs) in the Physiology of Inflammation. Front. Physiol. 9:101. doi: 10.3389/fphys.2018.00101 Keywords: aquaporins (AQPs), inflammation, homeostasis, inflammatory diseases, animals

# INTRODUCTION

The inflammatory process and the complex mechanisms governing its regulation have always attracted scientific interest, focusing on the role of the various parties involved. Inflammation is a multifaceted phenomenon involving the body's physiological response to injury. The inflammatory tissue damage can be determined, among others, by trauma, heat, chemicals, toxins, and microbes. Both acute and chronic inflammatory process implicates an enormous expenditure of metabolic energy, loss of function, tissue damage, and destruction, involving different immune cells. Many signals orchestrate all inflammatory responses and a pivotal role is played by the immune system. Generally, it can be divided into two interconnected subsystems. The innate or non-specific immune system includes cells and processes that protect the host from infections by pathogen organisms, detecting and signaling the occurrence of infection (Takeda et al., 2003; Medzhitov, 2007). These signals are necessary to trigger the inflammatory cascade and to activate the adaptive immune response, the second subsystem. During activation, numerous substances released by the injured tissues induce alterations to the surrounding uninjured tissues. Simultaneously, cells of hematopoietic origin are recruited to damaged sites in order to resolve the inflammation and initiate tissue repair. Several of the products released from these cells, differing in origin and composition, constitute the real messengers of the inflammation process, and have been often viewed as drug targets.

It is interesting to examine the mechanisms and the mediators underlying the initiation of inflammation starting by the alteration of cellular and tissue homeostasis: an important role is played by the membranes that contribute to maintain the equilibrium in the microenvironment. Indeed, the cells normally modulate their internal environment in response to external changes; the loss of their ability to regulate fluid movement through the membrane lead to severe alteration of cell physiology (Loo et al., 1996). This cellular disorder affects, as a direct consequence, numerous biochemical processes, including alterations of protein structure and function and hence enzyme activity (Brocker et al., 2012). During homeostasis perturbation, as in conditions characterized by local or systemic inflammation, disease susceptibility appears and this phenomenon is associated with abnormal ion transport (Kotas and Medzhitov, 2015). In addition, many inflammatory signals, including cytokines and chemokines, can modify tissue/cell homeostasis by changing their sensitivity to homeostatic signals or by modifying gate or channel access (Medzhitov, 2008). The swelling of cells and tissues with a surplus of extracellular fluids (edema), is a clear sign of homeostasis disturbance and inflammation.

Aquaporins (AQPs) represent a new class of proteins extensively distributed on cell membrane that form pores and primarily act in several transporting and trafficking processes. Growing data have addressed their possible involvement in inflammation, participating as regulators of innate host defense at cell membrane level. AQPs could be involved in inflammasome activation regulating cell volume (Compan et al., 2012). Indeed, NALP3 inflammasome activation is induced by sodium overload and water influx, both features of cell swelling (Schorn et al., 2011), that represents a clear inflammatory response involving pro-inflammatory cytokine synthesis.

Here, we report recent data on the potential role of AQPs in the inflammatory process, as an adaptive response to the loss of cellular homeostasis or tissue damage. In addition, the function or alteration of AQPs in different inflammatory diseases or animal models is also discussed.

# AQUAPORINS

AQPs constitute a group of integral membrane proteins characterized by six transmembrane helices, that are organized in monomers, dimers, and tetramers, forming pores in the membrane of biological cells (Verkman, 2013). The first aquaporin discovered, initially called CHIP28 and later renamed as Aquaporin (AQP)1, was isolated in human erythrocytes (Denker et al., 1988; Preston and Agre, 1991) and in renal proximal tubule membranes. Since this discovery, many other proteins belonging to the same aquaporin family were described not only in mammalian cells, but also in all kingdoms of life, including bacteria (Kayingo et al., 2001), plants (Maurel et al., 2008), and fungi (Pettersson et al., 2005). Forming pores at the level of biological membranes, AQPs act as selective channels allowing the water transportation (aquaporins) and small molecules or solutes (aquaglyceroporins) (Agre et al., 1998) (**Figures 1A,B**). In the great family of aquaporins constituted by 13 proteins, a newly group named unhortodox aquaporins has been defined (Ishibashi et al., 2011). These last proteins are less understood and, in part, differ from the other groups for their structure and subcellular distribution. The widely distribution of AQPs in cells and tissues has increased the scientific interest toward their structural and functional characterization contributing to strength the idea that the water permeability is required for a variety of physiological processes. This observation is based on the consideration that water constitutes the major component of a living organism and that continuous exchange of water takes place in almost all body-tissues.

Accordingly, many data demonstrate that AQPs show a role in maintaining the homeostasis of many physiological processes related to secretive and absorptive activities of several tissues, such as kidney, salivary gland, lung, skin, sweat glands, and intestine (Laforenza, 2012; Pelagalli et al., 2016b). Moreover, these proteins participate in maintaining a constant water homeostasis in whole organism as essential prerequisite for their life. More recently, emerging evidence have confirmed that other cellular processes are controlled by AQPs, including cell adhesion, signaling, volume regulation and protein expression (Kitchen et al., 2015).

Moreover, the ability of these channel proteins to regulate cell volume, as well as cell migration and apoptosis, all related to the inflammation process, makes them useful tools to investigate their relevant role in physiological and pathological status.

# AQUAPORINS AND INFLAMMATION

Immune system and host-pathogen communication work together in host-pathogen interplay.

In particular, cells involved in inflammation can undergo modifications of osmotic microenvironment causing an increase in cell hydraulic permeability and size and thus, alterations in cytoskeletal structure (Maidhof et al., 2014). In some cases, weak bacterial stimuli can induce tissue impairment, but also a more intense response and damage that results in a chronic inflammation.

Many articles have highlighted the possible involvement of AQPs in the development of inflammatory processes also considering that several of them are expressed in cells of innate and adaptive immunity (Ishibashi et al., 1998; Moon et al., 2004). In particular, AQPs are involved in the phagocytic functions, and also in specific processes related to immune cells (i.e., activation and migration) (De Baey and Lanzavecchia, 2000; Jablonski et al., 2004; Zhu et al., 2011; Rabolli et al., 2014).

# Aquaporins and Phagocytic Functions of Immune Cells

Cell volume and shape modifications of macrophages promptly occur during phagosome development (Clarke et al., 2010; Tollis et al., 2010), modulating water transport and volume regulation

necessary for phagocytic cup formation. The participation of AQP3 in phagosome formation has been proposed (Zhu et al., 2011), postulating a mechanism similar to that observed in immature human dendritic cells (De Baey and Lanzavecchia, 2000). This mechanism identifies the glycerol transport as a key element for macrophage phagocytosis. Indeed, in a model of bacterial peritonitis a greater mortality in AQP3-deficient mice was observed respect to wild-type mice (Zhu et al., 2011). In peritoneal macrophages obtained from AQP3(−/ −) mice, water and glycerol permeability was reduced compared to those obtained from AQP3(+/ <sup>+</sup>) mice. Moreover, glycerol supplementation partially recovered the ATP content and the impairment of macrophage function in AQP3(−/ <sup>−</sup>) mice. AQP3 was identified as a novel key element in macrophage immune function, facilitating water and glycerol transportation, and its subsequent participation in phagocytic and migration activity. A crosstalk in glycerol and glucose metabolic pathways was also evidenced in AQP3 effects. Similar data for glucose and glycerol content modification were also observed in AQP3 (−/ −) keratinocytes (Hara-Chikuma and Verkman, 2008), identifying a possible new role for glycerol in macrophage energy metabolism.

Recent evidence has indicated an innovative AQP9 role for bacteria (Pseudomonas aeruginosa)–macrophage communication and for the sensing system in this process (Holm et al., 2015). In particular, it has been demonstrated that macrophages presenting AQP9 expression and infected by P. aeruginosa undergo modification of AQP9, and subsequent water fluxes, affect their shape and protrusive activity. These results confirm the role of AQP9 in macrophages during infection, clarifying how these proteins, participating as mediators to relationship between bacteria and macrophages, can affect the development of infection, inflammation, and the progression of the disease.

# Aquaporin Involvement in Migration of Immune Cells

The first demonstration of AQPs involvement in cell migration was reported by Loitto et al. (2002) indicating an impaired neutrophil migration after AQP9 blockage. Subsequently, other studies confirmed the AQP role in cell migration, showing AQP1 and AQP4 localization at the leading edge in migrating CHO cells and astroglial cells, respectively (Saadoun et al., 2005a,b). Other data have largely demonstrated that several AQPs facilitate migration of immune cells (Papadopoulos et al., 2008). In particular, chemokine-dependent T cell migration requires AQP3-mediated hydrogen peroxide uptake (Hara-Chikuma et al., 2012), regulating downstream intracellular signaling in cutaneous immune response (Miller et al., 2010). AQP3 is also expressed in macrophages (Zhu et al., 2011) and is regulated by several factors and conditions (TNFα, PPARγ, calcium, and low pH) (Horie et al., 2009; Jiang et al., 2011). These results demonstrate AQP3 involvement in the inflammatory process.

More recently, a study focusing on AQP1 has demonstrated its effect on macrophage migration, suggesting that some phenotypic and migratory modifications of these cells may be regulated by this water channel that results crucial for the switch of M0/M2 phenotype (Tyteca et al., 2015).

# POTENTIAL ROLE OF AQUAPORINS IN DIFFERENT MODELS OF INFLAMMATION

# Potential Role of Aquaporins in Models of Lung Injury and Inflammation

Numerous evidence clearly demonstrates that the mammalian lung expresses at least three AQPs whose role in lung damage or inflammation has been in part investigated (**Table 1**).

In particular, AQP1 is expressed in microvascular endothelia and pneumocytes (Nielsen et al., 1993; Folkesson et al., 1994). AQP4 and AQP5 were detected in airway and alveolar epithelial cells, respectively (Nielsen et al., 1997) and their distribution and physiological role has been reviewed in lung (Verkman et al., 2000). Thereafter, Liu et al. (2007) evidenced AQP3 expression in healthy and cancer lung, demonstrating that this AQP is extensively expressed in respiratory tract regulating water homeostasis. AQP3 seem to be implicated into tumor differentiation and processes related to clinical stage in lung adenocarcinomas (Liu et al., 2007). On this regard, the specific distribution of various AQPs in lung adenocarcinoma (AQP1, AQP3, and AQP5) has suggested that these proteins could be involved in different and distinct aspects of the cancerous process. In particular: (1) AQP1 localized on lung capillaries could be involved in the development of angiogenesis; (2) AQP3 could participate in several regulatory pathways, while (3) AQP5 could promote cell proliferation and tumor invasion (Wang et al., 2015).

All the data available demonstrate an interesting contribution given by lung AQPs in regulating fluid trafficking between the air space and cellular, interstitial or vascular compartments. In addition, evidence shows that expression of AQPs is modulated by growth factors, inflammatory mediators, and osmotic stress in the respiratory physiology (King et al., 1997; Borok et al., 1998; Towne et al., 2000).

However, after pulmonary infection, numerous processes altering the lung physiology occur (Peteranderl et al., 2017). In particular, autocrine and/or paracrine mediators cause several pathophysiological modifications of the alveolar–capillary barrier and of epithelial ion channel and pump expression altering vectorial ion gradient. Among these mediators, proinflammatory cytokines (TGF-β, TNFα, interferons, IL-1β) are released after infection by different bacteria (i.e., Streptococcus pneumoniae, Klebsiella pneumoniae, or Mycoplasma pneumonia) or virus. These inflammatory players can induce edema formation and reduce alveolar fluid clearance, modifying the expression of: (i) epithelial Na,K-ATPase, (ii) epithelial ion channels, and (iii) fibrosis membrane conductance regulator (Peteranderl et al., 2017). This alteration can represent also the result of altered gas exchange involved in the modulation of the alveolar–capillary fluid homeostasis modulated by AQP5 and AQP4 (Musa-Aziz et al., 2009) or AQP1 (Al-Samir et al., 2016).

Towne et al. (2000) demonstrated a marked and significant reduction of AQPs (AQP1 and AQP5) in lung directly correlated with an increase in tissue wet-to-dry weight ratios in a model of mice infected with adenovirus. The reduction in AQP1 and AQP5 expressions was noticed at sites distant from that of infection, suggesting a humoral regulation of these AQPs. Albeit, these data did not define the precise activity of AQP1 or AQP5 in infective process, they give an important contribution in the field, confirming AQP involvement in the pathophysiological processes of the respiratory tract (King et al., 2000).

Further data supports these findings showing that Th2 cytokines and IL-4, both involved in mucin gene expression, are down-regulated in AQP5 knockout mouse (Karras et al.,


DSS, Dextran sodium sulfate; LPS, lipopolysaccharide; TNBS, 2,4,6-trinitrobenzene sulphonic acid; 5-FU, 5-floruracil; IHC, Immunohistochemistry; MPTP, 1-methyl-4-phenyl-1,2,3,6 tetrahydropyridine; PD, Parkinson disease; RT-PCR, real time-PCR; TEM, transmission electron microscopy; WB, western blotting.

2007), thus suggesting the contribution of this channel protein in effective Th2-driven responses to allergens. More recently, the relationship between the AQP5 deletion and elevated IFN-α and IL-2 production was evidenced, indicating that this protein acts in the shift from Th2 toward Th1 response in a murine model of mucous hyperproduction during antigen-induced airway inflammation (Shen et al., 2011). It is well-known that Th2 cytokines operate in these pathologies with different mechanisms such as eosinophil recruitment, airway hyper-responsiveness and mucus hypersecretion (Tomkinson et al., 2001; Walter et al., 2001).

More recently, AQPs were described as potential downstream targets of altered gene expression in several murine models of induced asthma (by allergen ovalbumin or IL-13) (Krane et al., 2009). AQP3 and AQP5 gene expressions appear as quite early modification in the lung response to mIL-13 induced airway constriction. Indeed, preclinical data showed that AQP3 potentiates allergic airway inflammation in OVA-induced asthma (Ikezoe et al., 2016). A significant reduction of airway inflammation observed in AQP3 deficient mice, respect to wildtype mice, was associated to in vivo and in vitro results, showing that an increase of chemokines (i.e., CCL24 and CCL22) was induced by AQP3 through a control mechanism of the cellular H2O<sup>2</sup> production in M2 polarized alveolar macrophages (Ikezoe et al., 2016).

# Involvement of Aquaporins in Neuroinflammation

Accumulating evidence in humans and animals supports physiological and pathological role of AQPs expression and function in the nervous system (**Table 1**). The potential contribute of these channel proteins in the neuroinflammation has been widely investigated, examining several diseases caused by a failure of innate immunity, such as neuromyelitis optica (NMO) and multiple sclerosis (Oklinski et al., 2016). The channel protein AQP4 is expressed in astrocytes in CNS and regulates the brain water flux, neuroexcitation, and astrocyte migration (Verkman et al., 2006). In fact, lesions observed in NMO patients show that specific autoantibodies targeting AQP4 are expressed on astrocytic membrane and thus, alter cell functions through different mechanisms. Among these, activation of complement, cellular cytotoxicity mediated by an antibodydependent mechanism, or both mechanisms were evidenced (Bennett et al., 2009; Bradl et al., 2009). AQP4 represents a specific target for NMO-IgG (Fukuda and Badaut, 2012; Hinson et al., 2012). Moreover, it has been clearly established that APQ4 is involved in neuroinflammation (i.e., water intoxication and ischemic stroke), evidencing a reduction of brain edema and swelling of pericapillary astrocytic foot processes in AQP4 deficient mice (Manley et al., 2000). These results indicate a key role for AQP4 in controlling brain water transport, and propose

that AQP4 blockage could be represent a new therapeutic strategy for ameliorate conditions of cerebral edema at the basis of several brain pathologies.

Papadopoulos et al. (2008) reviewed evidence that AQPs facilitate cell migration both in vitro and in vivo in a variety of cell types such as endothelial cells and astrocytes. In particular, AQP1 deletion diminishes endothelial cell migration, limiting tumor angiogenesis and growth, on the other hand, AQP1-expressing tumor cells have enhanced local infiltration and metastatic effects. AQP4 deletion reduces the migration of reactive astrocytes, damaging glial scarring after brain injury. AQPs modulating cell migration can be implicated in several processes such as angiogenesis, tumor metastasis, wound healing, glial scarring, and other events requiring prompt cell movement. The mechanisms by which AQPs act, is probably related to actin polymerization/depolymerization and variation of transmembrane ionic fluxes and osmolality (Papadopoulos et al., 2004). AQPs could thus increase osmotic water flow in cell protrusions of the plasma membrane that shape during cell migration. Indication for involvement of AQPs in a similar process, i.e., secretory vesicle exocytosis, has been previously described (Cho et al., 2002; Sugiya and Matsuki, 2006).

Pro-inflammatory role for AQP4 was confirmed by Li et al. (2011) demonstrating that LPS, administered by intracerebral injection, induced greater neuroinflammation in wild-type than in AQP4-knockout mice, and cytokine (TNFα and IL-6) production was reduced in astrocyte cultures obtained from AQP4-knockout mice. These data suggested that astrocyte swelling and cytokine release are AQP4-dependent in this cell type. In this way AQP4 is involved in the communication between microglia and astrocyte and their functions (**Figure 2**; Ikeshima-Kataoka, 2016). For all these reasons, the decrease in AQP4 water transport or AQP inhibitors could play a protective role in neuroinflammation, modulating brain edema and cell migration.

# Aquaporins in Bowel Diseases

AQPs expression and their relevant role in physiological and pathological processes have been evidenced in gastrointestinal tract of human and mammalian species (Laforenza, 2012; Pelagalli et al., 2016b). Gut involvement of AQPs was determined in many mechanisms that mediate water transport (intestinal permeability, and fluid secretion/absorption). Regarding AQPs regulation, it is known that different substances (hormones and dietary components) act modifying their expression and thus alter fluid homeostasis and several local mechanisms (De Luca et al., 2015; Squillacioti et al., 2015; Pelagalli et al., 2016a,b). In this regulation, a role for cAMP was also addressed (Hamabata et al., 2002).

Inflammatory bowel diseases (IBDs) are inflammatory relapsing diseases of gastrointestinal tract with a chronic aberrant stimulation of immune system, gut inflammation and leakage of fluid, solutes, and lipids in bowel mucosa, involving gut microflora (Mayer, 2010). IBDs have been also characterized for a dysregulation in electrolyte and water transport with resultant alteration of permeability and diarrhea (Dunlop et al., 2006; Zhou et al., 2009; Martínez et al., 2012), for this reason the relationship between colitis and AQPs have been extensively investigated (**Table 1**). In particular, the remarkable increase of intestinal membrane permeability observed in these diseases has suggested the participation of AQPs (**Figure 3**). Moreover, it is also well-known that cytokines, as important signaling molecules of the intestinal immune system, are correlated to the severity of inflammation (Kim, 2011; Strober and Fuss, 2011). Among them, TNFα and IL-1β play a pivotal role in gut inflammatory processes directly, influencing intestinal epithelial tissue behaving as a frontline between genetic, environmental, and immunological factors (Hering et al., 2012; Keita and Soderholm, 2012).

In 2007, Guttman et al. evidenced for the first time the direct correlation between AQPs and diarrhea, defining AQP contribution to diarrhea caused by attaching and effacing bacteria (i.e., enterohemorrhagic Escherichia coli and enteropathogenic E. coli) pathogenesis (Guttman et al., 2007). Very recently, Chao and Zhang (2017) evidenced a possible relationship between AQPs (AQP1, AQP3 and AQP8) expression and NF-κB pathway in a model of IBD. Numerous findings indicate NF-κB pathway as the main regulator of several processes (pro-inflammatory cytokine production, leukocyte recruitment, or cell survival), and its involvement in relation to AQPs (Ito et al., 2006; Hasler et al., 2008). Regarding the possible link between TNFα and AQPs, it has been evidenced that TNFα acts modulating AQP3 expression (down- or up-regulation), according to the cell type involved, through different signaling pathways (Tancharoen et al., 2008; Horie et al., 2009). More recently, it was demonstrated that AQP3 downregulation is mediated by the inhibition of constitutive transcriptional activity at the AQP3 promoter in HT-29 cells (Peplowski et al., 2017). In another cell line (CMT93) (Dicay et al., 2015) demonstrated that IFNγ limits epithelial AQP1 expression through the activation JAK/STAT3 pathway. In addition, the authors provided data that demonstrated a role for IRF-2 in the basal expression of AQP1, and that IFNγ was able to regulate AQP1 expression (Dicay et al., 2015).

Colitis, as gut IL-dependent inflammation, is mediated by a Th1 cell response and AQPs affect it, interfering in the proliferative activities of colon epithelial cells. In AQP3 null mice, dextran sodium sulfate (DSS) induced severe colitis, characterized also by hemorrhage in colon, marked epithelial cell loss and death after 3 days, while wild-type mice showed remarkably less severe colitis, surviving to >8 days (Thiagarajah et al., 2007). Moreover, in AQP3 null mice, cell proliferation was greatly reduced. A new role for AQP3 in enterocyte proliferation was likely related to its glycerol-transporting function. In fact, oral glycerol administration largely enhanced survival of AQP3 null mice and reduced the severity of colitis. These data identify AQP3 deficiency as the cause of a reduced AQP3-facilitated glycerol transport, compromising cell metabolism.

Recently, a detailed study focused on the possible role of AQPs in both severe IBDs (Crohn's disease and ulcerative colitis), demonstrating a different distribution of these channel proteins in the gut and the existence of a direct relationship between

intestine inflammation and physiological water/solute trafficking and regulation (Ricanek et al., 2015).

# Aquaporins and Bone and Cartilage Diseases

Recently, data on several inflammatory diseases affecting bone and cartilage has involved AQPs. It is well-known that the principal pathological phenomena associated with rheumatoid arthritis (RA) are characterized by enormously elevated levels of inflammatory cytokines secreted by activated B and T cells causing damage of the cartilage and bone. At the same time, different AQPs have been found in cartilage cells where they regulate the traffic of ions and molecules (Mobasheri et al., 2004) and thus, regulate the cartilage physiology.

In particular, Nagahara et al. (2010) evidenced that in synovial tissues from patients with osteoarthritis (OA) and RA, TNFα could regulate either AQP9 mRNA and protein expression (Nagahara et al., 2010). This result has suggested a particular role for cytokines that, altering the activity of glucose transporters, important for chondrocyte metabolism (Mobasheri et al., 2002; Richardson et al., 2003), could influence AQP function. According to this mechanism, AQP1 on chondrocyte membrane could act as regulator of metabolic or extracellular matrix water (Trujillo et al., 2004), suggesting that chondrocytes mightrespond to changes in their ionic and osmotic environment modifying volume regulatory behavior. The direct involvement of AQPs in the pathogenesis of this disease was investigated in a model of OA cartilage evaluating AQP1 mRNA by RT-PCR and demonstrating that up-regulation of AQP1 was related to the chondrocyte apoptosis (**Table 1**) (Gao et al., 2011).

Recently, for the first time, Cai et al. (2017), identified AQP4 as potential responsible of RA pathogenesis in adjuvant-induced arthritis (AIA) rat model. The results showed that the reduced mRNA levels of collagen type II and aggrecan, observed in cultured AIA chondrocytes, were reverted by acetazolamide treatment. AQP4 inhibition obtained with acetazolamide promoted extra cellular matrix production of AIA chondrocytes in vitro (Cai et al., 2017).

# INFLAMMATORY DISEASES IN DOMESTIC ANIMAL SPECIES: THE INVOLVEMENT OF AQUAPORINS

In the recent last years, many studies have examined inflammatory diseases in domestic animal species, albeit available data are limited considering several limitations for ethical problem in these species respect to laboratory animals. However, evidence indicate that animals as well as humans can suffer of several inflammatory diseases whose possible mechanisms are not always well-defined. Even if, different diseases have been investigated in domestic species, few are the studies regarding the possible link between AQPs and inflammation. The most investigated species is the dog most likely because it is very similar to humans.

In particular, inflammation-based diseases, in organs and systems, like gut, central nervous system, and lung have been investigated in dog species with the perspective to clarify their pathophysiology finalized to adequate therapeutic protocols. On this regard Cerquetella et al., 2010) studied some particular aspects regarding dysbiosis networks in dog IBDs, evidencing differences and similarities with humans. The results of this study providing new important contributes for translational medicine require further development of scientific research for understanding differences between dog and human in some bacteria species. In addition, an interesting study showed an increase of AQP4 and IL6 levels in cerebrospinal fluid (CSF) of dogs affected by idiopathic communicating internal hydrocephalus and a reduction of these proteins after ventriculoperitoneal shunting (Schmidt et al., 2016). In addition, a study on acute respiratory distress syndrome in beagle dogs showed a clear inflammatory process characterized by TNFα increase that can facilitate the secretion of cytokines, such as IL-1A, IL-6, and IL-10 (Zhao et al., 2012). Moreover, the decreased AQP1 and AQP5 expression observed as possible consequence of pulmonary capillary membrane barrier damage suggests their possible involvement in the regulation of these fluid trafficking mechanisms along this membrane.

Moreover, in avian species AQPs expression has been investigated at level of nasal gland and its fluid secretion. In particular, AQP1 and AQP5 seem to play a role in modulating nasal fluid secretion that it is always hypertonic, differently from vertebrates. (Müller et al., 2006).

# CONCLUSIONS

Described evidence suggest that AQPs are not only simple transporting proteins, since their dysregulation occurs in immune and epithelial cells in response to infectious and inflammatory stimuli.

The discovery of AQPs involvement in inflammation certainly can contribute to the knowledge of the complex mechanisms regulating host-pathogen communications. Overall, it seems clear that AQPs are new possible candidates as therapeutic potential target in modulating edema, cell migration and, inflammatory cytokines and mediators release.

Future studies are needed to better understand the molecular mechanisms driving osmotic stress-induced inflammatory response and to clarify the unravel signaling pathways involved in the regulation of AQPs functions. The acquisition of these basic skills could help to clarify the importance of osmotic imbalances not only in inflammation and inflammation based-diseases but also in cancer.

# AUTHOR CONTRIBUTIONS

RM: provided intellectual input, corrected drafts of the review and refined final version; CP: conceived some aspects of review and corrected its drafts; AP: conceived the review, wrote the original manuscript, corrected several drafts and refined final version.

# FUNDING

AP declare that manuscript realization has been supported in part by funds from the Department of Advanced Biomedical Sciences, University of Naples Federico II.

# ACKNOWLEDGMENTS

The work was supported in part by research funds from the Department of Advanced Biomedical Sciences, University of Naples "Federico II," Naples, Italy. We thank Prof. Arturo Brunetti for financial support and critical discussions.

# REFERENCES


but is inhibited after apoptotic volume decrease. Am. J. Physiol. Cell Physiol. 286, C975–C985. doi: 10.1152/ajpcell.00180.2003


molecular alterations in the tight junction signaling pathway that are associated with mucosal pathobiology and clinical manifestations. Am. J. Gastroenterol. 107, 736–746. doi: 10.1038/ajg.2011.472


an important role for H2O2? Eur. J. Gastroenterol. Hepatol. 20, 555–560. doi: 10.1097/MEG.0b013e3282f45751


**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 Meli, Pirozzi and Pelagalli. 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.

# L-Threonine Supplementation During Colitis Onset Delays Disease Recovery

Joana Gaifem1,2, Luís G. Gonçalves<sup>3</sup> , Ricardo J. Dinis-Oliveira4,5,6, Cristina Cunha1,2 , Agostinho Carvalho1,2, Egídio Torrado1,2, Fernando Rodrigues1,2, Margarida Saraiva7,8 , António G. Castro1,2 and Ricardo Silvestre1,2 \*

<sup>1</sup> Life and Health Sciences Research Institute, School of Health Sciences, University of Minho, Braga, Portugal, 2 ICVS/3B's – PT Government Associate Laboratory, Guimarães, Portugal, <sup>3</sup> Instituto de Tecnologia Química e Biológica António Xavier, Universidade NOVA de Lisboa, Oeiras, Portugal, <sup>4</sup> IINFACTS – Institute of Research and Advanced Training in Health Sciences and Technologies, Department of Sciences, CESPU, CRL, University Institute of Health Sciences, Gandra, Portugal, <sup>5</sup> UCIBIO, REQUIMTE, Laboratory of Toxicology, Department of Biological Sciences, Faculty of Pharmacy, University of Porto, Porto, Portugal, <sup>6</sup> Department of Public Health and Forensic Sciences, and Medical Education, Faculty of Medicine, University of Porto, Porto, Portugal, <sup>7</sup> Instituto de Investigação e Inovação em Saúde, Universidade do Porto, Porto, Portugal, <sup>8</sup> Instituto de Biologia Molecular e Celular, Universidade do Porto, Porto, Portugal

#### Edited by:

Maria Teresa Cruz, University of Coimbra, Portugal

#### Reviewed by:

Vicente Lahera, Complutense University of Madrid, Spain Diana Jurado Serra, University of Coimbra, Portugal

\*Correspondence:

Ricardo Silvestre ricardosilvestre@med.uminho.pt

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 09 March 2018 Accepted: 17 August 2018 Published: 05 September 2018

#### Citation:

Gaifem J, Gonçalves LG, Dinis-Oliveira RJ, Cunha C, Carvalho A, Torrado E, Rodrigues F, Saraiva M, Castro AG and Silvestre R (2018) L-Threonine Supplementation During Colitis Onset Delays Disease Recovery. Front. Physiol. 9:1247. doi: 10.3389/fphys.2018.01247 Dietary nutrients have emerged as potential therapeutic adjuncts for inflammatory bowel disease (IBD) given their impact on intestinal homeostasis through the modulation of immune response, gut microbiota composition and epithelial barrier stability. Several nutrients have already been associated with a protective phenotype. Yet, there is a lack of knowledge toward the most promising ones as well as the most adequate phase of action. To unveil the most prominent therapy candidates we characterized the colon metabolic profile during colitis development. We have observed a twofold decrease in threonine levels in mice subjected to DSS-induced colitis. We then assessed the effect of threonine supplementation in the beginning of the inflammatory process (DSS + Thr) or when inflammation is already established (DSS + Thr D8). Colitis progression was similar between the treated groups and control colitic mice, yet threonine had a surprisingly detrimental effect when administered in the beginning of the disease, with mice displaying a delayed recovery when compared to control mice and mice supplemented with threonine after day 8. Although no major changes were found in their metabolic profile, DSS + Thr mice displayed altered expression in mucin-encoding genes, as well as in goblet cell counts, unveiling an impaired ability to produce mucus. Moreover, IL-22 secretion was decreased in DSS + Thr mice when compared to DSS + Thr D8 mice. Overall, these results suggest that supplementation with threonine during colitis induction impact goblet cell number and delays the recovery period. This reinforces the importance of a deeper understanding regarding threonine supplementation in IBD.

Keywords: IBD, threonine, DSS-induced colitis, goblet cells, metabolomics, IL-22, mucin

# INTRODUCTION

fphys-09-01247 September 4, 2018 Time: 9:42 # 2

Inflammatory bowel disease (IBD) is a complex debilitating disorder of the gastrointestinal tract which comprises both Crohn's disease and ulcerative colitis. Despite the unclear etiology of IBD, several factors have been accounted as key for the development of the disease, such as genetics, immune system and environmental factors, namely diet and gut microbiota composition (Khor et al., 2011).

Dietary supplementation has emerged as a promising therapeutic practice in the prevention and treatment of IBD (Durchschein et al., 2016). Recent evidence has revealed that fiber-enriched diets promote protection against IBD development, since dietary fiber is mainly fermented by intestinal microbiota into short-chain fatty acids (SCFAs), such as butyrate, acetate and propionate (den Besten et al., 2013). The protective properties of these metabolites are widely described by their impact on immune cell activation and epithelial barrier stability (Kelly et al., 2015; Macia et al., 2015), with decreased levels of SCFAs being found in colon samples from IBD patients (Huda-Faujan et al., 2010). Other studies have also pointed out several specific amino acids that can improve intestinal homeostasis, mainly by boosting mucosal healing and regeneration. For instance, glutamine is known to promote protection in dextran sulfate sodium (DSS)- and 2,4,6-trinitrobenzenesulfonic acid (TBNS)-induced intestinal inflammation, acting via NF-κB downregulation (Kretzmann et al., 2008). Colitic mice orally administered with glutamine displayed suppressive Th1/Th17 immune responses and subsequently decreased inflammation when compared to mice fed with regular diet (Hsiung et al., 2014). Other amino acids have been also associated with a protective phenotype against colitis. Using distinct animal models of colitis, diets enriched in threonine, serine, proline and cysteine, given before and throughout disease development, have been shown to restore mucin synthesis and stabilization of gut microbiota (Faure et al., 2006). Similar findings were observed with the administration of a mixture of threonine, methionine and monosodium glutamate after colitis induction (Liu et al., 2013).

Several studies have so far addressed whether administration of specific nutrients may arise as a prophylactic and/or therapeutic approach. However, there is a lack of knowledge toward the most promising and adequate phase of action. Thus, we investigated the metabolic profile of mice developing colitis, aiming to identify variation of metabolites during inflammation. The identification of the most attractive potential targets for therapy and the definition of a time range more prone to potentiate the effects of their supplementation may be relevant for future applications in IBD prophylaxis and therapy.

# MATERIALS AND METHODS

# Animals

Seven to nine-week old C57BL/6J male mice were purchased from Charles River Laboratories and housed in i3S animal facilities, under pathogen free conditions, with food and water ad libitum. All experimental procedures were approved by the i3S Animal Ethics Committee and licensed by the Portuguese National Authority for Animal Health (DGAV) with reference 014811/2016-07-13.

# Colitis Induction

Dextran sulfate sodium (DSS; TdB Consultancy; 2% (w/v), molecular weight approximately 40000 Da) was administered in drinking water ad libitum for 5 days. Clinical signs of colitis were monitored daily and scored as a disease activity index (DAI; **Supplementary Table S1**).

# L-Threonine Administration

Mice were divided into DSS (control), DSS with L-Threonine (DSS + Thr) and DSS followed by L-Threonine administration at day 8 (DSS + Thr D8), as shown in **Figure 2A**. L-Threonine [Thr; Sigma–Aldrich; 0.166% (w/v) corresponding to 250 mg/Kg/day] was given in the drinking water ad libitum. The dose was chosen according to the daily intake in previous studies with rodents (Faure et al., 2006; Liu et al., 2013) and to a high-threonine human supplementation study (Pencharz et al., 2008). The safety of our protocol measuring biomarkers of renal and liver damage was evaluated to confirm the absence of toxicity (**Supplementary Figure S1**). Similar fluid intake was found among all groups.

# Metabolomic Analysis by Nuclear Magnetic Resonance (NMR)

Methanol/water extracts of colon were analyzed at an UltrashiedTM 800 Plus (Bruker) spectrometer as described in Graça et al. (2017). Metabolite concentrations were performed by integration of 1H-NMR resonances using TSP as reference.

# Quantitative Real-Time PCR (qPCR)

Total RNA was isolated from colonic samples (TripleXtractor, Grisp). As DSS inhibits both polymerase and reverse transcriptase activities, RNA was purified with lithium chloride, as in Viennois et al. (2013). qPCR was performed as described in Correia et al. (2017). The list of primers used is in **Supplementary Table S2**.

# Histology and Goblet Cell Count

Colons were fixed in 10% buffered formalin (Sigma–Aldrich) and embedded in paraffin. Sections of 5 µm were stained with hematoxylin/eosin and Alcian Blue/Periodic acid-Schiff. Goblet cell number was assessed for each experimental condition in a blinded fashion. Only crypts cut longitudinally from crypt opening to bottom were analyzed.

# Cytokine Quantification

Colonic explant cultures were performed as previously described (McNamee et al., 2011). Cytokine quantification was performed in supernatants by ELISA (Biolegend). Tissue explants were homogenized and total protein was measured using Bradford assay. The concentration of secreted cytokines in the supernatant was normalized to total tissue protein and expressed as picogram of cytokine per µg of total tissue protein.

Statistical Analysis

experiment is shown out of two. <sup>∗</sup>p < 0.05.

Metabolite differences were evaluated by ANOVA in R statistical software. Partial least squares – discriminant analysis (PLS-DA) models were performed in SIMCA software. Other statistical analyses were carried out with GraphPad Prism (version 6.01). For multiple group comparisons one-way ANOVA test with a Tukey multiple-comparison posttest was performed, while for multiple group comparisons with repeated measures twoway ANOVA test with a Tukey multiple-comparison posttest was applied. Data are presented as mean ± standard deviation (SD). Statistically significant values are: <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

# RESULTS AND DISCUSSION

To characterize the colon metabolic profile during colitis development, metabolites present on colonic extracts of mice prior (day 0) and after 5 days of DSS-induced colitis were analyzed by NMR (**Figure 1A**). These time points were selected since it allows the comparison of a homeostatic profile (day 0) against a period with established inflammation and lesion, yet reversible and treatable (day 5). Multivariate analyses of the metabolomic data only demonstrated minor alterations between profiles for day 0 and day 5 of colitis development (**Figure 1B**). However, univariate analyses performed on the metabolites allow to discriminate ADP and particularly threonine as significantly altered from day 0 to day 5 (**Figures 1C,D**). Among essential amino acids, threonine has a prominent role in maintaining a healthy gut. Threonine is able to generate the main three SCFAs, namely acetate, butyrate and propionate (Neis et al., 2015). In fact, it has been previously identified several biosynthetic genes for threonine metabolism in the human gut microbiota, suggesting the relevance of this amino acid for microbiota biology (Abubucker et al., 2012). SCFAs are described as important modulators of immune response, since they are ligands for G-protein-coupled receptor 43 (GPR43) that is expressed by immune cells on the lamina propria, such as regulatory T cells, regulating the proinflammatory responses in the intestine (Bollrath and Powrie, 2013). Moreover, threonine is vastly metabolized in the intestine for mucin synthesis (Faure et al., 2005). These proteins are paramount in intestinal stability, since the mucus layer in the colonic outer layer prevents the direct contact of luminal microorganisms with the epithelium (Johansson et al., 2014). It has also been suggested that threonine requirements are increased under pathological settings to maintain proper intestinal function, such as production and formation of the mucus layer (Remond et al., 2009). Therefore, by participating in the mucus layer synthesis and production of anti-inflammatory SCFAs, threonine metabolism by gut microbiota proves to be essential for gut barrier integrity and function. Our data show that threonine levels drop 2-fold during colitis development until day 5 (**Figures 1C–E**). Therefore, we hypothesized that threonine supplementation during active inflammatory disease could help to restore the intestinal homeostasis and thus present some therapeutic potential.

Previous studies have investigated threonine in combination with other amino acids as a potential candidate for therapy against colitis (Faure et al., 2006; Liu et al., 2013). Nevertheless,

not only single threonine supplementation was not evaluated before, but also there is scarce evidence regarding the most adequate time frame for its supplementation in colitis treatment. Accordingly, we addressed threonine supplementation in two distinct phases: (1) in parallel with the initiation of the inflammatory process, i.e., simultaneously with DSS administration (DSS + Thr), and (2) on day 8, when inflammation is already established (DSS + Thr D8). Mice subjected to DSS-induced colitis but without threonine supply were used as control (DSS) (**Figure 2A**). We observed that the three groups developed colitis with a similar progression profile. Nevertheless, after this time point, the recovery profile of colitic mice supplemented with threonine in the drinking water (DSS + Thr) was slower than that of the other two groups, showing statistically significant differences from day 10 to the final day of experiment when compared to control group (DSS). Besides, DSS + Thr mice had also a distinctive DAI score at day 11 and 12 when compared with mice that only received threonine after day 8 (DSS + Thr D8) (**Figure 2B**). Despite the divergent

phenotype, no major differences were found in colon length neither in the intestinal permeability (**Supplementary Figure S1**).

Threonine administration was surprisingly detrimental for disease recovery when given at the setting of the inflammatory process. To understand this phenotype, we first evaluated the potential alterations in the colon metabolic profile among the different groups. No significant differences were found between most of the metabolites. Particularly, the levels of the major SCFAs (i.e., acetate, butyrate, and propionate), normally associated with a protective phenotype, were similar among groups. Only succinate levels, which is as an important marker of inflammation promoting IL-1β induction in inflammatory contexts (Tannahill et al., 2013), were found to be markedly decreased in DSS + Thr D8 mice (**Figure 2C**). No major alterations were observed in the inflammatory infiltrate profile, tissue organization and hepatic and renal toxicity serum biomarkers (**Supplementary Figure S1**).

Threonine plays a major role in mucin synthesis and consequently in the formation of the mucus layer. Indeed, lack of threonine is also known to impair intestinal paracellular permeability and is associated with fewer goblet cells and mucus synthesis (Faure et al., 2005; Mao et al., 2011). The mucus layer serves as a barrier against microbial translocation to the lamina propria and therefore its integrity is paramount for intestinal homeostasis. When we analyzed the expression of mucin-encoding genes, we found that Muc2 expression is decreased in DSS + Thr mice when compared to mice that only received threonine at day 8 (DSS + Thr D8) (**Figure 2D**). Muc2 encodes for the oligomeric mucus gel-forming mucin 2 protein that is the major responsible for mucus synthesis. Indeed, impairment or total absence of the mucus layer is associated with severe colitis, as observed in Muc2-deficient mice (Van der Sluis et al., 2006). We also found that DSS + Thr mice display higher expression of Muc1, which has shown to contribute to intestinal inflammation and colon cancer progression (Baldus et al., 2004; Takahashi et al., 2015). We next quantified the number of goblet cells of the colonic mucosa. These are a secretory epithelial cell lineage found in both the small and the large intestines, whose major function is the production of mucus. By analyzing colon slices stained with Alcian Blue/Periodic acid-Schiff, we observed that DSS + Thr mice displayed significantly fewer goblet cells when compared to both DSS and DSS + Thr D8 (**Figures 2E,F**). Therefore, our results suggest that an alteration in mucus synthesis due to threonine administration during the onset of disease may impact intestinal integrity, by delaying the recovery of disease. Faure et al. (2006) have demonstrated that supplementation with diet containing higher doses of amino acids, including L-threonine, lead to an increase in goblet cell number, regulated mucin production in the colon and restored microbiota composition after DSS treatment in rats. Notwithstanding, not only the animal model is different, but also L-threonine was given before colitis induction, which may be underlying the distinctive outcome.

Previous studies have linked several cytokines to mucus production in the intestine (Parks et al., 2015). To examine the immunological profile of the three groups, cytokine levels in the colon were quantified. No major changes were observed between the groups for interleukin (IL)-1β, IL-12p70, IL-10, IL-17A/F and granulocyte-macrophage colony-stimulating factor (GM-CSF) levels (**Supplementary Figure S2**). Notwithstanding, the amount of IL-22 was significantly decreased in DSS + Thr mice when compared to DSS + Thr D8 mice (**Figure 2G**). IL-22 is a member of the IL-10 family of cytokines and has been vastly studied in the context of intestinal homeostasis. It can be produced by several cell types, such as T helper (Th1) 1, Th17, Th22 and innate lymphoid cells (ILCs), and present several roles in the gastrointestinal tract, such as tissue regeneration and maintenance of the intestinal epithelial barrier (Rutz et al., 2013). Thus, the decreased IL-22 levels may be associated with delayed recovery of the intestinal balance.

Overall, our data demonstrate that supplementation of threonine during colitis induction impairs goblet cell number, with concomitant decreased Muc2 expression and IL-22 production. These variations are likely to be the cause of delayed recovery observed in this situation. Interestingly, these effects are not seen when threonine is administered once colitis is established. Acute DSS-induced colitis is known to promote gut microbial dysbiosis (Munyaka et al., 2016). Threonine is metabolized by some intestinal commensal bacteria, leading to the production of several metabolites used for intestinal maintenance and to mediate immune responses (Neis et al., 2015). Thus, threonine supplementation during induction of colitis may impact differently the colonic microbiota populations present during the onset and upon the establishment of inflammation, having ultimately distinct effects in intestinal function. Further understanding of the mechanisms underlying threonine supplementation may give new insights on how dietary nutrients modulate the dynamic balance between microbiome, immune response and barrier function.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of European Council Directive (2010/63/EU) guidelines that where transposed into Portuguese law (Decree-Law n.◦ 113/2013, August 7th), i3S Animal Ethics Committee and licensed by the Portuguese National Authority for Animal Health (DGAV). The protocol was approved by the i3S Animal Ethics Committee and licensed by the Portuguese National Authority for Animal Health (DGAV) with reference 014811/2016-07-13.

# AUTHOR CONTRIBUTIONS

JG, ET, CC, AgC, FR, MS, AnC, and RS designed the experiments. JG, LG, RD-O, and RS performed the experiments. JG, LG, RD-O, and RS analyzed the data. JG, LG, RD-O, and RS interpreted the results. JG and RS drafted the manuscript and prepared the tables and figures. JG, LG, RD-O, ET, CC, AgC, FR, MS, AnC, and RS revised the paper and approved the final version of the manuscript.

# FUNDING

fphys-09-01247 September 4, 2018 Time: 9:42 # 6

This work was supported by the Northern Portugal Regional Operational Programme (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European Regional Development Fund (FEDER) (NORTE-01-0145-FEDER-000013) and the Fundação para a Ciência e Tecnologia (FCT) (contracts PD/BD/106053/2015 to JG via Inter-University Doctoral Programme in Ageing and Chronic Disease – PhDOC, IF/00021/2014 to RS, IF/01390/2014 to ET, IF/01147/2013 to RD-O, IF/00735/2014 to AgC, SFRH/BPD/96176/2013 to CC, and SFRH/BPD/111100/2015

# REFERENCES


to LG). MS is a FCT investigator. The NMR data was acquired at CERMAX (Centro de Ressonância Magnética António Xavier) which is a member of the National NMR network with the support of Project LISBOA-01-0145-FEDER-007660.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2018.01247/full#supplementary-material

intestinal epithelial HIF augments tissue barrier function. Cell Host Microbe 17, 662–671. doi: 10.1016/j.chom.2015.03.005


colitis, indicating that MUC2 is critical for colonic protection. Gastroenterology 131, 117–129. doi: 10.1053/j.gastro.2006.04.020

Viennois, E., Chen, F., Laroui, H., Baker, M. T., and Merlin, D. (2013). Dextran sodium sulfate inhibits the activities of both polymerase and reverse transcriptase: lithium chloride purification, a rapid and efficient technique to purify RNA. BMC Res. Notes 6:360. doi: 10.1186/1756-0500-6-360

**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 DS and handling Editor declared their shared affiliation at the time of the review.

Copyright © 2018 Gaifem, Gonçalves, Dinis-Oliveira, Cunha, Carvalho, Torrado, Rodrigues, Saraiva, Castro and Silvestre. 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.

# Could Sodium Chloride be an Environmental Trigger for Immune-Mediated Diseases? An Overview of the Experimental and Clinical Evidence

#### Eric Toussirot 1,2,3,4 \*, Matthieu Béreau<sup>5</sup> , Charline Vauchy 1,2 and Philippe Saas 1,2,6

1 Institut National de la Santé et de la Recherche Médicale CIC-1431, Centre d'Investigation Clinique Biothérapie, University Hospital Besançon, Besançon, France, <sup>2</sup> Fédération Hospitalo-Universitaire INCREASE, University Hospital Besançon, Besançon, France, <sup>3</sup> Rhumatologie, University Hospital Besançon, Besançon, France, <sup>4</sup> Département de Thérapeutique et EPILAB EA4266: "Epigénétique des Infections Virales et des Maladies Inflammatoires", Université Bourgogne Franche-Comté, Besançon, France, <sup>5</sup> Department of Neurology, University Hospital Besançon, Besançon, France, <sup>6</sup> Institut National de la Santé et de la Recherche Médicale U1098, Etablissement Français du Sang Bourgogne Franche-Comté, Université Bourgogne Franche-Comté, Interactions Hôte-Greffon-Tumeurs, LabEx LipSTIC, Besançon, France

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

David García-Bernal, Universidad de Murcia, Spain Mildred Audrey Pointer, North Carolina Central University, United States

> \*Correspondence: Eric Toussirot etoussirot@chu-besancon.fr

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 13 December 2017 Accepted: 06 April 2018 Published: 24 April 2018

#### Citation:

Toussirot E, Béreau M, Vauchy C and Saas P (2018) Could Sodium Chloride be an Environmental Trigger for Immune-Mediated Diseases? An Overview of the Experimental and Clinical Evidence. Front. Physiol. 9:440. doi: 10.3389/fphys.2018.00440 Immune mediated diseases (IMDs) are complex chronic inflammatory diseases involving genetic and environmental factors. Salt intake has been proposed as a diet factor that can influence the immune response. Indeed, experimental data report the influence of sodium chloride on the differentiation of naive CD4<sup>+</sup> T cells into IL-17 secreting T helper (Th) cells (Th17 cells), by a mechanism involving the serum glucocorticoid kinase-1 (SGK1) that promotes the expression of the IL-23 receptor (IL-23R). The IL-23/IL-23R is critical for pathogenic inflammatory Th17 cell differentiation. Experimental data in murine models of arthritis, colitis and encephalomyelitis corroborate these findings. This manuscript reviews the current knowledge on the effects of sodium chloride on innate and adaptive immunity. We also performed a systematic literature review for clinical studies examining the relationships between salt consumption and the development or the activity/severity of the most common IMDs mediated by the IL-23/Th17 pathway, i.e., rheumatoid arthritis (RA), multiple sclerosis (MS), and Crohn's disease (CD). Nine studies were found, 4 in RA, 4 in MS and 1 in CD. An association was found between developments of anti-citrullinated protein antibody (ACPA) positive RA in smokers and salt intake, but these results were not confirmed in another study. For MS, no association was observed in pediatric subjects while in adult patients, a link was found between salt intake and disease activity. However, this result was not confirmed in another study. These conflicting results highlight the fact that further evaluation in human IMDs is required. Moreover, physicians need to develop clinical trials with diet interventions to evaluate the impact of low salt intake on disease activity/severity of IMDs.

Keywords: autoimmune diseases, sodium chloride, IL-23, Th17, SGK1

# INTRODUCTION

Immune-mediated diseases (IMDs) are complex chronic inflammatory disorders involving the contribution of different predisposing factors. Indeed, both specific genetic backgrounds and environmental factors participate in their pathogenesis. Environmental factors have been implicated according to the prevalence of these diseases in certain countries or geographical areas, and/or through exposure of the population to local factors, but lifestyle habits as well as diet are also involved. The most common IMDs are type I diabetes, systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), psoriasis, multiple sclerosis (MS), and inflammatory bowel diseases, such as Crohn's disease (CD). Contributing environmental factors are specific for each IMD, and for instance, smoking is a well identified factor for the development of RA or CD, while ultraviolet exposure is involved in SLE. The influence of diet has long been suspected as a contributing pathogenic factor for IMD. For instance, a Mediterranean diet is thought to influence the occurrence of RA (Casas et al., 2017). Diet is also a modifiable environmental risk factor, and thus, the identification of a specific eating habit or food component in IMDs could help physicians in the prevention and treatment of these diseases. However, in general, the influence of diet on the pathogenesis of most IMDs is currently unknown.

It has recently been reported that salt in the diet could influence the development of certain IMDs, such as arthritis, CD, or MS (Croxford et al., 2013; van der Meer and Netea, 2013; Sigaux et al., 2017). Sodium is an extracellular cation that plays a fundamental role in the physiology of humans. Indeed, sodium is the most abundant cation in the organism, and is involved in osmotic balance, extracellular fluid volume as well as membrane potential of cells. Most dietary sodium is consumed as common salt i.e., sodium chloride. It is present in common food and is added when cooking or in food preparation. A high-salt diet has been implicated in certain pathological conditions, such as hypertension, cardiovascular and kidney diseases. It is well known that the mean daily salt intake is excessive in most western countries (between 7 and 18 g) and even higher in Asia, and in excess compared to the World Health Organization recommendations (<5 g daily; Brown et al., 2009). Recent data have highlighted the potential role of high-salt diet in the development of experimental IMD models, such as collagen-induced arthritis (CIA), experimental autoimmune encephalomyelitis (EAE), and experimental colitis. Indeed, mice receiving a high-salt diet have been shown to experience exacerbated colitis, CIA or EAE (**Table 1**). This has been associated with either an increase of inflammatory immune responses (e.g., M1 macrophages; Ip and Medzhitov, 2015; Hucke et al., 2016), pathogenic T cells (Kleinewietfeld et al., 2013; Wu et al., 2013, 2018; Wei et al., 2017; Aguiar et al., 2018), or autoantibodies (Sehnert et al., 2014) or a decrease of protective immune functions (e.g., regulatory CD4<sup>+</sup> T cell [Treg]; Hernandez et al., 2015; Wu et al., 2018), and IL-4/IL-13 activated M2 macrophages suppressive activity (Binger et al., 2015a). These findings suggest that high-salt diet may affect the functions of innate (i.e., macrophages; Binger et al., 2015a; Hucke et al., 2016) or innate lymphoid cells [ILC] ILC3 (Aguiar et al., 2018) and adaptative (i.e., CD4<sup>+</sup> T cells) immune cells. For instance, sodium chloride favored the polarization of pathogenic CD4<sup>+</sup> T cells toward a T helper 17 (Th17) phenotype (see below and **Table 1**). All these data are supported by studies performed in healthy volunteers showing that a controlled high-salt diet is correlated with an increase of circulating inflammatory monocytes counts, as well as pro-inflammatory cytokines (e.g., IL-6, IL-17, or IL-23) (Zhou et al., 2013; Yi et al., 2015). Recently, circulating Th17 cells were analyzed in an exploratory pilot cohort of 8 healthy male volunteers before and after a high-salt challenge (Yi et al., 2015). A significant increase of IL-17A<sup>+</sup> TNF<sup>+</sup> CD4<sup>+</sup> T cells was observed after this high-salt challenge. Altogether, this sheds new insights on the relationship between environmental factors, diet, and specific IMDs.

In this paper, we aim to provide an overview of the available experimental evidence as well as the clinical studies that were performed in this field and that found a link between salt intake and the development of IMDs.

# THE IL-23/TH17 PATHWAY, IL-17 AND TH17 CELLS: IMPLICATION IN IMDS

IMDs are complex chronic inflammatory diseases mediated by immune mechanisms involving innate immunity, as well as different T lymphocyte subsets. During the inflammatory process, naive CD4<sup>+</sup> T cells differentiate into various CD4<sup>+</sup> T cell subsets according to the environmental cytokine milieu. In the presence of IL-12, CD4<sup>+</sup> T cells differentiate into Th1 cells typically producing IFN-γ and TNF-α and expressing the T-bet transcription factor. This lymphocyte subset has been implicated in RA, CD, psoriasis, and also MS. In contrast, Th2 cells express GATA-3 transcription factor, produce IL-4 and IL-13 and play a role in allergic diseases, parasitic infections and to a certain extent in SLE. The so-called Th17 population is characterized by its production of IL-17, IL-21, IL-22, and IL-26. IL-17 corresponds to a family of cytokines which includes six members, IL-17A to IL-17F (Kikly et al., 2006), with Th17 cells producing IL-17A and IL-17F. Th17 cells have been reported to be implicated in the natural host defense by fighting against extracellular bacteria or fungi, in granulopoiesis (Gaffen et al., 2014), neutrophil migration and activation, but also in inflammatory processes and IMDs (Miossec et al., 2009). IL-17A and IL-17F are implicated in inflammatory responses associated with autoimmune conditions, but, in general, associated with other inflammatory cytokines produced by Th17 cells, such as IL-21 (Muranski and Restifo, 2013) and IL-22 (Bettelli et al., 2006). "Typical" or "conventional" Th17 cells have been described as a main—but not the only—cellular population responsible for IL-17 production. Indeed, a large variety of innate and adaptive immune cells are responsible for IL-17A and IL-17F production, including CD8<sup>+</sup> T cells, CD4−/CD8<sup>−</sup> α/β T lymphocytes, γ/δ T cells, NK cells, NKT cells, neutrophils, innate


ATIC, adoptive transfer-induced colitis consisting in infusion of naive CD4<sup>+</sup> T cells in Rag2-deficient mice; CIA, Collagen-induced arthritis; CNS, central nervous system; DSS, dextran sulfate sodium; EAE, experimental autoimmune encephalomyelitis; iNOS, inducible nitric oxide synthase; LPS, lipopolysaccharide; mLN, mesenteric lymph nodes; MOG, myelin oligodendrocyte glycoprotein; TNBS, trinitrobenzene-sulfonic acid.

lymphoid cells (ILC), and especially ILC3 cells, and mucosalassociated invariant T (MAIT) cells (Venken and Elewaut, 2015). The differentiation of naive CD4<sup>+</sup> T cells toward the Th17 subset involves the expression of the specific retinoic acid receptor- related orphan receptor γ-T (RORγt) transcription factor (Ivanov et al., 2006). The differentiation of Th17 is also dependent on the environmental cytokine milieu that includes IL-6, TGF-β, and IL-1β. IL-23, which comprises two chains (IL-23p19 and p40), is another master cytokine involved in naive CD4<sup>+</sup> T cell differentiation toward the Th17 subset (Korn et al., 2009; Gaffen et al., 2014; DuPage and Bluestone, 2016). IL-23 after interaction with its receptor sharing two chains, the IL-23R and IL-12Rβ1, favors Th17 stabilization (Langrish et al., 2005), but is also critical for maturation of pathogenic inflammatory Th17 cells by inhibiting IL-10 production (Gaffen et al., 2014). Indeed, Th17 cells represent heterogeneous Th cell subsets with pathogenic Th17 cells and regulatory anti-inflammatory Th17 cells (Gaffen et al., 2014). Moreover, Th17 cells are characterized by high plasticity and are related to both Th1 cells and Treg generated in the periphery (called peripheral Treg [pTreg]) (Korn et al., 2009). Pathogenic Th17 cells may acquire the Th1 T-bet transcription factor and the capacity to secrete IFN-γ to become highly inflammatory (Muranski and Restifo, 2013).

It is noteworthy that the IL-23/Th17 pathway is implicated in the pathophysiology of chronic IMDs in humans (Toussirot, 2012), as follows:



# RELATIONSHIPS BETWEEN SODIUM AND CELLS OF THE IMMUNE SYSTEM

Experimental data have reported in the nineties that mononuclear cells may release pro-inflammatory cytokines (e.g., IL-1β) when cultured in a hypertonic milieu (Shapiro and Dinarello, 1995). Later, it has been shown that sodium chloride favors pro-inflammatory M1 phenotype of macrophages (Hucke et al., 2016) and limits M2 macrophage activation (Binger et al., 2015a) (**Figure 1**). The pro-inflammatory effects of sodium chloride on M1 macrophages are observed in both mouse and human macrophages (Zhang et al., 2015). Several signaling pathways are triggered by elevated sodium chloride concentrations, namely: mitogen-activated protein kinases (MAPK), such as p38 kinase (Zhang et al., 2015; Hucke et al., 2016) and NF-κB (Hucke et al., 2016) (**Figure 1**). Macrophages may also activate the NLRP3 and NLRC4 inflammasomes in the setting of an osmotic milieu (generated for instance by high concentrations of sodium chloride) (Ip and Medzhitov, 2015). This leads to the release of IL-1β (Zhang et al., 2015). This cytokine IL-1β is critical for pathogenic human Th17 cell differentiation by inhibiting IL-10 and inducing IFN-γ (Sallusto, 2016). The direct implication of macrophages and inflammasome activation in high-salt diet-induced Th17 polarization has been demonstrated using caspase-1 deficient mice (Ip and Medzhitov, 2015). Thus, high-salt diet is responsible for innate immune cell activation that may in turn affect CD4<sup>+</sup> polarization. Besides this influence on innate immune cells, it has recently been demonstrated that salt can influence adaptive immunity by favoring the Th17 pathway. Indeed, two initial studies provide strong data demonstrating the direct influence of salt in the differentiation of naive CD4<sup>+</sup> T cells toward Th17 cells. Wu et al. performed a genome-wide mRNA analysis to identify the mechanisms that can explain the development of Th17 cells (Wu et al., 2013). Serum glucocorticoid kinase-1 (SGK1) was identified as a relevant candidate for IL-23R signaling and highly induced with Th17 differentiation. SGK1 expression is increased in specific conditions, including mineralocorticoid excess and hypertonicity, such as high salt exposure (Arora, 2013; Binger et al., 2015b). Indeed, SGK1 was expressed in Th17 cells and its expression strongly correlated with IL-23R signaling. This was shown by network analysis of the transcriptional changes in wild type animals and SGK1-deficient mice (Wu et al., 2013). Of interest, SGK1 was not equally expressed in other T cells subsets, especially Th1, Th2 and in vitro TGF-β-induced Treg (iTreg). IL-23 has the capacity to induce and maintain the expression of SGK1 in Th17 cells. Using protein-protein interaction analysis from a large database, FOXO1 was identified as a transcriptional factor that had a downstream role in the expression of SGK1. SGK1 promoted phosphorylation of FOXO1, which led to reduced FOXO1 activity and increased IL23r mRNA expression and induction of RORγt, a major regulator of Th17 differentiation (Wu et al., 2013). The second study by Kleinewietfeld et al. reported similar results: CD4<sup>+</sup> T cells cultured in a medium with increased salt concentration regulated SGK1 expression, and thus promoted Th17 differentiation (Kleinewietfeld et al., 2013). In addition,

it was shown that high salt condition activates the p38/MAPK pathway involving the nuclear factor of activated T cells, NFAT5, which in turn promotes activation of SGK1. Collectively, these data support a role for salt in SGK1 induction and the promotion of Th17 differentiation via signaling mechanisms involving FOXO1 and RORγt (**Figure 1**).

In parallel to these effects on the IL-23/Th17 pathway, salt exerts some influence on other T cell subsets, namely natural/thymic Treg (tTreg; i.e., generated in the thymus; Hernandez et al., 2015; Wu et al., 2018) and peripheral T reg (pTreg; i.e., differentiated in the periphery from naïve CD4<sup>+</sup> T cells; Wu et al., 2018). Hypertonic culture medium or a high-salt diet impairs Treg function (Hernandez et al., 2015). Sodium chloride increases the release of IFN-γ by Treg thus promoting their differentiation toward a Th1 phenotype. Conversely, reducing IFN-γ may restore the suppressive activity of Treg. These effects on Treg were mediated via SGK1 activity. A recent paper confirms this implication of SGK1 as a negative regulator of both tTreg and pTreg (Wu et al., 2018). In fact, SGK1 represses the expression of Foxp3- the master transcription factor regulating Treg differentiation and function—via the regulation of IL-23R and reduced FOXO1 nuclear exclusion (Wu et al., 2018). The salt-sensing transcription factor SGK1 plays thus a critical role in regulating the balance of pro-inflammatory Th17 and Treg (**Figure 1**). High-salt diet via SGK1 activation tips the balance in favor of pro-inflammatory Th17.

Efforts have been performed to link high salt diet with modification of T cell subsets in healthy volunteers. Indeed, data are also available in human healthy subjects. A high monocyte count associated with elevated levels of circulating proinflammatory cytokines (e.g., IL-6, IL-17A, or IL-23) has also been reported in normal subjects who have a high salt diet (Zhou et al., 2013; Yi et al., 2015). Conversely, a reduction in salt intake is associated with a decrease in inflammatory cytokine production, such as IL-6 and IL-23 (Yi et al., 2015). A recent well-controlled study performed in 8 healthy volunteers has demonstrated the increase of circulating Th17 cells after high-salt diet ingestion (Yi et al., 2015). This confirms experimental data and the induction of potential pathogenic humans Th17.

# INFLUENCE OF SALT DIET ON IMDS

The data reported above provide a strong rationale for examining the effects of sodium chloride on the development of IMDs, especially those with a Th17 contribution. We thus focussed our review on the diseases where IL-17A has proven to be a key player, i.e., RA, psoriasis, CD, and MS. We discuss here the available experimental data together with the clinical studies that were performed on this topic. For clinical studies, we reviewed the available literature in Medline via a Pubmed search for articles published between 2013 (date of the first experimental reports; Kleinewietfeld et al., 2013; Wu et al., 2013) and 2017, as well as abstracts from the American College of Rheumatology (ACR) meetings. The MeSH terms used were "salt" OR "sodium intake" OR "sodium chloride" OR "dietary sodium intake" OR "dietary sodium" AND "autoimmune diseases" OR "rheumatoid arthritis" OR "multiple sclerosis" OR "Crohn's disease" OR "psoriasis." To be selected for analysis, a study had to meet the following criteria: it had to be an epidemiological or pathophysiological study designed to examine the relationship between the development of RA, psoriasis, CD, or MS, and salt consumption, or to evaluate the relationship between salt diet and disease activity/severity. The Pubmed search retrieved 490 articles. After applying our selection criteria, 9 papers were selected (**Table 2**).

The results are discussed for each IMD, first experimental data followed by clinical studies that were found by our systematic literature review.

# Salt Diet in RA Experimental Data (Table 1)

Two studies were performed in murine arthritis and both were available as an abstract presented at the 2014 ACR meeting (Jung et al., 2014; Sehnert et al., 2014). These two studies were performed in the CIA model which shares with human RA several clinical, histopathological, and immunological features. These latter features consists in the breach of tolerance with the implication of pathogenic T cells, as well as the production of auto-antibodies against self-antigens, such as collagen. In the first study, a low salt diet was associated with decreased joint severity compared to a high salt diet. In addition, animals with a high salt diet had elevated levels of pathogenic antibodies (IgG2a) against the autoantigen, type II collagen (Sehnert et al., 2014). The second study confirms that mice fed with a high- salt diet had more severe joint disease. Splenocytes from these animals expressed a high level of RORγt and were likely to differentiate into Th17 cells (Jung et al., 2014).

## Clinical Studies

Four published studies were found (Salgado et al., 2015; Sundström et al., 2015; Jiang et al., 2016; Marouen et al., 2017). Three examined the relationship between the intake of sodium chloride and the development of RA, and one examined the relation between sale intake and disease severity. The first study used data from a large cohort of early RA from the Netherlands (EIRA study) and analyzed the impact of sodium consumption on the development of RA among smokers according to the level of sodium intake (Jiang et al., 2016). Sodium intake was evaluated by means of a self-report questionnaire. As expected, smoking status was found to be associated with anti-citrullinated protein antibody (ACPA) positive RA, but only in those with medium or high sodium intake [odds ratio (OR): 1.7 and 2.09 for ever-smokers and heavysmokers, respectively]. Conversely, medium or high sodium consumption was associated with an increased risk of ACPApositive RA among smokers only (OR 1.3 and 2.1 in eversmokers and heavy-smokers, respectively). A Swedish nested case -control study that included 386 subjects and 1,886 matched controls examined the interaction between dietary sodium, smoking and the risk of RA (Sundström et al., 2015). Sodium intake was quantified according to a food questionnaire. Dietary habits were collected before the onset of RA. The results

pro-inflammatory M1 macrophages and blunts the suppressive functions of IL-4 plus IL-13-activated M2 macrophages. A high salt intake/environment induces the activation of M0 macrophages via three different pathways: the p38 MAPK/cFos/Jun pathway, NF-κB activation which increases pro-inflammatory cytokine mRNAs including those coding pro-IL-1β, and finally, NLRP3/NLRC4 inflammasome which leads to caspase-1 activation responsible for pro-IL-1β cleavage into active IL-1β. IL-1β produced by macrophages exposed to a high salt environment may favor Th17 differentiation. In addition to macrophages, a high salt intake/environment directly affects CD4<sup>+</sup> T cells. The activation of p38 MAPK by sodium chloride promotes the activation of nuclear factor of activated cells 5 (NFAT5), and then serum glucocorticoid kinase-1 (SGK1). SGK1 induces the phosphorylation of the transcription factor forkhead box protein O1 (FOXO1), a nuclear factor that represses the expression of the Il23r gene. Conversely, ROR-γt is induced leading to the transcription of the Il23r gene, resulting in the expression of IL23R at the membrane surface favoring the differentiation of naïve CD4<sup>+</sup> T cells toward a Th17 phenotype, and thus production of IL17A and IL-17F. This favors IMD, such as experimental autoimmune encephalomyelitis (EAE), a murine model of multiple sclerosis (MS), colitis, or collagen-induced arthritis (CIA), an experimental model of rheumatoid arthritis (RA). Note that other cytokines can be released by Th17 cells, such as IL-21, IL-22, IL-1β, or IL-6. Furthermore, a high salt intake/environment inhibits the suppressive functions and the expression of FoxP3 via the SGK1/FOXO1 pathway.

showed no significant association between sodium intake and the development of RA when all individuals were examined. However, there was an increased risk of RA among smokers according to the tertile of sodium consumption, with those in the highest tertile having more than a 2-fold increase in risk compared to the subjects in the lowest tertile (OR 2.26; 95%CI 1.06–4.81). Thus, these 2 studies suggest an interaction between smoking and high sodium consumption for the development of RA. A third study performed in Spain gave contradictory results (Salgado et al., 2015). In that study, daily sodium intake was estimated from a validated questionnaire. This study included 18,555 subjects and among them, 392 developed RA. The results showed that total sodium intake was significantly associated with RA (OR 1.5; 95%CI: 1.1–2.1) but never smokers with high sodium intake had a higher risk than ever smokers (p = 0.007). Finally in a case-control study, sodium intake as evaluated by 24 h urinary sodium excretion was found to be increased in a small cohort of patients with early RA (Marouen et al., 2017). Sodium excretion was also greater in patients with radiographic erosions compared to those without.

# Salt Diet in Crohn's Disease Experimental Data (Table 1)

Normal intestinal lamina propria mononuclear cells—extracted from macroscopically and microscopically unaffected colonic samples of 9 patients undergoing resection for cancer colonexpress high quantities of IL-17A, IL-23R, TNF-α, and RORγt when exposed to increasing concentrations of sodium chloride (Monteleone et al., 2017). In a mouse model of colitis, mice receiving a high salt diet developed more severe colitis that was abrogated by a pharmacologic agent that controlled p38/MAPK, and thus SGK1 (Monteleone et al., 2017). Mice that were exposed to a high salt diet had an increased frequency of IL-17A producing cells in the intestinal lamina propria as compared to mice fed with a normal diet (Wei et al., 2017). As observed in the previous study, 2,4,6-Trinitrobenzenesulfonic acid (TNBS) induced colitis was exacerbated under a high salt diet, with a significant increase in Th17 responses in the colonic lamina propria. These results have been recently confirmed in another study by a different research group with a rapid induction of Sgk1 mRNA (90 min after high-salt diet consumption) in the colon (Aguiar et al., 2018). Furtehrmore, IL-10 deficient mice that

TABLE 2 | Clinical studies examining the relationships between salt intake and development or activity/severity of the immune-mediated diseases.


RA, rheumatoid arthritis; MS, multiple sclerosis; CD, Crohn's disease; CIS, clinically isolated syndrome; UC, ulcerative colitis; OR, Odds ratio [95% confidence interval]; ACPA, anti-citrullinated protein antibodies.

were exposed to a high-salt diet developed also a more severe spontaneous colitis (Tubbs et al., 2017).

## Clinical Studies

One study examined the relationship between salt consumption and CD (Khalili et al., 2016). The subjects evaluated were women from the Nurse Health Studies (NHS and NHSII studies), which included detailed information on lifestyle and diet. The effect of salt and potassium intake was evaluated in this prospective study and incident cases of CD were recorded. Assessment of diet was performed using a 161 item semi-quantitative food frequency questionnaire. While dietary intake of potassium was found to be a protective factor for the development of CD, dietary salt intake had no influence.

# Salt Diet in Multiple Sclerosis Experimental Data (Table 1)

In EAE, mice that were fed with a high salt diet developed more severe disease. High salt consumption also accelerated the onset of disease. In this setting, IL-17A expressing CD4<sup>+</sup> T cells markedly infiltrated the central nervous system (Kleinewietfeld et al., 2013). Moreover, in SGK1-deficient mice, there was reduced frequency and severity of EAE (Wu et al., 2013).

### Clinical Studies

Our literature search found 4 papers analysing the effect of dietary salt in MS, namely 2 in adult patients and 2 others in pediatric subjects (Farez et al., 2015; McDonald et al., 2016; Nourbakhsh et al., 2016; Fitzgerald et al., 2017). In a multicentre case-control study performed in pediatric MS, sodium intake was evaluated using a specific pediatric food questionnaire (McDonald et al., 2016). In this study, salt intake was not related to MS: there was no difference in mean sodium intake between cases and controls. These results were confirmed in another study that evaluated dietary sodium intake using the same questionnaire (Nourbakhsh et al., 2016). Dietary sodium intake was not found to be associated with time to relapse of neurological disease. In adult patients, one study reported a significant association between sodium intake and disease activity of MS as evaluated clinically (by the expanded disability status scale-EDSS) or by central nervous system MRI (Farez et al., 2015). Sodium intake was evaluated in urine samples. There was a positive correlation between exacerbation rates and sodium excretion. Moreover, neurological exacerbation rate was 2.75 to 3.95-fold higher in patients with medium or high sodium intake compared to patients in the low intake group. These results were not confirmed by the BENEFIT study (Fitzgerald et al., 2017). This study examined the relationship between conversion from an early form of MS (i.e., clinically isolated syndrome) to MS, and urinary sodium concentration. The results showed no association between sodium intake and the conversion to MS over a 5 year follow-up period (Hazard ratio 0.91; 95%CI: 0.67–1.24). In the same way, salt intake did not correlate with disease outcome such as clinical or MRI measurements.

# Salt Diet and Psoriasis

We found neither experimental nor clinical data examining the relationship between salt consumption and psoriasis.

# DISCUSSION

Sodium is an essential nutrient for the organism and is a major physiologic player. The description of its role as an environmental factor in inflammation, and especially in IMDs comes from several reports (Arora, 2013; Croxford et al., 2013; van der Meer and Netea, 2013). Indeed, strong experimental data support its implication as a driver of IL-17A production via IL-23R promotion. Indeed, the studies by Wu et al. and Kleinewietfeld et al. both reported that an increased salt concentration promotes Th17 differentiation by SGK1 involvement (Capon et al., 2007; Kleinewietfeld et al., 2013; Wu et al., 2013). A specific mechanism involving SGK1, FOXO1 and p38/MAPK/NAFT5 and IL-23R is thus described, providing an elegant demonstration of the relationship between salt and IL-17A production (Binger et al., 2015b). However, these results raise some questions, especially concerning the specific implication in human IL-23/Th17-mediated diseases.

One question is the consequence of a high salt diet on the blood or lymph node compartments. Indeed, besides the hypertonic and vasopressive effects on blood pressure, a high salt diet does not induce a high concentration in the blood or lymph nodes (van der Meer and Netea, 2013). Th17 cell differentiation is mainly performed in the secondary lymphoid organs, such as lymph nodes (Gaffen et al., 2014). Conversely, it may produce changes in the gut environment, and this may produce potential effects on local gut immune responses. In turn, this may represent a link between a diet factor, gut environment and flora (i.e., microbiota) and the immune system (Croxford et al., 2013). Modification of the microbiota is a well-described mechanism that has been implicated in the pathogenesis of diverse IMDs, especially CD. Of interest, a recent study reported the influence of a high salt diet on gut microbiota in mice, resulting in a depletion of Lactobacillus murinus. Treatment of mice with L. murinus prevented salt-induced aggravation of actively-induced EAE by modulating Th17 cells. A moderate high-salt intervention in a pilot study in humans reduced intestinal survival of Lactobacillus spp. and increased Th17 cells. These results indicate that high salt intake is connected to the gut-immune axis (Wilck et al., 2017).

Besides sodium chloride, are other salt components involved in the induction of the Th17 lymphocyte subset? For instance, potassium has been reported to be a protective cation for CD (Khalili et al., 2016). Since salt drives IL-23R expression, another question is the specific cells that could be sensitive to this environmental factor. Most experimental studies focussed the analysis on T lymphocytes, but IL-17A can be release by a wide range of other cells than Th17, including ILC3 and MAIT cells (Venken and Elewaut, 2015). These 2 cellular subsets are present in the gut, and thus, may be responsive to a salt diet. A recent study reports that ILC3 frequency is increased 2 weeks after high salt diet consumption in the mouse colon (Aguiar et al., 2018). This is an interesting track to be further explored. The experimental data reported an influence of salt in the initiating step of autoimmune diseases such as EAE or CIA, but what are the effects of salt/sodium chloride when the disease is at an established or chronic phase? Indeed, salt has a greater effect on initiation than on the maintenance of Th17 response (van der Meer and Netea, 2013).

Despite strong experimental reports, clinical studies did not report consensual results. Indeed, clinical evidence for the relationships between sodium consumption and development and/or activity/severity of IMDs is limited. Three studies reported such a link in the setting of RA (Sundström et al., 2015; Jiang et al., 2016; Sallusto, 2016; Marouen et al., 2017), and only in one study that was performed in adult patients with MS (Farez et al., 2015). One major limitation of these studies is the tool used for evaluating salt consumption, mainly a self-report questionnaire. A more accurate method involving quantitative measurements of urinary excretion was used in a limited number of studies (Farez et al., 2015; Fitzgerald et al., 2017; Marouen et al., 2017). In addition, the interaction between salt and other environmental factors (such as tobacco) was only examined in RA, giving contradictory results (Salgado et al., 2015; Sundström et al., 2015). This highlights the fact that there are indisputably multiple contributors to the development of IMDs, and they can interact with each other, as with tobacco and salt in RA. The interaction between salt and the genetic background of IMDs was not examined and it is certainly another potential avenue that warrants exploration, as it has previously been reported that there is a link between tobacco and the expression of the shared epitope for the risk of ACPA-positive RA (Klareskog et al., 2006). SGK1 gene polymorphism is another genetic background that merits investigation, despite negative findings in RA (Jiang et al., 2016). For MS, the study by Farez et al. provided evidence that salt may worsen disease activity (Farez et al., 2015), but these results were not confirmed in the BENEFIT study (Fitzgerald et al., 2017). Moreover, epidemiological data do not support a strong influence of salt in MS occurrence. Indeed, while people from Asian countries are exposed to a high salt diet, the incidence of MS in Asian countries is among the lowest worldwide (Bach, 2018). In addition, studies examining the effect of salt in psoriasis (but also in spondyloarthritis including psoriatic arthritis), an IMD with a strong Th17/IL-17A involvement, are currently lacking. Finally, a further avenue to explore is a diet-specific intervention in these different IMDs during clinical trials. Randomized trials with a low salt diet vs. normal diet in different populations worldwide would certainly provide relevant information on the specific influence of salt in IMDs, at least in terms of disease activity. In this regard, a low salt diet intervention was performed in patients with RA or SLE, and controlled by urinary excretion measurements (Scrivo et al., 2017). In both diseases, Th17 cells decreased while there was an increase in Treg cells. These results suggest that restricted sodium dietary intake may dampen the inflammatory response in RA and SLE patients (Scrivo et al., 2017). The influence of salt diet on the response to IL-23p40 or IL-17A blocking agents (ustekinumab and secukinumab/ixekizumab, respectively) in patients with psoriasis, CD or spondyloarthritis is another relevant question to examine.

# CONCLUSION

The influence of the environment on the development and clinical expression of IMDs is an exciting and challenging issue. The role of sodium chloride in IMDs such as RA, CD, psoriasis and MS is a relevant question for the pathogenesis of these diseases but also for the management of patients. Current experimental data strongly support a link between salt intake and Th17 differentiation. However, the specific influence of dietary salt intake on the expression of IMDs in humans is still not demonstrated, and requires further studies, especially clinical trials with nutritional interventions aimed at comparing low vs. high salt diets. The interaction between salt and other environmental factors, as well as the genetic background, is a promising avenue for further research, in order to improve our understanding of the

# REFERENCES


specific role of sodium chloride in the pathogenesis of IMDs.

# AUTHOR CONTRIBUTIONS

ET, MB, CV, and PS analyzed and discussed the literature and conceived the outline of the manuscript. ET and PS wrote the manuscript. All authors reviewed the manuscript and provided critical discussion and input.

# ACKNOWLEDGMENTS

This work is supported by the Agence Nationale de la Recherche (ANR) under the program Investissements d'Avenir with reference ANR-11-LABX-0021-LipSTIC and by the Region Bourgogne Franche-Comté (support to LipSTIC LabEX 2017).


**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 Toussirot, Béreau, Vauchy and Saas. 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.

# Role of the Inflammation-Autophagy-Senescence Integrative Network in Osteoarthritis

Claire Vinatier 1,2, Eduardo Domínguez <sup>3</sup> , Jerome Guicheux 1,2,4 and Beatriz Caramés <sup>5</sup> \*

1 INSERM, UMR 1229, Regenerative Medicine and Skeleton, University of Nantes, ONIRIS, Nantes, France, <sup>2</sup> University of Nantes, UFR Odontologie, Nantes, France, <sup>3</sup> Biofarma Research Group, Center for Research in Molecular Medicine and Chronic Diseases, University of Santiago de Compostela, Santiago de Compostela, Spain, <sup>4</sup> CHU Nantes, PHU4 OTONN, Nantes, France, <sup>5</sup> Grupo de Biología del Cartílago, Servicio de Reumatología. Instituto de Investigación Biomédica de A Coruña, Complexo Hospitalario Universitario de A Coruña, Sergas, A Coruña, Spain

Osteoarthritis is the most common musculoskeletal disease causing chronic disability in adults. Studying cartilage aging, chondrocyte senescence, inflammation, and autophagy mechanisms have identified promising targets and pathways with clinical translatability potential. In this review, we highlight the most recent mechanistic and therapeutic preclinical models of aging with particular relevance in the context of articular cartilage and OA. Evidence supporting the role of metabolism, nuclear receptors and transcription factors, cell senescence, and circadian rhythms in the development of musculoskeletal system degeneration assure further translational efforts. This information might be useful not only to propose hypothesis and advanced models to study the molecular mechanisms underlying joint degeneration, but also to translate our knowledge into novel disease-modifying therapies for OA.

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Phillip Trevor Newton, Karolinska Institutet (KI), Sweden Gautham Yepuri, University of Fribourg, Switzerland

#### \*Correspondence:

Beatriz Caramés beatriz.carames.perez@sergas.es

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 17 February 2018 Accepted: 22 May 2018 Published: 25 June 2018

#### Citation:

Vinatier C, Domínguez E, Guicheux J and Caramés B (2018) Role of the Inflammation-Autophagy-Senescence Integrative Network in Osteoarthritis. Front. Physiol. 9:706. doi: 10.3389/fphys.2018.00706

Keywords: inflammation, aging, senescence, autophagy, chondrocytes, cartilage, OA, therapeutics

# INTRODUCTION

Osteoarthritis (OA) is one of the most prevalent joint diseases and a leading cause of disability worldwide (GBD 2016 DALYs HALE Collaborators, 2017). Our understanding of the molecular mechanisms underlying OA, which is now considered an inflammation-associated multifactorial disorder, progressed substantially. However, no effective disease-modifying therapy is available yet and current pharmacological interventions only address pain and inflammation (Martel-Pelletier et al., 2016). Although the relationship between aging and the development of OA is not completely understood, it is becoming apparent that aging-related changes in the musculoskeletal system, in conjunction with mechanical injury and genetic factors contribute to OA (Loeser, 2009; Taniguchi et al., 2009). This novel vision of OA pathogenesis has led the OARSI consortium to re-evaluate the clinical definition to more accurately reflect the contribution of the underlying molecular mechanisms (Kraus et al., 2015).

The process of cellular senescence contributes to age-related dysfunction and chronic inflammation. Senescence essentially refers to the irreversible growth arrest that occurs when cells experience stress insults. Articular cartilage undergoes aging-associated changes in structure and biochemical composition which compromise biomechanical function and contribute to the formation of structural defects (Rahmati et al., 2017). Changes in cartilage aging include a reduction in chondrocyte density, proliferation and abnormal biosynthetic function (Loeser, 2009) as a response to extracellular stimuli in human (McAlinden et al., 2001) and mouse (Sandy and Plaas, 1986) articular chondrocytes. Senescent cells secrete proinflammatory cytokines, chemokines, and proteases, termed the senescence-associated secretory phenotype (SASP) (Kuilman et al., 2010; Soto-Gamez and Demaria, 2017). Through the inflammatory, growth-promoting, and remodeling factors that are secreted, this mechanism can explain how senescent cells alter tissue microenvironments, attract immune cells, and induce malignant phenotypes in neighboring cells (Acosta et al., 2013). Proteins that are associated with the SASP, such as IL-1β, TNF-α, IL-6, or matrix metalloproteinases (MMPs) increase with aging (Freund et al., 2010), and participate in the inflammation process. SASP is thus seen as a driver of age-related inflammation, at least under certain conditions (Rodier and Campisi, 2011). In line with these, the selective elimination of senescent cells or their deleterious effects is proposed as an intervention to reduce age-related chronic inflammation (Baar et al., 2017), enhance health span (Baker et al., 2011), and disrupt the link between aging and chronic musculoskeletal disease (Garcia-Prat et al., 2016).

Aging, which is a complex physiological process consisting of progressive decline of certain functions, coordination, loss of homeostasis, and physiological integrity (Lopez-Otin et al., 2013), is one of the major risk factors for OA. Like aging, OA is commonly described as the result of disruption of cartilaginous tissue homeostasis, where catabolic products such as damage-related molecular models (DAMP) accumulate in the joint and cause oxidative stress and inflammation (Haseeb and Haqqi, 2013). Articular chondrocytes rely on autophagy as the primary mechanism for maintaining normal function and survival (Terman et al., 2010). However, during aging, autophagy gradually decreases in chondrocytes thus inducing senescence, which ultimately results in increased OA severity (Carames et al., 2010).

Considering the growing body of evidence related to the role of inflammation-autophagy network in OA-associated chondrocyte senescence, this review describes the molecular pathways underlying age-associated changes in chondrocyte senescence that could lead to novel transformative therapies.

# SENESCENCE IN OA

# General Context

Cellular senescence also called replicative senescence is a stress response characterized by a persistent cell-cycle arrest. Two different observations have led to the discovery of senescence in the 1960s: Normal non-transformed cells have a limited number of division in culture and this maximal number of cell division reached by normal human cell in culture decrease with the age of the donor (Hayflick, 1965).

Cellular senescence is concomitantly considered as beneficial or detrimental for cells (Campisi and d'Adda di Fagagna, 2007). On the one hand, cellular senescence is considered beneficial notably through its property to suppress the development of cancer, its role as a pivotal mechanism during embryonic development (Munoz-Espin et al., 2013; Storer et al., 2013) or its role in wound healing at site of tissue injury (Demaria et al., 2014). On the other hand, senescence exerts detrimental actions through its promoting effect in aging and age-related diseases as evidenced by the accumulation of senescent cells in degenerated tissues (Baker et al., 2008; Campisi, 2014) and the exponential increase in the expression of some senescent markers thereby suggesting that senescent cells contribute to aging and shorten health span (Baker et al., 2016). Even if the underlying mechanisms of cellular senescence are not fully identified, they have been associated with telomere erosion, DNA damage, oxidative stress, and inflammation and led to the description of different types of senescence. The first one is telomere-initiated cellular senescence also called DNA Damage Response (DDR). The signaling pathway activated in DDR involves the stabilization of p53 after its phosphorylation by the ataxia telangiectasia mutated kinase (ATM) (d'Adda di Fagagna et al., 2003; Deng et al., 2008). The second type of senescence is stressinduced senescence also called premature senescence, which is considered a DDR-independent pathway. This premature senescence is consecutive to a variety of stress that remains poorly understood. The third one is oncogene-induced senescence which is due to the overexpression of several oncogenes such as RAS type family (Di Mitri and Alimonti, 2016).

## Senescence Features

Senescent cells in culture undergo morphological alterations, such as increased cell size, flattening, vacuolization, and accumulation of stress granules (Kuilman et al., 2010; Rodier and Campisi, 2011). These senescent cells also exhibit specific hallmarks including (i) growth arrest, (ii) cell death resistance, and (iii) an altered gene expression leading to the production of bioactive SASP.

**(i) The growth arrest**, which is a replication failure due to the expression of cell cycle inhibitors, occurs in senescent cells following various stimuli. These stimuli can consist in DNA damage, telomere shortening, oncogene activation for the most widely described, but other stimuli such as inflammation, reactive oxygen species (ROS), mitotic stress or unresolved unfolded protein response can also drive cell growth arrest. Following to these stimuli, two main signaling pathways are often involved, the DDR-dependent pathway P53-P21CIP1-retinoblastoma protein (pRB) and the DDR-independent pathway P16INK4A-pRB. These pathways include cyclin-dependent kinase inhibitors (CDKIs) such as P21CIP1 (CDKN1A) or P16INK4A (CDKN2A), which respectively inhibited cyclin-dependent kinase-2 (CDK2), CDK4, and CDK6. Through CDK4 and CDK6 inhibition, the p16INK4A conserves pRB in their repressive forms, thus preventing G1/S phase cell cycle progression (Gil and Peters, 2006). The p53-p21 pathway inhibits CDK2 that also maintains the repressive forms of pRB, but induced G1 or G2 phase cell cycle halt (Gil and Peters, 2006). While both DDR-dependent and DDR-independent pathways are intimately linked with various common upstream regulators and downstream effectors, p16INK4A/pRB pathway

**(ii) Cell death resistance** is also a common feature of senescent cells. Indeed, senescent human fibroblasts resist apoptosis induced by growth factor deprivation or oxidative stress (Hampel et al., 2005). Senescent cells also exhibit a noticeable resistance to mitochondria-mediated apoptosis in part by the maintenance or up-regulation of the anti-apoptotic protein B cell lymphoma 2 (BCL-2) family members (Ryu et al., 2007), known to inhibit the mitochondrial outer membrane permeabilization (MOMP) and the subsequent release of proapoptotic cytochrome C. This BCL-2-mediated resistance to apoptosis of senescent cells has been exploited by developing senolytic compounds such as Navitoclax or ABT-737 that target BCL-2 family members (Chang et al., 2016; Yosef et al., 2016; Zhu et al., 2016). In addition, senescent cells also resist to apoptosis through the interception of Fas ligand (FasL) by decoy receptor 2 (DCR2), which is overexpressed in senescent cells. Apoptosis of senescent cells may also be stopped through the P53-mediated up-regulation of p21 which directly inhibits caspase 3 activity (Tang et al., 2006).

**(iii) The Senescence-Associated Secretory Phenotype (SASP)** is a bioactive secretome produced by senescent cells which remain metabolically active despite they are not proliferative (Coppe et al., 2010a). SASP is associated with the production of proinflammatory cytokines (IL-6, IL-1α/β), chemokines [Monocyte chemotactic protein (MCP-1/CCL2)], IL-8/CXCL8, growth regulating oncogene (GROα), growth factors [TGF-β, VEGF, insulin-like growth factor (IGF)-binding protein 7 (IGFBP7) and amphiregulin] and proteases [Matrix Metalloproteases (MMP-3, TIMP-1)] that may have potent beneficial or deleterious effects on neighboring cells and the surrounding tissues (Soto-Gamez and Demaria, 2017). SASP, through the involvement of paracrine mediators, indeed propagates senescence from cell to cell thereby exacerbating the pro-aging effects of senescent cells. Among the SASP components, IGFBP-7, plasmin activator inhibitor (PAI-1), IL-6, CXCR2 ligands (IL-8 and GROα) are known to strengthen senescence either by autocrine (IL-6) (Kuilman et al., 2008) or paracrine (IGFBP7), effects (Wajapeyee et al., 2008; Acosta et al., 2013). The transient exposure to SASP is also conversely known to promote tissue regeneration or repair by the release of growth factors or proteases (Krizhanovsky et al., 2008; Demaria et al., 2014). Whereas persistent SASP signaling can result in deleterious effects such as chronic inflammation, SASP is also involved in the immune surveillance process and elimination of senescent cells by the immune system (van Deursen, 2014).

Whatever the senescence inducers (stress-induced, oncogeneinduced. . . ) and the associated signaling pathways, all SASP regulators converge to the nuclear factor-κB (NF-κB) signaling (Chien et al., 2011). This NF-κB signaling pathway is regulated by mechanistic target of rapamycin (mTOR)-dependent protein translation (Herranz et al., 2015; Laberge et al., 2015), and the cell surface-bound IL-1α (Orjalo et al., 2009). Others signaling pathways that converge to NF-κB activation are p38 mitogenactivated protein kinase (MAPK) (Freund et al., 2011), GATA4 which is stabilized by a defective p62-mediated autophagy (Kang C. et al., 2015), the phosphoinositide 3 kinase (PI3K) (Kortlever et al., 2006) pathways and the RIG-1/IRF pathways (Liu et al., 2011). NF-κb activation and the subsequent SASP production might also be initiated by oxidative stress and increased ROS (Chandrasekaran et al., 2017).

Moreover, advanced glycation end product (AGEs) as well as HMGB1 through their interaction with the Receptor for advanced glycation end product (RAGE) (Orlova et al., 2007) also activate NF-κβ pathway. Once translocated in the nucleus, NFκB binds to the promoter region of RAGE and enhances RAGE mRNA translation (Li and Schmidt, 1997).

# Identification of Senescent Cells

Owing to the lack of a specific "magic" marker to characterize or evaluate senescence in vivo, senescent cells have to be identified using an association of multiple markers including SAβ-gal or lipofuscin staining (Evangelou et al., 2017), the upregulated expression of p16INK4A, p21CIP1 or p19ARF and SASP (Coppe et al., 2008, 2010b). In addition to the presence of these partly selective markers, the decrease or loss of some other markers may also be instrumental to identify senescent cells. Among these markers, Nuclear High mobility group box 1 (HMGB1), which is localized in the nucleus in non-senescent cells, is secreted by senescent cells, making the loss of nuclear staining for HMGB1 a sign of senescence (Davalos et al., 2013). Similarly, the decreased expression of lamin B1 is also often used to address the presence of senescent cells (Freund et al., 2012). Below are the major markers classically used to identify senescent cells.

### **Senescence associated** β**-galactosidase**

Senescent cells exhibit increased levels of the lysosome enzyme acidic β-galactosidase called senescent-associated-βgalactosidase (SAβ-gal). Whereas, the normal β-galactosidase activity is detectable at pH 4.5, the SAβ-gal is only detectable at pH = 6 on fresh sample with the X-gal substrate (Itahana et al., 2007; Debacq-Chainiaux et al., 2009). Due to its easy detection, SAβ-gal is the most commonly used senescence marker. However, SAβ-gal has been also detected in immortalized cells, tumor cell lines, and even in normal cells when cultured in conditions known to improve lysosome activity (Young et al., 2009). The expression of SAβ-gal is now considered not specific to senescent cells (Itahana et al., 2007; Muller, 2009) and must be combined to other markers (see below) to properly identify such cells.

# **Lipofuscin**

Recently Lipofuscin, a non-degradable aggregate of oxidized lipids, proteins, oligosaccharides and metals which accumulates in the lysosomes, has been recently shown to amass with age in various post-mitotic tissues (Terman and Brunk, 2004b). It has been demonstrated that lipofuscin accumulation in senescent cells is a consequence of the senescent process (Jung et al., 2007). Georgakopoulou et al. have thus shown, through Sudan Black B (SSB) staining, that lipofuscin accumulates and co-localizes with SAβ-gal in senescent cells (Georgakopoulou et al., 2013). Interestingly, SBB staining of lipofuscin can also be achieved in paraffin-embedded materials allowing retrospective senescence analysis in archival samples.

### **Telomeres length**

Telomeres are tandem TTAGGG repeated sequences located at the ends of chromosome and associated with a protein complex named "shelterin" which cape the telomere end (de Lange, 2005). The shelterin complex is composed of six-subunit comprising TRF1, TRF2, POT1, TPP1, TIN2, and Rap. This protein complex is specifically associated with mammalian telomeres and allows cells to distinguish the natural ends of chromosomes from DNA damage sites. During aging or replication theses telomeres progressively shortens in somatic cells lacking the enzyme telomerase. With this shortening, the shelterin complex is displaced and expose the telomere end that become recognized by the DNA repair machinery as a double-strand DNA break and triggers a DNA Damage Response leading to senescence (Sfeir and de Lange, 2012). Moreover, cells lacking shelterin components, such as POT1 or TRF2, exhibit a DNA damage response and premature induction of senescence (Denchi and de Lange, 2007). Telomere shortening is also increased by oxidative stress (Ludlow et al., 2014) leading to senescence and accelerating aging (Aubert and Lansdorp, 2008). Inflammation through activity of NF-κB and the pro-inflammatory cytokines contribute to telomere attrition and the onset of senescence (Zhang et al., 2016).

## **Sirtuin (SIRT)**

Sirtuins are members of a family of evolutionarily conserved enzymes with NAD+-dependent deacetylase/deacylase activity. Sirtuins are involved in various biological role including cellular metabolism, genome stability and lifespan regulation (for review see Houtkooper et al., 2012). Among the seven sirtuin proteins (SIRT1–7) identified in mammals, SIRT1 is the most extensively studied. SIRT1 activation is known to prevent senescence by promoting P53 degradation (Solomon et al., 2006), whereas its inactivation increases the transcriptional activity of p53. These findings indicate that the SIRT1-p53 pathway is critical for regulating cellular senescence (Yu et al., 2017). SIRT 1 level has been shown to protect from stress induce premature senescence as well as replicative senescence (Ghosh and Zhou, 2015). SIRT 6 expression ameliorates lifespan in mice whereas its knock down induces premature death within a month (Mostoslavsky et al., 2006). SIRT6 has been shown to promote DNA damage repair process (Mao et al., 2011) and protect cells from telomere shortening (Michishita et al., 2008). SIRT6 deacetylates H3K18 to prevent mitotic errors and suppress senescence (Tasselli et al., 2017). SIRT6 also attenuates NF-κβ signaling by deacetylating histone H3 at K9 on the promoters of NF-κB target genes (Kawahara et al., 2009) suggesting that SIRT6 could regulate the SASP. As a whole, sirtuins seems to have a potential key role in protecting cells from premature senescence and accelerated aging.

# **High mobility group box-1 (HMGB1)**

HMGB-1 is a non-histone nuclear protein which sustains chromosome structure and stability. Under stress conditions, HMGB1 translocates from the nucleus to the cytosol and is then released into the extracellular space to function as an alarmin, a signal of tissue and cell damage (Ugrinova and Pasheva, 2017). In senescent cells, following activation of p53 HMGB-1 is secreted in the extracellular space (Davalos et al., 2013). Released HMGB-1 then stimulates NF-κβ transcriptional activity and stimulates SASP through its fixation on various receptors such as the receptor for advanced glycation end product or toll-like receptor 4. On the contrary nuclear HMGB-1 inhibits NF-κβ activity and the subsequent SASP (Davalos et al., 2013). Therefore, the loss of HMGB1 nuclear staining is an investigated marker in senescence characterization.

## **Caveolin-1 (CAV1)**

CAV1 is a major component of the caveolae structure and a transmembrane protein well-known to regulate numerous cellular processes such as cell growth and differentiation or endocytosis but also the stress-induced senescence (Zou et al., 2011; Nguyen and Cho, 2017). CAV1 is notably upregulated during replicative and premature senescence. CAV1 mediated signaling seems to be implicated in the promoting effect of oxidative stress on cell senescence, and overexpression of CAV1 induces senescence in age-related diseases (Galbiati et al., 2001; Bartholomew et al., 2009; Volonte et al., 2015). Paradoxically, CAV1 deficient mice exhibit aging related phenotype in various organs, such as neurodegeneration, fat atrophy, and premature aging (Head et al., 2010; Briand et al., 2011). Furthermore, CAV1 deficiency has recently been shown to also induce premature senescence via mitochondrial dysfunction and SIRT-1 inactivation (Yu et al., 2017).

# **p16**INK4<sup>A</sup>

The selective inhibitor of cyclin-dependent kinases CDK4 and CDK6, p16INK4A is the second most widely used senescent marker after SAβ-gal. The expression of p16INK4A is increased in senescent cells and participates in the cell cycle arrest (Alcorta et al., 1996). Aging organisms accumulate senescent cells expressing p16INK4A in various tissues (Krishnamurthy et al., 2004). This accumulation contributes to the onset of age-related disorders by compromising tissue function through the loss of tissue structural integrity and depletion of tissue-specific stem cell pools (Sharpless, 2004; Martinez-Zamudio et al., 2017a). Despite its well-defined role in senescence, the use of p16INK4A as a marker in mice has been hampered by the lack of specific and robust antibodies available for immunohistochemistry. The involvement of p16INK4A in aging and age-related diseases was however evidenced through the use of specific transgenic mice such as INK-ATTAC (Baker et al., 2011) and p16-3MR mice (Demaria et al., 2014). The INK-ATTAC mice exhibit a FK506 binding-protein–caspase 8-enhanced GFP (FKBP–Casp8-eGFP) fusion protein under the control of a p16INK4A promotor. These mice allow the detection and sorting of p16INK4A positive cells as well as their elimination by activation of caspase-8 after addition of AP20187 (Baker et al., 2011). The P16-3MR mice exhibit a trimodality reporter fusion protein, which contains functional domains of a synthetic Renilla luciferase (LUC), monomeric red fluorescent protein (mRFP), and truncated herpes simplex

virus 1 (HSV-1) thymidine kinase under the control of p16INK4A promoter allowing to concomitantly trace and kill p16INK4A expressing cells through administration of ganciclovir (Demaria et al., 2014). Of particular interest, these two models have greatly contributed to address the protective role of depleting p16INK4A expressing-senescent cells in aging and age-related diseases (Baker et al., 2011).

### **Nuclear foci: senescence associated heterochromatin foci (SAHF) and DNA scar**

During senescence onset in the cells, a drastic chromatin remodeling arises in the nucleus leading to the formation of DNA dense heterochromatic regions called **senescenceassociated heterochromatin foci (SAHF)** (Narita et al., 2003). SAHF can be determined by visualizing the reorganization of DAPI, as well as by immunoprecipitation of heterochromatinassociated histone H3K9Me2 and H3K9Me3 and its binding partner heterochromatin protein-1γ (HP-1γ) (Zhang et al., 2007). SAHF contain at least two other proteins, namely the histone variant MacroH2A and HMGA proteins (Narita et al., 2003; Zhang et al., 2007). Whereas HMGA proteins contribute to senescence-associated proliferation arrest (Funayama et al., 2006), MacroH2A protein confers to the chromatin a resistance to ATP-dependent remodeling and binding of transcription factors that in turn contribute to cell cycle exit (Zhang et al., 2005). The persistent DDR signaling triggered following DNA double-strand breaks also leads to the formation of nuclear foci, termed **DNA segments with chromatin alterations reinforcing senescence (DNA-SCARS)** where DDR proteins accumulate allowing the DNA-SCARS detection. These foci are detected through the staining of γH2AX, the phosphorylated form of histone H2AX. Following DNA damage, Histone H2AX is phosphorylated by the ATM/ATR protein and generates bright foci (Rogakou et al., 1998). The γH2AX thereafter recruits additional ATM/ATR complexes through involvement of two other proteins, the mediator of DNA damage checkpoint (MDC1) and p53-binding protein 1 (p53BP1). These two last proteins may also be used for the detection of DNA-SCARS (Martinez-Zamudio et al., 2017b). Recently new type of DNA damage consisting in cytosolic chromatin fragments that stands out from the nucleus of cells during senescence (Dou et al., 2017). This cytoplasmic chromatin fragment is recognized as double– strand DNA by cyclic GMP-AMP synthase (CGAS) a well-known pattern recognition receptor for foreign DNA in the cytosol (Stetson and Medzhitov, 2006). Once activated, CGAS activates the stimulator of interferon genes (STING) thereby instigating the SASP production (Gluck et al., 2017).

## **The senescence-associated secretory phenotype (SASP) components**

Even if the SASP components greatly vary as a function of cell type, some of them including the pro-inflammatory cytokines IL-6 and IL-8 appear highly conserved. IL-6 and IL-8 are thus considered as SASP-dependent senescence markers. In addition to IL-6 and IL8, other SASP components such as MMP-1 and−3, growth factors (GM-CSF, G-CSF, IGF), cytokines (IL-1α/β, TNF-α), and chemokines (MCP-1, MIP-1α, GROα/β) can be directly assayed in cell supernatants using Elisa or multiplex immunoassay.

# Relevance of Senescence in OA and Articular Cartilage Aging

Aging organisms accumulate senescent cells expressing p16INK4A in various tissues. This accumulation contributes to the onset of age-related disorders and morbidity by compromising tissue function and structural integrity (Martinez-Zamudio et al., 2017a). During aging and OA, chondrocytes exhibit many features of senescence, including growth arrest, presence of DDR and SASP production (Price et al., 2002). Indeed, chondrocytes like many other cell types and according to the Hayflick limit, can only undergo a restricted number of cell division (Evans and Georgescu, 1983). Consistently, aging chondrocytes exhibit a decreased proliferation capacity and extracellular matrix (ECM) synthesis (Guerne et al., 1995). With age articular cartilage undergo modification such as localized fibrillation and a decreased ability to respond to various anabolic stimuli, which weakens cartilage repair (Buckwalter and Mankin, 1998). Senescent cells have been observed near the OA lesion. This accumulation of senescent cells may predispose joints to OA development (Kozhemyakina et al., 2015). In addition to age-related senescence, a stress-induced senescence may be induced by pro-inflammatory cytokines on chondrocytes. The specific hallmarks of OA-associated chondrocyte senescence are discussed below.

# Oxidative Stress

Oxidative stress increases either when ROS production is improved or when antioxidants level is decreased, and intriguingly both phenomenon occur during aging (Finkel and Holbrook, 2000). Besides, elevated ROS levels have been shown to correlate with a decreased autophagy thereby suggesting that oxidative stress contributes to tissue homeostasis rupture and OA severity (Hui et al., 2016). This increased oxidative stress is also believed to result from a mitochondrial dysfunction, since OA chondrocytes exhibit reduced mitochondrial DNA content and reduced mitochondrial mass (Wang et al., 2015). This mitochondrial dysfunction is also accompanied by a decreased expression of sirtuin 1 (SIRT1) and peroxisome proliferator–activated receptor γ coactivator 1α (PGC-1α) both involved in mitochondrial biogenesis, and a down-regulation of the nuclear respiratory factors 1 and 2 (NRF1 and NRF2), which regulate antioxidant gene expression (Wang et al., 2015). This mitochondrial dysfunction was further confirmed in OA chondrocytes which exhibit a marked decrease in the complexes I, II, and III of the respiratory chain, triggering a reduction in the mitochondrial membrane potential (1ψm) and ATP synthesis (Maneiro et al., 2005). Interestingly, the pro-inflammatory cytokines IL-1β and TNFα, that are found at elevated concentration in the synovial fluid of OA joint, have been shown to alter mitochondrial activity through the reduction of respiratory chain complex I and the decrease in the 1ψm and ATP synthesis (Lopez-Armada et al., 2006). In line with these, elevated intracellular levels of ROS and particularly superoxide anion have been found in post-traumatic OA. This superoxide

anion accumulation is related to the downregulation of its degrading enzyme, the mitochondrial superoxide dismutase 2 (SOD2) (Koike et al., 2015). Consistently, the mitochondrial dysfunction associated with an imbalance between the ROS production and the antioxidant capacities of the cells is now considered a potent player in the onset and development of OA.

#### Senescent Markers in OA

Many of the above reported senescence-associated markers are overexpressed in OA-affected chondrocytes.

### **SA**β**-gal**

A positive SAβ-gal staining was observed in a subset of chondrocytes close to the lesion sites of mild, moderate and severe OA. Moreover, the level of SAβ-gal correlated with the severity of OA (Gao et al., 2016). Since SAβ-gal staining has been used in many studies in OA (Kim et al., 2014; Nagai et al., 2015; Platas et al., 2016; Jeon et al., 2017), it is now belonging to the conventional arsenal of senescence markers.

## **Telomere shortening**

It has been observed that with age, the average telomere length decrease in chondrocytes inducing replicative senescence and participating in the progression of OA (Martin and Buckwalter, 2001). However, since chondrocytes exhibit a poor division rate, telomere shortening may be more likely caused by chronic stress such as oxidative stress known to trigger DNA damage and cell senescence through ROS production (Martin et al., 2004; Davies et al., 2008; Brandl et al., 2011). Nonetheless, during OA the slight reduction of telomere length might be due to the slight increase in chondrocyte proliferation. Recently, a divergent subpopulation of chondroprogenitors with increased telomere erosion has been evidenced in OA cartilage (Fellows et al., 2017).

## **Sirtuins (SIRT)**

SIRT1-7 are NAD+-dependent deacetylase/deacylases that regulate a wide variety of biological functions. SIRT proteins, through the regulation of energy metabolism, contribute to cellular homeostasis and lifespan (Houtkooper et al., 2012). Among the seven members of the SIRT family, four (SIRT1, SIRT3, SIRT6, and SIRT7) have been studied in articular cartilage or chondrocytes as well as in OA. In the articular cartilage SIRT1 expression decrease with aging and cartilage-specific SIRT1 conditional knockout mice exhibit a more severe OA score at 1 year and an accelerated progression after post-traumatic OA (Matsuzaki et al., 2014b). Moreover, cartilage-specific SIRT1-conditional knockout mice exhibit increased chondrocyte apoptosis and MMP13 and ADAMTS5 expression levels. These data suggest that SIRT1 may have a preventive role in the development of OA via the suppression of the NF-κB pathway activation, a pathway also involved in the induction of SASP (Matsuzaki et al., 2014b). The protective role of SIRT1 was further confirmed by another study demonstrating that SIRT1 gene knock-out may exacerbate cartilage degeneration in OA by activating the SREBP2 protein-mediated PI3K/AKT signaling pathway (Yu et al., 2016). Consistent with this protective role of SIRT1, it has also been shown beneficial effects involving the inhibition of senescence through the activation of autophagy (Lim et al., 2017). Other members of the SIRT family, notably SIRT3, have also been associated with the development of OA. It has thus been demonstrated that SIRT3 is decreased in OA cartilage and associated with a concomitant reduction of SOD2 specificity and activity (Fu et al., 2016a). Consistent with this observation, the genetic ablation of SIRT3 has been found to accelerate the development of OA thereby strongly suggesting a protective role for SIRT3 (Fu et al., 2016a). SIRT6, which is localized to the nucleus and is involved in transcriptional silencing, genome stability, and longevity (Kanfi et al., 2012) has also been suggested as a protective factor in OA. SIRT6 is particularly implicated in the regulation of life span and aging through the regulation of NF-κB signaling and glucose homeostasis. SIRT6 level is significantly decreased in the articular chondrocytes of OA patients (Wu et al., 2015) and is preferentially expressed in cluster-forming chondrocytes of the superficial zone in OA cartilage (Nagai et al., 2015). SIRT6 inhibition is associated with an increased MMP1 and MMP13, decreased proliferation and increased number of SAβ-gal positive chondrocytes. Moreover, the senescent phenotype induced by SIRT6 inhibition has been further confirmed by the increase in P16INK4A expression and the presence of γH2AX foci (Nagai et al., 2015). In a second set of experiments, the overexpression of SIRT6 suppresses the replicative senescence of chondrocytes and reduces the expression of NF-κB dependent genes (Wu et al., 2015). Finally, the lentivirus-associated intra articular delivery of SIRT6 has been shown to protect mice from cartilage degeneration. As a whole, these data strongly suggest that the overexpression of SIRT6 can prevent OA development by reducing both the inflammatory response and chondrocytes senescence. The last SIRT member associated with OA is SIRT7. However, since SIRT7 KO mice are resistant to the development of age-associated OA and forced exercise-induced OA, SIRT7 is believed to exert detrimental effect on cartilage. Whereas its knockdown increases the deposition of a glycosaminoglycan-rich extracellular matrix, SIRT7 has also been shown to suppress the transcriptional activity of SOX9 thereby confirming its putative deleterious effect in OA and cartilage homeostasis (Korogi et al., 2018).

### **HMGB-1**

It is localized in the nucleus of non-senescent cells and secreted by senescent cells, making the loss of its nuclear staining a sign of senescence (Davalos et al., 2013). Immunohistochemistry analysis of synovial membrane reveals that HMGB1 expression shift from a strictly nuclear localization in healthy individuals to a nuclear and cytoplasmic localization in OA patients (Ke et al., 2015). Moreover, both the synovium and synovial fluid HMGB-1 levels were significantly higher in OA patients (Ke et al., 2015). In another study, a correlation between the total number of HMGB-1 positive chondrocytes and the OARSI scoring was found (Terada et al., 2011). It has notably been highlighted that in the highest OARSI grade, the number of chondrocytes with cytoplasmic HMGB-1 expression was increased in the deep layers of cartilage. Similarly, the level of expression of the receptor for advanced glycation end product (RAGE), a receptor able to

bind HMGB-1, also correlates with the severity of OA (Terada et al., 2011). Of particular interest in an OA context, HMGB-1 is released by chondrocytes or synoviocytes in response to pro-inflammatory cytokines such as IL-1β and TNF-α (Terada et al., 2011; Amin and Islam, 2014; Philipot et al., 2014). These results suggest that HMGB-1, as a pro-inflammatory cytokine, may play a crucial role in the progression of OA. In line with this observation, HMGB-1 is considered an alarmin that are released during tissue damage (Nefla et al., 2016). When secreted, alarmins enhance inflammatory response and catabolic process and participate in the progression of OA. HMGB-1 contains two DNA binding motifs, A and B box. The B box is considered the cytokine active domain of HMGB-1 which is involved in its pro-inflammatory effect, whereas the A box consists in a competitive binding site with antagonist properties. In a recent study, Fu et al. demonstrated that overexpression of HMGB1 A box inhibits the effect of IL-1β on chondrocytes through the suppression of HMGB1/TLR4/NF-κB pathway (Fu et al., 2016b). Recently, the loss of nuclear HMGB1 staining was clearly correlated with an increase in the P16INK4 senescent marker expression in post-traumatic OA chondrocytes (Jeon et al., 2017). Of particular clinical interest, the clearance of senescent cell was found to reestablish the nuclear staining of HMGB1 (Jeon et al., 2017) in the same post-traumatic OA models.

#### **CAV1**

CAV1 as described in the section Identification of Senescent Cells is involved in the senescence of various cell type (Zou et al., 2011). In chondrocytes, IL-1β and H2O2, both senescence inducer, increase CAV1 expression, and promote senescence phenotype such as telomere erosion, SAβ-gal activity and altered morphology through activation of the P38MAPK pathway (Dai et al., 2006; Yudoh et al., 2009). All these senescenceassociated phenotypic alterations induced by IL-1β and H2O<sup>2</sup> are consistently inhibited by an antisense oligonucleotide targeting CAV1 (Dai et al., 2006). These results strongly suggest that CAV1 is involved in IL-1β-induced senescence in chondrocytes. Finally, increased expression of CAV1 has also been observed in OA cartilage samples and is correlated with the disease severity (Min et al., 2015). However, considering the contradictory data regarding the effect of CAV1 overexpression or deficiency on senescence in other organs, the role of CAV1 in OA cartilage needs further investigations notably through the use of cartilage specific CAV1 conditional knockout mice.

## **p16**INK4<sup>A</sup>

It is physiologically involved in the processes of cartilage aging, and may be partly responsible for the senescence of chondrocytes as seen in OA (Zhou et al., 2004). Indeed, a significant increase in p16INK4a was detected in OA chondrocytes as compared to age-matched healthy chondrocytes in vivo and in vitro (Zhou et al., 2004). Moreover, P16INK4 silencing in OA chondrocytes has been found to increase chondrocyte proliferation and DNA synthesis and to decrease SAβ-gal staining (Zhou et al., 2004). The expression of P16INK4A has been also associated with the terminal differentiation of chondrocytes in vitro (Philipot et al., 2014) and IL-1β treatment of human OA chondrocytes induce the expression of P16INK4A, which leads to an increased production of OA-associated catabolic proteases MMP1 and MMP13 (Philipot et al., 2014). Recently, the causal role of P16INK4A positive cells in the OA onset has been demonstrated. Through the use of the P16-3MR transgenic mice, the kinetics of senescence after OA surgical induction was investigated. Using this murine model as well as the senolytic compound UBX0101, they undoubtedly demonstrate the beneficial effect of removing p16INK4A positive senescent cells in OA, as evidenced by a reduction in cartilage damage and the promotion of cartilage repair (Jeon et al., 2017).

### **SASP**

Senescent cells produce a bioactive secretome which participate in the propagation of senescence to neighboring cells (Hoare and Narita, 2013). Numerous SASP components described in senescent cells are also present in OA tissues or synovial fluid (Kapoor et al., 2011). The implication of some SASP factors in OA will be addressed below. The proinflammatory cytokines IL-6, IL-8, TNF-α, IL-1α, and IL-1β are particularly relevant in OA (Robinson et al., 2016). These cytokines have been shown to promote OA progression, notably by altering chondrocytes function and viability and by inducing synovitis (Kapoor et al., 2011). IL-1β and TNFα drives chondrocytes toward a senescent phenotype (Philipot et al., 2014) comprising the secretion of a SASP containing IL-6, IL-8, MMP-13, MMP-1, MMP-9, MMP-3, and MMP-14, ADAMTS4, 5, and 9 as well as PGE2 and NO which participate in pathogenesis. IL-1β and TNFα are the most widely used cytokines to induce an OA-like phenotype in chondrocytes in vitro (Fukui et al., 2003; Carames et al., 2008) and are now recognized as senescent inducers (Platas et al., 2016; Song et al., 2018). The effect of IL-1β and TNFα are mediated by the activation of p38 MAPK and NF-κB signaling pathways (Ovadya and Krizhanovsky, 2014). IL-6 level is increased in synovial fluid of OA patients (Kaneko et al., 2000) and IL-6 production is directly stimulated by IL-1β and TNF-α in chondrocytes (Guerne et al., 1990). However, the role of IL-6 in OA is controversial. On the one hand IL-6 upregulates MMP-1 and MMP-13 expression, and on the other hand, IL-6 intra articular injection in IL-6 deficient mice reduced cartilage degeneration (van de Loo et al., 1997). This opposite effect of IL-6 cytokine may be due to the type of IL-6 receptor engaged to mediate its signaling (membrane-bound or soluble form; Hunter and Jones, 2015). IL-8 is significantly increased in the synovial fluid of OA patients and induces chondrocytes hypertrophy (Takahashi et al., 2015). Matrix degrading enzymes such as a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS4 and ADAMTS5) and Metalloproteinases (MMP-1, MMP2, MMP-3, MMP-14) are also involved in OA pathogenesis. MMP-13 and MMP-1 are secreted by IL-1β-induced senescent chondrocytes, which promote the degradation of aggrecan and type II collagen (Philipot et al., 2014). ADAMTS4 is also induced by IL-1β and TNFα in chondrocytes. Many chemokines are produced in OA, but CCL2/MCP-1 is considered a major component of the SASP. CCL2 indeed up-regulates the expression of MMP13 and increase proteoglycan loss in vitro. MCP-1 synovial fluid level is positively

correlated with OA severity and pain (Miller et al., 2012; Li and Jiang, 2015). In line with this observation, mice lacking CCL2 or its receptor CCR2 have recently been found protected against OA (Raghu et al., 2017). VEGF is also a common component of the SASP, and VEGF, which is expressed in OA cartilage, participate in the dysregulation of bone remodeling through promoting osteophytes formation.

Reinforcing the idea of the production of a SASP by chondrocytes during OA, the NF-κB pathway described above as the main pathway governing the SASP production (Chien et al., 2011) is also an important signaling pathway in OA. Indeed, several studies reporting beneficial anti-inflammatory effects of drugs on OA are mediated through the inhibition of the activation of NF-κB pathway.

Specially promising are the beneficial effects obtained by selective removal of the senescent chondrocytes from OA patients, which decreases expression of senescent and inflammatory biomarkers and increases expression of ECM proteins (Jeon et al., 2017). There are in fact progenitors with senescence phenotypic characteristics (βgalactosidase activity and telomere shortening) derived from osteoarthritic cartilage (Fellows et al., 2017). Together, these preclinical data support the concept of selective targeting senescent chondrocytes as a novel therapeutic strategy for OA.

# AUTOPHAGY IN OA

# General Context

Autophagy is the cellular mechanism that degrades proteins and organelles to maintain cellular homeostasis and quality control (Sridhar et al., 2012). Physiological functions such as remodeling, differentiation, survival, death, senescence, and longevity are regulated by autophagy (Cuervo and Macian, 2012). The immune response is also regulated by autophagy in infection, inflammation and adaptive immunity, degrading intracellular pathogens, and presenting antigens and activating lymphocyte proliferation (Deretic et al., 2013; Arroyo et al., 2014). In normal conditions, autophagy constitutively occurs in all mammalian cells, and can be regulated as a consequence of stresses such as nutrient deprivation, heat, oxidation, and infection (Meijer and Codogno, 2009). Autophagy involves multiple steps including initiation, nucleation, elongation, maturation, and degradation. The steps involving the autophagosome formation result in the enzymatic degradation of the sequestered products that are transformed into nucleotides, amino acids, fatty acids, and sugar, which can be recycled into metabolic pathways to generate energy and build macromolecules (Glick et al., 2010). These rate-limiting steps are mediated by the autophagy-related genes (ATGs) (Mizushima et al., 2011; Abounit et al., 2012). Uncoordinated-51-like kinase (ULK) complex, composed by ULK1/2, Atg13, Focal Adhesion Kinase (FAK), Family Interacting Protein of 200 kDa (FIP200) and ATG101, regulates the induction of the autophagosome formation. This complex is the final target of Mammalian Target of Rapamycin (mTORC1) and Adenosine 5 ′Monophosphate-Activated Protein Kinase (AMPK) signaling cascades (Wesselborg and Stork, 2015). mTORC1 is the main inhibitory signal of autophagy, and is regulated by growth factors, energy levels and nutrient availability, among other stress signals. Signaling converges through Phosphatidylinositol 3- Kinase (PI3K), Mitogen-Activated Protein Kinase (MAPK), and p90 Ribosomal S6 kinase (RSK) pathways, each one inhibiting Tuberous Sclerosis Heterodimeric Complex (TSC1/TSC2) complex. TSC1/TSC2 complex inhibition will subsequently stimulate mTORC1 (Sengupta et al., 2010). In respect to energy status, the decreased ATP levels activate AMPK, which in turn activates the TSC1/TSC2 complex, resulting in mTORC1 inactivation (Wullschleger et al., 2006). Indeed, nutrient richconditions activate mTORC1 that conjugates with ULK complex, avoiding the conjugation of ULK with AMPK, and results in autophagy inhibition. On the other hand, in energy depletion and starvation, mTORC1 dissociates from ULK, allowing the association between AMPK and ULK which consequently results in autophagy induction (Papinski and Kraft, 2016). The nucleation is mediated by PI3K, composed by a central core complex resulting from the association between class III PI3K (Vacuolar Protein Sorting Protein 34, hVps34), beclin-1 and p150. This complex is responsible to the expanding phagophore (Funderburk et al., 2010). The elongation and enclosure of the autophagosome involves 2 ubiquitin-like conjugation systems: Atg12 and LC3/Atg8. These two conjugation systems are essential for autophagosome formation (Geng and Klionsky, 2008; Glick et al., 2010). Accumulation of undegraded substrates on lysosomes may have a great impact on cells and tissues, highlighting the importance of lysosome dysfunction and impaired proteostasis as a major outcome of failure of autophagy and a key feature of senescence. This particularly relevant in metabolic tissues such as liver and heart (Schneider et al., 2015; Gianfranceschi et al., 2016), but might also be true in musculoskeletal tissues.

# Relevance of Autophagy in Articular Cartilage and OA

Autophagy is essential to preserve the integrity and function of articular cartilage. Normal human cartilage expresses high levels of autophagy regulators, including ULK1, Beclin1, and LC3-II, suggesting that autophagy is a constitutively active mechanism in cartilage. Seminal studies in human OA cartilage and in an experimental OA, demonstrated that autophagy is abnormally reduced (Carames et al., 2010). Consistent with these findings, mTOR was overexpressed in human OA cartilage and mouse models of experimental OA (Zhang Y. et al., 2015). Moreover, a positive correlation of mTOR activation in both peripheral blood mononuclear cells (PBMC) and articular cartilage was found in end-stage OA patients (Tchetina et al., 2013; Zhang et al., 2017). On the other hand, LC3-II and Beclin1 are upregulated in OA chondrocytes (Sasaki et al., 2012), while OA tissues display numerous LC3 puncta via hypoxia-inducible factor 2 (Bohensky et al., 2009). A potential explanation could be the different location of harvested OA cartilage potentially corresponding to different disease stages (cartilage from lateral femoral condyles was identified as mild OA, while cartilage from medial femoral condyles was identified as severe OA), where autophagy markers

were differentially expressed within these two regions. Mild OA cartilage had a strong expression of autophagy markers compared to non-OA cartilage and severe OA cartilage. During OA progression, autophagy may act as an adaptive response in an attempt to protect cells. However, once severe OA is established, a decrease in autophagy is detected and could eventually contribute to OA progression (Sasaki et al., 2012).

A similar dual role of autophagy can be seen regarding its effect on cell death. Many authors consider the activation of autophagy as a protective mechanism that avoids chondrocyte death (Carames et al., 2012a,b; Sasaki et al., 2012), while others propose that reduction of autophagy is often accompanied with increased apoptosis (Carames et al., 2010; Zhang Y. et al., 2015). Moreover, the up-regulation of autophagy suppresses glucocorticoid-stimulated chondrocyte apoptosis (Liu et al., 2014), paradoxically constituting an alternative form of cell death generally observed in different pathological disorders (Levine and Yuan, 2005; Maiuri et al., 2007). Indeed, exposure to monosodium urate crystals promotes chondrocyte death through autophagy activation (Hwang et al., 2015). Young chondrocytes are protected from cell death in OA chondrocytes (Chang et al., 2013). Autophagy might play a possible dual role at different stages of progression: protective and a death-promoting in OA pathogenesis, where the death of chondrocytes appears to be associated to a complex interaction between autophagy and apoptosis across different stages of OA as well as different anatomical locations (Almonte-Becerril et al., 2010). These conflicting results may suggest that autophagy could promote either chondrocyte survival or death depending on donor age, the presence and stage of OA, and the type of autophagy induction (Hwang et al., 2015). In fact, this dual role of autophagy is consistent with studies in other cell types and tissues, mainly dependent on the type of stress stimuli (Chen et al., 2010). Notwithstanding, the relationship between autophagy and cell death is not fully understood and thus prompts additional studies to decipher the underlying cell signaling mechanisms (Musumeci et al., 2015).

There is a consensus at recognizing the protective role of autophagy under stress conditions. Both nutritional (starvation) and catabolic stresses (IL-1β or sodium nitroprusside, SNP) increase autophagy in chondrocytes (Sasaki et al., 2012). In cartilage challenged with mechanical injury, initial upregulation of LC3-II is detected at 24 h, while 48 and 96 h after injury, LC3- II levels appear to decrease (Carames et al., 2012b). Similarly, the biomechanical dental stimulation that leads to degradation of the cartilage from the temporomandibular joint also increases autophagy as an early response (Zhang et al., 2013). Intermittent cyclic mechanical tension (ICMT) leads to calcification of end plates of the intervertebral disc, which is responsible for its degeneration. Short-term ICMT increased autophagy and it was accompanied by an insignificant calcification of end plate chondrocytes. However, long term ICMT suppressed autophagy, leading to endplate chondrocyte calcification. Although the cartilage of intervertebral disc differs from articular cartilage, a similar mechanism may be involved (Xu et al., 2014; Xu, 2015). Furthermore, in response to energy stress through mitochondrial dysfunction, chondrocytes showed an early increase in autophagy as a compensatory mechanism. However, when prolonged stress exceeds cellular compensation, damage occurs (Lopez de Figueroa et al., 2015). Both energy and mechanical stress activate autophagy initially, but this is probably insufficient to protect cartilage in the long-term and eventually autophagy becomes defective. These results provide strong evidence that autophagy has an important role in protecting chondrocytes from different stressors and, therefore, can be involved in OA pathogenesis.

# Major Targeted Pathways to Elucidate the Relevance of Autophagy in Articular Cartilage and OA PI3K/AKT/mTOR

mTOR is been extensively studied as a key autophagy regulator in OA. Inducible cartilage-specific mTOR KO mice subjected to experimental OA by destabilizing the medial meniscus (DMM model) have less cartilage degradation, more proteoglycans and chondrocytes, as well as less signs of synovial fibrosis. Autophagy markers (i.e., ULK1, AMPK1, Atg5, LC3) are significantly increased, while apoptotic cells and catabolic factors, such as MMP-13 and MMP-induced type II collagen breakdown product C1, 2C, are reduced (Chen et al., 2013). In human disc cells, mTORC1/RAPTOR silencing protects against inflammation through AKT and autophagy induction (Ito et al., 2017). An interesting link between mTOR pathway and hormonal regulation was found in cartilage-specific Tsc1 KO and inducible Tsc1 KO mice, where articular chondrocyte proliferation and differentiation are activated to initiate OA, in part by downregulating FGFR3 and PPR (Zhang et al., 2017). Sestrins, a family of highly conserved stress-responsive proteins that are transcriptionally regulated by p53 and forkhead transcription factor (FoxO), protect cells from stress conditions by regulating AMPK and mTOR signaling. In aging and OA cartilage, sestrins are suppressed and promote autophagy by inhibiting mTOR, contributing to reduce homeostasis (Shen et al., 2017). Regulated in development and DNA damage response 1 (Redd1), an inhibitor of mTOR signaling, is regulated by ubiquitin ligases and is highly expressed in normal human cartilage and reduced in aging and OA (Alvarez-Garcia et al., 2016). In mice lacking Redd1, autophagy and mitochondrial biogenesis are reduced, which leads to OA (Alvarez-Garcia et al., 2017). The regulatory role of mTOR in autophagy also control inflammatory responses and can prevent both cartilage damage and OA (Salminen et al., 2012; Rahmati et al., 2016). Increased mTOR expression in peripheral blood mononuclear cells (PBMC) was associated with synovitis (Tchetina et al., 2013). In inflammatory arthritis in vivo, mTOR inhibition reduced osteoclast numbers and activity, protected against local bone erosions and cartilage damage, and decreased synovitis (Cejka et al., 2010). A similar association was observed in OA, where mTOR deletion reduced synovial inflammation and decreased IL-1β expression (Carames et al., 2012a). Moreover, chondrocytes from PPARγ KO mice showed an increase in inflammatory mediators (i.e., COX-2 and iNOS), enhanced expression of mTOR and a decrease in autophagy markers (Vasheghani et al., 2015). On the other hand, human OA chondrocytes treated with IL-1β show increased expression of mTOR in conjunction with increased expression of catabolic factors and a decreased expression of collagen type II, suggesting that pro-inflammatory cytokines can also alter mTOR pathway in OA (Zhang Y. et al., 2015). Viewed together, these studies suggest an important connection between mTOR, inflammation and cartilage damage.

### Nuclear Receptors and Transcription Factors

PPARγ plays a protective role in articular cartilage. PPARγ KO mice exhibited increased apoptosis as well as production of inflammatory and catabolic factors, and decreased expression of anabolic factors, resulting in an accelerated OA (Vasheghani et al., 2013, 2015). A significant reduction in LC3-II and increased mTOR expression were detected in those animals, while in chondrocytes, the restoration of PPARγ expression downregulates mTOR and up-regulates LC3-II expression. Catabolic and inflammatory factors are decreased, while anabolic factors are increased. PPARγ-mTOR double KO mice are protected from experimentally induced-OA. Degradation of cartilage is less severe, less proteoglycan and chondrocyte loss is observed, and these beneficial effects are associated with increased LC3-II and reduced MMP-13 expression. Thus, decreased PPARγ contributes to mTOR compensatory upregulation that is responsible for autophagy suppression, lead to chondrocyte death and increased catabolic activity, which ultimately accelerates OA. Both in vivo studies confirm that mTOR deletion protects against OA, thus reinforcing the role of decreased autophagy in OA development.

Impressive proof-of-concept in the context of cartilage biology and OA has recently been shown for the role of FoxO transcription factors as major regulators of autophagy, metabolism and aging of the joint (Matsuzaki et al., 2018). FoxO plays a key role in development, aging and longevity. Articular cartilage degradation is caused by age, genetic and environmental challenges, which ultimately lead to OA. In the absence of insulin and growth factor signaling, FoxO are translocated to the nucleus and result in the integration and activation of a cascade of key target genes that cause cell cycle arrest, stress resistance, and cell death (Webb and Brunet, 2014). FoxOs trigger autophagy in a variety of tissues, including skeletal muscle and cartilage. Human chondrocytes from patients suffering OA have FoxO1 and ATGs reduced, and restoring FoxO1 decreased inflammatory cytokines and up-regulated lubricin, a secreted proteoglycan that contributes to the lubrication and minimize friction of joint synovium. In this study, Matsuzaki and collaborators demonstrated the importance of FoxO on postnatal cartilage development, maturation, and homeostasis of cartilage. In young mice lacking FoxO1/3/4, cartilage was thicker and chondrocytes were more proliferative. Chondrocyte-specific FoxO-deficient mice exhibited severe joint damage with aging and increased cartilage degradation in response to surgically induced OA. More importantly, expression of superficial zone protein PRG4, was significantly reduced. Further studies to investigate the molecular mechanism of FoxO on cartilage superficial zone homeostasis indicate that FoxO1 is a transcriptional factor of PRG4 in cartilage. In vitro studies in IMACS and ATDC5 chondrogenic cells showed that overexpression of FoxO1 significantly upregulates the expression of PRG4. Moreover, previous studies shown reduced FoxO in aging and OA (FoxO 1, 3, and 4) associated with increased susceptibility to cell death induced by oxidative stress and reduced levels of antioxidant proteins as well as autophagy (Akasaki et al., 2014a,b). Thus, these results indicate the importance of FoxO in aging and OA and suggest that targeting FoxO transcription could be a novel strategy to prevent or delay OA progression.

p63 belongs to p53 family and play an important role in tissue development. Global and tissue-specific overexpression of p63α and p63γ mice with aging or surgically induced instability showed significant resistance to OA development and suppression of chondrocyte apoptosis (Taniguchi et al., 2017). Proapoptotic genes were increased in chondrocytes along with in the growth plate. In contrast, p53 was more predominant in the superficial zone of articular cartilage, as opposed to p63. Immunosuppressive drugs have deleterious effects on growth and senescence of articular chondrocytes from rabbit (Kang et al., 2016).

# Aging-Related Autophagy

Aging is one of the major risk factors for OA, precipitating onset predominantly among adults 60 years of age or older (Lotz and Loeser, 2012). The low turnover of ECM and articular cartilage likely increases the sensitivity to accumulate age-related changes, including a disturbed structural organization of ECM due to formation and accumulation of AGEs, cartilage calcification, and fibrosis, which alter cartilage mechanical properties (Roberts et al., 2016). Cellular age-related changes, those include senescence, reduction in the number of chondrocytes, mitochondrial dysfunction, and altered growth factors responsiveness. Alterations in protective mechanisms such as decreased antioxidant defense and reduced autophagy also occur with aging, disturbing the anaboliccatabolic equilibrium and reducing the remodeling and repairing ability of cartilage (Lotz and Loeser, 2012; Hui et al., 2016). Autophagy can regulate age-related changes in articular cartilage. Both aging human and mice chondrocytes exhibit a reduction in constitutive autophagy. Moreover, compromised autophagy as a consequence of aging precedes the decrease in cartilage cellularity and the onset of structural damage (Carames et al., 2010, 2015; Hui et al., 2016). Chondrocyte-specific ablation of autophagy gene Atg5 in mice leads to age-related OA but no alteration of injury-induced OA (Bouderlique et al., 2016). Altered tissue repair, misbalanced homeostasis and defective autophagy are important hallmarks of aged tissues in general (Cuervo et al., 2005; Cuervo, 2008). Consistently, the molecular mechanisms occurring during aging can explain the impairment in autophagy, including defects in induction and inefficient lysosomal clearance. For instance, the altered hormonal regulation of autophagy occurring in aged tissues can be a possible explanation for the compromised autophagy observed in old organisms caused by glucagon upregulation (Cuervo et al., 2005). Signaling pathways involving pro-longevity factors such as Sirtuin 1 (SIRT1), transcription factor forkheadbox O3 (FOXO3), and pro-senescence factors NF-kB and p53 are also known autophagy regulators and thus can be involved in cartilage aging (Salminen and Kaarniranta, 2009). Decreased autophagy can be explained by the impaired capacity of the lysosomes to fuse with autophagosomes and the failure of lysosome hydrolases that decrease the proteolysis efficiency of lysosomes (Cuervo, 2008). In fact, the accumulation in the lysosomes of undegradable materials that occurs with aging hampers the degradative capacity and results in decreased autophagy (Brunk and Terman, 2002a,b; Terman and Brunk, 2004a).

In cardiac and skeletal muscle of mice and humans with progeria, autophagy decreases with aging, while mTOR activity is elevated (Ramos et al., 2012; Sandri et al., 2013). Pharmacological and genetic inhibition of mTOR extends lifespan by unknown mechanisms (Lamming, 2016). mTOR inhibition represses autophagy promoting the degradation of aberrant proteins and damaged organelles, thus protecting from toxicity, and consequently slowing aging. Other processes including the regulation of protein synthesis, regulation of mitochondrial function, anti-inflammatory effects, increased stress resistance and preservation of stem-cells might also contribute to the prolongevity effects of mTOR inhibition (Johnson et al., 2013). Aging and OA are clearly intertwined, and autophagy might be a common mechanism involved in cartilage degradation in both conditions.

Kashin-Beck is a rare disease of the bone occurring in children, with high prevalence in certain areas of Asia. Functional genetic studies identified ATG4C as a novel susceptibility gene in Kashin-Beck (Wu et al., 2014). In normal human chondrocytes stimulated with IL-1β, Parkin, a component of the multiprotein proteasome complex, eliminates dysfunctional mitochondria promoting survival and protecting from apoptosis (Ansari et al., 2017). In age-related and surgical OA mice, Bach1 deficiency reduces the severity of articular cartilage damage probably by antioxidant effects of HO-1 and downregulation of ECM catabolic enzymes (Takada et al., 2015). Chondrocyte-derived extracellular organelles from adult porcine and osteoarthritic patients evaluated by light scatter nanoparticle counting have constitutive autophagy that is increased with rapamycin and suppressed by autophagy inhibitors and genetic silencing of ATG5, supporting the role of autophagy in cartilage disease and repair (Rosenthal et al., 2015).

### microRNA

Small non-coding RNA molecules have been extensively used as tools to study the post-transcriptional regulation of gene expression. miRNA have been manipulated in mice, articular cartilage and chondrocytes to test different hypothesis related to musculoskeletal system physiology and disease. In OA model mice and SW1353 human chondrosarcoma cells treated with interleukin-1β, miR-17-5p was decreased (Li et al., 2018). Autophagy was found suppressed in knee joints mainly through suppressing p62/SQSTM1, an autophagosome cargo protein that targets other proteins that bind to it for selective autophagy. In an hypoxic environment as occurs in aged articular cartilage, miR-146a targets related to inflammation Traf6 and IRAK1, as well as chondrocyte-specific Smad4 were down-regulated (Chen et al., 2017). Traf6 and IRAK1 were identified as regulators of autophagy. Previous reports found that miR-146a promote chondrocytes autophagy via depressing Bcl-2 (Zhang F. et al., 2015). In primary human and TC28a2 chondrocytes, miR-155 inhibited autophagy and contributed to the autophagy defects in OA (D'Adamo et al., 2016). In rats with OA, miR-4262 regulates chondrocyte fate by influencing PI3K/AKT/mTOR signaling (Sun et al., 2018). In OA chondrocytes and in zebra fish, Fis1 suppression induces accumulation and inhibition of lysosomes by altering miRNAs and energy signals (Kim et al., 2016). These results indicate that post-transcriptional targeting of autophagy with specific miRNAs might be considered as a promising preclinical model to study musculoskeletal physiology and disease.

These pathways span a broad range of biological processes, and therefore targeting specific pathways with small molecules should be taken into consideration regarding the wide endogenous functions of these pathways in non-target tissues. **Table 1** summarizes relevant advances in the autophagy-related molecular mechanisms in OA.

# CURRENT DEVELOPMENT IN INNOVATIVE THERAPEUTIC APPROACHES

Although many advances to understand the pathophysiological processes have been made, no effective treatments to stop or prevent OA exist. Few symptomatic treatments are available, mainly focusing on pain relief and improving joint function. Current clinical management relies in a combination of pharmacological approaches with surgical procedures. Also, decisions on treatment greatly depend on different patient-related factors, such as the occurrence of other co-morbidities and the presence of one or multiple affected joints. Pharmacological options include analgesic and anti-inflammatory agents primarily. When the goals are not achieved with conservative treatment and the disease progresses to diminished quality of life, surgical options are considered. However, the burden, risks and lack of long-term efficacy that come with surgical interventions, ultimately may affect the quality of life of patients and impact the health care system. Thus, efforts focused on generating strong preclinical evidence that can lead to the development of novel targeted therapies are necessary.

# Approaches Targeting Autophagy as OA Therapeutics

Pharmacological activation of autophagy with small molecules has been the subject of several studies showing in some cases some preclinical efficacy in OA (**Table 2**). Targeting PI3K/AKT/mTOR pathway to modulate autophagy stands out as a major focus of pharmacological interventions in preclinical studies. Rapamycin induces autophagy activation through the inhibition of mTORC1 in many cell and tissues (Galluzzi et al., 2017). It is used as an immunosuppressive in transplantation and autoimmune disorders, and in cancer due to its activity at inhibiting growth and proliferation of tumor cells


#### TABLE 2 | Preclinical treatments targeting autophagy in OA.


(Lamming et al., 2013). In articular cartilage, mTOR inhibition by Rapamycin protects from oxidative stress, and cell death (Carames et al., 2012b; Sasaki et al., 2012; Lopez de Figueroa et al., 2015), increases the expression of aggrecan and type II collagen, while decreases MMP-13 in OA chondrocytes (Zhang Y. et al., 2015). C57Bl/6J mice with experimentally induced OA treated with Rapamycin show autophagy activation, including increased LC3-II levels and inhibited rpS6 phosphorylation in articular cartilage. Treatment also protects against OA-like alterations, including the maintenance of cartilage cellularity and prevention of ECM damage. Furthermore, the severity of synovitis was attenuated and inflammatory cytokine IL-1β was decreased. These results suggest that induction of autophagy can significantly reduce the severity of experimental OA (Carames et al., 2012a). It is interesting to note that beneficial effects are also observed in meniscus, a specific type of cartilage not present in all human joints (Meckes et al., 2017). However, systemic long-term administration of Rapamycin is associated with side effects including edema, mucositis, hair and nail disorders, and dermatological effects, among others, making its utility problematic for chronic conditions (Lamming et al., 2013). In order to avoid these side effects, it has been suggested the intra-articular injection as a more appropriate route for clinical use. Indeed, intra-articular injection reduces articular degeneration and preserves hyaline cartilage, by downregulating cartilage degrading enzymes (MMP9 and MMP13), hypertrophyrelated genes (vascular endothelial growth factor-VEGF- and COL10A1), inflammation-related genes (IL-1β and IL-6), and stress-responsive genes (CCAAT/enhancer binding protein beta-C EBPβ and mTOR) and by upregulating Col2a1 (Matsuzaki et al., 2014a; Takayama et al., 2014). These studies show that direct administration of autophagy activators in the joint have protective effects, while potentially avoiding the side effects observed with systemic administration. Selective ATP-competitive inhibition of mTOR with Torin1 blocks phosphorylation of both mTORC1 and mTORC2 and induces autophagy by blocking signals required for cell growth and proliferation (Thoreen et al., 2009). Torin 1 treatment reduces severity of OA pathology in rabbits (Cheng et al., 2016a,b), highlighting again the important role of PI3K/AKT/mTOR in articular cartilage degeneration. However, it is important to consider the negative feedback loop of the PI3K/Akt pathway. mTOR signaling regulates protein synthesis and autophagy mediated, at least in part, through changes in the phosphorylation of downstream molecules such as p70S6K, 4E-BP1, and ULK1. Ultimately, the cascade leads to the inhibition of the insulin receptor substrate (IRS1), an upstream regulator of PI3K/Akt (Laplante and Sabatini, 2012). On the other hand, mTOR inhibition leads to an enhancement of PI3K/Akt signaling and increased MMP production by chondrocytes. Based on this, Chen et al. proposed the dual inhibition of both pathways as a more promising approach for OA to abolish the negative feedback mechanism and its possible side effects (Chen et al., 2013). On the other hand, excessive inhibition of mTOR gene expression could be harmful for tissues. OA outpatients with lower mTOR gene expression in PBMCs compared to healthy controls, exhibited more pain while standing and upon joint function, and have increased total joint stiffness than OA outpatients with higher mTOR gene expression (Tchetina et al., 2013). Inhibition of PI3K/AKT/mTOR promotes autophagy and attenuates inflammation in rat articular chondrocytes (Xue et al., 2017). Chronic upregulated autophagy has been associated with compromised lifespan and pathology. For instance, mouse models of premature aging show systemic degeneration and weakening of the musculoskeletal system partly due to DNA damage defects (Marino et al., 2008). Therefore, continuous autophagy activation might not likely be the best therapeutic approach. Instead, repairing autophagy defects, or restoring the normal levels of autophagy, could be a more efficient therapeutic strategy for age-related diseases. Sugars, as precursors of the synthesis of glycosylated proteins and lipids, have also been tested as therapeutic agents for OA. Glucosamine is a dietary supplement marketed to support the structure and function of joints. In chondrocytes and in articular cartilage, it protects nucleus pulposus cells through the inhibition of mTOR and consequent activation of autophagy (Carames et al., 2013; Jiang et al., 2014). Primary human chondrocytes treated with sucrose are protected from degradation, chondrocyte inflammation and death via PI3K/Akt/mTOR (Khan et al., 2017), and trehalose reduces ER and oxidative stress-mediated by autophagy stimulation (Tang et al., 2017). Although these results are promising, the time-dependent dual role of glucosamine (shortterm exposure increases autophagy, while long-term exposure have inhibitory effects (Kang Y. H. et al., 2015), making these data difficult to interpret. Interestingly, drugs widely used in the management of OA such as glucocorticoid dexamethasone have also shown a similar dual profile of action in chondrocytes via FOXO3 and in meniscus via inositol 1,4,5-trisphosphate receptor (Liu et al., 2014; Shen et al., 2015a,b). Perhaps, autophagy activation is an early protective response, whereas prolonged treatments eventually lead to permanent autophagy defect. These dual effects might be due to the particular metabolism and pharmacokinetics of these molecules and the wide interference with diverse metabolic and inflammatory pathways. Whether this dual effect is generally observed with such molecules potentially interfering with autophagy, remains to be clarified. Glucose metabolism might play a relevant role in cartilage as diabetic mice with experimental OA have is degenerated joint tissue and Rapamycin treatment decreases both pathology and inflammation (Ribeiro et al., 2016b). Indeed, insulin might exert catabolic effects in chondrocytes, inducing loss of proteoglycans and inflammation by downregulating autophagy, similarly to what is observed in chondrocytes from diabetic-OA patients (Ribeiro et al., 2016a). On the other hand, sirtuins have arisen as potential targets of phenol-like molecules such as Hydroxytyrosol to prevent inflammation by promoting cell autophagy (Cetrullo et al., 2016; Zhi et al., 2018).

Preclinical studies targeting autophagy and related molecular mechanisms in cartilage with advanced therapies medical products such as gene and cell therapy have also received some attention. Intra-articular treatment with Parathyroid hormone (1–34) alleviates cartilage degeneration by activating autophagy (Chen et al., 2018). Adiponectin protects chondrocytes from apoptosis related to oxidative stress by inducing autophagy possibly via PI3K/Akt/mTOR (Hu et al., 2017). Other biological products such as Platelet rich plasma (PRP), adipose-stem cells, and serum have also shown efficacy at protecting from degeneration and inflammation through autophagy-related mechanisms in cartilage disease (Jiang et al., 2016; Moussa et al., 2017; Zhang et al., 2018).

Gene therapy and genome editing by nucleases has dramatically expanded the toolbox for the identification and therapeutic validation of targets. Agents interfering with transcription to achieve therapeutic benefit have also received special attention. Nanoparticles targeting NF-κB not only are bioavailable, but reduce early apoptosis, synovitis, and

maintain cartilage homeostasis by enhancing AMPK signaling that ultimately attenuate inflammation (Yan et al., 2016). Other examples include miRNA targets (miR-20, miR-30b) that maintain cartilage homeostasis by enhancing autophagy in chondrocytes (Chen et al., 2016; He and Cheng, 2018). One would expect a dramatic transformation of drug discovery pipelines by the generation of engineered cell-based models with disease-relevant genetic modifications in the upcoming years.

# Current and Future Therapeutic Approaches of Inflammation in OA

OA is now considered an inflammation-associated multifactorial disorder. As compared to rheumatoid arthritis, the inflammation in OA is chronic, comparatively low grade and mainly mediated by the innate immune pathways including complement and pattern-recognition receptors such as damage associated molecular patterns (Robinson et al., 2016). The recent identification of novel regulators (Liu-Bryan and Terkeltaub, 2015) of inflammation in OA holds promise for the development of transformative therapies. Among these emerging regulators, the bioenergy sensors serine/threonine kinase AMPK and sirtuin-1 (SIRT1) have received particular attention. The role of AMPK and SIRT1 is actually not restricted to energy metabolism, but includes resistance to stress and inflammation. It is noteworthy their role in the tuning of mitochondrial activity to adapt energy consumption, in the resolving of inflammation via the down-regulation of NF-κB, in the control of matrix catabolic responses and last but not least in promoting autophagy. These wide range activities of AMPK and SIRT1 make it a promising entry point for the development of novel therapeutics. A large variety of AMPK activators such as aspirin (Hui et al., 2016), metformin or methotrexate are clinically used in inflammatory disorders (O'Neill and Hardie, 2013). Their use in OA is conceptually attractive and sound, and the results of randomized placebo-controlled trials will certainly bring conclusive data.

The positioning of such sensors at the crossroad of inflammation and autophagy paves the way of research avenues to identify the missing links between chronic inflammation and metabolism alterations related to aging, exercise, nutrition or even circadian rhythm (CR) (Berenbaum and Meng, 2016). CR not only interferes with several physiological processes ranging from behavioral and physiological activities such as sleep/wake cycles, blood pressure, body temperature and metabolism, but also contributes to the severity of aging-related disorders (Longo and Panda, 2016). CR is centrally regulated by a core pacemaker located in the hypothalamus and by an integrated network of peripheral oscillators in tissues and cell types. The coordinated activity of this network controls diurnal pattern of rest and activity in mammals. The pacemaker consists in a group of genes coding for transcriptional activators/repressors that interact in a 24 hourly-tuned feedback loop. The rhythmic activity of the positive regulators brain and muscle aryl hydrocarbon molecular pattern.

receptor nuclear translocator-like protein 1 (BMAL1) and circadian locomotor output cycles kaput (CLOCK) drives the expression of E-box promoter containing genes that are pivotal in cell differentiation. Among these clock-controlled genes, the negative regulators cryptochrome (CRY) and mammalian period (PER) in turn repress the expression of BMAL1 and CLOCK, thereby completing the feedback loop that cycles every 24 h. While it remained poorly understood, pain and stiffness in OA patients intriguingly follows a daily pattern notably for hand, knee and hip OA. On the other hand, it has been recently identified a circadian clock in chondrocytes (Gossan et al., 2013) that controls the expression of genes involved in cartilage homeostasis and can be altered during aging. Together these data provide a rationale for questioning the role of CR alteration in OA. Following the pioneering description of BMAL1/CLOCK in chondrocytes, it has been subsequently demonstrated that inflammatory cytokines such as IL-1β, conversely to TNF-alpha, severely affects the expression of BMAL1/CLOCK through functional interference with NF-kB (Guo et al., 2017). More recently, it has also been demonstrated that BMAL1, which is essential to maintain the differentiated phenotype of chondrocytes through its interaction with a large number of TGF-β signaling members (Akagi et al., 2017), was repressed in OA and associated with a switch of catabolic phenotype (Snelling et al., 2016). These converging reports have culminated in the description that genetic ablation of BMAL-1 specifically in mouse chondrocytes deeply affects intrinsic CR,

contribute to induce a catabolic profile in chondrocytes and leads to cartilage damage (Dudek et al., 2016). Interestingly, it has also been described that BMAL1 predisposes to a marked degeneration of the fibrocartilaginous intervertebral disc (Dudek et al., 2017). Altogether, these data support the notion that disruption of chondrocyte CR may predispose individuals to the onset of cartilage disorders. In search of novel therapeutic interventions in OA, it seems thus reasonable to speculate that the development of drugs interfering with the peripheral clock in cartilaginous tissues may open new therapeutic spaces in the management of OA.

Consistent with this idea, the recent description of a molecular dialog between clock gene BMAL1 and the energy/nutrient sensor SIRT1 has shed further light on the clinical translatability of this therapeutic concept (Dvir-Ginzberg et al., 2016). SIRT1 is well-known to form a regulatory complex with CLOCK/BMAL1 that represses clock gene expression notably in muscle and heart. Such a crosstalk has also been described in human OA cartilage (Yang et al., 2016), where a direct SIRT1-mediated circadian regulation of BMAL1 occurs. Strikingly and of particular clinical applicability, the sirtuin-activating molecule resveratrol, also known as an interacting agent with the aryl hydrocarbon receptor, is well-known to exert beneficial effects on cartilage anabolism. To address the clinical efficacy of such a molecule, the effects of oral resveratrol are currently being assessed in a phase 3 double blind placebo-controlled and randomized clinical trial (NCT02905799-Arthrol) in OA patients.

Also in line with the concept that cartilage intrinsic CR may be a pragmatic entry point for the development of integrative therapeutics in OA, is the intriguing connection between circadian genes and autophagy. As extensively described above, autophagy is a key process in the onset of OA and is thus considered a promising therapeutic target. Of interest, BMAL1 has recently been shown to positively regulate autophagy through mTOR in cardiomyocytes (Qiao et al., 2017). While it remains to be determined whether such a relationship also operates in cartilage, it raises the possibility that restoring the cartilage CR may contribute to increase autophagy processes which in turn contribute to limit cell stress, inflammation and catabolism. Among the molecules that could interfere with central and peripheral CR, melatonin has been widely studied in the context of rheumatologic disorders (Jahanban-Esfahlan et al., 2017). Melatonin is a circadian multi-tasking hormone mainly produced in the pineal gland but also in a wide range of other tissues. Melatonin exhibits diverse biological activities such as anti-oxidation, anti-inflammation or antiapoptosis through its interaction with specific plasma membrane or nuclear receptors and intracellular targets. In the context of OA, intraarticular injection of melatonin protected against cartilage damages induced by partial meniscectomy in rabbits (Lim et al., 2012). Deciphering the underlying mechanisms has revealed that the anti-inflammatory effect of melatonin could be mediated by a SIRT1-dependant inhibition of the NF-kB pathway (Lim et al., 2012). Consistent with a role of cartilage clock in OA, low doses of oral melatonin, in association with exercise, was found to concomitantly restore the OA-associated reduced levels of clock-controlled genes and reduce the severity of cartilage damages in a murine model of collagenase-induced OA (Hong et al., 2017). This converging body of evidence suggesting that melatonin exerts a protective effect in OA remains however to be consolidated through well-conducted in vitro and in vivo experiments in animal model of OA or in clinical trial, notably with respect to the ambiguous role of SIRT1 in mediating the pro-inflammatory effects of IL-1β or H2O<sup>2</sup> in chondrocytes (Guo et al., 2017; Hong et al., 2017).

Despite some areas of darkness still subsist notably regarding the molecular links between CR, SIRT1/AMPK energy sensors, and autophagy, these data reinforce the notion that local clock genes may be considered as clinically relevant targets for the development of novel therapeutic intervention. See **Figure 1** summarizing hypothetical mechanisms underlying the role of the circadian rhythm in OA.

# CONCLUSIONS

Failures of autophagy homeostasis lead to cell dysfunction senescence and death and initiate the musculoskeletal degeneration and inflammation characteristic of rheumatic

# REFERENCES

diseases. In this sense, preclinical models of such fundamental homeostasis mechanisms might create opportunities to build biobanks with patient derived material that can be used to perform drug screens and facilitate drug development. Clinical and preclinical models include blood, tissues from human joints and experiments in rodents showing that aging and OA of articular cartilage are invariably associated with defective autophagy and inflammation. Studying these dimensions of cartilage aging, chondrocyte senescence, inflammation and autophagy could facilitate to open the pharmacological space to novel disease-modifying therapies for OA. In this review, we highlighted what is currently known about preclinical models of OA homeostasis, the major targets and pathways identified, and its potential value for translation. See **Figure 2** illustrating the interrelation between autophagy, senescence and inflammation in OA.

# REVIEW CRITERIA

Data for this review were collected by searching in PubMed database for articles published from 1965 to 2018 using the following keywords "osteoarthritis," "inflammation," "aging," "senescence," "SASP," "chondrocyte," "homeostasis," "autophagy," "therapeutics," "metabolism," "circadian clock" alone or in combination. English-language original publications and Review articles were selected on the basis of their relevance for the inclusion in the bibliography.

# AUTHOR CONTRIBUTIONS

All authors approved the final version to be published. CV, ED, JG, BC: manuscript design and drafting and revising the manuscript.

# FUNDING

This study was supported by Instituto de Salud Carlos III, Ministerio de Economía y Competitividad, Spain and Fondo Europeo de Desarrollo Regional (FEDER), Una manera de hacer Europa, PI14/01324. BC was supported by Miguel Servet Type II Program- CPII16/00045-A, Instituto de Salud Carlos III, Ministerio de Economía y Competitividad, Spain and thank the FOREUM Foundation for Research in Rheumatology for their support. JG and CV would like to thank the Fondation arthritis, the French Society of Rheumatology, the ROAD Network and the FOREUM-Foundation for Research in Rheumatology for their supports. ED would like to thank to Ministerio de Economía y Competitividad Spain and Fondo Europeo de Desarrollo Regional (FEDER) (Project Code: RTC-2015-4207-1) and the Kaertor Foundation for their support.

Abounit, K., Scarabelli, T. M., and McCauley, R. B. (2012). Autophagy in mammalian cells. World J. Biol. Chem. 3, 1–6. doi: 10.4331/wjbc.v3.i1.1

Acosta, J. C., Banito, A., Wuestefeld, T., Georgilis, A., Janich, P., Morton, J. P., et al. (2013). A complex secretory program orchestrated by the inflammasome controls paracrine senescence. Nat. Cell Biol. 15, 978–990. doi: 10.1038/nc b2784


knee osteoarthritis in rats via autophagy. J. Appl. Physiol. 124, 1177–1185. doi: 10.1152/japplphysiol.00871.2017


apoptosis in human osteoarthritic cartilage. Exp. Cell Res. 352, 146–156. doi: 10.1016/j.yexcr.2017.02.012


patients and disease activity. Arthritis 2013:461486. doi: 10.1155/2013/ 461486


cyclic mechanical tension induced calcification. Bone 66, 232–239. doi: 10.1016/j.bone.2014.06.018


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

The handling Editor declared a past co-authorship with one of the authors BC.

Copyright © 2018 Vinatier, Domínguez, Guicheux and Caramés. This is an openaccess 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.

# Obesity, Metabolic Syndrome, and Musculoskeletal Disease: Common Inflammatory Pathways Suggest a Central Role for Loss of Muscle Integrity

Kelsey H. Collins 1,2, Walter Herzog1,2, Graham Z. MacDonald<sup>1</sup> , Raylene A. Reimer 1,3 , Jaqueline L. Rios 1,2,4, Ian C. Smith<sup>1</sup> , Ronald F. Zernicke1,5,6 and David A. Hart 1,2,7,8 \*

<sup>1</sup> Human Performance Laboratory, Faculty of Kinesiology, University of Calgary, Calgary, AB, Canada, <sup>2</sup> McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB, Canada, <sup>3</sup> Department of Biochemistry and Molecular Biology, University of Calgary, Calgary, AB, Canada, <sup>4</sup> CAPES Foundation, Brasilia, Brazil, <sup>5</sup> Departments of Orthopaedic Surgery and Biomedical Engineering, School of Kinesiology, University of Michigan, Ann Arbor, MI, United States, <sup>6</sup> Department of Surgery, Department of Physiology and Pharmacology, University of Calgary, Calgary, AB, Canada, <sup>7</sup> Department of Family Practice, The Centre for Hip Health and Mobility, University of British Columbia, Vancouver, BC, Canada, <sup>8</sup> Alberta Health Services Bone and Joint Health Strategic Clinical Network, Calgary, AB, Canada

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Oreste Gualillo, Servicio Gallego de Salud, Spain Francis Berenbaum, Université Pierre et Marie Curie, France Tom Appleton, University of Western Ontario, Canada

> \*Correspondence: David A. Hart hartd@ucalgary.ca

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 30 October 2017 Accepted: 05 February 2018 Published: 23 February 2018

#### Citation:

Collins KH, Herzog W, MacDonald GZ, Reimer RA, Rios JL, Smith IC, Zernicke RF and Hart DA (2018) Obesity, Metabolic Syndrome, and Musculoskeletal Disease: Common Inflammatory Pathways Suggest a Central Role for Loss of Muscle Integrity. Front. Physiol. 9:112. doi: 10.3389/fphys.2018.00112

Inflammation can arise in response to a variety of stimuli, including infectious agents, tissue injury, autoimmune diseases, and obesity. Some of these responses are acute and resolve, while others become chronic and exert a sustained impact on the host, systemically, or locally. Obesity is now recognized as a chronic low-grade, systemic inflammatory state that predisposes to other chronic conditions including metabolic syndrome (MetS). Although obesity has received considerable attention regarding its pathophysiological link to chronic cardiovascular conditions and type 2 diabetes, the musculoskeletal (MSK) complications (i.e., muscle, bone, tendon, and joints) that result from obesity-associated metabolic disturbances are less frequently interrogated. As musculoskeletal diseases can lead to the worsening of MetS, this underscores the imminent need to understand the cause and effect relations between the two, and the convergence between inflammatory pathways that contribute to MSK damage. Muscle mass is a key predictor of longevity in older adults, and obesity-induced sarcopenia is a significant risk factor for adverse health outcomes. Muscle is highly plastic, undergoes regular remodeling, and is responsible for the majority of total body glucose utilization, which when impaired leads to insulin resistance. Furthermore, impaired muscle integrity, defined as persistent muscle loss, intramuscular lipid accumulation, or connective tissue deposition, is a hallmark of metabolic dysfunction. In fact, many common inflammatory pathways have been implicated in the pathogenesis of the interrelated tissues of the musculoskeletal system (e.g., tendinopathy, osteoporosis, and osteoarthritis). Despite these similarities, these diseases are rarely evaluated in a comprehensive manner. The aim of this review is to summarize the common pathways that lead to musculoskeletal damage and disease that result from and contribute to MetS. We propose the

overarching hypothesis that there is a central role for muscle damage with chronic

**82**

exposure to an obesity-inducing diet. The inflammatory consequence of diet and muscle dysregulation can result in dysregulated tissue repair and an imbalance toward negative adaptation, resulting in regulatory failure and other musculoskeletal tissue damage. The commonalities support the conclusion that musculoskeletal pathology with MetS should be evaluated in a comprehensive and integrated manner to understand risk for other MSK-related conditions. Implications for conservative management strategies to regulate MetS are discussed, as are future research opportunities.

Keywords: joint diseases, muscle, bone, tendon, NFkB, MAPK

# METHODOLOGY

The studies presented in this review were identified through PubMed Searches and the review of relevant papers in the area of diet-induced obesity, musculoskeletal health, musculoskeletal disease, and inflammation.

# INTRODUCTION TO METABOLIC SYNDROME AND OBESITY

Metabolic syndrome (MetS) is a cluster of conditions visceral obesity, hypertension, dyslipidemia, and elevated fasting glucose—that increase an individual's risk for diabetes and cardiovascular complications (Alberti and Zimmet, 1998; Manuel et al., 2014). Human metabolism has evolved to efficiently convert chemical energy obtained through the consumption of food into thermal and chemical energy. Our body's metabolic pathways have developed to provide energy to tissues in times of physical threat and survival, or to efficiently conserve energy in times of food deprivation. Today, westernized societies have an abundance of food (food security) and many individuals have little need to perform physical activity. This combination has led to excessive nutrient storage, placing significant stress on our metabolic pathways, and leading to an increase in the prevalence of disease stemming from metabolic dysfunction (Miranda et al., 2005).

Concordant with the rise in MetS prevalence, there is also a global increase in the prevalence of musculoskeletal (MSK) diseases and disorders (Wearing et al., 2006). Recent evidence demonstrates that metabolic complications also increase the risk for the most prominent MSK diseases, such as sarcopenic obesity (muscle loss in obesity), osteoporosis, tendinopathy, and osteoarthritis, conditions which contribute significantly to disability and time lost from work. The resultant damage and pain associated with these conditions likely develops through low-level systemic inflammation (Hoy et al., 2014; Smith et al., 2014; **Figure 1**) in addition to loading due to obesity (Felson et al., 1988), and reduced ability to withstand loading due to sarcopenia.

MSK disease is of particular concern, for example, as osteoarthritis-related walking disability significantly increases risk for all-cause mortality and cardiovascular events, when controlling for other cofounders (Hawker et al., 2014). This suggests that MSK disability associated with MetS can contribute to the worsening of MetS through sedentary behavior. Although obesity has received considerable attention regarding chronic cardiovascular conditions and diseases, as well as diabetes, the MSK complications that result from obesity associated MetS are less frequently discussed and rarely evaluated comprehensively.

Muscle mass is a key predictor of longevity in older adults (Srikanthan and Karlamangla, 2014), and since muscle is highly plastic and undergoes regular remodeling, it is a vulnerable tissue in a chronic low-level inflammatory environment, such as that seen with metabolic dysfunction (Tidball, 2005; Fink et al., 2014; D'Souza et al., 2015; Collins et al., 2016a). For example, intramuscular lipid deposits increase with obesity and are also positively correlated with insulin resistance (Akhmedov and Berdeaux, 2013; Addison et al., 2014; Fellner et al., 2014), linking structural alterations with altered capacity for glucose homeostasis. Functionally, impaired muscle integrity, persistent atrophy, and lipid accumulation in muscle are risk factors for tendinopathy (Meyer and Ward, 2016), osteoporosis (Ormsbee et al., 2014), osteoarthritis (Lee et al., 2012), and integrity of a motion segment, such as the leg (**Figure 2**).

# A ROLE FOR METABOLIC DISTURBANCE IN MOTION SEGMENT TISSUE DAMAGE

A hypothesis has emerged proposing that metabolically mediated damage to MSK tissues may be one additional component of the MetS, as adipose-based inflammation links obesity, MetS, and MSK tissue damage (Hart and Scott, 2012; Zhuo et al., 2012). Increased visceral adiposity is linked to induction of increased levels of catabolic mediators (Fontana et al., 2007) and ultimately tissue damage (**Figures 1** – **3**). Additionally, the presence of hypertension may be linked to tissue damage through vasoconstriction and ultimately depriving tissues of appropriate nutrient exchange (McMaster et al., 2015). Moreover, high cholesterol is speculated to link dysregulated lipid metabolism and endothelial dysfunction and has been directly associated with tissue damage in tendons (Tilley et al., 2015), decreased bone mineral density (Makovey et al., 2009), and osteoarthritis damage (Farnaghi et al., 2017). In parallel, investigations into the impact of low-level systemic inflammation from metabolic disturbance on the onset of sarcopenic obesity tendinopathy, osteoporosis, and osteoarthritis have been conducted. Many of these authors report associations between low-level systemic inflammation from diet-induced obesity (DIO) with MSK disease outcomes (**Table 1**), however these diseases are seldom evaluated

comprehensively to evaluate a common inflammatory pathway to disease induction and progression.

As such, we suggest a hypothesis linking altered muscle integrity to direct and indirect consequences on the motion segment (**Figures 2**, **3**). An obesogenic diet, resulting in over nutrition and development of a low-level systemic inflammation, acutely challenges associated tissues of the motion segment and can result in positive adaptation (i.e., through dynamic compensation that helps the tissue accommodate metabolic challenge), as well as negative adaptation (i.e., vulnerability to deleterious changes in tissue integrity). With these initial challenges, such tissue adaptive responses are balanced and help preserve the associated tissues and motion segment integrity. However, with chronic exposure to obesogenic diet and its inflammatory consequences, tissues demonstrate dysregulated repair, an imbalance toward negative adaptation resulting in regulatory failure, tissue damage (i.e., ectopic lipid storage and tissue fibrosis), and failure of motion segment integrity progressing to loss of function and ultimately, disease.

# AIM AND SCOPE OF REVIEW

The aim of this review is to summarize the links between induction of local and systemic inflammation, DIO, and a central role for muscle integrity in the inflammation-based pathogenesis of these MSK diseases. Early loss of muscle integrity can have both direct and indirect "ripple" effects downstream on tendons and bones, as well as the functioning of multi-tissue complex joints (e.g., the joint as an organ, **Figure 2**) (Frank et al., 2004; Loeser et al., 2012). As muscle plays a critical role in the mechanical and biological homeostasis of bones (through the muscle tendon unit indirectly and directly through the muscle/bone interface), tendons (directly through the muscle tendon unit), and joints (through loading and stabilization), we suggest that with the influence of dysregulated muscle loading, inflammation, and altered muscle integrity, failure of motion segment integrity is induced and exacerbated.

This review is focused on outcomes from studies using preclinical DIO models, as the gradual and progressive pathway toward MetS afforded by DIO provides critical insight into the short- and long-term pathophysiology, in addition to the phenotype of MSK diseases (Buettner et al., 2007; Nilsson et al., 2012). Evaluating diet-induced alterations allows for linking results across systems from food, through the gut and the associated microbiome, to early and late tissue-based changes, as well as the inclusion of potential epigenetic outcomes regarding temporal relations for the onset and progression of MSK disease.

# OBESITY AND IMPACT ON MUSCLE INTEGRITY

# Sarcopenic Obesity

Sarcopenic obesity, or low muscle mass and quality with increased fat mass, is not only associated with poor physical function (Zamboni et al., 2008), but also results in additional weight gain and an 8 to 11-fold increase in the risk for three or more additional physical disabilities (Baumgartner, 2000). Sarcopenic obesity was first defined clinically using two criteria: (a) an individual who is −2 standard deviations in muscle mass index (muscle mass/height<sup>2</sup> ) compared to healthy, samesex younger individuals; and (b) an individual who has a body fat percentage greater than the median body fat percentage for each sex (males: muscle mass index <7.26 kg/m<sup>2</sup> with body fat percentage >27%, females: muscle mass index <5.45 kg/m<sup>2</sup> body fat percentage >38%) (Baumgartner et al., 1998; Baumgartner, 2000). An alternative criteria for sarcopenic obesity involves falling below a linear regression-based threshold amount of lean mass given an amount of fat mass (Stenholm et al., 2008).

Sarcopenic obesity can predict disability and loss of activity in elderly adults (Baumgartner et al., 2004), and sarcopenic

obesity is more closely associated with MetS than either obesity or sarcopenia alone. This suggests important roles for both fat accumulation and muscle loss in the etiology of MetS (Lim et al., 2010). Although the age-dependent declines in muscle structure and strength are well-documented, the mechanism by which MetS results in sarcopenic obesity remains to be clarified (Kob et al., 2014). As sarcopenic obesity results in disability, loss of activity, altered mechanical loading, and altered biological function in the muscle due to lipid deposition and its sequelae (Ormsbee et al., 2014), it is likely that sarcopenic obesity is central to the development of other musculoskeletal pathologies. In fact, data from our laboratory in a rat model revealed changes in the integrity of specific muscles as early as 3-days on a high-fat/high-sucrose (HFS) diet (Collins et al., 2016c), and correspond to long-term changes in systemic inflammation and gut microbiota (Collins et al., 2015a). These data support the notion that muscle may be among the first MSK tissues affected by DIO, and inflammation likely plays a substantial role in this loss of integrity.

# Dysregulated Tissue Regeneration of Muscle—Primary Tissue Damaged by MetS

Muscle fiber damage happens on a daily basis and is generally considered to be a beneficial stimulus, leading to growth and adaptation through muscle regenerative processes (Karalaki et al., 2009). In muscle, monocyte and macrophage recruitment, as well as phagocytosis of necrotic material, occurs within the first 24 hours. Muscle is repaired through a series of tightlycontrolled inflammatory processes (Akhmedov and Berdeaux, 2013). Specifically, muscle regeneration is a multistep process involving degeneration, regeneration, and remodeling, ultimately restoring structure and function (Laumonier and Menetrey, 2016).

The three most active cells in the regeneration of skeletal muscle are macrophages, satellite cells, and fibroblasts

(Akhmedov and Berdeaux, 2013). The metabolic complications associated with obesity can result in an inappropriate temporal recruitment of these cells, which in turn leads to impaired angiogenesis and myocyte formation, while promoting the deposition of fibrotic and adipose tissue, ultimately leading to a reduction in structural integrity and functional capacity of a muscle (Karalaki et al., 2009).

With DIO, metabolic dysfunction and the presence of chronic low level inflammation can impair the "normal" inflammatory response and the regenerative capacities of skeletal muscles, resulting in a pseudo-injury (Collins et al., 2016c; **Figure 3**). For example, elevated levels of leptin, a satiety hormone now appreciated to have a role in low-level systemic inflammation, can impair angiogenesis, leading to tissue ischemia (Brown et al., 2015). IL-6 expression at the muscle level is a key mediator of macrophage infiltration and muscle repair (Zhang et al., 2013), while elevated expression of TGF-β promotes an increased fibrotic tissue deposition (Laumonier and Menetrey, 2016).

Efficient muscle regeneration can be attributed to satellite cells being readily available, and the cells' ability to re-establish residual pools to support multiple rounds of regeneration (Karalaki et al., 2009). Satellite cells are limited by the complex physiological environment in which they interact, an environment that can be significantly altered in individuals with obesity (D'Souza et al., 2015). A pathological host environment can limit a satellite cell's ability to be activated, proliferate, and differentiate into a muscle fiber (D'Souza et al., 2015; Meng et al., 2015). This was elegantly shown by Boldrin and colleagues through the transplantation of satellite cells from mdx mice, a genetic mouse model of muscular dystrophy, into a neutral environment (Boldrin et al., 2015). Despite the impairment of mdx satellite cells as a result of being in the pathogenic environment of an mdx mouse, following transplantation, mdx satellite cells were fully capable of being activated, and could proliferate and differentiate into a fully functional muscle fiber (Boldrin et al., 2015). Macrophages may also inhibit satellite cell activity, suggesting another mechanism by which low-level systemic inflammation may inhibit muscle repair (Tidball and Villalta, 2010).

Impairments in satellite cell activity have been reported to promote fibro/adipogenic progenitor cells (FAP) that normally aid in muscle regeneration, to differentiate into fibroblasts and/or adipocytes (Chapman et al., 2016). FAPs have also been demonstrated to be a source of intramuscular lipid deposition with rotator cuff/supraspinatus tendon injury (Liu et al., 2016). FAPs are thought to be vulnerable to reprogramming in TABLE 1 | Examples of MSK damage resulting from DIO.


the presence of low-level inflammation, which may result in increased lipid and fibrosis deposition, and may limit reversibility of fibrosis (Mann et al., 2011). Thus, chronic obesity and associated MetS, with inflammation and fatty infiltration of muscles, may lead to a compromise in the regeneration of muscle integrity.

In addition to the environmental challenges to the muscle regeneration posed by metabolic disturbance, skeletal muscle from individuals with obesity displays a greater number of glycolytic-fibers vs. oxidative-fibers when compared to a healthy individual (Pattanakuhar et al., 2017). Since oxidativefibers generally contain a greater number of satellite cells relative to glycolytic fibers (Karalaki et al., 2009), individuals with obesity may also have fewer satellite cells. Based on this information, impairments in satellite cell function as a result of alterations in cell metabolism, a reduction in cell number, suppression of cell activation, depleted cell reserves, and impaired cell proliferation and differentiation may lead to impairments in muscle fiber regeneration (Akhmedov and Berdeaux, 2013).

Work from our laboratory has demonstrated that the oxidative soleus muscle is protected against HFS-induced damage over short-term and long-term exposures in a rat model (Collins et al., 2017b; **Figure 4**). By 3-days on HFS, dynamic increases in mRNA levels for superoxide dismutase (SOD2) in HFS animals implicate compensatory oxidative stress scavenging in the soleus muscle compared to control animals. By 2 weeks on HFS, increased mRNA levels for oxidative capacity [succinate dehydrogenase (SDH)] were detected compared to chow-fed controls, suggesting one adaptation strategy that the soleus muscle may employ with HFS metabolic challenge. Although the precise mechanism(s) by which the soleus is protected from metabolic disturbance-induced muscle damage remains to be clarified, it appears that increasing the oxidative capacity and the oxidative stress scavenging ability of muscles (i.e., with aerobic exercise) may be a beneficial strategy for mitigating obesity-induced muscle damage and its consequences.

Inflammation related to obesity can also impair myocyte remodeling as a result of a reduction in protein synthesis due to elevated TNF-α levels (Brown et al., 2015). Furthermore, there is evidence for adipocyte-muscle cross talk in vitro, whereby adipocyte-derived inflammation can contribute to inflammation and atrophy in muscle cells subjected to a metabolically dysfunctional environment, possibly through IGF-1 (Pellegrinelli et al., 2015). Impaired protein synthesis can also prevent muscle from properly adapting to mechanical stimuli. Brown et al. (2015) demonstrated this inability, showing that following muscle damage, obese muscle displayed no adaptations, while lean mice displayed an increase in muscle wet weight and muscle fiber hypertrophy (Brown et al., 2015). Potential contributing factors to a reduction in protein synthesis may be elevated lipid metabolites resulting from impaired mitochondrial function contributing to elevated TNF-α levels, which are known to have inhibitory effects on IGF-1 (Akhmedov and Berdeaux, 2013). Decreases in IGF-1 result in the inhibition of the muscle growth signaling pathway (IGF-1 → P13K → Akt → mTOR), effectively blunting muscle protein synthesis (Brown et al., 2015).

Furthermore, increased myostatin levels can also contribute to impaired growth in obese muscle. Myostatin is not only significantly up-regulated in obese skeletal muscle, but in adipose tissue as well, further inhibiting myogenesis, providing another avenue through which potential muscle-adipose cross talk may occur (Karalaki et al., 2009).

# Functional Muscle Damage with Metabolic Derangement

Generally speaking, adults with obesity are reported to have significantly higher absolute strength in lower limb muscles, but lower strength when normalized to body mass (Tomlinson et al., 2016). When the upper limb muscles are evaluated, there are no statistical significant differences between individuals with obesity and normal-weight controls (Tomlinson et al., 2016). Potentially, the characterization of obesity by body mass, which is common in these studies, may be inappropriately representing body composition. We have shown, in a healthy and overweight population cohort, that body mass index (BMI) inappropriately estimates body composition in 30% of the population, with a specific disparity between BMI and body composition measurements in healthy females (Collins et al., 2017c). Also, there is a lack of data describing the effect of obesity on muscle integrity, and a lack of consistent protocols to assess muscle strength (Tomlinson et al., 2016). However, data from our lab (Collins et al., 2016a,c, 2017b) and others, in rodents (Ciapaite et al., 2015) and large mammals (Clark et al., 2011), have demonstrated deleterious alterations in muscle structural integrity with metabolic disturbance. Computational approaches modeling the gastrocnemius muscle have demonstrated that whole-muscle force is dependent on muscle integrity, specifically regarding reductions of muscle force due to intramuscular lipid (Rahemi et al., 2015). A zebrafish model of diet-induced obesity further demonstrated that obesity induces decreases in locomotor performance, isolated muscle isometric stress, workloop power output, and muscle relaxation rates (Seebacher et al., 2017). Of note, these decrements in performance and function were not reversed with weight loss, generating interesting questions about the potential reversibility of impaired muscle function with obesity.

Additional sources of muscle damage and altered repair are advanced glycation end products (AGEs), which accumulate over time due to increased availability of glucose and hyperglycemia (**Figure 5**). Dietary AGEs can interfere with muscle healing and impair contractile function in a mouse model of obesity (Egawa et al., 2017). HFS diets can induce hyperglycemia in rodents, further linking diet-induced obesity to AGEs (Sumiyoshi et al., 2006). The receptor for AGEs on macrophages, called RAGE, is associated with a pro-inflammatory state, and RAGE/AGEs are reported to be involved in the onset and progression of metabolic disturbance, insulin resistance, and adipokine expression (Leuner et al., 2012; Hofmann et al., 2014). AGEs have been implicated in macrophage polarization toward M1 pro-inflammatory phenotypes, pro-inflammatory IL-6 secretion in adipose tissues, and initiating inflammatory cascades (Bopp et al., 2008; Frommhold et al., 2011; Nativel et al., 2013; Jin et al., 2015; Son et al., 2016) [i.e., NF-κB, p38 mitogen-activated protein kinase (MAPK)]. Reactive AGEs can also cross-link with collagen fibers, which subsequently can affect the fiber's mechanical and biological properties (Abate et al., 2013). Although AGE collagen cross-linking can be reversed (Asif et al., 2000), it is unclear whether glycation itself can be reversed. Of note, endurance exercise has been shown to attenuate AGEs in cardiac muscle of rats (Wright et al., 2014).With weight loss, reversal of metabolic dysfunction may not be fully achievable, and weight re-gain is common (Fothergill et al., 2016). As mentioned above, to what degree reversibility of skeletal muscle damage may be achieved in this context remains to be determined (**Figures 2** – **4**), and several factors could contribute to irreversibility (Mann et al., 2011). Likely, clarifying relations between the time of exposure to low-level systemic inflammation and muscle damage should be determined, as impaired muscle integrity has been observed following short-term exposure to a metabolic challenge, likely before "full" metabolic derangement has been achieved (Fink et al., 2014; Collins et al., 2016c). Also, some alterations in satellite cell function can be irreversible (Sacco and Puri, 2015). Generally, DNA methylation, or epigenetic changes, is an actively researched area being explored in the context of muscle (Carrió and Suelves, 2015), and DNA methylation is an important step in muscle cell differentiation (Brunk et al., 1996). However, it is likely that epigenetic changes, which are dynamic and induced by environmental changes, can induce cellular reprogramming, and thus could potentially limit the reversibility of muscle damage, even if the MetS-related inflammation is controlled (Carrió and Suelves, 2015). Investigations to clarify the role of methylation in physiological and pathophysiological muscle changes may provide valuable insight into the reversibility potential of muscle damage.

# Dysregulated Homeostasis of Associated Motion Segment Tissues (Tendon, Bone, Cartilage, and Joint) with Obesity: Acute and Chronic

Muscle, tendon, cartilage, and bone tissues repair and regenerate over different timelines, and these different repair timelines may dictate vulnerability or resistance to damage with metabolic disturbance involving inflammatory processes. Both biological and mechanical stimuli are critical to homeostatic tissue regulatory mechanisms. Across tissues, remodeling rates slow

with age. As discussed above, muscle tissue is likely the most vulnerable MSK tissue to perturbation with metabolic challenge. Repair of such tissues in the face of an active inflammatory environment is likely compromised, and resolution of repair is particularly challenging (Hart et al., 2004).

Altered muscle integrity can challenge the motion segment in a dynamic manner, until the associated tissues ultimately begin to fail (**Figures 2**, **3**). Specifically, tendon healing is slower compared to muscle. In tendon, tenocytes migrate to the wound, initiating type III collagen synthesis. After several weeks to months, remodeling of tendinous tissue occurs such that collagen fibers are aligned in the direction of stress, and maturation occurs over the course of a year. Bone mechanotransduction, initiating the maintenance/remodeling processes (Scott et al., 2008), requires a series of events that can require almost 24-h to ultimately result in the synthesis of bone matrix proteins (Robling and Turner, 2009). Cartilage is considered to have limited intrinsic repair properties, although with the facilitation of autologous stem cells, biologics, and other transplants, some repair may be achievable (Chu et al., 2010). The impaired regenerative capacity of tendon, bone and cartilage likely indicate a lack of reversibility with altered integrity and subsequent compromise in function, underscoring a need for strategies aimed at primary prevention (before it occurs), secondary prevention (limiting the amount of damage as it occurs), tertiary prevention (prevention of progression of damage to disease), or reversal of damage.

# DIET-INDUCED OBESITY AND ELEMENTS OF THE DYSREGULATED INFLAMMATORY RESPONSE

# Inflammatory Mediator Alterations in Obesity

#### Cytokines and Chemokines

Activation of the innate immune system is critical in the pathogenesis of type 2 diabetes and tissue damage. Some of the key signaling pathways involved in this process are nuclear factor –κB (NF-κB), c-Jun N-terminal Kinase (JNK), and the NLRP3 inflammasome, all resulting in transcription of proinflammatory cytokines (Lackey and Olefsky, 2015). NF-κB can be activated by elevated levels of pro-inflammatory cytokines such as TNF-α, which is increased in the adipose tissue of obese and diabetic animals. Neutralizing TNF-α has been shown to reduce insulin resistance (Hotamisligil et al., 1993). Additionally, JNK activity increases in tissues that are sensitive to insulin, is activated by ER stress, and directly inhibits insulin signaling (Gual et al., 2005; Lackey and Olefsky, 2015). The inflammasome is a protein complex that matures and secretes inflammatory cytokines, such as IL-1β and IL-18. Similar to NFκB, the inflammasome can be activated by pro-inflammatory cytokines, lipopolysaccharide (LPS), and some forms of low density lipoproteins (LDLs). LPS, low density lipoproteins, and the AGE-product of LDLs can signal through Toll-like Receptor-4 (TLR-4) (Hodgkinson et al., 2008), activating IL-1β and IL-18 signaling.

TLR-4 recognition of saturated fatty acids is necessary to enable NF-κB signaling and induce expression of proinflammatory cytokines (TNF-α, IL-6, and MCP-1) (Jialal et al., 2014; Lackey and Olefsky, 2015). TLRs sense pathogen-associated molecular patterns and damage-associated molecular patterns (PAMPs), and regulate the inflammatory responses to mitigate tissue repair (Lee et al., 2013). TLR-4 gene expression and protein content are increased in muscle from patients with obesity and diabetes, potentially contributing to insulin resistance, as well as compromised muscle integrity observed with metabolic disturbance (Reyna et al., 2008). In the context of OA, TLRs can modulate the catabolic pathways and maintain joint homeostasis (Houard et al., 2013). Of interest, the lubricating molecule proteoglycan-4 (PRG-4) has been shown to modulate the inflammatory response through competitive inhibition of TLR-4 receptors, suggesting that as PRG-4 concentration is reduced with OA severity, inflammatory modulation may also be reduced, contributing to the progression of OA (Iqbal et al., 2016). In tendinopathy, there may not be a clear role for TLRs, as catabolic processes in Achilles tendinopathy seem to occur independently of TLR4-induced gene expression from IL-1β or TNF-α (de Mos et al., 2009). In bone, LPS-induced activation of TLR-4 in neutrophils is reported to upregulate the catabolic RANKL osteoclast cascade, linking TLR based inflammation to increased bone resorption (Chakravarti et al., 2009).

Appropriate levels of pro-inflammatory cytokines are necessary for tissue homeostasis. However, exogenous exposure to pro-inflammatory cytokines, or endogenous high levels of pro-inflammatory cytokines, are associated with damage across all musculoskeletal tissues. In muscle, TNF-α, IL-1β, and IL-6 activate transcription of MuRF-1 and MAFBx/atrogin-1, the key muscle atrophy pathway, through IGF/Akt-1 (Akhmedov and Berdeaux, 2013). For example, IL-6 and TNF-α inhibit bone-forming osteoblast cells, and NF-κb, RANKL, TNF-α, IL-6, M-CSF, and MCP-1 can contribute to osteoclast recruitment, maturation, and inhibit osteoclast apoptosis (Roy et al., 2016). In tendon, pro-inflammatory cytokines (IL-1β, TNF-α, IL-6) can disrupt tissue homeostasis, by inducing extracellular matrix degradation, induce other pro-inflammatory cytokines, which result in necrotic and apoptotic cell changes, and affect collagen and elastin expression (Schulze-Tanzil et al., 2011). Cytokine effects in tendon are modulated by mechanical loading in this complex mechano-biological environment (Killian et al., 2012).

The knee joint is a complex organ system with many resident inflammatory cells, particularly in the synovium and infrapatellar fat pad. These tissues secrete pro-inflammatory cytokines, like IL1-α and IL-1β (Sanchez-Adams et al., 2014). In humans, systemic and synovial fluid inflammatory profiles can differentiate between patients based on OA severity, suggesting that cytokine levels may help define OA pathogenesis (Heard et al., 2013). In addition to the infrapatellar fat pad, there is also a synovial adipose depot, and these two depots differ significantly from each other in lean individuals with OA (Harasymowicz et al., 2017). Moreover, infrapatellar fat pads and synovium adipose depots differed in adipocyte size, fibrosis, and macrophage infiltration in OA patients with obesity compared to lean OA patients (Harasymowicz et al.,

2017). High glucose concentrations can contribute to increased chondrocyte responsiveness to cytokines, increased levels of reactive oxygen species, leading to an overproduction of IL-6 and PGE2 (Laiguillon et al., 2015). These cytokines can also contribute to mitochondrial dysfunction within the joint, and, in turn, mitochondrial dysfunction can amplify chondrocyte responsiveness to cytokines (Vaamonde-García et al., 2012).

for intramuscular fat; Picro, Picrosirius stain for collagen).

Unpublished data from the author's laboratories suggest that intramuscular lipid deposits and fibrotic material in the vastus lateralis muscle of the knee are associated with the presence of metabolic-induced knee joint damage after a 12-week obesity induction period (**Figure 6**). Based on the inflammatory mediator profile of joints undergoing damage (Collins et al., 2015a,b, 2016b), inflammation is likely playing a significant role in this process and in the relations between muscle damage and joint damage. Ongoing efforts will probe the role of associated inflammation on muscle integrity in the onset and progression of such metabolic joint damage and whether it leads to overt OA.

Given the systemic-to-local hypothesis in metabolic OA (Zhuo et al., 2012; Collins et al., 2016b; Berenbaum et al., 2017), we have evaluated hip, knee, and shoulder joints using the HFS rat model system for 12-weeks (Collins et al., 2017a). A HFS diet, in the absence of trauma, resulted in significant increases in joint damage/OA-like changes in the shoulder and knee joints of rats after a standard 12-week obesity induction period. The hip joint, however, was not significantly affected by DIO, which is consistent with findings from human epidemiological studies. Total joint damage, assessed by adding the individual Modified Mankin Scores across all three joints, was increased in DIO animals compared to chow-fed animals, and was associated with the percentage of body fat. Positive significant predictive relations for total joint score were found between body fat and two serum mediators (IL-1α and VEGF). These data suggest that systemic inflammatory alterations from DIO in this model system may result in a higher incidence of knee, shoulder, and multi-joint OA-like/joint damage with the HFS diet over the long term (Collins et al., 2017a). Due to the preliminary nature of these studies, longitudinal experiments with multiple time points are needed to validate these proposed relations. If these relations are supported in future studies, then not all joints are affected equally by obesity-associated inflammation in MetS. Studying these relations are an ongoing direction of current research.

### Adipokines

Adipocytes release adipokines such as leptin, adiponectin, visfatin, and resistin as a signaling mechanism in addition to passively storing energy, which can cause and exacerbate chronic low-level systemic inflammation (Gomez et al., 2009). Adipokines have been shown to induce pro-inflammatory mediators in activated CD4+ T cells from osteoarthritis patients, demonstrating that systemic mediators may play a role in osteoarthritis (Scotece et al., 2017). After interaction with such activated CD4+ T cells, chondrocytes demonstrate increased expression of MMP-13 and decreased expression of collagen-2 and aggrecan.

Leptin is a satiety hormone that increases in a nearlinear fashion to body fat (Friedman and Halaas, 1998). It inhibits appetite and regulates body weight, energy expenditure, and maintains glucose homeostasis. However, with metabolic disturbance, individuals become leptin resistant. Leptin is a class-1 cytokine secreted from adipose tissue, and high concentrations of leptin are associated with musculoskeletal tissue damage (Zabeau et al., 2003; Collins et al., 2015a,b, 2016b,c). Leptin activates downstream pro-inflammatory pathways (IL-2, IFNγ) and inhibits the anti-inflammatory pathway (IL-4) (Lechler et al., 1998). Aside from being produced by adipose tissue, leptin levels can be increased by TNF-α, IL-1, and LPS, creating a positivefeedback loop with low-level chronic inflammation (Grunfeld et al., 1996).

Leptin signaling is critical to conventional muscle maintenance (Akhmedov and Berdeaux, 2013), and basal levels of satellite cells are reduced in animals with impaired leptin



Data are shown as fold change ± standard error. DIO n = 12–14; chow n = 5–7.\* Indicates p < 0.05 vs. control; \*\*Indicates p < 0.01 vs. control, # Indicates different in SF between DIO and chow.

signaling (Peterson et al., 2008). However, hyperphysiological levels of leptin also stimulate the proliferation and activation of macrophages, which may be a mechanism by which leptin concentration influences tissue damage (Santos-Alvarez et al., 1999). Leptin may also positively upregulate myostatin (Rodríguez et al., 2015), a member of the TGF-β superfamily that negatively regulates muscle mass and growth, and a molecule which is up-regulated in muscle from individuals with metabolic disturbance (Hittel et al., 2009). Upregulation of follistatin has been an effective treatment for muscle degenerative diseases by increasing muscle growth due to its ability to inhibit myostatin, a negative regulator of muscle mass. Follistatin also alleviates synovitis and mitigates OA-like changes from inflammatory arthritis in mice (Yamada et al., 2014). As muscle has a role in OA pathogenesis, protecting both muscles and joints with follistatin represents an attractive therapeutic opportunity for MetS-induced musculoskeletal damage.

Leptin also may be involved in mediating the pathogenesis of osteoarthritis in humans (Fowler-Brown et al., 2014) and other animals (Griffin et al., 2009, 2010, 2012; Collins et al., 2015a,b, 2016b). Leptin is found in synovial fluid of humans (Lübbeke et al., 2013) and other animals (Collins et al., 2015a,b, 2016b). Unpublished mRNA data (**Table 2**); from our laboratory indicate increased mRNA expression levels for leptin in the fat pad and synovium of rats with DIO compared to those on a chow diet, after a 12-week obesity induction period.

Additionally, leptin may be involved in mediating the pathogenesis of osteoarthritis in humans (Fowler-Brown et al., 2014) and other animals (Griffin et al., 2009, 2010, 2012; Collins et al., 2015a,b, 2016b). Leptin is found in synovial fluid of humans (Lübbeke et al., 2013) and other animals (Collins et al., 2015a,b, 2016b). Increased mRNA levels for leptin were accompanied by increased leptin in the serum and synovial fluid (Collins et al., 2016b). After 28-weeks of DIO, the fat pad and synovium demonstrated disparate up-regulation of IL-1β, although IL-1β was detected in the synovial fluid of these animals (Collins et al., 2015a). These findings suggest that the fat pad and synovium contribute to increased pro-inflammatory synovial fluid profiles (Collins et al., 2016b). However, cartilage explants are not substantially damaged when exposed to physiological levels of leptin (Griffin et al., 2010), despite reports of associations between leptin and MMPs (Koskinen et al., 2011), calling into question the direct involvement of leptin in eliciting cartilage damage.

In the context of bone, leptin influences the formtion and resorption of mineralized tissue by increasing the activity of osteoclasts through RANKL (Ducy et al., 2000). However, there are also reports that leptin supports bone growth and bone MSC differentiation into osteoblasts (Thomas et al., 1999), so the precise mechanism for leptin's effects on bone is unclear, but likely is dependent on exposure and dose (Kawai et al., 2009). Low-level systemic inflammation and leptin can negatively influence tendon structural integrity (Abate, 2014; Abate et al., 2016), can result in accelerated heterotopic ossification in tendon tissues (Jiang et al., 2017) and may be associated with increases in tendon ruptures (Ji et al., 2010). Taken together, increases in systemic leptin—or mediators downstream of leptin—appear to have deleterious effects on all major musculoskeletal tissues. These data support the notion that catabolic activity as a result of increased leptin is an important pathway to clarify in the context of global musculoskeletal integrity (Griffin et al., 2009). More details regarding the interface among leptin signaling, inflammation, metabolism, and musculoskeletal disorders are detailed elsewhere (Abella et al., 2017b).

Resistin is an adipokine involved in insulin resistance, inflammation, and energy homeostasis. Serum resistin levels are reported to be derived from visceral adipose tissue (Milan et al., 2002). There is conflicting evidence as to whether resistin is associated with bone mineral density (Mohiti-Ardekani et al., 2014). Within the joint, synovial fluid levels of resistin are associated with inflammatory and catabolic molecules in the joints of human osteoarthritis patients (Koskinen et al., 2014). However, resistin and visfatin, demonstrated positive predictive relations with recovery from upper extremity soft tissue disorders such as tendinopathy, and are thought to be related to antiinflammatory response mechanisms (Rechardt et al., 2014).

Adiponectin, another adipokine, is derived from visceral fat, and is thought to serve a protective role on cardiovascular health and glucose homeostasis (Milan et al., 2002). As body fat decreases, adiponectin levels generally increase, and adiponectin may modulate adipose tissue regulation via NF-κB (Ajuwon and Spurlock, 2005). Specifically, adiponectin treatment in obese mice increases bacterial clearance and hematopoietic progenitor proliferation in bone marrow (Masamoto et al., 2016). Adiponectin was shown to be significantly correlated with bone mineral density in a group of osteoporotic and healthy patients (Mohiti-Ardekani et al., 2014). Adiponectin has also been proposed as a systemic biomarker of OA. Plasma adiponectin was significantly higher in a population of OA patients and was also higher in women with erosive hand OA compared to patients with non-erosive OA (Filková et al., 2009). In a separate cohort of hand OA patients, the individuals with the highest levels of adiponectin demonstrated a decreased risk for hand OA progression (Yusuf et al., 2011). However, the study populations were not the same in these two reports (i.e., total numbers, males and females vs. females alone, and use of European populations potentially differing genetically) and therefore, additional research needs to be performed to better evaluate this relationship. Additionally, in a study evaluating 76 males in Thailand with knee OA revealed that levels of adiponectin in synovial fluid correlated with disease severity (Honsawek and Chayanupatkul, 2010). In muscle, adiponectin has been shown to increase fatty acid oxidation and glucose uptake, and to attenuate local inflammation (Nigro et al., 2015). Adiponectin has been proposed as a treatment for diabetic tendinopathy (Rothan et al., 2013), suggesting it may be protective and have favorable anti-inflammatory effects across MSK tissues.

Progranulin is a more recently identified adipokine that may also have anti-inflammatory characteristics. mRNA levels for progranulin are increased in cartilage, synovium, and infrapatellar fat pads from OA patients, and mRNA levels are increased in response to pro-inflammatory stimulation (Abella et al., 2016). Specifically, progranulin has been shown to counteract pro-inflammatory molecule expression (i.e., NOS, MMP-13) induced by IL-1β and the LPS-TLR-4 axis (Abella et al., 2016). Attstrin, a progranulin-derived peptide, is a promising therapeutic candidate for osteoarthritis (Abella et al., 2017a) given that progranulin can counteract IL-1 driven inflammation through TNFR1 in human chondrocytes, and intraarticular injection of progranulin-derived attstrin prevented OA-progression in a surgical model of murine OA (Xia et al., 2015).

# Obesity and Involvement of Cells of the Inflammatory System Macrophages

Generally speaking, macrophages are dichotomized into M1 (pro-inflammatory) and M2 (anti-inflammatory) phenotypes based on their activity during tissue repair processes (Novak and Koh, 2013). However, it is understood that this dichotomy represents an oversimplification based on in vitro data which may not accurately represent in vivo states (Martinez and Gordon, 2014). Macrophages are described to demonstrate a high degree of functional plasticity and their phenotypes can change based on environmental stimuli (Stout and Suttles, 2004). It is likely useful to consider macrophage activation as a spectrum rather than a binary categorization (Mosser and Edwards, 2008), but for the purposes of this review, and concordant with the current musculoskeletal literature, macrophage activity and these relations are generalized using the M1/M2 paradigm.

Macrophages are derived from monocyte precursor cells. Tissues have resident macrophages, which are responsible for general tissue maintenance. These macrophages are described as alternatively activated, or M2 macrophages, and are induced by TGF-β, IL-4, and IL-13 (Gordon and Martinez, 2010). M1 macrophages (classically activated macrophages) are key phagocytes within tissues, and are induced through IFN-γ activation and LPS-induced TLR signaling or from detection of pathogen-associated molecular patterns (Lampiasi et al., 2016). With obesity, stress, loading, or tissue and niche specific control mechanisms, M2-type macrophages may experience a phenotypic shift. This shift may be influenced by exposure to certain cytokines through their general signaling mechanisms or in the presence of other conditions, including IL-6 (Braune et al., 2017) or TNF-α (Wu et al., 2015). AGEs can also play a role in M2 to M1 polarization (Jin et al., 2015) (**Figure 5**). Typically, an imbalance in the ratio of M1:M2 macrophages is considered maladaptive, creating an imbalance toward tissue degradation and an absence of adequate repair. MAPK is a key pathway in macrophage-mediated inflammatory responses and may play a significant role in diseases mediated by macrophages (Yang et al., 2014).

Within 24 hours following muscle damage, thousands of macrophages have infiltrated the damaged tissue (Tidball, 2005; Grounds, 2011). Pro-inflammatory M1-macrophages first perform cell lysis, removing necrotic muscle tissue debris (Laumonier and Menetrey, 2016). Anti-inflammatory M2 macrophages infiltrate the damaged site once M1-macrophages have removed necrotic debris (∼48 hours following injury), helping resolve the inflammatory response while promoting myogenesis (Akhmedov and Berdeaux, 2013). Through the secretion of a number of growth factors, macrophages promote angiogenesis (FGF, TGF-β), synthesis of ECM proteins (TGFβ), and activation of satellite cells, all promoting myogenesis (Grounds, 2011).

It is possible that an increased presence of M1 macrophage cells in a tissue is an early sign of disrupted tissue homeostasis and repair. With diet-induced obesity, M1-type macrophages are present in the quadriceps muscle of mice (Fink et al., 2014), rats (Collins et al., 2016c), and humans (Fink et al., 2014; Khan et al., 2015). In bone, osteoclasts are considered M1-type macrophages. With a high-fat diet, bone loss is accelerated in young mice due to increased osteoclastogenesis (Shu et al., 2015), and as noted, pro-inflammatory cytokines increase osteoclast activity in humans with obesity through activation of the NFκB/RANKL pathway (Cao, 2011). In the context of the joint as an organ (Frank et al., 2004), the infrapatellar fat pad does not demonstrate inflammation or M1 polarization prior to knee OA in mice (Barboza et al., 2017). Synovial membrane samples from OA patients reveal markers for M1 macrophages (IL-1α, IL-1β, and TNF-α) and decreased levels of the M2 marker IL-1RA (Smith et al., 1997). As such, it is challenging to determine if levels of M1/M2 macrophages are a cause or consequence of disease. In rabbits, intra-articular injections of the proinflammatory mediators IL-1 and TNF-α stimulate M1 activation and degrade cartilage, suggesting macrophage polarization may occur in synovial cells (Pettipher et al., 1986; Henderson and Pettipher, 1989). Conditional macrophage depletion in obese mice with a knee injury does not mitigate OA severity (Wu et al., 2017b). Moreover, in mice where macrophages were depleted, there was an increase in neutrophils and CD3+ T cells compared to control animals (Wu et al., 2017b). These data suggest that a potential redundancy may exist between macrophages and other inflammatory cells. Macrophages may modulate inflammatory homeostasis within the joint OA and obesity, and that a basal level of macrophages are likely needed to maintain joint health (Wu et al., 2017b). While this study is not conclusive, it demonstrates that the relations among macrophage polarization and presence in obesity and OA are complex. It is important to note that what constitutes a healthy or maladaptive balance in macrophage phenotype in one tissue or disease may differ from another tissue or disease process. As such, these findings are likely context- and tissue-dependent. More studies are needed to better understand whether synovial macrophage polarization, mitigation/modulation, or ablation is a promising therapeutic target in OA pathogenesis (Sun et al., 2016).

# SPECIFIC CELL INVOLVEMENT IN METABOLIC OVERLOAD WITH OBESITY

Many cells (e.g., adipocytes, hepatocytes, and myocytes) contribute to the conversion of food into chemical energy and can be placed under significant metabolic stress when exposed to a typical western-style diet. Significant increases in the postprandial flux of metabolic substrates results in drastic nutrient spikes characterized by elevated glucose, free fatty acid (FFA), and triglyceride (TG) levels (Weiss et al., 2013). Chronic nutrient overload places metabolic pathways under significant stress, overwhelming subcutaneous adipose depots, leading to adipocyte mitochondrial dysfunction and impaired insulin sensitivity (Miranda et al., 2005). Normally, insulin inhibits lipolysis and promotes glucose transport. However, during chronic metabolic overload, mitochondrial dysfunction can disrupt adipocyte insulin signaling pathways, leading to impaired glucose transport into adipocytes and the inability to suppress cell lipolysis, further increasing blood glucose and free fatty acid (FFA) levels in circulation (Miranda et al., 2005). As a result of increased circulating blood glucose and FFA levels, hepatocytes and myocytes are placed under great metabolic stress.

# Mitochondrial Dysfunction and Adipose Adaptations with Metabolic Challenge

With the onset of metabolic challenge, the imbalance of redox states and mitochondrial dysfunction may be instrumental to the development of insulin resistance and MetS (Long et al., 2015). With nutrient overload, mitochondrial dysfunction in adipose tissue will give rise to an increase in inflammatory mediators (adipokines) resulting in tissue remodeling, and potentially cell death (Trayhurn, 2005; Vernochet et al., 2014). Adipokine dysregulation can result in the fibrosis, or inadequate healing response, of adipose tissue (Kwon et al., 2012), as well as liver (Chiang et al., 2011). As adipose depots become inflamed and fibrotic, they become limited in their energy storing and endocrine function, resulting in the deposition of ectopic lipid in other tissues coupled with the presences of insulin resistance (i.e., muscle and liver) (Sun et al., 2013). Altered mitochondrial function, ectopic lipid formation, and incomplete or inadequate healing responses are hallmarks of MSK diseases, and are likely initiated by low-level systemic inflammation.

# Mechanisms of Tissue Metabolic Dysfunction with Obesity

Once subcutaneous adipose depots become overwhelmed, lipid spill-over leads to ectopic lipid accumulation in visceral adipose tissue depots and other insulin sensitive tissues, such as muscle (Miranda et al., 2005). In muscle, the nutrient overload placed on myocytes, similar to adipocytes, can lead to metabolically dysfunctional cells, further contributing to the already dysregulated inflammatory environment. The complexity and interconnectedness between these processes and mitochondrial dysfunction are illustrated in **Figure 7**.

# DIETARY CONSIDERATIONS, NUTRIENT CONTENT, AND OBESITY PHENOTYPES

While discussing inflammation from DIO, it is important to consider the composition of the experimental diets, as the dietary profile and energy density of a given diet likely will impact the resultant inflammatory profile in each preclinical model (reviewed in Hariri and Thibault, 2010). For example, diets rich in fat elicit different types of glucose intolerance compared to diets rich in sugar, whereby dietary fat largely compromises muscular glucose tolerance, and sugar challenges liver lipogenic and gluconeogenic enzyme activities (Sumiyoshi et al., 2006). As such, different dietary composition may yield different findings or differences in the severity of disease outcomes.

Fat and sugar dietary constituents can contribute to different animal phenotypes, whereby animals on a high sugar diet do not necessarily gain weight, but demonstrate a conversion of lean body mass to fat body mass (Sumiyoshi et al., 2006). Diets high in saturated fat, however, generally result in increased body fat and body mass, although the type of fat employed (i.e., omega-3 fats) can also have anti-inflammatory effects, and can modulate body fat and body mass (Schemmel et al., 1969; Sumiyoshi et al., 2006; Hariri and Thibault, 2010; Wu et al., 2014, 2017a). To add to the complexity of this matter, many animal species, like humans, demonstrate obesity prone and obesity resistant phenotypes (Schemmel et al., 1969). Arguably, skeletal muscle oxidative metabolism (Shahrokhi et al., 1993) and genetic differences in fat oxidation (Ji and Friedman, 2007; Marrades et al., 2007) may affect whether an animal is prone or resistant to obesity. This feature allows for the experimental evaluation of animals that are exposed to the same amounts of an obesogenic diet, but develop disparate increases in body mass. With MSK disease, this is an important consideration, as tissue damage can be evaluated between animals with similar mass, but increased body fat and lowlevel inflammation (Hariri and Thibault, 2010). Epidemiological studies reveal MetS can also occur in individuals with typical body weight or that are underweight (e.g., HIV-infected patients), with increases in body fat and metabolic derangement, which underscores the relevance of evaluating obesity resistant animals as a surrogate for this phenomenon to understand MSK diseases in the context of MetS (Drelichowska et al., 2015).

# Nutrient Sensing, Mitochondrial Dysfunction, and Metabolic Dysfunction

Mitochondria are the primary cellular organelles that generate adenosine triphosphate (ATP). Mitochondrial dysfunction is attributed to excessive nutrient processing, resulting in the uncoupling of oxidative phosphorylation, effectively increasing reactive oxygen species (ROS) generation and decreasing ATP production (Weiss et al., 2013). The free radical theory states that an imbalance between the generation of ROS and antioxidants produced by peroxisomes can result in ROS stealing electrons from other cellular sources, resulting in cellular damage (Ashok and Ali, 1999). Under nutrient overload, as seen in obesity, peroxisomal numbers and/or activity can be impaired as a result of micronutrient deficiencies and/or excessive inflammatory cytokine (e.g., TNF-α) accumulation resulting from elevated oxidative stress and the accumulation of metabolic by-products.

When exposed to nutrient overload, nutrient sensors [i.e., AMP-activated protein kinase (AMPK), sirtuins (SIRT), and mechanistic target of rapamycin, (mTOR)] become impaired, impair mitochondrial function, and can contribute to systemic inflammation, adiposity, and lipotoxicity. AMPK activity also plays a role in bone homeostasis, as an essential mediator of fat and glucose metabolism on bone remodeling (Jeyabalan et al., 2012; reviewed in Finkel, 2015). In skeletal muscle, leptin has been shown to activate AMPK (Salminen and Kaarniranta, 2012), inhibiting FA synthesis, increasing FA oxidation and glucose uptake (Minokoshi et al., 2002). Moreover, limiting nutrient supply and the associated increase in SIRT-1 activity are hypothesized to enhance muscle cell proliferation, while nutrient overload, or age-related loss of SIRT-1, is expected to provide an unfavorable environment for cell proliferation (Akhmedov and Berdeaux, 2013). Likely, reduced cell proliferation occurs, in part, through age-related loss of SIRT, potentially as a consequence of reduced AMPK activity, which may result in loss or dysfunction of mitochondrial activity (Wang et al., 2015; Berenbaum et al., 2017).

In chondrocytes from human OA patients, mitochondrial activity is deficient, leading to a catabolic cascade of responses. Activating AMPK/SIRT-1, mitochondrial master regulator PGC-1a, and FOXO3A mediates chondroprotection, suggesting these pathways may also be critical to maintaining mitochondrial function in chondrocytes, which is critical to preserving cartilage tissue structure (Zhao et al., 2014; Wang et al., 2015; Berenbaum et al., 2017). Mitigating age-related changes in oxidative stress in cartilage (i.e., through SIRT-3) protects against OA-progression and osteoclastogenesis in bone, suggesting that improving the resistance of cartilage to oxidative stress may be one therapeutic target for OA, or for maintaining bone homeostasis (Fu et al., 2016; Huh et al., 2016).

# Reactive Oxygen Species (ROS) and Oxidative Stress Alterations with Obesity

As a by-product of mitochondrial metabolism and homeostasis, ROS are produced and are highly regulated. With cell damage, mitochondrial dysfunction, or oxidative stress, ROS levels accumulate and are associated with the onset of metabolic dysfunction (Serra et al., 2013). ROS accumulation can lead to: lipid peroxidation and disruption of the cellular membrane; ER stress, resulting in protein mis-folding and unfolding, and a decrease in protein synthesis, ultimately rendering cells incapable of clearing misfolded proteins; and activation of Caspase-3 and cell apoptosis (Weiss et al., 2013). Furthermore, oxidative stress in tissues is reported to be an important pathogenic mechanism of MetS onset (Furukawa et al., 2004).

Increased fat mass has been linked to increased systemic markers of oxidative stress in humans and mice (**Figure 7**) (Furukawa et al., 2004). For example, increased peroxide and reduced endothelial nitric oxide synthase have been associated with cancellous bone loss in an obesity model (Ohnishi et al., 2009). In bone, reactive oxygen species and oxidative stress are critical to osteoclast differentiation. As such, ROS may contribute to osteoporosis and bone catabolism, through activation of RANKL—or receptor activator of NF-κB ligand- influencing osteoclast activity, as well as other pathways downstream such as NF-κB (Callaway and Jiang, 2015). With increased systemic inflammation associated with obesity, bone marrow macrophages and their progenitors can be increased (Singer et al., 2014), with related stimulation of osteoclastogenesis and reduced osteoblast development (Kyung et al., 2009; Halade et al., 2011). Yue and co-workers also reported that leptin produced from obese adipose tissue can bind to leptin receptors on mesenchymal stem cells promoting differentiation to adipocytes and inhibiting osteoblast formation (Yue et al., 2016). The central effects of leptin can also promote cancellous bone loss via the sympathetic nervous system (Ducy et al., 2000; Takeda et al., 2002).

In skeletal muscle, although low-levels of ROS are necessary for force production, high levels of ROS can result in great susceptibility to fatigue and contractile dysfunction (Powers et al., 2011) and are associated with reduced muscle repair capacity (Kozakowska et al., 2015). There may be a link between dysfunctional repair and overuse tendinopathies through ROS production, but this idea remains to be tested experimentally (Bestwick and Maffulli, 2004; Longo et al., 2008). Although the signaling pathway between ROS and osteoarthritis is not clear, there are examples of antioxidant therapy (e.g., dietary polyphenols and hyaluronic acid) that are useful in human studies (reviewed in Lepetsos and Papavassiliou, 2016).

# Lipotoxicity and Insulin Resistance with Obesity

Accumulation of lipids and lipid by-products can result in dysfunction in myocyte metabolic pathways. Lipid concentration is considered a strong indicator of the sensitivity of myocytes insulin-mediated pathways, as well as adipocyte and hepatocyte insulin sensitivity, significantly affecting substrate metabolism and circulating metabolite concentrations (Weiss et al., 2013). The adverse metabolic effects associated with ectopic lipid storage are supported by lipodystrophy studies, showing that muscle lipid accumulation can result in severe insulin resistance and diabetes (Weiss et al., 2013). In addition to the liver, primarily utilized for short term energy storage when circulating nutrient levels are elevated, skeletal muscle is a primary site of glucose uptake and utilization (Yu et al., 2012). Skeletal muscle insulin resistance largely contributes to whole body glucose levels and the presence of a pro-inflammatory environment, as the body is comprised of 40–50% muscle by weight (DeFronzo and Tripathy, 2009; Weiss et al., 2013). In response to low-level inflammation from DIO, extracellular matrix remodeling may contribute to the onset of insulin resistance, thereby inducing collagen synthesis and muscle fibrosis, and contributing to decreased insulin signaling (Williams et al., 2015).

Insulin resistance can also impact tissues of the joint organ system, because insulin resistance has been linked to chondrocyte dysfunction, and insulin appears to have a protective role for synoviocytes, whereby insulin blunts TNF-induced matrix metalloproteinase release (Hamada et al., 2015). However, patients with diabetes exhibit increased synovial levels of TNF-α and macrophages, suggesting that insulin resistance may impair the protective effect of insulin in the joint (Hamada et al., 2016), and greater insulin resistance is related to lower femoral neck strength (Srikanthan et al., 2014).

# Accumulation of Metabolic By-Products and Toxic Lipid Metabolites in Obesity

Elevated levels of oxidative stress can result in incomplete substrate oxidation. These impairments can lead to the excessive accumulation of toxic lipid metabolites (e.g., diacylglycerol, fatty acetyl CoA, and ceramides) and ROS, all natural by-products of cellular metabolism (Broussard and Devkota, 2016). Fatty acid trafficking in muscle may be one of the key factors involved in the onset of insulin resistance, by changing the availability of substrates involved in formation and clearance of harmful lipid intermediates (diacylglycerides and ceramide) (Mittendorfer, 2011). The balance between synthesis and breakdown of these intermediates may influence the balance between the formation and breakdown of FA and TGs, resulting in the storage and development of lipid depots in non-adipose tissues (Mittendorfer, 2011) from alterations in clearance mechanisms.

# Hyperglycemia and Advanced Glycation End Products with Obesity

Advanced glycation end products (AGEs) have a pronounced effect on many proteins, particularly collagen. Generally, AGEs accumulate in tissues and form cross-links with targeted proteins, alter cell structure, and interact with receptors that induce oxidative stress and inflammation, resulting in tissue damage (Sanguineti et al., 2014), **Figure 5**. Specifically, AGEs can alter the physiological failure behavior of some tissues, including tendons (Fessel et al., 2014), by increasing lateral spacing, accumulating cross-links between collagen fibrils, and reducing mechanical properties of tissues by reducing sliding behavior (Gautieri et al., 2017). Moreover, AGEs affect growth and modulate the physiological processes of OA (Franke et al., 2009), and damageassociated molecular patterns (DAMPS) can bind to RAGE and contribute to the pathogenesis of OA (Rosenberg et al., 2017). AGEs can induce muscle atrophy in a RAGE-mediated AMPK down-regulated manner (Chiu et al., 2016) and may affect skeletal muscle growth and contractile function in mice fed a diet high in AGEs (Egawa et al., 2017). In bone, AGEs are associated with decreased bone mineral density and impaired bone quality, eliciting oxidative stress and inflammatory responses in bone cells, while altering material properties of bone collagen fibers via cross-linking (Yamamoto and Sugimoto, 2016).

# OBESITY, BACTERIAL LIPOPOLYSACCHARIDE (LPS), AND GUT MICROBIOTA

It is well-established that the gut microbiota is altered with DIO in a manner that promotes a pro-inflammatory environment (Cani et al., 2009; Cândido et al., 2017). Through changes in tight junction proteins and intestinal barrier integrity, components of gram-negative bacteria may leak into the systemic circulation, resulting in increased systemic LPS concentrations in obese animals including humans (**Figure 8**) (Cani et al., 2008). Systemic LPS, inflammation, and metabolic disturbance resulting from altered gut microbiota and a leaky gut are believed to be key initiating factors leading to insulin resistance (Cani et al., 2007). In the context of musculoskeletal disease, an impaired gut barrier function has been implicated in rheumatoid arthritis and OA, and may present a viable therapeutic approach for disease management (Scher and Abramson, 2011). However, to date, cause and effect relations between rheumatoid arthritis and the gut microbiota are still being explored (Bravo-Blas et al., 2016).

We were among the first to identify a significant linkage between constituents of the gut microbiota (Methanobrevibacter and Lactobacillis spp.) and musculoskeletal changes in metabolic OA/joint damage severity (Collins et al., 2015a) and the onset of compromised muscle integrity (Collins et al., 2016c). These findings link systemic (serum) and local (synovial fluid) to LPS and musculoskeletal changes with diet-induced obesity (Huang et al., 2016) (**Figure 9**). Using a fecal transplant intervention, lipid profiles can be transferred from donors to germ-free hosts, and altered muscle integrity can be recapitulated, suggesting that the gut microbiota may have a direct effect on muscle development and integrity (Yan et al., 2016). The gut-OA pathophysiological link indicates that dietary and microbiotabased interventions may be important therapeutic opportunities that should be evaluated in the context of MSK health (Collins et al., 2015a; Huang et al., 2016; Berenbaum et al., 2017). There is a need for studies detailing the causal mechanisms of interactions among the gut-microbiota, LPS, and MSK changes. However, given the short-term changes in gut microbiota composition, as well as the decrease in variance in relative abundance across species, studies promoting gut microbiota diversity could be useful in this context. One such candidate is evaluating the impact of prebiotic fiber, which may positively modulate the gut microbiota to promote improved host interactions (Bomhof et al., 2014; Paul et al., 2016; Nicolucci et al., 2017; Parnell et al., 2017).

# IMPLICATIONS FOR CONSERVATIVE CARE STRATEGIES RELATED TO METS AND INFLAMMATION

Information regarding pharmacologic management for obesity and metabolic syndrome can be found elsewhere (Apovian et al., 2015), and promising therapeutic targets are discussed throughout this review (and highlighted in **Table 3**).

The inflammatory interface between damage to MSK components and MetS should be a critical consideration, as protecting and preserving MSK structural integrity could be a key approach to conservative management of MetS. Presently, clinical guidelines indicate that weight loss and exercise are good evidence-based conservative approaches for MSK conditions (McAlindon et al., 2014). In the context of osteoarthritis specifically, the combined effect of 18-months of dietary modulation and a combination program involving both aerobic and strength training-based exercise was found to induce weight loss, reduce knee compressive forces, reduce serum IL-6 levels, decrease infrapatellar fat pad size, decrease pain, and increase function in overweight or obese adults with knee OA (Messier et al., 2013; Beavers et al., 2015; Pogacnik Murillo et al., 2017). As such, careful characterization of exercise and dietary interventions in human and other animal models is needed to implement these strategies.

Although there are several candidates for dietary intervention, prebiotic fiber and probiotic supplementation targeting the microbiome (Nicolucci et al., 2017; Parnell et al., 2017), and decreased intake of dietary sugar (Te Morenga et al., 2013) are three safe, readily available, and translatable dietary interventions to protect against MSK damage due to metabolic disturbance that warrant further investigation. As the onset of MSK damage with metabolic disturbance does not have a readily identified starting point, studies are needed to evaluate the efficacy and utility of these dietary modulations in the patient population of interest (e.g., in prevention vs. reversibility and acute vs. chronic), as well as unintended consequences of these modifications. It would be critical to monitor muscle, tendon, bone, and cartilage tissues simultaneously while evaluating these dietary interventions, to facilitate an understanding of the benefits and interconnectedness of these MSK tissue changes and diseases.

# OPPORTUNITIES AND FUTURE RESEARCH DIRECTIONS

There are several gaps in the current understanding of the effects of low-level systemic inflammation vs. metabolic disturbance on musculoskeletal health. Many of these gaps are summarized in **Table 3**. Whether muscle is a primary target tissue in the motion segment, and whether muscular changes directly result TABLE 3 | Remaining questions, research agenda, and evidence-supported candidate pathways, targets, and conservative care opportunities.

#### Remaining Questions:


#### Research Agenda:


#### Evidence-Supported Candidate Pathways:


#### Evidence-Supported Candidate Targets:


in subsequent changes in bone, tendon, and joints, remains to be assessed and clarified. The studies presented here underscore the need for evaluating multiple musculoskeletal tissues and diseases in concert with MetS (as well as defining which elements of the MetS are responsible for which consequences of obesity). This approach will facilitate an understanding of how damage in each tissue develops with respect to the other tissues comprising a motion segment, how the different tissues may contribute to homeostasis or damage of the motion segment, and what role that MetS/inflammation plays in manifesting the risks of tissue damage. At present, it is unclear if: MetSrelated inflammation increases in severity with chronicity; systemic inflammation is constant with continued metabolic disturbance; or MetS-related inflammation is dynamic and fluctuating. It may be useful to evaluate tissues as "sources" or "targets" of low-level systemic inflammation (**Figure 9**). This conceptual framework is demonstrated in the data from our previous studies, which suggest a systemic-to-local timecourse shift from serum to synovial fluid, and serum to muscle changes (Collins et al., 2016a,b). Briefly, metabolic dysregulation initiates an inflammatory cascade, whereby an early perturbation in a presumptive inflammatory source (i.e., visceral fat and gut microbiota) results in target tissue (i.e., muscle) adaptation and subsequent damage. Although the source of the inflammation is unclear, as are the specific pathways involved, target tissue inflammation and damage may affect subsequent associated motion segment tissues (i.e., bone, tendon, and joints), and these relations are affected by many factors (i.e., the dynamic make-up/profile of the inflammatory components, time of exposure, and severity of the metabolic disturbance).

Furthermore, adaptation and impact of sources of inflammation on target tissues are important knowledge gaps that need to be addressed. It would be of value to determine if metabolic dysfunction and regulation are integrated and interdependent, or if source tissues and target tissues of the motion segment are regulated and affected differently and independently, at least in a first approximation. Identifying the potential for reversibility of tissue damage with dietinduced obesity is another critical area of research that should be addressed. The mechanism(s) by which target tissue or motion segment tissue changes may contribute to MetS and inflammation are incompletely understood at this time. Time-course studies are needed to answer these questions.

Discussion of the impact of metabolic disturbance on neuroregulatory systems, and a role for neuroinflammationmediated influence on loss of musculoskeletal integrity was limited in this review, but this may be a potentially important area for future research. Nearly all of the tissues of a motion segment discussed are innervated except for adult cartilage, and thus, neuro-regulation and how it is affected primarily and secondarily by MetS associated inflammation, may be critical in understanding the underlying mechanistic considerations. In particular, how these changes in neuroregulation and neuroinflammation contribute to a loss of muscle integrity is an interesting area for future research.

# SUMMARY

Studies presented in this review implicate several pathways that may be critical for the onset and progression of systemic inflammation due to metabolic disturbance and musculoskeletal damage. Some of the pathways that could be involved are mitogen-activated protein kinases (MAPK, p38 MAPK, JNK), myeloid differentiation primary response gene 88 (MyD88), NFκB, and the NLRP3 inflammasome. Likely, some or all of these pathways may be activated in parallel creating a potentially positive-feedback based redundant system. Investigations targeting specific pathways in the context of metabolic disturbance and MSK disease will provide much needed mechanistic insights to better understand the consequences of diet-induced obesity.

A motion segment, such as a leg is comprised of a number of interdependent components (i.e., muscles, tendons, bones, and articular joints) that rely on the integrated biological and mechanical integrity of all tissues for optimal function. Thus, the direct and indirect impact of dysregulation via MetS and associated inflammation on any one element may have a "ripple" effect that may be compounded over time. There are many important and relevant gaps in this research area, and addressing them will provide valuable insights into the relations among the motion segment, common inflammatory pathways, and resulting MSK disease.

# AUTHOR CONTRIBUTIONS

KC was responsible for conception and design, collection, and assembly of data analysis and interpretation of the data, drafting of the article, critical revision of the article for important intellectual content and final approval of the article. WH, GM, RR, JR, IS, and RZ were responsible for conception and design, collection and assembly of data analysis, interpretation of the data, critical revision of the article for important intellectual

# REFERENCES


content and final approval of the article. DH was responsible for conception and design, analysis and interpretation of the data, drafting of the article, critical revision of the article for important intellectual content and final approval of the article.

# ACKNOWLEDGMENTS

This work was supported by the Canadian Institutes of Health Research # RT736475 and MOP 115076, the Canada Research Chair Programme, the Alberta Innovates Health Solutions Osteoarthritis Team Grant, Alberta Innovates Health Solutions, Alberta Health Services Strategic Clinical Network Program, Canadian Institutes of Health Research Banting and Best Canada Graduate Scholarship, and the Killam Foundation. Due to space and scope limitations, we apologize to authors in this area whose work we were unable to include in the present review.


infrapatellar adipose tissue depots and their response to class II and III obesity in patients with OA. Arthritis Rheumatol. 69, 1396–1406. doi: 10.1002/art.40102


**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 Collins, Herzog, MacDonald, Reimer, Rios, Smith, Zernicke and Hart. 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.

# Human Milk and Donkey Milk, Compared to Cow Milk, Reduce Inflammatory Mediators and Modulate Glucose and Lipid Metabolism, Acting on Mitochondrial Function and Oleylethanolamide Levels in Rat Skeletal Muscle

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

Luciana Lopes Guimaraes, Universidade Santa Cecilia, Spain Vicente Lahera, Complutense University of Madrid, Spain Antonio Longo, Università degli Studi di Catania, Italy

# \*Correspondence:

Maria P. Mollica mariapia.mollica@unina.it

† These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 19 June 2017 Accepted: 10 January 2018 Published: 30 January 2018

#### Citation:

Trinchese G, Cavaliere G, De Filippo C, Aceto S, Prisco M, Chun JT, Penna E, Negri R, Muredda L, Demurtas A, Banni S, Berni-Canani R, Mattace Raso G, Calignano A, Meli R, Greco L, Crispino M and Mollica MP (2018) Human Milk and Donkey Milk, Compared to Cow Milk, Reduce Inflammatory Mediators and Modulate Glucose and Lipid Metabolism, Acting on Mitochondrial Function and Oleylethanolamide Levels in Rat Skeletal Muscle. Front. Physiol. 9:32. doi: 10.3389/fphys.2018.00032 Giovanna Trinchese1†, Gina Cavaliere1†, Chiara De Filippo<sup>1</sup> , Serena Aceto<sup>1</sup> , Marina Prisco<sup>1</sup> , Jong Tai Chun<sup>2</sup> , Eduardo Penna<sup>1</sup> , Rossella Negri <sup>3</sup> , Laura Muredda<sup>4</sup> , Andrea Demurtas <sup>4</sup> , Sebastiano Banni <sup>4</sup> , Roberto Berni-Canani <sup>3</sup> , Giuseppina Mattace Raso<sup>5</sup> , Antonio Calignano<sup>5</sup> , Rosaria Meli <sup>5</sup> , Luigi Greco<sup>3</sup> , Marianna Crispino<sup>1</sup> and Maria P. Mollica<sup>1</sup> \*

<sup>1</sup> Department of Biology, University of Naples Federico II, Naples, Italy, <sup>2</sup> Biology and Evolution of Marine Organisms, Stazione Zoologica Anton Dohrn, Naples, Italy, <sup>3</sup> European Laboratory for Food Induced Diseases, Department of Translational Medical Sciences, University of Naples Federico II, Naples, Italy, <sup>4</sup> Dipartimento di Scienze Biomediche, Università degli Studi di Cagliari, Cagliari, Italy, <sup>5</sup> Department of Pharmacy, University of Naples Federico II, Naples, Italy

Scope: Milk from various species differs in nutrient composition. In particular, human milk (HM) and donkey milk (DM) are characterized by a relative high level of triacylglycerol enriched in palmitic acid in sn-2 position. These dietary fats seem to exert beneficial nutritional properties through N-acylethanolamine tissue modulation. The aim of this study is to compare the effects of cow milk (CM), DM, and HM on inflammation and glucose and lipid metabolism, focusing on mitochondrial function, efficiency, and dynamics in skeletal muscle, which is the major determinant of resting metabolic rate. Moreover, we also evaluated the levels of endocannabinoids and N-acylethanolamines in liver and skeletal muscle, since tissue fatty acid profiles can be modulated by nutrient intervention.

Procedures: To this aim, rats were fed with CM, DM, or HM for 4 weeks. Then, glucose tolerance and insulin resistance were analyzed. Pro-inflammatory and anti-inflammatory cytokines were evaluated in serum and skeletal muscle. Skeletal muscle was also processed to estimate mitochondrial function, efficiency, and dynamics, oxidative stress, and antioxidant/detoxifying enzyme activities. Fatty acid profiles, endocannabinoids, and N-acylethanolamine congeners were determined in liver and skeletal muscle tissue.

Results: We demonstrated that DM or HM administration reducing inflammation status, improves glucose disposal and insulin resistance and reduces lipid accumulation in skeletal muscle. Moreover, HM or DM administration increases redox status, and mitochondrial uncoupling, affecting mitochondrial dynamics in the skeletal muscle. Interestingly, HM and DM supplementation increase liver and muscle levels of the N-oleoylethanolamine (OEA), a key regulator of lipid metabolism and inflammation.

**107**

Conclusions: HM and DM have a healthy nutritional effect, acting on inflammatory factors and glucose and lipid metabolism. This beneficial effect is associated to a modulation of mitochondrial function, efficiency, and dynamics and to an increase of OEA levels in skeletal muscle.

Keywords: human milk, donkey milk, mitochondrial functions, mitochondrial dynamics, oxidative stress, inflammation

# INTRODUCTION

Human milk (HM), the natural food of infants, provides an adequate supply of all nutrients necessary to support postnatal growth and development. In addition, it provides immunomodulant components and plays a key role in preventing inflammation and metabolic diseases throughout life (Agostoni et al., 2013; Cacho and Lawrence, 2017). Interestingly, it has been demonstrated that breastfed individuals have lower rate of obesity and type 2 diabetes than those fed with infant formula (Owen et al., 2006). Such effects have been attributed to appetite regulation and reduced weight gain in breast-fed children and/or to the presence of unique nutrients or bioactive constituents in HM. In fact, metabolism-regulating hormones, such as leptin, adiponectin, resistin and ghrelin, detected in HM (Lönnerdal, 2013), may be involved in the regulation of growth in infancy and might influence the programming of energy balance with long-term consequences on health (Agostoni et al., 2013). In addition, the peculiar high concentration in HM of palmitic acid in the sn-2 position of triacylglycerol backbone seems to play a crucial role in the regulation of energy and lipid metabolism (Innis, 2016). Interestingly, also donkey milk (DM) is characterized by a high concentration of palmitic acid in sn-2 position of triacylglycerol, and is recognized as the best potential substitute for HM due to its remarkable nutritional value coupled with good palatability and reduced allergenicity (Tafaro et al., 2007; Fiocchi et al., 2010). Recently, we have demonstrated that palmitic acid in sn-2 position of triacylglycerol is able to modulate N-acylethanolamine levels in rat tissues, which may possibly explain its effects on lipid and energy metabolism (Carta et al., 2015). Among N-acylethanolamine, palmitoylethanolamide (PEA) and oleylethanolamide (OEA), are PPARα endogenous agonists. PPARs are a family of nuclear receptors that regulate fatty acid metabolism in different tissues. In particular, PPARα activation increases peroxisomal and mitochondrial oxidation and leads to anti-inflammatory and antioxidant effects (Fu et al., 2005; Lo Verme et al., 2005; Contreras et al., 2013).

Recently, we have demonstrated that HM or DM administration improves liver inflammatory state, enhancing hepatic mitochondrial functions and uncoupling (Trinchese et al., 2015). Mitochondrial uncoupling, dissipating the proton gradient across the inner membrane, creates a futile cycle of glucose and fatty acid oxidation without generating ATP (Nedergaard et al., 2005; Tseng et al., 2010), and thereby increases lipid oxidation and reduces intracellular lipid content (Harper et al., 2008). Promoting this inefficient metabolism that generates heat instead of ATP, mitochondrial uncoupling can serve as a potential treatment for obesity. Therefore, drugs, nutrients, or natural molecules modulating the mitochondrial function and efficiency can be considered as potential treatment for obesity and insulin resistance promoting an inefficient metabolism (Li et al., 2000). To ensure a healthy mitochondrial population, cells are equipped with various quality control systems, that regulate mitochondrial shape, function, and mass. These regulatory systems include mitochondrial biogenesis and dynamics (Liesa and Shirihai, 2013). In particular, mitochondrial dynamic is a tightly regulated process controlled by fusion and fission events (Detmer and Chan, 2007). During mitochondrial fusion, optic atrophy (OPA)1 is responsible for mitochondrial inner membrane fusion, and two other proteins mitofusins 1 and 2 (Mfn1 and Mfn2) mediate outer membrane fusion. On the other hand, dynamin-related protein (DRP)1 and fission protein (Fis)1 are involved in mitochondrial fission (Detmer and Chan, 2007). Usually, a shift toward fusion optimizes mitochondrial function and plays a beneficial role in the maintenance of long-term bioenergetic capacity. In contrast, a shift toward fission results in numerous mitochondrial fragments contributing to elimination of irreversibly damaged mitochondria through autophagy. Not surprisingly, alteration of mitochondrial dynamics has profound impact on several pathological conditions including metabolic diseases such as obesity and type 2 diabetes (Sebastián et al., 2012; Schneeberger et al., 2013; Zorzano et al., 2015). Skeletal muscle is one of the major targets of insulin accounting for high percentage of the hormone-dependent glucose uptake. Indeed, metabolic impairment in skeletal muscle strongly affects glucose disposal in the whole-body. Moreover, skeletal muscle is a chief determinant of resting metabolic rate, whose reduction is associated with weight gain (Ravussin et al., 1988). Recent data have demonstrated that the insulin-resistant condition in skeletal muscle is characterized by alterations of mitochondrial function and efficiency (Cavaliere et al., 2016).

Here, we examined the effects of supplementation of cow milk (CM), HM, and DM on systemic and skeletal muscle inflammation, insulin sensitivity, endogenous lipids, and endocannabinoids levels in skeletal muscle and liver, and mitochondrial functions in skeletal muscle.

# MATERIALS AND METHODS

# Animals and Chemicals

All chemicals were purchased from Sigma–Aldrich (St. Louis, MO, USA), unless specified otherwise. Male Wistar rats (60 days old; 345 ± 7 g; Charles River, Calco, Lecco, Italy) were caged in a temperature-controlled room and exposed to a daily light–dark cycle (12/12 h) with free access to chow diet and drinking water. The rats were divided into four experimental groups (n=7), three of them were supplemented with equicaloric intake (82 kJ) of raw CM, DM, or HM drinking 21, 48, or 22 ml/day, respectively. The animals were treated for 4 weeks. The last group did not receive milk supplement and was used as control. Despite the different volumes used, the energy density provided by the different milk supplements was kept virtually the same (**Table 1**). DM from the Ragusana breed and CM were obtained from "Azienda Agricola Insalata" (Oliveto Citra, SA, Italy). Human milk was kindly provided by the milk bank "Banca del Latte Materno" Ospedale Colle dell'Ara—Chieti-Italy. At the end of the treatments, the animals were anesthetized by intra-peritoneal injection of chloral hydrate (40 mg/100 g body weight), and blood was taken from the inferior cava. Skeletal muscle was removed from the hind leg for analysis, and the samples not immediately used for mitochondrial preparation were stored at −80◦C. All experiments were conducted in compliance with national guidelines for the care and use of research animals (D.L. 116/92, implementation of EEC directive 609/86).

# Serum and Tissue Parameters

Plasma concentrations of triglycerides and cholesterol were measured by colorimetric enzyme reaction by use of commercial kits (SGM Italia, Italy and Randox Laboratories ltd., United Kingdom). Instead, specific ELISA kits were used to measure the serum and tissue levels of IL-1α, IL-10 (Thermo Scientific, Rockford, IL, USA), TNF-α, and monocyte chemoattractant protein-1 (MCP-1) (Biovendor R&D, Brno, Czech Republic), adiponectin and leptin (B-Bridge International Mountain View, CA).

# Oral Glucose Tolerance Test and Insulin Tolerance Test

For oral glucose tolerance test, rats were allowed to fast overnight and then orally dosed with glucose (3 g/kg body weight) dissolved in water. For insulin tolerance test, fasting rats (5 h)


Values represent means ± SEM from n = 7 animals/group. Labeled means without a common letter differ (P < 0.05).

were intraperitoneally injected with insulin (homolog-rapidacting, 10 unit/kg body weight in sterile saline; Novartis, Basel, Switzerland). Blood was collected by direct flow from a small tail cut, before and after the treatments at certain intervals, and the glucose levels were determined by glucose monitor calibrated for rats (BRIO, Ascensia, NY), and the insulin levels by ELISA (Mercodia rat insulin; Mercodia, Uppsala, Sweden). Basal fasting values of serum glucose and insulin were used to calculate Homoeostatic Model Assessment (HOMA) index as [Glucose (mg/dL) <sup>∗</sup> Insulin (mU/L)]/405 (Cacho et al., 2008).

# Mitochondrial Parameters and Basal and Inducible Proton Leak

Mitochondrial isolation, oxygen consumption, and proton leakage measurements were performed as previously reported (Cavaliere et al., 2016). Oxygen consumption (polarographically measured using a Clark-type electrode) was measured in the presence of substrates and ADP (state 3) or with substrates alone (state 4), and respiratory control ratio (RCR) was calculated. Mitochondrial proton leakage was assessed by a titration of the steady-state respiration rate as a function of the mitochondrial membrane potential in skeletal muscle mitochondria. Carnitine palmitoyl-transferase (CPT) system, aconitase, and superoxide dismutase (SOD) specific activity were measured spectrophotometrically as previously reported (Flohe and Otting, 1984; Alexson and Nedergaard, 1988; Hausladen and Fridovich, 1996). Rate of mitochondrial H2O<sup>2</sup> release was assayed by following the linear increase in fluorescence (ex 312 nm and em 420 nm) due to the oxidation of homovanillic acid in the presence of horseradish peroxidase (Barja, 1998).

# Lipid Content, Redox Status, and Nrf2-Activated Enzyme Activities in Skeletal Muscles

Total lipid content in the skeletal muscle was estimated by using the Folch method. Briefly, the skeletal muscle was weighed, chopped into small pieces, thoroughly mixed, and finally homogenized with water (volumes equal to twice the skeletal muscle weight) in a Polytron homogenizer. On appropriate aliquots of the homogenate, lipid content was determined gravimetrically after extraction in chloroform–methanol and evaporation to constant weight by a rotating evaporator (Folch et al., 1957).

TABLE 2 | The sequences (5′ -3′ ) of the primers used in Real Time PCR.


Reduced (GSH) and oxidized (GSSG) glutathione concentrations in skeletal muscle were measured with the dithionitrobenzoic acid (DTNB)-GSSG reductase recycling assay (Bergamo et al., 2007), and the GSH/GSSG ratio was used as an oxidative stress marker. To investigate the possible involvement of NF-E2-related factor 2 (Nrf2) in the diet-induced stress, cytoplasmic extracts were prepared from rat skeletal muscle. The enzymatic activities of glutathione S-transferases (GSTs) and quinone-oxidoreductase (NQO1) were evaluated spectrophotometrically in cytoplasmic extracts (Benson et al., 1980; Habig and Jakoby, 1981; Levine et al., 1990).

# Lipid Analysis

Aliquots of total lipid extract from tissues were mildly saponified as previously described (Banni et al., 1996) in order to obtain free fatty acids for HPLC analysis. Separation of unsaturated fatty acids was carried out with an Agilent 1100 HPLC system (Agilent, Palo Alto, Calif., USA) equipped with a diode

common letter differ (P < 0.05).

array detector as previously reported. Spectra (195–315 nm) of the eluate were obtained every 1.28 s and were electronically stored. These spectra were acquired to confirm the identified HPLC peaks (Melis et al., 2001)**.** Quantification of anandamide (AEA), 2-arachidonoyl glycerol (2-AG), PEA, and OEA was

conducted by liquid chromatography-atmospheric pressure chemical ionization-mass spectrometry (LC-APCI-MS) using as internal standards their deuterated homologs as previously described (Piscitelli et al., 2011).

# Real Time PCR Analysis

Total RNA was extracted from skeletal muscle using the TRIzol reagent (Ambion). After DNase treatment, RNA was quantified and reverse-transcribed (1 µg) using the Advantage RT-PCR kit (Clontech). For the evaluation of mitochondrial fission and fusion gene transcription, we used murine primers, whose sequences are shown in **Table 2**. qPCR was performed as previously described (Trinchese et al., 2015).

# Western Blot Analysis

Tissues were homogenized in lysis buffer (150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, Tris-HCl 20 mM, pH 7.5) containing protease and phosphatase inhibitors. Proteins (15– 30 µg/lane) were separated on 10–12% SDS-polyacrilamide gel and transferred to PVDF membranes (GE Healthcare). Blocking (5% non-fat milk, 30 min) and immunoreaction (o/n) were performed in TBST buffer (0.1% Tween, 150mM NaCl, 10mM Tris-HCl, pH 7.5) at room temperature. The primary antibodies were anti-Mitofusin1, anti-Mitofusin2, anti-DRP1 (all Santa Cruz), and anti-β-actin (BD Biosciences). Signals were visualized by horseradish peroxidase-linked secondary antibodies in enzyme-linked chemiluminescence (ECL, Millipore). Relative protein levels were normalized with β-actin levels on the same membrane.

# Transmission Electron Microscopy

Samples of soleus muscle were cut into ∼1 mm<sup>3</sup> fragments and fixed in 2.5% glutaraldehyde in 0.1M phosphate buffer, (pH 7.4) for 4 h at 4◦C. After washing in phosphate buffer, tissues were post-fixed with 1% osmium tetroxide in phosphate buffer for 2 h at 4◦C. Samples were dehydrated through graded alcohols (50, 70, 90, and 100%) and propylene oxide, and subsequently embedded in EMbed 812 resin (Electron Microscopy Science) (48 h at 60◦C). Ultrathin (70 nm thick) sections were cut at a Sorvall Porter-Blum ultramicrotome, contrasted with uranyl acetate and lead citrate. The ultrastructure of the sample was visualized on a transmission electron microscope Philips 208S at Centro


PEA, N-palmitoylethanolamine; AEA, anandamide; 2AG, 2-arachidonoyl glycerol. Values represent means ± SE from n = 7 animals/group.

TABLE 3 | Tissue levels of AEA, PEA, and 2AG.

(P < 0.05).

di Servizi Metrologici Avanzati (CeSMA, University of Naples Federico II); micrographs were acquired with a Mega View II Soft Imaging System camera. Mitochondrial compartment volume density was measured using the stereological point-sampling technique (Weibel et al., 1966; Altunkaynak et al., 2012). For each experimental group, 10 random images were taken at 25,000 x magnification. A grid with equally and symmetrically spaced intersection points was overlaid on each micrograph using the Image J1.50 software (NIMH). Mitochondrial volume density for each micrograph was calculated as the ratio of points overlapping the cellular compartment of interest and the total number of grid points overlapping the cytoplasm area. To measure mitochondrial fusion and fission percentage, the number of mitochondria undergoing fusion and fission on total mitochondrial number in the visual field was counted in 10 images for each experimental group.

# Statistical Analysis

Data were analyzed by one-way ANOVA followed by the Bonferroni post-hoc test. Labeled means without a common letter

differ (P < 0.05). Analyses were performed and plotted using GraphPad Prism (GraphPad Software, San Diego, CA, USA).

# RESULTS

# Body Weight and Food Intake

As expected, although different milk treatments provided similar total food intake, CM-treated animals exhibited a significant increase in body weight and body weight gain when compared with control, DM- and HM-treated rats (**Table 1**).

# Serum Metabolites and Inflammatory Parameters

Serum parameters were reported in **Figure 1**. Triglycerides significantly decreased in the DM-fed animals (**Figure 1A**), while cholesterol and MCP1 serum levels were not affected by different

presented as means ± SEM from n = 7 animals/group. Labeled means without a common letter differ (P < 0.05).

milk treatments (**Figures 1B,D**, respectively). TNF-α and IL 1 levels significantly decreased in both serum and tissue of HMand DM-fed animals in comparison with control and CMfed rats (**Figures 1C,E**, respectively). IL 10 level increased in the serum of all treated groups (HM>DM>CM) compared to control (**Figure 1F**), while IL 10 level increased in the skeletal muscle of DM and HM compared to control and CM group (**Figure 1F** inset). Leptin concentration was significantly higher in CM-treated animals compared to the other groups; consistently, adiponectin concentration decreased in CM-fed animals, whereas it significantly increased in the DM- and HMtreated rat (**Figures 1G,H**). Accordingly, the leptin/adiponectin (L/A) ratio significantly increased in CM-treated rats compared to the other two groups (**Figure 1I**). Glucose levels were significantly lower in HM- and DM-fed rats compared to the other groups, while insulin levels were significantly lower only in HM rats (**Figure 2A**). Therefore, compared with controls, a marked reduction of HOMA index was observed in the DM and HM groups (HM<DM; **Figure 2B**). Moreover, HM- and DMtreated rats exhibited a higher tolerance to glucose load than CM-treated and control, (evaluated as "area under the curve" AUC of glucose levels during time); in agreement, the HM- and DM-fed animals showed a significant reduction of insulin release (AUC) after glucose load in comparison with the other groups (HM<DM) (**Figures 2C,D**). In addition, insulin tolerance test revealed better glucose reduction (AUC HM<DM), following insulin administration, in DM- and HM-fed rats compared with the CM-treated or control groups (**Figure 2E**).

# Skeletal Muscle and Liver Lipid Analysis

Fatty acid analysis showed a significant increase of OEA in the liver and skeletal muscle of rats fed with HM and DM compared to CM or control animals (**Figure 3**), while the other amides and 2-AG did not change significantly, although PEA showed an increasing trend (**Table 3**).

# Evaluation of Mitochondrial Parameters in Skeletal Muscle

When succinate was used as a substrate, state 4 respiration increased in mitochondria isolated from the DM- and HMtreated animals compared to the other groups (**Figure 4A**). In presence of palmitoyl-carnitine, state 4 increased in all three milk-treated groups, with the highest rate being observed in the HM animals (**Figure 4B**). Mitochondrial state 3 using both substrates, increased in the three milk-treated groups, with the highest oxidation rates observed again in the HMtreated group (**Figures 4A,B**). CPT activity increased in the DMand HM-treated animals compared to control and CM-treated rats (**Figure 4C**). The RCR values did not significantly change among the groups, confirming the high-quality of mitochondrial preparations (**Figures 4A,B** insets). Mitochondria from control and CM-treated rats exhibited comparable kinetic curves of basal proton-leakage, that was significantly increased in DM-treated rats and even more in HM-treated ones (**Figure 4D**). Likewise, mitochondrial free fatty acid (FFA)-induced proton leakage was significantly increased in the DM- and HM-treated animals (HM>DM), but the effects on mitochondrial leakage in the CMtreated animals was the opposite, as it was slightly decreased compared to control (**Figure 4E**). Thus, the HM-treated animals exhibited the highest oxygen consumption to maintain the same membrane potential. H2O<sup>2</sup> yield increased in CM-fed animals compared to controls, while it significantly decreased in DMand at higher extent in HD-treated rats (**Figure 4F**). SOD activity was significantly higher in both DM- and HM-treated rats compared to control and CM-treated animals (**Figure 4G**). Similarly, aconitase activity was significantly increased in the

DM-treated animals and, even more, in the HM-treated ones. In contrast, the CM-treated animals displayed a significant decrease in aconitase activity (**Figure 4H**). Altogether, these data suggest that the administration of DM and HM may result in diminishing the mitochondrial ROS emission.

# Antioxidant/Detoxifying Defenses

The lipid content in skeletal muscle of CM-treated rats was substantially higher than in the other groups (**Figure 5A**). Dietary supplementation with DM or HM improved the antioxidant state and cytoprotective enzyme activities. Indeed, while GSH levels significantly increased in DM or HM group, compared to the other groups (**Figure 5B**), GSSG content was reduced, resulting in an increased GSH/GSSG ratio in the HM and DM groups (HM>DM) (**Figures 5C,D**). In addition, the activities of GST and NQO1 were significantly higher in the DM- or HM-treated rats than in the CM-treated and control rats (**Figures 5E,F**).

# Mitochondrial Dynamics

The mRNA levels of Opa1, Mfn1, and Mfn2 significantly increased in HM-fed rats compared to other groups (**Figures 6A–C**). On the other hand, Drp1 and Fis1 mRNA levels in DM and HM significantly decreased when compared to CM or control groups (**Figures 6D,E**). These results were consistent with the morphological data on the ultrastructure of

the skeletal muscles, in fact, mitochondria from DM and HM groups appeared more abundant and larger than those in the control group (**Figure 7A**). This was confirmed by a higher

fusion percentage in HM and DM (**Figure 7B**) and a lower percentage of fission in HM (**Figure 7C**), as well as an increased mitochondrial volume density in DM and HM (**Figure 7D**). Moreover, the mitochondria of DM and HM appeared more electron-dense compared to those in CM and control groups. At the molecular level, the expression of two marker proteins for mitochondrial fusion, MFN1 and MFN2, was evidently increased in the DM and HM in comparison with control and CM animals (**Figures 8A,B**), while the expression levels of the marker protein for fission, DRP1, did not significantly change among all groups (**Figure 8C**).

# DISCUSSION

The main finding of this study is that oral supplementation with HM and DM of rat diet affects glucose and lipid metabolism, modulating pro- and anti-inflammatory serum and tissue mediators. In particular, HM and DM exert their beneficial effects on skeletal muscles by reducing fat accumulation and modulating mitochondrial function, efficiency, and dynamics associated to an improvement of the redox status and activation of detoxifying enzymes (Nrf2 pathway). Our study highlights the ability of HM, and to a lesser extent of DM, to control glucose homeostasis, improving HOMA index and glycaemic and insulin response, as demonstrated by the responses to glucose load and insulin tolerance tests. These results are consistent with other data indicating that HM, compared with CM, lowers insulin blood levels (Gunnerud et al., 2012). It has been shown that insulin response to glucose can be modulated not only by the lactose component, but also to the whey fraction, or the bioactive peptides that are present in the milk (Nilsson et al., 2004, 2007). Here, we have shown that the modulation of glucose metabolism by HM and DM, strongly correlates with the reduction of inflammatory mediators in serum and tissue. Moreover, we have shown that the modulation of glucose metabolism may be related to adiponectin and leptin balance. These two hormones, derived from fat cells, are involved not only in glucose and lipid metabolism, controlling energy homeostasis, but also in the modulation of inflammation (Minokoshi et al., 2002; Yamauchi et al., 2002). With the accumulation of fat mass, leptin level increases, while adiponectin decreases. Accordingly, CM-treated group, characterized by higher body lipid levels compared with the HM and DM groups (Trinchese et al., 2015), showed increased leptin level and decreased adiponectin level. Interestingly, HM and DM-treated animals, with a body lipid level similar to control (Trinchese et al., 2015), displayed significantly increased adiponectin in their sera compared with other groups. In some animal models, a decrease in adiponectin level was found to parallel to a reduction in insulin sensitivity and to precede the onset of type 2 diabetes (Chakraborti, 2015). Adiponectin secretion is inhibited by several factors including high level of TNF-α and oxidative stress (Chakraborti, 2015). Hence, our data, showing increased adiponectin level in HM and DM groups, associated with decreased level of TNF-α, may indicate a lower grade of inflammation in these animals. It has been proposed that a useful index of metabolic diseases is the ratio of leptin to adiponectin (L/A), which shows a better correlation to insulin resistance than the level of leptin and adiponectin alone (Oda et al., 2008). Our data indicates a significantly lower L/A ratio in DM and HM groups, that correlates with the modulation of glucose metabolism in these animals. Adiponectin and leptin have been recently detected in HM (Lönnerdal, 2013), and it has been postulated that these adipokines may be involved in the regulation of growth in infancy and influence the programming of energy balance with long-term consequences on health (Guardamagna et al., 2012; Agostoni et al., 2013). Indeed, it has been demonstrated that breastfeeding is associated with a reduced risk of type 2 diabetes, with lower blood glucose and serum insulin concentrations during infancy and with lower insulin levels in adulthood (Owen et al., 2006). Our study suggests that DM affects glucose metabolism in a similar way HM does, although the biologically active components of DM remain largely unknown.

Interestingly, HM and DM have high concentration of palmitic acid in the sn-2 position of triacylglycerol backbone (Innis, 2016), that can be responsible for the increased levels of OEA that we found in liver and skeletal muscle of HM and DM animals. These increased levels of OEA may have important implications in terms of energy metabolism and inflammation. Indeed, OEA, recently identified as an important regulator of inflammation, is an endogenous ligand of the nuclear receptor PPAR alpha, which regulates fatty acid metabolism in different tissue. In particular, PPARα activation increases peroxisomal and mitochondrial oxidation and leads to anti-inflammatory and antioxidant effects (Contreras et al., 2013; Pontis et al., 2016).

Considering the central role of the skeletal muscle in lipid and glucose metabolism, we evaluated the effects of different milks on skeletal muscle mitochondria in terms of their metabolic function, efficiency and dynamics. We observed that DM, and even more HM, had several beneficial effects. For instance, both HM- and DM skeletal muscle mitochondria showed increased respiratory capacity and improved redox status even when the ability to utilize fat as a metabolic fuel was increased. The increased mitochondrial fatty acid oxidation observed in the skeletal muscle is likely to be related to an enhancement of CPT activity, which would further increase entry of longchain FFAs into mitochondria. The consequent increase in lipid oxidation is apparently sufficient to handle the decreased load of skeletal muscle FFAs. In addition to stimulating fatty acid oxidation, HM induced a less efficient utilization of lipid

substrates through the stimulation of thermogenic mechanisms such as proton-leakage. This decline in mitochondrial energy efficiency may also contribute to fat burning. Therefore, we hypothesized that HM-fed animals, similarly to DM-treated rats, might be protected from development of obesity and insulin resistance through these mechanisms. The increase in proton-leakage has a further metabolic implication that is the maintenance of mitochondrial membrane potential below the critical threshold for ROS production (Skulachev, 1991; Mailloux and Harper, 2011). Accordingly, in HM- and DMtreated animals, we observed a decrease in H2O<sup>2</sup> yield, an enhancement of antioxidant defense mechanisms (aconitase inhibition, SOD activity), and an improvement of redox status (GSH/GSSG ratio), as well as increased activities of detoxifying enzymes (GST-NQO1) which are at least in part attributable to the activation of the Nrf2–ARE pathway. Another important observation in our study is that DM and HM improve glucose metabolism modulating mitochondrial dynamics. Mitochondrial dynamic is a complex process that impacts on mitochondrial metabolism through complex modulation of proteins that play a key role in fusion and fission machinery. The fusion protein Mfn2, a potent modulator of mitochondrial metabolism, also controls cell metabolism and insulin signaling by limiting reactive oxygen species production, and is subjected to tight regulation. Indeed, proinflammatory cytokines, glucocorticoids, or lipid availability block its expression, whereas exercise and increased energy expenditure promote its upregulation (Zorzano et al., 2015; Schrepfer and Scorrano, 2016). Our results show that HM and DM decrease Drp1 and Fis1 mRNA levels, while HM increases the levels of mRNA coding for mitochondrial fusion proteins (OPA1, Mfn1, and Mfn2), suggesting a shift toward fusion in these animals. The increase in mitochondrial fusion in the skeletal muscles of HM and DM animals was confirmed by western blot and electron microscopy analysis. This latter analysis also revealed that mitochondria of DM and HM are more abundant, larger and more electron-dense than those in the CM and control animals. The higher electrondensity of DM and HM could be linked to mitochondrial state, in particular to the inner membrane condensation (Mannella, 2006; Sancho et al., 2014). Condensed state (alternative to orthodox state) has been associated to more active mitochondria, suggesting that these organelles regulate their shape to adjust their activity with metabolic conditions (Mannella, 2006). Based on these data, we propose that the shift toward fusion, upon DM and HM administration, contribute to the improvement of glucose metabolism in these animals. Accordingly, previous data indicated that mitochondrial fragmentation and enhancements of fission machinery negatively affected glucose metabolism, leading to an increase in ROS formation due to alterations in mitochondrial electron transport and coupling (Yu et al., 2008; Westermann, 2012). Our data indicate that both HM and DM have a protective role on metabolism, with HM having higher beneficial effects. These potentiated effects of HM may depend on molecular mechanisms involving modulation of fusion/fission processes and consequently diminished levels of inflammation, ROS, and improved glucose metabolism. A shift toward fission results in numerous mitochondrial fragments contributing to the elimination of irreversibly damaged mitochondria through autophagy. At the same time, the mixing of the matrix and the inner membrane allows the respiratory machinery components to cooperate most efficiently, reducing the uncoupling effect. On the other hand, a shift toward fusion optimizes mitochondrial function and thereby plays a beneficial role in maintaining longterm bioenergetics capacities. Indeed, fusion favors generation of interconnected mitochondria, which may be responsible for a decrease in mitochondrial efficiency stimulating the uncoupling effect that contributes to the dissipation and rapid provision of energy. Based on these findings, we can identify the activation of mitochondrial fusion and the uncoupled respiration as part of a complex protective mechanism that reduces oxidative stress and maintains healthy and functional mitochondria (Li et al., 2000; Liesa et al., 2009; Liesa and Shirihai, 2013). In agreement with this hypothesis, Mfn2 depleted cells exhibited a decreased mitochondrial proton leak and increased bioenergetics efficiency (Bach et al., 2003), suggesting that the loss-of-function of Mfn2 in obesity condition may enhance bioenergetics efficiency and thereby contribute to obesity by reducing energy expenditure and increasing fat energy store (Liesa et al., 2009). In conclusion, our study highlights that dietary supplementation with HM or DM decreases inflammatory factors, modulates lipid and glucose metabolism, and increases OEA tissue levels (**Figure 9**). Interestingly, we indicate that the beneficial effects elicited by HM and DM are, at least in part, mediated by their ability to improve redox status, modulating Nrf2 activation, mitochondrial uncoupling, and dynamics.

# ETHICS STATEMENT

The procedures reported here, involving animals and their care, were approved by the Institutional Committee on the Ethics of Animal Experiments (CSV) of the University of Naples Federico II and by the Ministero della Salute.

# AUTHOR CONTRIBUTIONS

MPM, LG, and MC: Conceived the original idea, designed, and supervised the whole study; GT, GC, CDF, SA, MP, EP, LM, and AD: Performed the experiments; MPM and MC: analyzed and interpreted data; MPM, JTC, RN, SB, AC, GMR, RM, RBC, LG, and MC: Critically revised the manuscript for intellectual content; MPM and MC: wrote the paper.

# ACKNOWLEDGMENTS

The authors thank Dr. Roberta Scognamiglio for technical assistance.

# REFERENCES


**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 Trinchese, Cavaliere, De Filippo, Aceto, Prisco, Chun, Penna, Negri, Muredda, Demurtas, Banni, Berni-Canani, Mattace Raso, Calignano, Meli, Greco, Crispino and Mollica. 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.

# Obesity, Fat Mass and Immune System: Role for Leptin

Vera Francisco<sup>1</sup> \* † , Jesús Pino<sup>1</sup>† , Victor Campos-Cabaleiro<sup>1</sup> , Clara Ruiz-Fernández<sup>1</sup> , Antonio Mera<sup>2</sup> , Miguel A. Gonzalez-Gay<sup>3</sup> , Rodolfo Gómez<sup>4</sup> and Oreste Gualillo<sup>1</sup>

<sup>1</sup> The NEIRID Group (Neuroendocrine Interactions in Rheumatology and Inflammatory Diseases), Servizo Galego de Saude and Instituto de Investigación Sanitaria de Santiago, Santiago University Clinical Hospital, Santiago de Compostela, Spain, <sup>2</sup> Servizo Galego de Saude, Division of Rheumatology, Santiago University Clinical Hospital, Santiago de Compostela, Spain, <sup>3</sup> Epidemiology, Genetics and Atherosclerosis Research Group on Systemic Inflammatory Diseases, Hospital Universitario Marqués de Valdecilla, Universidad de Cantabria and IDIVAL, Santander, Spain, <sup>4</sup> Musculoskeletal Pathology Group, Servizo Galego de Saude and Instituto de Investigación Sanitaria de Santiago, Santiago University Clinical Hospital, Santiago de Compostela, Spain

# Edited by:

Joaquin Garcia-Estañ, Universidad de Murcia, Spain

#### Reviewed by:

Deanne Helena Hryciw, Griffith University, Australia Antonio La Cava, University of California, Los Angeles, United States Giuseppe Matarese, Università degli Studi di Napoli Federico II, Italy

#### \*Correspondence:

Vera Francisco vlgfrancisco@gmail.com

†These authors have contributed equally to the realization of this work.

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 18 December 2017 Accepted: 11 May 2018 Published: 01 June 2018

#### Citation:

Francisco V, Pino J, Campos-Cabaleiro V, Ruiz-Fernández C, Mera A, Gonzalez-Gay MA, Gómez R and Gualillo O (2018) Obesity, Fat Mass and Immune System: Role for Leptin. Front. Physiol. 9:640. doi: 10.3389/fphys.2018.00640 Obesity is an epidemic disease characterized by chronic low-grade inflammation associated with a dysfunctional fat mass. Adipose tissue is now considered an extremely active endocrine organ that secretes cytokine-like hormones, called adipokines, either pro- or anti-inflammatory factors bridging metabolism to the immune system. Leptin is historically one of most relevant adipokines, with important physiological roles in the central control of energy metabolism and in the regulation of metabolism-immune system interplay, being a cornerstone of the emerging field of immunometabolism. Indeed, leptin receptor is expressed throughout the immune system and leptin has been shown to regulate both innate and adaptive immune responses. This review discusses the latest data regarding the role of leptin as a mediator of immune system and metabolism, with particular emphasis on its effects on obesity-associated metabolic disorders and autoimmune and/or inflammatory rheumatic diseases.

Keywords: adipokines, adipose tissue, immunometabolism, leptin, metabolism, rheumatic diseases, rheumatoid arthritis, Type 2 diabetes mellitus (T2DM)

**Abbreviations:** ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; AMPK, adenosine monophosphate-activated protein kinase; Arg-1, arginase-1; BMDM, bone marrow-derived macrophages; BMI, body mass index; cAMP, cyclic adenosine monophosphate; CCL, CC-chemokine ligand; CD, cluster of differentiation; COX-2, cyclooxygenase-2; CPCs, chondrogenic progenitor cells; CRP, C-reactive protein; DAMPs, damage-associated molecular patterns; DCs, dendritic cells; ERK, extracellular signal-regulated kinase; GLUT, glucose transporter; HFD, high-fat diet; HLA-DR, human leukocyte antigen-antigen D related; HSC, hepatic stellate cell; ICAM, intercellular adhesion molecule; IFNγ, interferon γ; IL, interleukin; iNOS, inducible nitric oxide synthase; IPFP, intrapatellar fat pad; IRS, insulin receptor substrate; JAK, Janus kinase; JNK, c-jun N-terminal kinase; KC, kupffer cells; LEPR, leptin receptor; LPL, lipoprotein lipase; LPS, lipopolysaccharide; MAPK, mitogen-activated protein kinase; MCP-1, monocyte chemoattractant protein 1; MEK, mitogen-activated protein kinase kinase; MIP-1α, macrophage inflammatory protein-1 alpha; miRNA, microRNA; MMPs, matrix metalloproteinases; NAFLD, non-alcoholic liver disease; NASH, non-alcoholic steatohepatitis; NF-κB, nuclear factorκB; NO, nitric oxide; NOS2, nitric oxide synthase 2; OA, osteoarthritis; PAMPs, pathogen-associated molecular patterns; PGE2, prostaglandin E2; PI3K, phosphoinositide 3-kinase; PKC, protein kinase C; PMA, phorbol myristate acetate; PMNs, human polymorphonuclear neutrophils; PPARα, peroxisome proliferator-activated receptor alpha; RA, rheumatoid arthritis; RORγt, retinoic acid-related orphan receptor gamma t; ROS, reactive oxygen species; SLE, systemic lupus erythematosus; SREBP-1c, sterol regulatory element-binding protein-1c; STAT, signal transducer and activator of transcription; T2DM, type 2 diabetes mellitus; TGF-β, transforming growth factor; Th, T helper cells; TLR, toll-like receptor; TNF-α, tumour necrosis factor-α; Treg, T regulatory cells; VAT, visceral adipose tissue; VCAM, vascular cell adhesion molecule; WAT, white adipose tissue.

Obesity, the greater public health problem in the western world, is associated with high-incident chronic autoimmune and inflammatory pathologies, such as T2DM, NAFLD, OA, and RA, thus having a huge social and economic impact (Zhang et al., 2014). Adipose tissue, initially considered as a simple energy storage tissue, is now recognized as an active endocrine organ and a bona fide immune organ, constituted not only by adipocytes but also by fibroblasts, endothelial cells and a wide array of immune cells (adipose tissue macrophages, neutrophils, mast cells, eosinophils, T and B cells that maintains tissue homeostasis in lean individuals (Huh et al., 2014; Vieira-Potter, 2014). The adipocyte expansion caused by positive energy balance leads to adipocyte hypoxia, apoptosis, and cell stress, ultimately resulting in the expression of chemoattractant molecules and infiltration of inflammatory cells (Vieira-Potter, 2014). The obese adipose tissue is also characterized by a markedly deregulated production of adipose tissue-derived factors, i.e., adipokines, a growing family of low molecular weight, biologically active proteins with pleiotropic functions (Al-Suhaimi and Shehzad, 2013). Adipokines are crucial players not only in energy metabolism but also in inflammation and immunity, most of them being increased in obesity and contributing to the associated 'low-grade inflammatory state' (Tilg and Moschen, 2006).

Leptin was discovered in 1994 by the group of Jeffrey Friedman (Zhang et al., 1994) and is the best-characterized member of adipokine family. Encoded by LEP gene (the human homolog of murine ob gene), leptin is a 16 kDa non-glycosylated protein mainly produced by adipocytes, but also by skeletal muscle, intestine, brain, joint tissues and bone (Scotece et al., 2014). This adipokine exerts its physiological activity through its receptor (LEPR or Ob-R), a class I cytokine receptor family from diabetes (db) gene (Münzberg and Morrison, 2015). There are at least six LEPR isoforms that differ in the length of the cytoplasmic domain: a soluble isoform, four short isoforms, and a long isoform, which has the full intracellular domain that allows the transduction of leptin signal via JAK and STAT signaling pathways (Frühbeck, 2006). Alternatively to canonical JAK/STAT pathway, LEPR could activate ERK 1/2, p38 MAPK, JNK, PKC, and PI3K/Akt pathways (Zhou and Rui, 2014) (**Figure 1**). This hormone, together with other regulatory molecules, has a central role in appetite and body weight homeostasis by inducing anorexigenic factors (as cocaine-amphetaminerelated transcript) and suppressing orexigenic neuropeptides (as neuropeptide Y) on hypothalamus (Al-Suhaimi and Shehzad, 2013; Rosenbaum and Leibel, 2014). Therefore, central leptin resistance, caused by impairment of leptin transportation, leptin signaling and leptin target neural circuits, is considered the main risk factor for the obesity pathogenesis.(Al-Suhaimi and Shehzad, 2013; Rosenbaum and Leibel, 2014). Interestingly, leptin release is modulated in a circadian rhythm manner, which has been correlated with sweet taste recognition (Nakamura et al., 2008). Moreover, leptin also affects other physiological functions, namely bone metabolism, inflammation, infection and immune responses (Scotece et al., 2014) (**Figure 2**). Accordingly, LEPR is expressed in across the cells of innate and adaptive immune system, evoking leptin as a crucial linker of neuroendocrine and immune systems (Carlton et al., 2012; Procaccini et al., 2017).

This review summarizes the latest data regarding the role of leptin as a mediator of innate and adaptive immune cells activity, and its effects on obesity-associated metabolic disorders, namely T2DM and NAFLD, and autoimmune and/or inflammatory rheumatic diseases, such as OA and RA.

# LEPTIN AND IMMUNOMETABOLISM

The rising prevalence of obesity in western society is paralleled with a significant augment in autoimmune diseases. Accordingly, numerous association studies had demonstrated that overweight is implicated in a higher risk of developing multiple sclerosis (Kavak et al., 2015), RA (Ajeganova et al., 2013), and psoriasis (Duarte et al., 2013). On the other hand, malnutrition/starvation has been long related to increased susceptibility to infectious diseases (Taylor et al., 2013; Jones et al., 2014). These observations bring out immune response as a highly energy-dependent biological process that is dependent on an adequate food intake and metabolism. In fact, a recent article critically discuss the correlation between autoimmunity and overnutrition or metabolic pressure (De Rosa et al., 2017). At the interface of the historically distinct fields of immunology and metabolism, immunometabolism has emerged as a new research discipline (Mathis and Shoelson, 2011). In the last years, it has been evidenced that metabolic status of immune cells directly determines their function and differentiation, thus affecting immunity and tolerance, as well as the failure of the immune response in autoimmune pathologies (Gaber et al., 2017). In fact, innate and adaptive immune cells adapt to altered tissue microenvironment, characterized by hypoxia and nutrient competition, by reprogramming their metabolism (Gaber et al., 2017), and failure in this metabolic reconfiguration ultimately leads to a deregulated immune response and pathology (Gaber et al., 2017).

Leptin, the forerunner of adipokine family, is a key sensor of energy metabolism and a cornerstone in the regulation of metabolism-immune system interplay. Malnutrition results in hypoleptinemia, while obesity leads to hyperleptinemia, both conditions affecting the immune response in an opposite manner. In particular, obese subjects demonstrated decreased levels of Treg (central players in the control of peripheral immune tolerance), which are inversely correlated with leptin levels and BMI (Matarese et al., 2010). In malnutrition, altered T cell function and metabolism was associated with decreased leptin levels (Cohen et al., 2017). Leptin and LepR-deficient mouse models presented augmented number and activity of Treg cells together with a resistance to autoimmune diseases, and leptin replacement rescues Treg cell levels to wild-type mice values (Matarese et al., 2010). Accordingly, human T cell activation and production of cytokines can be induced after incubation with 10 ng/mL exogenous leptin following nutritional rehabilitation (Rodríguez et al., 2007). Furthermore, the central effect of leptin on the hypothalamus is mediated, at least in part, by inhibition of hypothalamic-pituitary-adrenal axis and activation of the

Additionally, leptin induces the activation of SHP2, which then recruits the adaptor protein Grb2 to prompt activation of Ras/Raf/MAPK signaling cascade. Leptin also mediated phosphatidylinositol-3-kinase (PI3K)/Akt activation via insulin receptor substrate 1/2 (IRS1/2) and protein tyrosine phosphatase 1B (PTP1B) acts as a negative regulator of leptin signaling through JAK2 dephosphorylation.

sympatho-adrenal axis, having the sympathetic nervous system a function in the central control of the immune system (Pérez-Pérez et al., 2017). Moreover, most immune cells express LEPR at their surface, which evokes a straight action of leptin in the modulation of the immune response (Procaccini et al., 2017).

# LEPTIN AND INNATE IMMUNITY (FIGURE 3)

# Granulocytes (Neutrophils, Eosinophils, and Basophils)

Human polymorphonuclear neutrophils express LEPR (Caldefie-Chezet et al., 2001), but only the short-form (Ob-Ra) was been detected (Zarkesh-Esfahani et al., 2004). Although the short-form of LEPR lacks most of the intracellular domain of the receptor, it is enough to signal through MAPK pathways, enhancing CD 11b expression and preventing apoptosis, but not through JAK-STAT pathways as long-form of LEPR (Bjorbaek et al., 1997; Zarkesh-Esfahani et al., 2004). Leptin is likely to act as a survival cytokine for neutrophils. At 500 nM, leptin delayed the cleavage of Bid and Bax, mitochondrial release of cytochrome C and second mitochondria-derived activator of caspase, as well as the activation of caspase-3 and caspase-8 (Bruno et al., 2005). PI3K, NF-κB, and MAPK pathways were involved in the anti-apoptotic activity of leptin in human neutrophilsin vitro (Bruno et al., 2005; Sun et al., 2013). Additionally, leptin (250 ng/ml) stimulated the release of oxygen radicals, such as superoxide anion and hydrogen peroxide, by PMNs (Caldefie-Chezet et al., 2001, 2003).

There is strong evidence for an effect of leptin on neutrophil chemotaxis and infiltration. Leptin (50 ng/ml) mediated the migration of human neutrophils in vitro, through activation of p38 MAPK and Src kinases (Montecucco et al., 2006), and by indirect mechanisms via TNF-α released by monocytes (Caldefie-Chezet et al., 2001; Zarkesh-Esfahani et al., 2004), having no secretagogue properties (no detectable [Ca2+]i mobilization, oxidant production, or β2-integrin upregulation) (Montecucco et al., 2006). Otherwise, leptin inhibits neutrophil chemotaxis to classical chemoattractants, like interleukin (IL)-8 (Montecucco et al., 2006). Murine neutrophils with Q223R LEPR mutation had reduced chemotaxis toward leptin (Naylor et al., 2014), while

regulation, via activation of leptin receptor (LEPR).

neutrophils from human volunteers and wild-type C57BL/6 mice migrated toward leptin in a dose-dependent manner, requiring JAK2/PI3K signaling (Ubags et al., 2014). Nevertheless, one study found that physiological concentrations of leptin (1–100 ng/ml) do not affect human neutrophils, and high leptin concentrations induced survival and changes in neutrophils proteome, but no effect on chemotaxis was observed (Kamp et al., 2013). In vivo studies clarified the effect of leptin in neutrophils. It was observed that neutrophil populations were enhanced in rats with high-fat-diet induced obesity, compared with control diet rats (do Carmo et al., 2013), and neutrophils from obese subjects displayed elevated superoxide release and chemotactic activity (Brotfain et al., 2015). Furthermore, leptin administration (50 µg) increased pulmonary neutrophilia in Escherichia coli pneumonia murine model as well as in healthy mice (Ubags et al., 2014).

Alike neutrophils, both human eosinophils and basophils expressed LEPR on the cell surface (Bruno et al., 2005; Suzukawa et al., 2011). In eosinophils, leptin (50 ng/ml) enhanced the release of pro-inflammatory cytokines IL-1β and IL-6, and chemokines IL-8, growth-related oncogene-α and MCP-1 (Wong et al., 2007). It also modulated the surface expression of adhesion molecules; in particular, up-regulates ICAM-1 and CD18, and suppress ICAM-3 and L-selectin (Wong et al., 2007). Treatment of human eosinophils with recombinant leptin in vitro delayed apoptosis via JAK, NF-κB, and p38 MAPK signaling pathways, suggesting leptin as a survival cytokine (Wong et al., 2007), similar to neutrophils (Bruno et al., 2005). Furthermore, leptin also stimulated chemokinesis (Wong et al., 2007) and enhanced chemotactic migration of eosinophils isolated from human peripheral blood, in a dose-dependent manner, however, the underlying mechanisms remain unclear (Kato et al., 2011). In obese individuals, eosinophils demonstrated greater adhesion and chemotaxis toward eotaxin and RANTES (CCL5), compared with non-obese healthy volunteers (Grotta et al., 2013).

In human basophils, leptin treatment (10 nM) induced a strong migratory response, promoted the secretion of type 2 cytokines IL-4 and IL-13, and up-regulated the cell surface expression of CD63, which may have an exacerbating action on allergic inflammation (Suzukawa et al., 2011). Moreover, leptin is a survival-enhancing factor of human basophils, as aforementioned for eosinophils and neutrophils. Although leptin was a weak effect on direct induction of basophil degranulation, it potently primed basophils for enhanced degranulation in response to aggregation of IgE or its high-affinity receptor FcεRI (Suzukawa et al., 2011).

Altogether, these findings suggest leptin as a potent activator of neutrophils, eosinophils, and basophils through its positive action in cell survival, cytokines release and chemotaxis.

# Monocytes and Macrophages

activates B cells to secrete cytokines and modulates B cell development.

Both isoforms of LEPR are expressed in PBMCs, being lower in cells from obese individuals compared with lean subjects (Tsiotra et al., 2000). Functional LEPR was also expressed in macrophages (O'Rourke et al., 2001). The effect of leptin on monocytes and macrophages has been well-established since its first evidence in Santos-Alvarez et al. (1999). Leptin promoted the proliferation of human circulating monocytes in vitro as well as its activation through induction of TNF-α and IL-6 production, and stimulation of surface markers, namely CD25, HLA-DR, CD38, CD71, CD11b, CD11c, and CD16 (Santos-Alvarez et al., 1999; Cannon et al., 2014). Moreover, leptin potentiated the stimulatory effect of LPS or PMA on human monocytes (Santos-Alvarez et al., 1999), and increased CCLs in cultured murine macrophages, being JAK2-STAT3 signaling pathway involved (Kiguchi et al., 2009). Leptin (625 nM) also augmented the production of several inflammatory mediators in monocytes/macrophages, such as interleukin 1 receptor antagonist (IL-1Ra) (Gabay et al., 2001), interferon-c-inducible protein (Meier et al., 2003), leukotrienes (Mancuso, 2004), nitric oxide (Dixit et al., 2003), and pro-inflammatory cytokines, namely TNF-α, IL-6, IL-1β, and resistin (Tsiotra et al., 2013; Scotece et al., 2014; Inzaugarat et al., 2017). By contrast, it was reported that 1 µg/ml leptin had no effect on IL-1β secretion but enhanced IL-18 in the THP-1 murine monocytic cell line. These apparent discrepancies could be species-specific (human vs. murine cells) and/or leptin treatment-dependent (1 µg/ml for 24 h vs. 1 µg/ml for 3 h). Additionally, recombinant leptin increased the expression of TLR2, but not TLR4, in human monocytes (Jaedicke et al., 2013).

A dose-dependent effect of leptin as a trophic factor to prevent apoptosis was found in serum-deprived human monocytes, being this effect mediated by the p42/p44 MAPK pathways (Najib and Sánchez-Margalet, 2002). Leptin (2 nM) stimulated the oxidative burst in monocytes (Sánchez-Pozo et al., 2003), and increased LPL expression through oxidative stress- and PKC-dependent pathways (Maingrette and Renier, 2003). Moreover, leptin promoted the phagocytosis of apoptotic cells by macrophages from lupus mice, via modulation of cAMP levels (Amarilyo et al., 2014). Leptin also promoted a

defensive environment against Leishmania donovani infection by induction of macrophage phagocytic activity and intracellular ROS generation (Dayakar et al., 2016). Accordingly, macrophages from knocked-out LepR Tyr 985 mice presented reduced phagocytosis and killing activity of Klebsiella pneumoniae that was associated with diminished ROS production (Mancuso et al., 2012). It has been demonstrated that leptin-mediated protein radical formation, tyrosine nitration and activation of KCs are caused by peroxynitrite formation, which exacerbated NASH in diet-induced obese mice (Chatterjee et al., 2013). Concerning to chemoattractive activity, it was verified that leptin induced in vitro chemotactic responses for monocytes and macrophages (Curat et al., 2004; Gruen et al., 2007), via intracellular calcium influx, JAK/STAT, MAPK and PI3K pathways (Gruen et al., 2007). However, it has been reported that hematopoietic LEPR deficiency in mice did not change macrophage accumulation in WAT after diet-induced obesity versus wild-type mice (Gutierrez and Hasty, 2012). Likely, compensatory in vivo effects of other cytokines (like IL-1 or TNF-α) present in WAT could occur in obese individuals and which are absent in in vitro assays.

Leptin treatment (50 ng/ml) of human macrophages in culture, induced 'alternatively activated' or M2-phenotype surface markers, but they were able to secrete M1-typical cytokines (TNF-α, IL-6, IL-1β, IL-1ra, IL-10, MCP-1, and macrophage inflammatory protein 1-alpha (MIP-1α)), suggesting a role for leptin in the phenotype of macrophages found in adipose tissue (Acedo et al., 2013).In macrophages, leptin also triggered catecholamine-dependent increases in cAMP-histone deacetylase 4 signaling pathway, that reduced inflammation in adipose tissue (Luan et al., 2014). Additionally, leptin increased the expression of LEPR in M2 macrophages and stimulated IL-8 expression via p38 and ERK signaling pathways (Cao et al., 2016). In tumor-associated macrophages, leptin induced the expression of IL-18 via NF-kB, possible contributing to tumor progression (Li K. et al., 2016).

Macrophages are indirectly regulated by leptin through mast cells (Zhou et al., 2015). In particular, leptin expression was reduced in both human and mouse mast cells from lean adipose tissue compared with obese individuals. Leptin deficiency led to the anti-inflammatory activity of mast cells and, consequently, to a shift in macrophage polarization from M1 to M2; in vitro co-cultures of mast cells with BMDM increased IL-4-mediated arginase-1 and IL-10 expression, and suppressed LPS-mediated iNOS and IL-6 expression (Zhou et al., 2015). Furthermore, reduction of mast cells in leptin-deficient ob/ob mice exacerbated obesity and diabetes, indicating an important role of mast cells in obesity-related inflammation through its reactivity to leptin levels (Zhou et al., 2015).

Biologic drugs used for the treatment of psoriatic arthritis, namely adalimumab (an anti-TNF-α monoclonal antibody) and ustekinumab (a monoclonal antibody against the p40 subunit of IL-12 and IL-23), augmented LEPR expression in THP-1 human macrophages (Voloshyna et al., 2016). However, only ustekinumab was able to increase the expression of leptin, suggesting a novel mechanism for this biological drug. Further mechanistic studies focused on the leptin pathway could have potential therapeutic action in common obesity-related complications of psoriasis (Voloshyna et al., 2016). The establishment of LepR-deficient macrophage cell line DB-1, derived from differentiated bone marrow cells of Lepr-knockout mice, provide a powerful tool to study the role of leptin and its receptor in obesity-associated inflammation and immune system deregulation (Dib et al., 2016).

# NK Cells

The role of leptin in regulating NK cell development and activation was first verified in obese Lepr-deficient (db/db) mice, which showed decreased NK cell function (Tian et al., 2002). In this animal model, the population of NK cells in bone marrow was impaired through an increase in apoptotic rate, and recombinant leptin (200 ng/ml) significantly enhanced the survival of immature NK cells from wild-type mice via modulation of Bcl-2 and Bax gene expression (Lo et al., 2009). Furthermore, leptin administration (500 µg/kg) led to a higher activity of NK cells in lean animals (Nave et al., 2008). Consistently, human NK cells expressed functional long- and short-form of LEPR that influenced NK cell cytotoxicity through STAT3 activation and, consequently, transcription of genes encoding IL-2 and perforin (Zhao et al., 2003).

The above-mentioned results indicated that leptin signaling is required for normal NK cell immune function. However, there are some controversial findings concerning the time of leptin treatment in vitro. Short-term stimulation of human NK cells with leptin (50 nM) raised the secretion of IFNγ and cytotoxicity (Wrann et al., 2012; Laue et al., 2015). By contrast, long-term exposure to leptin decreased NK cell proliferation and immune function (Wrann et al., 2012). Obesity is partially characterized by a state of long-term, highly elevated leptin exposure, and NK cells from obese animals were significantly resistant to leptin stimulation (Nave et al., 2008), which could explain the functional desensitization of NK cells after long-term exposure. Accordingly, exposure of NK-92 human cell line to hyperleptinemia (similar to that observed in obese individuals) led to metabolic activation of NK-92 cells after 24 h, but there is a reduction of cell metabolism after 96 h (Lamas et al., 2013). Furthermore, obese individuals have lower NK function compared to lean individuals (Laue et al., 2015) and, after weight loss, the decrease of plasma leptin levels is accompanied by a restoration of IFNγ production by NK cell (Jahn et al., 2015; Bähr et al., 2017; Favreau et al., 2017).

Overall, leptin signaling seems to be necessary for normal NK cell immune function, increasing the immune activity and cell proliferation, and reducing the apoptotic rate of NK cells. Long-term exposure to hyperleptinemia, observed in obesity, has been associated with decreased NK immune activity possibly due to the development of leptin resistance. Further studies are needed to better understand the correlation between leptin levels and NK cell development and function, as well as the potential implications in obesity.

# Dendritic Cells

fphys-09-00640 May 30, 2018 Time: 18:27 # 7

Human DCs, both immature and mature DCs, present functional active LEPR with the capacity to signal STAT-3 phosphorylation (Mattioli et al., 2009). Leptin (10 nM) acted as an activator of human DCs, evidenced by up-regulation of IL-1β, IL-6, IL-12, TNF-α and MIP-1α production, improvement of immature DCs migration (Mattioli et al., 2008; Al-Hassi et al., 2013) and their chemotactic responsiveness, licensing them toward Th1 priming (Mattioli et al., 2008). Moreover, leptin treatment promoted DC survival through decreased apoptosis via activation of NF-κB and PI3K-Akt signaling pathways, with a parallel increase of bcl-2 and bcl-x<sup>L</sup> gene expression (Lam et al., 2006; Mattioli et al., 2009).

Lepr-deficient db/db mouse bone marrow culture displayed a reduced number of DCs, attributable to dysregulation of Bcl-2 genes and a consequent increase of apoptosis (Lam et al., 2006). Moreover, DCs from db/db mice possessed markedly reduced expression of co-stimulatory molecules and a Th 2-type cytokine profile, with a poor capacity to stimulate allogeneic T cell proliferation (Lam et al., 2006). Consistently, db/db DCs demonstrated down-regulation of PI3K/Akt and STAT-3 pathways (Lam et al., 2006). Lep-deficient ob/ob mice presented a reduced expression of DC maturation markers (CD40, CD80, and CD86), decreased production of inflammatory cytokines (IL-12, TNF-α, and IL-6), and augmented TGF-β production, but ob/ob mice-derived DCs were more efficient in inducing Treg or Th17 cells than wild-type animals (Moraes-Vieira et al., 2014). In DCs from ob/ob mice, leptin deficiency resulted in defective antigen presentation function toward Leishmania donovani, which was not reversed by leptin treatment (Maurya et al., 2016). Conversely, one report verified no changes in the phenotype, activation, antigen processing or presentation of DCs from leptin-knockout mice, but these cells showed an enhanced ability to activate T cells, suggesting that leptin may dampen T-cell responsiveness in the physiological context (Ramirez and Garza, 2014). Diet-induced obesity in mice fed with HFD results in an elevation of serum leptin levels and splenic CD11c<sup>+</sup> DCs, with diminished DC cell stimulatory capacity, being these effects distinct from that caused by HFD alone in obese-resistant mice (Boi et al., 2016).

Altogether, these data demonstrated the important role of leptin in DC activation, chemoattraction, and survival, with possible implications in DC maturation and migration. Given the ability of DCs to orchestrate immune response and promote potent immunogenic responses through activation of T cell immunity, DCs-based immunotherapies to elicit immunity against cancer and infectious diseases are currently being developed. In particular, DCs can be differentiated ex vivo, exposed to antigens and induced to mature in the presence of adjuvants. Then, the mature DCs are injected into the patient and migrate to the lymph nodes to present antigens to T cells. Thus, the modulation of DCs maturation and activity by leptin is of most importance considering a potential application of leptin in immunotherapeutic approaches and as novel adjuvant immunopotentiator in vaccination protocols employing ex vivo generated autologous DCs.

# LEPTIN AND ADAPTIVE IMMUNITY (FIGURE 3)

The role of leptin in adaptive immunity has first evidenced working with ob/ob and db/db mice, which showed thymus atrophy, T-cell lymphopenia, and impaired delayed-type hypersensibility (Lord et al., 1998; Howard et al., 1999; Matarese, 2000). Moreover, chronic leptin administration (1 µg/g body weight) reversed immunosuppressive status and thymic atrophy of ob/ob mice (Lord et al., 1998, Nature; Howard et al., 1999, J. Clin. Inv.). Since then, the role of leptin in T and B cell populations have been extensively studied.

# T Cells

T lymphocytes expressed the long form of LEPR (higher in peripheral CD4+ than in CD8+ T cells) (Lord et al., 1998; Kim et al., 2010), with signaling capacity to activate JAK-STAT pathway (Sanchez-Margalet and Martin-Romero, 2001). Consequently, leptin modulated cell proliferation, responsiveness, and polarization of T cells. Leptin dosedependently promoted the proliferation of human naïve (CD45RA+) CD4+ T cells, whereas it minimally affected memory (CD45RO+) CD4+ T cells proliferation (Lord et al., 1998, 2002). Additionally, morbidly obese children, who were congenitally deficient in leptin, presented a decreased number of circulating CD4+ T cells, as well as impaired T cell proliferation and cytokine release, which were reversed by administration of recombinant human leptin (Farooqi et al., 2002). Moreover, leptin inhibited autophagy in human CD4+CD25− conventional T cells via mTOR pathway (Cassano et al., 2014), which emerged as the potential link between immunity and nutritional status (Procaccini et al., 2012).

## T Helper Cells

Leptin also promoted CD4+ T cell polarization toward a Th1 response (which secretes IFNγ and IL-2) rather than Th2 response (which secretes IL-4) (Martín-Romero et al., 2000). Accordingly, under Th2-polarizing conditions, the in vitro leptin treatment decreased IL-4-producing T cells and inhibited T cell proliferation (Batra et al., 2010). However, it was recently reported that in vivo leptin-deficiency attenuated allergic airway inflammation and that high leptin levels associated with obesity promoted proliferation and survival of Th2 lymphocytes, as well as the production of type 2 cytokines, altogether contributing to allergic responses (Zheng et al., 2016). Besides that, leptin was involved in thymus morphology and functions (Lamas et al., 2016), particularly in thymocyte differentiation of double positive CD4+CD8+ T cells into single positive CD4+ T cells (Kim et al., 2010).

IL-17-producing Th cells (Th17) have a crucial role in the promotion and maintenance of inflammatory and autoimmune pathologies. Leptin was demonstrated to increase Th17 population and responsiveness in SLE, via retinoic acid-related

orphan receptor (ROR)γt (Yu et al., 2013; Fujita et al., 2014; Reis et al., 2015). In collagen-induced arthritis mouse model, articular injection of leptin (5 µg) increased the number of Th17 in the joint tissue, resulting in exacerbating joint inflammation, and consequently early onset of arthritis and increased disease severity (Deng et al., 2012). Leptin, in concentrations similar to that found in blood during pregnancy, promoted the differentiation of peripheral blood CD4+ cells to Th17 cells, but suppressed the formation of Treg cells in vitro (Orlova and Shirshev, 2014). CD4+ T cell-derived leptin, but not plasma leptin, were positively correlated with the percentage of Th17 cells or RORγt levels in chronic lymphocytic thyroiditis, an organ-specific autoimmune disease (Wang et al., 2013). Furthermore, Lepr-deficient CD4+ T cells verified a reduced capacity for Th17 differentiation, via down-regulation of STAT3 activation (Reis et al., 2015).

### T Regulatory Cells (Treg)

Leptin also regulated CD4+CD25+ Treg proliferation (Matarese et al., 2010). Treg lymphocytes play a critical role in controlling the inappropriate immune responses characteristic of autoimmune diseases and allergy. In humans, leptin negatively affected the proliferation of Foxp3+CD4+CD25+ Treg; in vitro leptin neutralization, during anti-CD3 and anti-CD28 stimulation, led to the proliferation of the isolated human Treg cell (De Rosa et al., 2007). Obese individuals presented a reduced number of CD4+CD25+CD127-Foxp3+ Treg cells, which was correlated with body weight, BMI, and plasma leptin levels (Wagner et al., 2013). Moreover, leptin-deficient mice presented an increased percentage of peripheral Treg, compared with wild-type mice, which is reversed after leptin administration (De Rosa et al., 2007). It was verified that leptin played an important role in Treg dysfunction in patients with pulmonary arterial hypertension (Huertas et al., 2016). Accordingly, Lepr-deficient rats developed less severe hypoxiainduced pulmonary hypertension and were protected against decreased Treg function after exposure to hypoxia (Huertas et al., 2016). In SLE, the disease-associated higher leptin serum levels were negatively correlated with disease severity and number of Treg cells (Ma et al., 2015; Margiotta et al., 2016; Wang et al., 2017), and fasting-induced hypoleptinaemia was related to Treg population recovery in lupus-prone mice (Liu et al., 2012). Leptin-deficient ob/ob mice and a mouse model of lupus with leptin deficiency demonstrated increased frequency of Tregs cells (Fujita et al., 2014; Lourenço et al., 2016). These data evidenced the potential of anti-leptin-based approaches for immune system-dysregulated pathologies associated with reduced Treg function, such as SLE, obesity, T2DM, and metabolic syndrome.

T cell metabolism is directly related to its function (MacIver et al., 2013); effector T cells, such as Th1 and Th17, demand a high glycolytic metabolism to fuel proliferation and function, while Treg cells require oxidative metabolism to fuel suppressive activity. Recently, leptin was found to directly promote T-cell glycolytic metabolism and consequently induce Th17 cell differentiation, being Treg cells unchanged, in a mouse model of experimental autoimmune encephalomyelitis (Gerriets et al., 2016). Leptin regulated glucose metabolism partly by upregulation of glucose transporter Glut1 (Saucillo et al., 2014). Moreover, fasting led to decreased ability of T cells to secrete IL-2 and IFNγ, and inability to up-regulate glucose uptake and glycolytic flux (Saucillo et al., 2014), while Treg expansion was increased (Liu et al., 2012); leptin administration (1 µg/g body weight) rescued peripheral T cell function and metabolism in fasted mice (Saucillo et al., 2014). Likely, fasting was extensively reported to be associated with immune deficiency and increased susceptibility to infection (Gerriets and MacIver, 2014). Thus, leptin seems to provide a key link between nutritional status and inflammatory T cell responses (Gerriets et al., 2016; De Rosa et al., 2017).

Altogether, these data revealed the ability of leptin to increase immune activity by modulation of T cell number and function. Leptin can promote proliferation of naive T cells, as well as Th1 and Th17 proliferation and cytokine production. Moreover, leptin decreases Treg cell proliferation. Considering the regulatory effects of leptin on Th17 and Treg populations, revoking leptin signaling might be a potential therapeutic approach for inflammation and autoimmunity.

# B Cells

In contrast to macrophages and T cells, little is known about the role of leptin in the B lymphocytes development and function. B cells expressed the long form of the LEPR, suggesting a direct effect of leptin on B cell function (Busso et al., 2002). Accordingly, db/db and ob/ob presented a reduced number of peripheral blood and bone marrow B lymphocytes, which was recovered after leptin treatment (Bennett et al., 1996; Claycombe et al., 2008). Conversely, db/db mice presented an increased absolute number of B cells in the peritoneal cavity (Jennbacken et al., 2013), and the increase of leptin was correlated with a decrease in B cells of mice with unbalanced diets (carbohydraterich and fat-rich) (Martínez-Carrillo et al., 2015). Thus, further investigation is needed to better clarify the role of leptin in lymphopoiesis. Leptin promoted B cell homeostasis through inhibition of apoptosis and induction of cell cycle entry via Bcl-2 and cyclin D1 activation (Lam et al., 2010). Furthermore, leptin dose-dependently activated human peripheral blood B cells, inducing the secretion of pro-inflammatory cytokines, namely TNF-α and IL-6, and the anti-inflammatory cytokine IL-10, via JAK-STAT and p38MAPK-ERK1/2 signaling pathways (Agrawal et al., 2011). Likewise, leptin (50 ng/ml) activated and induced the production of higher amounts of TNF-α, IL-6, and IL-10 by B cells from aged subjects compared to young individuals (Gupta et al., 2013), which is associated with leptin-mediated STAT3 phosphorylation (Gupta et al., 2013; Frasca et al., 2016).

Leptin can also modulate B cell development – decrease pro-B, pre-B and immature B cells and increase mature B cellsin bone marrow of fasted mice, characterized by low serum leptin levels (Tanaka et al., 2011). Leptin administration reversed the starvation-induced lymphopenia of bone marrow B cells, indicating an important role of central leptin in the immune system (Tanaka et al., 2011; Fujita et al., 2012). Moreover, leptin might regulate B cell activity in obesity (Nikolajczyk, 2010; Frasca et al., 2016). In particular, B cells were described to

accumulate in murine VAT and to critically regulate T2DMassociated inflammation through activation of CD8+ and Th1 cells and release of pathogenic antibodies (Winer et al., 2011; DeFuria et al., 2013).

In summary, leptin can increase B cell population by augmenting proliferation and reducing the apoptotic rate, activate B cell to secrete pro-, anti- and regulatory cytokines, and also modulate B cell development.

# LEPTIN AND IMMUNE-METABOLIC PATHOLOGIES

# Leptin and Obesity-Associated Metabolic Disorders

Obesity is associated with life-threatening co-morbidities, including insulin resistance, T2DM, NAFLD and steatohepatitis (NASH) (Lebovitz, 2003; Kamada et al., 2008; Klöting and Blüher, 2014; Fasshauer and Blüher, 2015). Adipokines, in particular leptin, mediate the crosstalk between adipose tissue and metabolic organs (especially liver, muscle, pancreas and central nervous system) (Cao, 2014). Thus, leptin has emerged as a significant pathological component in the development of metabolic disorders (DePaoli, 2014).

### Type 2 Diabetes Mellitus

T2DM is the most significant obesity-associated metabolic disorder and their prevalence is increasing worldwide in parallel (Bhupathiraju and Hu, 2016). Leptin has been proposed as a therapeutic target of T2DM, for its impact on food intake and body weight as well as its potential to improve insulin action (Kalra, 2009). Interestingly, leptin-deficient mice (Pelleymounter et al., 1995) and human (Farooqi et al., 1999, 2002; Ozata et al., 1999) have diabetic features, which were reversed with leptin replacement. The anti-diabetic effect of leptin is mediated by activation of IRS-PI3K pathway that improved insulin sensitivity in peripheral tissues (Morton et al., 2005). Activation of JAK2/IRS/PI3K/Akt signaling pathway by leptin and insulin triggers the translocation of glucose transporter type 4 (GLUT4) from cytosol to cell surface, and glucose uptake (Benomar et al., 2006; Zhao and Keating, 2007). Moreover, in the liver, leptin deficiencies decrease AMPK activity (Namkoong et al., 2005), which is also involved in glucose homeostasis regulation (Schultze et al., 2012). Leptin has also been implicated in the regulation of insulin secretion by pancreatic β-cells (Kulkarni et al., 1997), as well as in peripheral insulin resistance (Silha et al., 2003; Yadav et al., 2011). However, clinical trials to evaluate the potential of leptin monotherapy in obese humans with T2DM failed to demonstrate therapeutic activity acutely or chronically, with no observation of important weight loss or metabolic improvements (insulin sensitization, amelioration of glucose and lipid metabolism) (Mittendorfer et al., 2011; Moon et al., 2011; Wolsk et al., 2011). In this context, unresponsiveness to leptin – leptin resistance, caused by hyperleptinemia observed in obese humans, should be considered (Frederich et al., 1995). Further understanding of leptin resistance mechanisms could enable new leptin targeted therapies for obesity and diabetes in specific subsets of patients. In fact, leptin therapy improved the diabetic condition in children (Farooqi et al., 1999, 2002) and adults (Licinio et al., 2004) with familial leptin deficiency, and in lipoatrophic diabetes (Oral et al., 2002).

Importantly, T2DM is associated with altered components of immune system, including modified levels of specific chemokines and cytokines, changed number and activation status of leukocytes populations and increased apoptosis and tissue fibrosis (Donath, 2014), all promoted by obesity-associated inflammation in adipose tissue (Donath and Shoelson, 2011). Given the modulator action of leptin in innate and adaptive immune system (deeply described above), it is rational to see leptin as a linker of T2DM development, not only with metabolism but also with inflammation. Indeed, in patients with newly diagnosed T2DM, leptin levels were correlated with CRP, an inflammatory marker broadly evaluated for its association with risk factors for T2DM pathology (Morteza et al., 2013).

### Non-alcoholic Fatty Liver Disease

NAFLD, the major cause of chronic liver illness in developed countries, comprises a wide group of pathologies primary caused by a buildup of fat in the liver that spans from simple steatosis to NASH, liver fibrosis, cirrhosis, and hepatocellular carcinoma (Tiniakos et al., 2010). Given that NAFLD is increasing worldwide and it is associated with high-incident extra-hepatic complications such as obesity, T2DM, cardiovascular diseases and chronic kidney disease (Byrne and Targher, 2015; Polyzos et al., 2015), great efforts have been made in the last years to unravel the mechanisms underlying the disease pathophysiology and further development of effective NAFLD therapies.

Hepatic inflammation and hepatocyte injury and death are hallmarks of NAFLD/NASH. Fat overload by hepatocytes causes lipotoxicity and the release of DAMPs which activated Kupfer cells (KC; specialized liver macrophages) and HSC promoting inflammation and fibrosis, respectively. KC activation plays a central role in NAFLD pathophysiology through the production of pro-inflammatory cytokines and chemokines such as TNF-α, IL-1β, IL-6, CCL2 and CCL5, that contributed to leukocyte infiltration and inflammatory necrosis of hepatocytes, and fibrogenesis (Arrese et al., 2016). Dysbiosis of the gut microbiota may also conduct to KC activation through PAMPs, which originate in the gut and reach liver via portal circulation due to altered intestinal permeability. Other immune cells have been implicated in the NAFLD pathophysiology, although its role is less clear (Arrese et al., 2016). NK cells were impaired in experimental NASH, while natural killer T cells (unique immune cell subtype that expresses NK cells surface markers as well as T-cell antigen receptor) are depleted in steatosis but increased later during disease progression likely leading to inflammation and fibrosis in NASH via the production of IL-4, osteopontin, and IFN-γ (Tajiri and Shimizu, 2012; Tian et al., 2013). Neutrophils exacerbate the ongoing inflammation through macrophage recruitment and cell damage via the release of myeloperoxidase, ROS, and elastase (Xu R. et al., 2014). The role of DCs in NASH is complex and somehow controversial. DCs rapidly infiltrate into the liver in experimental NASH exhibiting an activated immune phenotype, but its depletion exacerbates

hepatic inflammation (Tacke and Yoneyama, 2013; Arrese et al., 2016). B- and T-cells also contributed to hepatic inflammation via secretion of pro-inflammatory cytokines that stimulated KC activation (Arrese et al., 2016).

Considering the strong metabolic and inflammatory components of NAFLD, leptin is regarded as a key regulator of NAFLD physiopathology (Polyzos et al., 2015). Leptin seems to feature a dual activity in NAFLD experimental models by exerting an early protective anti-steatosis effect in the initial stages of the disease, and a late pro-inflammatory and pro-fibrogenic action, when the disease persists or progress (Polyzos et al., 2015). In leptin-resistant Zucker fa/fa diabetic fatty rats, the expression of SREBP-1c (master regulator of glucose metabolism, and lipid and fatty acid production) is increased in liver (Kakuma et al., 2000), and infusion of adenovirus-leptin decreased hepatic triglyceride synthesis and β-oxidation via SREBP-1c down-regulation and PPARα up-regulation (Lee et al., 2002), thus preventing hepatic lipid accumulation.

Through anti-steatotic effect, leptin can ultimately lead to hepatic detrimental effects. Leptin activated HSCs, leading to up-regulation of pro-inflammatory and pro-angiogenic factors expression (like angiopoietin-1 and vascular endothelial growth factor), as well as collagen α1 and tissue inhibitor of metalloproteinase-1, ultimately acting as hepatic fibrogenesis inducer (Polyzos et al., 2015). Activated HSCs were able to secrete leptin, thus establishing a vicious cycle that further promotes liver fibrosis (Polyzos et al., 2015). Moreover, leptin was reported as potent HSCs mitogen and to prevent HSCs apoptosis, hence promoting the pathogenesis of hepatic fibrosis. Leptin (200 nM) increased the expression of TGF-β1 in KCs and sinusoidal endothelial cells, and connective tissue growth factor in KCs (Ikejima et al., 2002), being KCs-HSCs cross-talk proposed for liver fibrosis (Wang et al., 2009). Additionally, leptin is involved in macrophage-mediated KCs activation via induction of oxidative stress in macrophages (Chatterjee et al., 2013). Compared to healthy subjects, NAFLD patients demonstrated an increased leptin-stimulated TNFα and ROS production in peripheral monocytes, as well as IFNγ production in circulating CD4+ cells (a marker of Th1 differentiation) (Inzaugarat et al., 2017). Altogether, these data elucidated the role of leptin in NAFLD by modulation of HSCs, KCs, and inflammatory cells response.

In ob/ob mice, the congenital absence of leptin abrogated the development of CCl4-induced hepatic fibrosis comparing to lean littermates, which is reverted by leptin treatment (100 ng/ml) (Saxena et al., 2002). Likewise, xenobiotics- or thioacetamideinduced hepatic fibrosis was prevented in Zucker fa/fa rats, being involved the activation of HSCs and expression of procollagen-I and TGF-β1 (Ikejima et al., 2005). In humans, the role of leptin is controversial. Although leptin serum levels were initially related with hepatic steatosis but not with necroinflammation or fibrosis (Chitturi et al., 2002), later studies failed to demonstrate any significant association (Tsochatzis et al., 2009). Recombinant leptin has been successfully used in the treatment of insulin resistance and hepatic steatosis in patients with lipodystrophy and NASH (Oral et al., 2002; Petersen et al., 2002; Javor et al., 2005). However, large-scale and well-designed prospective cohort studies are necessary to deeply elucidate the role of leptin in hepatic lipid handling, inflammation, and fibrosis along with the identification of NAFLD patients subsets that may benefit from therapies directed to leptin system.

# Leptin in Rheumatic Diseases

Leptin has been described as a key factor in the pathophysiology of rheumatic diseases due to its capability to modulate bone and cartilage metabolism and to influence innate and adaptive immune responses (**Figure 4**).

### Osteoarthritis

Osteoarthritis, the most common joint disease, is a painful and debilitating illness characterized by progressive degeneration of articular joints. Initially seen as simply "wear and tear" disease, OA is currently considered a complex and multifactorial pathology triggered by inflammatory and metabolic imbalances that affect the entire joint structure (articular cartilage, meniscus, ligaments, bone, and synovium) (Loeser et al., 2012). Leptin levels are increased in serum, infrapatellar fat pad (IPFP), synovial tissues, and cartilage of OA patients compared to healthy individuals (Dumond et al., 2003; de Boer et al., 2012; Conde et al., 2013). Accordingly, leptin-deficient or LepR-deficient mice developed extreme obese phenotype without increased incidence of knee OA, suggesting that leptin signaling is essential to the development and progression of obesity-associated OA (Griffin et al., 2009). Furthermore, long form of LEPR was found to be expressed in human cartilage cells – chondrocytes (Figenschau et al., 2001).

Some initial findings suggested an anabolic role of leptin in cartilage. In particular, exogenous leptin administration (30 µg) stimulated proteoglycan and growth factors (insulin-like growth factor-1 and TGF-β) synthesis in rat knee-joint cartilage (Dumond et al., 2003). However, most of the studies reported a catabolic role of leptin underlying OA pathogenesis. A recent study determining the gene expression profile of leptin-induced articular rat cartilage by microarray analysis, associated the upregulation of matrix metalloproteinases (MMPs), inflammatory factors, growth factors and osteogenic genes with leptin-induced OA phenotype (Fan et al., 2018). Our group demonstrated that leptin (400 or 800 nM), in synergy with IFNγ or IL-1β, activated type 2 nitric oxide synthase (NOS2) via JAK2, PI3K and MAPK (MEK1 and p38) pathways, in cultured human and murine chondrocytes (Otero et al., 2003, 2005, 2007). Nitric oxide (NO), is a well-known pro-inflammatory mediator which lead to joint degradation through induction of chondrocyte phenotype loss, apoptosis and metalloproteinase (MMP) activity (Rahmati et al., 2016). Leptin (800 nM), alone or in combination with IL-1β, also induced the expression of COX-2 and the production of PGE2, IL-6, and IL-8 in cartilage explants of OA patients and human primary chondrocytes (Vuolteenaho et al., 2009; Gomez et al., 2011), revealing that leptin contributed to the pro-inflammatory environment of OA cartilage. It was also demonstrated that leptin (500 ng/ml) enhanced IL-6 production, mediated by chondrocyte-synovial fibroblast cross-talk, in OA patients (Pearson et al., 2017). Moreover, leptin modulated the production of inflammatory mediators by immune cells. In

particular, the production of IL-6, IL-8, and CCL3 were increased by leptin in CD4+ T cells from OA patients, but not from healthy subjects (Scotece et al., 2017); thus, demonstrating new insights into the role of leptin in the immune system and OA pathophysiology.

Leptin can directly induce the expression of MMPs that are involved in OA-related joint destruction, like MMP-1 (also known as interstitial collagenase), MMP-3 (also known as stromelysin), and MMP-13 (also known as collagenase), via NFκB, PKC, and MAPK pathways (Bao et al., 2010; Koskinen et al., 2011; Hui et al., 2012). MMP-2 (72 kDa type IV collagenase), MMP-9, and disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS) 4 and ADAMTS5, were also increased by leptin, while fibroblast growth factor 2 and proteoglycan were down-regulated (Bao et al., 2010; Conde et al., 2011b). Leptin (800 nM) can perpetuate the cartilagedegradation processes due to induction of VCAM-1, an adhesion molecule responsible for leukocyte and monocyte chemotaxis and infiltration to inflamed joints, via JAK2 and PI3K pathways in chondrocytes (Conde et al., 2012; Vestweber, 2015). SOCS-3 was pointed as a regulator of leptin-induced expression of MMP-1, -3, and -13, and pro-inflammatory mediators IL-6, NO and COX-2 (Koskinen-kolasa et al., 2016). Additionally, leptin increased the production of other adipokines, namely lipocalin-2, by human cultured chondrocytes (Conde et al., 2011a).

MicroRNAs, small single-stranded non-coding segments of RNA, are increasingly recognized as regulatory molecules involved in disease processes, including OA, inflammation, and obesity (Marques-Rocha et al., 2015; Deiuliis, 2016; Nugent, 2016). miR-27 was found to be decreased in OA chondrocytes and to directly targeted the 3<sup>0</sup> -untranslated region of leptin (Zhou et al., 2017). Furthermore, the injection of OA rats with miR-27 lentiviral overexpression vector resulted in decreased levels of IL-6 and -8, as well as MMP-9 and -13, thus indicating the protective action of miR-27 in OA, possibly by targeting leptin.

Chondrogenic progenitor cells as cartilage seed cells are crucial to maintain cartilage homeostasis and replace damaged tissue (Seol et al., 2012). Leptin (50 ng/ml) can reduce CPCs migratory ability and their chondrogenic potential, and augment CPCs osteogenic transformation, hence changing CPC differentiation fate (Zhao et al., 2016). CPC cell cycle arrest and senescence are also induced by leptin (Zhao et al., 2016). Furthermore, leptin influenced the regulation of bone metabolism through induction of abnormal osteoblast function, which is associated to joint destruction in OA patients (Findlay and Atkins, 2014; Conde et al., 2015). The augmented production

of leptin by OA subchondral osteoblasts is related with in vitro elevated levels of alkaline phosphatase, osteocalcin, collagen type 1, and TGF-β1, all being responsible for dysregulated osteoblast function (Mutabaruka et al., 2010). Additionally, bone morphogenic protein (BMP)-2 is increased in leptin-stimulated human primary chondrocytes (Chang et al., 2015). Leptin also suppressed bone formation in vivo (Ducy et al., 2000), but there are some discrepancies with in vitro results.

Taking together, this evidence indicated a key role of leptin in OA pathophysiology by influencing pro-inflammatory status, cartilage catabolic activity, as well as cartilage and bone remodeling. However, studies in a large cohort of patients are needed to better clarify the leptin significance in the development and progression of OA.

#### Rheumatoid Arthritis

Rheumatoid arthritis is a chronic inflammatory joint illness characterized by synovial membrane inflammation and hyperplasia ("swelling"), production of autoantibodies, namely rheumatoid factor and anti-citrullinated protein antibody – autoimmune disease, destruction of cartilage and bone ("deformity"), and systemic features including skeletal, cardiovascular, pulmonary, and psychological complications (McInnes, 2011; Smolen et al., 2016). Evidencing the crucial role of immune system in RA pathology, RA-associated synovitis comprises both innate immune cells (like monocytes, DCs, and mast cells) and adaptive immune cells (like Th1, Th17, and B cells) (McInnes, 2011; Smolen et al., 2016). As described above, leptin modulates neutrophils chemotaxis, activates proliferation and phagocytosis of monocytes and/or macrophages, regulates NK cytotoxicity, induces proliferation of naïve T cells, promotes Th1 cell immune response and down-regulates Th2 cell immune response. Moreover, leptin modulates the activity of Treg cells, which are potent inhibitors of autoimmunity, thus having a potential implication in RA pathophysiology (Toussirot et al., 2015).

Several studies have found a positive correlation between serum and synovial leptin levels and RA pathology (Otero et al., 2006; Targonska-Ste˛pniak et al., 2008 ´ ; Yoshino et al., 2011; Olama et al., 2012), but there are controversial results (Anders et al., 1999; Popa et al., 2005; Hizmetli et al., 2007; Oner et al., 2015). Differing results may be due to the relatively small sample size, inconsistency of the baseline characteristics of participants (age, race, disease duration, BMI, ...), co-existence of other auto-immune diseases, employment of different methods to measure leptin levels in RA patients, or underlying patients treatments that intervene with the endocrine system. The present consensus is that leptin levels are elevated in RA patients, and serum and synovial fluid levels of leptin were associated with disease duration and parameters of RA activity (Olama et al., 2012; Lee and Bae, 2016), although large cohorts studies are necessary. Experimental animal models of arthritis had demonstrated the leptin action in joint inflammation. In particular, compared with control mice, leptin-deficient mice presented a less severe antigen-induced arthritis, decreased levels of TNF-α and IL-1β in knees synovium, and an impaired antigen-specific T cell proliferative response with lower IFN-γ and higher IL-10 production, which indicates a shift toward Th2 cell response (Busso et al., 2002). Accordingly, injection of leptin (5 µg) into the knee joint of collagenimmunized mice augmented arthritis severity, accompanied by elevated synovial hyperplasia and joint damage through enhancement of Th17 cell response (Deng et al., 2012). In fact, clinical trials using a humanized anti-IL-17 monoclonal antibody added to oral disease-modifying anti-rheumatic drugs, demonstrated improved signs and symptoms of RA, indicating the therapeutic potential of IL-17-directed strategies (Genovese et al., 2010).

Since leptin modulates the immune system, as well as insulin resistance and metabolic disorders like metabolic syndrome and obesity, all RA-associated conditions, this adipokine represents an attractive therapeutic target for RA. Accordingly, reducing leptin levels in RA patients by fasting improved the clinical symptoms of the disease (Fraser et al., 1999). Amidst the possible therapeutic approaches to antagonize leptin actions in RA, are leptin mutants with antagonist activity, and monoclonal antibodies against human LEPR or leptin itself (Tian et al., 2014). Interestingly, clinical studies evaluating the effect of drug modulators of insulin sensitivity (affected by leptin levels as described above), such as PPARγ agonists, are ongoing to provide new potential treatment to improve the inflammatory status and cardiovascular outcome in RA patients (Chimenti et al., 2015). Further understanding of leptin mechanisms would be of utmost importance for RA treatment.

Hence, leptin can be pointed as a link between immune tolerance, metabolic function, and autoimmunity, and leptin signaling-directed strategies could provide future innovative therapies for autoimmune disorders like RA.

### Systemic Lupus Erythematosus

Systemic lupus erythematosus is a chronic autoimmune disorder of unclear etiology characterized by hyperactive T and B cells, autoantibody production, deposition of immune complex, elevated blood levels of pro-inflammatory cytokines and multisystem organ damage, encompassing from mild manifestations (non-erosive arthritis or skin rash) to lifethreatening complications (lupus nephritis, neuropsychiatric disorders, cardiovascular disorders, and metabolic syndrome). Although the pathogenesis of SLE is poorly understood, genetic, hormonal and environmental factors have been implicated in the onset of this heterogeneous disease, which predominantly affects women of childbearing age (Gatto et al., 2013; Liu and La Cava, 2014).

Several studies suggest the implication of adipokines, namely leptin, in the pathogenesis of SLE. Although some reports found no statistical association between disease activity and leptin levels (Li H.M. et al., 2016), recently, a meta-analysis of eighteen studies determined that serum/plasma leptin levels were significantly elevated in SLE patients (Lee and Song, 2018). Furthermore, leptin has been suggested as a player that affects the cardiovascular risk in SLE patients. Accordingly, leptin and HFD induced proinflammatory high-density lipoproteins and atherosclerosis in BWF1 lupus-prone mice, and leptin levels were correlated with BMI, disease activity index (SLEDAI), as well as

insulin and CRP levels, all CVD risk factors, in SLE patients (Xu W.D.et al., 2014).

The role of leptin in SLE development has been investigated using leptin-deficient (ob/ob) mice treated with lupus-inducing agent (Lourenço et al., 2016). Leptin deficiency protected mice from the development of autoantibodies as well as renal disease, and elevated the levels of Treg cells, compared with wild-type controls. Moreover, in (NZBxNZW)F1 lupus-prone mice, leptin administration accelerated the development of autoantibodies and renal disease, while leptin antagonism delayed disease progression (Lourenço et al., 2016). At cellular level, leptin promoted Th1 responses in human CD4+ T cells and in lupusprone mice via RORγ transcription, whereas leptin neutralization inhibited Th17 responses in autoimmune-prone mice (Yu et al., 2013). Additionally, fasting induced hypoleptinemia or leptindeficient mice demonstrated decreased levels of Th17 and elevated levels of Treg cells (Liu et al., 2012). In SLE, apoptotic cells represent the major source of self antigens that promote and fuel autoimmune responses. Leptin promoted T cell survival and proliferation of autoreactive T cells in mice with an autoreactive T cell repertoire, including (NZBxNZW)F1 lupus-prone mice (Amarilyo et al., 2013). Leptin also promoted phagocytosis of apoptotic cells by macrophages in lupus-prone mice, which increase the availability of apoptotic-derived antigens to T cells and subsequent development of self-antigen-reactive T cells (Amarilyo et al., 2014).

Altogether, these data support the involvement of leptin in the development of SLE. However, further investigations are needed to fully understand the role of leptin in SLE and thus, explore this adipokine as potential therapeutic target of SLE.

# CONCLUSION AND FUTURE OUTLOOK

Obesity and its comorbidities, such as T2DM, non-alcoholic fatty acid liver disease, OA and RA, reached epidemic proportions and are still rising in developing countries. Anti-obesity therapeutic options have provided only a limited long-term efficacy (lifestyle changes, physical activity, diet, and pharmacotherapies) or are not completely safe (bariatric surgery) (Zhang et al., 2014). Therefore, it has become increasingly relevant to disclose new clinical biomarkers and to develop innovative therapeutic strategies for obesity-associated pathologies and chronic inflammation.

The adipose tissue-derived factor leptin has been emerged as a key regulator of nutritional state and metabolism, as well as a modulator of immune system activation and innate-adaptive frontier; thus bridging obesity with metabolic disorders (T2DM and NAFLD) and inflammatory pathologies that affect bones and joints (OA and RA). Consequently, plasma leptin concentration could be a biological marker of the inflammatory status and the onset and evolution of pathologies associated with dysregulation of the immune system, and hereafter evaluations will be essential to establish leptin as clinical biomarker. Moreover, control of bioactive leptin levels by high-affinity leptin-binding molecules, miRNAs targeting leptin, LEPRs antagonists or monoclonal humanized antibodies against LEPR are likely to be feasible therapeutic approaches (Otvos et al., 2011). Recombinant leptin is already available for use in patients with leptin congenital deficiency while the synthetic leptin analog metreleptin has been approved for lipodystrophy treatment (Tchang et al., 2015). Importantly, the development of antibodies that could crossreact with endogenous leptin and cause an effective leptindeficient state responsible for the loss of efficacy and infection has become a significant concern (DePaoli, 2014). However, given the pleiotropic action of leptin, a systematic approach to modulate their levels and thus prevent obesity-associated disorders might be, for the moment, unavailable. Instead, strategies targeting leptin's actions precisely and in specific immune cell subpopulations, or targeting of specific receptor isoforms, could be a potential viable option to add novel therapeutic agents against immune-metabolic pathologies. It is now clear that leptin is an important regulator of metabolic status and influence inflammatory and immune responses in several diseases. Nonetheless, leptin network is complex and a lack of a full understanding of leptin's immunomodulatory mechanisms in almost all of the cells of immune system and its potential side effects are still problems that need to be figured out in drug discovery. Further insights into the pathophysiological role of leptin in the immune system and in obesity-associated disorders will be of great importance for the development of novel therapeutic approaches for these diseases.

# AUTHOR CONTRIBUTIONS

VF and JP have made a substantial contribution to acquisition and analysis of data and critically revised it. VC-C, CR-F, AM, MG-G, and RG have been involved in drafting the manuscript and revising it critically for important intellectual content. OG made a substantial contribution to conception and design of the review article, drafting the manuscript, and critically revising it. All authors approved the final version to be published.

# FUNDING

OG is Staff Personnel of Xunta de Galicia (Servizo Galego de Saude, SERGAS) through a research-staff stabilization contract (ISCIII/SERGAS). VF is a "Sara Borrell" Researcher funded by ISCIII and FEDER. RG is a "Miguel Servet" Researcher funded by Instituto de Salud Carlos III (ISCIII) and FEDER. OG, MG-G, and RG are members of RETICS Program, RD16/0012/0014 (RIER: Red de Investigación en Inflamación y Enfermedades Reumáticas) via Instituto de Salud Carlos III (ISCIII) and FEDER. The work of OG and JP (PIE13/00024 and PI14/00016, PI17/00409), and RG (PI16/01870 and CP15/00007) was funded by Instituto de Salud Carlos III and FEDER. OG is a beneficiary of a project funded by Research Executive Agency of the European Union in the framework of MSCA-RISE Action of the H2020 Program (Project No. 734899). The funders had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.

# REFERENCES

fphys-09-00640 May 30, 2018 Time: 18:27 # 14



neuroendocrine/metabolic dysfunction of human congenital leptin deficiency. J. Clin. Invest. 110, 1093–1103. doi: 10.1172/JCI15693





and transcriptional signature controlling CD4+ effector T cell responses. J. Immunol. 189, 2941–2953. doi: 10.4049/jimmunol.1200935



identify subjects at increased metabolic and cardiovascular risk. Obesity 21, 461–468. doi: 10.1002/oby.20087


tissue during obesity. Cell Death Dis. 7:e2167. doi: 10.1038/cddis.20 16.54


**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 Francisco, Pino, Campos-Cabaleiro, Ruiz-Fernández, Mera, Gonzalez-Gay, Gómez and Gualillo. 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.

# Relaxin-2 in Cardiometabolic Diseases: Mechanisms of Action and Future Perspectives

Sandra Feijóo-Bandín1, 2 \*, Alana Aragón-Herrera<sup>1</sup> , Diego Rodríguez-Penas <sup>1</sup> , Manuel Portolés 2, 3, Esther Roselló-Lletí 2, 3, Miguel Rivera2, 3, José R. González-Juanatey 1, 2 and Francisca Lago1, 2

<sup>1</sup> Cellular and Molecular Cardiology Research Unit, Institute of Biomedical Research and University Clinical Hospital, Santiago de Compostela, Spain, <sup>2</sup> Centro de Investigación Biomédica en Red de Enfermedades Cardiovasculares, Madrid, Spain, <sup>3</sup> Cardiocirculatory Unit, Health Research Institute of La Fe University Hospital, Valencia, Spain

Despite the great effort of the medical community during the last decades, cardiovascular diseases remain the leading cause of death worldwide, increasing their prevalence every year mainly due to our new way of life. In the last years, the study of new hormones implicated in the regulation of energy metabolism and inflammation has raised a great interest among the scientific community regarding their implications in the development of cardiometabolic diseases. In this review, we will summarize the main actions of relaxin, a pleiotropic hormone that was previously suggested to improve acute heart failure and that participates in both metabolism and inflammation regulation at cardiovascular level, and will discuss its potential as future therapeutic target to prevent/reduce cardiovascular diseases.

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Chen Huei Leo, Singapore University of Technology and Design, Singapore Elizabeth Anne Schroder, University of Kentucky, United States

#### \*Correspondence:

Sandra Feijóo-Bandín sandra.feijoo@gmail.com

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 29 May 2017 Accepted: 03 August 2017 Published: 18 August 2017

#### Citation:

Feijóo-Bandín S, Aragón-Herrera A, Rodríguez-Penas D, Portolés M, Roselló-Lletí E, Rivera M, González-Juanatey JR and Lago F (2017) Relaxin-2 in Cardiometabolic Diseases: Mechanisms of Action and Future Perspectives. Front. Physiol. 8:599. doi: 10.3389/fphys.2017.00599

Keywords: cardiovascular diseases, relaxin-2, metabolism, inflammation, therapy, heart

# INTRODUCTION

In the last decades, cardiovascular diseases (CVDs) have remained as the first cause of death worldwide, being their prevalence boosted every year mainly due to our new way of life, based on the increased intake of cheaper energy-dense food and a sedentary lifestyle (Pérez-Martínez et al., 2017; WHO | Cardiovascular diseases (CVDs), 2017). Hand in hand with this increase in the prevalence of CVDs goes the increase in obesity (WHO | Obesity Overweight, 2016), which not only is a risk factor for CVDs by itself, but also promotes the development of other CVDs comorbidities/risk factors, including hypertension, insulin resistance, dyslipidemia, type 2 diabetes mellitus (T2DM) or the increase in systemic inflammation (Tune et al., 2017). In particular, the combination of abdominal obesity, hypertension, hyperglycemia and dyslipidemia is known as metabolic syndrome (Matsuzawa et al., 2011; Wiernsperger, 2013), and due to the increasing evidences relating the presence of metabolic syndrome to the development of cardiovascular events such as myocardial infarction or stroke, this state is now termed cardiometabolic syndrome (Wiernsperger, 2013). The main therapeutic approach to treat the cardiometabolic syndrome is focused on restoring the metabolic disorder to a normal state through weight reduction and the prescription of drugs such as anti-diabetics, statins, anti-inflammatories or anti-hypertensives (Duprez and Toleuova, 2013; Ginsberg, 2013; Wiernsperger, 2013; Soare et al., 2014; Desouza et al., 2015). Unluckily, the therapeutic approaches available nowadays to treat the pathologies that define the cardiometabolic syndrome are not sufficient, since this syndrome alters different metabolic pathways, mainly those regarding glucose and lipid metabolism, and affects diverse organs/tissues,

**142**

including the liver, the muscles or the fat tissue, and, moreover, each individual can show different metabolic abnormalities (Wiernsperger, 2013). Thus, there is an urge to understand the signaling pathways of the different contributors to the development of cardiometabolic diseases (CMDs) in an attempt to find new possible targets that with their therapeutic modulation could improve CMDs treatment and/or prevention.

In this line, in the last years the obesity and the adipose tissue have received considerable attention regarding their potential contribution to the development of CMDs. It is well established that the adipose tissue functions as an endocrine organ by secreting a number of proteins/hormones (adipokines) mainly implicated in the regulation of metabolism and in the control of the inflammatory response (Mancuso, 2016). Obesity induces an imbalance in the adipokine production in favor of proinflammatory adipokines and in detriment of anti-inflammatory adipokines, leading to a low grade of chronic inflammation that promotes both systemic metabolic dysfunction and CVDs (Nakamura et al., 2014; Molica et al., 2015). In fact, inflammation is nowadays recognized as a central player in the development of CVDs and its complications (Ruparelia et al., 2016), and the study of this kind of hormones that influence metabolism and inflammation and which have been shown to have effects at cardiovascular level (not only adipokines, but also other hormones, such as ghrelin, which is mainly produced by the stomach (Lilleness and Frishman, 2016), nesfatin-1, that is widely expressed in the body, including the brain and the heart (Feijóo-Bandín et al., 2015), or prokineticin, secreted by immune cells and reproductive organs, and expressed in heart and kidney apart from the adipose tissue Nebigil, 2017) has raised a great interest among the scientific community regarding their potential role in the development/prevention of CMDs (Ingelsson et al., 2008; Athyros et al., 2010; Gonzaga et al., 2014; Chiara et al., 2015; Prinz and Stengel, 2016; Colldén et al., 2017). Hence, the study of this kind of proteins/hormones that participate in the regulation of metabolism and/or inflammation can shed light in the understanding of how cardiometabolic diseases behave, and contribute to the developing of new therapeutic approaches.

Relaxin is a hormone that was first identified as a reproductive hormone implicated in vasoregulation during pregnancy and the softening of the tissues of the birth canal during delivery (Bani, 1997), but that has been recently suggested to participate in metabolism regulation and to exert protective effects at cardiovascular level. This review outlines the functions of relaxin as a new potential metabolic hormone with cardiovascular actions and discusses its potential as future therapeutic target to prevent/reduce CMDs.

# RELAXIN

Relaxin is a 6 kDa hormone identified and named in 1926 by Frederick Hisaw due to its ability to induce the relaxation of the pubic ligaments and the softening of pubic symphysis just prior to delivery in pregnant gophers and guinea pigs (Hisaw, 1926; Wilkinson et al., 2005). Subsequently, different relaxin genes were discovered, so that nowadays the relaxin peptide family consists of seven peptides: relaxin (RLN)-1, RLN-2, RLN-3/insulin-like peptide (INSL)-7, and INSL3-6 (Bathgate et al., 2013). In humans and higher primates, there are three RLN genes: RLN-1, RLN-2, and RLN-3; however, the function of RLN-1 is unclear and it may even represent a pseudogene in these species. In contrast, other mammals have only the RLN-1 and RLN-3 genes. Importantly, the RLN-2 gene in humans and the RLN-1 gene from other mammals are equivalent and encode the relaxin peptides that circulate in blood during pregnancy, being RLN-2 in humans and great apes and RLN-1 in other non-primate species commonly referred to as relaxin (Wilkinson et al., 2005; Bathgate et al., 2013; in this review we will refer to RLX-2 and RLX-1 as relaxin), while the peptide encoded by the RLN-3 gene is a neuropeptide in all species (**Figure 1**; Dschietzig, 2014).

Although human relaxin was originally discovered as a hormone mainly secreted by the corpus luteum of the ovary that regulates the adaptive changes in pregnancy, it is also produced in non-pregnant women, and males also produce relaxin, being identified the presence of the relaxin peptide in prostate. Relaxin mRNA was also detected in other tissues, such as the endometrium, decidua, placenta, mammary gland, brain and the heart (Bathgate et al., 2013).

Initially, the different members of the relaxin family were discovered due to their similar structure and their roles in reproduction. However, nowadays we know that they participate in a wide range of physiological functions apart from reproduction, including stress, fear and anxiety responses, behavioral activation, mood, reward, depression, addiction, feeding behavior, metabolism, water drinking behavior, learning and emotional memory or somatosensory motor behavior (Gundlach et al., 2013). In some cases, they are expressed and have well conserved roles in different species, like RLX-3 and INSL-3, but in others, such as RLX-2, the expression and function differ between species (Bathgate et al., 2013).

# Relaxin Activation and Receptors

Relaxin is a two-chain peptide with a structure and processing similar to insulin. It is produced as a pro-hormone, containing a signal sequence and a B-C-A domain configuration, and after processing by prohormone convertases, the C domain is removed and three disulphide bonds are formed between six highly conserved cysteine residues in the A and B chains. Thus, the mature relaxin is constituted by the A and B chains with three disulphide bonds, like insulin (James et al., 1977; Bathgate et al., 2013). Human relaxin gene structure was first identified in 1983, showing a highly conserved sequence within the B-chain (R-X-X-X-R-X-X-I/V-X) that was later found to be indispensable for the binding to relaxin receptors (Hudson et al., 1983; Bathgate et al., 2013). All of the members subsequently discovered of the relaxin family retain the relaxin-like pre-prohormone structure and are predicted or proven to have the same processing and structure (Bathgate et al., 2013).

Although relaxin peptides are structurally related to insulin, they have low sequence similarity and bind to a different type of receptors. Relaxin peptides activate a group of four G protein-coupled receptors (GPCRs): the relaxin family peptide receptors (RXFP) 1-4; whereas insulin activates tyrosine kinase

receptors (Wilkinson et al., 2005; Siddle, 2011; Bathgate et al., 2013). Relaxin 1/2 and INSL-3 bind to RXFP-1 and RXFP-2, respectively. RXFP-1 activation triggers signaling pathways mainly related to the generation of second messengers like nitric oxide (NO) or cyclic adenosine monophosphate (cAMP) (and the subsequent activation of protein kinase A (PKA) and the cAMP-response element (CRE)-mediated transcription), and also stimulates the phosphorylation of mitogen-activated protein (MAP) kinases like ERK1/2 or AKT, while RXFP-2 activation only induces cAMP and CRE-dependent gene transcription (Bathgate et al., 2013). RLN-3 and INSL-5 activate RXFP-3 and RXFP-4 respectively, which inhibit cAMP production and activate MAP kinases. The receptors for INSL-6 and INSL-4 remain currently unknown (**Figure 2**; Bathgate et al., 2013).

Different studies have shown a wide distribution of relaxin receptors in humans and murine, including ovary, prostate, brain, kidney, liver, pancreas, skeletal muscle, ligament, tendon, joint tissues, thymus, thyroid, adrenal glands, heart, arteries and veins (Halls et al., 2007; Clifton et al., 2014; Jelinic et al., 2014; Kim et al., 2016). Thus, this wide distribution of relaxin receptors in different species supports the potential pleiotropic effects of relaxin.

# Relaxin as a Metabolic Hormone

The similar structure of relaxin and insulin suggests that relaxin might have some important actions in regulating energy metabolism. In this line, it has been shown the expression of relaxin receptors in important organs for insulin action, like the pancreas, liver or muscle (Halls et al., 2007), and there exist some studies that propose relaxin as a new potential energy-regulating peptide. In fact, relaxin has been shown to promote glycogen depletion and to induce morphological changes of hepatocytes, which are consistent with functional activation, in both male and non-pregnant female rats (Bani et al., 2001b), and the peripheral infusion of relaxin in C57BL/6J mice has been shown to enhance insulin-stimulated muscle glucose uptake when animals are fed with normal diet but not when fed with high fat diet, and to reverse diet-induced insulin resistance in those fed with high fat diet, suggesting that relaxin can be an effective new molecule to revert muscle insulin resistance (Bonner et al., 2013). One of the mechanisms that trigger insulin resistance is the inflammation produced in obesity, due to the impairment of the adipose tissue homeostasis and the imbalance on the adipokine production toward a pro-inflammatory profile (Blüher, 2016). So far, little is known about the mechanisms through which relaxin seems to ameliorate insulin resistance and more studies are needed, but the control of the inflammatory processes linked to the development of impaired insulin sensitivity could be one of the involved pathways and is an interest field of study regarding the possible relaxin use as a therapy in CMDs.

Relaxin circulating concentrations in women with T2DM have been found to be negatively correlated to pancreatic βcells activity, but positively correlated to insulin sensitivity and to other factors that are closely related to pancreatic function and insulin sensitivity, like fasting circulating concentrations of insulin, total cholesterol and LDL cholesterol, or C peptide, suggesting that relaxin may protect against insulin resistance in women with T2DM (Szepietowska et al., 2008). Serum relaxin concentrations have also been shown to be elevated in pregnant women with T1DM (Whittaker et al., 2003), in pregnant women with early gestational diabetes mellitus (Alonso Lopez et al., 2017), in non-pregnant women with metabolic

syndrome (Ghattas et al., 2013). However, in a study composed by men and women, circulating relaxin concentrations were lower in patients with T2DM than in controls but not related to component traits in patients with diabetes such as cholesterol, triglycerides, fasting blood glucose or fasting insulin (Zhang et al., 2013). This difference could be explained due to the sample size and the gender differences between studies.

Despite de fact that (a) relaxin shares its structure with insulin, (b) it seems to improve insulin resistance, and (c) its circulating levels are altered in diabetes, it is not clear whether or not relaxin could share with insulin the capacity to decrease blood glucose levels. Although, in healthy C57BL/6J mice fed with high fat diet (Bonner et al., 2013) and in genetically diabetic db+/db + mice (Bitto et al., 2013) the treatment with relaxin was shown to decrease fasting blood glucose levels (Bonner et al., 2013), there exist other studies with diabetic animal models in which relaxin do not normalize circulating glucose concentrations (Dschietzig et al., 2015; Ng et al., 2017) or glycosylated hemoglobin (Wong et al., 2013). Thus, more studies are needed to elucidate if relaxin could indeed mimic insulin and decrease circulating glucose levels. In regard with this aspect, even though insulin and relaxin activate a different type of receptors, both insulin receptors and RXFP-1 signaling can converge in the activation of AKT (Zaid et al., 2008; Bathgate et al., 2013; Sun et al., 2016; Ogunleye et al., 2017), which is a key kinase implicated in glucose transporter-4 (GLUT-4) mobilization to the cell membrane and glucose uptake in different tissues, contributing to the lowering of glucose circulating concentrations (Sakamoto and Holman, 2008). This observation suggests that relaxin could potentially participate in the regulation of blood glucose levels, but more studies are needed to clarify this issue.

Regarding food intake regulation, the neuropeptide RLN-3 is the one of the relaxin family that was by far more studied. Centrally administered RLN-3 in rats has been shown to increase water intake, food intake and body weight in males (McGowan et al., 2005; Hida et al., 2006; Otsubo et al., 2010). When comparing the different effect of RLN-3 central injection between sex, female rats show a higher increase in food intake and in body weight gain compared to males, and it induces a different corticotrophin-releasing hormone (CRH) expression pattern in different paraventricular hypothalamic nucleus (PVN) areas between male and female rats, an effect suggested to mediate the different response to RLN-3 on food intake behavior between sex (Lenglos et al., 2015). On the other hand, the intake of rewarding substances, like sucrose or alcohol, has been shown to increase endogenous RLN-3 levels in the brain (Ryan et al., 2014); and in rats with diet-induced obesity (DIO) it was shown that central RLN-3 expression is constitutively increased, and that refeeding after food deprivation stimulates the orexigenic effect of RLN-3 through the increase of RXFP-3 expression in brain areas that regulate food intake (Lenglos et al., 2014). The orexigenic effect of RLN-3 through its cognate receptor RXFP-3 has been proved to be mediated by the hyperpolarization and consequent inhibition of the majority of putative magnocellular PVN neurons, including cells producing the anorexigenic neuropeptides, oxytocin and vasopressin (Kania et al., 2017).

Although there exist more studies so far focused on the neuropeptide RLN-3 role regarding food intake behavior than on relaxin, these works could provide some clues and open a new line of study concerning the possible role of relaxin in food intake regulation and its concomitant effect on CMDs. In fact, and on the contrary of RLN-3, central and peripheral administration of RLN-2 in ad libitum-fed male rats has been shown to reduce food intake (McGowan et al., 2010).

Relaxin has been identified as a secreted factor in porcine adipose tissue (Hausman et al., 2006), and to induce hypertrophy in mammary and parametrial adipose tissue in female mice and in 3T3-L1 preadipocytes (Bianchi et al., 1986; Bani et al., 1989; Pawlina et al., 1989), and to promote lipid deposition in the parametrial adipose tissue in mice (Bani et al., 1989). Although the adipose tissue seems to be a target organ for relaxin, it is unknown whether or not the different adipose tissue depots express relaxin receptors, and whether or not relaxin could participate in the regulation of adipose tissue homeostasis in terms of growth, energy metabolism or even adipokine secretion, with the concomitant effect on systemic inflammation and the development of CMDs.

# Relaxin and Relaxin Receptors Expression in the Cardiovascular System

Relaxin and its receptors are widely located in different cardiovascular tissues. It has been demonstrated the expression of RXFP-1 in rodents in the aorta, vena cava, mesenteric artery, mesenteric vein, femoral artery, femoral vein, small pulmonary arteries and small renal arteries (Novak et al., 2006; Jelinic et al., 2014), as well as in cardiomyocytes (Moore et al., 2009) and in cardiac atrial and left ventricle tissue, with higher expression in atria (Osheroff and Ho, 1993; Hsu et al., 2000; Kompa et al., 2002; Krajnc-Franken et al., 2004; Scott et al., 2004). Moreover, relaxin shows a specific and high-affinity binding to its receptors in the atrium in both male and female rat heart (Osheroff et al., 1992). RXFP-1 has also been detected in human heart (Hsu et al., 2002; Dschietzig et al., 2011), again with higher expression in atria, and its expression is enhanced by α1-adreonreceptors stimulation but suppressed by β1-adrenoreceptors activation in cultured rat cardiomyocytes and in transgenic mouse hearts with cardiacrestricted overexpression of subtypes of adrenoceptors (Moore et al., 2009, 2014).

Likewise relaxin receptors, relaxin is also expressed in cardiovascular tissues. In rodents, relaxin has been detected in thoracic aortas, mesenteric arteries, small renal arteries, in rat heart tissue and in cultured cardiomyocytes derived from the atria of neonatal rats, which secrete relaxin in detectable amounts (Taylor and Clark, 1994; Gunnersen et al., 1995; Novak et al., 2006), and RLX-3 is also detected in the atria and ventricle in mice and rats (Bathgate et al., 2001; Kompa et al., 2002). In humans, relaxin was demonstrated to be expressed in atrial and ventricular cardiac tissue (Dschietzig et al., 2001).

# Relaxin Effects at Cardiovascular Level

Relaxin has been demonstrated to participate in the cardiovascular and hemodynamic changes required to adapt the cardiovascular system to pregnancy, so that during pregnancy take place increases in plasma volume, cardiac output or heart rate, and decreases in blood pressure and vascular resistance (Bathgate et al., 2013). However, relaxin can also regulate cardiovascular function in men and non-pregnant women at different levels, modulating blood pressure, inflammation, cell injury/death, fibrosis, hypertrophy or angiogenesis (Teichman et al., 2010; Leo et al., 2016a).

# Relaxin Effects on Vasculogenesis and Vascular Function

Relaxin is able to stimulate the formation of new blood vessels, not only in pregnancy but also in tumorigenesis or ischemic wounds, through the upregulation of vascular endothelial growth factor (VEGF) transcripts (Shirota et al., 2005; Silvertown et al., 2006; Segal et al., 2012; Bitto et al., 2013; Unemori et al.). And in genetically diabetic mice, relaxin not only increases new vessel formation but also improves the impaired wound healing, suggesting that it could be beneficial in diabetes-related wound disorders (Squadrito et al., 2013).

The endothelial cells are the key regulators of the vascular tone through the production and secretion of vasoactive substances, including vasodilator factors such as NO, prostacyclin (PGI2), kinins (bradykinin), or endothelium-derived hyperpolarizing factors (like K<sup>+</sup> ions), and vasoconstrictor agents such as endothelin-1, thromboxane A or angiotensin II (AngII) (Félétou and Vanhoutte, 2006; Su, 2006).

A big number of studies have shown that relaxin promotes vasodilation through a mechanism that involves NO production in a wide range of organs/tissues, not only in reproductive organs such as the mammary glands (Bani et al., 1988) or the uterus (Vasilenko et al., 1986; Bani et al., 1995), but also in nonreproductive tissues like the mesocaecum (Bigazzi et al., 1986), kidney (Danielson et al., 1999, 2000; Novak et al., 2001; Conrad et al., 2004; McGuane et al., 2011), subcutaneous fat (McGuane et al., 2011) or liver (Bani et al., 2001a).

In the heart, relaxin increases coronary flow in normal and hypertensive rats (Bani-Sacchi et al., 1995; Masini et al., 1997; Debrah et al., 2005b). Relaxin has also been shown to decrease systemic arterial resistance and to increase global artery compliance in rats (Conrad et al., 2004; Debrah et al., 2005a,c, 2006; Conrad and Shroff, 2011), as well as it reverses large artery remodeling and improves arterial compliance in senescent spontaneously hypertensive rats (Xu et al., 2010). In pregnant relaxin-deficient mice, relaxin administration for 5 days has been shown to prevent vascular dysfunction in mesenteric arteries and to ameliorate the increased responsiveness of small mesenteric arteries to the vasoconstrictor AngII, suggesting that relaxin could alleviate maternal systemic vascular dysfunction associated with hypertensive diseases in pregnant women (Marshall et al., 2017).

Recombinant human relaxin in co-treatment with high doses of glucose for 3 days was also demonstrated to prevent vascular dysfunction in the mouse aorta through a mechanism that reverts the reduced sensitivity to the endothelium-dependent agonist acetylcholine induced by high glucose, and that ameliorates PGI<sup>2</sup> production (Ng et al., 2016). In streptozotocin induced diabetic mice, relaxin treatment for 2 weeks reversed diabetes-induced endothelial dysfunction in terms of endothelial vasodilator function in mesenteric arteries and aorta by increasing NO and PGI<sup>2</sup> mediated relaxation, but it did not affect endotheliumderived hyperpolarizing factors (Ng et al., 2017).

Acute infusion of relaxin (3 h) in healthy male rats has also been shown to increase in the mesenteric artery basal NOS activity and to reduce endothelin-1 dependent contraction, and this vasodilator effect was sustained for 24 h due to the following increase in PGI2/bradykinin production, even though the absence of circulating levels of relaxin at 24 h (Leo et al., 2014). Similarly, in male rats continuously infused with relaxin, it was shown an increase in the endothelial vasodilator function in arteries, but not in veins, through the production of NO and the increase of eNOS activity at 48 h, a mechanism reverted at 72 h, but at this time, relaxin induced a transition to PGI<sup>2</sup> and bradykinin production, a mechanism suggested by the authors to be key to sustain vascular response to relaxin in time (Leo et al., 2016b). The same effects are observed when relaxin is administered chronically (5 days) in male rats: relaxin reduces wall stiffness and increases volume compliance in mesenteric arteries through the increase of bradykinin-mediated relaxation, involving enhanced NO production but not endothelium-derived K <sup>+</sup> hyperpolarization, and in this study PGI<sup>2</sup> production was not observed (Jelinic et al., 2014). On the other hand, in blood-perfused hamster cremaster muscle preparations in situ, relaxin induced a rapid (seconds), transient vasodilation in transverse and branch arterioles through NO production and K <sup>+</sup> hyperpolarization, while the smallest ramification of the arteriolar tree was not responsive to relaxin (Willcox et al., 2013). However, it was also shown that 48 h intravenous relaxin infusion in healthy rats does not significantly alter resting outer diameter or pressure-induced myogenic tone in the mesenteric vasculature despite enhancing the contribution of NO through increased endothelial NO synthase (eNOS) dimerization (Jelinic et al., 2017).

Taken these results all together, it seems clear that relaxin has a potent vasodilator effect, and that contributes to ameliorate endothelial dysfunction in cardiometabolic scenarios such as hypertension or diabetes. It was recently suggested that endothelial cells have functional heterogeneity depending on the tissue, being determined by mechanical and metabolic stimuli, as well as by the characteristic microenvironment of each tissue (Potente and Mäkinen, 2017), and also between sex (Mudrovcic et al., 2017). Thus, the differences observed regarding timing and the specific pathways activated by relaxin in the different studies could be due not only to the different experimental designs and animal models used or relaxin doses, but also to a different response by endothelial cells from different tissues/physiopathological conditions to relaxin.

Apart from the regulation of the vascular tone, endothelial cells mediate other functions, such as the preservation of blood fluidity, the formation of new blood vessels, platelet function, vascular smooth muscle cell growth and migration or the regulation of the inflammatory response (Jensen and Mehta, 2016; Incalza et al., 2017). Under pathological scenarios associated with a pro-inflammatory profile, such as obesity, diabetes, hypertension or dyslipidemia, the endothelial cells are influenced by cytokines and external stimuli to change into a pro-inflammatory and pro-coagulant state, characterized by the expression of cell-surface adhesion molecules required for the recruitment and attachment of inflammatory cells, which lead to clot generation, increasing the thrombotic risk as a consequence of increased blood thrombogenicity or impaired fibrinolysis (Incalza et al., 2017; Montecucco et al., 2017). Thus, and although little is known so far, the proved effect of relaxin on regulating endothelial function suggests that relaxin could also ameliorate the inflammatory response in the vascular system under pathological conditions, and this opens a promising new field of study of relaxin regarding its potential role as a regulator of cardiovascular inflammation. In fact, in human endothelium and vascular smooth muscle cells, relaxin was already proved as a potent inhibitor of early vascular inflammation, decreasing the expression of endothelial adhesion molecules, cytokine expression and suppressing monocyte adhesion to the endothelium (Brecht et al., 2011), a result also observed in vivo in female apolipoprotein E-deficient mice fed with a high-fat and cholesterol-rich diet for 6 weeks, in which relaxin treatment for the last 4 weeks reduced vascular oxidative stress, improved endothelium-dependent vasodilatation, reduced the development of the atherosclerotic plaque, decreased circulating concentrations of the cytokines interleukin (IL)-6 and IL-10, and down-regulated the angiotensin II type 1a receptor in the aorta, but in this study authors did not find differences in vascular macrophage, T-cell or neutrophil infiltration, nor in collagen/vascular smooth muscle cell content between relaxin treated and control mice (Tiyerili et al., 2016).

## Chronotropic and Inotropic Effects of Relaxin in the Heart

In the heart, relaxin has powerful positive chronotropic and inotropic effects. It has been shown to induce an increase in the contraction force and rate in isolated rat atria, and in conscious normotensive and spontaneously hypertensive rats relaxin increases heart rate without alter urine or blood volume, mean arterial pressure of water and food intake (Kakouris et al., 1992; Ward et al., 1992; Toth et al., 1996). In rat perfused hearts, relaxin infusion has been shown to induce the release of the atrial natriuretic peptide (ANP) along with the increase in heart rate through a mechanism that involves protein kinase C (PKC) activation (Toth et al., 1996), and in isolated murine cardiac myofilaments relaxin increases cardiac myofilaments force through a PKC-dependent pathway that leads to the increase of myofilament Ca2<sup>+</sup> sensitivity (Shaw et al., 2009). As well, in rat isolated hearts it causes a dosedependent tachycardia in both intact preparations and those in which the atria had been removed, suggesting that relaxin acts on both the atrial and ventricular pacemakers to increase the heart rate (Thomas and Vandlen, 1993). In fact, it was demonstrated in single cells isolated from the sinoatrial node in rabbits that relaxin is able to enhance L-type Ca2<sup>+</sup> current through a mechanism dependent on cAMP formation and PKA activity (Han et al., 1994). In human myocardium, relaxin has positive inotropic effects in atrial tissue, without differences between control and failing hearts, through a mechanism that involves PKA activation and a decrease in the transient K<sup>+</sup> outward current, an effect partially blunted by the pretreatment with pertussis toxin and the inhibition of phosphoinositide-3 kinase (PI3K) in non-failing hearts but notably suppressed in failing myocardium (Dschietzig et al., 2011). However, in this study, relaxin did not show any inotropic effects in ventricular myocardium.

### Relaxin and Ischemia-Reperfusion Injury

Relaxin has been extensively proven to protect the heart against damage induced by ischemia/reperfusion. The process of ischemia/reperfusion induces the generation of O2-derived free radicals, that contribute to the peroxidation of cell membrane lipids and to damage the mitochondrial function, and the overload of Ca2+, which alters myofilaments contractile function and triggers proteolytic cascades, leading to cell injury (Anderson et al., 2012; Bompotis et al., 2016). In isolated guinea pig heart, relaxin was shown to protect myocardium from ischemia/reperfusion injury by decreasing the peroxidation of cell membrane lipids and Ca2<sup>+</sup> overload, as well as the hypercontraction of myofibrils, mitochondrial swelling and accumulation of dense granules in the mitochondrial matrix, through a mechanism that involves NO production (Masini et al., 1997).

As well, in a swine model of acute myocardial infarction, relaxin injection during reperfusion caused a reduction in circulating markers of myocardial injury, as troponin T, creatin kinase-MB or myoglobin, and in tissue malondialdehyde (an end product of lipid peroxidation) and Ca2<sup>+</sup> (mediate cardiomyocyte injury), caspase-3 (implicated in cardiomyocyte apoptosis), and myeloperoxidase (which recruits inflammatory leukocytes), and improved cardiac contractile function (Perna et al., 2005). According to this, other authors have found that relaxin also protects from the damage induced by ischemia and reperfusion in rat heart by a similar mechanism, so that intravenous relaxin injection 30 min before ischemia diminished the extension of the damaged areas, ventricular arrhythmias, the recruitment and accumulation of neutrophils and morphological signs of myocardial cell injury, by decreasing oxygen-derived free radicals, preventing the Ca2<sup>+</sup> overload in the myocardial tissue, and reducing hypercontraction of myofibrils, mitochondrial calcification, and cell necrosis (Bani et al., 1998). Moreover, in rats with isoproterenol-induced myocardial injury, it was found a compensatory up-regulation of myocardial relaxin expression, and when relaxin was co-administered with isoproterenol for 10 days, it attenuated myocardial injury and fibrosis, and improved cardiac function (Zhang et al., 2005). In this line, in a rat model with myocardial infarction, subcutaneously administrated relaxin during 2 weeks was probed to attenuate tachyarrhythmia and cardiac dysfunction in the healing infarcted heart, to reduce the dispersion of action potential duration in post-infarcted hearts, to reduce myocardial apoptosis and cardiac fibrotic collagen deposition and to inhibit protein expression levels of tumor growth factor (TGF) β1, α-SMA, and type I collagen (Wang et al., 2016). In a different murine model of myocardial infarction, it has been shown that relaxin administration (1 h prior to ischemia or as a reperfusion therapy) attenuates myocardial ischemia/reperfusion injury by reducing infarct size and left ventricular dysfunction after 24 h through a mechanism that involves eNOS signaling and the attenuation of the activation of the Nod like receptor containing a pyrin domain-3 (NLRP3)-inflammasome (Valle Raleigh et al., 2017), which is a macromolecular structure that functions as a platform for the production of pro-inflammatory cytokines of the IL-1 family (i.e., IL-1b and IL-18) and is involved in the impairment of heart function and remodeling after myocardial injury (Mezzaroma et al., 2011; Lamkanfi and Dixit, 2012; Bracey et al., 2013).

After the induction of myocardial infarction in swine and rats, the transplantation of skeletal myoblasts overexpressing relaxin was probed as an effective treatment to increase vascularization, increase collagen turnover, reduce fibrosis, and improve left ventricular function, compared to non-overexpressing skeletal myoblast (Formigli et al., 2007; Bonacchi et al., 2009).

### Relaxin and Atrial Fibrillation

Atrial fibrillation (AF), defined as a supraventricular tachyarrhythmia due to uncoordinated atrial activation with deterioration of the atrial mechanical function, is nowadays one of the cardiovascular events that are causing an extremely costly public health problem, being sex, age and hypertension the main risk factors for its development (Fuster et al., 2006). Thus, the understanding of the mechanisms related to AF development/prevention is of a great interest.

In spontaneously hypertensive rats, relaxin treatment for 14 days was shown to suppresses AF through the inhibition of fibrosis and hypertrophy, and the increase in conduction velocity, and in human cardiomyocytes derived from inducible pluripotent stem cells, relaxin treatment for 48 h was probed to up-regulate voltage-gated Na<sup>+</sup> channels, a mechanism suggested by the authors to participate in the suppression of AF (Parikh et al., 2013), a result also observed in aged rats (Henry et al., 2015). As well, in mice with myocardial infarction, relaxin treatment after myocardial infarction for 14 days also reduces AF through the decrease on fibrosis and hypertrophy, the increase in conduction velocity, and, moreover, the decrease of the proinflammatory cytokine IL-1β expression (Beiert et al., 2017).

In humans, circulating relaxin was found to be increased in patients with AF and to be associated with serum concentration of fibrosis-related markers, as well as with the occurrence of heart failure in AF patients (Zhou et al., 2016).

# Relaxin Effects on Cardiac Cells

Apart from the numerous studies regarding relaxin effects on the cardiovascular system and in heart physiology, there are also some reports concerning relaxin direct effects on cardiomyocytes and cardiac fibroblasts.

In neonatal rat atrial and ventricular fibroblasts in culture, relaxin was shown to decrease collagen secretion and deposition by the inhibition of fibroblasts proliferation and differentiation, and the enhancement of matrix metalloproteinase activity, an effect also observed in two models of cardiac fibrosis in vivo, in which relaxin is able to revert collagen overexpression (Samuel et al., 2004; Mookerjee et al., 2005; Wang et al., 2009). In cardiac fibroblasts, relaxin co-treatment with high glucose was suggested to inhibit high glucose-associated cardiac fibrosis partly through the decrease in total expression and translocation of PKCβ2 (Su et al., 2014).

In mouse neonatal immature cardiomyocytes, relaxin promotes cell proliferation and maturation (Nistri et al., 2012), an effect that is also potentiated when are co-cultured with relaxin overexpressing skeletal myoblasts (Formigli et al., 2009). In fact, relaxin has been shown to potentiate intercellular coupling between myoblasts and cardiomyocytes by up-regulating the transcellular exchange of regulatory molecules between both cell types (Formigli et al., 2005).

Relaxin was demonstrated to inhibit the ability of cardiac fibroblast-conditioned medium to induce hypertrophy in cardiomyocytes, and to directly attenuate apoptosis induced by oxidative stress and by high glucose exposure in cardiomyocytes through a protective mechanism that involves AKT and ERK activation (Moore et al., 2007), and through the inhibition of both extrinsic and intrinsic pathways of apoptosis and endoplasmic reticulum stress (Zhang et al., 2015).

# RELAXIN AS A FUTURE THERAPY FOR CARDIOMETABOLIC DISEASES: LIGHTS AND SHADOWS

Due to the important relaxin effects not only on the cardiovascular system but also in the development of metabolic disorders that suppose risk factors for the development of cardiovascular diseases (**Figure 3**), relaxin has been considered in the last years as a really promising cardiometabolic hormone that with its therapeutic modulation could help to prevent/treat cardiovascular diseases. Although relaxin has been probed to participate in the pathophysiological processes that lead to CMDs, is just in the scenario of acute heart failure where relaxin has created interest as a therapeutic agent. In this line, human recombinant relaxin (serelaxin/RLX030) has been under commercial development by Novartis Pharma A.G. (Basel, CHE) and it was first tested in healthy or hypertensive rodents and

humans, proving its capacity to increase systemic vasodilatation, global arterial compliance, cardiac index and stroke volume, and to decrease arterial stiffness (Du et al., 2014). Subsequently, its safety, tolerability and beneficial effect was tested in phase I and II clinical trials in stable and acute heart failure patients (Dschietzig et al., 2009; Teerlink et al., 2009; Sato et al., 2015), and in 2013 there were published the results of the phase-III multicenter, randomized and placebo-controlled (RELAX-AHF) trial (Teerlink et al., 2013), consisted in 1,161 acute heart failure patients; 581 patients treated with serelaxin and 580 patients receiving placebo, showing that the infusion of serelaxin for 48 h improved dyspnea, and reduced heart failure events, congestion, the length of hospital stay and the intensive care, as well as it reduced cardiovascular and all-cause mortality, blood pressure, and renal adverse events compared with placebo, independently of having preserved or reduced left ventricle ejection fraction (Teerlink et al., 2013; Filippatos et al., 2014).

Despite these encouraging results, a real-world patients (5,856) study designed to further analyze the RELAX-AHF results showed that only 23% of all consecutive patients hospitalized with acute heart failure met criteria of the RELAX-AHF trial, and that the mortality rates were lower in participants of ongoing randomized clinical trials in comparison with real-world acute heart failure patients (Spinar et al., 2017).

Recently, it has been developed the RELAX-AHF-2 study to corroborate the promising results of serelaxin observed in the RELAX-AHF. RELAX-AHF-2 is a multicenter, randomized, double-blind, placebo-controlled, event-driven, phase III trial involving ∼6,800 patients hospitalized for acute heart failure with persistent dyspnea and pulmonary congestion, elevated natriuretic peptide levels, mild-to-moderate renal impairment, and systolic blood pressure ≥125 mmHg. The primary objectives of this study are to probe that serelaxin is superior to placebo in decreasing 180 days cardiovascular death, and the reduction of occurrence of worsening heart failure through day 5. Key secondary endpoints include 180 day all-cause mortality, composite of 180 day cardiovascular death or rehospitalization due to heart/renal failure, and in-hospital length of stay during index acute heart failure (Teerlink et al., 2017). Although the results from this study have not been published yet, Novartis has recently provided a report announcing that the RELAX-AHF-2 do not confirm the efficacy of serelaxin in acute heart failure, so that it does not meet its primary endpoints of reduction in cardiovascular death through day 180 or reduced worsening heart failure through day 5 (Novartis provides update on Phase

# REFERENCES


III study of RLX030 (serelaxin) in patients with acute heart failure | Novartis)<sup>1</sup> . Thus, the real effect of serelaxin as an improver of heart failure should be deeply studied.

# CONCLUSION

Relaxin is a cardiometabolic hormone with important impact on the cardiovascular pathophysiology. Although relaxin beneficial effects on acute heart failure patients have been previously proved, nowadays its beneficial effect is under controversy due to the contradictory results found between the RELAX-AHF and both the RELAX-AHF-2 and a real-world patients study. Moreover, the precise mechanism through which relaxin act in different CMDs is not known yet, neither the mechanisms that regulate relaxin and its receptor expression in the different tissues in which they are produced. Furthermore, it was also suggested that different concentrations of relaxin can activate its receptor in a different way (Bathgate et al., 2013), so that the regulation of relaxin effects in different tissues depending on its concentration could be difficult to comprehend. Overall, it seems clear that relaxin is a new potential candidate as a therapeutic agent to treat/prevent cardiometabolic diseases, so that it has clear effects on vascular function, has positive chronotropic and inotropic effects in the heart, and prevents ischemia/reperfusion injury and atrial fibrillation. Although, further studies are needed, it also seems to be a potential regulator of metabolism, so it could regulate the metabolic disturbances observed in CVDs.

# AUTHOR CONTRIBUTIONS

SF, JG, and FL: Manuscript redaction and revision. AA, DR, MP, ER, and MR: Manuscript redaction.

# FUNDING

This work was supported by the "Fondo de Investigaciones Sanitarias del Instituto de Salud Carlos III" Madrid, Spain [PI15/00681 and CIBER de Enfermedades Cardiovasculares (CIBERCV)]; and European Regional Development Fund (FEDER).

Curr. Drug Targets 11, 122–135. doi: 10.2174/1389450107900 30992


<sup>1</sup>Novartis provides update on Phase III study of RLX030 (serelaxin) in patients with acute heart failure | Novartis Available online at: https://www.novartis. com/news/media-releases/novartis-provides-update-phase-iii-study-rlx030 serelaxin-patients-acute-heart [Accessed May 4, 2017].


relaxin in conscious rats. Am. J. Physiol. Regul. Integr. Comp. Physiol. 279, R1298–R1304.


to adult male rats inhibits food intake. Diabetes. Obes. Metab. 12, 1090–1096. doi: 10.1111/j.1463-1326.2010.01302.x


**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 Feijóo-Bandín, Aragón-Herrera, Rodríguez-Penas, Portolés, Roselló-Lletí, Rivera, González-Juanatey and Lago. 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.

# Mediators and Patterns of Muscle Loss in Chronic Systemic Inflammation

Sandra Pérez-Baos, Iván Prieto-Potin, Jorge A. Román-Blas, Olga Sánchez-Pernaute, Raquel Largo\* and Gabriel Herrero-Beaumont

Bone and Joint Research Unit, Service of Rheumatology, IIS-Fundación Jiménez Díaz, Autonomous University of Madrid, Madrid, Spain

Besides its primary function in locomotion, skeletal muscle (SKM), which represents up to half of human's weight, also plays a fundamental homeostatic role. Through the secretion of soluble peptides, or myokines, SKM interacts with major organs involved in metabolic processes. In turn, metabolic cues from these organs are received by muscle cells, which adapt their response accordingly. This is done through an intricate intracellular signaling network characterized by the cross-talking between anabolic and catabolic pathways. A fine regulation of the network is required to protect the organism from an excessive energy expenditure. Systemic inflammation evokes a catabolic reaction in SKM known as sarcopenia. In turn this response comprises several mechanisms, which vary depending on the nature of the insult and its magnitude. In this regard, aging, chronic inflammatory systemic diseases, osteoarthritis and idiopathic inflammatory myopathies can lead to muscle loss. Interestingly, sarcopenia may persist despite remission of chronic inflammation, an issue which warrants further research. The Janus kinase/signal transducer and activator of transcription (JAK/STAT) system stands as a major participant in muscle loss during systemic inflammation, while it is also a well-recognized orchestrator of muscle cell turnover. Herein we summarize current knowledge about models of sarcopenia, their triggers and major mediators and their effect on both protein and cell growth yields. Also, the dual action of the JAK/STAT pathway in muscle mass changes is discussed. We highlight the need to unravel the precise contribution of this system to sarcopenia in order to design targeted therapeutic strategies.

#### Keywords: skeletal muscle, turnover, anabolism, catabolism, sarcopenia, myokines, inflammation

Skeletal muscle (SKM) is a vast organ, which accounts for 40% of total weight in non-obese population (Janssen et al., 2000). The high metabolic activity of muscle cells, or myocytes, not only provides the necessary contraction for locomotion, but also fuels other organs' functions. Through the secretion of soluble peptides called myokines, the SKM interacts with surrounding fat, bones and skin, as well as with principal organs, including the cardiovascular system, brain, digestive tract and glands (Hartwig et al., 2014; Giudice and Taylor, 2017). In addition, some myokines exhibit autocrine/paracrine actions in the muscle, thereby helping sustain its normal growth (Pedersen and Febbraio, 2008). Considering the active interplay between SKM and other tissues, the impact of physical inactivity on general metabolism can be envisaged. In short, lack of exercise

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Luca De Toni, Università degli Studi di Padova, Italy Anna Picca, Università Cattolica del Sacro Cuore, Italy Massimo Negro, Centro di Medicina dello Sport Voghera, Università degli Studi di Pavia, Italy

> \*Correspondence: Raquel Largo rlargo@fjd.es

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 20 December 2017 Accepted: 04 April 2018 Published: 24 April 2018

#### Citation:

Pérez-Baos S, Prieto-Potin I, Román-Blas JA, Sánchez-Pernaute O, Largo R and Herrero-Beaumont G (2018) Mediators and Patterns of Muscle Loss in Chronic Systemic Inflammation. Front. Physiol. 9:409. doi: 10.3389/fphys.2018.00409

**155**

results in a lower insulin sensitivity, changes in the postprandial lipid profile, and accumulation of visceral adipose tissue (Pedersen and Febbraio, 2012). These effects are conceived to be the result of an evolutionary positive selection of pro-inflammatory pathways as well as of genes favoring gluconeogenesis, insulin resistance and fat storage. While formerly considered advantageous traits, in populations exposed to famines and epidemics, they have turned into a burden in modern ages, furthering cardiovascular diseases, because of lifestyle changes (Tuomilehto et al., 2001; Nocon et al., 2008).

Not only does SKM adjust the individual's metabolic activities, but it also suffers the consequences of perturbations in the systemic milieu. The SKM response to environmental cues is regulated by the hypothalamus, which integrates endocrine and immune signals, as well as information about physical activity and nutritional state (Clegg et al., 2013). Thus, levels of some nutrients –like glucose, fatty acids and amino acids– along with leptin and additional adiposity-related hormones (Roh et al., 2016) act as inputs for the elaboration of brain responses controlling energy expenditure, food intake, insulin secretion and glucose/fatty acid turnover in SKM (Roh et al., 2016). Muscle homeostasis requires fine-tuning of both protein turnover and cell growth in order to adapt its response to particular needs of the individual, without affecting muscle mass balance.

There is a dramatical shift in SKM homeostasis toward muscle loss during chronic inflammation. It is thought that an overacting immune system can divert energy expenditure and lead to a shortage of stored reserves affecting general metabolism (Straub, 2017). In spite of their relevance, these collateral effects are often forgotten in the assessment of prevalent conditions, while there is an increasing need of accurately measuring SKM response to stress induced by a variety of insults. However, the pathogenesis of systemic inflammation-related SKM damage, which has been termed cachexia or more accurately sarcopenia, has not until recently been understood. In this regard, sarcopenia can be observed in chronic debilitating diseases, as well as in typical inflammatory conditions, like rheumatoid arthritis, or in the proximity of injured joints, as in the case of osteoarthritis. It can also be the result of primary SKM autoimmune diseases. In the same way that the precipitating conditions are quite distinct in nature, so is the pattern of sarcopenia associated to them, while it also depends on the severity of the injury. All these factors make of sarcopenia a complex entity influencing general health which should be addressed in the therapeutic management of diseases.

This article gives an overview on triggers and mediators of SKM protein synthesis and cell turnover, especially focusing on clinical situations associated to muscle loss. The impact of systemic inflammation on muscle mass is discussed, looking into the molecular signals which disrupt muscle homeostasis in the different models of sarcopenia. In particular, the dual role of the Janus kinase/signal transducer and activator of transcription (JAK/STAT) pathway is underlined.

# MUSCLE PROTEIN TURNOVER

The major driver of SKM anabolic activity is the phosphatidylinositol 3-kinases (PIK3)/Akt signaling pathway. Along with exercise, insulin and insulin growth factor (IGF)-1 can induce phosphorylation of the insulin receptor substrate (IRS)-1 through the binding of specific receptors, subsequently activating PI3K (Sandri et al., 2013). In turn, the end product phosphatidylinositol-3, 4, 5 -trisphosphate (PI3P) facilitates membrane anchorage of Akt and its phosphorylation by 3′ phosphoinositide-dependent protein kinase-1 (PDK-1), upon which Akt enables activation of the mechanistic target of rapamycin (mTOR). The latter acts as downstream effector of the anabolic pathway through both the stimulation of ribosomal protein S6 kinase beta-1 (RS6K, or 70S6K), and the inhibition of 4E-binding protein 1 (4E-BP1) (Glass, 2003) (**Figure 1**). As we will discuss, Akt also shuts down the expression of forkhead O (FoxO), a transcriptional activator of musclespecific E3 ubiquitin ligases involved in protein catabolism (Mammucari et al., 2007; Schiaffino et al., 2013). On the other hand, testosterone can induce muscle hypertrophy through the PI3K/Akt/mTOR pathway or also using androgen receptors (Basualto-Alarcón et al., 2013; Hughes et al., 2016).

Conversely, myostatin (MSTN) –or growth differentiation factor-8 (GDF-8)– is a member of the transforming growth factor β (TGF-β) family, and a negative regulator of muscle growth. The MSTN-Smad2/3 route inhibits Akt-dependent protein synthesis and growth of mature muscle cells (Morissette et al., 2009; Schiaffino et al., 2013; Brooks and Myburgh, 2014; Retamales et al., 2015). This myokine binds to activin type 2 receptors (ACVR2) and promotes cachexia-related catabolism, via FoxO-dependent induction of atrogenes (McFarlane et al., 2006) (**Figure 1**). As it has been shown both in rodents and humans, training acts as a MSTN repressor. This fact has been put in relationship with the benefit of endurance exercise on metabolism. On the other hand, both muscle and serum of obese individuals appear to be comparatively enriched in MSTN (Hittel et al., 2009).

# Regulatory Networks of Protein Turnover

The concurrence of pro/anti-anabolic and pro/anti-catabolic factors, not only within the muscle milieu but also in the systemic environment, determines the yield of muscle turnover (Reed et al., 2012; Schiaffino et al., 2013). Usually, up-regulation of

**Abbreviations:** PDK-1, 3′ -Phoshphoinositide-Dependent Protein Kinase-1; 4E-BP1, 4E-Binding Protein 1; PRR, Activate Pattern Recognition Receptors; ActRII, Activin Type II Receptors; ADM, Amyopathic Dermatomyositis; ATG, Autophagy-Related Genes; CRP, C-Reactive Protein; CA, Chronic Arthritis; ER, Endoplasmic Reticulum; ERK, Extracellular Signal–Regulated Kinase; FoxO, Forkhead O; GDF-8, Growth Differentiation Factor-8; GH, Growth Hormone; IMM, Idiopathic Inflammatory Myopathies; IGF, Insulin Growth Factor; IRS-1, Insulin Receptor Substrate-1; IFN, Interferon; IL, Interleukin; JAK, Janus Kinase; MHC, Major Histocompatibility Complex; mTOR, Mammalian Target of Rapamycin; MAPK, Mitogen Activated Protein Kinases; MAFbx, Muscle Atrophy F-Box Protein; MuRF-1, Muscle Ring Finger-1; MyoD, Myogenic Differentiation; Myf5, Myogenic Factor 5; MSTN, Myostatin; NFκB, Nuclear Factor KB; OA, Osteoarthritis; 70S6K, P70-S6 Kinase; Pax7, Paired Box Transcription Factor 7; PI3K, Phosphatidyilinositol 3 Kinase; PS, Primary Sarcopenia; RA, Rheumatoid Arthritis; RS, Rheumatoid Sarcopenia; RS6K, Ribosomal Protein S6 Kinase Beta-1; SC, Satellite Cells; SS, Secondary Sarcopenia; STAT, Signal Transducer and Activation Of Transcription; SKM, Skeletal Muscle; TLR, Toll-Like Receptors; TGF, Transforming Growth Factor; TNF, Tumor Necrosis Factor; UPS, Ubiquitin-Proteasome System

IL-6-dependent activation of STAT-3 contributes to myogenic differentiation and SC proliferation in homeostatic conditions. Myostatin (MSTN) activates Smad2/3, which activates the ubiquitin proteasome system (UPS). Smad2/3 also inhibits the PI3K/Akt/mTOR signaling pathway, thus diminishing myoblast differentiation and protein synthesis. UPS is also activated by Forkhead O (FoxO), p38 mitogen activated protein kinases (MAPK) and nuclear factor κB (NFκB), which also induces a decay in MyoD mRNA, thus leading to a decrease in myoblast differentiation. IL-6-induced JAK2/STAT-3 signaling impairs myogenesis in a catabolic scenario.

involved in protein catabolism. IRS-1 also activates the extracellular signal–regulated kinase (ERK), leading to satellite cell self-renewal and myoblast proliferation.

protein synthesis is accompanied by a reduction of protein degradation (Schiaffino et al., 2013). Likewise, under catabolic stress, a well-orchestrated network blocks new protein synthesis in order to limit energy expenditure. Nonetheless, if conditions are favorable, the amino acids released during proteolysis will boost new protein synthesis (Tran et al., 2007; Zoncu et al., 2011; Perl, 2015).

The role of microRNAs in muscle anabolic fine-tuning is noteworthy. MicroRNAs are small non-coding RNAs which act as negative regulators of gene expression. Muscle specific microRNAs, known as myomiRs, target myostatin, TGF-β and Akt-dependent pathways. In turn, these factors modify the expression rate of different myomiRs (Butz et al., 2012; Hitachi and Tsuchida, 2014; Jung and Suh, 2015). Their isolation from muscle exosomes suggests that myomiRs can be released and exert autocrine/paracrine actions in surrounding cells, in this way facilitating a synchronous muscle growth response (Demonbreun and McNally, 2017; Fry et al., 2017).

# CELL GROWTH

# Myogenesis

Although anabolic factors have commonly a role in myogenesis, processes of cell growth and protein synthesis have to be considered as independent parts of the SKM metabolism. In particular, myogenesis accounts for the capacity of cell renewal and differentiation during repair processes.

This capacity is based on the existence of a niche of quiescent myogenic cells, known as satellite cells (SC), between the outer coat of sarcolemma and the basal lamina of myofibers (Morgan and Partridge, 2003). These cells are characterized by the expression of paired box transcription factor (Pax) 7 and myogenic factor (Myf) 5 and have a crucial role in selfrenewal of SKM. Upon injury, SC undergo proliferation and differentiation into myoblasts (Pax7+/Myf5+/MyoD+), which eventually lose expression of Pax7 and upregulate myogenin (Pax7-/Myf5+/MyoD+/Myogenin+). The latter determines cessation of cell proliferation and progression to terminal differentiation, maturation and fusion of cells into new myofibers (Morgan and Partridge, 2003).

The JAK/STAT pathway and its triggering cytokine IL-6 stand as key myogenic factors. The intracellular network exerts a relevant homeostatic role in healthy SKM. Normal proliferating myoblasts have been found to exhibit high levels of phospho-STAT-3, both in vivo (Kami and Senba, 2002) and in culture (Spangenburg and Booth, 2002; Yang et al., 2009). According to the studies performed by Hoene and co-workers in primary mouse myoblasts and in C2C12 cells, the activation of STAT-3-SOCS-3 in response to an autocrine action of IL-6 is needed for their differentiation (Hoene et al., 2013). In the same way, IL-6 stimulation of STAT-3 is required for SC proliferation following SKM overload. Moreover, not only does IL-6 stimulate myogenesis, but it also exerts some key metabolic actions (Pedersen and Febbraio, 2008; Muñoz-Cánoves et al., 2013) and blocks the production of pro-apoptotic cytokines like tumor necrosis factor alpha (TNF) (Schindler et al., 1990). Altogether, these actions place IL-6 as a pivotal myokine released from muscle cells upon exercise.

Myoblast proliferation is subject to a tight control, which prevents alterations in muscle mass volume during repair processes (Verdijk et al., 2007; McFarlane et al., 2011; Fry et al., 2015; Demonbreun and McNally, 2017). Experiments conducted in the rodent C2C12 and L6 myoblast cell lines have shown that IGF-1 plays opposite roles at distinct stages of myocyte differentiation. While the peptide initially exerts a mitogenic effect, it inhibits growth and promotes cell differentiation at later phases (Engert et al., 1996; Florini et al., 1996). In turn, these polar activities are driven by different signaling pathways. Extracellular signal-regulated kinase (ERK) appears to be involved in cell proliferation (Rosenthal and Cheng, 1995; Coolican et al., 1997), whereas PI3K/Akt/p70S6K has been shown to mediate IGF-1-induced myogenin expression (Xu and Wu, 2000) (**Figure 1**). Similarly, the JAK/STAT pathway may promote either proliferation or differentiation depending on the molecular isoforms involved in its activation, as we will discuss (Sun et al., 2007; Wang et al., 2008; Diao et al., 2009).

# REGULATORY CATABOLIC PROCESSES

Metabolic disturbances and stress trigger numerous repair mechanisms in SKM aimed at restoring homeostasis. Intracellular proteolytic complexes, including calpains, the endoplasmic reticulum (ER) stress response, caspase cascades, the autophagic machinery and the ubiquitin-proteasome system (UPS) can participate in such response. Dysregulation and/or perpetuation of these reparative mechanisms may result in muscle wasting, either by increasing proteolysis or initiating muscle cell apoptosis.

# The Ubiquitin-Proteasome System (UPS)

The UPS accounts for the principal mechanism of protein degradation in muscle cells (Glass, 2010; Milan et al., 2015). Two muscle-specific E3 ligases, muscle ring finger-1 (MuRF-1) and muscle atrophy F-box protein (MAFbx) –or atrogin-1– (Schiaffino et al., 2013) have been identified and their expression is commonly used as a marker of UPS activity in secondary sarcopenia models (Milan et al., 2015). UPS is regulated by FoxO (Sandri et al., 2004; Tournadre et al., 2017), p38 mitogen activated protein kinase (MAPK) (Zhang and Li, 2012), STAT-3 (Bonetto et al., 2012) and nuclear factor κB (NFκB) (Cai et al., 2004). Furthermore, STAT-3 and NFκB down-regulate MyoD expression, and, consequently, decrease myoblast differentiation (Guttridge et al., 2000) (**Figure 1**). While FoxO is blocked by Akt, the other mediators are activated in response to proinflammatory molecules (Schiaffino et al., 2013). Notably, insulin peripheral resistance, which shuts the IGF1-Akt pathway down, also raises UPS activity (de Alvaro et al., 2004; Brown et al., 2009) (**Figure 1**).

# Autophagy

Autophagy is carried out by a cascade of proteolytic reactions which degrade and recycle malfunctioning organelles, proteins and other cytoplasmic molecules, using the lysosomal machinery and a multimolecular complex (Mizushima and Komatsu, 2011). This highly conserved mechanism of homeostasis allows cells to survive under stress conditions (Cuervo, 2008). In addition, defects in autophagy associate with muscle disease and inflammation (Levine and Kroemer, 2008). However, overinduction of autophagy is also responsible for muscle loss. Up to 35 different autophagy-related genes (ATG) encoding the autophagy machinery have been found. The inhibition of mTOR complex (mTORC)1 activates the initiation steps of autophagy, inducing autophagosome nucleation and substrate recognition. This is followed by the recruitment of different ATG products that facilitate protease digestion of the cargo (Portal-Núñez et al., 2015). Likewise, muscle mass can also be regulated by mitochondrial dysfunction, ER stress or myocyte apoptosis (Nagaraju et al., 2005; Busquets et al., 2007; Nogalska et al., 2007; Marzetti et al., 2009, 2010; Deldicque, 2013).

# Endoplasmic Reticulum (ER) Stress

ER and mitochondria cooperate in SKM homeostasis. Disturbances of their balance can be induced by protein misfolding, starvation or energy deprivation, and lead to an ER stress response. Consequently, ER activates the so-called 'unfolded protein response', consisting in the upregulation of chaperones and other enzymes that participate in adequate protein folding. If the insult persists, the stress response can be unable to maintain cell homeostasis and give rise to the activation of NF-κB and other inflammatory pathways. Uncontrolled ER stress eventually lead to myocyte death by mechanisms which include apoptosis, autophagy and necrosis (Deldicque, 2013). Furthermore, it has been proposed that ER stress could indirectly contribute to muscle wasting through the blockage of mTORC1, thus creating a state of anabolic resistance (Deldicque, 2013).

# Apoptosis

Myofiber loss accounts for the key process leading to sarcopenia during aging, skeletal muscle immobilization, muscular dystrophy and other inflammatory conditions (Dupont-Versteegden, 2006). The specific mechanisms of cell death in SKM remain largely unknown and will not be addressed in this review. Noteworthy, Dupont-Versteegden and coworkers suggested that classical apoptotic pathways could be of little relevance in muscle. Alternatively, additional molecules might exert a key role in triggering regulatory cell death processes (Dupont-Versteegden, 2006). Of note, being myofibers multinucleated syncytia, nuclear elimination does not necessarily carry cell destruction. In this regard, muscle loss can rely on loss of myonuclei or of their associated cytoplasmic domains (Allen et al., 1999).

# SARCOPENIA, THE SKM RESPONSE TO SYSTEMIC INFLAMMATORY STRESS

Systemic inflammation affects all body systems and organs, including the SKM. Muscle response can follow different patterns, namely primary sarcopenia, secondary sarcopenia, and those respectively found in osteoarthritis and in idiopathic inflammatory myopathies. These four inflammation-related clinical settings of muscle wasting are summarized in **Table 1** and will be addressed below.

# Primary Sarcopenia

The condition known as primary sarcopenia (PS) is associated with the aging and the frailty syndrome and is defined as a progressive and generalized loss of SKM mass and/or strength leading to a significant functional impairment (Cruz-Jentoft et al., 2010). This muscle response is considered to be the consequence of aging/disease interactions at multiple systems (Cruz-Jentoft et al., 2010). Thus, PS identifies a poor health status frequently associated with disability, increasing risk of falls and fractures, and potentially dragging the elderly to dependence. On these grounds, its appearance leads to a decreased life expectancy (Cruz-Jentoft et al., 2010).

This type of sarcopenia represents a failure in muscle anabolic processes. In addition, it is likely that a mild persistent inflammatory status could play a role in its pathogenesis, promoting a catabolic scenario (Ali and Garcia, 2014).

Morphologically, atrophy of type II (fast-twitch, highly ATPconsuming) fibers can be observed, following a loss of myocyte proteins, organelles and cytoplasm size (Muscaritoli et al., 2010), and the accumulation of muscle fat. The origin of PS has been put in relationship with age-dependent functional changes in mitochondria, ER, cells and tissues. Furthermore, as suggested by studies conducted in aged mice, the niche of SC could decline with age and be insufficient for nuclear replacement (Brack et al., 2005). Additional factors are related to lack of mobility, neurodegeneration, nutritional deficiencies and hormone decrease. Progressive testosterone deficiency increases peripheral resistance to insulin and IGF-1 –both potent activators of the Akt/mTOR pathway– yielding a lower synthesis and a higher degradation of muscle proteins, through the activation of FoxO (Sandri et al., 2004). Likewise, an age-dependent impairment in the GH/IGF-1 axis has been reported. A fall in gene expression of GH receptors in the elderly inversely correlates with serum MSTN levels (Perrini et al., 2010). This fact could hamper both synthetic processes and cell renewal (Taylor et al., 2001) –and the same effects could be expected from the lowering of sex hormone levels –. Of all myokines, IGF-1 is regarded as a pivotal mediator of muscle growth because of its effect on SC proliferation, and its concentration is inversely associated with the development of PS. However, recent experiments have suggested that deficiency of GH/IGF-1 could increase longevity in animals (Sattler, 2013). Indeed, some human studies have also drawn controversial results, a matter which has diverted the attention from the GH/IGF-1 axis to MSTN up-regulation, as the cornerstone of the anti-anabolic response of PS.

On the other hand, although clearly regarded as an anabolic resistant process, there is also an age-related low-grade chronic inflammation that may contribute to muscle wasting in PS (Clegg et al., 2013) through the up-regulation of pro-inflammatory cytokines, like TNF, IL-1 and IL-6 (Krabbe et al., 2004; Maggio et al., 2006; Drummond et al., 2009; Ali and Garcia, 2014). This view is supported by the finding of differential levels of serum biomarkers between active and non-active elders, with an enhancement of IL-8, myeloperoxidase and TNF in the latter group (Marzetti et al., 2014).

Increased levels of IL-6 appear to be particularly associated to a higher disability in the elderly (Barbieri et al., 2003; Maggio et al., 2006). In spite of its homeostatic role in SKM, both driving myogenesis and mediating IGF-I anabolic actions (Barbieri et al., 2003; Maggio et al., 2006; Mammucari et al., 2007; Schiaffino et al., 2013), over-expression of IL-6 is known to impair myocyte functions (Roubenoff, 2014). Indeed, its levels along with those of TNF-R1 out of a group of 15 different NF-kB-dependent molecules, have been found to be the best predictor of mortality in the elderly (Varadhan et al., 2014). Altogether this suggests that the inflammatory response could account for a therapeutic target in PS.

# Secondary Sarcopenia (SS)

Secondary sarcopenia (SS) occurs in the context of chronic illnesses, like cancer, renal/respiratory failure or inflammatory diseases. A paradigm of this SKM response is rheumatoid sarcopenia –or sarcopenia during rheumatoid arthritis– (RS) (Morley et al., 2006; von Haehling et al., 2016). Sarcopenia is a prominent feature of the more generalized rheumatoid cachexia syndrome, thus termed because it targets major organs and immune cells along with the SKM, leading to a profound loss of cell mass in all these tissues. In this regard, two types of cachexia have been associated to RA; the 'classic' type –which resembles those found in cancer and AIDS– and shows both muscle and fat mass loss, and the more typical 'rheumatoid' type –which results in a reduced muscle mass but an increase in fat volume– (Giles et al., 2008; Lemmey et al., 2009; Summers et al., 2010; Tournadre et al., 2017). Other forms of SS are also frequently referred to as cachexia (Roubenoff et al., 1992), due to their predominant catabolic component (Masuko, 2014). However, there are marked differences between RS and the other forms of SS (Summers et al., 2010) because of the high inflammatory status which characterizes RS, and the relevant role of the adaptive immune system in this disease.

The prevalence of RS cannot be determined with accuracy, largely due to a lack of consensus in establishing a clear cutoff in body composition for its diagnosis. Besides, the syndrome can pass unnoticed in patients with stable weight resulting from the increase in fat volume, which can mask muscle mass loss (Summers et al., 2010). It has though been estimated to be present in 38% of patients with active rheumatoid arthritis and in 10–20% of those with well-controlled disease (Elkan et al., 2009; Phillips et al., 2009; Summers et al., 2010). The appearance of RS does not carry the same poor prognosis as SS associated to AIDS or cancer. Still, its development impacts both life expectancy and quality of life, particularly due to its association with metabolic syndrome, cardiovascular disease, and weakness, in an independent fashion of disease severity (Kotler, 2000; Walsmith and Roubenoff, 2002; Fukuda et al., 2013).

Why does RS run an independent course from disease activity has been difficult to understand, since the condition is principally unchained by systemic inflammation. It appears that the concurrence of a variety of signals triggers RS, the most relevant of them being inflammation-driven increased metabolism, a reduced anabolic activity, the coexistence of malnutrition, peripheral insulin resistance and lack of exercise. Although RS has not been primarily associated with an impairment of anabolic processes, low IGF-1 has been shown to parallel muscle mass loss in experimental models (Soto et al., 2001; Castillero et al., 2009a)


and also in the patients, after adjusting for age, sex, and adiposity (Lemmey et al., 2009; Baker et al., 2015). In addition, it has been proposed that rheumatoid arthritis-associated insulin resistance indirectly promotes muscle wasting, because of the physiological anti-catabolic effect of insulin (Walsmith and Roubenoff, 2002). However, further studies are needed to clarify the involvement of this pathway in RS (Lemmey et al., 2009).

Pro-inflammatory cytokines such as TNF, IL-1β and IL-6, have been proved to exert a critical contribution to the hypercatabolic state found in these patients (Roubenoff, 2014). Although RS pathogenesis is poorly understood, elevated serum levels of these cytokines are known to activate UPS. In experimental cachexia, an upregulation of muscle-specific E3 ubiquitin ligases MuRF-1 and atrogin-1 has been shown, followed by an increased myofibrillar proteolysis, through the activation of NF-κB (Li et al., 2009; Varadhan et al., 2014; Little et al., 2017). These data are consistent with the over-activation of UPS during SS, as has also been found in muscle wasting associated to experimental arthritis (Castillero et al., 2009a; Little et al., 2017). In the same way, the increased activity of FoxO resulting from MSTN-dependent PI3K/Akt inhibition enhances the expression of atrogenes, thus promoting protein catabolism during cachexia (McFarlane et al., 2006); however, the role of this transcriptional regulator in RS is not well defined.

Our group recently conducted studies to look into muscle response to chronic arthritis in a rabbit model which emulates RS. The diseased animals had a reduction in weight and muscle size, and an up-regulation of atrogin-1 both in muscle and in synovium, altogether indicating an increased protein breakdown. Strikingly, there was a paradoxical decrease in MSTN expression, along with a reduction in phospho-STAT-3 levels, which pointed to the existence of a compensatory anabolic activation (Little et al., 2017). This pattern of response suggests that the inflamed muscle could contribute to the process of SS through a mechanism of autocrine atrophy triggered by the release of muscle-derived, pro-inflammatory mediators (Little et al., 2017).

It can be envisaged how devastating the effects of atrogenes can be on the musculature of elderly RS patients (Guadalupe-Grau et al., 2015). Also interesting is that both muscle mass loss and IGF-1 content are remarkably returned to normal in patients with RS following a high intensity training program (Lemmey et al., 2009). This underlines the therapeutic goal of breaking down anabolic resistance in RS. On the other hand, standard antirheumatic therapies should not be expected to prevent RS, since changes in body composition can be observed in patients with low disease activity (Metsios et al., 2007). Previous work in our laboratory disclosed a protective role of celecoxib in the development of sarcopenia in rabbits (Romero et al., 2010). The selective inhibition of COX-2, and COX-2 derived products such as PGE2, yielded a reduction in both systemic inflammation and NF-kB activation, together with an amelioration of weight loss in arthritic rabbits. As suggested by previous studies, the inhibitory effect of NSAIDs on NF-kB signaling might be responsible for the suppression of muscle wasting induced by the activation of the ubiquitin-proteasome pathway (Wyke et al., 2004).

As regards myogenesis impairment, there are no conclusive data about its direct contribution to RS. Low MyoD and myogenin levels have been observed in a rat model of arthritis, arguing for this possibility (de Oliveira Nunes Teixeira et al., 2013). In this regard, both TNF and IFNγ can impair myogenesis, as shown in C2C12 myoblasts and in mouse muscle, where the cytokines suppressed MyoD and myogenin through an NFκB-dependent pathway (Guttridge et al., 2000). By contrast, Castillero and co-workers showed increased expression of these myogenic factors in arthritic rats (Castillero et al., 2009a,b).

# OA Sarcopenia

The effect of osteoarthritis (OA) represents a third subset of muscle metabolic adaptive response to systemic inflammation. A variety of factors are involved in the pathogenesis of this complex syndrome, including biomechanical stress, senescence, hormonal status and inflammatory mediators, while the genetic background could also be relevant (Herrero-Beaumont et al., 2009). While all joint tissues are targeted by OA, the initial insult can be localized at any of them, and render different syndromes (Roman-Blas et al., 2016). As the disease progresses into advanced phases, the OA syndrome is usually more uniform, although interspersed flares in relationship with acute injuries, can occur (Castañeda et al., 2014). On the whole, the disease can be considered a protean long-course process, with many possible shifts of phenotype along time.

In the same way, inflammation in OA can take many forms. Typically, there is a low-grade inflammatory status in the OA joint, which can get temporarily higher in response to different triggers, such as capsular sprains, micro-trauma, and the presence of crystals or additional danger associated molecular patterns (DAMPs). All these factors are known to couple innate receptors, such as PI3K and the toll-like family (TLRs) (Gómez et al., 2015). Alternatively, biomechanical sensors can also contribute to these flares, since their signaling network interlinks with major inflammatory pathways.

As regards OA-related sarcopenia, it does not only target the neighboring muscles of affected joints but can also involve the whole SKM. However, the underlying mechanisms regulating joint-muscle crosstalk are not yet fully understood. A close relationship between impaired SKM functions and knee OA has been observed, along with an enhanced expression of FoxO1 reflecting lower muscle strength (Levinger et al., 2016). Since cartilage and SKM cells share some cellular pathways, paracrine communication between them remains conceivable, in addition to the influence of a close anatomical proximity (De Ceuninck et al., 2014). On the other hand, individual factors associated to OA, such as lack of physical activity and obesity can play an indirect role in the pathogenesis of OA-related sarcopenia, while they also account for an increased cardiovascular risk and a shortened life expectancy (Yoshimura et al., 2012). This places adipokines as principal actors in this model of SKM adaptive response to 'low-grade inflammation' (Scotece et al., 2011, 2017). Adipokines are hormone-like fat-derived factors which contribute to maintain the low-grade inflammatory status in patients with metabolic syndrome (Gómez et al., 2015). Adiponectin and leptin are by far the best characterized adipokines. They both stimulate glucose uptake and fatty acid oxidation both in muscle and in adipose tissue (Kalinkovich and Livshits, 2017). Of them leptin exerts a pro-inflammatory role and adiponectin is mainly regulatory. While serum levels of adiponectin decrease in relationship with OA, age and obesity, serum leptin is enhanced in patients with OA in parallel with the accumulation of adipose mass (Poonpet and Honsawek, 2014; Kalinkovich and Livshits, 2017). Other relevant adipokines reported to be increased in OA are resistin (Koskinen et al., 2014), which drives pro-inflammatory actions in human SKM (Carey et al., 2006) and chemerin (Ma et al., 2015), which has been shown to inhibit myogenesis and induce adipogenesis in C2C12 myoblasts (Li et al., 2015).

Both OA and sarcopenia are prevalent in the elderly (Felson et al., 1987; Fielding et al., 2011). OA is frequently considered as part of the metabolic syndrome and of senescence (Herrero-Beaumont et al., 2009). The severe peri-articular sarcopenia found in OA could be partly due to inactivity, but also due to low-grade persistent systemic inflammation, which is a feature of both syndromes (Krabbe et al., 2004; Scanzello and Loeser, 2015). In fact, sarcopenic obesity is more closely associated with knee OA than non-sarcopenic obesity, thus supporting the tight relationship between muscle metabolism and inflammation in this disease (Lee et al., 2012).

It has been suggested that sarcopenia could not only be a trigger of OA, but also a worsening factor for its progression (De Ceuninck et al., 2014). Muscle transcription factors associated to inflammation, such as STAT-3 and NFκB, correlate with the grade of joint dysfunction, disability and gait impairment in the patients (Levinger et al., 2011b). Similarly, an inverse correlation has been observed between muscular resistance of hamstrings and serum IL-6 levels in elderly women with OA (Santos et al., 2011), suggesting a role for this cytokine in OA sarcopenia. In fact, Levinger and co-workers found augmented levels of IL-6, STAT-3, SOCS-3 and NFκB, among others, in the vastus lateralis of patients with knee OA (Levinger et al., 2011a). The relevance of IL-6 in OA pathogenesis is further sustained by its association with cartilage loss and radiographic knee OA (Livshits et al., 2009; Stannus et al., 2010). The cytokine could therefore account for a pivotal link between OA and sarcopenia, and provide an attractive therapeutic target, although further research is warranted.

# Sarcopenia Associated With Idiopathic Inflammatory Myopathies (IMM)

Finally, a distinct pattern of muscle wasting can be observed in patients with idiopathic inflammatory myopathies (IMM). These systemic autoimmune diseases are characterize by weakness, muscle inflammation (Day et al., 2017) and fat mass gain (Cleary et al., 2015), without an increase in serum CRP levels (Hanisch and Zierz, 2015; Malik et al., 2016). Both the innate and the adaptive immune systems are involved in their pathogenesis. On one hand, there is a deep disturbance in the adaptive immune responses, with activation of auto-reactive cytotoxic T cells along with production of autoantibodies (Dalakas, 2010). On the other, an infiltration of inflammatory cells is responsible for cell death, and disruption of the normal muscle architecture. Local production of cytokines participates in IMM pathogenesis, both promoting cell damage and impairing muscle cell function, although some of them, such as IL-1, TNF or IL-15, appear to have a role in repair stages as well. Therefore, the complexity of muscle cytokine networks deserves especial attention at designing therapeutic strategies (Loell and Lundberg, 2011). Also to underline is the fact that muscle function restriction often persists in the patients in the long-term, despite the achievement of remission with immunosuppressive treatments (Loell et al., 2016). There appears to be therefore a permanent footprint of a previous muscle injury, which might also be present in the other three subsets of sarcopenia.

In these heterogeneous muscle disorders myocyte degeneration and necrosis occur mainly due to directed self-reactivity of CD8+ T cells. Sarcolemma disruption provokes the release of myoglobin and creatine kinase (Chargé and Rudnicki, 2004). Antigen-specific cytotoxic T lymphocytes (CTL) migrate through the endothelial wall and directly bind to muscle fibers aberrantly expressing major histocompatibility complex (MHC)-I molecules on their surface, through their T-cell receptors. Upon presentation of muscle antigens, these infiltrating CD8+ T cells undergo clonal expansion (Rayavarapu et al., 2013). Histologically, affected muscles are characterized by perivascular cell infiltration, predominantly consisting of CD8+ T cells invading and surrounding healthy-appearing muscle fibers. Direct cytotoxicity to muscle cells can then take place through the release of perforin granules. Cell infiltration is located in different regions of the muscle fascicles (i.e. interfascicular septae, periphery of the fascicle, epimysium, endomysium, etcetera) depending of the clinical subgroup (Vattemi et al., 2014). Cytokines, such as IFN-γ, IL-1, and TNF released by activated T cells, may enhance MHC class I up-regulation and T-cell cytotoxicity. There is also a shift toward Th17 CD4+ T cell differentiation, which furthers the autoimmune process (Tews and Goebel, 1996). However, as mentioned before, it has been reported that IL-1 and TNF may exert a role in muscle regeneration (Loell and Lundberg, 2011). Some of the above-mentioned cytokines signal through the NF-kB pathway and/or are controlled by this transcription factor, which indeed seems to have a key pathogenic role in IIM. It is well known that NF-kB becomes activated both in inflammatory cells and in myocytes and enhances MHC-I expression in muscle fibers. This event has been associated with ER stress, which fuels MSTN production and muscle injury (Tews and Goebel, 1996; Nagaraju et al., 2005; Nogalska et al., 2007; Creus et al., 2009). As has been previously mentioned, NF-kB is known to induce atrogene expression and loss of MyoD messenger RNA (Guttridge et al., 2000; Cai et al., 2004). On the other hand, MSTN exerts important anti-anabolic effects by blocking the PI3k/Akt/mTOR signaling pathway (Morissette et al., 2009; Schiaffino et al., 2013; Brooks and Myburgh, 2014; Retamales et al., 2015). In conclusion, these procatabolic and anti-anabolic mediators can elicit IIM-dependent sarcopenia.

# THE DUAL ROLE OF JAK/STAT AND IL-6 IN SKM

While the JAK/STAT pathway has been reported to play a crucial role in myogenesis, its precise contribution to muscle wasting and regeneration is yet to be defined. According to results drawn by different studies, it can be argued that the activation of JAK/STAT results in different effects in healthy and in injured muscle cells. This is likely due to the magnitude and duration of the induction.

As previously mentioned, the IL-6/STAT-3/SOCS-3 axis contributes to myogenic differentiation in physiological situations (Hoene et al., 2013). By contrast, experimental murine models of rheumatoid cachexia exhibit a profound total and muscle and total weight loss and elevated IL-6 levels are consistently found along time. IL-6 involvement could be determinant to muscle atrophy (Bonetto et al., 2012), and drive muscle wasting through the activation of UPS (Bodell et al., 2009). Furthermore, chronically elevated IL-6 levels might reflect a feedback mechanism triggered by an impairment in IL-6 dependent signaling. To our knowledge, whether this phenomenon of 'IL-6 resistance' takes place during muscle wasting has not been addressed. However, it is conceivable that IL-6 deregulation during disease could follow a similar pattern to those of insulin or leptin tissue resistance.

Hypothetically, during RS this phenomenon could be due to an overexpression of SOCS-3 in muscle cells. Consistently, muscle SOCS-3 overexpression is known to abrogate leptindependent STAT-3 phosphorylation, in this way favoring leptin resistance (Bjørbaek et al., 1998). Interestingly, increases in SOCS-3 could also inhibit the insulin receptor, thus promoting insulin resistance as well (Ueki et al., 2004). Also of interest would be the assessment of soluble gp130/sIL-6R in RS muscle, since this decoy receptor could account for an additional mechanism of IL-6 hyporesponsiveness (Jostock et al., 2001). The conflicting effects of IL-6 on muscle certainly represent a novel avenue for further research, which could be termed the 'IL-6 paradox'.

In line with this findings, Tierney et al. and Price et al. reported that the enhancement of JAK2/STAT-3 signaling impaired myogenesis in aging mice, likely due to a loss of SC self-renewal capability (Tierney et al., 2014; Price et al., 2015). Subsequently, STAT-3 has been hold responsible for driving SS. Bonetto and co-workers found that STAT-3 and its responsive genes were upregulated in mice with cancer-associated cachexia (Bonetto et al., 2011). In these mice, they found that constitutive activation of the transcription factor worsened their wasting status, and that JAK1/2 or STAT-3 blockade could revert this situation (Bonetto et al., 2012). As has been shown in transfected C2C12 myoblasts, overexpression of STAT-3 results in its direct interact with MyoD, thus inhibiting myogenic differentiation. Reciprocally, MyoD was shown to decrease STAT-3 activity (Kataoka et al., 2003).

Although there are scant studies looking into this signaling cascade in OA related sarcopenia, a role of the IL-6/STAT-3/SOCS-3 axis in muscle wasting following 'chronic lowgrade inflammation' can be foreseen. Of note, in addition to experimental findings, levels of these mediators were increased in the vastus lateralis of patients with knee OA (Levinger et al., 2011a).

Globally, these data point to a role of STAT-3 in normal regeneration of healthy muscle. Nonetheless, its chronic activation during aging or chronic inflammation could carry an impairment of SC self-renewal, a lower proliferation of myoblasts and an overall abnormal muscle repair.

Among all the JAK and STAT molecular mediators, STAT-3 has received major attention in the study of healthy and damaged muscle. This transcription factor appears, nonetheless, to play distinct roles on myogenesis depending on the JAK protein kinase activated upstream. In brief, the JAK2/STAT2/STAT-3 pathway has been found to exert a pro-differentiation effect in primary muscle cells and C2C12 myoblasts (Wang et al., 2008), whereas JAK1/STAT-1/STAT-3 was shown to boost myoblast proliferation through the regulation of cell cycle-associated genes' expression. As reported, the latter could also prevent myoblast premature differentiation, thus acting as a checkpoint during myogenesis (Sun et al., 2007). In a subsequent study, the same group observed that SOCS-1 and SOCS-3 intercepted JAK1/STAT-1/STAT-3 signaling, thereby promoting myogenic differentiation (Diao et al., 2009).

Very little is known about JAK/STAT activation in IIM. According to Illa and co-workers patients with dermatomyositis could show an increased STAT-1 activation in myofibers (Illa et al., 1997). Presumably, the enhancement of JAK/STAT dependent transcription by local cytokine networks drives muscle wasting in IIM. The overall extent of muscle wasting is likely determined by activation of specific JAKs and STATs triggered by a particular group of cytokines. Whether these changes can be reverted with treatment, as well as the effects of different JAK inhibitors depending on their specificity for particular JAK/STAT cascades, deserves further research.

# CONCLUSIONS AND FUTURE PERSPECTIVES

A tight regulation of protein turnover and cell growth is crucial to maintain homeostasis in SKM. Chronic systemic inflammation, however, provokes a dramatical shift in this balance, thus compromising SKM mass. Behind this fine-tuning, there is a very intricate network of catabolic and anabolic signaling pathways, which are summarized in **Figure 1**. In short, induction of the Akt/mTOR pathway not only enhances myogenesis and protein synthesis, but also slows down catabolic UPS and the autophagic machinery. Conversely, MSTN exerts important antianabolic and pro-catabolic effects, through the suppression of the Akt/mTOR axis and the induction of UPS, respectively. Pro-inflammatory cytokines mostly act as pro-catabolic and anti-anabolic factors, with IL-6 exerting a dual role in SKM turnover.

It has not been until recently that the pathogenesis of chronic inflammation-related sarcopenia has been understood. Consequently, the SKM response to the inflammatory milieu is often unadvertised in the assessment of prevalent conditions. In addition, despite the recent advances providing insight on the relationship between chronic inflammation and sarcopenia, there are still substantial gaps, which may account for the lack of treatments for sarcopenia. Intriguingly, muscle mass loss does not seem to revert despite the achievement of remission. Therefore, the mechanisms behind this permanent footprint need to be approached in future studies.

While innate and adaptive immune responses are extremely energy-consuming, metabolic inflammation does not lead to an increased energy expenditure. Notwithstanding this fact, metabolic inflammation is often said to resemble a smoldering fire, which is difficult to extinguish (Hotamisligil, 2006). Moreover, chronic low-grade inflammation and metabolic dysfunction drive the development of the most prevalent chronic diseases, particularly those targeting the musculoskeletal system such as OA (Hotamisligil, 2006; Robinson et al., 2016). Of interest, they also significantly contribute to the fragility syndrome of the elderly. As is widely acknowledged, major factors associated with accelerated atherosclerosis are the components of the metabolic cluster, namely hypertension, high blood glucose, excess in waist fat, and abnormally increased lipid levels. Obesity, which shows a higher prevalence in OA than in non-OA age-matched individuals, could be the most relevant of these traits.

Additional pathways of muscle loss are unchained by anorexia, asthenia and inactivity, all of which are typical features of chronic systemic inflammation (Phillips et al., 2009). These processes are, in fact, shared by the four previously mentioned clinical settings (Huffman et al., 2017). In this regard, as has been recently reviewed in depth by Dalle and coworkers, exercise and dietary interventions have proved beneficial against the anabolic resistance of the elderly (Dalle et al., 2017). On the other hand, considering the impact of chronic inflammation on the development of this resistance, it appears that anti-inflammatory

# REFERENCES


therapies could provide an extra benefit to lifestyle and nutrition changes on restoring muscle mass in the aged population (Dalle et al., 2017). Similarly, not only suppression of disease activity but also exercise and a well-suited nutritional management have been underlined as major strategies against RS (Masuko, 2014).

An increasing attention is being currently paid to the JAK/STAT pathway as a promising target for the treatment of muscle wasting diseases. The product of this signaling cascade seems to differ, likely depending on the magnitude and duration of the input. It would be of high interest to assess therapeutic effects of JAK/STAT inhibitors not only on inflammatory parameters, but also on muscle mass and function both in clinical and preclinical studies, since they certainly stand as promising drugs in the management of sarcopenia.

On the whole, to look into muscle involvement in distinct types of systemic conditions may help identifying patterns of muscle classic inflammatory response (i.e. IMM), low-grade chronic inflammation (i.e. PS or OA) or severe chronic systemic inflammation (i.e. RS).

# AUTHOR CONTRIBUTIONS

SP-B, RL and GH-B were in charge with conception and design; SP-B, IP-P, and GH-B: were involved in drafting the article; SP-B, IP-P, JR-B, OS-P, RL, and GH-B: revise it critically for important intellectual content and approved the final version to be published.

# FUNDING

This work was partially supported by grants from the Instituto de Salud Carlos III (PI13/00570; PI15/00340 and PI16/00065), co-funded by Fondo Europeo de Desarrollo Regional (FEDER).


and disuse atrophy. Exp. Biol. Med. 238, 1421–1430. doi: 10.1177/1535370 213505961


potential therapeutic targets in osteoarthritis. Biochem. Pharmacol. 108, 1–10. doi: 10.1016/j.bcp.2015.12.012


myopathies and differential diagnosis. Autoimmun. Highlights 5, 77–85. doi: 10.1007/s13317-014-0062-2


**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 Pérez-Baos, Prieto-Potin, Román-Blas, Sánchez-Pernaute, Largo and Herrero-Beaumont. 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.

# Syngeneic B16F10 Melanoma Causes Cachexia and Impaired Skeletal Muscle Strength and Locomotor Activity in Mice

Fabrício A. Voltarelli 1, 2, Fernando T. Frajacomo1, 3 \*, Camila de Souza Padilha<sup>1</sup> , Mayra T. J. Testa<sup>1</sup> , Paola S. Cella<sup>1</sup> , Diogo F. Ribeiro<sup>1</sup> , Donizete X. de Oliveira<sup>1</sup> , Luciana C. Veronez <sup>4</sup> , Gabriela S. Bisson<sup>4</sup> , Felipe A. Moura<sup>1</sup> and Rafael Deminice<sup>1</sup>

 Department of Physical Education, Faculty of Physical Education and Sport, State University of Londrina, Londrina, Brazil, Department of Physical Education, Faculty of Physical Education, Federal University of Mato Grosso, Cuiabá, Brazil, Program of Molecular Carcinogenesis, Brazilian National Institute of Cancer, Rio de Janeiro, Brazil, <sup>4</sup> Department of Maternal-Infant Nursing and Public Health, Ribeirao Preto College of Nursing, University of São Paulo, São Paulo, Brazil

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Paula Tavares, University of Coimbra, Portugal Kate T Murphy, University of Melbourne, Australia Craig Andrew Goodman, Victoria University, Australia

\*Correspondence:

Fernando T. Frajacomo ffrajacomo@inca.gov.br

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 12 May 2017 Accepted: 05 September 2017 Published: 29 September 2017

#### Citation:

Voltarelli FA, Frajacomo FT, Padilha CS, Testa MTJ, Cella PS, Ribeiro DF, de Oliveira DX, Veronez LC, Bisson GS, Moura FA and Deminice R (2017) Syngeneic B16F10 Melanoma Causes Cachexia and Impaired Skeletal Muscle Strength and Locomotor Activity in Mice. Front. Physiol. 8:715. doi: 10.3389/fphys.2017.00715 Muscle wasting has been emerging as one of the principal components of cancer cachexia, leading to progressive impairment of work capacity. Despite early stages melanomas rarely promotes weight loss, the appearance of metastatic and/or solid tumor melanoma can leads to cachexia development. Here, we investigated the B16F10 tumor-induced cachexia and its contribution to muscle strength and locomotor-like activity impairment. C57BL/6 mice were subcutaneously injected with 5 × 10<sup>4</sup> B16F10 melanoma cells or PBS as a Sham negative control. Tumor growth was monitored during a period of 28 days. Compared to Sham mice, tumor group depicts a loss of skeletal muscle, as well as significantly reduced muscle grip strength and epididymal fat mass. This data are in agreement with mild to severe catabolic host response promoted by elevated serum tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6) and lactate dehydrogenase (LDH) activity. Tumor implantation has also compromised general locomotor activity and decreased exploratory behavior. Likewise, muscle loss, and elevated inflammatory interleukin were associated to muscle strength loss and locomotor activity impairment. In conclusion, our data demonstrated that subcutaneous B16F10 melanoma tumor-driven catabolic state in response to a pro-inflammatory environment that is associated with impaired skeletal muscle strength and decreased locomotor activity in tumor-bearing mice.

Keywords: cancer, melanoma, inflammation, muscle strength, general locomotor activity

# INTRODUCTION

Cancer cachexia is a complex syndrome characterized by progressive weight loss, anorexia, muscle loss and weakness. Muscle wasting have been emerging as one of the principal components of cancer cachexia, leading to progressive impairment of work capacity (Fearon et al., 2011) a significant hallmark of poor prognosis in cancer patients (Christensen et al., 2014). The causal understanding of the muscle wasting in cancer cachexia is complex; the pro-inflammatory scenario however, is thought to play a prominent role (Ebrahimi et al., 2004). Elevated TNF-α, IL-6 and IL-1β

**169**

are associated with weight loss, hypercatabolism, reduced food intake and muscle loss in cancer patients (De Larichaudy et al., 2012; Yuan et al., 2015). Thus, elevated inflammatory interleukins and muscle wasting during tumor development can significantly modulate muscle strength and locomotor capacity in tumorbearing mice, some of the most critical endpoints associated with cancer morbidity and mortality (Christensen et al., 2014).

Melanoma is the deadliest form of skin cancer. When diagnosed at localized stages (I and III), the five-year survival rates correspond to 98%; however, this dramatically drops to 5– 19% when metastases is presented (Sandru et al., 2014; American Cancer Society, 2017). Despite early stages melanomas are not related with weight loss, the appearance of metastatic and/or solid tumor melanoma is related to cachexia development (Kawamura et al., 1999b; Das et al., 2011). Melanoma metastasis has a simultaneous lymphatic and hematogenous spread, with the potential to metastasize in any organ (Belhocine et al., 2006). Recent data has been demonstrated lung, gastrointestinal tract, extra-regional lymph nodes, liver and prostate are the principal target of metastatic melanoma; all of than with elevated rate of cachexia (40 to 80%) and mortality (von Haehling et al., 2016).

To better unveil melanoma tumor biology and tumor-host interaction, a number of murine models have been widely used in preclinical melanoma studies (Becker et al., 2010; Kuzu et al., 2015). The most widely used cell type in the melanoma model is the B16 cell line and two subclones (F1 and F10) that spontaneously form tumors after syngeneic transplantation in C57BL/6J mice (Fidler and Nicolson, 1976). The syngeneic model involves the induction of tumor cells into the same species and genetic background (Darro et al., 2005; Bobek et al., 2010). These models have important benefits over xeno-transplantation or genetically modified models since mice possess a normal immune system that opens a useful platform for new immunotherapy and adequate host response (Becker et al., 2010). However, a few number of studies have demonstrated B16 cells induces cachexia in mice following intraperitoneal inoculation through higher lipase lipoprotein activity (Kawamura et al., 1999a,b; Das et al., 2011). Indeed, there is a paucity of information dealing with tumor-host interaction in mice carrying B16F10 melanoma cells regarding cancer cachexia, specially the interaction among cancer cachexia components and skeletal muscle strength and locomotor activity. Thus, we aimed to investigate the cachexia development response to B16F10 tumor cells inoculation and it interactions with skeletal muscle adaptations (mass and strength) and mice locomotor activity. We hypothesized muscle loss and pro-inflammatory condition induced by tumor development play a pivotal role on reduced muscle strength and decreased locomotion and exploratory activity in B16F10 tumor-bearing mice.

## METHODS

## Animals

Twenty male C57BL/6 mice, 6–8 weeks old, were initially obtained from the Central Creation Unit of University of Sao Paulo (USP), Ribeirao Preto, Brazil and were used in this study. Mice were maintained under controlled temperature (24 ± 1 ◦C), light (12 h of light/12 h of darkness), relative air humidity (60- 70%) and allowed free access to water and food. After 1 week of local animal care facility adaptation, animals were randomly divided into two groups: Sham inoculated mice (n = 8) and tumor-bearing mice (n = 12). Animals were monitored three times a week for body weight, food and water intake, and tumor dimensions and euthanized 28 days after tumor cells inoculation. The daily food intake was calculated by the difference between the weight of the received food in their home cage hopper (lab balance Shimadzu BL320H) and the weight of the remained food 24 h after. The current study was approved by the Ethics Committee in Animal Experimentation of the State University of Londrina (#28336.2014.38), which follows the recommendations of the Brazilian Code for Use of Laboratory Animals (Law No. 6638, of May 8th, 1979 and Decree No. 26645 of July 10th, 1934).

# Tumor Cells Inoculation

B16F10 melanoma cells were cultured in RPMI 1640 medium (Gibco, Invitrogen) supplemented with 10% of fetal bovine serum (Gibco, Invitrogen), 100 µg/ml of streptomycin and 100 units/ml of penicillin at 37◦C and 5% of CO2. For inoculation, cells were removed from culture flasks by adding 0.05% of trypsin solution, centrifuged and resuspended in sterile PBS in order to obtain a solution containing 5 × 10<sup>4</sup> cells/ml. Cell viability was determined by trypan blue exclusion. Finally, C57BL/6 mice were subcutaneously injected with 5 × 10<sup>4</sup> cells/animal (0.1 ml) into the right thigh. As a negative control, the sham group was inoculated with 0.1 ml PBS only. Two of initial 12 tumor-bearing animals died during the study; other two animals did not develop any apparent tumors after B16F10 inoculation and were excluded from the study. Therefore, a total of eight tumor-bearing mice were considered for the study; the same number of animals was considered for the sham group.

# Tumor Assessment

Tumor growth was monitored by digital caliper (Sagyma Plus, 0– 150 mm) three times a week using two-dimensional measures. Tumor volume was calculated using a standard solid tumor formula V = 1/2∗(D∗d2) (Goto et al., 2000); being V, volume, D higher diameter and d lower diameter. The same examiner performed all measures in order to minimize bias.

# Maximal Muscle Strength Analysis

After 2 weeks of adaptation to protocol, the grip strength was determined using a dynamometer EEF 305 Grip Strength Meter (Insight <sup>R</sup> , Ribeirão Preto, Brazil) containing a pyramidal platform adapted to forelimbs. The test consisted of pulling the tail of the animal in order to cause their limbs to touch the grid of the equipment, which has a traction strength sensor. Three measures were performed in each mouse, with corresponding intervals of 15–12 s between sets, and with the tension administered gradually and consistently with the limitations of each animal. The quantitative data used corresponds to the mean strength of three attempts performed by the animals. The tests were carried out under the same experimental conditions.

# Locomotion and Exploratory Activity

Locomotion and exploratory activity was determined using an open field arena (40 × 40 cm<sup>2</sup> ) that was divided equally into lines drawn on the chamber floor for visual scoring of the activity by the experimenter (Prut and Belzung, 2003). Each mouse was placed in the center of the field arena and was allowed to freely explore the chamber for 30 s, followed by a 5-min test length. Overall locomotor activity (measured with video track digital camera, Longitech, C920, 30 Hz fixed above 60 × 60 cm rigid box) was recorded, as well as the amount of time and distance spent and numbers of entries in the center area. During the tests, the video sequences were stored in a personal computer for analysis. Using an automatic tracking method via DVideo software interface (Figueroa et al., 2003, 2006), we obtained the trajectories of mice. The two automatic procedures of segmentation-background subtraction and tracking-generated the mouse's positions as a function of time over the entire test. Background subtraction is a very common method used TABLE 1 | Initial body mass, tumor-free body mass, relative weights of the liver, muscle, epididymal fat pads and tumor mass, as well as cachexia index determined for tumor-bearing and Sham mice.


Values expressed as mean ± SD. Data evaluated by T-test for independent samples; adopted value of significance of P < 0.05.

for segmenting moving objects, which consists of the difference between a set of images and its background model.

Before the tests, we obtained the coordinates of four specific points relative to a coordinate system associated with the box. The corresponding projections of these points in the image were determined with DVideo software. The homography parameters of the image-object transformation were then calculated based on the DLT (Direct Linear Transformation) proposed by Abdel-Azis and Karara (1971) and mouse two-dimensional (2D) coordinates relative to the box coordinate system were obtained. Mice naturally prefer to be near a protective wall rather than be exposed to danger in an open area; however a competing foraging will motivate them to explore (Hefner et al., 2007).

# Euthanasia and Preparation of Tissues

The animals were euthanized using anesthetic overdose with ketamine 0.02 g and xylazine 0.004 g. Subsequently, 1 ml of blood was collected from the abdominal aorta, stored in heparinized tubes and centrifuged for plasma separation. Plasma samples were stored at −80◦C for further analysis. Liver, epididymal fat pads and soleus and gastrocnemius muscles were carefully dissected and weighted. The sum of both skeletal muscles (P) was used as muscle mass parameter. Tumors were carefully dissected from the upper-left flank for weighing, and were subsequently cut in half and stored in liquid nitrogen. The cachectic state was determined using the cachexia index that consider the body weight gain of control mice and the tumor mass in tumor-bearing mice, according to the following equation:[(Initial body weight - final body mass + (mass of the tumor) + control weight gain) x 100 / (initial body mass + mass gain control group)] (Martins et al., 2016). Cachexia index is a tool to determine weight loss related to control animals. Animals were considered cachectic when presented weight body loss >5% (Fearon et al., 2011).

# Biochemical Analysis

For biochemical analysis, plasma lactate dehydrogenase (LDH; Commercial Kit Labteste/LDH Liquiform/Ref: 86-2) was

measured using a plate reader (Epoch <sup>R</sup> Bio Tek, Winnoski, USA). The concentrations of plasma cytokines TNF-α (EBioscience Ref: 88-7340-88), interleukin-6 (EBioscience Ref: 88-7064-88) and interleukin-10 (EBioscience Ref: 88-7105- 88) were determined using ELISA commercial available kits in a plate reader (Epoch <sup>R</sup> Bio Tek, Winnoski, USA).

# Statistical Analyses

The statistical analyses were performed using IBM SPSS Software (version 21). The normality was evaluated using the Kolmogorov-Smirnov test. Normal data were analyzed by Student's T-test for independent samples and the results expressed as Mean ± Standard Deviation (SD). Food intake and tumor volume over time were compared using ANOVA two way differences b-test. The Pearson correlation coefficient test was used to determine association among skeletal muscle strength, locomotor-like activity and cachexia characteristics (body weight loss, reduced adipose tissue and skeletal muscle loss, elevated inflammatory interleukins and anorexia). In all cases, the differences were considered significant when P < 0.05.

# RESULTS

Tumor mass was apparent only after 19 days after cell inoculation in average. The animals were monitored until euthanasia when the tumor mass reached 11.72% of total body weight. After careful analysis, we did not find any metastases in target tissues such as lung, stomach and intestines, extra-regional lymph nodes, liver, pancreas and spleen. Tumor development also caused decreased food intake over the third and fourth weeks (**Figure 1**), without changes in water intake.

The tissue responses of the tumor-bearing animals and the sham group are shown in **Table 1**. Clearly, the muscle and epididymal fat content were significantly decreased (P < 0.05) in tumor-bearing mice compared to the sham group. Spleen was 2-fold larger in tumor-bearing compared to Sham mice. Tumorfree body weight was not different, despite the P trend = 0.07 (**Table 1**). Cachexia index was 8.3%, higher than 5% established by as Fearon et al. (2011) as cachexia definition. Liver relative weight was not affected by tumor growth.

**Figure 2** shows the data related to the systemic response of experimental groups. The tumor-bearing mice presented elevated LDH plasma levels. Plasma inflammatory cytokines were significantly elevated in the tumor bearing group; TNF-α was over two-fold and IL-6 was over eight-fold higher in tumorbearing mice compared to Sham animals. However, there was no difference between groups in relation to IL-10.

**Figure 3** presents the results of the locomotor and exploratory test as well as the gripping strength test. The overall distance and distance traveled in the center were significantly (P < 0.05) lower

(40 and 56%, respectively) in tumor-bearing mice compared to sham. Tumor-bearing mice also spent significantly (P < 0.05) more time stopped than sham mice. In addition, tumor growth promoted a significantly decreased muscle gripping strength compared to sham mice.

The Pearson correlation test demonstrated reduced skeletal muscle mass and body weight loss were significantly (P <0.05) associated to gripping strength. Elevated plasma TNF-α was inversely (P < 0.05) associated to gripping strength. Also, reduced skeletal muscle mass and elevated plasma TNFα and IL-6 were determinant for the impaired locomotor activity parameters (**Figure 4**). In addition, correlation test also demonstrated negative association between decreased skeletal muscle mass and elevated plasma TNF-α (r = −0.71; P < 0.05) and IL-6 (r = −0.57; P < 0.05)

# DISCUSSION

Our data demonstrated B16F10 tumor cell inoculation promoted cachexia characterized by weight loss, skeletal muscle wasting, adipose tissue loss and anorexia. Fat tissue significantly targeted presenting a 33.06% reduction in tumor-bearing mice compared to controls. The catabolic state appears to be intrinsic to the anorexia state and systemic organic pro-inflammatory response against tumor progressive growth. In addition, tumor growthdriven inflammation and muscle loss impaired voluntary muscle strength and decreased locomotor activity in tumor-bearing mice. Although previously studies have demonstrated B16F10 intraperitoneal inoculation was able to induce cancer cachexia (Kawamura et al., 1999a,b). Our data have demonstrated therefore, both muscle wasting and inflammatory condition

mediated tumor growth play an important role in skeletal muscle strength loss and locomotor activity impairment.

Preclinical tumor models indicate that body and tissue wasting can impact functional parameters (Chen et al., 2015). Skeletal muscle loss appears to be the most essential component associated with poor outcome in cancer cachexia (Fearon et al., 2011). Our data highlighted reduced grip strength and decreased locomotor activity in the tumor group challenged with B16F10 cells, were associated to loss of muscle mass. These data demonstrated tumor growth promotes a critical modification in skeletal muscle that compromised animal's locomotor activity. Tumor-bearing mice inoculated with LLC cells demonstrated atrophy and lower grip strength (Chen et al., 2015). However, the skeletal muscle loss causative interaction is poorly investigated. Our results are particularly relevant since muscle loss has not been linked with lower muscle function only (Gabriel et al., 2006; Jorgensen et al., 2010). Due to complex coordination among neural, mechanic and physiological functions, muscle strength is highlighted as the most critical functional endpoint to survival and hospital discharge (Mendes et al., 2014; Joglekar et al., 2015).

Tumor-driven inflammatory factors may act as mechanisms underlying lipolysis and myolysis in wasting conditions (Der-Torossian et al., 2013; Laine et al., 2013; Inacio Pinto et al., 2015). Previously studies have demonstrated inflammatory cytokines play a critical role in muscle wasting by stimulating transcription factor nuclear factor-kB (NF-kB), which targets key skeletal muscular genes of catabolic cascade and up-regulate ubiquitin-proteasome pathway for protein catabolism (Zhou et al., 2016). In the subcutaneous tumor model, tumor-bearing mice show elevated TNF-α, IL-6 and activin A levels that up regulate ubiquitin ligases such as muscle ring finger-1 (MuRF1) and Atrogin-1 mRNA expression in gastrocnemius muscles (Matsuyama et al., 2015). Our data elucidates elevated plasma levels of TNF-α and IL-6 after 28 days of transplantation was associated to skeletal muscle loss, as demonstrated by negative association between decreased skeletal muscle mass and elevated plasma TNF-α (r = −0.71; P < 0.05) and IL-6 (r = −0.57; P < 0.05). Our study also demonstrated spleen was 2-fold larger in tumor-bearing compared to Sham mice which is a good indicator of systemic inflammation state. Indeed, chronic inflammatory condition caused by tumor development is thought to play a prominent role on muscle loss in cancer cachexia (De Larichaudy et al., 2012; Yuan et al., 2015). Although previous studies have suggested that IL-10 plays an important role in cachexia from different cancer types (Ebrahimi et al., 2004; Sun et al., 2010), our model promoted no changes in IL-10 plasma levels. In addition, taking the negative association between inflammatory interleukins and locomotor activity parameters (**Figure 4**), we may affirm that skeletal muscle wasting, loss of muscle strength and impaired locomotor activity imposed by B16F10 melanoma is inherent to the systemic pro-inflammatory condition promoted by tumor growth. Despite de association demonstrated among loss of muscle mass, elevated interleukin and impaired locomotion activity, is important to say that skeletal muscle tissue mass only is not the best predictor of muscle wasting. The absence of skeletal muscle cross sectional area and cardiac mass appears to be the principal limitation of the present study.

In addition to the catabolic-related inflammation scenario, a higher production of LDH is characteristic of a glycolytic phenotype reprogram. According to the Warburg effect, cancer cells undergo glycolysis rather than mitochondrial phosphorylation as a major source of energy (Vander Heiden et al., 2009), which may be directly linked with an activation of oncogenes and lower expression of tumor suppressors (Dang and Semenza, 1999; Allison et al., 2014). In melanoma patients, LDH is a stronger predictive factor of lower overall survival (Kelderman et al., 2014; Diem et al., 2015). However, the link between LDH and functional impairments in cancer is not clear. Indeed, melanoma cells favor a glycolytic phenotype and pro-inflammatory environment that may exert a key role in lipolysis, lower muscle strength and significant loss of mobility.

Anorexia is also an important component of cancer cachexia. Anorexia-related to cancer primary cause is often increased pro-inflammatory state and/or increased lactate production (Ezeoke and Morley, 2015). Administration of cytokines and lactate to rodents has been demonstrated to reduce food intake (Silberbauer et al., 2000; Patra and Arora, 2012). These two factors then modulate central nervous system neurotransmitter cascades, especially in hypothalamus, resulting in a reduction in satiation (Ezeoke and Morley, 2015). Therefore, is reasonable to affirm that elevated circulating interleukin and LDH activity demonstrated in our study may play a pivotal role in anorexia development induced by B16F10 melanoma development. Indeed, anorexia in cancer cachexia may also contribute significantly to muscle wasting and impaired locomotor activity demonstrated in our study. We cannot determine however,

# REFERENCES


the significance plot of each cachexia component to locomotor activity impairment; it must be considered in future studies.

In conclusion, our data demonstrated that a subcutaneous B16F10 melanoma model promotes a catabolic state probably in response to a pro-inflammatory environment. Elevated inflammatory interleukins and muscle wasting induced by tumor development appears to play a important role on the impairment of muscle strength and decreased locomotor activity in tumorbearing mice, some of the most critical endpoints associated with cancer morbidity and mortality. This model may be a valuable tool in testing new therapeutic approaches and future translational perspectives.

# AUTHOR CONTRIBUTIONS

FF and RD conceived of the studies. LV, GB, and FM provided essential equipment, expertise and resources. FF, CP, MT, PC, DR, and DdO performed experiments. FV, FF, and RD analyzed and interpreted data. FV, FF, and RD wrote and edited the manuscript.

# FUNDING

Supported by grants from: Fundação Araucária, Brazil; Coordenação de Aperfeiçoamente de Pessoal de Nível Superior - Capes PVE, Brazil Proc. Proc. 88881.068035/2014-01; Conselho Nacional de Desenvolvimento Científico e Tecnológico-CNPq, Brazil.


**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 PT and handling editor declared their shared affiliation.

Copyright © 2017 Voltarelli, Frajacomo, Padilha, Testa, Cella, Ribeiro, de Oliveira, Veronez, Bisson, Moura and Deminice. 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.

# Identification of Biomarkers Correlated with the TNM Staging and Overall Survival of Patients with Bladder Cancer

Sheng Li 1, 2, 3†, Xiaoping Liu3†, Tongzu Liu<sup>1</sup> , Xiangyu Meng<sup>3</sup> , Xiaohong Yin<sup>3</sup> , Cheng Fang<sup>3</sup> , Di Huang<sup>3</sup> , Yue Cao<sup>3</sup> , Hong Weng<sup>1</sup> , Xiantao Zeng1, 3 \* and Xinghuan Wang1, 3 \*

<sup>1</sup> Department of Urology, Zhongnan Hospital of Wuhan University, Wuhan, China, <sup>2</sup> Department of Biological Repositories, Zhongnan Hospital of Wuhan University, Wuhan, China, <sup>3</sup> Center for Evidence-Based and Translational Medicine, Zhongnan Hospital of Wuhan University, Wuhan, China

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

Alex Zhavoronkov, Biogerontology Research Foundation, United Kingdom Monica Catarina Botelho, Istituto Nacional de Saúde, Portugal

#### \*Correspondence:

Xinghuan Wang wangxinghuan@whu.edu.cn Xiantao Zeng zengxiantao1128@163.com

† These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 02 May 2017 Accepted: 08 November 2017 Published: 28 November 2017

#### Citation:

Li S, Liu X, Liu T, Meng X, Yin X, Fang C, Huang D, Cao Y, Weng H, Zeng X and Wang X (2017) Identification of Biomarkers Correlated with the TNM Staging and Overall Survival of Patients with Bladder Cancer. Front. Physiol. 8:947. doi: 10.3389/fphys.2017.00947 Objective:To identify candidate biomarkers correlated with clinical prognosis of patients with bladder cancer (BC).

Methods: Weighted gene co-expression network analysis was applied to build a co-expression network to identify hub genes correlated with tumor node metastasis (TNM) staging of BC patients. Functional enrichment analysis was conducted to functionally annotate the hub genes. Protein-protein interaction network analysis of hub genes was performed to identify the interactions among the hub genes. Survival analyses were conducted to characterize the role of hub genes on the survival of BC patients. Gene set enrichment analyses were conducted to find the potential mechanisms involved in the tumor proliferation promoted by hub genes.

Results: Based on the results of topological overlap measure based clustering and the inclusion criteria, top 50 hub genes were identified. Hub genes were enriched in cell proliferation associated gene ontology terms (mitotic sister chromatid segregation, mitotic cell cycle and, cell cycle, etc.) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (cell cycle, Oocyte meiosis, etc.). 17 hub genes were found to interact with ≥5 of the hub genes. Survival analysis of hub genes suggested that lower expression of MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1were associated with better overall survival of BC patients. BC samples with higher expression of hub genes were enriched in gene sets associated with P53 pathway, apical junction, mitotic spindle, G2M checkpoint, and myogenesis, etc.

Conclusions: We identified several candidate biomarkers correlated with the TNM staging and overall survival of BC patients. Accordingly, they might be used as potential diagnostic biomarkers and therapeutic targets with clinical utility.

Keywords: bladder cancer, biomarkers, WGCNA

# INTRODUCTION

Bladder cancer (BC) is the second most frequent genitourinary malignancy and the sixth most common malignancy in men. BC represents a spectrum of disease ranging from superficial, well-differentiated disease, which does not significantly affect the survival of BC patients, to highly fatal tumors for which longterm survival may be dismal (So, 2016; Ghervan et al., 2017). With the aging of population, the incidence of BC is rising year by year, and BC in elder patients will become even more frequent and evolved into a public health challenge in future. For patients with superficial BC, telescopic removal of the cancer (transurethral resection of bladder tumor, TURBT) followed by instillation of chemotherapy or vaccine-based therapy into the bladder with prolonged telescopic checking of the bladder are usually recommended, and the 5-year overall survival for these patients reaches 90%, while about 40–80% of these patients will develop disease recurrence or progression (Malmström et al., 2017). For patients with invasive BC, radical cystectomy plus pelvic lymph node dissection (PLND) followed by neo-adjuvant chemotherapy is recommended as a standard of care, and once it becomes metastatic cancer, the 5-year overall survival for patients with invasive BC is a dismal 6% (Salama et al., 2016; Sargos et al., 2016).

Biomarkers are biological substances whose detection indicates a particular disease state. So far, a variety of biomarkers have been introduced in daily clinical practice, including risk assessment, screening, differential diagnosis, determination of prognosis, prediction of response to treatment, and monitoring of progression of disease. Thus, identification of biomarkers that are associated with clinical outcomes of patients with BC might be of clinical significance (Giunchi et al., 2017).

Currently, several screening algorithms based on gene expression data, including Gene Set Enrichment Analysis (GSEA), Signalling Pathway Impact Analysis (SPIA), Topology Gene-Set Analysis, and DEGraph, and in silico Pathway Activation Network Decomposition Analysis (iPANDA), etc., have been introduced for both academic and commercial purpose, and most of these algorithms are intended to identify differentially expressed genes between groups of samples (Ozerov et al., 2016). Weighted gene co-expression network analysis (WGCNA), a systems biology algorithm, can be applied to describing the correlation patterns among genes across microarray samples, finding and summarizing modules of high related genes, and relating modules to certain clinical phenotype (Zhang and Horvath, 2005; Yip and Horvath, 2007). During the network construction, highly co-expressed genes are connected in the network and grouped into modules, thus, different modules included different functionally related genes and the most central and connected genes are treated as hub genes (Zhang and Horvath, 2005; Yip and Horvath, 2007). Correlation networks facilitate network based gene screening methods that can be used to identify candidate biomarkers or therapeutic targets. Therefore, WGCNA is usually used to facilitate the screening or identification of candidate biomarkers or therapeutic targets. Tumor node metastasis (TNM) 2009 (7th edition) was recommended for the BC, and previous studies demonstrated that patients with an infiltrative pattern had better survival than those with other pattern types (PDQ Adult Treatment Editorial Board, 2002; Yaxley, 2016). In the present study, the WGCNA algorithm was applied to identify candidate biomarkers for BC based on the TNM staging of BC patients.

# METHODS

# Data Sources and Data Preprocessing

Gene expression profile of GSE13507 (Kim et al., 2010; Lee et al., 2010) was downloaded from Gene Expression Omnibus (GEO) database. GSE13507 is a microarray dataset containing 165 primary BC samples, 23 recurrent non-muscle invasive tumor tissues, 58 normal looking bladder mucosae surrounding cancer and 10 normal bladder mucosae, the clinical characteristics of included samples are attached as well. In the present study, only the 165 primary BC samples were included for subsequent analysis. Preprocessed gene expression profile of GSE13507 was obtained from GEO database for our WGCNA analysis. Probesets were filtered by their variance across all samples, only probes with variances ranked in top 10,000 were selected for subsequent analyses.

# Co-expression Network Construction and Detection of Hub Genes

The R package "WGCNA" (Langfelder and Horvath, 2008) was used to build co-expression network for the filtered gene expression matrix. Before building a co-expression network, we applied sample networks method introduced by Oldham et al for outlier detection. We designated sample as outlying, if the Z.K value was below −2.5. The soft threshold power β was selected based on the scale-free topology criterion introduced by Zhang and Horvath (2005). We calculated Pearson's correlations between each gene pair to determine concordance of gene expression to generate a matrix of adjacencies, and then the adjacencies were transformed into topological overlap matrix (TOM) (Li and Horvath, 2009). Next, we performed average linkage hierarchical clustering based on the TOM-based dissimilarity with a minimum module size of 30 and a medium sensitivity of 2, and other parameters were designated as default. After relating modules to clinical traits and calculating the associated Gene Significance (the correlation between the genes and the trait) and Module Membership (the correlation of the module eigengene and the gene expression profile), we detected top 50 hub genes using a networkScreening function based on Gene Significance and Module Membership and genes with q.Weighetd value less than 0.001 were finally regarded as hub genes (Dong and Horvath, 2007).

# Functional Annotation of Hub Genes

WebGestalt (WEB-based GEne SeT AnaLysis Toolkit) (Zhang et al., 2005; Wang et al., 2013) was used to conduct gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (http://www.kegg.jp/kegg/) pathway enrichment analysis of hub genes. Search Tool for the Retrieval of Interacting Genes (STRING) database (Szklarczyk et al., 2015) was use to construct a protein-protein interaction (PPI) network, and then the PPI network was visualized using Cytoscape (Shannon et al., 2003).

# Survival Analysis

Another BC microarray study GSE19915 (Lindgren et al., 2010), which included 144 BC samples and 12 normal samples and the associated clinical characteristics including gender, age, biopsy Gleason score, survival status, follow-up, etc., was regarded as a validation cohort. PROGgenesV2 (Goswami and Nakshatri, 2014) was applied to conduct log-rank based survival analyses to compare the overall survivals of particular comparing groups defined based on the medians of the expression level of hub genes. Difference with statistical significance was defined as P < 0.05.

# Gene Set Enrichment Analysis (GSEA)

Bladder cancer (BC) microarray datasets GSE31684 (Riester et al., 2012, 2014) was used to conduct GSEA (Subramanian et al., 2005) analysis of hub genes. BC samples were divided into a particular hub gene high expression group and low expression group based on the median expression of this hub gene. Differences at nominal P < 0.05 and FDR (false discovery rate) less than 25% were defined as statistical significance.

# RESULTS

# Results of Data Preprocessing, Co-expression Network Construction and Hub Genes Identification

Gene expression profile of GSE13507 was obtained from GEO database, and probes with variances ranked in top 10,000 were used in the subsequent analyses. After outlier detection, one sample was excluded for further analysis (**Supplementary Figure 1**). As shown in **Figure 1A**, β = 14, the lowest power for which the scale-free topology fit index reaches 0.9, was selected for the subsequent adjacencies calculation. Based on the results of TOM based clustering, we obtained 11 gene modules as shown in **Figure 1B**. After the modules were related to the TNM stages of BC patients, top 50 hub genes identified based on the corresponding Gene Significance and Module Membership were summarized in Supplementary Table 1.

# Gene Ontology Enrichment Analysis of Hub Genes

To get a primary understanding of the biological relevance of the hub genes, we first performed GO enrichment analysis of the hub genes. As shown in **Table 1**, top 10 enriched GO terms are listed. The hub genes mainly enriched in "mitotic sister chromatid segregation," "mitotic cell cycle," "nuclear division," "sister chromatid segregation," "cell cycle," "chromosome segregation," "mitotic nuclear division," "cell cycle process," "organelle fission," and "cell division."

# KEGG Pathway Enrichment Analysis of Hub Genes

Moreover, as shown in **Table 2**, the KEGG pathway enrichment analysis of the hub genes indicated that these genes were mainly enriched in "Cell cycle," "Oocyte meiosis," "MicroRNAs in cancer," "Protein digestion and absorption," and "Progesteronemediated oocyte maturation."

# PPI Network Analysis of Hub Genes

Hub genes were mapped to STRING database. As shown in **Figure 2**, a total number of 244 pairs of PPIs were obtained from STRING database,and 17 hub genes ("AURKB," "TOP2A," "PRC1," "KIF4A," "CENPF," "NUSAP1," "KIF2C," "CEP55," "CDCA8," "CENPA," "CCNB2," "FOXM1," "UBE2C," "CDC45," "KIF15," "BUB1B," and "CDCA3"), interacted with ≥5 of the hub genes, were at the hub of the PPI network.

# Survival Analysis of Hub Genes

To perform a further validation of the hub genes and identify potential biomarkers for BC, survival analyses of the hub genes were conducted using PROGgenesV2. As shown in **Table 3**, lower expression of 11 hub genes (including MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1) were significantly associated with better overall survival of BC patients.

# GSEA Analysis of Hub Genes that were Significantly Correlated with Overall Survival of BC Patients

To characterize the potential mechanisms involved in the influences on overall survival of the above 11 hub genes. GSEA was conducted based on the expression of MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1. As shown in Supplementary Table 2, BC samples in MMP11 high expression group were most significantly enriched in P53 pathway; BC samples in COL5A2 high expression group were most significantly enriched in apical junction (Supplementary Table 3); BC samples in CDC25B, CENPF, TPX2, CDCA8, and FOXM1 high expression groups were most significantly enriched in mitotic spindle (Supplementary Tables 4–8); BC samples in TOP2A, CDCA3 and TK1 high expression groups were most significantly enriched in G2M checkpoint (Supplementary Tables 9–11); BC samples with AEBP1 high expression were most significantly enriched in myogenesis (Supplementary Table 12).

# DISCUSSION

As mentioned above, BC is one of the most common cancers and significant progress has been made during the past decade. The prognosis of patients with BC, especially invasive BC, remains poor. The recurrence rate for superficial BC is about two-thirds, and, despite advances in management, some patients still develop stage progression. Meanwhile, approximately 30% of muscleinvasive BCs are associated with occult distant metastasis at the time of diagnosis, which led to a dismal 5-year survival of patient with muscle invasive BC (Kamat et al., 2016). Biomarker is defined as a substance found in tissue, blood, or other body fluids that might be a sign of cancer or noncancerous conditions (Mohammed et al., 2016). Previous studies suggested that biomarkers, including BC biomarkers, exhibited several features: prognostic, predictive, and pharmacodynamic. The biological

functions of prognostic biomarker, predictive biomarker, and pharmacodynamic biomarker were predicting the natural course of cancers, evaluating the probable benefit of a particular treatment, and assessing the treatment effects of a drug on a tumor and determine the proper dosage of a new anticancer drug, respectively (Kojima et al., 2016; Mohammed et al., 2016; Nandagopal and Sonpavde, 2016). Nowadays, many biomarkers and their corresponding targeted agents have been determined for the diagnosis and treatment of BC. In the present study, we identified 50 hub genes that were significantly correlated with the TNM staging of BC patients using WGCNA.

As we know, TNM staging, devised by Pierre Denoix, describes the size of the original (primary) tumor and whether it has invaded nearby tissue (T), describes nearby (regional) lymph nodes that are involved (N), and describes distant metastasis (M) (Denoix, 1946). The functional enrichment analysis of hub genes that were correlated with the TNM stages indicated that these genes were enriched in cell proliferation associated GO terms (mitotic sister chromatid TABLE 1 | GO oncology analysis of top 50 hub genes.


#### TABLE 2 | KEGG pathway analysis of top 50 hub genes.


segregation, mitotic cell cycle and cell cycle, etc.) and KEGG pathways (cell cycle, Oocyte meiosis, etc.). Thus, we speculated that these hub genes affected the TNM staging of BC patients through promoting the proliferation of BC cells.

The PPI network analysis of the hub genes suggested that a total number of 17 hub genes interacted with ≥5 of the hub genes were at the hub of the PPI network. Previous studies demonstrated that functional partnerships and interactions that occurred between proteins were at the core of cellular processing and their systematic characterization helped to provide context in molecular systems biology. Thus, the 17 hub genes, as mentioned above, might play an important role in the biological process of BC cells.

Our survival analyses of the hub genes indicated that lower expressions of 11 hub genes were correlated with relatively shorter overall survivals of patients with BC. Meanwhile, our GSEA results indicated that BC samples with higher expression of these genes were enriched in genes sets that were

correlated with biological behaviors of tumor cells (such as P53 pathway, Apical junction, mitotic spindle, G2M checkpoint, Myogenesis, etc.).

As for the 11 survival associated hub genes, we conducted a literature review of these genes. MMP11 is a member of the matrix metalloproteinase (MMP) family involved in the breakdown of extracellular matrix in normal physiological

TABLE 3 | Survival of hub genes, high expression of MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1 were correlated with better overall survival of patients with bladder cancer.


processes, such as embryonic development, reproduction, and tissue remodeling, as well as in disease processes, and several studies demonstrated that overexpression of MMP11 were correlated with many cancers (including colon cancer, laryngeal cancer, breast cancer, prostate cancer, etc.) (Deng et al., 2005; Kou et al., 2013; Roscilli et al., 2014; Li et al., 2015; Pang et al., 2016); COL5A2 encodes an alpha chain for one of the low abundance fibrillar collagens, and previous studies suggested that COL5A2 was associated with malignancy in colorectal cancer, breast cancer, osteosarcoma, etc. (Fischer et al., 2001; Vargas et al., 2012; Wu et al., 2014); CDC25B is a member of the CDC25 family of phosphatases, and previous studies suggested that elevated expression of CDC25B promoted the proliferation of gastric cancer, esophageal carcinoma, renal cell carcinoma, etc. (Yu et al., 2012; Wang et al., 2015; Leal et al., 2016); TOP2A encodes a DNA topoisomerase, and previous studies suggested that TOP2A overexpression was a poor prognostic factor in patients with breast cancer, colorectal cancer, prostate cancer, and nasopharyngeal carcinoma, etc., (de Resende et al., 2013; Lan et al., 2014; Tarpgaard et al., 2016; Zheng et al., 2016); CENPF is a component of the nuclear matrix during the G2 phase of interphase, and previous studies demonstrated that CENPF was associated with the tumor progression of patients with prostate cancer, hepatocellular carcinomas, and breast cancer, etc. (Brendle et al., 2009; Kim et al., 2012); Cell division cycle associated 3 (CDCA3) is a part of the Skp1 cullin-F-box (SCF) ubiquitin ligase and refers to a trigger of mitotic entry and mediates destruction of the mitosis inhibitory kinase, and previous studies demonstrated overexpression of CDCA3 promoted progression of several cancers (oral cancer, prostate cancer, and lung cancer, etc.) (Uchida et al., 2012; Chen et al., 2013; O'Byrne et al., 2016); TK1(thymidine kinase 1) encodes a cytosolic enzyme that catalyzes the addition of a gamma-phosphate group to thymidine, and previous studies proved that high TK1 expression is a predictor of poor survival in patients with pT1 of lung adenocarcinoma, BC, gastrointestinal cancer, etc. (Jagarlamudi et al., 2015; Du et al., 2016); Targeting protein for Xenopus kinesin-like protein 2 (TPX2) is microtubule-associated protein and impacts spindle assembly in human cells (Huang et al., 2014), and previous studies proved that TPX2 expression was associated with poor survival in gastric cancer, breast cancer, colon cancer and esophageal cancer, etc. (Wei et al., 2013; Liu et al., 2014; Yang et al., 2015; Tomii et al., 2017); The cell division cycle associated 8 (CDCA8) gene encodes a component of the chromosomal passenger complex and plays an important role in mitosis, and overexpression of CDCA8 was reported in some human cancers, demonstrating that CDCA8 was required for the growth and progression of several cancers such as breast cancer, gastric cancer, lung carcinogenesis (Hayama et al., 2007; Yan et al., 2012; Jiao et al., 2015); Several studies demonstrated the AEBP1 upregulation conferred acquired resistance to BRAF (V600E) inhibition in melanoma and was associated with bad survivals of patients with glioma (Ladha et al., 2012; Hu et al., 2013); The protein encoded by FOXM1 is phosphorylated in M phase and regulates the expression of several cell cycle genes, such as cyclin B1 and cyclin D1, and previous studies suggested that FOXM1 promoted tumor progression of multiple cancers including gastric cancer, ovarian cancer, cervical cancer, colorectal cancer and breast cancer, etc. Trichostatin A potentiates TRAIL-induced antitumor effects via inhibition of ERK/FOXM1 pathway in gastric cancer (Barger et al., 2015; Yau et al., 2015; Zheng et al., 2015; Li et al., 2016; Wang et al., 2016; Song et al., 2017). In summary, all the conclusions were consistent with the results of survival analysis and GSEA analysis that high expressions of MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1 were correlated with worse overall survival of BC patients and BC samples with relatively higher expression of these genes were enriched in gene sets that were associated with cell proliferation.

In conclusion, the identified 50 hub genes that were closely correlated with the TNM staging of BC patients, and 11 hub genes (MMP11, COL5A2, CDC25B, TOP2A, CENPF, CDCA3, TK1, TPX2, CDCA8, AEBP1, and FOXM1) of which were significantly correlated with the overall survival of BC, which could be candidate biomarkers for BC. Meanwhile, further in vivo and in vitro studies were needed to make clear the exact molecular mechanisms that affected the growth of BC cells.

# AUTHOR CONTRIBUTIONS

XW designed the study. SL and XL collected, analyzed and interpreted the data. XL, TL, XM, XY, CF, DH, and HW participated in revising the manuscript. XZ and YC participated in the study design and helped to draft the manuscript. All authors read and approved the final manuscript.

# FUNDING

This work was supported by the National Key Research and Development Plan of China (Grant No. 2016YFC0106300) and the National Natural Science Foundation of China (Grant No. 81772730) and Wuhan Clinical Research Center for Male Urogenital Cancer (Grant No. 303-230100055) and the 351 Talent Project of Wuhan University (Luojia Young Scholars: SL).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2017.00947/full#supplementary-material

# REFERENCES


Supplementary Figure 1 | Cluster tree of bladder cancer samples. The leaves of the tree correspond to the bladder cancer sample. The first color band

underneath the tree indicates which arrays appear to be outlying. The sixth color band represents TNM staging (red indicates high values). Similarly, the remaining color-bands color-code the numeric values of physiologic traits.


**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 Li, Liu, Liu, Meng, Yin, Fang, Huang, Cao, Weng, Zeng and Wang. 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.

# The Evaluation of Flow-Mediated Vasodilation in the Brachial Artery Correlates With Endothelial Dysfunction Evaluated by Nitric Oxide Synthase Metabolites in Marfan Syndrome Patients

Oscar Lomelí<sup>1</sup>† , Israel Pérez-Torres<sup>2</sup>† , Ricardo Márquez<sup>3</sup> , Sergio Críales<sup>4</sup> , Ana M. Mejía<sup>5</sup> , Claudia Chiney<sup>6</sup> , Enrique Hernández-Lemus<sup>7</sup> and Maria E. Soto<sup>3</sup> \*

#### Edited by:

Oreste Gualillo, Servicio Gallego de Salud, Spain

#### Reviewed by:

Paul Kenneth Witting, The University of Sydney, Australia Anjali Mishra, The Ohio State University, United States

\*Correspondence:

Maria E. Soto mesoto50@hotmail.com †Shared first authorship

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 15 September 2017 Accepted: 02 July 2018 Published: 21 August 2018

#### Citation:

Lomelí O, Pérez-Torres I, Márquez R, Críales S, Mejía AM, Chiney C, Hernández-Lemus E and Soto ME (2018) The Evaluation of Flow-Mediated Vasodilation in the Brachial Artery Correlates With Endothelial Dysfunction Evaluated by Nitric Oxide Synthase Metabolites in Marfan Syndrome Patients. Front. Physiol. 9:965. doi: 10.3389/fphys.2018.00965 <sup>1</sup> Department of Echocardiography, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>2</sup> Department of Pathology, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>3</sup> Department of Immunology, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>4</sup> Department of Computed Tomography, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>5</sup> Blood Bank, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>6</sup> Central Laboratory, National Institute of Cardiology "Ignacio Chávez", Mexico City, Mexico, <sup>7</sup> Computational Genomics Division, National Institute of Genomic Medicine, Mexico City, Mexico

Marfan syndrome (MS) is of the most common connective tissue disorders. Although most patients have mutations in the fibrillin-1 gene (FBN1) and more than 1,700 mutations have been described, there are no mutations in less than 10% of patients. Aortic dilation is the most important complication; it involves chronic inflammatory processes and endothelial dysfunction. Prospective study from March 2015 to January 2017, in a cohort of 32 patients of MS confirmed by Ghent criteria and 35 controls of both genders, with a median age of 26 years (18–56). Patients had no comorbidities such as diabetes, hypertension, and/or neoplasms. They were not being treated with statin, NSAIDs, calcium antagonists, oral nitrates, and/or beta-blockers during 7 days prior to the study and patients with smoking history in the last 4 years. Controls were matched by age and gender. We analyzed endothelial dysfunction by flow-mediated vasodilation in the brachial artery, determining the maximum peak flow in the reactive hyperemia phase with a Philips Envisor device with Doppler capability. Its correlation with serum levels of biological markers that could participate in endothelial dysfunction pathways such as NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, NO<sup>−</sup> 2 , citrulline, TNFα, IL-1, IL-6, IL-10, IL-8, osteopontin, ICAM, VCAM, and NO<sup>−</sup> 3 /NO<sup>−</sup> <sup>2</sup> was determined. Endothelial dysfunction was found in 21 MS patients (65%). The aortic annulus (AAo) was of 27 mm (22–40) and 24 mm (22–30) (p = 0.04) in MS patients with and without dysfunction. The level of NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, was of 108.95 ± 12.05 nM/ml in controls vs. 170.04 ± 18.76 nM/ml in MS (p = 0.002), NO<sup>−</sup> <sup>2</sup> was of 33.78 ± 3.41 vs. 43.95 ± 2.59 nM/ml (p = 0.03), citrulline 62.65 ± 3.46 vs. 72.81 ± 4.35 µMol/ml (p = 0.06). VCAM median was 39 pg/ml (0–86) vs. 32 pg/ml (11–66) (p = 0.03), respectively. The correlation of VCAM with triglycerides (TG) was of 0.62 (p = 0.005). There were no differences in TNFα, IL-1, IL-6, IL-8, IL-10, and osteopontin. MS endothelial dysfunction is related to aortic diameters, and increased levels of VCAM, <sup>L</sup>-citrulline and NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, NO<sup>−</sup> 2 . VCAM-1 has a significant correlation with TG and could play a significant role in endothelial dysfunction.

Keywords: Marfan syndrome, flow-mediated vasodilation, endothelial dysfunction, inflammation, nitric oxide

# INTRODUCTION

fphys-09-00965 August 17, 2018 Time: 18:41 # 2

Marfan syndrome (MS) is rare disease with a dominant autosomal hereditary pattern that has an overall incidence of 3/10,000. It is related to 1,700 mutations in the fibrillin-1 (FBN1) gene (Dietz, 2017). The FBN1 gene is 250 kb long; it is composed of 65 exons, and is located in chromosome 15q-21.1. From the mutational repertoire, up to 25% of mutations can be de novo mutations and about 10% have not been identified (Judge and Dietz, 2005). FBN1 is essential component of elastic and inelastic connective tissues. FBN1 helps in the transference of the hemodynamic burden and in the alignment of the fibers along the direction of parietal stress. Therefore, FBN1 is involved in the protective mechanisms that prevent over-distension of elastin, improving arterial elasticity. This can result in flow-mediated vasodilation (Eberth et al., 2009). The augmented activity of the TGFB signaling pathway may lead to elastic fiber disruption and to an increase in collagen reservoirs in MS (Yang H.H. et al., 2010).

The ascending aorta is commonly affected by dilation and/or dissection in this disease, that constitute the main causes of morbidity and mortality. Experimental studies in vascular hemodynamics in homozygotic mutant mice to mgR have shown mechanic alterations secondary to vascular structural changes that compensate for the lost elasticity, hence maintaining intravascular hemodynamic homeostasis (Eberth et al., 2009). Ultrasound measurements in MS patients have shown a delayed expansion and synchrony during systole in carotid arteries.

Endothelial dysfunction has been proposed as a mechanism accounting for aortic dilation (Takata et al., 2014). There is an increase in the inducible nitric oxide synthase (iNOS) in animal models (Yang H.H. et al., 2010) and other studies have described diminished levels of phosphorylated eNOS and augmented levels of iNOS. The increase in iNOS was associated with over-production of NO in SM patients. Excessive iNOSdriven NO production causes cellular damage via accumulation of peroxynitrites (ONOO−). Peroxynitrites are associated to the inflammatory pathway that is one of the main players in the formation of aortic aneurysms in MS patients (Soto et al., 2016a).

Increased vasodilation in MS may be related to other aspects besides NO availability. Murine model studies have shown participation of the cyclooxygenase pathway. Diminished levels of alpha 2 thromboxane, mild expression of type 1- cyclooxygenase and of an increase in type 2- cyclooxygenase also play an important role (Soto et al., 2018). These factors lead to an increase in I<sup>2</sup> prostaglandin levels. These factors have as a consequence an overall diminution of the contraction of the thoracic aorta and to a severe compromise its structure and function (Chung et al., 2007b).

In addition, an animal model of MS, showing IL-6 deficiency, partially preserved the structure of the extracellular matrix, suggesting a role for IL-6 in the pathologic remodeling of the aortic wall (Ju et al., 2014). An increase in IL-6 in the adventitia has been found in the dissection site of human aneurisms (Doyle et al., 2012). A study done in IL-1β deficient mice, revealed diminished aneurism progression, low levels of cytokines and metalloprotease 9 (MMP9). These results point out that IL-1β may be a potential target for the treatment of aneurisms in the thoracic aorta (Johnston et al., 2014). Furthermore, IL-10 levels are significantly reduced in MS patients when compared to controls subjects (Kadoglou et al., 2012).

In spite of these findings in the aortic disease of MS patients, it remains unclear whether these mechanisms are the only ones associated with endothelial dysfunction. Abnormal response to flow mediated vasodilation could also be related to the mechano-transductional mechanisms (Wilson et al., 1999). The production and release of NO responds to biomechanical effects and therefore, the study of endothelial dysfunction by bidimensional ultrasound and Doppler sonography of the brachial artery might be useful for the assessment of flow-mediated vasodilation. Friction-force stimulus generated by sudden blood flow over the brachial artery might lead to endothelial NO release with a concomitant measurable vasodilation (Corretti et al., 2002). However, since NO production in serum is unlikely to happen, in the present study we evaluated serum levels of NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, NO<sup>−</sup> 2 and citrulline, as alternative biomarkers to detect endothelial dysfunction and their correlation to flow-mediated vasodilation.

# MATERIALS AND METHODS

This is a comparative and prospective cohort observational study that took place between the years 2015–2016.

# Population Under Study

We included MS patients, evaluated by a rheumatologist using Ghent criteria (Loeys et al., 2010). Male and female subjects with an age above 18 years, without chronic or acute disease, or neoplasms were included. All individuals had a 7 day wash-out of statin, NSAIDs, calcium antagonists, oral nitrates and beta blockers, with negative serology for HCV, HBV, HIV, syphilis, and Chagas disease. Controls were paired by age and gender, from blood bank volunteers. All participating healthy subject were given a clinical record and were subjected to physical exploration to determine the absence of clinical MS criteria. Controls were also analyzed by the same serologic tests to discard infection.

# Exclusion Criteria

fphys-09-00965 August 17, 2018 Time: 18:41 # 3

For MS cases, patients that had not suspended statins, NSAIDs, calcium antagonists, oral nitrites, and beta-blockers 7 days prior to sampling or ultrasound; subjects with a previous aortic surgery or with associated comorbidities such as diabetes, thyroid disease, arterial hypertension, coronary disease, peripheral arterial disease, angioplasty of the upper limbs, cervical sympathectomy, smoking or that were unwilling to sign informed consent were excluded.

For controls, exclusion criteria were: individuals with first and second- degree familiar relationships with MS or similar diseases or thoracic aneurism related maladies. Pregnant women and those in menopause or menstruating were also excluded.

# Laboratory Tests

HDL and LDL lipoprotein, triglycerides (TG), and serum glucose were tested. Sample obtainment and flow-mediated vasodilation were performed the same day with a maximum delay of 2 h between sample obtainment and cabinet studies to evaluate endothelial dysfunction. Both studies were done under fasting conditions. For the determination of NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, NO<sup>−</sup> 2 and citrulline it was required that both, patients and controls did not perform physical exercise during 24 h previous to sample obtainment. Consumption of flavonoids, theobromine, some fruits and vegetables, olive oil, fish oil, beef or pork, red wine, chocolate, coffee, cocoa, soy, or tea was avoided by all included subjects.

### Ethics Statement

The study was carried out according to the international ethical standards and the General Health Law, as well as according to the Helsinki Declaration, modified at the Congress of Tokyo, Japan. Also, this protocol (PT 15-15) was approved by the local ethical committee. All of the patients and controls read and signed an informed consent form.

# Flow-Mediated Vasodilation

Flow-mediated vasodilation was determined by brachial artery ultrasound sonography. The study took place in the radiology department at our institution, in a temperature controlled room (21–23◦C), under fasting conditions including both liquids and solids, for a minimum of 8 h, and after a 7 day wash out of the already mentioned anti-inflammatory drugs. Four hours previous to the study, subjects had not ingested caffeine, c vitamin, smoked or chewed tobacco and they had restrained from physical exercise. Women were not studied during menstruation. Ultrasound studies were performed with a Phillips Envisor device that is able to measure colored and pulsatile Doppler (to determine the maximum peak of flow in the reactive hyperemia phase). The device was synchronized with a monitor that is able to register cardiac frequency and a high frequency transducer (7–12 MHz) was used. The subject under study was placed in supine decubit position with the arm in a comfortable position to allow for brachial artery detection. The subjects remained at rest for 10 min prior to basal image acquisition. A clear segment of the artery was found by locating the anterior and posterior portions of the intima and by using grayscale bidimensional ultrasound over a longitudinal plane to the artery (5 cm) (see **Figure 1**).

To create a stimulus similar to that producing flow-like vasodilation, a sphygmomanometer was used, placing it over the antecubital pit (or below) to take a basal image of the artery. Next, the sphygmomanometer was inflated to at

FIGURE 1 | Shows a segment of the artery located at the anterior and posterior portions of the intima. This image was obtained by using grayscale bidimensional ultrasound over a longitudinal plane to the artery.

least 50 mmHg over the systolic pressure of the patient for 5 min. A continuous image was recorded for 30 s prior to the beginning of deflation and finished 2 min after deflation completion (in total, 150 s were recorded). Continuous velocity recording of this period (deflation or reactive hyperemia) is shown in **Table 4**. Images taken at 60 and 120 s after deflation were analyzed. The tele-diastolic period, and the peak of the R wave (four measurements of the lumen of the vessel were made and the average was taken for the analysis of fractional changes) were taken into account for the measurements. The percentage rate of change in diameter of the brachial artery was calculated as follows (**Table 2**):

(Maximum diastolic diameter −

Basal diastolic diameter) × 100.

# Basal Diastolic Diameter

An increase greater or equal to 10% in the size of the diastolic diameter of the brachial artery as related to the basal image, was taken as evidence of response to the maneuver, 60 s into the reactive hyperemia period (after sphygmomanometer deflation) with the subjects being their own controls.

# Sample Size

Since it was an exploratory study, there is no previous evidence to support for statistical sample size calculations. The number of MS patients entering the National Institute of Cardiology "Ignacio Cháve" (NIC) varies from 15 to 2 per year. Age- and genderpaired healthy controls were included for comparison. Average flow velocity (AFV) with a fractional change of less than 10% in the diastolic diameter of the brachial artery is considered as a sign of true damage when its change is around 0.02. In contrast, for a healthy population, the fractional change in AFV with a change in diameter is greater than 10%. To this end, we recruited 33 cases and 33 controls adjusted for losses. The model is as follows:

$$\mathbf{H}\_{\bullet} \colon \mathbf{p} = \text{po versus } \mathbf{p} \neq \text{po}$$

$$A = 0.051 - \beta = 0.80$$

P<sup>o</sup> = 0.10

$$P\_1 = \text{ } 0.02$$

$$n = \frac{popo\left[\text{Z}\_{1-\alpha/2} + \text{Z}\_{1-\beta}\sqrt{\frac{p\_1q\_1}{popo}}\right]^2}{p\_1 - po}$$
 
$$n = \frac{0.10\left(0.90\right)\left[\text{Z}\_{0.975} + \text{Z}\_{0.80}\sqrt{\frac{0.02\left(0.98\right)\_1}{0.10\left(0.90\right)}}\right]^2}{\left(0.08\right)^2}$$

# Variable Specification

### Endothelial Dysfunction

Endothelial dysfunction was assessed by brachial artery ultrasound and defined as a response lower than 10% in the enlargement of the diastolic diameter (as compared to basal measurement) in the 60 s of the reactive hyperemia (posterior to sphygmomanometer deflation).

### Ascending Aorta Dilation

Ascending aortic dilation was defined as an absolute value greater than 34 mm for males and greater than 31 mm for females.

### Citrulline Determination

Hundred microliter of serum were incubated for 30 min at 37◦C after addition of 50 µL urease 12 mg/mL and of 3 mL of chromic mixture. The chromic mixture consisted of 25% H2SO4, 20% H3PO4, 9.24 µM FeCl3·6H2O, 0.125% 2,3-butanedione monoxime, and 0.0075% thiosemicarbazide mixed by vortex and incubated at 100◦C for 5 min. The samples were cooled to room temperature and the color developed was measured at 530 nm. The calibration curve was made with a standard solution of L-citrulline 1 µmol/L from Sigma-Aldrich (St. Louis, MO, United States) (Pérez-Torres et al., 2016).

### Nitrate and Nitrite Quantification

For NO<sup>−</sup> 3 /NO<sup>−</sup> 2 quantification, 100 µl of previously de-proteinized serum were incubated with 50 µl Cu-Cd for 30 min. The mixture was centrifuged at 850 g, at room temperature for 5 min. and the supernatant was recovered and incubated in the presence 100 µl of sulfanilamide 1% and 100 µl of N-naphthyl-ethyldiamine 0.1%. The total volume was adjusted to 1 ml (Pérez-Torres et al., 2016). For NO<sup>−</sup> 2 , 100 µl of previously de-proteinized serum were added to 100 µl of sulfanilamide 1% and 100 µl of N-naphthyl-ethyldiamine 0.1% and the total volume was adjusted to 1 ml. Both quantifications were measured at 540 nm. The calibration curve was obtained using a solution of KNO<sup>2</sup> at concentration ranging from 5 to 0.156 nM.

#### Inflammatory Interleukins

Levels of human inflammatory interleukin mediators were measured with Quantikine ELISA assays (R&D systems,


BMI, body mass index. Gender expressed as frequency and percentage, other variables as mean and standard deviation (SD). A value of p = 0.05 was considered statistically significant.


M, men; W, women; FH, family History. Systemic Positive Score is a total ≥7/20 points.

Minneapolis, MN, United States), using specific kits to TNFα (Cat DTA00C), IL-1β (Cat. DLB50), IL-6 (Cat. D6050), IL-8 (Cat. D8000C), IL-10 (Cat. D1000B), Osteopontin (Cat. DOST00), VCAM (Cat. DVC00), and ICAM (DCD540). Briefly, in a 96 well polystyrene microplate coated with a primary anti-cytokine monoclonal antibody, 200 µl of standard (reference curve), samples or control were added to each well, in duplicates. The plate was incubated for 2 h at room temperature. After that, each well was aspirated and washed with 300 µl of wash buffer. This step was repeated three times, and then the plate was inverted and blotted against clean paper towels. After that, 200 µl per well of secondary antibody anti-cytokine conjugate to horseradish peroxidase were added, the plate was then incubated for 2 h at room temperature. Next, the plate was aspirated and washed again four times. Two hundred microliter of substrate solution were added to each well. The plate was incubated for 20 min at room temperature and finally 50 µl of stop solutions were added to each well. Plates were read at 450 nm using a microplate reader Opsys MR (Dynex Technologies, Chantilly, VA, United States).

# Statistical Analysis

All variables were assessed and subjected to a Kolmogorov– Smirnov test to determine their distribution. Based on this, the appropriate central tendency and dispersion measures for their descriptive analysis were set.

Bivariate analysis of the relevant variables and statistical inference (where applicable) were carried out via either c<sup>2</sup> or exact Fisher tests. Central tendency measures were evaluated by Student's t-tests or Mann–Whitney U-tests for continuous variables and c<sup>2</sup> or exact Fisher tests for the categorical ones. Associations were performed via Pearson correlation linear models. Data analysis was carried out using SPSS version 22.

# RESULTS

A total of 67 subjects were studied, of which 38 (57%) were female and 29 (43%) male. Median age was of 26 years with minimum and maximum values of 13–56. Demographic data are shown

in **Tables 1**, **2**. A total of 32 MS patients were considered of which 21 (66%) presented familiar background of MS and 20 (63%) suffered from crystalline luxation. 10 males and 13 females showed aortic dilation with a mean dilation value of 46 ± 11 mm, whereas the non-dilated individuals showed an average value of 30 ± 2 mm. All of the MS cases showed a systemic score greater than 7 points, with a median of 10 and extreme values (7–15 points) (**Table 2**). Mitral prolapse was found in 14 patients (43.8%). Serum biomarkers for cases and controls are shown in **Table 3**.

In the flow-mediated dilation analysis, we found a basal diameter of 0.33 cm (Minimum: 0.31–Maximum: 0.50) for the controls vs. 0.35 cm (0.26–0.56) for MS patients (p = 0.04).

Regarding diastolic diameter after the first minute, there were no significant differences; the median for controls was of 0.37 cm (0.26–0.48) vs. 0.38 cm (0.28–0.59) in MS patients (p = 0.49).

Fractional change between basal diameter and diastolic diameter at 1 min was significantly different between groups with 12.5% ± 7.1 for controls and 5.3% ± 8.5 for MS subjects (p = 0.001, **Table 4**). However, the fractional change after 2 min did not show significant differences between cases and controls (6.2% ± 7.4 vs. 5.1% ± 6.6, p = 0.49, respectively). Twentyone of the MS patients showed endothelial dysfunction and the percentage rate of change for flow-mediated dilation was 1.2% (−13 to 9.4%) for the ones with dysfunction and there were 11 cases without dysfunction. For these, the change was of 12.6% (10–23%; p = 0.01).

The levels of NO<sup>−</sup> 3 /NO<sup>−</sup> 2 and NO<sup>−</sup> 2 in controls was 108.95 ± 12.05 and 33.78 ± 3.41 nM/ml in MS patients with 170.04 ± 18.76 and 43.95 ± 2.59 nM/ml (p = 0.002 and p = 0.03, respectively, **Table 3**). Citrulline levels in controls were of 62.65 ± 3.46 vs. 72.81 ± 4.35 µMol/ml in MS patients (p = 0.06, **Table 3**).

We found an inverse correlation of HDL and citrulline levels for patients with MS and endothelial dysfunction R = −0.50 (p = 0.01). Similarly, there was a correlation for cholesterol vs. citrulline levels R = −0.43 (p = 0.03), as well as a positive correlation between citrulline and ICAM, R = 0.54 (p = 0.04),


Values expressed as median (minimum value − maximum) and mean ± standard deviation (SD). Tests to compare measures of central tendency were Student's t and Mann–Whitney's U. A value of p ≤ 0.05 was considered significant.

osteopontin R = 0.33 (p = 0.07) and VCAM R = 0.33 (p = 0.06). There was also a significant correlation between the levels of citrulline and the diameter of the ascending aorta R = 0.62 (p = 0.04). In patients with endothelial dysfunction, there was also an inverse correlation between HDL and the dilation of the sine tubular junction, R = −0.46 (p = 0.03) and with the ascending aorta diameter R = −0.39 (p = 0.07). TG showed a direct correlation with the diameter of the aorta at the level of the sine tubular junction R = 0.58 (p = 0.006). The levels of NO<sup>−</sup> 3 /NO<sup>−</sup> 2 in patients with endothelial dysfunction showed a direct correlation with IL-1. R = 0.55 (p = 0.01). In patients without endothelial dysfunction, there was a significant inverse correlation between citrulline levels and total cholesterol R = −0.57 (p = 0.03).

# DISCUSSION

# Flow Mediated Vasodilation

Flow mediated vasodilation has been used as a tool to detect endothelial dysfunction in individuals with cardiovascular risk and to prevent macro and microvascular events (Al et al., 2001). Endothelial dysfunction in MS subjects is present even before a structural change in the vessels can be detected. Although, the two mechanisms may be thought as independent, structural alterations are associated with high levels of ONOO<sup>−</sup> and with a poor response to vasodilating drugs. Thus, a cycle of inflammation and damage to the elastic fibers in the arterial vessels might be established (Pereira et al., 1999).

This work is one of the few studies in human MS subjects in which endothelial dysfunction was found in 62% and correlated to the presence of aortic dilation, in up to 45% of them. This value is higher than the one found in subjects without endothelial dysfunction (p = 0.01). This finding confirms what has been suggested in previous studies (Takata et al., 2014).

It is worth mentioning that not all of the patients presented flow-mediated endothelial dysfunction. In 11 cases, there was almost the same percentage change in the dilation than in the control group (12.6 vs. 12.5%), unlike the findings in the study by Wilson et al. (1999). However, we found that these MS patients with no endothelial dysfunction had a lower diameter of the aorta at the level of the annulus (ring) (24 vs. 27 mm, p = 0.04). They also showed a tendency to have a greater diameter in all other segments of the aorta similar to those found in patients with endothelial dysfunction. This observation is comparable to those of Takata et al. (2014).

This study showed a difference in the size of the basal diameter of the brachial artery between MS and controls. This result is relevant since the control group actually responded in the maximal hyperemia phase. This means that controls also have a change in flow-mediated vasodilation. However, this response differs from vasodilation in MS patients in that it is only present after 1 min. This result contrasts with the reports by Wilson et al. (1999).

NO measurements are quite relevant to evaluate the functional status of the endothelium. However, NO has a very short mean life, which makes it difficult to



A value of p ≤ 0.05 was considered significant.

fphys-09-00965 August 17, 2018 Time: 18:41 # 7

measure it in serum. For this reason, we measured the metabolites NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio and NO<sup>−</sup> 2 (Baylis and Vallance, 1998).

The aortic tissue of MS patients has showed high levels of ONOO<sup>−</sup> and activity of the iNOS (Soto et al., 2014) associated with the development of aortic aneurisms. Here, we showed a tendency to an increased citrulline level of up to seven times in the serum of subjects with endothelial dysfunction related to increased lipid levels. Citrulline is a metabolite of the NO synthases pathway. We also observed increased NO<sup>−</sup> 3 /NO<sup>−</sup> 2 in MS patients independently of their endothelial dysfunction status. This indicates that structural changes in the arteries of MS patients prevent a response to NO generated by the friction forces (either by blood flow or vortices inside the lumen of the vessel) (Soto et al., 2016b). The axial pressure exerts perpendicular stress in the vessel (Syyong et al., 2009) as well as structural alterations in the endothelial cells (Chung et al., 2007a). Here, we established a correlation between the serum levels of NO metabolites and endothelial dysfunction assessed by flow-mediated vasodilation for the first time. Previous studies by our group have shown the involvement of the glutathione (GSH) system, that becomes exhausted with a high activity of glutathione reductase (GR) and a diminished activity of Glutathione-S-transferase (GST) and glutathione peroxidase (GPX), as well as an increase in lipoperoxydation, which was associated with ONOO<sup>−</sup> increase (Zúñiga-Muñoz et al., 2017) in MS aortic tissue. These findings all together point to key mechanisms in the promotion of molecular and structural alterations in the thoracic aorta of MS patients.

Studies in animal models of MS have found that there is vasomotor dysfunction in the thoracic aorta, which may be associated with the accumulation of ONOO<sup>−</sup> (Yang H.H. et al., 2010; Soto et al., 2016a), leading to endothelialdependent vasodilation and vasoconstriction changes. These changes compromise smooth muscle contractility and increase vessel rigidity (Jondeau et al., 1999). The association between flow-mediated dilation and the diameter of the ascending aorta has been found to be negatively correlated in MS subjects with aortic dilation (Takata et al., 2014). Aside from static structural alterations like changes in the diameter of the aorta, there are other changes associated with endothelial dysfunction and bad prognosis. MS patients have a positive correlation between carotid pulse pressure (as a proxy for central pulse pressure, as a parameter for aortic distortion) and the diameter of the ascending aorta, independent of age and body surface, with negative correlation in healthy subjects. These findings associate flow-mediated endothelial dysfunction with both, static and dynamic structural alterations of the arterial vessels (Jondeau et al., 1999).

# Dyslipidemia Is Associated With Endothelial Dysfunction

When we evaluated dyslipidemia and its correlation with the endothelial function, we found that there is a tendency of lower HDL levels (down to 19 times less), as well as high LDL and TG levels in subjects with endothelial dysfunction than in subjects without it, and this finding is consistent with the findings of Liao et al. (1995). We also found evidence of statistically significant negative correlation between low HDL levels and increase of the aortic diameter in the sino-tubular segment. This dilation was positively correlated with increase in TG. It remains unclear whether the clinical relevance of these findings, however, could be associated with some implications of the alterations in the synthesis and bioavailability of NO during endothelial dysfunction and dyslipidemia. Saturated fat inhibits the production of NO, whereas polyunsaturated fat favors it, by mechanisms yet unknown. In contrast with polyunsaturated fatty acids, oleic acid inhibits the activity of eNOS, leading to lower synthesis of NO by this pathway. However, it may increase the activity of iNOS, thus favoring NO synthesis (Soto et al., 2016a). Indeed, NO formed by this pathway increases inflammation. It has been reported that MS is associated with oxidative stress and the presence of this state, may favor the oxidation of NO to ONOO−, leading to an increase of metalloproteinases, TGFβ and the degradation of the elastic fibers which in turn, favor the development of aortic aneurisms and rupture via a negative feedback mechanism (Yang W.I. et al., 2010). The negative effect of hypercholesterolemia over NO synthesis and endothelial dysfunction is well-documented. Oxidized LDL may act via a multitude of mechanisms, namely, inhibition of arginine transport from the blood plasma to the vascular endothelium, lowering of eNOS synthesis, interference with intracellular trafficking of eNOS from the endoplasmic reticulum to the membrane caveolae, as well as an increase in the intracellular asymmetric dimethyl arginine ADMA concentration and lowering the levels of reduced coenzyme BH<sup>4</sup> (Liao et al., 1995).

Also, the increase in the intra endothelial concentrations of cholesterol favors the synthesis of caveoline-l. This protein binds eNOS forming an inactive complex. These processes constitute the basis for the biological foundation for the positive effect of statins over the endothelial function (Koh et al., 2004). We found a moderate correlation between citrulline and LDL and TG that showed a trend to increase 8 and 19-times higher

in subjects with MS and endothelial dysfunction, respectively. A moderate correlation (R = 0.5) was also found in relation to HDL. Here, the lesser HDL, the greater citrulline and NO<sup>−</sup> 3 /NO<sup>−</sup> 2 levels. Also during the inflammatory process, cell adhesion molecules are involved in the initiation and progression of atherosclerosis, as pro-inflammatory and pro-atherogenic proteins (Galkina and Ley, 2007). Furthermore was evaluated, the association of the intracellular adhesion molecules ICAM-1 and VCAM-1 with other biomarkers. In spite of not having found statistically significant differences, between cases and controls subjects or between patients with and without endothelial dysfunction, there was a moderate correlation in the patients with endothelial dysfunction and citrulline. We might explain these facts by the presence of a number of mediators, like inflammatory cytokines, TNFα, growth factors TGF-1β, free fatty acids, advanced non-enzymatic glycosylation products, LDL and angiotensin 1 (AT1), that act by stimulating their receptors at the cellular membrane. For instance, AT1 stimulation by angiotensin II promotes the synthesis of phospholipases C and D, leading to the formation of diacylglycerol and inositol triphosphate (Abe and Berk, 1998). Then, the Ca2<sup>+</sup> release that these mediators provoke, leads to the activation of protein kinase C that in turn stimulates the NADPH oxidase enzyme complex, generating the formation of reactive oxygen species (ROS). ROS activate nuclear factor NF-KB allowing the expression of pro-inflammatory genes such as cytokines and chemokines like the monocyte chemotactic factor-1 (Hulsmans and Holvoet, 2010). Once activated, these mediators lead to the expression of cell adhesion molecules such as ICAM-1, VCAM-1, and E selectin at the level of the endothelial surface (Bedard and Krause, 2007).

Reactive oxygen species formation is also able to activate protein phosphorylation processes, diminishing dephosphorylation, inhibiting tyrosinphosphatase activity and the favoring the formation of mitogen activated protein kinases that are also able to activate NF-KB (Guzik and Harrison, 2007). ROS also oxidate LDL, thus augmenting their atherogenic potential and residence time at the vascular intima by binding to proteoglycans. HDL oxidation driven by ROS diminishes their anti-inflammatory properties and reverses cholesterol transport capacities (Ragbir and Farmer, 2010). Furthermore, ROS molecules inactive NO by binding and decoupling the eNOS as observed in endothelial dysfunction, but it can also increase the oxidation of NO favoring the development of ONOO<sup>−</sup> (Gryglewski et al., 1986). Given our results, we have reasons to believe that this interaction represents the more prevalent mechanism behind endothelial dysfunction (Guzik et al., 2002). In addition, ROS coming from macrophages localized in the vascular intima promote the activation of MMPs leading to the degradation of the collagen capsule and rupture of the plaque (when it exists). The role that FBN-1 mutation plays and the factors and mechanisms associated with damage leading to MS are still to be completely unveiled by studying animal models and in clinical studies on human subjects (Comeglio et al., 2007).

Currently, we have evidence to believe that FBN-1, TGB-1, and TGB-2 are key starting points to the processes defining the MS phenotype. This, along with changes in the extracellular matrix and interaction with the signaling molecules just described, have provided some clues that, still need to be complemented with clinical and translational studies, allowing convergence of the clinician and basic scientists' views on the disease triggering mechanisms.

For instance, it is well-known that other mechanisms contributing to aortic damage in MS patients, like oxidative stress and lipid deregulation, show a correlation with endothelial dysfunction (Zúñiga-Muñoz et al., 2017). These (basic) findings could be used to redesign medical management of the patients, leading to improved prognosis and evolution. Patients could profit from timely anti-oxidant, statin medication as well as nutritional and exercise plans evaluated by means of clinical assays. It is highly relevant to study tissues. We consider that endothelial dysfunction is one of the first manifestations of vascular disease. Endothelial cells have gene expression that leads to alterations in the synthesis and processing of a highly regulated protein, which show a correlation with specific mechanic/hemodynamic physicochemical processes leading to adaptive responses (Davies and Tripathi, 1993). These responses give as a result important changes in the shape, orientation and organization of endothelial cells (Reidy and Langille, 1980), as well as in changes in the ionic response to flow variations which, in turn, decrease eNOS and NO expression that generate flow-mediated vasodilation (Sumpio et al., 1988). Several factors modify these functions at the vascular endothelium, leading to endothelial dysfunction, and to disequilibrium on the bioavailability of active substances of endothelial origin predisposing to inflammation, vasoconstriction and increased vascular permeability (Badimon et al., 1992; Dejana, 1996).

# CONCLUSION

In MS, there is flow-mediated endothelial dysfunction, which is correlated with an increase of the aortic diameter, NO<sup>−</sup> 3 /NO<sup>−</sup> 2 ratio, NO<sup>−</sup> 2 and lipids. Endothelial dysfunction is present even before it can be detected by structural changes in the vessels and could be determined by a non-invasive technique (ultrasound). Therefore, this study suggests that this approach should be implemented during the initial diagnostic phase. Also this study, suggest that use of timely antioxidant therapy, combined with nutritional and exercise regimes, and counseling that should be evaluated by means of randomized clinical trials.

# AUTHOR CONTRIBUTIONS

OL performed endothelial dysfunction studies and built databases. IP-T performed citrulline, NO<sup>−</sup> 2 and NO<sup>−</sup> 3 /NO<sup>−</sup> 2 assays, interpreted results, and manuscript reviser. RM performed and analyzed interleukin measurements. SC performed and analyzed computer tomography studies. AM contributed to blood analyses. CC contributed to general laboratory analyses. EH-L contributed to molecular biology discussion and manuscript writing. MS designed the study, diagnosed the patients, coordinated the general project,

REFERENCES

fphys-09-00965 August 17, 2018 Time: 18:41 # 9


supervised students and directed manuscript writing. All authors read and approved the final version of the manuscript.


cyclic tensional deformation. J. Vasc. Surg. 7, 130–138. doi: 10.1016/0741- 5214(88)90386-2


syndrome. J. Am. Soc. Echocardiogr. 23, 1310–1316. doi: 10.1016/j.echo.2010. 08.022

Zúñiga-Muñoz, A. M., Pérez-Torres, I., Guarner-Lans, V., Núñez-Garrido, E., Velázquez, E. R., Huesca-Gómez, C., et al. (2017). Glutathione system participation in thoracic aneurysms from patients with Marfan syndrome. Vasa 46, 177–186. doi: 10.1024/0301-1526/a00 0609

**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 Lomelí, Pérez-Torres, Márquez, Críales, Mejía, Chiney, Hernández-Lemus and Soto. 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.

# Lipid and Non-lipid Factors Affecting Macrophage Dysfunction and Inflammation in Atherosclerosis

Mark S. Gibson, Neuza Domingues and Otilia V. Vieira\*

Lysosomes in Chronic Human Pathologies and Infection, Faculdade de Ciências Médicas, Centro de Estudos de Doenças Crónicas, NOVA Medical School, Universidade NOVA de Lisboa, Lisbon, Portugal

Atherosclerosis is a chronic inflammatory disease and a leading cause of human mortality. The lesional microenvironment contains a complex accumulation of variably oxidized lipids and cytokines. Infiltrating monocytes become polarized in response to these stimuli, resulting in a broad spectrum of macrophage phenotypes. The extent of lipid loading in macrophages influences their phenotype and consequently their inflammatory status. In response to excess atherogenic ligands, many normal cell processes become aberrant following a loss of homeostasis. This can have a direct impact upon the inflammatory response, and conversely inflammation can lead to cell dysfunction. Clear evidence for this exists in the lysosomes, endoplasmic reticulum and mitochondria of atherosclerotic macrophages, the principal lesional cell type. Furthermore, several intrinsic cell processes become dysregulated under lipidotic conditions. Therapeutic strategies aimed at restoring cell function under disease conditions are an ongoing coveted aim. Macrophages play a central role in promoting lesional inflammation, with plaque progression and stability being directly proportional to macrophage abundance. Understanding how mixtures or individual lipid species regulate macrophage biology is therefore a major area of atherosclerosis research. In this review, we will discuss how the myriad of lipid and lipoprotein classes and products used to model atherogenic, proinflammatory immune responses has facilitated a greater understanding of some of the intricacies of chronic inflammation and cell function. Despite this, lipid oxidation produces a complex mixture of products and with no single or standard method of derivatization, there exists some variation in the reported effects of certain oxidized lipids. Likewise, differences in the methods used to generate macrophages in vitro may also lead to variable responses when apparently identical lipid ligands are used. Consequently, the complexity of reported macrophage phenotypes has implications for our understanding of the metabolic pathways, processes and shifts underpinning their activation and inflammatory status. Using oxidized low density lipoproteins and its oxidized cholesteryl esters and phospholipid constituents to stimulate macrophage has been hugely valuable, however there is now an argument that only working with low complexity lipid species can deliver the most useful information to guide therapies aimed at controlling atherosclerosis and cardiovascular complications.

Keywords: chronic inflammation, atherosclerosis, oxidized lipids, macrophage heterogeneity, lysosome dysfunction

#### Edited by:

Maria Teresa Cruz, University of Coimbra, Portugal

#### Reviewed by:

Dmitri Sviridov, Baker Heart and Diabetes Institute, Australia Kevin Woollard, Imperial College London, United Kingdom

> \*Correspondence: Otilia V. Vieira otilia.vieira@nms.unl.pt

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 12 January 2018 Accepted: 14 May 2018 Published: 26 June 2018

#### Citation:

Gibson MS, Domingues N and Vieira OV (2018) Lipid and Non-lipid Factors Affecting Macrophage Dysfunction and Inflammation in Atherosclerosis. Front. Physiol. 9:654. doi: 10.3389/fphys.2018.00654

**195**

# ATHEROSCLEROSIS INVOLVES MULTIPLE EXAMPLES OF DYSREGULATED CELL FUNCTION

# Excess Lipid Loading Induces Lysosome Dysfunction

Atherogenesis occurs when pro-inflammatory monocytes (Mo) are recruited to the intimal layer of the medium and large size arteries. Here, they differentiate into macrophages (MO) and encounter a plethora of modified lipid species such as oxidized low-density lipoprotein (oxLDL). MOs employ mainly scavenger receptors to facilitate the uptake of these modified lipoproteins, leading to the formation of foam cells (Libby et al., 2011). These lipid-rich cells secrete multiple pro-inflammatory mediators that propagate the development of the necrotic core of an atherosclerotic plaque, increasing its vulnerability. This stage also involves significant contributions from the T-cell populations residing in the lesion, however, their involvement in the inflammatory process is beyond the scope of this review and will not be discussed further. Long before this stage is reached, dysregulated lipid metabolism serves to establish the disease. Early atherogenesis involves the development of fatty streaks whereby Mo-derived MOs internalize retained ApoBcontaining lipoproteins, which become degraded in lysosomes, with excess free cholesterol (FC) trafficked to the endoplasmic reticulum (ER) (**Figure 1**). LDL particles ingested via the LDL receptor are degraded and cholesteryl esters (CE) are hydrolyzed in lysosomes to FC and fatty acids. FC is trafficked to peripheral cellular sites by a mechanism involving the proteins Niemann Pick C 1 and 2 (NPC1 & NPC2) (Ouimet et al., 2011) (**Figure 1**). Delivery of FC to the ER leads to downregulated LDL receptor expression and endogenous cholesterol synthesis through suppression of the sterol-regulatory element-binding protein (SREBP) pathway (Brown and Goldstein, 1997). In the ER, FC is esterified by acyl CoA:cholesterol acyltransferase (ACAT), and the resulting CE is packaged into cytoplasmic lipid droplets, which are a hallmark of foam cells (Brown et al., 1980). Two major pathways facilitate cytoplasmic CE clearance. The first involves the hydrolysis of cytoplasmic CE by cholesterol ester hydrolase (NCEH) and the resulting free cholesterol is mobilized away from the ACAT pool (Brown et al., 1980) and made available for efflux via ATP-binding cassette transporter A1 (ABCA1), scavenger receptor class B type I (SR-BI), and aqueous diffusion. Alternatively, cytoplasmic CEs in lipid droplets are packaged into autophagosomes, which are trafficked to lysosomes, where the CE is hydrolyzed by lysosomal acid lipase (LAL), generating FC for ABCA1-dependent efflux (Ouimet et al., 2011). This process is induced upon MO cholesterol loading, with lysosomal hydrolysis crucial for the mobilization of lipid droplet-associated cholesterol for reverse cholesterol transport (Ouimet and Marcel, 2012). At this stage, lysosomal hydrolysis and sterol clearance is effective. However, the progression of fatty streak lesions to unstable plaques is characterized by the substantial accumulation of CEs and FC in lysosomes (Miller and Kothari, 1969; Fowler et al., 1980; Jerome and Lewis, 1990), indicating a failure to adequately hydrolyse and clear them. This confirms lysosome dysfunction is a key event in late-stage atherosclerotic disease. This phenomenon has been observed in vivo and replicated in vitro. In pigeons, cholesterol trapped in the lysosomes of lesional foam cells remained trapped, even after intentionally reducing plasma cholesterol returned this parameter to normal levels and cytoplasmic CE droplets had been cleared (Jerome and Lewis, 1990). The loss of lysosomal hydrolysis was also demonstrated in THP-1 MOs exposed to mildly oxLDL or aggregated LDL (aggLDL). Initially, CE hydrolysis within lysosomes and FC efflux from this organelle was not inhibited. After prolonged exposure to these lipoproteins (>48 h), however, CE hydrolysis became increasingly inhibited and lysosomes began to accumulate CE (Yancey and Jerome, 2001; Griffin et al., 2005). A further study suggested the inhibition of lysosomal function is a general effect not related to the oxidation status of a given lipid. Here, the loss of lysosome function, including a reduction in LAL-dependent CE hydrolysis over time, was verified in MOs loaded with either mildly oxLDL or aggLDL or cholesteryl ester-rich lipid dispersions (Cox et al., 2007).

Furthermore, lysosomal dysfunction in oxLDL-loaded MOs is irreversible (Jerome et al., 2008). Collectively, these data indicate that inhibition of LAL activity, as well as of other lysosomal enzymes such as cathepsins (O'Neil et al., 2003), are the result of permanent alterations in lysosome function following excess lipid loading. The net effect is an accumulation of lysosome cargo (**Figure 1**). These findings are in agreement with data produced in our group using cholesteryl hemisuccinate (Chems), a member of the cholesteryl hemiester family. Cholesteryl hemiesters are produced as result of CE oxidation and have been detected in human atheromata (Hutchins et al., 2011) and oxLDL (Kamido et al., 1995). MOs exposed to LDL enriched in Chems demonstrated an irreversible accumulation of undigested lipid in enlarged lysosomes, which had increased Chems content in lysosomal membranes (Estronca et al., 2012; Domingues et al., 2017).

# Functional Impairment of Lysosomes Stimulates Inflammasome Assembly and Pro-inflammatory Cytokine Release

Lysosome function in both health and disease is intrinsically linked to cytokine release. Cytokines are synthesized by MOs after cell activation and secreted via the constitutive (or continuous) secretory pathway or by non-conventional secretion. The majority of cytokines expressed in MOs are processed and transported through the constitutive pathway; however, some require non-conventional secretion for cellular release. Secretory lysosomes facilitate this secretion and are also able to degrade inflammatory cytokines to regulate the immune response to an external stimulus, such as lipopolysaccharide (LPS) and adenosine triphosphate (ATP) (Murray and Stow, 2014). Lysosomes modulate cytokine production in other fundamental ways. For instance, TMEM9B, a glycosylated protein localized in lysosomal membranes regulates Tumour necrosis factor (TNF-) induced Interleukin-6 (IL-6) and IL-8 mRNA expression. It is also necessary for TNF-, IL-6 and Toll-like receptor 2- (TLR2-), TLR3-, and TLR4- induced IL-8 expression, demonstrating

FIGURE 1 | Functional impairment of lipid-engorged lysosomes stimulates inflammasome assembly and IL-1β release. The uptake of lipid and lipoprotein molecules occurs by a variety of scavenger receptor (SR), and Toll-like receptor (TLR)-mediated mechanisms such as phagocytosis or macropinocytosis. In early atherogenesis, monocyte (Mo)-derived macrophages (MOs) retain lipids and lipoproteins, which become degraded in lysosomes, with excess free cholesterol (FC) trafficked to the endoplasmic reticulum (ER). Cholesteryl esters (CE) are hydrolyzed in lysosomes by lysosomal acid lipase (LAL) to FC and fatty acids. FC is trafficked to the ER via Niemann Pick C1 (NPC1), where it becomes esterified by acyl CoA:cholesterol acyltransferase (ACAT), forming CE, which are packaged into cytoplasmic lipid droplets. This is a distinctive feature of foam cell formation. It is induced upon MO cholesterol loading, with lysosomal hydrolysis vital for the mobilization of lipid droplet-associated cholesterol for reverse cholesterol transport. This mechanism of sterol clearance is initially effective, however, when fatty streak lesions degenerate into unstable plaques, a substantial accumulation of CEs, their oxidized derivatives (oxCEs), and FC occurs in lysosomes due to inadequate hydrolysis and clearance. Lysosome dysfunction in excess lipid-loaded MOs is irreversible and is characterized by an inhibition of LAL and cathepsin activity, due to permanent alterations in lysosome function, resulting in the accumulation of cargo. Dysfunctional lysosomes become ruptured, releasing cathepsins, reactive oxygen species (ROS) and other molecules into the cytoplasm. These activate the NLRP3 inflammasome, leading to the maturation and release of IL-1β. Five potential mechanisms of IL-1β release have been described including the exocytosis of secretory lysosomes, exosomes and microvesicle shedding. Lys, lysosome; MVB, multivesicular body.

its control of pro-inflammatory signaling cascades (Dodeller et al., 2008). Additionally, lysosomes can down-regulate TLR9 mediated proinflammatory cytokine and type I IFN production in MOs by degrading the receptor following its translocation from the ER (Yao et al., 2009).

IL-1β has a fundamental role in establishing and driving the pathogenesis of atherosclerosis. It stimulates Mos and MOs as well as endothelial cells (EC) and smooth muscle cells (SMC) to secrete proinflammatory cytokines and chemokines (Libby, 2017). These cells also release increased amounts of specific matrix metalloproteinases (MMPs) in response to IL-1. These include MMPs with defined roles in EC erosion, SMC proliferation, remodeling and migration, and MO-mediated plaque rupture. IL-1 induced chemokines attract phagocytes to a plaque. Effects on cardiomyocyte function are also known (Libby, 2017). The significance of its role in atherosclerosis formed the basis of the CANTOS clinical trial (Ridker et al., 2017), which is discussed in a subsequent section within this article entitled Therapeutic Interventions to Treat Atherosclerosis. Given this critical role, here we will discuss the impact of lysosome function and dysfunction on its release. Five potential mechanisms of IL-1β release have been described (Eder, 2009; Martín-Sánchez et al., 2016), one of which is the exocytosis of secretory lysosomes (**Figure 1**). Early evidence of this process came from Mos in which the lysosomal membrane marker LAMP-1 co-localized with intracellular pro-IL-1β, pro-caspase-1 and cathepsin D. The latter was subsequently detected extracellularly (Andrei et al., 1999). In MOs, but not Mos, two distinct signals are required for IL-1β release. The first of these is a "priming" signal (Signal 1), which leads to synthesis of a biologically inactive precursor form of pro-IL-1β. Examples of lipid ligands that prime MOs are discussed in the section entitled A Broad Spectrum of Lipid and Lipoprotein Species Modulate Inflammation in Macrophages. To produce a mature, bioactive form of IL-1β, it is necessary for the pro-peptide to be cleaved by caspase-1 (Black et al., 1988). To synthesize a mature form of caspase-1, a multiprotein inflammasome complex needs to be activated and formed in the cytosol. Inflammasome assembly requires a second signal (Signal 2), which can be delivered by a number of different sources, including dysfunctional lysosomes (Tall and Yvan-Charvet, 2015) (**Figure 2**). Differences between the two cell types are believed to reflect an adaptation by monocytes to facilitate their role in patrolling a typically pathogen-free environment, hence a requirement to rapidly respond to danger signals. Macrophages, by contrast, are tissue resident so are almost continually exposed to a range of foreign antigens. Being overly sensitive to their presence would be dangerous for the host (Van De Veerdonk et al., 2011). Sterol-loaded lysosomal membranes have increased permeability, causing their contents to leak into the cytosol. This has been shown in MOs that demonstrated a leakage of lysosomal enzymes after cells had been incubated with oxLDL, a mixture of cholesterol oxidation products or cholesterol crystals (Li et al., 1998; Yuan et al., 2000; Emanuel et al., 2014). OxLDL entry into MOs led to the formation of cholesterol crystals after only 1 h (Duewell et al., 2010). These crystals grew in magnitude over time, were deposited in phagolysosomes and ruptured lysosomal membranes. Both oxLDL and exogenously applied cholesterol crystals induced Il-1β release in the absence of other stimuli, demonstrating they provide both signals (1 and 2) required to release this cytokine in MOs. In cathepsin B and L knockout (KO) bone marrow derived MOs (BMDM), crystal-stimulated Il-1β release is diminished but not abolished, confirming these cysteine proteases derived from dysfunctional lysosomes provide some of signal 2 (Duewell et al., 2010). The precise mechanism of cathepsin-mediated NACHT, LRR and PYD domains-containing protein 3 (NLRP3) inflammasome assembly remains elusive. A potential consequence of leaky lysosomes is an increase in the excretion of lysosomal enzymes, which may explain the presence of extracellular LAL in atherosclerotic lesions (Tapper and Sundler, 1992; Hakala et al., 2003).

In addition to the unregulated leaking of lysosomal proteases, particulate matter also triggers K<sup>+</sup> efflux from the cell (**Figure 2**). This is essential for the activation of NLRP3 inflammasomes induced by lysosomal destabilization (Muñoz-Planillo et al., 2013; He Y. et al., 2016). The mechanism linking particulate matter-induced lysosomal rupture to K<sup>+</sup> efflux is undetermined at present. Calcium phosphate also exists as particulate matter, and has been shown to accumulate in atherosclerotic lesions (Hirsch et al., 1990). Calcium phosphate crystals also initiate NLRP3 inflammasome assembly through lysosomal membrane rupture and cathepsin B release. They facilitate the release of both IL-1β and IL-1α in MOs (Usui et al., 2012). Data produced in our group observed that in MOs incubated with Chems, lysosomes become dysfunctional, and cells become overtly proinflammatory, releasing IL-1β in addition to other inflammatory cytokines (Domingues et al., 2017). However, in contrast to oxLDL and cholesterol crystals, Chems does not deliver both of the signals required for IL-1β release, with exposure to a low dose of LPS also necessary. In this experiment, a non-toxic concentration of Chems was used and it may be possible that with higher concentrations, lysosomes become more damaged, providing the second signal for IL-1β release.

# ER Stress and Mitochondrial Dysfunction Occur in Response to Lipid Loading

Under homeostatic conditions, the ER regulates the synthesis, processing, and folding of secretory proteins. An imbalance occurs when the synthesis of proteins overwhelms the folding capacity of the ER, leading to the accumulation of misfolded and unfolded proteins in the lumen. This is defined as ER stress and is implicated in the pathogenesis of many inflammatory disorders. Under ER stress conditions, a compensatory system called the unfolded protein response (UPR) is mobilized to mitigate ER stress signaling to restore homeostasis. Under pathologically chronic ER stress, cell dysfunction and disease manifest, potentially leading to cell death. Three upstream ER stress transmembrane sensor proteins respond to the presence of misfolded and unfolded proteins by triggering the UPR. They are activating transcription factor 6 (ATF6), inositol-requiring enzyme 1α (IRE1α), and protein kinase Rlike endoplasmic reticulum kinase (PERK). The latter augments the translation of ATF4, inducing the production of C/EBPhomologous protein (CHOP), which instigates a number of restorative functions to temper transient ER stress (Tabas, 2010). One of the now well-established consequences of ER stress is activation of the NLRP3 inflammasome (Menu et al., 2012). The precise molecular mechanism underlying this pathway is poorly understood, though it may be UPR independent (Menu et al., 2012). Regardless of the mechanism, ER stress is evidently a critical factor in the progression of atherosclerosis and has been reviewed extensively (Hotamisligil, 2010; Tabas, 2010). Through determining CHOP expression (amongst other markers) in MOs from early fatty streaks and advanced lesions, ER stress was manifestly elevated during the progression of atherosclerosis in chow- or Western diet-fed ApoE−/<sup>−</sup> mice. CHOP expression levels were found to increase with disease severity (Feng et al., 2003; Zhou et al., 2005). Similar findings were subsequently revealed in humans with unstable plaques (Myoishi et al., 2007). ER stress activates SREBP2, which delivers a source of "Signal 2" to activate assembly of the NLRP3 inflammasome (Reboldi et al., 2014) (**Figure 2**).

Mitochondrial dysfunction is also prominent during atheroprogression. Amongst the multitude of processes that occur within mitochondria, oxidative phosphorylation, through the production of ATP, liberates moderate (physiological) levels of superoxide, the bulk of which are converted to hydrogen peroxide by superoxide dismutase. This process remains homeostatic under normal conditions, however, under pathophysiological conditions; chronic overproduction of ROS arises from excessive mitochondrial oxidative stress. The link between mitochondrial oxidative stress (MitoOS)-induced pathology and atherosclerosis was made in an elegant study using transgenic mice in which MitoOS had been suppressed in

FIGURE 2 | Mechanisms of inflammasome activation in atherosclerosis. To facilitate the release of IL-1β (& IL-18) in macrophages (MO), two distinct signals must be delivered. Firstly, cells need to be primed (Signal 1) to activate the NF-κB-responsive genes NLRP3 and IL-1β. In atherosclerosis, a broad range of lipid and lipoprotein agonists provides this signal by activating cell surface pattern recognition receptors (PRRs). 1-palmitoyl-2-arachidonoyl-sn-3-glycero-phosphorylcholine (oxPAPC), 1-palmitoyl-2-glutaroyl-sn-glycero-phosphatidylcholine (PGPC), and 1-palmitoyl-2-(5-oxovaleroyl)-sn-glycero-phosphatidylcholine (POVPC) prime cells through CD14. Toll-like receptor 2 (TLR2) has also been described as an oxPAPC receptor. Free fatty acids (FFA) prime MOs through a TLR2-TLR4 signaling complex. The ApoAI moiety in High-density lipoprotein (HDL) can co-activate peritoneal MOs along with Lipopolysaccharide (LPS) through TLR2 or TLR4. Oxidized LDL (oxLDL) can prime MOs via signaling through a CD36-TLR4-TLR6 receptor complex. CD36 can additionally promote oxLDL uptake and its conversion into cholesterol crystals within phagolysosomes. The second signal (Signal 2) is required for assembly of the canonical NACHT, LRR and PYD domains-containing protein 3 (NLRP3) inflammasome complex. A number of different activating stimuli are able to deliver this signal. These include increased sterol-regulatory element-binding protein 2 (SREBP2) activity, mitochondria-derived reactive oxygen species (ROS), lysosome damage leading to cathepsin release, and exogenously-derived ATP from adjacent necrotic cells. Activation of surface P2X receptors leads to K<sup>+</sup> efflux through ion channels. Inflammasome activation produces mature caspase-1 which cleaves pro-IL-1β (or pro-IL-18), forming mature bioactive IL-1β that is released from the cell. Cathepsin inhibitors diminish IL-1β release by inhibiting NLRP3 inflammasome activation. Other chemicals shown to inhibit the NLRP3 inflammasome include MCC950, which also reduces atherosclerosis in mice, and CY-09 that works through an unknown mechanism. SCAP, SREBP cleavage-activating protein; INSIG, insulin-induced gene protein.

lesional MOs. These mice exhibited a reduction in lesional area of the aortic root, a lower abundance of Mos in these lesions and a concomitant reduction in Ccl2 and RelA phosphorylation, indicating the NF-κB signaling pathway responsible for proinflammatory cytokine production was also downregulated. Mitochondrial damage or dysfunction delivers, through the production of ROS, another source of "signal 2" to initiate inflammasome assembly (**Figure 2**).

Uncoupling oxidative phosphorylation in mitochondria reduces ATP production and may also decrease ROS production in a cell. Despite this, a study looking at the effects of oleic acid in mice found that this unsaturated fatty acid induced mitochondrial respiratory uncoupling and inflammasome independent IL-1α release to drive atherogenesis (Freigang et al., 2013). Il-1α release was mediated by an increase in calcium efflux from the mitochondria to process inactive pro-IL-1α via calcium-dependent calpains. Intriguingly, this study described mitochondrial dysfunction in the absence of overt ROS-mediated IL-1β production with IL-1α the more prominent cytokine (Freigang et al., 2013). This suggests the considerable effort undertaken to decipher the mechanisms of IL-1β production in atherosclerotic MOs may only tell part of the story.

Some of the research into these organelles is now aiming to identify specific dietary factors that promote ER stress, mitochondrial oxidative stress and/or ROS overproduction. For example, ROS-mediated NLRP3 inflammasome activation was established in BMDMs exposed to long-chain saturated fatty acids (SFA) (Wen et al., 2011). More recently, SFAs were shown to induce an ER stress response via an inositol-requiring enzyme 1-α (IRE1-α)-dependent pathway in BMDMs, promoting NLRP3 activation and IL-1β release (Robblee et al., 2016), though the long chain SFA palmitate does not induce these changes by directly binding TLR4 (Lancaster et al., 2018). An assessment of additional ligands that may affect these organelles is likely to be carried out in the near future.

# Apoptosis and Inefficient Efferocytosis Contribute to Plaque Progression

In advanced atherosclerotic lesions, MO apoptosis, often caused by ER stress and lysosome dysfunction; and plaque necrosis are prominent features of the disease pathology (Subramanian et al., 2015; Tabas and Bornfeldt, 2016). Both caspase-1-dependent pyroptosis (Lin et al., 2013b) and receptor-interacting protein kinase 3 (RIP3)-dependent necroptosis (Lin et al., 2013a) are additional forms of lipid-mediated cell death to have been observed in vitro, and are likely to also contribute to lesional MO death in vivo.

Necrotic cores arise following the death of advanced lesional MOs and within these lipid-rich regions, another example of cell dysfunction occurs to exacerbate plaque stability. The major functional feature of MOs is phagocytosis, which under infectious conditions involves degrading pathogenderived antigens, which become presented on their surface by MHC molecules. In atherosclerotic plaques, MOs perform a related yet distinct function known as efferocytosis. It involves removing apoptotic and other dying MOs by engulfment to prevent the establishment of a necrotic core, resolving inflammation. Generally, efferocytic MOs are more resistant to lipid accumulation, have an alternative anti-inflammatory phenotype, and reside in more stable regions of a plaque (Chinetti-Gbaguidi et al., 2015). In advanced plaques, efferocytosis becomes defective, leading to an accumulation of apoptotic MOs contributing to plaque necrosis and increased inflammation (Tabas, 2010).

# THE INFLAMMATORY STATUS OF LESIONAL MACROPHAGES IS REGULATED AT MULTIPLE LEVELS

MO-driven inflammation is fundamental to the progression of atherosclerosis, but precisely how these cells become polarized and evolve is an open debate. Data acquired over the past decade and a half has revealed a considerable amount of MO heterogeneity exists within atherosclerotic plaques (Chinetti-Gbaguidi et al., 2015). Depending on the lesional stimulus encountered, Mos differentiate toward a particular phenotype, which is characterized by cytokine release, surface markers, lipid or iron handling capabilities, and functions such as phagocytic capacity (Colin et al., 2014). MOs exhibit considerable plasticity and are thought to be able to switch their phenotype and inflammatory status in a plaque upon exposure to alternative stimuli (Chinetti-Gbaguidi et al., 2015). It is important to remember these phenotypes have been defined in vitro, under simplified conditions, and usually in response to a single stimulus. Given the complexity of the lesional microenvironment, they are unlikely to exist in their purest form in plaques and instead represent snapshots of a whole spectrum of states (Tabas and Bornfeldt, 2016). This viewpoint is gathering momentum (Nahrendorf and Swirski, 2016; Tabas and Bornfeldt, 2016), and mirrors a similar move by the field to redefine MO nomenclature according to functional status (Mosser and Edwards, 2008; Martinez and Gordon, 2014; Murray et al., 2014). So a more realistic view of plaque MO phenotype is a fluid, continually evolving entity, with the inflammatory status determined by a balance between the multiple signals they receive in vivo, lipid uptake and cellular metabolism of fatty acids (Tabas and Bornfeldt, 2016). Historically, a combination of technical limitations and a generally reductive experimental approach have failed to adequately evaluate plaque MO heterogeneity in vivo. Using a pre-defined panel of markers or secreted molecules such as cytokines, usually chosen to fit with the data from in vitro studies, is inherently biased and has undoubtedly limited the discovery of novel phenotypes. This has led to interpreting in vivo function by analogy. Highresolution data on other cell types known to reside in lesions has also been lacking. A brand new study has used single cell RNA sequencing to characterize aortic MOs and dendritic cells (DCs), revealing hitherto unknown levels of diversity amongst these cell types and their distinct subsets in vivo (Cochain et al., 2018). Alternative future approaches may include mass cytometry (Spitzer and Nolan, 2016) and ribosome profiling (Brar and Weissman, 2015), however, the latter has yet to be developed at the single cell level. Coupling this with a thorough metabolic and epigenetic characterization would greatly advance our understanding of how these cells become dysregulated under disease conditions.

# Macrophage Immunometabolism in Response to Lipids

Our understanding of what constitutes inflammation and how it is controlled is also changing (Editorial, 2017). Inflammation and metabolism are interminably entwined with both inflammatory and metabolic phenotypes capable of regulating one another. It is now well-known that MO phenotype and function are controlled by mitochondria acting as the hub for intrinsic metabolic pathways (O'Neill and Pearce, 2016). Glycolysis promotes proinflammatory functions in LPS + IFNγ-activated MOs, whilst in alternatively, IL-4 polarized MOs, oxidative phosphorylation regulates their anti-inflammatory phenotype (O'Neill and Pearce, 2016). This picture has become a little blurred more recently with both metabolic processes now recognized as having dichotomous roles in driving these opposing strands of inflammation (Van Den Bossche et al., 2017). A mechanistic understanding of how other ligands drive immunometabolism in MOs is currently lacking as most of the data accrued thus far is from in vitro polarized cells. A handful of studies have begun to assess the impact of lipid species on MO metabolism and inflammation in atherosclerosis (Bories and Leitinger, 2017). For instance, oxLDL-stimulated MOs are pro-inflammatory and exhibit enhanced glycolysis (Tawakol et al., 2015). This study also identified hypoxia-inducible factor 1-α (HIF1α) as the key driver of glucose uptake and the expression of glycolytic enzymes HK2 and 6- phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3). Another study also found that glutamine facilitates the lipotoxic effects of dietary fatty acids in MOs. In its absence, a reduction in lysosome dysfunction, inflammasome assembly and cell death was observed (Wen et al., 2011; He L. et al., 2016). In IL-4 polarized MOs, fatty acid oxidation following exogenous lipid uptake and lysis are important for the engagement of elevated oxidative phosphorylation and to preserve their antiinflammatory phenotype (Huang et al., 2014). Most of this data, however, has been derived in vitro, so the immunometabolic status of plaque MOs and how this evolves within the dynamic lesional microenvironment is mostly unknown. Consequently, the impact it may have on the progression of atherosclerosis remains scant.

# Dysregulated Inflammation in Foam Cells

Lesional MO abundance correlates with the progression and stability of plaques. The chronic inflammatory state within a plaque has a significant impact on this, as lipid-loaded foam cells are known to release inflammatory cytokines (Moore et al., 2013) and may also undergo necrosis to propagate the disease. The progression of inflammation in cells that are already full of lipid has therefore received some attention. A study on this theme (Spann et al., 2012) revealed that in peritoneal foam cells removed from LDLR KO mice on a high cholesterol, high fat diet; a significant number of pro-inflammatory genes had down-regulated expression. Most of these genes were highly expressed in response to a Tlr4 agonist in a previous study referenced by the authors. Amongst those demonstrating the highest degree of transcriptional suppression were Il-1β, Cxcl9, and Cxcl10. These findings were consolidated in vitro using murine peritoneal MOs (pMOs) and human Mo-derived MOs. In both cell types, cholesterol loading followed by Tlr4 stimulation [with Kdo(2)-lipid A, KLA, a TLR4-specific agonist] led to a significant inhibition of Cxcl9 and Cxcl10 expression, compared with KLA-treated cells (Spann et al., 2012). A similar trend was observed in elicited pMOs loaded with acetylated-LDL (Suzuki et al., 2012). In this study, a panel of LPS-inducible pro-inflammatory genes had down-regulated expression when subsequently challenged with this agonist, compared with cells that were only exposed to LPS (Suzuki et al., 2012). In human M-CSF-differentiated MOs converted into foam cells with acetylated LDL, a comparable picture emerged. Foam cells, along with untreated MOs, were exposed to polarizing stimuli (LPS & IFNγ) after which, the expression of several key pro-inflammatory cytokines was quantified. For every gene examined, lipid loading led to lower relative expression levels (Da Silva et al., 2016). A recent study attempted to decipher the mechanism underlying this trend. In line with previous reports, a general blunting of the pro-inflammatory response in lipid-loaded MOs was observed (Jongstra-Bilen et al., 2017). Specifically, exposure to oxLDL inhibited pro-inflammatory cytokine expression in pMOs subsequently stimulated with Tlr2, Tlr3, Tlr4, or Tlr9 agonists (Jongstra-Bilen et al., 2017). Mechanistically, the NF-kB RelA/p65 subunit demonstrated reduced binding to Il-6 and Ccl5 promoters, which was partially reversed with a broad-spectrum histone deacetylase inhibitor (Jongstra-Bilen et al., 2017). These results confirm that oxLDL can instigate epigenetic changes that regulate inflammation, corroborating previous in depth studies (Chen et al., 2011; Reschen et al., 2015). Further studies could seek to better characterize these foam cells and also identify whether different lipid species can dampen a common cohort of lipid handling and proinflammatory genes. In assessing why peritoneal foam cells had inhibited pro-inflammatory cytokine expression, (Spann et al., 2012) speculated that exogenous stimuli from within the artery wall might elicit a pro-inflammatory state in lesional MOs. Regardless, the suppressive effect of lipid loading on inflammation is clear though may be context-dependant with cell origin, the limitations of in vitro conditions, and technical differences in oxLDL preparation potentially affecting outcomes.

# Senescent Plaque Macrophages Are Dysfunctional

Despite the very clear and reproducible effect described above, numerous studies over many years have reported that lipid-laden plaque MOs can be markedly pro-inflammatory in progressing plaques. Recent research has elegantly demonstrated a link between their inflammatory capacity, advanced plaque stability and cell senescence (Childs et al., 2016). The study revealed that pro-inflammatory foamy MOs rapidly emerge and colonize fatty streak lesions in the inner curvature of the aortic arch in LDLR KO mice on a high fat diet. After only 9 days, some of these cells expressed senescence markers and by selectively eliminating them, lesional pro-inflammatory marker expression and streak size was reduced (Childs et al., 2016). In established plaques, removing senescent cells also diminished pro-inflammatory marker expression, reduced the size and amount of lesions and halted disease progression. Senescent cells isolated and flow sorted from advanced lesions expressed far higher levels of proinflammatory markers when compared with non-senescent cells (Childs et al., 2016). This study adds another layer of complexity to our understanding of inflammation in atherosclerotic MOs. It does, though, carry a degree of controversy in the context of other recent studies. For instance, plaque inflammation is greatly enhanced by proliferating MOs that drive and increase the severity of atherosclerosis (Tang et al., 2015). By contrast, senescent cells do not divide but they are also inflammatory and drive atherosclerosis (Childs et al., 2016) in the absence of proliferation, which was not quantified. It is possible that smooth muscle cell-like MO infiltration may be involved and contribute to perpetuating inflammation in the senescent cell-rich plaque areas.

# Circadian Rhythms Regulate Inflammatory Function

An important concept under an increasing amount of scrutiny is the circadian control of inflammation. A landmark study on the link between the circadian clock and immune function revealed that the rhythmic control of cytokine release in MOs is profound. The release of Il-6 & Tnfα in LPS-stimulated ex vivo murine MOs was shown to oscillate, confirming its regulation is cell intrinsic, not systemic. Furthermore, multiple genes at all cellular stages (surface, cytoplasmic, nuclear) of LPS-induced cytokine expression were evidently under circadian control (Keller et al., 2009). Recent examples of circadian rhythms pertaining to key processes that underpin atherogenesis and related vascular diseases have emphasized its importance. Individuals that suffer an acute myocardial infarction early in the morning experience greater damage to the heart and have a poorer recovery and long-term prognosis. This is attributable to a circadian controlled fluctuation in neutrophil recruitment to the heart, which is governed by oscillations in their surface expression of CXCR2 (Schloss et al., 2016). This study refers only to the circadian control of immune function affecting cardiac events. Yet, the strong connection between atherosclerosis and vascular disease opens up the possibility that the circadian control of inflammatory lesional components may also contribute to the timing and severity of an event, along with the known morning blood pressure surge (Marfella et al., 2007). In support of this, many fundamental components driving atherogenesis, plaque progression and plaque stability are known to be under circadian control. These include the expression of intercellular adhesion molecule 1 (ICAM-1) and several chemokines in endothelial cells; metalloproteinase expression in VSMCs; and cytokine, chemokine, TLR, and lipid handling genes in MOs (Mcalpine and Swirski, 2016). Phagocytosis, an indispensable function performed by MOs, has also recently been shown to be under circadian control (A-Gonzalez et al., 2017). Phagocytic MOs were shown to be anti-inflammatory, had dampened Il-1β expression regardless of the tissue source, and maintained tissue homeostasis by removing dead or dying cells–a process called efferocytosis (which has been reviewed extensively elsewhere, Tabas, 2010). This limited inflammatory cytokine production, thereby reducing leukocyte migration (A-Gonzalez et al., 2017). Whilst the relevance of this study was not discussed in the context of atherosclerosis, the function of efferocytic plaque MOs is conspicuously similar to the data they acquired from phagocytic MOs in other tissues, further emphasizing the probable circadian influence on the disease.

# A BROAD SPECTRUM OF LIPID AND LIPOPROTEIN SPECIES MODULATE INFLAMMATION IN MACROPHAGES

The biological properties of lipids and lipoproteins have been studied for decades. A major focus of this research has been to decipher their inflammatory capacity and, consequently, the impact they may have on atherosclerosis and other chronic diseases. Given the increasing desire for more translational studies and improved therapies, efforts to comprehend the fundamental elements of cell and organelle dysfunction underlying inflammatory changes are underway. One of the most important concepts to have recently emerged in the field of immunology is that of innate immune memory or "trained immunity." This arm of the immune system has historically been viewed as non-specific and lacking the capacity for memory. Trained immunity dictates that the innate immune response has the ability to adapt (Netea and Van Der Meer, 2017). Given the long-term chronic inflammation that characterizes atherosclerosis and the initial priming signals mediated by lipids, there was inevitably an interest in evaluating the effect in this context. So, do lipid signals train cells of the innate immune response? It appears that they do as exposure to atherogenic mediators induces a long-term activated phenotype in innate immune cells. Mechanistically, there are quantifiable changes in epigenetic programming and immunometabolism that define this altered phenotype, which modulates the perpetual inflammatory environment within a plaque. These findings are in contrast with those discussed in the subsection entitled "Dysregulated Inflammation in Foam Cells," and this may be related to the extent (and type) of lipid loading in the cells examined. Dietary factors are also known to affect the trained phenotype (Christ et al., 2016). Therefore, a better characterization of well-known and more recently discovered lipid mediators of inflammation and precisely how they alter cell function is essential. This is relevant for "priming" signals that rapidly affect transcription and intracellular signals; and also for metabolic, epigenetic, and inflammatory effects that occur following lipid ingestion and processing. This section will discuss how a selection of these lipids and lipoproteins affect inflammation and the potential consequences for cell physiology.

Lipids and lipoproteins undergo considerable modification following oxidative damage, rendering them profoundly immunogenic. Oxidized host molecules have an altered appearance or conformation due to the exposure of distinct moieties known as oxidation-specific epitopes (OSEs) (Miller et al., 2011). These OSEs operate as de facto damage-associated molecular patterns (DAMPs), and are therefore recognized by host pattern recognition receptors (PRRs), similar to how bacterial pattern-associated molecular patterns (PAMPs) are recognized by TLRs (Miller et al., 2011). Despite this, these DAMPs do not faithfully recapitulate the effects of microbial PAMPs. They are recognized, for instance, by TLRs & scavenger receptors, on host cells, inducing pro-inflammatory cytokine production. However, the precise nature of these interactions differs from pathogen recognition (Miller and Shyy, 2017). When lipids are oxidized to the extent that they become fragmented, producing "end products" of lipid oxidation, they are consistently pro-inflammatory. Some of the best-known examples are reviewed in detail elsewhere (Miller and Shyy, 2017).

# Oxidized Low-Density Lipoprotein and Cholesterol Crystals

To better understand lipid-mediated sterile immune responses in vivo, many researchers have established in vitro models using oxLDL, given its prevalence in atherosclerotic patients. Similar to "glaucoma," oxLDL is an umbrella term that describes a vast mixture of over 3000 molecules; comprised of apolipoprotein B-100, cholesteryl esters, free cholsterol, phospholipids, and triglycerides (Levitan et al., 2010). Although oxLDL is usually reported as being pro-inflammatory, particularly in Mos and MOs (Stewart et al., 2010; Sheedy et al., 2013; Tiwari et al., 2014; Rhoads et al., 2017), there are context-dependent examples of where it is not, as discussed in the previous section. As well as the cell types used, alternative outcomes may also be a consequence of the different methods used to modify LDL in vitro. It is probable that across the many studies published, no two batches of oxLDL are exactly alike, with variable amounts of oxidized immunogenic components affecting the intensity of inflammatory responses in MOs. The study describing how oxLDL promotes inflammation (Duewell et al., 2010) is discussed in the section entitled Atherosclerosis Involves Multiple Examples of Dysregulated Cell Function. Another key concept established by this study was the activation of NLRP3 inflammasome assembly by oxLDL & cholesterol crystals for both Il-1β release in vitro and atherogenesis in vivo. The mechanism of Nlrp3 activation was elucidated further in mouse MOs in vitro. OxLDL uptake, formation of cholesterol crystals and subsequent inflammasome priming and Il-1β release were dependent on Cd36 (Sheedy et al., 2013). In peritoneal MOs, Nlrp3 priming also required Tlr4 and Tlr6 and was driven by OxLDL-induced ROS production (Sheedy et al., 2013). Subsequent studies have reinforced the importance of the NLRP3 inflammasome as a critical driver of the disease (Paramel Varghese et al., 2016; Fuster et al., 2017), and therefore a key target for therapeutic intervention (Sheridan, 2017). Indeed, selectively inhibiting Nlrp3 restricts oxLDL uptake by THP-1 MOs, thus impairing foam cell formation. This inhibition also led to the downregulation of Cd36, which, in the context of the findings by Sheedy et al. (2013), almost certainly limited the appearance of foam cells (Chen et al., 2018). Several recent reviews have also focussed on NLRP3 as nexus between oxidized lipids and inflammatory cytokine release in atherosclerosis (Baldrighi et al., 2017; Patel et al., 2017; Hoseini et al., 2018). In contrast to the majority of available literature, one study found that in ApoE−/<sup>−</sup> Nlrp3−/<sup>−</sup> C57/BL6 mice on a high fat diet there was no difference in the degree of atherosclerosis between these and wild-type controls after 11 weeks. This suggested that the NLRP3 inflammasome had no impact on atherogenesis (Menu et al., 2011). This outcome has been directly contradicted in a recent study using a highly specific NLRP3 inflammasome inhibitor (MCC950) (Van Der Heijden et al., 2017). Here the data showed that atherosclerosis in ApoE−/<sup>−</sup> mice receiving MCC950 is reduced after 4 weeks of lesional development compared with controls. These mice were fed a Western diet containing cholesterol and butter. There are tangible differences between the methods used in these two studies, which likely affected the outcome. MCC950-mediated NLRP3 inhibition has already been shown to be highly specific in vivo and in ex-vivo murine Mos and MOs (Coll et al., 2015; Primiano et al., 2016), though the mechanism of its action remains elusive. A recent study has described another specific NLRP3 inhibitor (CY-09). This compound targets the inflammasome by inhibiting NLRP3 oligomerization and ASC recruitment and may offer a safe and selective therapeutic benefit to patients with atherosclerosis in future (Jiang et al., 2017). The use of cholesterol crystals was central to defining the role of NLRP3 in atherosclerosis (Duewell et al., 2010) and dietary cholesterol has also been shown to induce intestinal inflammation in vivo in zebrafish (Progatzky et al., 2014). The dietary form seems to lack the potency of crystals, as it requires a priming signal delivered by commensal microbiota to induce NLRP3-dependent inflammation (Progatzky et al., 2014).

# High Density Lipoprotein

Whilst LDL and its oxidized derivatives are typically proinflammatory, high-density lipoprotein (HDL) is commonly perceived as being anti-inflammatory, with a wealth of research supporting this view (Rye and Barter, 2014). HDL has a range of prominent functions, one of which is reverse cholesterol transport (Heinecke, 2012). Two membrane ATPbinding cassette transporters, ABCA1 and ABCG1, upregulated in cholesterol-rich MOs are responsible for mobilizing HDL to the liver via the circulation. The importance of HDL in limiting inflammation has been demonstrated in mice that have these transporters deleted in their MOs (ABCDKO mice), but not in their stem and progenitor cell populations. In LDLR−/<sup>−</sup> mice transplanted with ABCDKO bone marrow (BM), both inflammation and atherosclerosis are increased (Westerterp et al., 2013). In plaque MOs lacking these transporters, proinflammatory cytokine (Il-1, Il-6) and chemokine [Ccl3 (Mip1α), Ccl2 (Mcp1)] expression is significantly elevated (Westerterp et al., 2013). Directly challenging cells with HDL is an alternative approach that has been used to study its impact on inflammation. In MOs loaded with acetylated LDL, subsequently pre-treated with HDL then stimulated with LPS, a significant inhibition of LPS-mediated gene expression was observed (Suzuki et al., 2010). HDL selectively inhibited anti-viral response genes from the type I IFN response pathway that were TRAM/TRIFdependent. Some of these genes, such as Nos2 and Cxcl10, have described roles in promoting inflammation. Notably, pretreatment with HDL did not inhibit MyD88-dependent pathway genes, which included the pro-inflammatory cytokines Il-1b, Il-6, and Tnfa (Suzuki et al., 2010). A further study characterized the molecular mechanism underlying the HDL-driven suppression of inflammation (De Nardo et al., 2014). Two forms of HDL were used to similar effect—native HDL (nHDL) isolated from human plasma donors, and reconstituted HDL [HDL; pure human apolipoprotein A-I (ApoA-I) with phospholipids]. In mice injected with nHDL and subsequently injected with the Tlr9 agonist CpG-DNA, there were significantly lower levels of serum Il-6 and Il-12p40 1 h later. The identical result was observed with HDL, as well as lower levels of Tnf and Ccl2, but also Il-10. These data were validated in vitro in BMDMs pre-treated with HDL and subsequently challenged with a range of Tlr agonists. Mechanistically, HDL was shown to drive the expression of the transcriptional repressor, activating transcription factor 3 (ATF3), to downregulate pro-inflammatory cytokine production (De Nardo et al., 2014).

Two prominent studies have presented conflicting data, both convincingly showing HDL is pro-inflammatory in mouse MOs. The first of these (Smoak et al., 2010) used Apo-AI (the major protein constituent of HDL) to stimulate pMOs from Tlr2, Tlr4, MyD88, Tirap, and Trif knockout mice alongside wild type controls. All of these receptors and signaling adaptors respond to Apo-AI. MyD88−/<sup>−</sup> pMOs were also used to show ApoA-I induces release of Il-1α, Il-1β, Il-6, Tnf-α, Cxcl1, Cxcl2, and Ccl2 (De Nardo et al., 2014). The second, very recent assessment of HDL revealed that its pro-inflammatory capacity is cell-typedependant (Van Der Vorst et al., 2017). In endothelial cells and smooth muscle cells, the same group previously demonstrated that HDL is anti-inflammatory (Bursill et al., 2010; Van Der Vorst et al., 2013). Yet, they showed it is evidently pro-inflammatory in mouse MOs. HDL in isolation induced dose-dependent increases in Tnfα and Il-12 transcription, but the opposite trend was observed for Il-10. BMDMs stimulated with a range of nHDL concentrations then a low fixed [LPS] dose-dependently released Tnfα & Il-12. Intriguingly, BMDMs stimulated with a fixed concentration of either nHDL, reconstituted HDL (rHDL; ApoA-I complexed with 1-palmitoyl-2-linoleoyl-PC) or commercially purchased HDL (cHDL) then a low fixed [LPS] released Tnfα & Il-12 but failed to do so without the LPS present, suggesting HDL and LPS synergistically regulate cytokine release (Van Der Vorst et al., 2017). These findings were broadly replicated in vivo. In line with this finding, minimally modified LDL (mmLDL) and a low dose of LPS synergistically increase chemokine release in MO cell lines (Wiesner et al., 2010). Some of the experiments carried out by De Nardo et al, were replicated using soybean phosphatidylcholine rHDL, but got contrasting data (Van Der Vorst et al., 2017). They saw a similar reduction in Il-12 and IL-10 release, but Tnfα release and Atf3 expression were unaffected. They also speculated that the antiinflammatory function of the rHDL used in the De Nardo study was caused by the soybean content of the preparation (Van Der Vorst et al., 2017). Regardless, it is clear that understanding the biology of HDL remains a work in progress, with both pro- and anti-inflammatory functionality and its capacity to be dysfunctional identified and reviewed many years ago (Navab et al., 2009).

# Phospholipids

Oxidized phospholipids (oxPLs), formed following polyunsaturated fatty acid oxidation, are another major class of lipids that have been extensively studied in the context of inflammation and atherosclerosis. Their biology is also complicated and they initiate pro-inflammatory responses in a context-dependent manner. The transcriptome and functional phenotype of BMDMs exposed to the oxidation products of 1-palmitoyl-2-arachidonoyl-sn-3-glycero-phosphorylcholine (oxPAPC) is markedly different to that generated by LPS + IFNγ

(M1) and IL-4 (M2) polarization stimuli (Kadl et al., 2010). These MOs, termed Mox, express a narrow range of pro-inflammatory markers (Il-1β, Cox-2, Cxcl2, Cxcl1) (Kadl et al., 2010, 2011), though the amount of Il-1β released is considerably less than from M1 MOs (Kadl et al., 2010), however, this is commensurate with the typically lower levels of inflammation seen in chronic diseases. Tlr2−/<sup>−</sup> mice were used to confirm the OxPAPC receptor (Kadl et al., 2011). Interestingly, the inflammatory bioactivity was contained in the "long chain" fractions of OxPAPC, but not the short chain fractions or individual oxidized phospholipid constituents POVPC, PGPC, lysoPC (all short chain). Using heme oxygenase-1 (HO-1) as the Mox phenotypedefining marker, this population accounted for a third of all the MOs found in the aortas of LDLR−/<sup>−</sup> mice following a 30-week "atherogenic" diet (Kadl et al., 2010). The subsequent discovery that an anti-inflammatory subtype, termed Mhem (Boyle et al., 2012), also highly expresses Hmox-1 may indicate a lower abundance of Mox MOs in those mice. Furthermore, there may be undiscovered novel or transient phenotypes that express HO-1.

An intriguing pair of recent studies has further uncovered the complexities of oxPAPC bioactivity. Both reports are shaped around the release of OxPAPC from dying cells and how this affects adjacent cells, and whilst neither explicitly discusses atherosclerosis, their findings are applicable to the plaque microenvironment. Firstly, mouse DCs briefly primed with LPS then stimulated with OxPAPC, release Il-1β in an Nlrp3 dependent manner (Zanoni et al., 2016). When stimulated with either ligand in isolation, this cytokine is not released. This combined LPS/OxPAPC treatment induces something the authors describe as a "hyperactive state" where Il-1β release occurs in the absence of cell death. Active components of OxPAPC (KOdiA-PC, POVPC, PGPC) also induced Il-1β release in DCs, but not in isolation (LPS priming was required). Different forms of OxPAPC were also used, a commercially purchased form and a form the authors synthesized to be enriched in PEIPC, a known bioactive constituent of OxPAPC, and both induced Il-1β release to the same extent. Intriguingly, OxPAPC could not induce Il-1β release in MOs, whether primed or not (Zanoni et al., 2016). The same group further explored this final observation in a subsequent report. Here, they showed that the minor components (<10%) of OxPAPC, PGPC, and POVPC, are able to hyperactivate LPS-primed MOs, inducing Il-1β release, but OxPAPC itself cannot (Zanoni et al., 2017). As with DCs, the priming signal is essential for cytokine release. This study also identified CD14 as an OxPAPC receptor, whose endocytosis leads to OxPAPC delivery, hyperactivation, and Il-1β release.

A fascinating aspect of OxPL bioactivity is its ability to simultaneously induce inflammation and inhibit LPS signaling through TLR4 and TLR2. This has been known for many years (Bochkov et al., 2002; Walton et al., 2003) and occurs in Mos (Von Schlieffen et al., 2009). Multiple components of the extracellular signaling pathway are inhibited by oxPPLs, including CD14, LPS binding protein and MD-2 (Erridge, 2009). So what is the relevance of this in vivo? A subsequent study addressed this by demonstrating that oxPLs are far more potent at inhibiting LPS-mediated inflammation than their own ability to induce it (Oskolkova et al., 2010). This study examined a range of oxPLs aswell as components of oxPAPC (POVPC, PGPC, PEIPC) with all eliciting similar responses in human endothelial cells (HUVECs). Mice injected with LPS and/or oxPAPC replicated the trend observed in vitro (Oskolkova et al., 2010). Components of oxPAPC were present at significantly higher concentrations in the atheroma than in the plasma of humans and mice, so given their more prominent pro-inflammatory function at high concentrations, they may exacerbate the inflammatory plaque microenvironment. This of course relies upon their ability to induce release in the absence of a priming signal, which, except for the oxPAPC-induced Il-1β release seen in MOs (Kadl et al., 2010), was not apparent in the other studies referenced here. There may be other TLR4 ligands in the plaque microenvironment that fulfill this function or subclinical endotoxemia, known to aggravate atherosclerosis may provide this stimulus.

# Cholesteryl Esters

The diversity and sheer number of lipid species present in vivo alongside the difficulty in identifying some of them before refined methods became available, has inevitably meant some lipids have only recently been found whilst others have received little attention. A good example of this are oxidized cholesteryl ester (oxCE) molecules, which have only come under greater scrutiny over the past decade. They are abundantly present in oxLDL and are bioactive, inflammatory components of mmLDL, able to induce Cxcl2 release in mouse MOs (Harkewicz et al., 2008). Using High Performance Liquid Chromatography (HPLC) coupled with tandem Mass Spectrometry, a large and diverse collection of oxidized CEs, many of which were novel, were detected in human atheromata samples relatively recently (Hutchins et al., 2011; Choi et al., 2013). This study uncovered a greater abundance and diversity of oxCE species than had been previously believed to exist. From a therapeutic point of view, understanding the inflammatory capacity of oxCEs is a coveted aim and whilst some studies exist, they are currently thin on the ground. Amongst the extensively oxCEs to have been identified, is cholesteryl (9,11)-epidioxy-15-hydroperoxy-(5Z,13E)-prostadienoate (BEP-CE), produced following oxidation of cholesteryl arachidonate with 15-lipoxygenase. This highly oxygenated cholesteryl ester was shown to be pro-inflammatory in mouse BMDM. It induced Tlr4 dimerization and signaling and the release of Cxcl2. In Syk−/<sup>−</sup> and Tlr4−/<sup>−</sup> mice, cytokine release was abolished (Choi et al., 2013). Cholesteryl esters hydroperoxides are abundantly present in oxLDL and are bioactive, inflammatory components of mmLDL, able to induce Cxcl2 release in mouse MOs (Harkewicz et al., 2008). By contrast; 9-oxononanoyl cholesterol (9-ONC), an oxoester from cholesteryl linoleate identified in human atherosclerotic lesions (Hoppe et al., 1997), is not. It is derived from cholesteryl linoleate, upregulates the anti-inflammatory cytokine TGFβ in J774A.1 MOs (Sottero et al., 2005), and increases the expression and protein synthesis of TGFβ and the TGFβ receptor in human U937 promonocytic cells (Gargiulo et al., 2009). Despite these findings, both 9-ONC studies only examined TGFβ and presented no evidence of pro-inflammatory marker quantification. Another oxCE identified in oxLDL (Kobayashi et al., 2001) but which has received little attention is 7-Ketocholesteryl-9-carboxynonanoate. The structure of this lipid only differs from 9-ONC (Boechzelt et al., 1998) by two additional oxygen atoms, yet it induces a range of transcripts in J774A.1 MOs that are consistent with a pro-inflammatory role. These include IL-1R1, ICAM-1, CCR2, and NF-κB, although the PCR array used in this experiment also showed upregulated IL-4 and did not show a significant increase in the expression of several inflammatory cytokines and chemokines, including Il-1β (Huang et al., 2010).

Research in our laboratory has focused on the biology of a novel class of stable end products of CE oxidation named cholesteryl hemiesters (Estronca et al., 2012; Domingues et al., 2017). The only commercially available member of this family, cholesteryl hemisuccinate, induces pro-inflammatory cytokine (Il-1β, Il-6, Tnfα) release in BMDM, but does not affect Il-10 levels relative to native LDL controls (Domingues et al., 2017).

Like many of the lipid classes discussed above, exposure to a low dose of LPS was required for cytokine release. In zebrafish on a Chems-enriched diet, the number of myeloid cells that infiltrate the vasculature is significantly higher than in control or cholesterol-fed fish (Domingues et al., 2017). It can be argued, correctly, that Chems is not a major natural product of LDL oxidation. However, it may be expected that the major products would have the same effects as Chems (ongoing work).

# FUNDAMENTAL DIFFERENCES IN THE EXPERIMENTAL APPROACHES USED TO STUDY LIPID-MEDIATED INFLAMMATION CAN LEAD TO ALTERNATIVE OUTCOMES

The broad spectrum of lipid classes and products used to model atherogenic, proinflammatory immune responses has facilitated a greater understanding of some of the underlying molecular mechanisms. The field is now awash with data, yet there are emerging concepts that warrant an even deeper understanding in the hope that improved, and perhaps personalized, therapies can be developed to reduce the inflammatory burden, thereby limiting the pathogenesis of atherosclerosis. It is clear from the literature that multiple alternative approaches are often being used to answer the same or similar questions. A good example of this is the different reagents that are used to generate MOs, with either MO Colony-Stimulating Factor (M-CSF) or L-Cell Conditioned Media (LCCM; which contains M-CSF) or Granulocyte-MO Colony-Stimulating Factor (GM-CSF) added to human Mo or mouse bone marrow cultures. This is important because all three of these cytokine options produce MOs with markedly different transcriptomes, including differentially expressed inflammatory and lipid handling genes. Studies carried out a decade ago or earlier showed that human Mos differentiated with GM-CSF generate pro-inflammatory (M1-like) MOs, whilst those exposed to M-CSF develop an anti-inflammatory (M2-like) phenotype (Verreck et al., 2004; Xu et al., 2007; Waldo et al., 2008). A similar trend was also observed in mouse BMDM (Fleetwood et al., 2007, 2009). A comprehensive assessment of the transcriptomes of GM-CSF and M-CSF differentiated human blood Mos and mouse bone marrow was more recently undertaken using microarrays and qPCR (Lacey et al., 2012). Comparing the expression levels of TNFα, IL-1β, IL-12p35, IL-12p40, IL-23p19, IL-8, CCL2, and IL-10 between GM-CSF and M-CSF-differentiated human MOs revealed the only transcripts significantly different between the two populations were CCL2 and IL-10, both elevated in M-CSFderived MOs (Lacey et al., 2012). Significant differences were observed between the transcriptomes of GM-CSF and M-CSFderived mouse MOs. Amongst the top 100 most differentially expressed genes, the relative expression levels of several cytokine and chemokine receptors, some chemokines and notably Il-1b and Vcam1 were substantial. The latter two are particularly relevant in atherosclerosis. Strikingly, only 17% of the genes differentially regulated by human GM-CSF and M-CSF had a conserved pattern of transcriptional control in the mouse MOs, suggesting that observations in mouse BMDM may not necessarily translate to human cells. Interestingly, the same group showed in an earlier study that type I IFNs, known to be present in LCCM (Warren and Vogel, 1985), grossly affect the basal transcript levels of a number of genes, including some proinflammatory chemokines, expressed in MOs differentiated with M-CSF (Fleetwood et al., 2009), i.e., the CSF in LCCM.

As a master regulator of inflammation, IL-1β bioactivity fundamentally affects atherosclerosis as discussed already. A very recent study has revealed some stark, CSF-dependent differences in how and when IL-1β is produced by human MOs, which have implications for atherosclerosis research in vitro (Budai et al., 2017). In it, both GM-CSF and M-CSF-derived MOs release similar amounts of IL-1β, yet the release kinetics are dissimilar. M-CSF-differentiated MOs stimulated with LPS and either ATP or nigericin release a high concentration of IL-1β after just 2 h, which declines substantially after 6 h, returning to baseline levels after 12 h (Budai et al., 2017). By contrast, when GM-CSF was used, IL-1β release was more gradual, peaking between 6 and 24 h yet continuing to remain well above baseline at this latter time point. Perhaps most intriguing of all was the observation that only a priming signal, in this case LPS, was required for release in GM-CSF-polarized MOs. The authors demonstrated IL-1β release in the absence of "signal 2," and constitutive caspase-1 activity in these cells (Budai et al., 2017). This phenomenon has previously only been shown in Mos (Netea et al., 2009).

Another potentially critical factor in understanding the inflammatory response to lipids is the type of MOs that are used. It is apparent in the literature, and clear from the references cited in this article, that both pMOs and BMDMs are used to characterize lipids in vitro. Alternative MO phenotypes have distinct functional characteristics yet the differences that may exist between these populations are not usually acknowledged or examined. Based on this logic, a recent study found that when comparing pMO relative to BMDM, in either untreated MOs (basal expression) or oxLDL-transformed foam cells, substantial differences in cell surface marker, and inflammatory cytokine gene expression levels were evident (Bisgaard et al., 2016). Collectively these data expose major differences between the alternative MO populations that are used to study lipid-mediated inflammation. Most in vitro studies into lipid and lipoproteinmediated inflammation use one of five alternative ex-vivo MO types. Differentiation of blood or bone marrow-derived Mos produces three of these types in response to either L-cell conditioned media (containing M-CSF + TypeI IFNs) or M-CSF or GM-CSF. Each of these types possess a distinct transcriptional phenotype that has some overlap with the other two CSF-derived types (Fleetwood et al., 2007, 2009; Lacey et al., 2012). Small and large peritoneal MOs are also used for in vitro studies. They are transcriptionally distinct from one another and from their bone marrow-derived counterparts. When any of these five types of MOs are exposed to the same exogenous signal, it leads to their polarization, producing novel phenotypes that differ from one another as per the original differentiation signals.

In addition to primary cells, there are many MO cell lines used to study lipid-mediated inflammation, and some of them are perhaps unsuitable. Whilst cell lines offer a more homogenous population, most of them are already polarized toward a particular phenotype. This is important as these cells are almost certain to respond differently to a given lipid compared with a more naive population such as M0 primary MOs. Furthermore, beginning an experiment with polarized cells, e.g., with an M1 like phenotype, isn't very physiological as Mos recruited to the intima first differentiate into M0 MOs followed by polarization according to the signals received from the microenvironment. Some cell lines have also been shown to lack specific genes, have low expression levels of key molecules or have certain signaling pathways switched off. A good example here is the RAW MO cell line that lacks the intracellular inflammasome component ASC (Pelegrin et al., 2008) and is therefore unable to release Il-1β. This is significant because of the profound link between lysosome/mitochondrial dysfunction, metabolism, and inflammation (Guo et al., 2015; Van Den Bossche et al., 2017). These might seem like easy and obvious statements to make; however there continue to be reports describing a particular lipid inducing a particular effect, yet, the experimental conditions used can sometimes seem questionable.

Other experimental parameters are subject to variation and can be critical when studying inflammation. One example is the vehicle used to deliver a lipid. This is illustrated rather well with 7-ketocholesterol (7KC), which is found abundantly in oxLDL. Feeding the U937 Mo line either with 7KC or oxLDL induces a loss of cell viability. Both also cause glutathione loss and lead to increased 7KC content in these cells. By contrast, when acLDL is used to deliver this sterol to U937s, its uptake is high yet the toxicity to U937 cells is very low (Rutherford and Gieseg, 2012). Another example is the use of serum in cellular assays. In response to oxPAPC stimulation, one study showed that serum components were essential for the expression of inflammatory mediators IL-1β and COX-2, but not for HO-1 in BMDM (Kadl et al., 2011). Serum starvation is a commonly employed procedure in vitro, however, it has the potential to alter the outcome of an experiment (Pirkmajer and Chibalin, 2011). It may therefore be prudent when studying lipid-driven inflammation to always work with both serum-free and cultures with serum present if there is a need or desire to examine serum-starved cells.

The precise composition of lipid mixtures used to characterize inflammation also seems to vary greatly. For instance, with a complex mixture such as LDL, variable standards and methodologies of oxidation are employed. These include the use of 12/15-lipoxygenase (LO), copper, or osmium tetraoxide amongst others and have been reviewed extensively elsewhere (Levitan et al., 2010; Tsimikas and Miller, 2011). With no single or standard method of derivatization, some variation in the bioactivity of certain oxidized lipids have been reported. No two batches of oxLDL can be exactly alike, and it is often the case that post-oxidation analysis of a lipid mixture is not carried out. So in some studies, the exact composition of what is being added to the cells is not known. Given the array of immunogenic species that exist within oxLDL, with variable amounts of oxidized components likely between batches, this is certain to affect the intensity of pro- or anti-inflammatory responses in MOs or other cells. Consequently, there is a spectrum of reported inflammatory phenotypes for commonly studied lipid mixtures. This makes deciphering the impact upon other critical related parameters, such as immunometabolism, difficult. It is important to note there are many excellent examples of groups and studies (Harkewicz et al., 2008; Kadl et al., 2010; Oskolkova et al., 2010; Van Der Vorst et al., 2017; Zanoni et al., 2017) that employ stringent, robust methodologies. It would be greatly beneficial if some form of standardization or guidelines, in line with studies such as these, could be introduced.

Several of the studies reviewed in this article have highlighted the benefit of assessing the function of individual lipid components. The most striking of these perhaps being where oxPAPC and its minor fractions hyperactivate DCs to release IL-1β, however, only the minor fractions activate MOs (Zanoni et al., 2017). All of these reports make a strong case for refining the way in vitro experiments are carried out, by moving away from using complex mixtures and instead switching to individual lipid species or low complexity mixtures enriched with a single lipid component. In doing so, it is going to become easier to associate specific effects with specific lipid structures, which will improve the development of both generic and personalized therapies aimed at reducing or controlling atherosclerosis and other inflammatory diseases.

# THERAPEUTIC INTERVENTIONS TO TREAT ATHEROSCLEROSIS

By restoring or augmenting cellular function, or by inhibiting the factors that disturb it, numerous therapeutic interventions have been developed to treat atherosclerosis. Anti-inflammatory therapies are particularly ubiquitous and have a range of targets, with variable levels of efficacy. These include nutraceuticals, which are ingested as dietary supplements and have shown beneficial reductions in ROS activity, endothelial exocytosis, MO migration, and promote reverse cholesterol transport (Moss and Ramji, 2016). 3-Hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors, also termed statins are the most extensively used class of cholesterol lowering drugs that specifically target the mevalonate pathway. They were also recently shown to significantly and continuously suppress plaque inflammation by inhibiting MO proliferation (Tang et al., 2015). The use of cyclodextrin has also shown considerable benefit, through regression of plaque size. It increases the solubility of cholesterol, reducing the damaging effects of cholesterol crystals; and through targeting a gene set under the control of liver-X-receptors, promotes cholesterol efflux and anti-inflammatory effects (Zimmer et al., 2016). Epigenetic reprogramming can augment inflammation in trained immune cells and whilst atherosclerosis-specific therapies are not currently available to remodel methylated DNA, histone deacetylase inhibitors offer a potential novel option to reduce inflammation and are being experimented in some preliminary studies (Christ et al., 2016). A long standing and gradually improving treatment option is to directly block cytokines or the NLRP3 inflammasome, using either drugs or monoclonal antibodies (Toldo and Abbate, 2018). Controlling the biology of Interleukin-1 beta (IL-1β), which drives the progression of atherosclerosis (Libby et al., 2014; Libby and Hansson, 2015), leading to cardiovascular complications, has been of particular interest. A large-scale clinical trial (named CANTOS) that ended in early 2017 aimed to directly gauge the influence of IL-1β on atherothrombosis (Ridker et al., 2017). Enrolled patients received quarterly doses of canakinumab, a monoclonal antibody directed against IL-1β, for 3 years. Clinical outcomes included a significant decrease in non-fatal myocardial infarction, in non-fatal stroke and in cardiovascular death. These general results were underpinned by considerable reductions in selected inflammatory biomarkers C-reactive protein and IL-6, however, HDL and LDL levels were unaffected whilst triglycerides rose slightly (Ridker et al., 2017). The success of this trial reinforced the theory that IL-1β-mediated inflammation is a major driver of this disease. Blocking just a single molecule is unlikely to restrict all of the signaling pathways that lead to the production of additional inflammatory mediators. Other members of the IL-1 family including IL-1α (Freigang et al., 2013), IL-18 (Wang et al., 2015) as well as IL-6 (Schieffer et al., 2004) and TNFα (Brånén et al., 2004) are known to participate in atherogenesis and offer attractive future targets. PCSK9, a serine protease that binds to the extracellular domain of the LDL receptor targeting it for lysosomal degradation, also mediates inflammation in atherosclerotic plaques, including the synthesis of pro-inflammatory cytokines such as IL-1α, IL-6, and TNFα (Wicinski et al., 2017). A monoclonal antibody to PCSK9 (evolocumab) was recently used in a clinical trial (Sabatine et al., 2017), which led to substantial reductions in LDL cholesterol levels and the incidence of cardiovascular events. No data was presented on the impact upon inflammation, but it is highly likely to have been reduced, contributing to the improved clinical outcomes. Finally, a small but growing number of studies are exploring the benefits of restoring lysosome function through biogenesis. The activation of lysosome biogenesis by TFEB overexpression suppresses cholesterol crystal-induced NLRP3 inflammasome, attenuating the progression of atherosclerosis, in vitro and in vivo (Emanuel et al., 2014). More recently, TFEB overexpression was shown to reverse the gradual decline in the autophagy–lysosome system that occurs in lipid loaded cells with dysfunctional lysosomes, abolishing apoptosis and reducing IL-1β production, and consequently atherosclerosis (Sergin et al., 2017).

# SUMMARY

The purpose of this review was to offer a perspective on how cell dysfunction and inflammation contribute to atherosclerosis. We described how critical changes in organelle function, particularly those affecting lysosomes, can transform the inflammatory status of macrophages, the most prominent cell type in atherogenesis and which has a major influence on lesional inflammation, plaque progression and stability. By discussing lipid and non-lipid factors that can regulate inflammation, including the perhaps under appreciated role of circadian rhythms, we outlined some probable targets for more precise therapeutic interventions. Of course a major area of modern biomedical research is understanding the biology of lipid and lipoprotein factors that contribute to atherosclerosis and other chronic diseases, and some of the more prominent compounds studied were reviewed. In carrying out such studies, there are evident discrepancies in the literature, that in many cases are caused by alternative approaches to asking the same question. We made a point of highlighting where we believe these problems exist and hope that some form of standardization or guidelines could be introduced into the field to address them. Finally, we discuss recent advances in generic treatment options for atherosclerosis, which ideally could form one aspect of more personalized future novel therapies.

# AUTHOR CONTRIBUTIONS

All of the authors listed were involved in the preparation of the manuscript. ND carried out some of the experimental work referred to in the passages describing data produced in our group.

# REFERENCES


# FUNDING

This work was supported by—iNOVA4Health— UID/Multi/04462/2013, a program financially supported by FCT (Foundation for Science and Technology of the Portuguese Ministry of Science and Higher Education) through national funds and co-funded by FEDER under the PT2020 Partnership and PROGRAMAS DE ATIVIDADES CONJUNTAS, Reference: N◦ 03/SAICT/2015. ND holds a PhD fellowship from FCT, reference: SFRH/ BD/52293/2013.

# ACKNOWLEDGMENTS

The authors would like to thank Professor Winchil Vaz for advice on aspects of the manuscript content.


of macrophages via the transcriptional regulator ATF3. Nat. Immunol. 15, 152–160. doi: 10.1038/ni.2784


human atherosclerotic lesions. Arterioscler. Thromb. Vasc. Biol. 23, 1430–1436. doi: 10.1161/01.ATV.0000077207.49221.06


autoantibody-dependent uptake of oxidized low density lipoprotein by macrophages. J. Lipid Res. 42, 697–709.


inflammation downstream of type I interferon. Science 345, 679–684. doi: 10.1126/science.1254790


oxidized phospholipids. Arterioscler. Thromb. Vasc. Biol. 29, 356–362. doi: 10.1161/ATVBAHA.108.173799


**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 Gibson, Domingues and Vieira. 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.

# Macrophage Depletion Lowered Blood Pressure and Attenuated Hypertensive Renal Injury and Fibrosis

Lei Huang1,2, Aimei Wang<sup>2</sup> , Yun Hao<sup>2</sup> , Weihong Li<sup>2</sup> , Chang Liu<sup>3</sup> , Zhihang Yang<sup>1</sup> , Feng Zheng<sup>4</sup> and Ming-Sheng Zhou1,2 \*

<sup>1</sup> Department of Physiology, Shenyang Medical University, Shenyang, China, <sup>2</sup> Department of Physiology, Jinzhou Medical University, Jinzhou, China, <sup>3</sup> Department of Endocrinology, First Affiliated Hospital of Jinzhou Medical University, Jinzhou, China, <sup>4</sup> Department of Nephrology, Second Affiliated Hospital of Dalian Medical University, Liaoning, China

Monocyte/macrophage recruitment is closely associated with the degree of hypertensive renal injury. We investigated the direct role of macrophages using liposome-encapsulated clodronate (LEC) to deplete monocytes/macrophages in hypertensive renal injury. C57BL/6 mice were treated with a pressor dose of angiotensin (Ang, 1.4 mg/kg/day) II plus LEC or the PBS-liposome for 2 weeks. Ang II mice developed hypertension, albuminuria, glomerulosclerosis, and renal fibrosis. LEC treatment reduced systolic blood pressure (SBP), albuminuria, and protected against renal structural injury in Ang II mice. Ang II significantly increased renal macrophage infiltration (MOMA2<sup>+</sup> cells) and the expression of renal tumor necrosis factor α and interleukin β1, which were significantly reduced in Ang II/LEC mice. Ang II increased renal oxidative stress and the expression of profibrotic factors transforming growth factor (TGF) β1 and fibronectin. Ang II also inhibited the phosphorylation of endothelial nitric oxide synthase [phospho-endothelial nitric oxide synthesis (eNOS), ser1177]. LEC treatment reduced renal oxidative stress and TGFβ1 and fibronectin expressions, and increased phospho-eNOS expression in the Ang II mice. In Dahl rats of salt-sensitive hypertension, LEC treatment for 4 weeks significantly attenuated the elevation of SBP induced by high salt intake and protected against renal injury and fibrosis. Our results demonstrate that renal macrophages play a critical role in the development of hypertension and hypertensive renal injury and fibrosis; the underlying mechanisms may be involved in the reduction in macrophage-driven renal inflammation and restoration of the balance between renal oxidative stress and eNOS. Therefore, macrophages should be considered as a potential therapeutic target to reduce the adverse consequences of hypertensive renal diseases.

Keywords: macrophage, proinflammatory cytokines, hypertension, renal injury, angiotensin II

# INTRODUCTION

Hypertension is a major risk factor for nephrosclerosis and end-stage renal diseases (Appel et al., 2010). Despite extensive studies, the mechanisms by which hypertension causes renal injury are complicated and are not completely understood. Hypertensive nephropathy is characterized with renal inflammation, glomerular sclerosis, vascular hypertrophy, glomerular, and interstitial fibrosis

#### Edited by:

Geoffrey A. Head, Baker Heart and Diabetes Institute, Australia

#### Reviewed by:

Jacqueline Kathleen Phillips, Macquarie University, Australia Leonardo Roever, Federal University of Uberlandia, Brazil

> \*Correspondence: Ming-Sheng Zhou zhoums1963@163.com

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 25 January 2018 Accepted: 16 April 2018 Published: 07 May 2018

#### Citation:

Huang L, Wang A, Hao Y, Li W, Liu C, Yang Z, Zheng F and Zhou M -S (2018) Macrophage Depletion Lowered Blood Pressure and Attenuated Hypertensive Renal Injury and Fibrosis. Front. Physiol. 9:473. doi: 10.3389/fphys.2018.00473

**213**

(Yamashita et al., 2002; Liang et al., 2016). Recent studies suggest that immune cells, particularly for monocyte/macrophage lineage, play a critical role in the pathogenesis of hypertensive renal injury (Tian and Chen, 2015; Wenzel et al., 2016)

The mononuclear phagocyte system is the first line of immune cells in response to tissue injury, and plays an important role in tissue homeostasis and immune and nonimmune-mediated tissue injury and repair (Mosser and Edwards, 2008; Mantovani et al., 2013). Infiltrated and activated monocytes/macrophages release chemokines and cytokines which may cause renal inflammation, endothelial dysfunction, glomerular and tubule sclerosis, and fibrosis (Wang and Harris, 2011). Macrophage accumulation is correlated with the degree of renal dysfunction and/or severity of renal fibrosis in several animal models of kidney diseases including glomerulonephritis, ischemia-reperfusion renal injury, and diabetic nephropathy (Wang and Harris, 2011; Zhang M.Z. et al., 2012; You et al., 2013; Huen and Cantley, 2017). Increased infiltration of monocytes/macrophages into the kidney has also been reported in various animal models of hypertension. Immunosuppressive therapies reduce blood pressure and the number of infiltrated macrophages in the kidney and improve renal function in a variety of experimental hypertensive animals (Tian et al., 2007; McMaster et al., 2015). Several studies using chemical and genetic modification of macrophages have demonstrated that elimination of macrophages improves endothelial function and reduces vascular oxidative stress and the deleterious cardiac and vascular remodeling in the hypertensive animals (Wenzel et al., 2011; Thang et al., 2015; Kain et al., 2016; Wenzel et al., 2016). However, there are still no studies that have manipulated the level of macrophages to examine their direct role in hypertensive renal damage.

Inappropriate activation of renin–angiotensin (Ang) system is implicated in the pathogenesis of hypertension and hypertensive kidney damage (Zhou et al., 2003; Schulman et al., 2006). Ang II facilitates the infiltration of monocytes/macrophages into the kidney and induces renal inflammation and injury (Rudemiller et al., 2016). Ang II increases renal oxidative stress, glomerular sclerosis, and renal fibrosis (Xia et al., 2014). It has been reported that reduced renal macrophage recruitment by the blockade or knockout of the CC chemokine receptor can attenuate Ang II-induced renal injury (Elmarakby et al., 2007; Liao et al., 2008). Liposome-encapsulated clodronate (LEC) is widely used for in vivo depletion of macrophages from various organs and tissues (Kitamoto et al., 2009). After being engulfed by macrophages, LEC accumulates in macrophages and induces macrophage apoptosis (Jordan et al., 2003). Here, we used chemical depletion of mononuclear phagocyte lineage by administrating LEC into two hypertensive animal models of Ang II mice and Dahl salt-sensitive (DS) rats to examine the direct role of macrophages in hypertensive renal injury. Our results support the idea that renal macrophages are main contributors to hypertensive renal injury and fibrosis.

# MATERIALS AND METHODS

# Animal and Experimental Protocols

Six-week old male C57BL/6 mice or DS rats were purchased from Beijing Charles River Animal Laboratory (Beijing, China). All animal studies complied with the international standards stated in the Guide for the Care and Use of Laboratory Animals. All animal protocols were approved by the Institutional Animal Care and Use Committee of Jinzhou Medical University. The animals adapted to the new environment for 2 weeks before the experiments were performed.

# Animal Studies in Ang II Mice

The mice were randomly divided into four groups and treated for 2 weeks: (1) normotensive control (Ctr): sham surgery with implantation of an empty osmotic mini-pump plus an injection of PBS-liposome treatment (n = 8); (2) normotensive mice treated with LEC (LEC, Liposoma B.V., Amsterdam, Netherlands): implantation with an empty osmotic mini-pump plus LEC treatment (n = 8); (3) Ang II-infused mice (Ang II): implantation with an osmotic mini-pump of Ang II (1.4 mg/kg body weight/day, Sigma-Aldrich, St. Louis, MO, United States) plus PBS-liposome treatment (n = 8); (4) Ang II-infused mice treated with LEC (Ang II/LEC): implantation with an osmotic mini-pump of Ang II plus LEC treatment (n = 8). To implant the osmotic mini-pump (Alzet model 1002D, DURECT Inco., Cupertino, CA, United States), the mice were anesthetized using sodium pentobarbital (50 mg/kg I.P.). It has been reported that the mice receiving a sustained infusion of Ang II at this high dose can develop hypertension and renal injury (Liao et al., 2008; Zhang W. et al., 2012; Xiao et al., 2015). Macrophage depletions were performed by tail vein injections of LEC at the dose of 50 mg/kg body weight. LEC injection was done 1 day before mini-pump implantation and the injections were repeated every 3 days until the end of the experiments. The mice in the control group received a tail vein injection of PBS-liposome at similar interval and injection volume. At day 2 of LEC injection, blood was collected from tail vein for blood smears. The smear was stained with Giemsa (GS500; Sigma-Aldrich, St. Louis, MO, United States) and the amounts of monocytes, lymphocytes, and granulocytes were characterized according to their nuclear morphology, and counted by a blind observer. A total of 350 cells per smear were counted. Systolic blood pressure (SBP) was determined in the conscious mice using the tail cuff method (Softron Biotechnology Co., Ltd., Beijing, China) as described in our previous study (Zhou et al., 2015). Briefly, SBP was measured in a quiet and dark room. The mice were trained daily for 5 consecutive days prior to the implantation of the mini-pump. SBP was measured at three time points: at baseline (before Ang II administration or LEC injection), the fifth day, and the end of 2 weeks after the infusion of Ang II. At least five successive readings were recorded and averaged for each mouse. The day before the mice were euthanized, urine was collected by squeezing the animal bladder to stimulate urination, and the urine was collected on a metal plate. The ratio of urine albumin/creatinine was determined by albumin-to-creatinine ration assay kit

following the manufacturer's instructions (Shanghai Haoran Bio-Technology Co., Ltd., Shanghai, China). The mice were euthanized by overdose anesthetic agent (sodium pentobarbital 100 mg/kg I.P.).

# DS Rat Model of Salt-Sensitive Hypertension

DS rats were randomly divided into three groups and treated for 4 weeks as follows: (1) NS: the rats were fed a normal salt diet (0.5% NaCl, n = 8) plus PBS-liposome treatment; HS: the rats were fed a high salt (HS) diet (4% NaCl, n = 8) plus PBS-liposome treatment (n = 8); (3) HS/LEC: the rats were fed a HS diet plus LEC treatment. The macrophage deletions were performed by tail vein injections of LEC at the dose of 20 mg/kg body weight. LEC injections were started the day before the rats were given with HS treatment and repeated every 3 days until the end of the experiments. SBP was measured by tail cuff method as described above.

# Renal Histological Examination

The renal tissues were fixed in 4% paraformaldehyde in phosphate-buffered saline. The specimens were embedded in paraffin and cut into 4 µm thick section. Periodic acid-Schiff (PAS) staining was performed to evaluate the renal glomerular injury including glomerulosclerosis and mesangial matrix expansion (Zheng et al., 2004, 2006). The slides were photographed using an Olympus DS-41 microscope equipped with an Olympus DP-72 camera. A minimum of 20 images in one slide with at least one glomerulus per field were examined. A dark purple color in the glomeruli in each field was recognized as sclerosis. Percentage sclerotic area in glomeruli in each field was analyzed using the Image Probe Plus 6.0 image analysis system. The percentage area of sclerosis for 20 images in one kidney section was averaged as one single sample. Masson-Trichrome (Sigma-Aldrich, St. Louis, MO, United States) was carried out to evaluate renal fibrosis. Semi-quantitative analysis of the collagen contents in the renal tissues was assessed by evaluating percentage of positive stained areas with Image Pro Plus image analysis system. All tissue samples were evaluated independently by two reviewers who were not aware of the groups to which the animals belonged.

# Immunofluorescence Analysis

Renal **s**ections (4 µm) were cut from paraffin embedded tissues for immunofluorescence analysis. After deparaffinization and hydration, renal sections were microwaved for 30 min at 60◦C for antigen retrieval. The sections were incubated with primary antibody against monocyte/macrophage marker 2 (1:100 dilution with TBST buffer, MOMA2, Abcam Inco.) overnight at 4◦C, followed by incubation at 37◦C for fluorescein (FITC)-conjugated goat anti-rat lgG (1:200 dilution with TBST buffer, Protein Tech.) for 1 h. MOMA2 is a marker of monocyte/macrophage in mice (Yun et al., 2004; Daniel et al., 2017). The nuclei were stained by counter-staining with DAPI. The section in negative control was only incubated FITCconjugated lgG without primary antibody incubation. No fluorescence was detected in the negative control section. Monocytes/macrophages (MOMA2 positive cells) in renal tissues were viewed using a fluorescence microscope, and the MOMA2 positive cells were counted by two experienced reviewers who were blind to experimental groups. Twenty images in each section were examined and the number of positive cells per image was expressed by per mm<sup>2</sup> area of the image.

# Nicotinamide Adenine Dinucleotide Phosphate (NADPH) Oxidase Assay

Nicotinamide Adenine Dinucleotide Phosphate oxidase activity in renal homogenates was determined by lucigenin-enhanced chemiluminescence (ECL) in the presence of NADPH substrate as previously described (Zhou et al., 2005). In brief, 20 µl of renal homogenates was added into 50 mmol/l phosphate buffer (PH 7.4) containing 1 mmol/l EGTA as an assay solution, the reaction was triggered by adding NADPH (100 µmol/l) substrate and lucigenin (5 µmol/L). The data were expressed as counts/mini/mg protein.

#### Determination of In Situ O − <sup>2</sup> Production by Confocal Fluorescence Microscope

In situ O − 2 production in renal tissues was determined by a confocal fluorescence microscope using oxidative fluorescent dihydroethidine (Sigma-Aldrich, St. Louis, MO, United States) as previously described (Zhou et al., 2003). In brief, the fresh renal tissues were embedded in OCT compound and cut into 5-µm-thick sections. The slides were submerged in 2 µmol/l dihydroethidine in HEPES buffer and incubated at 37◦C for 30 min. The images were obtained with a Bio-Rad MRC-1024 laser scanning confocal microscope. A double-blind design was used to evaluate the image oxidative florescence intensity; the average fluorescent intensities were used for image quantification.

# Western Blot

Renal tissues were homogenized with lysis buffer containing 1 mmol/l PMSF, 10 µg/ml aprotinin, and 10 µg/ml leupeptin. After the homogenization, an aliquot of supernatant was taken for protein measurement with Bio-Rad protein assay. Thirty micrograms of protein was separated by SDS–PAGE and transferred to a nitrocellulose membrane. The membranes were incubated with blocking solution (5% milk added to TBST buffer) at room temperature for 1 h, then incubated with primary antibodies against phospho-endothelial nitric oxide synthesis (eNOS, Ser-1177, Cell signaling), NADPH oxidase subunits gp91phox and p22phox, tumor necrosis factor (TNF) α, interleukin (IL)1β, transforming growth factor (TGF) β1, and fibronectin (Santa Cruz Biotechnology Inc.) at 4◦C overnight (all these primary antibodies were diluted by blocking solution in 1:500) and washed three times with TBST containing Tween 20, then the membranes were incubated with horseradish peroxidase-conjugated secondary antibody (1:2000 dilution using blocking solution) for 1 h at room temperature. The signals were detected by ECL using hyperfilm and ECL reagent (Santa Cruz Biotechnology Inc.). The membranes were reblotted for β-actin (1:500 dilution, Santa Cruz Biotechnology Inc.), to serve as a loading control. The data were normalized to β-actin and expressed as fold change versus control group.

# Statistical Analysis

fphys-09-00473 May 3, 2018 Time: 17:36 # 4

The results were expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using SPSS 16.0 statistical software package (SPSS, Inc., Chicago, IL, United States), and statistical significance of difference was determined by two-way ANOVA with Bonferroni's correction for multiple comparisons. Values were considered significant when p < 0.05.

# RESULTS

# LCE Treatment Lowered Blood Pressure and Attenuated Ang II-Induced Albuminuria

There was no significant difference in SBP among all groups of mice at baseline. SBP was significantly increased at the fifth day of Ang II infusion and maintained the high blood pressure for 2 weeks (187 ± 4 vs. 110 ± 3 mmHg in control, p < 0.05). LEC treatment significantly reduced SBP (154 ± 2 vs. 187 ± 4 mmHg in Ang II group, p < 0.05) in the Ang II-infused mice but not in control mice (**Figure 1A**). In DS rats, HS intake for 4 weeks significantly increased SBP (183 ± 4 vs. 141 ± 3 mmHg in NS, p < 0.05), LEC treatment significantly lowered SBP in HS rats (159 ± 3 vs. 183 ± 4 mmHg in HS, p < 0.05, **Figure 1B**). Infusion of Ang II for 2 weeks resulted in a significant increase in the ratio of albuminuria/creatinine (83.6 ± 2.3 vs. 33.9 ± 4.9 µg/mg creatinine in control group, p < 0.05). LEC treatment significantly reduced albuminuria in the Ang II infused mice but not in control mice (**Figure 1C**).

# LEC Effectively Reduced Circulating Monocytes and Renal Macrophages

It has shown that LEC is an effective drug to deplete monocytes/macrophages via the induction of monocyte/macrophage apoptosis after it is engulfed by macrophages (van Rooijen and Hendrikx, 2010). In the present study, the mice were administrated with LEC at similar ways with the dose and the interval as reported (Jordan et al., 2003; Kitamoto et al., 2009; Moore et al., 2015; Kain et al., 2016). Consistent with these previous studies, the Giemsa-staining showed that LEC treatment reduced more than 70% circulating monocytes without significant changes in granulocyte and lymphocyte populations in the mice (**Figure 1D**). In DS hypertensive rats, LEC treatment also reduced the numbers of circulating monocytes by 67% without significant changes in other circulating cell populations (**Figure 1E**). Immunofluorescence showed that Ang II significantly increased the number of monocyte/macrophage marker MOMA2 positive cells in the renal tissues, LEC treatment significantly reduced the number of renal MOMA2 positive cells in both normotensive and Ang II-hypertensive mice (**Figure 2**), suggesting that LEC effectively and selectively targets monocyte/macrophage lineage.

# Depletion of Macrophages by LEC Protected Against Ang II-Induced Renal Injury and Fibrosis

To assess the effect of LEC treatment on Ang II-induced renal injury, renal tissues were stained with PAS and percentage area of glomerular sclerosis was calculated. As shown in **Figure 3**, compared with control mice, Ang II significantly increased glomerulosclerosis and glomerular matrix expansion, which was significantly attenuated by LEC treatment. Renal fibrosis is a hallmark of chronic hypertensive nephropathy. Inappropriate activation of RAS mainly by Ang II actions is considered as a major mechanism to contribute to renal fibrosis in hypertensive diseases (McMaster et al., 2015; Xiao et al., 2015). To determine the effect of macrophage depletion on renal fibrosis, Masson-Trichrome staining was performed. The quantification of collagen-stained area showed that Ang II infusion for 2 weeks significantly increased renal positive collagen-stained area and LEC treatment significantly attenuated the collagen-stained positive area (**Figure 4**). Although some investigators have reported that wild-type mice are to some extents resistance to Ang II-induced renal injury (Wesseling et al., 2005), an excess of Ang II can induce hypertension and profound renal damage (Zhang W. et al., 2012; Polichnowski et al., 2015; Xiao et al., 2015). The present study also supports that sustained infusion of Ang II at high dose can induce renal damage.

# LEC Treatment Attenuated Renal Injury and Fibrosis in Hypertensive DS Rats

To confirm the renal protective effects of LEC, we used PAS staining and Masson-Trichrome staining to assess glomerular injury and renal fibrosis in hypertensive DS rats, respectively. As shown in **Figure 5**, hypertensive DS exhibited significant increase in glomerular sclerotic area with glomerular matrix expansion, which was significantly attenuated by LEC treatment. Renal fibrosis as demonstrated by collage-staining positive area was also increased in hypertensive DS rats, which reduced in HS/LEC rats (**Figures 5C,D**).

# LEC Effectively Reduced the Expression of Proinflammatory Cytokines TNFα and IL1β

It has been shown that macrophage-derived proinflammatory cytokines such as TNFα and IL1β play a crucial role in the induction of renal inflammation and injury (Vesey et al., 2002; Zhang et al., 2014). As shown in **Figure 6**, the protein expression of renal TNFα and IL1β was significantly increased in the Ang II mice, which was significantly reduced in the Ang II/LEC mice.

# LEC Reduced the Expressions of TGFβ1 and Fibronectin

Transforming growth factor β1 and its downstream molecule fibronectin are important fibrotic factors which are considered as key mediators for renal fibrosis. As shown in **Figure 7**, Ang II infusion significantly increased the expressions of TGFβ1 and fibronectin, which were attenuated in the Ang II/LEC mice.

FIGURE 1 | Treatment with liposome-encapsulated clodronate (LEC) reduced systolic blood pressure (SBP) in the angiotensin (Ang) II mice (A) and DS rats (B) and attenuated the ratio of albuminuria/creatinine (C) in the Ang II mice. Quantification of monocyte, lymphocyte, and granulocyte population in the circulating peripheral blood smear stained with Giemsa in Ang II mice (D) and hypertensive DS rats (E). Ctr, control mice treated with PBS liposome; LEC, normal mice treated with LEC; Ang II, the mice treated with Ang II (1400 ng/min/mg body weight); Ang/LEC, the mice treated with Ang II plus LEC treatment; NS, DS rats fed a normal salt (0.5% NaCl) diet; HS, DS rats fed a HS (4% NaCl) diet. N = 8, <sup>∗</sup>p < 0.05, vs. control group or NS group, #p < 0.05, vs. Ang II group or HS group.

# Depletion of Macrophages by LEC Reduced Renal Oxidative Stress and Increased p-eNOS Expression in Ang II Hypertensive Mice

The macrophages are major producers of reactive oxygen species (ROS) in the infiltrated tissues (Moore and MacKenzie, 2009; McMaster et al., 2015). Ang II increases ROS production via stimulation of NADPH oxidase (Rajagopalan et al., 1996). We assessed renal oxidative stress using confocal fluorescence microscope, NADPH assay, and the protein expression of NADPH oxidase subunits gp91phox and p22phox. As shown in **Figure 8**, renal oxidative fluorescence densities stained by dihydroethidine were significantly increased in the Ang

II mice, and were significantly reduced in the Ang II/LEC mice. NADPH oxidase activity (**Figure 9A**) and the protein expressions of NADPH oxidase subunits gp91phox and p22phox (**Figures 9B,C**) were also significantly increased in the Ang II mice, and reduced in the Ang II mice treated with LEC. Phospho-eNOS (p-eNOS) at ser1177 is an active form of eNOS. The expression of p-eNOS and the ratio of p-eNOS/eNOS were significantly decreased in the Ang II hypertensive mice, which was also partly revised by LEC treatment (**Figure 9D**).

# DISCUSSION

Hypertensive renal diseases are associated with the accumulation of macrophages in the kidney (You et al., 2013; Xiao et al., 2015). However, the direct role of macrophages in hypertensive renal diseases is not established. In the present study, we have demonstrated that: (1) the infusion of Ang II in the mice increased monocyte/macrophage recruitment in the kidney, LEC effectively and selectively targeted the circulating monocytes and reduced renal macrophages and inflammation; (2) the depletion of macrophages by LEC significantly reduced blood pressure and renal morphological injury and fibrosis in two hypertensive models of Ang II mice and DS rats; and (3) the depletion of macrophages attenuated Ang II-induced renal oxidative stress and preserved eNOS phosphorylation. These results support a direct role for macrophages in the pathogenesis of hypertensive renal damage.

Circulating monocytes and tissue macrophages play complex roles in the pathogenesis of hypertension (McMaster et al., 2015). In the vasculature and kidney, macrophage-derived ROS and inflammatory cytokines induce endothelial and epithelial dysfunction, respectively, resulting in vascular oxidative stress and impairment of sodium excretion, which may cause endothelial dysfunction and renal dysfunction, and contribute to hypertension (Wenzel et al., 2016). Using the mice lacking the macrophage colony stimulating factor (m-CSF), who are deficient in monocytes and macrophages, De Ciuceis et al. (2005) showed that these animals exhibited minimal elevation of blood pressure in response to chronic Ang II infusion and had preserved endothelium-dependent vasodilatation of the resistance mesenteric vessels. Ang II-induced hypertension is

associated with increased inflammatory response and immune cell infiltration in the vasculature and the kidney (Rudemiller et al., 2016). Immunosuppressive therapy (Guan et al., 2013; Shah et al., 2015) or macrophage depletion (Moore et al., 2015) has been shown to rapidly inhibit Ang II-induced blood pressure elevation. The present study showed that the depletion of macrophages by LEC induced a rapid reversal of high blood pressure in Ang II mice. Our result demonstrates that the antihypertensive effects of LEC may be attributed in part to its inhibition of renal inflammation.

Macrophages have remarkable plasticity and heterogeneity (Mosser and Edwards, 2008). Macrophages mediate renal injury, mainly through classically activated (inflammatory) M1 macrophages (Mosser and Edwards, 2008; Mantovani et al., 2013). M1 macrophages are an important source of inflammatory cytokines in the infiltrated tissues (Wang et al., 2016; Wenzel et al., 2016). It has been shown that the mice lacking proinflammatory Th1 cells that produce interferon-γ (IFN-γ) and TNF-α were protected from hypertensive damage to the kidney glomerulus despite a preserved blood pressure response to Ang II (Zhang et al., 2014). TNFα is a pleiotropic cytokine which can control the production of other cytokines such as TGF1β in an autocrine and paracrine fashion (Awad et al., 2015). TNFα has been shown to play a critical role in diabetic and hypertensive

FIGURE 8 | The macrophage depletion by LEC reduced oxidative fluorescence densities in the Ang II mice. (A) Representative images of renal sections stained with DHE for the evaluation of oxidative fluorescence density evaluated by confocal fluorescence microscope. (B) Quantification of average fluorescence densities in each

group of mice. N = 8, <sup>∗</sup>p < 0.05, vs. control group, #p < 0.05, vs. Ang II group.

Oxidative stress and inflammation are integrals of hypertension-induced renal injury (Awad et al., 2015; Ratliff et al., 2016; Pushpakumar et al., 2017). Infiltrated macrophages are major sources of ROS in the infiltrated tissues. Ang II increases oxidative stress by stimulating NADPH oxidase in the renal cells and/or the infiltrated macrophage (Keidar et al., 1995; Xu et al., 2016), NADPH oxidase in the phagocytes expresses large amount of gp91phox subunits (Moore and MacKenzie, 2009). Here we showed that the depletion of macrophages by LEC reduced Ang II-induced renal oxidative stress, inflammation, and renal injury in the Ang II mice. Consistent with our findings, a recent study (Pushpakumar et al., 2017) showed that the Ang II mice with Toll-like receptor (TLR) 4 deficiency, an important signaling molecule for activation of innate immune system, exhibited less renal injury and oxidative stress, suggesting an important link between renal inflammation, oxidative stress, and renal injury. Furthermore, we showed that the depletion of macrophages restored eNOS phosphorylation in the Ang II mice. We and others have demonstrated that the balance between eNOS-derived NO and oxidative stress is critical for the maintenance of cardiovascular and renal homeostasis (Patzak and Persson, 2007; Zhou et al., 2008). These results suggest that renal macrophages may promote renal injury by the induction of renal inflammation and the imbalance between renal oxidative stress and NO system.

# Limitations

The present study has several limitations. First, depletion of macrophages resulted in a significant reduction in SBP in two hypertensive animals with over 30 mmHg of maximal reduction in SBP. It is well known that hemodynamics are an important factor driving hypertensive renal injury (Polichnowski et al., 2015). Ang II-induced renal injury has

BP-dependent and BP-independent components (Polichnowski and Cowley, 2009; Polichnowski et al., 2015) and in the present study, we did not control for BP changes in the LEC-treated mice. Therefore, the possibility that reduction in SBP per se may to some extents contribute to the amelioration of renal injury and dysfunction in hypertensive animals cannot be excluded. Next, we did not examine other immune cells or macrophage subset in the kidney; it has been reported that other immune cells such as T-lymphocytes may also play the role in the Ang II-induced hypertension and end organ damages (Barhoumi et al., 2011; Mian et al., 2016). However, because LEC selectively targets monocyte/macrophage lineage (Kain et al., 2016), the renal protection of LEC may mainly attribute to its depletion of macrophages other than the depletion of other immune cells. Third, we used lucigenin-ECL-based superoxide detection to measure NADPH oxidase activity. Although the method is frequently used by many other investigators, whether the method is specific for NADPH oxidase determination needs further clarification (Rezende et al., 2016). Finally, the present study was a prevention study but not an interventional study. Although the results from the present study support the notion that renal macrophages play an important role in hypertensive renal injury, treatment was initiated before the induction of disease and an interventional study may be required to provide evidence as to whether or not this approach can be used clinically to treat established disease.

# Future Directions

Although our studies provide evidence that depletion of monocytes/macrophages leads to obvious renal benefits, targeting all monocytes/macrophage would be of little therapeutic value because of the systemic immunosuppressive effects. Different macrophage subsets may play different role for renal diseases (Zhang M.Z. et al., 2012; Lech et al., 2014). Future study should identify the role of renal monocyte/macrophage subsets in hypertensive renal diseases, and refine techniques by using liposomes to target specific monocytes/macrophages for treatment of hypertensive or other renal diseases.

# CONCLUSION

The present study provides experimental evidence that renal macrophages are main mediators of renal injury and fibrosis and blood pressure elevation in two hypertensive animal models. Our data suggest that the underlying mechanisms may be macrophage production/release of cytokines and an imbalance between oxidative stress and NO in the kidney. While extrapolation between experimental animal studies and human hypertension is always speculative, since many of the animal models have been developed using the etiological factors which have been hypothesized to have a contributory role in human hypertension (Leong et al., 2015). Our studies

support the notion that targeting renal macrophages (or macrophage subsets) might have important clinical and therapeutic implications for the treatment of hypertensive renal diseases.

# AUTHOR CONTRIBUTIONS

LH contributed to the conception and design of the work; acquisition of data, analysis, and interpretation of data; statistical analysis. AW contributed to the acquisition of data, analysis, and interpretation of data and statistical analysis. YH, WL, CL,

# REFERENCES


ZY, and FZ contributed to the acquisition of data, analysis, and interpretation of data. M-SZ contributed to the conception and design of the work, analysis, and interpretation of data and drafted the manuscript.

# FUNDING

This work was supported by the grant from the National Natural Science Foundation of China (Nos. 81470532 and 81670384) and an award for distinguished professor in Liaoning Province to M-SZ.



in angiotensin II-induced hypertension. Circ. Res. 117, 547–557. doi: 10.1161/ CIRCRESAHA.115.306010


**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 Huang, Wang, Hao, Li, Liu, Yang, Zheng and Zhou. 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.

# Paracrine Anti-inflammatory Effects of Adipose Tissue-Derived Mesenchymal Stem Cells in Human Monocytes

Maria I. Guillén1,2, Julia Platas<sup>1</sup> , María D. Pérez del Caz<sup>3</sup> , Vicente Mirabet<sup>4</sup> and Maria J. Alcaraz<sup>1</sup> \*

1 Instituto Interuniversitario de Investigación de Reconocimiento Molecular y Desarrollo Tecnológico (IDM), Universitat Politècnica de València – Universitat de València, Valencia, Spain, <sup>2</sup> Department of Pharmacy, Faculty of Health Sciences, CEU Cardenal Herrera University, Valencia, Spain, <sup>3</sup> Department of Plastic Surgery and Burns, Hospital Universitario y Politécnico La Fe, Valencia, Spain, <sup>4</sup> Valencia Transfusion Center, Valencia, Spain

The inflammatory process is an essential phenomenon in the induction of immune responses. Monocytes are key effector cells during the inflammatory process. A wide range of evidence indicates that mesenchymal stem cells from adipose tissue (ASC) are endowed with immunomodulatory capacity. However, the interaction between ASC and monocytes in the innate immune response is not well understood. The aim of this work was to investigate the possible paracrine anti-inflammatory effects of ASC in human monocytes. Monocytes were isolated from buffy coats and ASC from fat of non-obese patients. Conditioned medium (CM) from ASC in primary culture was used. We have assessed the effects of CM on the production of inflammatory mediators, degranulation, migration, phagocytic activity, senescence, oxidative stress, mitochondrial membrane potential and macrophage polarization. We have shown that ASC exert paracrine antiinflammatory actions on human monocytes. CM significantly reduced the production of TNFα, NO and PGE<sup>2</sup> and the activation of NF-κB. In addition, we observed a significant reduction of degranulation, phagocytic activity and their migratory ability in the presence of the chemokine CCL2. The senescence process and the production of oxidative stress and mitochondrial dysfunction were inhibited by CM which also reduced the production of TNFα by M1 macrophages while enhanced TGFβ1 and IL-10 release by M2 macrophages. This study have demonstrated relevant interactions of ASC with human monocytes and macrophages which are key players of the innate immune response. Our results indicate that ASC secretome mediates the anti-inflammatory actions of these cells. This paracrine mechanism would limit the duration and amplitude of the inflammatory response.

Keywords: mesenchymal stem cells, inflammation, monocytes/macrophages, oxidative stress, inflammatory mediators

# INTRODUCTION

Numerous investigations have demonstrated the high potential of mesenchymal stem cells (MSC) for the development of therapeutic strategies in tissue repair and control of inflammatory conditions (Law and Chaudhuri, 2013). Many reports have also shown that modulation of inflammation may contribute to the beneficial effects of MSC which could depend on the

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

# Reviewed by:

Jose A. Garcia-Sanz, Consejo Superior de Investigaciones Científicas (CSIC), Spain Bernhard H. Rauch, University of Greifswald, Germany Slavko Mojsilovic, University of Belgrade, Serbia

> \*Correspondence: Maria J. Alcaraz maria.j.alcaraz@uv.es

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 15 December 2017 Accepted: 14 May 2018 Published: 31 May 2018

### Citation:

Guillén MI, Platas J, Pérez del Caz MD, Mirabet V and Alcaraz MJ (2018) Paracrine Anti-inflammatory Effects of Adipose Tissue-Derived Mesenchymal Stem Cells in Human Monocytes. Front. Physiol. 9:661. doi: 10.3389/fphys.2018.00661

**224**

production of soluble factors or cell-cell contact (Sheng et al., 2008; Prockop and Oh, 2012). It is known that MSC exert immunomodulatory effects on innate and adaptive immune systems (reviewed in Law and Chaudhuri, 2013; Tofiño-Vian et al., 2018). The immunomodulatory and anti-inflammatory properties of MSCs have supported studies on cellular therapy for inflammatory autoimmune diseases such as rheumatoid arthritis, systemic lupus erythematosus and systemic sclerosis. In particular, human adipose tissue-derived mesenchymal stem cells (ASC) have demonstrated an interesting immunomodulatory potential (reviewed in Mattar and Bieback, 2015). Recently, we have observed that the in vivo anti-inflammatory effects of ASC can be reproduced by the administration of their conditioned medium (CM) in the zymosan-injected air pouch model (Carceller et al., 2015). We have also shown the anti-inflammatory and anti-senescence effects of CM from human ASC in osteoarthritic chondrocytes (Platas et al., 2013, 2016). A better knowledge of ASC paracrine properties may help to develop novel approaches for the treatment of inflammatory conditions.

To gain further insight into the paracrine effects of ASC, we have focused this study on human monocytes and macrophages which play a central role in innate immunity. These cells produce a wide range of inflammatory mediators subjected to regulatory mechanisms. Monocyte activation helps cells to remain viable in inflammatory microenvironments but a resolution failure results in continuous inflammation. Therefore, an exaggerated or prolonged activation leads to self-amplifying stimulation of immune cells and damaging effects on different cell types which are involved in the pathogenesis of chronic inflammatory diseases such as rheumatoid arthritis, inflammatory bowel disease, neurodegenerative disorders, atherosclerosis, etc. (Greaves and Channon, 2002; Parihar et al., 2010). In the present work, we have extended our studies on ASC paracrine effects, by characterizing the regulation of relevant inflammatory responses and major functions of human monocytes and macrophages by CM from ASC.

# MATERIALS AND METHODS

# Isolation and Culture of Cells

The design of the work was approved by the Institutional Ethical Committees (University of Valencia and La Fe Polytechnic University Hospital, Valencia, Spain). Samples were obtained from donors after they provided informed written consent according to the Helsinki Declaration of 1975, as revised in 2013. Adipose tissue was obtained from healthy non-obese adults who had undergone abdominoplasty (11 women and 2 men, aged 54.1 ± 7.4 years, mean ± SEM). Samples were washed with phosphate-buffered saline (PBS), minced, digested at 37◦C for 1 h with 1% of type I collagenase (Gibco, Life Technologies, Madrid, Spain), and filtered through a 100 µm cell strainer (BD Biosciences Durham, NC, United States). Cells were then washed with DMEM/HAM F12 (Sigma-Aldrich, St. Louis, MO, United States) containing penicillin (500 U/ml) and streptomycin (500 U/ml), seeded onto tissue culture flasks (350,000/25 cm<sup>2</sup> ) in medium supplemented with 15% human serum from whole-blood donations of AB-bloodgroup-typed donors according to the criteria of Valencia Transfusion Center (Valencia, Spain), and incubated with 5% CO<sup>2</sup> at 37◦C. When the cells reached the semi-confluence, tissue culture plates were washed to remove any residual non-adherent cells. The phenotype of ASC was analyzed by flow cytometry (FACS-Canto II, BD Biosciences, San Jose, CA, United States) with specific antibodies, anti-CD105-PE, anti-CD90PerCP-eFluo710, anti-CD34APC (eBioscience, Inc., San Diego, CA, United States), and anti-CD45-PE (BD Pharmingen, BD Biosciences), and cellular viability with propidium iodide. More than 98% of viable cells were positive for CD105 and CD90, and negative for CD45 and CD34. CM was collected from cells at passages 0 and 1 at 48 h of culture, pooled, centrifuged, and stored at −80◦C in sterile conditions. We performed a cytokine profiling of CM using the RayBio <sup>R</sup> Human Cytokine Antibody Array C6 (RayBiotech, Norcross, GA, United States) according to manufacturer's instructions. Detection of chemiluminescence was performed by the AutochemiTM System with the Labworks 4.6 program (UVP Inc., Upland, CA, United States). Image J program (NIH, Bethesda, MD, United States) was used for analysis of results. Compared with control medium, the array revealed in CM elevated signals for interleukin (IL)-6 and IL-10, CXCL6, CCL7, CCL22, and CCL23.

Human monocytes were isolated from buffy coats provided by the blood bank Valencia Transfusion Center. Samples were mixed with DMEM/HAM F12 containing penicillin (500 U/ml) and streptomycin (500 U/ml) in a 1:1 ratio and centrifuged for 15 min at 400 × g and 18–20◦C. The pellet was resuspended in the above medium and this suspension was added to tubes containing Ficoll-Paque Premium 1.073 (GE Healthcare, Barcelona, Spain) in a slow stream to maintain the gradient. The tube was then centrifuged for 40 min at 400 × g and 18–20◦C. The top layer was aspirated and the mononuclear cell fraction was collected from the interface. Cells were washed with medium, viability was assessed by the Trypan blue method and then they were seeded at 10<sup>6</sup> /ml in medium supplemented with 10% human serum. After 2 h incubation, cells were washed with medium and adherent cells were characterized by flow cytometry using a FACS-Canto II cytometer (BD Biosciences), anti-CD45-PE (BD Pharmingen, BD Biosciences) and anti-CD14-PE (eBioscience, Inc.) antibodies and propidium iodide. More than 98% of viable cells were positive for CD45 and CD14. To perform the experiments, cells were incubated in medium supplemented with 10% human serum and stimulated with different agents and for different times, as indicated, in the presence or absence of CM (100% of medium, 0.4 ml for 24-well plates, 1 ml for 6-well plates).

# MTT Assay

The mitochondrial reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) to formazan as an indicator of cell viability was assayed in monocytes treated with CM or medium in the presence or absence of lipopolysaccharide (LPS, from Escherichia coli 0111:B4, Sigma-Aldrich), (1 µg/ml) and

incubated at 37◦C for 24 or 72 h. MTT (200 µg/ml) was then added and incubation proceeded for 2 h. Medium was removed and cells were solubilized in dimethyl sulfoxide (100 µl) to quantitate formazan at 550 nm using a Victor3 microplate reader (PerkinElmer Spain, Madrid, Spain).

# Determination of TNFα, NO and PGE<sup>2</sup> Production by Human Monocytes

Monocytes were incubated with CM in the presence or absence of LPS (1 µg/ml) at 37◦C for 24 h. Supernatants were used to measure tumor necrosis factor-α (TNFα) by an ELISA assay (Invitrogen, Thermo Fisher Scientific), with sensitivity of 4 pg/ml, nitric oxide (NO) production by fluorometric determination of nitrite levels (Misko et al., 1993) using a Victor3 microplate reader (PerkinElmer Spain), and prostaglandin E<sup>2</sup> (PGE2) by radioimmunoassay (Moroney et al., 1988).

# Myeloperoxidase

Monocytes were incubated with CM and/or 12-Otetradecanoylphorbol-13-acetate (TPA) (300 nM, Sigma-Aldrich) at 37◦C for 3 or 24 h. Supernatants were used to measure myeloperoxidase activity by using 3,3<sup>0</sup> ,5,5<sup>0</sup> -tetramethylbenzidine (Sigma-Aldrich) as substrate as previously described (De Young et al., 1989). Absorbance at 450 nm was quantified in a Victor3 microplate reader (PerkinElmer Spain).

# Cell Migration

Cell migration was assayed with 8-µm-pore size Transwell migration chambers (Thermo Fisher Scientific) (Bronkhorst et al., 2014). Monocytes (10<sup>6</sup> cells) in 1 ml of DMEM/Ham's F-12 with antibiotics and 10% human serum or in 1 ml of CM were added to the upper chamber. In the lower chamber, the chemokine CCL2 (100 ng/ml, Peprotech EC Ltd, London, United Kingdom) was added to the medium. Heparan sulfate proteoglycan could influence the effects of CM on cell migration (Shute, 2012). Therefore, we included experimental groups treated with CM previously incubated with anti-heparan sulfate proteoglycan antibody (Sigma-Aldrich, clone A7L6, 10 µg/ml) for 2 h. Cell migration was allowed to proceed for 72 h at 37◦C and 5% CO2. Then, inserts were separated and migrated cells were washed with PBS, observed in a microscope Leica DM IL LED (Leica Microsystems, Solms, Germany), and photographed with a Leica DFC450 C Digital Microscope Camera using Leica Application Suite software. Cells were quantified using the Cell counter complement of ImageJ (NIH, United States).

# Phagocytosis

Monocytes were cultured for 7 days in medium with 10% human serum (Musson, 1983) to assess the phagocytosis of fluorescent beads by flow cytometry and confocal microscopy. For flow cytometry, cells were seeded at 10<sup>6</sup> cells/well in 6-well plates, incubated for 24 h with CM or medium and then fluorescent beads were added at 2 concentrations (10<sup>7</sup> and 5 × 10<sup>7</sup> /ml) of FluoSpheres <sup>R</sup> (Molecular Probes Thermo Fisher Scientific) and incubations proceeded for 3 h. Cells were then washed with PBS, trypsinized and resuspended in PBS to measure the fluorescence (excitation 580 nm/emission 605 nm) in a FACS-Canto II (BD Biosciences) flow cytometer. For confocal microscopy, cells were seeded at 1.2 × 10<sup>5</sup> cells/well in 8-well Lab-tek microchambers (Thermo Fisher Scientific) and incubated with medium or CM and fluorescent beads as indicated above. Then, samples were washed with PBS, and incubated with anti-CD45-FITC antibody overnight at 4◦C. Slides were mounted in ProLong <sup>R</sup> Gold with DAPI (Molecular Probes; Invitrogen) and observed in a confocal microscope (Olympus FV1000). The percentage of phagocytosis was calculated using the number of cells with engulfed fluorescent beads and the total cell number.

# Senescence-Associated β-Galactosidase (SA-β-Gal) Assay

Monocytes were seeded at 20 × 10<sup>3</sup> cells/well in Lab-tek chambers (Thermo Fisher Scientific) and incubated with CM in the presence or absence of LPS (1 µg/ml) at 37◦C for 3 days. SA-β-Gal activity was measured using the cellular senescence staining kit (Cell Biolabs, San Diego, CA, United States). Cells were fixed with 0.25% glutaraldehyde in PBS for 5 min at room temperature and incubated with staining solution at 4 ◦C overnight. Slides were mounted in Prolong Gold antifade reagent with DAPI (Molecular Probes, Invitrogen, Thermo Fisher Scientific) and examined under a microscope (Leica DM IL LED). Slides were photographed with a Leica DFC450 Digital Microscope Camera using the Leica Application Suite software.

# Oxidative Stress

Monocytes were seeded into 6-well plates at a density of 10<sup>6</sup> cells/well in medium with 10% human serum and incubated until semi-confluence. The medium was then replaced by CM in treated wells or by the medium in controls. After 24 h incubation, cells were stimulated with 1 µg/ml of LPS (Sigma-Aldrich) for 30 min. After washing with medium without phenol red, a solution of dihydrorhodamine (5 µM, Sigma-Aldrich) in this medium was added and cells were incubated for 15 min at 37◦C. The supernatant was then discarded, cells were washed several times with PBS and resuspended in PBS to measure the fluorescence (excitation 485 nm/emission 534 nm) in a FACS-Canto II (BD Biosciences) flow cytometer.

# Mitochondrial Transmembrane Potential

The mitochondrial transmembrane potential (1ψm) was assessed with the JC-1 probe (5,5<sup>0</sup> ,6,6<sup>0</sup> -tetrachloro-1,1<sup>0</sup> ,3,3<sup>0</sup> tetraethyl-benzamidazolylcarbocyanine iodide, Thermo Fisher Scientific). This lipophilic membrane-permeant cation exhibit potential-dependent accumulation in mitochondria, indicated by a fluorescence emission shift from ∼525 nm (monomeric form) to∼590 nm (aggregated form). Monocytes were seeded into 6-well plates (10<sup>6</sup> cells/well) in medium with 10% human serum and incubated until semi-confluence. The medium was then replaced by CM in treated wells or by the medium in controls. After 24 h incubation, cells were stimulated with 1 µg/ml of LPS (Sigma-Aldrich) for 30 min. Cells were trypsinized, resuspended in 1 ml of PBS, and incubated with 10 µg/ml of JC-1 dye for 10 min at 37◦C and 5% CO2. Then, cells were washed and resuspended in PBS. Both red and green fluorescence emissions were analyzed by flow cytometry using an excitation wavelength of 488 nm and observation wavelengths of 530 nm for green fluorescence and 585 nm for red fluorescence, in a Becton Dickinson FACS-Canto II cytometer (BD Biosciences).

# NF-κB Activation

fphys-09-00661 May 30, 2018 Time: 8:52 # 4

Monocytes were incubated with CM in the presence or absence of LPS (1 µg/ml) at 37◦C for 20 h. Nuclear factor-κB (NF-κB) binding to DNA was quantified by ELISA in nuclear extracts using the Nuclear Extract Kit Active Motif for nuclei extraction followed by TransAM p65 NF-κB Activation Assay kits (Active Motif Europe, Rixensart, Belgium), according to the manufacturer's recommendations.

# Macrophage Polarization

Human monocytes were isolated from buffy coats and seeded as we have described before. After 2 h incubation, the adherent cells were washed with medium and incubated for 6 days in medium with 5% human serum and 25 ng/ml of human recombinant macrophage colony stimulating factor (Promokine, Heidelberg, Germany) to obtain non-polarized macrophages Mφ (Bertani et al., 2017). M1 polarization from Mφ was induced by supplementation of the medium with human recombinant interferon-γ (10 ng/ml, Promokine) and LPS (100 ng/ml, Sigma-Aldrich) for 48 h. M2 polarization from Mφ was obtained by supplementing cells with human recombinant IL-4 (20 ng/ml, Promokine) for 48 h. Mφ, M1, and M2 macrophages were incubated in the presence or absence of CM from ASC for 24 h and cytokines were measured in supernatants by ELISA. The assay for TNFα has been indicated above. The levels of transforming growth factor β1 (TGFβ1) and IL-10 were determined by ELISA assays from eBioscience (Labclinics, Barcelona, Spain), with sensitivities of 8 and 2 pg/ml, respectively.

# Statistical Analysis

The data were analyzed by one-way analysis of variance followed by Sidak's test using the GraphPad Prism 7.0 software (Graph Pad Software, La Jolla, CA, United States). A p-value of less than 0.05 was considered to be significant.

# RESULTS

# Production of Inflammatory Mediators

First, we assessed the possibility of a cytotoxic effect of CM or LPS in our experimental conditions. No decrease in cell viability by the MTT method was observed after incubation of human monocytes with CM in the presence or absence of LPS for the times used in our experiments (data not shown). We then investigated the influence of CM on the production of inflammatory mediators. Cells were stimulated with LPS in the presence or absence of CM, and culture medium was sampled at 24 h to measure the accumulation of inflammatory mediators. **Figure 1** shows that LPS induced the production of TNFα. Although CM did not modify basal levels of TNFα, it significantly decreased the production of this

cytokine induced by LPS (**Figure 1A**). Nitrite accumulation, an index of NO synthesis, was measured in the culture medium by a fluorometric method. As shown in **Figure 1B**, nitrite levels in nonstimulated cells were similar to those of CM. Nevertheless, we observed a 2.5-fold higher amount of nitrite in the medium after 24 h of monocyte stimulation

compared to control (B: nonstimulated cells); ∗∗P < 0.01 compared to LPS.

with LPS and the levels of this mediator were significantly reduced in the presence of CM. LPS strongly induced PGE<sup>2</sup> production (**Figure 1C**). Treatment of nonstimulated cells with CM enhanced the levels of PGE<sup>2</sup> whereas CM significantly reduced the accumulation of this eicosanoid in the presence of LPS stimulation (**Figure 1C**).

# Myeloperoxidase Release

fphys-09-00661 May 30, 2018 Time: 8:52 # 5

**Figure 2A** shows that LPS induced the release of myeloperoxidase into the culture medium by 3.5-fold after 3 h of stimulation. Treatment with CM significantly decreased this process by 48%. A similar stimulation was observed after 24 h of incubation with LPS (**Figure 2B**) but CM treatment resulted in a higher reduction of myeloperoxidase release (by 61%). These results demonstrate that CM exerts inhibitory effects on monocyte degranulation.

# Cell Migration

The migration of human monocytes was analyzed using transwell migration chambers and the monocyte-attracting chemokine CCL-2 as the stimulus. This chemokine plays a major role in regulating the movement of myeloid cells into inflammatory sites (Hayashida et al., 2001). CCL-2 significantly promoted migration when compared to nonstimulated monocytes (**Figure 3**). Incubation in the presence of CM resulted in a significantly lower number of recruited monocytes compared with CCL-2 controls. Therefore, migration was reduced to a level below that of nonstimulated cells. To exclude a possible influence of

FIGURE 2 | Myeloperoxidase release by human monocytes. Myeloperoxidase activity in cell culture supernatants was measured by a fluorometric procedure after 3 h (A) or 24 h (B) of TPA stimulation in the presence or absence of CM (mean ± SD from 4 separate experiments with cells from separate donors). ++P < 0.01 compared to control (B: nonstimulated cells); ∗∗P < 0.01 compared to TPA. RFU, relative fluorescence units.

control (B: nonstimulated cells); <sup>∗</sup>P < 0.05, ∗∗P < 0.01 compared to CCL2. (B) Representative images. Microscopic magnification of the objective lens 20 × . Bar = 100 µm.

heparan sulfate proteoglycan on the observed effects of CM, some incubations were performed after the neutralization of this proteoglycan with a specific antibody. As shown in **Figure 3**, this treatment did not modify the inhibitory effect of CM. These results indicate that CM is able to down-regulate the chemotactic response induced by CCL2 in human monocytes.

# Phagocytosis

Non-opsonic phagocytosis can trigger the release of inflammatory mediators and contribute to tissue injury (Ofek et al., 1995). As shown in **Figures 4A,B**, fluorescent beads at the concentrations used (F1: 10<sup>7</sup> /ml and F2: 5 × 10<sup>7</sup> /ml) were phagocytosed in a concentration-dependent manner by human monocytes. When cells were treated with CM we observed significant reductions in this process by 60 and 40%, respectively.

# SA-β-Gal Activity

Senescent cells exhibit increased cytoplasmic activity of SA-β-Gal (Dimri et al., 1995). To examine whether CM may affect senescence, we characterized this process by the presence of SA-β-Gal–positive cells. Incubation of monocytes with LPS for 3 days induced a significant increase in the percentage of SA-β-Gal-positive cells (85%) compared with 36% in

nonstimulated controls (**Figures 5A,B**). Treatment with CM induced a significant reduction (47%) of LPS effects on this marker of senescence.

# Oxidative Stress

Oxidative stress plays an important role in the induction of premature senescence and contributes to tissue injury in inflammation (Ben Porath and Weinberg, 2005). The change in ROS levels following exposure of human monocytes to LPS was assessed and the results are shown in **Figure 6A**. LPS stimulation of human monocytes resulted in a 4.5-fold increase in oxidative stress and this process was significantly inhibited (65%) by CM treatment.

# Mitochondrial Membrane Potential

Because we observed an inhibitory effect of CM on oxidative stress, we were interested to determine whether CM could modify

the changes in mitochondrial membrane potential induced by LPS. The probe JC-1 was used to measure changes in the mitochondrial membrane potential (19) of human monocytes. **Figure 6B** shows that LPS enhanced the number of cells with a low mitochondrial membrane potential whereas CM treatment counteracted the effects of LPS.

# NF-κB Activation

NF-κB is the main transcription factor involved in the synthesis of inflammatory mediators induced by LPS and cytokines. We have determined the influence of CM on the binding of p65 NFκB to DNA in the nucleus of human monocytes stimulated with

LPS. As shown in **Figure 7**, LPS significantly increased NF-κB binding to DNA whereas in monocytes treated with CM, we observed a significant reduction of this process.

# Macrophage Polarization

Non-differentiated (Mφ) as well as classically (M1) and alternatively (M2) polarized monocyte-derived macrophages were incubated in the presence or absence of CM. As shown in **Figure 8**, Mφ macrophages released into the medium TNFα accompanied by very low levels of TGFβ1 or IL-10. We observed that differentiation into M1 macrophages resulted in a significant enhancement of TNFα production (**Figure 8A**) whereas M2 differentiation led to significant increases in TGFβ1 (**Figure 8B**) and IL-10 (**Figure 8C**) levels. Treatment with CM inhibited TNFα release in M1 macrophages and enhanced TGFβ1 and IL-10 levels in non-differentiated and M1 macrophages. In M2 macrophages, CM significantly increased the release of TGFβ1 whereas the levels of IL-10 were not modified. As reported previously (Bertani et al., 2017), M1 macrophages presented a higher number of spindle shaped cells compared with M2 macrophages which showed a more spread morphology (**Figure 8D**). Treatment of M1 with CM modified cell morphology toward the Bφ phenotype.

# DISCUSSION

A wide range of factors present at the inflammatory microenvironment activate monocyte to migrate, phagocytose and generate ROS and pro-inflammatory mediators (Shi and Pamer, 2011) which exert autocrine and paracrine stimulatory effects leading to an amplification loop to perpetuate inflammation (Bardelli et al., 2012). The results presented in this paper show that CM from ASC down-regulates the activation of monocytes/macrophages induced by different types of stimuli and controls a number of relevant pro-inflammatory functions.

LPS recognition by Toll-like receptor-4 stimulates downstream signaling pathways including NF-κB and mitogen-activated protein kinases to induce the synthesis of a variety of pro-inflammatory molecules such as cytokines, PGE<sup>2</sup> and NO (Akira and Takeda, 2004). In this study we report that CM attenuates the release of crucial mediators of inflammatory responses. Therefore, CM reduced the production of TNFα, a cytokine with a central role in inflammation and tissue injury which has become an important target for the development of effective therapeutic agents in rheumatoid arthritis and other chronic inflammatory conditions (Beutler, 1999). Besides, CM reduced the production of PGE<sup>2</sup> stimulated by LPS. This eicosanoid exerts pro-inflammatory effects with vasodilation, oedema formation and synthesis of matrix metalloproteinases (Martel-Pelletier et al., 2003) although in some circumstances it also exhibits immunomodulatory properties (Harris et al., 2002). CM also inhibited NO production, an important mediator that may promote inflammation in mononuclear cells (Frieri, 1998). It is interesting to note that elevated serum levels of NO have been found in inflammatory arthritis patients with severe disease activity and they correlate with monocyte expression of inducible NO synthase (Pham et al., 2003). In addition, we have shown that CM decreases the DNA binding activity of NF-κB which could be an important mechanism for its anti-inflammatory effects on human monocytes. This is in line with our previous

reports showing that the inhibition of this transcription factor mediates the anti-inflammatory effects of human ASC CM on osteoarthritic chondrocytes (Platas et al., 2013) and also of mouse ASC CM in the zymosan-injected mouse air pouch (Carceller et al., 2015).

Recruitment of monocytes is necessary to control infections, but it also contributes to the pathogenesis of inflammatory diseases (reviewed in Shi and Pamer, 2011). It is known the important role of monocyte infiltration in inflamed tissues where they mediate tissue injury. We have shown that CM reduces monocyte migration induced by the potent chemoattractant CCL2. Enhanced levels of this chemokine have been demonstrated in inflammatory conditions, e.g., in synovial fluid of rheumatoid arthritis patients (Akahoshi et al., 1993). CCL-2 induces a quick calcium influx and cell activation and polarization which contribute to the establishment of inflammation in concert with additional signals (Mukaida et al., 1998). We have also investigated the possible effect of CM on another cellular function, the non-opsonic phagocytosis of inert particles. Our results indicate that CM is able to modulate the phagocytic properties of human monocytes/macrophages, which may contribute to the control of the inflammatory response as phagocytic stimuli can trigger or potentiate the production of inflammatory mediators (Corradin et al., 1991).

Monocytes are activated by a variety of stimuli to produce superoxide anion in a process primarily mediated by the NADPH oxidase complex (Cathcart, 2004). ROS act as signaling molecules that regulate cell growth, adhesion, differentiation, senescence, and apoptosis. On the other hand, there are synergistic interactions between ROS and inflammatory agents to mediate the inflammatory response with promotion of endothelial dysfunction, migration of leukocytes across the endothelium and tissue injury (Mittal et al., 2014). In addition, NO and superoxide interact to form the potent oxidant peroxynitrite able to interact with lipids, DNA, and proteins leading to cell damage.

Peroxynitrite may also be involved in NF-κB activation and cytokine release in human monocytes (Matata and Galinanes, 2002) and represents an important pathogenic mechanism in chronic inflammatory diseases (Pacher et al., 2007). Our data also indicate that CM reduces myeloperoxidase release by human monocytes. This observation may be relevant in relation with the control of oxidative stress as myeloperoxidase catalyzes the production of potent oxidants such as hypochlorous acid from hydrogen peroxide and chloride anion. These oxidants amplify the potency of ROS and have been implicated as mediators of oxidative tissue damage and cellular dysfunction in the development of many inflammatory conditions (Rayner et al., 2014).

ROS can induce and stabilize cell senescence, a process characterized by mitochondrial dysfunction and elevated ROS production which is related to chronic inflammatory diseases (reviewed in Correia-Melo et al., 2014). Cell senescence may contribute to the development of chronic inflammatory diseases (Mytych et al., 2017). In particular, accumulation of senescent monocytes has been associated to chronic inflammation in conditions such as atherosclerosis (Merino et al., 2011). Inflammatory and oxidative and nitrosative stress by low-dose LPS can induce premature senescence due to DNA damage, proliferation inhibition, or conversion of protective monocytes to cells showing a secretory and pro-inflammatory phenotype that may alter cell function and enhance adhesion to endothelial cells (Mytych et al., 2017). Our data indicate that CM from ASC protects human monocytes from the pro-senescence effects of LPS.

Macrophages play a key role in chronic inflammatory diseases and upon activation are a main source of TNFα in inflamed tissues. In addition, these cells release ROS, nitrogen species, PGs and matrix-degrading enzymes, and contribute to phagocytosis and antigen presentation (Haringman et al., 2005). In the presence of different pathophysiological conditions and microenvironments, macrophages can acquire distinct functional phenotypes. Classically (M1) and alternatively (M2) polarized macrophages possess pro-inflammatory and anti-inflammatory and reparative functions, respectively. Our results indicate that ASC exert paracrine actions on

# REFERENCES


differentiated monocyte-derived macrophages to potentiate antiinflammatory cytokines markers of M2 macrophages while simultaneously reducing the pro-inflammatory cytokine TNFα marker of M1 macrophages. Therefore, ASC may down-regulate the inflammatory response and favor the development of homeostasis and repair processes. These findings are in line with reports of M2 macrophage polarization by ASC, a property responsible for accelerated wound healing in animal models (Zheng et al., 2015).

The secretome of ASC has a complex composition including soluble factors and microparticles (Tofiño-Vian et al., 2018). Although some molecules such as IL-10 may contribute to the anti-inflammatory properties of ASC CM, recent data suggest a more relevant role for extracellular vesicles. In fact, we have demonstrated that microvesicles and exosomes from ASC CM are the main anti-inflammatory mediators in human osteoarthritic osteoblasts (Tofiño-Vian et al., 2017). We are performing further studies to know the complex mechanisms involved in the regulation of the inflammatory process by the ASC secretome. The results of our study suggest that ASC paracrine actions may determine the evolution of inflammation by modulating the functions of monocytes/macrophages which play an important role in the innate immune response.

# AUTHOR CONTRIBUTIONS

MIG and MJA participated in the design of research. JP, MDPdC, and VM performed the experiments. MIG, JP, and MJA performed the data analyses. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated.

# FUNDING

This work has been funded by grants SAF2017-85806-R (MINECO and FEDER), PROMETEOII/2014/071 (Generalitat Valenciana) and PRCEU-UCH20/11.


Arterioscler. Thromb. Vasc. Biol. 24, 23–28. doi: 10.1161/01.ATV.0000097769. 47306.12


**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 Guillén, Platas, Pérez del Caz, Mirabet and Alcaraz. 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.

# Role of Toll-Like Receptor 4 on Osteoblast Metabolism and Function

Ana Alonso-Pérez<sup>1</sup>† , Eloi Franco-Trepat<sup>1</sup>† , María Guillán-Fresco<sup>1</sup> , Alberto Jorge-Mora1,2 , Verónica López<sup>3</sup> , Jesús Pino2,3, Oreste Gualillo<sup>3</sup> and Rodolfo Gómez<sup>1</sup> \*

<sup>1</sup> Musculoskeletal Pathology Group, Laboratory 18, Institute IDIS, Servicio Galego de Saúde, Santiago de Compostela, Spain, <sup>2</sup> Division of Traumatology, Santiago University Clinical Hospital, Santiago de Compostela, Spain, <sup>3</sup> NEIRID LAB, Laboratory 9, Institute IDIS, Servicio Galego de Saúde, Santiago de Compostela, Spain

#### Edited by:

Geoffrey A. Head, Baker Heart and Diabetes Institute, Australia

#### Reviewed by:

Mikaël M. Martino, Monash University, Australia Samuel García Pérez, University Medical Center Utrecht, Netherlands Han Wang, Soochow University, China

#### \*Correspondence:

Rodolfo Gómez rodolfobahamonde@gmail.com; rodolfo.gomez.bahamonde@ sergas.es

†These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 15 January 2018 Accepted: 18 April 2018 Published: 08 May 2018

#### Citation:

Alonso-Pérez A, Franco-Trepat E, Guillán-Fresco M, Jorge-Mora A, López V, Pino J, Gualillo O and Gómez R (2018) Role of Toll-Like Receptor 4 on Osteoblast Metabolism and Function. Front. Physiol. 9:504. doi: 10.3389/fphys.2018.00504 Inflammation is a process whose main function is to fight against invading pathogens or foreign agents. Nonetheless, it is widely accepted that inflammation takes part in multiple processes in a physiological or pathophysiological context. Among these processes the inflammation has been closely related to bone metabolism. It is well-known that in systemic inflammatory diseases such as rheumatoid arthritis the inflammatory environment contributes to the reduction of the bone mineral density. This has been further evidenced in different animals models of osteoporosis where the deletion of key inflammatory molecules dramatically reduced the bone loss. On the contrary, it is also well-known that certain degree of inflammation is required to allow bone fractures healing. In fact, excessive use of anti-inflammatory drugs inhibits bone fracture consolidation. The innate immune responses (IIRs) contribute to the development and maintenance of the inflammation. These responses have been observed in cells of the musculoskeletal system. Chondrocytes and osteoblasts are equipped with the molecular repertoire necessary to setting up these IIR, including the expression of several toll-like receptors. Specifically, toll-like receptor 4 (TLR4) activation in mesenchymal stem cells, osteoblasts, and osteocytes has been involved in catabolic and anabolic process. Accordingly, in this review we have summarized the current knowledge about the physiology of TLR4, including its signaling, and its endogenous agonists. In addition we have focused on its role on osteoblast metabolism and function.

Keywords: osteoblast, TLR4, inflammation, LPS, osteoclast, MSCs, osteoclastogenesis, bone resorption

# INTRODUCTION

Bone is a mineralized connective tissue that exerts important functions. It provides rigidity to de body, protects soft tissues, and contributes to the locomotion. Moreover, it stores calcium and phosphate harboring also the bone marrow (Datta et al., 2008). Bone is a dynamic tissue continuously being formed and resorbed (bone remodeling). This process is required to maintain the structural integrity of the skeleton allowing the repair of damaged tissue as well as the homeostasis of calcium and phosphorous metabolism (Teitelbaum, 2000). The adequate balance between bone destruction and bone formation determinates the correct bone metabolism (Velasco and Riancho, 2008).

**234**

Bone tissue contains multiple types of cells including osteoblasts, osteoclasts, osteocytes, immune cells, adipocytes, stem cells, etc. (Florencio-Silva et al., 2015). In contrast to osteoclasts, responsible for bone resorption, osteoblasts are the only cells in charge of bone formation; they synthesize almost all of the constituents of the bone matrix and regulate its mineralization. Once the bone matrix is completely formed, osteoblasts differentiate into osteocytes, which play major roles in the regulation of calcium homeostasis and bone remodeling (Ralston and Helfrich, 2012). Osteoblasts and osteoclasts, along with osteocytes form the bone-remodeling unit (Rosen and Bouxsein, 2006).

Inflammation has been closely related to bone metabolism (Claes et al., 2012). It is well-known that in systemic inflammatory diseases such as rheumatoid arthritis, pancreatitis, and others the inflammatory environment contributes to the reduction of bone mineral density and, therefore, to the development of osteoporosis (Hardy and Cooper, 2009; Haas et al., 2015). It is widely accepted that an excessive amount of pro-inflammatory cytokines in these pathologies promotes osteoclastogenesis (Hardy and Cooper, 2009). The increased osteoclastogenesis in turn involves the imbalance between bone formation and bone resorption (Hardy and Cooper, 2009). Nonetheless, the link between an altered bone metabolism and inflammation is not limited to the systemic inflammatory diseases. In fact, this relationship has also been observed in certain metabolic diseases such as obesity (Cao, 2011) and type II diabetes mellitus (Alblowi et al., 2009). The connection between inflammation and bone metabolism has been further evidenced in different animal models of osteoporosis where the deletion of the receptor for key inflammatory cytokines, like interleukin-1 (IL1) and tumor necrosis factor (TNFα), dramatically reduced the bone loss (Vargas et al., 1996). On the contrary, it is also wellknown that certain degree of inflammation is required to allow bone fractures healing (Cottrell and O'Connor, 2010; Claes et al., 2012). In fact, excessive use of anti-inflammatory drugs inhibits bone fracture consolidation (Cottrell and O'Connor, 2010). As a result, a fine regulation of the inflammatory environment is required to preserve bone homeostasis and bone regenerative properties (Claes et al., 2012).

The innate immune responses (IIRs) contribute to the development and maintenance of inflammation. These responses are tightly regulated by the toll-like receptor (TLRs) family. Among these receptors toll-like receptor 4 (TLR4) stands out. This receptor is expressed in the musculoskeletal system where it plays a key role in the regulation of the inflammatory environment (Gómez et al., 2015). Accordingly, in this review we have summarized the current knowledge about the physiology of TLR4, including its signaling, and its endogenous agonists. In addition, we have focused on its role on osteoblast metabolism, viability, inflammatory responses, and function.

# TOLL LIKE RECEPTORS

Human body is constantly defending itself from highly changing pathogens, and other different harmful agents. The innate immune system has evolved in this environment selecting the adaptability as an essential feature. This condition has involved that, in order to improve its efficiency, the immune system recognizes diverse biological patterns conserved across multiple pathogens rather than specific molecules. These patterns are known as pathogens-associated molecular patterns (PAMPs) (Janeway, 2013).

The receptors that recognize these structures are called pattern-recognition receptors (PRRs). They take part in the first and non-specific response of the immune system. We can find them at three different locations, secreted at the extracellular space, in the cytoplasmic membrane, or as intracellular molecules. For example, lipopolysaccharide (LPS) binding protein, or C-reactive protein (CRP) are secreted PRRs, and nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs) are intracellular PRRs (Feerick and McKernan, 2017). Instead TLRs, depending on its type and state of activation, can be found in the cytoplasmic membrane or at intracellular endosomes membranes (Chen and DiPietro, 2017). Despite these different locations the PRRs maintain structural and functional similarities, and they are highly conserved receptors.

Toll-like receptors-mediated immune responses control inflammation-related catabolism including cell dedifferentiation, induction of matrix metalloproteinases (MMPs) or inhibition of the expression of certain structural proteins. These receptors recognize PAMPs but some of them are also sense by damage-associated molecular patterns (DAMPs). These are host-derived molecules generated by damaged tissues. Among the 10 different TLRs described in humans so far, TLR4 is the TLR that detects more DAMPS (**Figure 1**). This precise TLR is the main PRR for LPS, but it is also activated by DAMPS related with different musculoskeletal pathologies like rheumatoid arthritis (RA) or osteoarthritis (OA) (i.e., Chen et al., 2007; Abdollahi-Roodsaz et al., 2011; Goldring and Scanzello, 2012).

# TLR4

TLR4 main expression is in myeloid origin cells like monocytes, macrophages, granulocytes, and also in the spleen (Medzhitov et al., 1997; Vaure and Liu, 2014). However, these are not the only cells where it is expressed, intestinal epithelium cells, brain endothelium cells, and adipocytes are able to react against LPS through TLR4 signaling (Vaure and Liu, 2014). Interestingly, it is noteworthy that chondrocytes (Wang et al., 2011), osteoblasts (Kikuchi et al., 2001) and synoviocytes (Midwood et al., 2009) express TLR4 receptor as well. This, along with the fact that TLR4 detects DAMPs, frequently associated with musculoskeletal pathologies, have supported the implication of this TLR in the pathophysiology of the musculoskeletal system (Midwood et al., 2009; Wang et al., 2011). Specifically, TLR4 has been linked to diseases like rheumatoid arthritis, osteoarthritis, and osteoporosis, where bone metabolism is altered (Abdollahi-Roodsaz et al., 2007; Kim et al., 2009; Gómez et al., 2015). Modulation or inhibition of TLR4 has been suggested as a treatment for these diseases. For instance, the inhibition of TLR4

in rheumatoid arthritis animal models, characterized by local and systemic reduction in bone mineral density, suppressed the severity of the disease (Abdollahi-Roodsaz et al., 2007). Moreover, TLR4 targeting has also been recently proposed as a potential treatment for osteoarthritis (Gómez et al., 2015), which is associated with subchondral bone alterations including osteophytes formation (bone spurs). In addition, in animal models of osteoporosis the pharmacological inhibition of its progression was also associated with an inhibition of TLR4 signaling (Vijayan et al., 2014).

Human TLR4 was the first characterized among all the mammalian TLRs (Medzhitov et al., 1997). This type I transmembrane protein is formed by 839 aminoacids, and is encoded by the gene located on chromosome 9q32-q33 (Keshava Prasad et al., 2009). TLR4 spans the cytoplasmic membrane. In contrast, the nucleic-acid-binding TLRs (TLR9, TLR8, TLR7, and TLR3) are confined to the membrane of intracellular endosomes (Chaturvedi and Pierce, 2009; Gangloff, 2012). TLR4, like all the TLRs and IL1 receptor (IL1R), are equipped with the same Toll/IL-1 receptor (TIR) domain (Keshava Prasad et al., 2009). Besides this, TLR4 has an extracellular leucine-rich repeat domain (Keshava Prasad et al., 2009). TLR4 has specific sites that have been associated with its activation and cellular localization through post-translational glycosylation and phosphorylation. Some of these modifications are essential for the correct function of this receptor. For instance, its glycosylation at Asn526 and Asn575 are vital for the expression of TLR4 on the cell surface. Likewise, the response of human cell lines to the PAMP LPS is blocked by the absence of two or more N-glycosylation sites in the TLR4 ectodomain (da Silva Correia and Ulevitch, 2002). TLR4 modifications are not limited to its extracellular domain since they also occur in the intracellular TIR domain (Medvedev et al., 2007; Raijmakers et al., 2010). In fact, in human cells phosphorylation of TLR4 at Tyr674 and Tyr680 are crucial for the correct signal transduction of this receptor (Medvedev et al., 2007).

# TLR4 SIGNALING

In contrasts to other TLRs, TLR4 activates two different signaling pathways after its dimerization. The myeloid factor 88 (MyD88) dependent and independent pathways (**Figure 2**). The canonical pathway, or MyD88-dependent pathway, is shared with the IL1R and all the TLRs. It later activates the alternative pathway or MyD88 independent pathway.

MyD88-dependent signaling pathway is initiated in the extracellular space. The activation starts by the recruiting of several co-factors including TIR-domain-containing adapter protein (TIRAP, also known as MAL) and MyD88 (Motshwene et al., 2009; Lin et al., 2010). Once this pathway is activated, MyD88 polymerizes and interacts with the intracellular IL1 receptor-associated kinases (IRAKs) 1 and 2 (Motshwene et al., 2009; Lin et al., 2010). The auto-phosphorylation of the IRAKs and the activation of TNF-receptor-associated factor 6 (TRAF6) (Medzhitov, 2001) triggers the signaling of phosphatidylinositol-4,5-biphosphate 3 kinases (PI3Ks), mitogen-activated- protein kinases (MAPKs), and the key pro-inflammatory transcription factor nuclear factor kappa B (NFκB) (Medzhitov, 2001). NFκB activates and leads to the induction of pro-inflammatory cytokines and proinflammatory mediators through the increased expression of certain enzymes such as the inducible nitric oxide synthase (iNOS) and the cyclooxygenase-2 (COX2) (Goldring and Goldring, 2007). Afterward, the receptor is internalized and activates the second pathway (Tanimura et al., 2008; Gangloff, 2012).

MyD88-independent signaling pathway takes place after MyD88-dependent pathway activation due to TLR4 internalization (Gangloff, 2012). This pathway involves the adapters TIR-domain-containing adapter protein inducing interferon-beta (TRIF) and TRIF-related adapter molecule (TRAM) (Kagan et al., 2008). The recruitment of TRIF by TRAM switches on the signaling proteins TRAF6, receptor-interacting serine/threonine-protein kinase 1 (RIPK1) and TANK binding kinase 1 (TBK1) (Gangloff, 2012). The activation of this pathway activates different transcription factors including NFκB, interferon-regulatory factor 3 (IRF3), interferon-regulatory factor 7 (IRF7), and their associated gene expression signature, which are mainly characterized by type I interferon gene expression (Gangloff, 2012). This alternative pathway is TLR4 and TLR3 specific (Kagan et al., 2008; Tanimura et al., 2008; Zanoni et al., 2011).

# TLR4 AGONISTS

Apart from LPS, TLR4 binds to multiple and diverse PAMPs (Chen et al., 2007) and DAMPs (**Figure 2**). These DAMPS include tenascin C (Midwood et al., 2009), oxidized LDL (Stewart et al., 2010), monosodium urate crystals (Scott et al., 2006), serum amyloid A (de Seny et al., 2013), amyloid-β (Stewart et al., 2010), hyaluronic acid (Liu-Bryan and Terkeltaub, 2010), fibronectin EDA (Chen et al., 2007), heparan sulfate (Chen et al., 2007), fibrinogen (Chen et al., 2007), some free fatty acids as lauric acid (Lee et al., 2004; Frommer et al., 2013), α1 microglobulin (Sohn et al., 2012), α2 macroglobulin (Sohn et al., 2012), high mobility group protein B1 (HMGB1) (Liu-Bryan and Terkeltaub, 2010; Yang et al., 2010), heat shock protein 60 (HSP60) (Kim et al., 2009) (Pevsner-Fischer et al., 2007), and vitamin-D-binding protein (Sohn et al., 2012). There also are some drugs that activate TLR4 such as opioids (morphine and oxycodone), or buprenorphine (Hutchinson et al., 2010).

Though the mechanisms to explain TLR4 promiscuity are not known, some hypotheses have been posed (Gómez et al., 2015). For instance, heterodimers formed with other TLRs (Stewart et al., 2010), as well as TLR4 interaction with several coreceptors and accessory molecules could be an explanation of the wide range of molecules recognized by this receptor (Scott et al., 2006; Stewart et al., 2010; Tsukamoto et al., 2010). MD-2, CD14, CD36 are some of these co-receptors that increase responsiveness to antigens or the wiliness to form favorable heterodimers (Scott et al., 2006; Stewart et al., 2010; Tsukamoto et al., 2010). One example is the fact that the CD36 mediated the formation of the heterodimer TLR4-TLR6 that bound oxidized LDL (Stewart et al., 2010). However, TLR4 homodimer could not recognize oxidized LDL as an agonist (Stewart et al., 2010). Notwithstanding, not every accessory molecules and co-receptors are indispensable. Some of them only improve DAMPs and PAMPs recognition, or the velocity of the signaling (da Silva Correia, 2001; Tsukamoto et al., 2010).

Sometimes certain DAMPs preparations used for research can be contaminated by some low-level of endotoxin (Tsan and Gao, 2007). This contamination could induce the activation of TLR4. Nevertheless, there are a pool of studies carried out in musculoskeletal tissues that include endotoxin contamination determination, and evidence how TLR4 is involved in the recognition of many DAMPs (Okamura et al., 2001; Pevsner-Fischer et al., 2007; Schelbergen et al., 2012; Sohn et al., 2012; de Seny et al., 2013; Frommer et al., 2013).

# TLR4 EXPRESSION IN THE BONE COMPARTMENT

TLR4 is highly expressed in human bone marrow (Uhlén et al., 2015). This is in line with the fact that TLR4 is highly expressed by immune cells (Shi et al., 2006), but also with the expression of TLR4 in adipocytes (Shi et al., 2006) and osteoblasts that also are present in the bone marrow (Kikuchi et al., 2001; Gasper et al., 2002; Zou et al., 2003; Nemoto et al., 2008). Human and mouse mesenchymal stem cells (MSCs), which are osteoblasts and adipocytes precursor cells, also express the mRNA of TLR4 (Pevsner-Fischer et al., 2007; Mo et al., 2008; Tomic et al., 2011; Chen et al., 2014; Huang et al., 2014; Albiero et al., 2015). Specifically, in human MSCs TLR4 expression was found induced along their osteoblastic differentiation (Mo et al., 2008).

Mesenchymal stem cells have been proposed as a tool to help in bone regeneration. MSCs used to this purpose can be obtained from different sources like bone marrow, adipose tissue, dental pulp, dental follicle, periodontal ligament, Wharton's Jelly, or umbilical cord blood (UCB). According to this, it is noteworthy that the pattern of expression and activation of TLR4 is different across all these MSCs. In fact, TLR4 expression and activation in UCB-derived MSCs was found different from the one exhibited by MSCs from the bone marrow (van den Berk et al., 2009). Dissimilar expression of TLR4 was also observed in dental pulp MSCs in comparison to dental follicle MSCs (Tomic et al., 2011). This divergent expression correlated with different outcomes upon TLR4 activation like the diverse production

of the transforming growth factor beta (TGF-β) (Tomic et al., 2011).

Interestingly, long-term activation of TLR4 in human bone marrow MSCs was related to a reduction in its expression (Mo et al., 2008). On the other hand in human bone marrow MSCs TLR4 activation by the DAMP HSP60 up-regulated TLR4 expression (Kim et al., 2009). This was also observed in mouse MSCs where TLR4 activation also up-regulated its own expression (Huang et al., 2014). Moreover TLR4 activation in the mouse osteoblastic cell line MC3T3-E1 induced its own expression (Liu et al., 2016). These apparent divergent results regarding TLR4 expression upon its stimulation might be explained according to different incubation times, agonist concentration, as well as cell-specific TLR4 expression rates.

# TLR4 AND OSTEOBLAST-MEDIATED INFLAMMATION

Excessive inflammation has been associated to bone loss (Inada et al., 2006). Interestingly, despite the structural role often depicted for osteoblasts they are also capable of mounting an

inflammatory response (Gasper et al., 2002). TLR4 plays a key role in this activity (**Figure 3**). It was observed in human and mouse osteoblasts that TLR4 activation up-regulated the expression of C-X-C motif chemokine ligand 10 (CXCL10) (Gasper et al., 2002; Nakao et al., 2013), a known chemotactic factor (Gasper et al., 2002). TLR4 activation in human osteoblast cells also induced iNOS activity and nitric oxide production (Sosroseno et al., 2009), a known inflammatory mediator (Wu et al., 2007). This effect required the signaling through protein kinase A (PKA), PKC, phospholipase A2 (PLA2), lypoxygenase, tyrosine kinase activity, and the accumulation of cAMP (Sosroseno et al., 2009). TLR4 activation also increased in mouse osteoblasts the expression of COX2 and membranebound prostaglandin E synthase 1 (mPGES-1) as well as the production of TNFα (Zou et al., 2003), C-C motif chemokineligand 2 (CCL2) (Nakao et al., 2013), C-X-C motif chemokine ligand 1 (CXCL1) (Nakao et al., 2013), and prostaglandin E 2 (PGE2) (Inada et al., 2006). Regarding this, it was determined in the femur of mice defective for mPGES-1 that PGE2 production was required for the TLR4-mediated bone loss (Inada et al., 2006). In line with these TLR4-mediated pro-inflammatory activities, it was described in primary rat osteoblasts that TLR4 activation by the DAMP HMGB1 promoted the nuclear translocation of NFκB (Li et al., 2016), a key transcription factor involved in inflammation development (Clancy et al., 2004; Wen et al., 2006). Nonetheless, in the mouse osteoblastic cell line MC3T3-E1 this effect of HMGB1 was not attributed to TLR4 activation (Qiu et al., 2016). Despite this and underpinning the inflammatory role proposed for TLR4 in osteoblasts, it was observed in mouse osteoblasts that the inhibition of the formation of the TLR4-MyD88 complex significantly blunted the inflammatory responses elicited by this receptor (Nakao et al., 2013).

As observed in osteoblasts, in mouse MSCs TLR4 activation by PAMPs or DAMPs induced the secretion of IL6 (Pevsner-Fischer et al., 2007; Huang et al., 2014; He et al., 2015), IL1β (Huang et al., 2014; He et al., 2015), TNFα (Huang et al., 2014), as well as the nuclear translocation of NFκB (Pevsner-Fischer et al., 2007; Huang et al., 2014). Similar results were observed in human periodontal ligament MSCs where TLR4 activation increased the expression of IL6 and IL8 mRNAs (Albiero et al., 2015). Likewise, in human UCB MSCs TLR4 activation induced IL1β (Zhang et al., 2015), Interferon γ (INFγ) (Zhang et al., 2015), IL6 (van den Berk et al., 2009) and IL8 production (van den Berk et al., 2009).

# TLR4 EFFECT ON CELL VIABILITY AND PROLIFERATION

TLR4 activation, apart from its associated inflammatory effects, it has also been involved in the regulation of cell viability (Kim et al., 2009). Specifically, in human osteoblast-lineage cells it induced the caspase-dependent intrinsic apoptotic pathway, as well as the activation of p38 kinase and NFκB (Kim et al., 2009). Consistent with this TLR4 activation mediated the anticitrullinated protein antibodies (ACPA) induced apoptosis of human osteosarcoma cells (SaOs-2 cell line) (Lu et al., 2016). However other reports revealed that TLR4 activation enhanced mouse osteoblast precursor cells proliferation (MSCs) (Wang et al., 2009; Huang et al., 2014; He et al., 2015) and protected them from oxidative stress-induced apoptosis, through a PI3K/Akt dependent mechanism (Wang et al., 2009). In contrast, in mouse MSCs TLR4 activation did not affect their proliferation or apoptosis rate (Chen et al., 2014). In the same way, in primary rat osteoblasts as well as in primary fetal rat calvaria osteoblasts

TLR4 activation did not affect their proliferation rate or viability (Kadono et al., 1999; Li et al., 2016). Moreover, a recent report determined that TLR4 activation had no effect on the viability of mouse MSCs (He et al., 2015) as well as in MSCs derived from human periodontal ligament (Albiero et al., 2015).

All these studies depict an apparent contradictory scenario. However, the majority of reports performed with TLR4 agonists (DAMPs or PAMPs) on different osteoblastic lineage cells (MSCs and osteoblasts) revealed no alterations on cell viability (apoptosis or non-programmed cell death) (Kadono et al., 1999; Huang et al., 2014; Albiero et al., 2015; He et al., 2015; Li et al., 2016). These results were consistent across diverse species.

Nonetheless, the effect of TLR4 activation on cell proliferation is something more controversial. While some studies reported that TLR4 did not affect cell proliferation (Kadono et al., 1999; Huang et al., 2014; Albiero et al., 2015; Li et al., 2016), others revealed opposite results (Wang et al., 2009; He et al., 2015). These discrepancies might be related to the specific agonist used to activate the receptor (DAMPs or PAMPs). Also, some divergent outcomes in terms of proliferation upon TLR4 activation might be attributable to the dissimilar expression of TLR4 across the investigated cell types (Tomic et al., 2011). In addition, the different effect of TLR4 activation on cell proliferation might be due to the variation in TLR4 agonists concentrations (Wang et al., 2009). Supporting this idea Wang et al. (2009) described that lower LPS concentrations promoted MSCs proliferation, while higher LPS concentration exhibited the opposite effect.

# TLR4 EFFECT ON OSTEOBLAST MEDIATED OSTEOCLASTOGENESIS

The osteoclast is a multinucleated cell type formed by the fusion of monocytes/macrophages. Osteoclasts main function is the resorption of the bone, a process that mainly consists of the digestion of the mineral and organic matrix of the bone (Indo et al., 2013). The receptor activator of nuclear factor kappa-B ligand (RANKL), a cytokine of the TNF family, is a key factor in this process. RANKL in the presence of the macrophage-colony stimulating factor (M-CSF) boost the osteoclast generation (Indo et al., 2013; Charles and Aliprantis, 2014). Conversely, osteoprotegerin (OPG), a decoy receptor for RANKL, is a factor that blocks this process (Lacey et al., 2012; Charles and Aliprantis, 2014).

In vitro and in vivo experiments have extensively associated TLR4 agonism to the stimulation of the osteoclastogenesis (Sismey-Durrant and Hopps, 1987; Orcel et al., 1993; Hayashi et al., 2004). Accordingly, it was demonstrated in C3H/HeJ mice, which have a mutated TLR4, that activation of this receptor is associated to bone resorption (Nakamura et al., 2008). Moreover, TLR4 activation has also been related with several activities involved in osteoblast-mediated osteoclastogenesis (Shi et al., 2006). In fact, its activation in mouse osteoblasts induced the expression of RANKL mRNA (Kikuchi et al., 2001; Zou et al., 2003; Tang et al., 2011) and protein (Tang et al., 2011) without modifying the expression of OPG (Kikuchi et al., 2001; Zou et al., 2003). This induction was mediated by the extracellular signal-regulated kinase (ERK) (Kikuchi et al., 2001), the c-Jun N-terminal kinase (JNK) (Tang et al., 2011), and the PKC (Kikuchi et al., 2001). Moreover RANKL induction was independent of other inflammatory factors associated to TLR4 activation, such as TNFα and the PGE2 (Kikuchi et al., 2001). Together with RANKL up-regulation, TLR4 activation in mouse osteoblasts also induced the expression of M-CSF (Zou et al., 2003), which supported that TLR4 activation in osteoblasts may contribute to bone resorption (Kikuchi et al., 2001). In line with this, coculture of mouse primary osteoblast and hematopoietic cells in the presence of a TLR4 agonist stimulated the formation of osteoclasts (Yang et al., 2005). The effect of TLR4 activation on the osteoclastogenesis was attributed to osteoblasts TLR4 activation (Zhuang et al., 2007) because its activation in mouse bone marrow monocytes (BMMs) (osteoclast precursors) without co-culturing with osteoblasts failed to promote the formation of osteoclasts (Zhuang et al., 2007; Liu et al., 2009). Liu et al further observed that TLR4 activation inhibited osteoclastogenesis from mouse BMM but stimulated from those pre-treated with RANKL or co-cultured with osteoblasts (Liu et al., 2009). Interestingly these authors found that RANKL-mediated BMM commitment to osteoclasts was a prerequisite for TLR4-induced osteoclastogenesis (Liu et al., 2009). Conversely, the priming of mouse BMM by

FIGURE 4 | Osteoclastogenesis regulated by osteoblasts. TLR4 plays a key role in the cell fate of bone marrow monocytes (BMMs). During the first stage of commitment, TLR4 activation alone, promotes the conversion of these precursor cells into macrophages, and blocks osteoclastogenesis. However, osteoblasts can shift this fate to the formation of osteoclasts. TLR4 activation of osteoblasts induces the production of receptor activator of nuclear factor kappa-B ligand (RANKL) and macrophage-colony stimulating factors (M-CSF). The presence of these factors during the commitment phase prevents the macrophage fate and drives to osteoclast formation. Instead during the development stage, TLR4 activation triggers both cell fates. Thus, osteoclasts formation mediated by TLR4 depends on the presence or absence of osteoblast-derived RANKL during the commitment stage.

TLR4 activation blocked RANKL-mediated osteoclastogenesis (Liu et al., 2009) (**Figure 4**).

# TLR4 EFFECT ON OSTEOBLAST DIFFERENTIATION AND METABOLISM

Several reports have described anabolic properties of TLR4 activation on osteoblast (Mo et al., 2008; van den Berk et al., 2009; He et al., 2015; Ma et al., 2017). This activation in human primary osteoblasts up-regulated the expression of key osteoblastic markers (Ma et al., 2017). It was also observed that prolonged TLR4 activation of human MSCs up-regulated their osteoblastic differentiation without affecting their proliferation rate (Mo et al., 2008). However, this long-term activation of TLR4 was also related to a reduction in their expression of TLR4 (Mo et al., 2008). Nonetheless, other reports in human UCB MSCs (van den Berk et al., 2009), human periodontal ligament MSCs (Albiero et al., 2015), adipose tissue MSCs (Raicevic et al., 2012), and bone marrow MSCs (Raicevic et al., 2012) revealed that TLR4 activation enhanced their osteoblastic differentiation (van den Berk et al., 2009; Raicevic et al., 2012). Likewise, in mouse MSCs TLR4 activation also promoted their osteoblastic differentiation as well as its proliferation (He et al., 2015). In these cells TLR4 activation up-regulated wingless-INT (Wnt) family member 3A (Wnt3a) and Wnt family member 5A (Wnt5a), two major factors involved in the commitment of MSCs toward the osteoblastic cell fate (He et al., 2015). siRNA inhibition of these factors blocked the proliferative and osteoblastogenic activities of TLR4 activation suggesting that wingless-INT (Wnt) signaling was the driving force underlying to the TLR4 activation (He et al., 2015). According to this MSCs from MyD88−/<sup>−</sup> mice lacked the capacity to differentiate into osteoblasts cells (Pevsner-Fischer et al., 2007), which suggested that TLR4 associated MyD88-dependent signaling pathway was required to achieve the osteoblastic phenotype (Pevsner-Fischer et al., 2007).

In contrast to these data attributing an anabolic or proosteoblastogenic role to TLR4 activation, other authors observed that bone healing was accelerated in TLR4−/<sup>−</sup> mice after a skull lesion (Wang et al., 2017). However, a similar effect was also observed in a myeloid cell-specific TLR4 knockout mouse, which suggested that the anabolic activities observed in the TLR4−/<sup>−</sup> mice were not related to the absence of TLR4 in the osteoblasts (Wang et al., 2017). Martino et al. (2016) did not find the same effect in different TLR4−/<sup>−</sup> mice. Nonetheless, since IL1R and TLR4 share their MyD88-dependent signaling pathway, it is of interest that the same authors also found that MyD88 deficient mice, as well as IL1R−/<sup>−</sup> mice, exhibited a faster bone regeneration than their wild type littermates (Martino et al., 2016). Likewise, MyD88 deficient mice were resistant to PAMPinduced bone loss (Madeira et al., 2013), which was associated to less osteoclast formation as well as increased expression of osteoblastic markers (Madeira et al., 2013). The TLR4/MyD88 independent signaling pathway, which involves the signaling through TRIF, was not involved in the TLR4-mediated bone loss since TRIF−/<sup>−</sup> mice were not resistant PAMP-induced bone loss (Madeira et al., 2013). This was consistent with other reports about TRIF-independent inflammatory responses elicited by TLR4 activation in mouse osteoblasts (Sato et al., 2004). According to the TLR4-mediated inhibition of certain anabolic processes, TLR4 activation in differentiating mouse primary osteoblasts (Bandow et al., 2010), mouse osteoblastic cell line MC3T3-E1 (Liu et al., 2016), or in mouse MSCs (Chen et al., 2014) inhibited the matrix mineralization (Bandow et al., 2010), whereas this activation did not inhibit the matrix mineralization of mouse MyD88−/−-derived primary osteoblasts (Bandow et al., 2010). Moreover, unlike occurred in MyD88−/<sup>−</sup> osteoblasts, TLR4 activation of wild type osteoblasts up-regulated the mRNA of the activating transcription factor 4 (ATF4), as well as down-regulated osteoblastic transcription factors like the runt related transcription factor 2 (Runx2), and osterix (Sp7) (Bandow et al., 2010). Similar results were observed in primary rat calvaria osteoblasts (Kadono et al., 1999). In these cells TLR4 activation inhibited the expression of different osteoblastic markers, including alkaline phosphatase (ALP) osteocalcin, and osteopontin (Kadono et al., 1999). In the same way TLR4 activation in the osteoblast cell line MC3T3-E1 inhibited ALP activity and the mRNA expression of ALP, osteocalcin, and Runx2 (Liu et al., 2016). Interestingly, it was depicted that the inhibitory effect of TLR4 activation on mouse MSCs osteoblastic differentiation was mediated by the inhibitory crosstalk of the TLR4/MyD88/NFκB pathway over the anabolic BMP/Smad pathway (Huang et al., 2014).

Experiments performed in human primary osteoblasts (Muthukuru and Darveau, 2014) with diverse PAMP preparations that presented different effects on TLR4 function revealed that while strong TLR4 agonism involved inhibition of the osteoblastic markers, weak TLR4 agonism or TLR4 antagonism up-regulated them (Muthukuru and Darveau, 2014). Interestingly, it was suggested that the effect of TLR4 activation or inhibition on the osteoblastic markers was associated to the inflammatory response elicited (Muthukuru and Darveau, 2014). According to this, strong TLR4 agonists at very low concentrations also exhibit an anabolic effect on the osteoblastic markers (Muthukuru and Darveau, 2014).

# EFFECT OF TLR4 ON OTHER OSTEOBLAST ACTIVITIES

Osteoblasts can also contribute to bone resorption through the production of key degradative proteinases such as the matrix metalloproteinases 13 (MMP13) (Gao et al., 2013). In this sense it is noteworthy that TLR4 activation also up-regulated the expression of this catabolic factor (Gao et al., 2013).

Hyperlipidemic or hyperglycemic environments can alter osteoblasts and bone metabolism. TLR4 activation has been related to these alterations (Moriya et al., 2014; Rendina-Ruedy et al., 2016). Specifically, in rat osteoblasts TLR4 activation upon stimulation with palmitate, a hyperlipidemic environment, induced the secretion of the vascular endothelial growth factor 120 (VEGF120) (Moriya et al., 2014), which has been related with an abnormal bone metabolism. Moreover in a mouse animal model of high-fat diet-induced glucose intolerance

TLR4 defective signaling (C3H/HeJ mice) was associated to a later onset and reduced bone alterations (Rendina-Ruedy et al., 2016).

In contrast to this, in co-culture experiments performed with human osteoblasts and endothelial cells it was observed that TLR4 activation enhanced angiogenesis (Ma et al., 2017), which is a key component of bone repair (Hankenson et al., 2011).

Osteoblast migration is a key process in skeletal development as well as in bone regeneration (Li et al., 2016). TLR4 activation in mouse MSCs inhibited the migration ability of these cells (Chen et al., 2014). However, other authors observed in primary rat osteoblasts that TLR4 activation by the DAMP HMGB1 promoted their migration (Li et al., 2016). Nonetheless, in the mouse osteoblastic cell line MC3T3-E1 this effect of HMGB1 was not attributed to TLR4 activation (Qiu et al., 2016).

# FUTURE DIRECTIONS

Inflammation regulation is required to achieve a healthy bone metabolism. Excessive inflammation or inhibition of inflammatory responses have been linked to bone resorption and altered bone fracture healing. Therefore, future research should be focused on the characterization of the magnitude and duration of TLR4 agonism associated to these bone alterations. Likewise, as it was established for osteoclasts, it would be necessary to address how TLR4 activation affects osteoblasts differentiation at each stage of the process. Moreover, considering that excessive TLR4 signaling implies bone destruction further studies aimed to investigate novel therapeutic weapons should be carry on.

# CONCLUSION

Inflammation plays a key role in bone metabolism. Bone inflammatory responses are partially meditated by PRRs such as TLR4. This receptor that recognizes DAMPS and PAMPS has been related with the onset and development of different musculoskeletal pathologies where bone physiology is altered. According to this, TLR4 is expressed in cells that directly participate in bone metabolism; namely osteoblasts, osteoclasts, and MSCs. Nonetheless, its expression is uneven across these

# REFERENCES


cells. Despite this, TLR4 activation in osteoblasts and MSCs mediates a similar production of multiple cytokines, chemokines, and inflammatory mediators.

TLR4 signaling in osteoblasts and MSCs has also been involved in the regulation of cell viability and proliferation. However, these effects were not consistent across different studies. This might be explained by confounding factors. Some of these factors could also be responsible for the un-consistent effects of TLR4 activation on osteoblasts differentiation, where either anabolic or catabolic effects have been reported.

It is well-known that osteoblasts contribute to osteoclastogenesis. Interestingly, activation of TLR4 in osteoblasts promotes this process despite that its activation in osteoclast precursor cells inhibits their commitment.

Altogether, these data suggest that TLR4 might be a potential target to modulate bone metabolism. However, further studies are required to elucidate the precise role of this receptor on osteoblasts.

# AUTHOR CONTRIBUTIONS

AA-P, EF-T, and RG: study design and conception. AA-P, EF-T, MG-F, AJ-M, VL, JP, OG, and RG: declare that they collaborate in the search and selection of the papers, as in manuscript drafting. All of them have seen and approved the final version.

# FUNDING

This research was supported by research grants from Fondo de Investigación Sanitaria funded by the Instituto de Salud Carlos III (ISCIII) and FEDER (PI16/01870, CP15/00007, PI14/00016, and PI17/00409). RG is funded by ISCIII and SERGAS through a Miguel Servet programme. OG is staff personnel of SERGAS through a research-staff stabilization contract (ISCIII/SERGAS). RG and OG are members of the RETICS program, RD16/0012/0014 (RIER: Red de Investigación en Inflamación y Enfermedades Reumáticas) from ISCIII. OG is beneficiary of a project funded by REA of European Union, MSCA-RISE-H2020 program (Project number 734899).

loss of cartilage in diabetic fracture healing. Am. J. Pathol. 175, 1574–1585. doi: 10.2353/ajpath.2009.090148




stimulation on membrane-associated CD14-expressing cells. Int. Immunol. 22, 271–280. doi: 10.1093/intimm/dxq005


**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 Alonso-Pérez, Franco-Trepat, Guillán-Fresco, Jorge-Mora, López, Pino, Gualillo and Gómez. 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.

# Toll-Like Receptors, Inflammation, and Calcific Aortic Valve Disease

Carmen García-Rodríguez 1,2 \*, Iván Parra-Izquierdo<sup>1</sup> , Irene Castaños-Mollor <sup>1</sup> , Javier López 2,3, J. Alberto San Román2,3 and Mariano Sánchez Crespo<sup>1</sup>

1 Instituto de Biología y Genética Molecular, CSIC-Universidad de Valladolid, Valladolid, Spain, <sup>2</sup> Centro de Investigación Biomédica en Red de Enfermedades Cardiovasculares (CIBERCV), Madrid, Spain, <sup>3</sup> Hospital Clínico Universitario, Valladolid, Spain

Inflammation, the primary response of innate immunity, is essential to initiate the calcification process underlying calcific aortic valve disease (CAVD), the most prevalent valvulopathy in Western countries. The pathogenesis of CAVD is multifactorial and includes inflammation, hemodynamic factors, fibrosis, and active calcification. In the development of CAVD, both innate and adaptive immune responses are activated, and accumulating evidences show the central role of inflammation in the initiation and propagation phases of the disease, being the function of Toll-like receptors (TLR) particularly relevant. These receptors act as sentinels of the innate immune system by recognizing pattern molecules from both pathogens and host-derived molecules released after tissue damage. TLR mediate inflammation via NF-κB routes within and beyond the immune system, and play a crucial role in the control of infection and the maintenance of tissue homeostasis. This review outlines the current notions about the association between TLR signaling and the ensuing development of inflammation and fibrocalcific remodeling in the pathogenesis of CAVD. Recent data provide new insights into the inflammatory and osteogenic responses underlying the disease and further support the hypothesis that inflammation plays a mechanistic role in the initiation and progression of CAVD. These findings make TLR signaling a potential target for therapeutic intervention in CAVD.

Keywords: Toll-like receptor (TLR), inflammation, NF-κB, osteogenesis, aortic valve interstitial cell (VIC), calcific aortic valve disease (CAVD)

# INFLAMMATION AND CALCIFIC AORTIC VALVE STENOSIS

Calcific aortic stenosis is the final step of calcific aortic valve disease (CAVD), a slowly progressive complex process which begins with alterations in the valve leaflets that may lead to the development of left ventricular outflow obstruction (Miller et al., 2011; Rajamannan et al., 2011; Pawade et al., 2015). It has been related to atherosclerosis, inflammation, and hemodynamic factors, but active calcification is a feature characteristic (O'Brien, 2006; Yetkin and Waltenberger, 2009). Its prevalence is high and depends on age, affecting more than 2% of people older than 75 years and 8% of people older than 84 (Roberts and Ko, 2005; Nkomo et al., 2006; Go et al., 2014; Otto and Prendergast, 2014). Once symptoms appear the disease progresses rapidly, and most patients undergo either surgical aortic valve replacement or transcatheter aortic valve implantation. An estimated 42,000 and 65,000 valvular implants are performed per year in Europe and the USA, respectively (Bridgewater et al., 2011; Go et al., 2014; Otto and Prendergast, 2014), thus explaining the social impact and growing healthcare costs associated with CAVD.

#### Edited by:

Alexandrina Ferreira Mendes, University of Coimbra, Portugal

#### Reviewed by:

Joshua D. Hutcheson, Florida International University, United States Cynthia St. Hilaire, University of Pittsburgh, United States

#### \*Correspondence:

Carmen García-Rodríguez cgarcia@ibgm.uva.es

#### Specialty section:

This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology

Received: 08 January 2018 Accepted: 23 February 2018 Published: 12 March 2018

#### Citation:

García-Rodríguez C, Parra-Izquierdo I, Castaños-Mollor I, López J, San Román JA and Sánchez Crespo M (2018) Toll-Like Receptors, Inflammation, and Calcific Aortic Valve Disease. Front. Physiol. 9:201. doi: 10.3389/fphys.2018.00201

**246**

García-Rodríguez et al. TLR, Inflammation, and CAVD

CAVD is no longer considered a degenerative process related to aging, but an actively regulated process in which many pathogenetic factors remain unknown (O'Brien, 2006; Rajamannan et al., 2011; Towler, 2013). This change in the paradigm is based on three set of data: (1) epidemiological associations of risk factors and higher prevalence and faster progression of CAVD; (2) histopathologic identification in excised stenotic valves of chronic inflammation features, lipoprotein deposition, renin-angiotensin system components, and molecular mediators of calcification; and 3) elucidation of cell-signaling pathways and genetic factors linked to valve disease pathogenesis (O'Brien, 2006). From a clinical point of view, it would be very important to deepen in the knowledge of the disease and identifying potential novel therapeutic targets to delay its progression. Despite extensive research efforts including several randomized controlled trials (Cowell et al., 2005; Rossebø et al., 2008; Chan et al., 2010), no pharmacotherapy strategies currently exist to prevent or treat CAVD.

During the last decades, many studies have shown the relationship between classical atherogenic risk factors and valve calcification that might explain their frequent coexistence in clinical practice. Clinical factors associated with the development of CAVD found in the Cardiovascular Health Study are similar to the cardiovascular risk factors: older age, male gender, serum lipoprotein and LDL levels, height, hypertension, metabolic syndrome and smoking (Stewart et al., 1997). Even though CAVD and atherosclerosis share some ethiopathological features, recent reports have disclosed significant differences in calcification progression in the late stages and the lack of effect of statins in CAVD progression in the SALTIRE trial (Cowell et al., 2005; O'Brien, 2006; Hjortnaes et al., 2010; Dweck et al., 2013).

The inflammatory response is a hallmark of CAVD, which has been extensively demonstrated (O'Brien, 2006; Coté et al., 2013; Mathieu et al., 2015). In the 90s, some studies identified the presence in aortic valve lesions of macrophages and T lymphocytes and expression of chronic inflammation effector molecules like interleukins (IL)-1 and 2, class II human leukocyte antigen, and HLA-DR. More recently, mast cells, proinflammatory cytokines, IL-1β and tumor necrosis factor α have been linked to the disease (O'Brien, 2006; Hjortnaes et al., 2010; Dweck et al., 2013). Other mechanisms include the deposition of the atherogenic lipoproteins LDL and lipoprotein(a) in stenotic aortic valves and the activation of the renin-angiotensin system. Remarkably, the typical calcification during CAVD is considered an inflammation-dependent process (Coté et al., 2013; Dweck et al., 2013). New insights using novel imaging approaches that allow simultaneous visualization of inflammation and early mineralization support the paradigm of inflammation-dependent calcification and reveal that inflammation precedes calcification. This suggests three phases for CAVD progression (Otto, 2008; Aikawa and Otto, 2012; Pawade et al., 2015), inflammation being involved in the initiation and propagation phases, whereas in the last phase, calcification rather than inflammation is predominant.

# TLR, INFLAMMATION, AND DISEASE

The innate immune receptors termed Toll-like receptors (TLR) belong to a phylogenetically ancient system specialized in the recognition of conserved motifs present in pathogens, the socalled pathogen associated molecular patterns (PAMP), and of endogenous molecules released upon tissue injury (DAMP, also known as alarmins) (Kawai and Akira, 2010). The TLR family consists of membrane-spanned receptors recognizing a broad range of ligands, each member sensing a specific set of molecular patterns. So far, 10 TLRs have been identified in humans and 13 in mammalians (Kawai and Akira, 2010). TLR4, the first one described, is the receptor for lipopolysaccharide (LPS), a gramnegative bacterial toxin (**Figure 1**). TLR2 and its co-receptors TLR1/6 detect lipoproteins and lipopeptides, and TLR5 senses flagellin. TLR2/4 can also sense DAMPs, including heat-shock proteins, high-mobility group box 1 (HMGB1), reactive oxygen intermediates, and extracellular matrix breakdown products (Ionita et al., 2010). Finally, viral-derived molecules and host nucleic acids are sensed by nucleic acid-sensing TLRs localized in the endosomal compartment, i.e., TLR3, TLR7-9 (Takeuchi and Akira, 2009). Upon ligand binding, TLR dimerization activates common signaling pathways via the adaptor MyD88, except TLR3 that uses the adaptor TRIF, leading to the induction of either pro-inflammatory molecules through the activation of NFκB or antiviral molecules via interferon regulatory factor (IRF) routes (**Figure 1**).

TLR-mediated inflammation is necessary for homeostasis and control of tissue damage; however, improper activation leads to chronic inflammation, thus promoting deleterious effects. A significant amount of evidence, including human genetic studies, supports the involvement of TLR in a variety of diseases such as sepsis, asthma and autoimmune diseases (Drexler and Foxwell, 2010; Netea et al., 2012). In the cardiovascular system, recent studies support the emerging role of TLR in inflammatory diseases like atherosclerosis and ischemia/reperfusion injury (Mann, 2011), and more recently CAVD (Mathieu et al., 2015).

# TLR AND CAVD PATHOGENESIS

The finding of pathogen cargo in stenotic aortic valve cusps prompted researchers to investigate the putative role of TLR in CAVD. Although a straightforward association between infective agents and valve calcification is lacking, Chlamydia pneumoniae and bacteria associated with chronic periodontal infection have been detected in stenotic valves at the level of the valvular fibrosa (Juvonen et al., 1997; Nakano et al., 2006; Skowasch et al., 2009; Edvinsson et al., 2010), and inoculation of oral bacteria has been found to cause aortic valve calcification in a rabbit model of

**Abbreviations:** BGN, Biglycan; BMP-2, Bone morphogenetic protein-2; CAVD, calcific aortic valve disease; DAMP, danger/damage associated molecular patterns; HMGB1, high-mobility group box 1; IL, interleukin; IRF, interferon regulatory factor; LPS, Lipopolysaccharide; oxLDL, oxidized low-density lipoprotein; PAMP, pathogen associated molecular patterns; TLR, Toll-like receptor; VIC, aortic valve interstitial cell.

NF-κB signaling, and most likely IRF activation, thus promoting up-regulation of inflammatory and osteogenic mediators. Putative endosomal TLR4 signaling and TLR4-independent mechanisms are not depicted. ALP, alkaline phosphatase; ICAM-1, Intercellular Adhesion Molecule 1; PGN, peptidoglycan.

recurrent low-grade endocarditis (Cohen et al., 2004). However, the potential pathogenic role of transient bacteremia associated with mucous membrane trauma and/or the ensuing cytokine response has not been translated into clinical practice.

Innate receptors have been posited as the molecular hubs between pathogen-derived molecules, inflammation and CAVD (Mathieu et al., 2015). Particularly, the association between TLR and CAVD is supported by around 30 articles conducted in cellular and animal models. This section summarizes current evidences supporting the key role of TLR in driving inflammation and valve calcification in response to several stimuli. In the aortic valve, TLR are present not only in infiltrating immune cells but also in resident cells, where TLR4 is the most abundant subtype (Meng et al., 2008; Yang et al., 2009; López et al., 2012). To note, most evidences are from aortic valve interstitial cells (VIC), a model used to study aortic valve inflammation and calcification (Mohler et al., 1999; Osman et al., 2006; Rutkovskiy et al., 2017) that shows a myofibroblast-like phenotype when cultured on plastic plates, a reason why they are termed activated VIC or aVIC (Liu et al., 2007; Wang et al., 2014).

# TLR2/4 via MyD88/NF-κB Dependent Routes

The first members of the TLR family found to be expressed in aortic valve tissue and VIC were TLR4 and TLR2 (Meng et al., 2008), which are up-regulated in VIC explanted from calcified valves (Yang et al., 2009; López et al., 2012). In healthy human VICs, receptors are functional and mediate their effects via the MyD88 adaptor and the NF-κB route (Meng et al., 2008). Notably, the association of TLR with CAVD pathology has been established by using several pathogen-derived as well as endogenous molecules along the last decade (**Figure 1**).

## PAMPs and Nonsterile Inflammation

Several highly conserved motifs present in pathogens but not in the host have been associated with inflammation and osteogenesis in the aortic valve via TLR.

Meng et al., first reported that the archetypal TLR4 agonist LPS acts as a proinflammatory factor in healthy human VIC by triggering NF-κB signaling and the subsequent induction of cytokines and adhesion molecules (Babu et al., 2008; Meng et al., 2008). These findings were later confirmed by our group and expanded to prostanoids (López et al., 2012). Moreover, VICs from calcified human valves are more responsive to LPS (Yang et al., 2009; Fernández-Pisonero et al., 2014). Remarkably, TLR4/2 ligands are also pro-calcific factors, as judged from several evidences emerged from bovine, porcine, and human VIC, as well as from mice models. Microarray and proteomic analysis revealed that LPS promotes a pro-calcific phenotype in human and clonal bovine VICs (Babu et al., 2008; Meng et al., 2008; Bertacco et al., 2010), where it functions as a procalcific factor by activating osteogenic mediators such as bone morphogenetic protein-2 (BMP-2), Runt-related transcription factor 2, and alkaline phosphatase. Moreover, LPS promoted in vitro calcification in studies using either β-glycerophosphate (Yang et al., 2009; López et al., 2012) or elevated phosphate levels (Rattazzi et al., 2008) as a phosphate source. In keeping with this, TLR2 ligands such as peptidoglycan, characteristic of the outer membrane of gram-positive bacteria, and synthetic peptides like Pam3CSK4 promote a pro-calcific phenotype in VIC (Yang et al., 2009; López et al., 2012). Even more, recent data highlight the in vivo role of TLR4/2 in osteogenesis since LPS promotes early aortic valve leaflet thickness increase in mice, and TLR4/2 deficiency abrogates high fat diet-induced aortic valve lesions. (Zeng et al., 2017). Interestingly, the response to TLR4/2 ligands differs according to cardiac valve site, being their effects stronger in aortic VICs as compared to pulmonary, mitral or tricuspid VICs (Yang et al., 2009; López et al., 2012; Venardos et al., 2014). Finally, a recent report hypothesizes that TLR-mediated effects are prevented in infants by a protective mechanism involving STAT3 activation that is absent in adults (Deng et al., 2015).

Together, evidences disclose a cardiac valve site-specific association between TLR4/2-NF-κB, inflammation and osteogenesis, and support an inflammation-driven calcification model in CAVD. Further studies on quiescent VIC cultured within 3D hydrogels (Hjortnaes et al., 2016) and on mice models to test the in vivo relevance are warranted. To note, additional LPS endosomal recognition via TRIF and TLR4-independent mechanisms cannot be ruled out (Tan and Kagan, 2014).

### DAMPs and Sterile Inflammation

Endogenous molecules released upon tissue injury induce and perpetuate the so-called sterile inflammation (Ionita et al., 2010). Here, we describe several DAMPs recently linked to CAVD pathogenesis that mediate their effects via TLR.

Biglycan (BGN) is a small proteoglycan widely distributed within tissues that appears to be dysregulated in pathological conditions (Schaefer and Iozzo, 2008), including its overexpression in valves from CAVD patients (Derbali et al., 2010). Moreover, soluble BGN acts as a pro-inflammatory inducer through TLR pathways in human VICs by inducing lipid-modifying enzymes and cytokine expression via TLR2/4 (Derbali et al., 2010; Song et al., 2014). Song and colleagues further reported its pro-inflammatory and pro-osteogenic activities via a mechanism involving TLR2 and MAPK/ERK signaling (Song et al., 2012, 2014). A follow-up study demonstrated BGN-mediated pro-osteogenic reprogramming in VIC and identified BMP-2 and transforming growth factor-β1 as the molecular mediators (Song et al., 2015).

HMGB1 is a regulatory nuclear protein that when secreted extracellularly acts as a pro-inflammatory cytokine (Yang et al., 2005). Recent studies in patients and animal models have associated this protein with CAVD, since tissue and plasma levels of HMGB1 are increased in patients with CAVD (Wang et al., 2016a), and can be detected in the secretory granules of endothelial and interstitial cells explanted from diseased valves (Passmore et al., 2015). Additional evidences with recombinant HMGB1 show its pro-osteogenic activity by increasing osteogenic markers and calcium deposition in human VIC (Wang et al., 2016b). Moreover, the role of TLR4 and its transducers NF-κB and JNK, is supported by in vitro and in vivo studies, where HMGB1 pro-osteogenic activity was markedly decreased by gene silencing and TLR4-deficiency (Wang et al., 2016b; Shen et al., 2017).

Matrilin-2, an extracellular protein expressed in different tissues (Deák et al., 1999), accumulates in the calcific nodules of human aortic valves (Li et al., 2017). Moreover, matrilin-2 enhances osteogenic activity via TLR2/4 in VICs, as demonstrated by using both gene silencing and neutralizing antibodies, being the effects regulated by the NF-κB family of transcription factors and NFATc1 (Li et al., 2017).

Altogether, several DAMPs promote sterile inflammation and osteogenesis via TLR4/2-NF-κB routes, which warrants further investigation to test the in vivo relevance. Additional DAMPs associated to CAVD pathogenesis, i.e., galectin-3 and heat-shock protein (Skowasch et al., 2009; Sádaba et al., 2016) might mediate their effects through TLR signaling. It is plausible to think that the increased levels of DAMPs reported in calcified valves may act as a pro-calcific loop, thus contributing to disease progression.

# TLR3 and the TRIF/IRF Dependent Routes

The TLR3 ligand dsRNA is present in virus and can also be produced under replication of positive-strand RNA viruses, dsRNA viruses, and DNA virus, i.e., poliovirus, coxsackievirus, and encephalomyocarditis virus, in the host. An endogenous source may be tissue damage or necrosis (Gantier and Williams, 2007). The first association of viral-derived molecules with inflammation and calcification of human VIC through TLR3 was reported using polyinosinic acid: polycytidylic acid to mimic dsRNA effects (López et al., 2012). Zhan et al., later confirmed dsRNA-mediated up-regulation of inflammatory mediators and pro-osteogenic activity, and further demonstrated by genesilencing and neutralizing antibodies the involvement of TLR3- TRIF non canonical NF-κB signaling pathways as well as the ERK route (Zhan et al., 2015, 2017). Moreover, recent in vivo data presented at European Society of Cardiology Congress 2017 have associated TLR3 to the onset of calcific aortic valve disease by using TLR3- and ApoE-deficient mice models (Tepekoylu et al., 2017).

In summary, these findings posit the association between TLR and osteogenesis with no reported differences between sterile and non-sterile inflammation, and support an inflammationdriven calcification model in CAVD. Future studies are needed to analyze putative mechanistic differences associated to different TLR pathways and their in vivo relevance.

# MODULATION OF TLR SIGNALING IN THE AORTIC VALVE: CROSSTALK WITH MEMBRANE RECEPTORS

TLR activation relevant to CAVD can be modulated by several inflammatory mediators at various levels, but remarkably all converge on the NF-κB route. This section includes TLR modulators reported to date in the context of aortic valve physiopathology (**Figure 1**).

# Lipoproteins and Lipid Components

Among CAVD risk factors are high levels of circulating oxidized LDL (ox-LDL), which also accumulate in diseased aortic valves (Côté et al., 2008; Yeang et al., 2016). oxLDL treatment of human VIC modulates LPS effects by synergistically increasing pro-osteogenic genes (Zeng et al., 2014). Pharmacological and gene silencing approaches revealed a mechanism involving the modulation of the Notch-NF-κB axis (Zeng et al., 2014). Remarkably, blocking TLR2/4 with neutralizing antibodies abolished oxLDL-induced activation of NF-κB and ERK1/2 (Zeng et al., 2017). Sphingosine 1-phosphate, a lipoprotein lipid component acting via G-protein coupled receptors, modulates TLR activation in human VICs, macrophages and aortic endothelial cells (Dueñas et al., 2008; Fernández-Pisonero et al., 2012, 2014). In healthy human VICs, TLR-sphingosine 1-phosphate receptor interplay leads to the potentiation of inflammatory, angiogenic, and osteogenic responses through NFκB and p38/MAPK signaling (Fernández-Pisonero et al., 2014).

# Notch

Genetic studies in humans and experiments in mice models suggest the association between the transmembrane receptor Notch and CAVD (Garg et al., 2005; Nigam and Srivastava, 2009). Recent studies provide evidence of the interplay between TLR and Notch pathways on the expression of inflammatory and osteogenic mediators and highlight the importance of their crosstalk in VICs from calcified valves, which have elevated levels of Notch1 (Zeng et al., 2012, 2013). Moreover, the mechanism of TLR4-Notch1 interplay includes the modulation of NF-κB and BMP-2 through a process dependent on Notch 1 cleavage and nuclear translocation (Zeng et al., 2012, 2013).

# Cytokines

IL-37, an anti-inflammatory cytokine that mediates its effects through IL-18 receptor and suppression of NF-κB function (Dinarello et al., 2016), is expressed in human VIC and downregulated in calcified valves (Zhan et al., 2017). Two recent studies uncovered its anti-inflammatory and anti-osteogenic

FIGURE 2 | Inflammation in CAVD progression. The scheme shows the role of inflammation in the different stages of CAVD, at both valve and cellular levels. (1) Valve endothelial layer disruption leads to the recruitment of immune cells and oxLDL. This inflammatory milieu promotes the activation and differentiation of quiescent VIC. These cells express functional TLR, the activation of which induces inflammation and osteogenic reprogramming. (2) In the propagation phase, inflammation promotes the development of microcalcifications. (3) Large scale calcification leads to leaflet thickness and valve dysfunction.

functions by modulating LPS-induced responses in human VIC (Zeng et al., 2017; Zhan et al., 2017). IL-37 negatively modulates AVIC osteogenic responses to both PAMPs such as TLR4/2 agonists, and DAMPs like oxLDL, by inhibiting ERK1/2 and NF-κB activities. Remarkably, IL-37 transgenic mice are protected against early aortic valve lesions induced by prolonged exposure to proinflammatory agents such as LPS (Zeng et al., 2017). Interestingly, IL-37 anti-inflammatory effect is specific of TLR4/2 ligands, most likely by regulating the MyD88-mediated canonical activation of NF-κB (Zhan et al., 2017).

Collectively, evidences provide mechanistic insights into the crosstalk between TLR and membrane receptors relevant to CAVD. Whether this crosstalk occurs at the level of trafficking or signaling and how these multiple pathways integrate in vivo needs to be elucidated. Their regulation may lead to identify therapeutic targets for the suppression of valvular inflammation.

## CONCLUSIONS AND PERSPECTIVES

Recent data highlight a potentially important link between the TLR-NF-κB axis, inflammation, and CAVD pathogenesis. Accumulating evidences suggest that TLRs could be significant contributors to the pathogenesis of valvular inflammation, and the ensuing inflammation-driven calcification processes in aortic valves (**Figure 2**). The key question, whether TLR signaling is involved in the onset of clinically relevant calcification, needs to be investigated using in vivo models recapitulating hyperphosphatemic and/or inflammatory calcification triggers, with an emphasis on the analysis of cell phenotypes to elucidate osteogenic and/or dystrophic mechanisms (Hutcheson et al.,

## REFERENCES


2017). Future studies will benefit from new approaches like optical molecular imaging for the detection of early-stage calcification and 3D cultures of VIC (New and Aikawa, 2011; Hjortnaes et al., 2016), and will help to clearly define the therapeutic potential of TLRs and their signaling transducers in the initial stages of CAVD. The major challenge of targeting this pathway will be minimizing harmful innate immune responses, while preserving appropriate innate immune defense mechanisms.

# AUTHOR CONTRIBUTIONS

CG-R: Contributed to the design and the writing of the manuscript, and to figure preparation; IP-I: Drafted the manuscript and designed figures; IC-M, JL, JAS, and MS: Drafted and revised the manuscript.

# FUNDING

Sources of funding include grants from the Instituto de Salud Carlos III (Spanish Ministry of Health): PI14/00022, CIBERCV; the Ministry of Economy and Competitivity of Spain, cofunded by the European Social Fund: SAF2013-44521- R; Junta de Castilla y León: GRS 1432/A/16, BIO/VA47/14, BIO/VA36/15, CSI035P17; Fundación Domingo Martínez. IP-I was a fellow from the University of Valladolid cofunded by Banco de Santander; IC-M was supported by a pre-doctoral fellowship from Regional Government of Castilla y León and the European Social Fund co-funded by CSIC.


common link with calcification and inflammation? Eur. Heart J. 34, 1567–1574. doi: 10.1093/eurheartj/eht034


**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 García-Rodríguez, Parra-Izquierdo, Castaños-Mollor, López, San Román and Sánchez Crespo. 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.

digital media

of impactful research

article's readership