# NEUROINFLAMMATION AND ITS RESOLUTION: FROM MOLECULAR MECHANISMS TO THERAPEUTIC PERSPECTIVES

EDITED BY : Morena Zusso, Stefano Moro, Pietro Giusti and Leanne Stokes PUBLISHED IN : Frontiers in Pharmacology and Frontiers in Neruoscience

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-854-3 DOI 10.3389/978-2-88963-854-3

#### 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

# NEUROINFLAMMATION AND ITS RESOLUTION: FROM MOLECULAR MECHANISMS TO THERAPEUTIC PERSPECTIVES

Topic Editors: Morena Zusso, University of Padua, Italy Stefano Moro, University of Padua, Italy Pietro Giusti, University of Padua, Italy Leanne Stokes, University of East Anglia, United Kingdom

Citation: Zusso, M., Moro, S., Giusti, P., Stokes, L., eds. (2020). Neuroinflammation and its Resolution: From Molecular Mechanisms to Therapeutic Perspectives. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-854-3

# Table of Contents


Jianbo Zhang, Xia Shi, Na Hao, Zhi Chen, Linjie Wei, Liang Tan, Yujie Chen, Hua Feng, Qianwei Chen and Gang Zhu


Michele Malaguarnera, Marta Llansola, Tiziano Balzano, Belén Gómez-Giménez, Carles Antúnez-Muñoz, Núria Martínez-Alarcón, Rahebeh Mahdinia and Vicente Felipo


Jiaying Wang, Jiajia Wang, Jincheng Wang, Bo Yang, Qinjie Weng and Qiaojun He

*73 Icariin Attenuates M1 Activation of Microglia and A*b *Plaque Accumulation in the Hippocampus and Prefrontal Cortex by Up-Regulating PPAR*g *in Restraint/Isolation-Stressed APP/PS1 Mice*

Yihe Wang, Tianrui Zhu, Min Wang, Feng Zhang, Guitao Zhang, Jing Zhao, Yuanyuan Zhang, Erxi Wu and Xiaohong Li

*89 Natural Diterpenoid Oridonin Ameliorates Experimental Autoimmune Neuritis by Promoting Anti-inflammatory Macrophages Through Blocking Notch Pathway*

Lu Xu, Lei Li, Chen-Yang Zhang, Hermann Schluesener and Zhi-Yuan Zhang


Olena Lykhmus, Lyudmyla Koval, Larysa Voytenko, Kateryna Uspenska, Serhiy Komisarenko, Olena Deryabina, Nadia Shuvalova, Vitalii Kordium, Alina Ustymenko, Vitalii Kyryk and Maryna Skok

### *128 A Bcr-Abl Inhibitor GNF-2 Attenuates Inflammatory Activation of Glia and Chronic Pain*

Gyun Jee Song, Md Habibur Rahman, Mithilesh Kumar Jha, Deepak Prasad Gupta, Sung Hee Park, Jae-Hong Kim, Sun-Hwa Lee, In-Kyu Lee, Taebo Sim, Yong Chul Bae, Won-Ha Lee and Kyoungho Suk

*141 Protective Effects of* Phyllanthus amarus *Against Lipopolysaccharide-Induced Neuroinflammation and Cognitive Impairment in Rats*

Akilandeshwari Alagan, Ibrahim Jantan, Endang Kumolosasi, Satoshi Ogawa, Maizaton Atmadini Abdullah and Norazrina Azmi

*153 Antidepressant-Like Activity of Myelophil* via *Attenuation of Microglial-Mediated Neuroinflammation in Mice Undergoing Unpredictable Chronic Mild Stress*

Jin-Seok Lee, Won-Young Kim, Yoo-Jin Jeon, Sung-Bae Lee, Dong-Soo Lee and Chang-Gue Son

*165 Annexin A1 Bioactive Peptide Promotes Resolution of Neuroinflammation in a Rat Model of Exsanguinating Cardiac Arrest Treated by Emergency Preservation and Resuscitation*

Qing Ma, Zhiquan Zhang, Jae-Kwang Shim, Talaignair N. Venkatraman, Christopher D. Lascola, Quintin J. Quinones, Joseph P. Mathew, Niccolò Terrando and Mihai V. Podgoreanu

*180 Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-*k*B Signaling*

Fei Zhang, Yuan Zhang, Ting Yang, Ze-Qing Ye, Jing Tian, Hai-Rong Fang, Juan-Juan Han, Zhe-Zhi Wang and Xing Li

*194 Corrigendum: Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-*k*B Signaling*

Fei Zhang, Yuan Zhang, Ting Yang, Ze-Qing Ye, Jing Tian, Hai-Rong Fang, Juan-Juan Han, Zhe-Zhi Wang and Xing Li

*195 Microglial Drug Targets in AD: Opportunities and Challenges in Drug Discovery and Development*

Knut Biber, Anindya Bhattacharya, Brian M. Campbell, Justin R. Piro, Michael Rohe, Roland G.W. Staal, Robert V. Talanian and Thomas Möller

*214 Neuroinflammation as a Common Feature of Neurodegenerative Disorders*

Leonardo Guzman-Martinez, Ricardo B. Maccioni, Víctor Andrade, Leonardo Patricio Navarrete, María Gabriela Pastor and Nicolas Ramos-Escobar

*231 N-3 Polyunsaturated Fatty Acids and the Resolution of Neuroinflammation*

Corinne Joffre, Charlotte Rey and Sophie Layé

*247 NVP-BEZ235 (Dactolisib) has Protective Effects in a Transgenic Mouse Model of Alzheimer's Disease*

Paula Maria Quaglio Bellozi, Giovanni Freitas Gomes, Leonardo Rossi de Oliveira, Isabella Guimarães Olmo, Érica Leandro Marciano Vieira, Fabíola Mara Ribeiro, Bernd L. Fiebich and Antônio Carlos Pinheiro de Oliveira

*258 The Role of Intercellular Adhesion Molecule-1 in the Pathogenesis of Psychiatric Disorders*

Norbert Müller

*267 Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats* via *Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC*

Ke Fang, Hua-Rong Li, Xing-Xing Chen, Xin-Ran Gao, Ling-Ling Huang, An-Qi Du, Chuan Jiang, Hua Li and Jin-Fang Ge

*280 Corrigendum: Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats* via *Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC*

Ke Fang, Hua-Rong Li, Xing-Xing Chen, Xin-Ran Gao, Ling-Ling Huang, An-Qi Du, Chuan Jiang, Hua Li and Jin-Fang Ge

# Editorial: Neuroinflammation and Its Resolution: From Molecular Mechanisms to Therapeutic Perspectives

Morena Zusso1\*, Leanne Stokes <sup>2</sup> , Stefano Moro<sup>1</sup> and Pietro Giusti <sup>1</sup>

<sup>1</sup> Department of Pharmaceutical and Pharmacological Sciences, University of Padua, Padua, Italy, <sup>2</sup> School of Pharmacy, University of East Anglia, Norwich, United Kingdom

Keywords: neuroinflammation, neurodegenerative diseases, microglia, cytokines, therapeutic targets

Editorial on the Research Topic

Neuroinflammation and Its Resolution: From Molecular Mechanisms to Therapeutic Perspectives

Neuroinflammation, the complex immune response of the central nervous system (CNS), when sustained, is a common denominator in the etiology and course of all major neurological diseases, including neurodevelopmental, neurodegenerative, and psychiatric disorders (e.g., Alzheimer's disease, AD; Parkinson's disease, PD; multiple sclerosis, MS; motor neuron disease; depression; autism spectrum disorder; and schizophrenia). Cellular (microglia and mast cells, two brain-resident immune cells, together with astrocytes) and molecular immune components (e.g., cytokines, complement and patternrecognition receptors) act as key regulators of neuroinflammation (Skaper et al., 2012). In response to pathological triggers or neuronal damage, immune cells start an innate immune response with the aim to eliminate the initial cause of injury. However, when the cellular activity becomes dysregulated, it results in an inappropriate immune response that can be injurious and affect CNS functions. Thus, limiting neuroinflammation and microglia activity represents a potential strategy to alleviate neuroinflammationrelated diseases.

The Research Topic collects 20 manuscripts, divided into five sections, that include both original research articles and reviews of the emerging literature and explore the role of neuroinflammation in various neurological diseases. There is particular attention dedicated to the relevant research exploring the mechanisms and mediators involved in the resolution of neuroinflammation. Our aim was to generate a valuable discussion contributing to identify new therapeutic targets in brain damage and providing new drug development opportunities for the prevention and treatment of CNS diseases involving neuroinflammation.

### PATHOPHYSIOLOGY OF NEUROINFLAMMATION

Guzman-Martinez et al. presented an overview of the various biochemical pathways involved in the neuroinflammatory cascade, focusing on the role of neuroinflammation in the process of protein alteration implicated in various neurodegenerative diseases, such as AD, PD, and Huntington disease. Herradon et al. highlighted the crosstalk between peripheral inflammation and neuroinflammation. The authors described cytokines such as pleiotrophin and midkine as key mediators in modulating both peripheral inflammation and neuroinflammation and they discussed the role of these cytokines in mediating both chronic neuroinflammatory conditions (i.e., neurodegenerative diseases) and peripheral inflammation. Furthermore, the existing therapeutics

#### Edited and reviewed by:

Nicholas M. Barnes, University of Birmingham, United Kingdom

\*Correspondence: Morena Zusso morena.zusso@unipd.it

#### Specialty section:

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

Received: 06 March 2020 Accepted: 26 March 2020 Published: 08 April 2020

#### Citation:

Zusso M, Stokes L, Moro S and Giusti P (2020) Editorial: Neuroinflammation and Its Resolution: From Molecular Mechanisms to Therapeutic Perspectives. Front. Pharmacol. 11:480. doi: 10.3389/fphar.2020.00480 for overt neuroinflammation are reviewed and new perspectives suggested for the development of innovative strategies that focus on the modulation of the pleiotrophin-midkine pathway.

Oxidative stress is a critical and common feature of a wide spectrum of CNS pathologies and has been strongly implicated in microglial activation and neuroinflammation. Malko et al. provided an extensive review focusing on the role of the transient receptor potential melastatin-related 2 (TRPM2) channel, an oxidative stress-sensitive calcium-permeable cationic channel, in the activation of microglia and neuroinflammation. The authors emphasized that in CNS diseases microglia-related neuroinflammation is strongly attenuated in TRPM2- KO mice or in animals treated with a TRPM2 inhibitor. These results suggested that microglial TRPM2 represents a prospective novel therapeutic target for neuroinflammatory CNS diseases.

In the attempt to find new antiinflammatory treatments, Joffre et al. described the antineuroinflammatory and neuroprotective effects of n-3 long chain polyunsaturated fatty acids and the derived specialized proresolving mediators, with the aim to identify molecules derived from these fatty acids as possible protagonists of the antiinflammatory effects of polyunsaturated fatty acids.

### NEUROINFLAMMATION IN AD AND MEMORY IMPAIRMENT

AD is the most common age-related neurodegenerative disease and is characterized by progressive memory decline and cognitive dysfunction. The impact of chronic inflammation is of relevance in the pathophysiology of AD, considering that subjects affected by inflammatory diseases have an increased risk of developing AD (Heneka et al., 2015). Dysregulated immunoactivity in AD has been widely studied and several papers point to microglia as legitimate drug targets in AD. In this context, Biber et al. provided a concise and informative review that reconciled some of the most recent updates in the understanding of microglia biology with drug discovery, noting that the dynamic properties of microglia and their location in the CNS make targeted small molecule therapies relatively difficult to develop. The authors proposed several microglia target candidates (e.g., CD33, kynurenine pathway, Toll-like receptors, the potassium channel KCa3.1, etc.), with the optimistic hope that one of them could be used in developing drugs useful for AD therapy.

In the attempt to understand the mechanisms underlying the link between neuroinflammation and AD, several authors have studied the effects of some small molecules in animal models of AD. Bellozi et al. evaluated the effects of a 14 days oral treatment with dactolisib, a dual phosphatidylinositol 3 kinase (PI3K)/mammalian target of rapamycin (mTOR) inhibitor, in a transgenic mouse model of AD overexpressing the amyloid precursor protein (APP). Among the different pathways involved in the maintenance and progression of AD, an abnormal and continuous activation of PI3K/protein kinase B (Akt)/mTOR signaling contributes to disease progression, because of the disrupted clearance of amyloid b (Ab) and tau, synaptic loss, and cognitive decline (O'neill, 2013; Heras-Sandoval et al., 2014). Dactolisib reduced social memory impairment, microglia activation in CA3 region of hippocampus, and the levels of hippocampal interleukin (IL)-10, suggesting that an adequate control of PI3K/Akt/mTOR pathway activation might have the potential to ameliorate AD signs and symptoms. Ano et al. demonstrated interesting effects of a short-term intake of iso-aacids, which are bitter components in beer, in the reduction of the hippocampal inflammation and neural hyperactivation and memory impairment in a transgenic mouse model of AD.

Considering that chronic psychosocial stress represents a risk factor for AD, being associated with cognitive deficits, microglia priming, and inflammatory responses in adult brain (Piirainen et al, 2017), Wang Y. et al. focused on the relation between social stress and impairment of learning and memory in AD. The authors showed that icariin, a natural flavonoid extracted from Epimedium brevicornum Maxim (a traditional Chinese herb), attenuated restraint/isolation stress-induced memory damage, Ab accumulation, and neuroinflammation by reducing microglia M1 activation and activating peroxisome proliferator-activated receptor g in APP/PS1 mice.

Alagan et al. showed the neuroprotective and antiinflammatory effects of an ethanol extract of Phyllanthus amarus in an in vivo rat model of lipopolysaccharide (LPS)-induced memory impairment and neuroinflammation. Lykhmus et al. presented an interesting study on the effects of mesenchymal stem cells (MSCs) in a mouse model of memory impairment induced by LPS. Intravenously injection of MSCs or their conditioned media prevented a<sup>7</sup> nAChR decrease, Ab accumulation, and episodic memory decline induced by a single dose of systemic LPS in mice, suggesting that MSCs could be a potential therapeutic tool to treat neuroinflammatory-related cognitive pathology.

### NEUROINFLAMMATION IN PSYCHIATRIC DISORDERS

Increasing evidence indicates that dysregulated or enhanced neuroinflammation is closely linked with the pathogenesis of psychiatric disorders, such as schizophrenia, anxiety and depression (Müller and Bechter, 2013; Weber et al., 2017). In a timely and comprehensive review, Müller considered the role of the intracellular adhesion molecule-1 (ICAM-1) in psychiatric illness. ICAM-1, expressed in microglia, astrocytes, and in endothelial cells of the CNS, is of particular interest from two interrelated reasons: it is a cell bound adhesion molecule, intimately associated with the blood brain barrier, and as a soluble protein it has been measured in blood or cerebrospinal fluid as a biomarker for several mental health disorders.

Depression is a recurrent, common, and potentially lifethreatening psychiatric disease related to multiple causes. Hyperactivity of the hypothalamic-pituitary-adrenal axis, involved in the pathophysiology of depression, induces microglia activation and the overproduction of proinflammatory cytokines in the brain (Brites and Fernandes, 2015). Extensive research in recent years has been concentrated on the immuno-inflammatory mechanisms for the treatment of depression. Indeed, many antidepressants are endowed with antiinflammatory effects (Kenis and Maes, 2002). However, current antidepressants have important limitations, due to significant adverse reactions, poor compliance, and a high risk of relapse following drug withdrawal (Berwian et al., 2017). At present, natural plant products are offering attractive alternatives in antidepressant drug development. In this context, Ke et al. examined the effect of quercetin, a natural polyphenol with antiinflammatory and neuroprotective effects, in a rat model of LPS-induced depression. The authors showed that quercetin attenuated the inflammation induced depression-like behavior and improved learning and memory in LPS-challenged rats. These effects were associated with the reversal effects of quercetin on LPS-induced alteration of expression of brain derived neurotrophic factor, Copine 6, and TREM1/2 in the hippocampus and in the prefrontal cortex. Lee et al. found that myelophil, a 30% ethanol extract of Radix Astragali and Radix Salviae, reversed the depressive behavior in a mouse model of unpredictable chronic mild stress. Myelophil also decreased the over-activation of microglia and the inflammatory response in the hippocampus, and reversed the reduction of serotonergic function in the dorsal raphe nuclei and neurogenesis in the subgranular zone of hippocampus.

### NEUROINFLAMMATION IN MS AND EXPERIMENTAL AUTOIMMUNE NEURITIS

MS is a CNS autoimmune and neurodegenerative disease that affects ~2.5 million people worldwide and causes a significant financial burden. Several experimental and clinical studies have revealed that microglia/macrophages actively participate in the course of the disease. In this context, Wang J. et al. provided a comprehensive review that attempted to discuss the role of macrophages and microglia in both healthy CNS and in MS. The authors discussed the possibility of a differential manipulation of CNS resident microglia and infiltrating macrophages to potentially target different mechanisms operating in MS and realize efficient treatments for the disease. Zhang F. et al. demonstrated that scopoletin, a phenolic coumarin found in some plants, improved the severity of the disease and prominently decreased inflammation and demyelination in an experimental autoimmune encephalomyelitis mouse model of MS, via nuclear factor-kB signaling.

Xu et al. examined the effect of oridonin, a diterpenoid compound extracted from Rabdosia rubescens, in an experimental autoimmune neuritis animal model. Oridonin ameliorated inflammatory disease progression and favored the disease outcome by inducing the switch of macrophages toward the antiinflammatory polarization state. Thus, oridonin could be considered a possible therapeutic candidate of inflammatory neuropathies.

### NEUROINFLAMMATION IN CHRONIC PAIN, HYPERAMMONIA, CARDIAC ARREST, AND INTRACEREBRAL HEMORRHAGE

Chronic pain is caused by nerve damage which occurs during nerve compression, diabetes, inflammation, and shingles virus infection (Campbell and Meyer, 2006). Cytokines, chemokines, prostaglandins, and nitric oxide released from activated microglia and astrocytes in the dorsal horn of the spinal cord play important roles in the pathogenesis of chronic pain (Skaper et al., 2018). Therefore, several studies have targeted activated microglia to reduce pain hypersensitivity. In light of this, the manuscript by Song et al. showed the positive effect of GNF-2, a selective allosteric inhibitor of Bcr-Abl, initially developed as an anticancer drug, on neuroinflammation and associated pain pathogenesis, using different animal models of pain.

Malaguarnera et al. studied the effect of bicuculline, a GABAA receptor antagonist, on hyperammonemic rats. In the hippocampus of these animals, GABAergic tone was increased contributing to some aspects of neuroinflammation. Glutamate transmission was altered and spatial learning and memory as well as anxiety were impaired. As expected, the treatment with bicuculline reduced hippocampal astrocytosis, but failed to control glial activation. Furthermore, the antagonist reduced the GABAergic tone and reversed the expression of the GluA1 and GluA2 subunits of AMPA receptors and of the NR2B subunit of NMDA receptors. The treatment with bicuculline also improved spatial learning, working memory and decreased anxiety in these rats.

The manuscript by Ma et al. explored the capacity of a bioactive annexin A1 short peptide to resolve neuroinflammation in a rat model of exsanguinating cardiac arrest treated by emergency preservation and resuscitation. The attenuation of cortical cell death and reduction of several biomarkers of neuroinflammation induced by the peptide were associated with the increased expression of sirtuin 3 and the upregulation of antioxidant pathways.

Zhang J. et al. indicated that a systemic treatment with simvastatin reduced polymorphonuclear neutrophil infiltration into brain tissue, ameliorated brain edema, and reduced proinflammatory mediator expression in perihematomal area in an animal model of neuroinflammation after intracerebral hemorrhage.

In conclusion, the Research Topic "Neuroinflammation and its resolution: from molecular mechanisms to therapeutic perspectives" presents the current state of knowledge about the involvement of microglia and neuroinflammation in neurological disorders and provides several new perspectives and therapeutic approaches for the development of novel pharmacological agents aimed at counteracting neuroinflammatory diseases.

Finally, it should be noted that in six manuscripts (~30%) of this Research Topic, the antiinflammatory effect of plant-derived products has been investigated. There is a high rate of failure in development of drugs for neurological diseases and for AD, in particular. New treatments are urgently needed, but even if progress is being made, we probably need to define new targets and developing new specific agents. In that regard, natural products could be a valuable source to provide candidates to be used in CNS diseases.

### AUTHOR CONTRIBUTIONS

All the authors have contributed to the composition and the revision of this Editorial Article and approved it for publication.

### REFERENCES


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

Copyright © 2020 Zusso, Stokes, Moro and Giusti. 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.

# Simvastatin Reduces Neutrophils Infiltration Into Brain Parenchyma After Intracerebral Hemorrhage via Regulating Peripheral Neutrophils Apoptosis

Jianbo Zhang<sup>1</sup>† , Xia Shi<sup>2</sup>† , Na Hao<sup>3</sup> , Zhi Chen<sup>1</sup> , Linjie Wei<sup>1</sup> , Liang Tan<sup>1</sup> , Yujie Chen<sup>1</sup> , Hua Feng<sup>1</sup> , Qianwei Chen1,4 \* and Gang Zhu<sup>1</sup> \*

<sup>1</sup>Department of Neurosurgery, Southwest Hospital, Third Military Medical University (Army Medical University), Chongqing, China, <sup>2</sup>Department of Nutrition, Southwest Hospital, Third Military Medical University (Army Medical University), Chongqing, China, <sup>3</sup>Department of Orthopedics, Chongqing Hospital of Traditional Chinese Medicine, Chongqing, China, <sup>4</sup>Department of Neurosurgery, The 452 Hospital of Western Air Force, Chengdu, China

Statins, known for their lipid-lowering effects, also have immunomodulatory properties. This study aims to examine whether systematic simvastatin administration could decrease polymorphonuclear neutrophils (PMNs) infiltration into brain tissue, as well as alleviate neuroinflammation in a rat model of intracerebral hemorrhage (ICH). The ICH model was induced in adult male Sprague–Dawley rats by an injection of autologous blood. Animals randomly received simvastatin (i.p. 2 mg/kg) or vehicle daily from 5 days before ICH until sacrificed. Routine blood counts, brain water content, neurological scoring, immunofluorescence and RT-PCR were conducted to evaluate the antiinflammatory effect of simvastatin following ICH. Furthermore, flow cytometric and western blotting analysis were implemented for elucidating the mechanisms involved in simvastatin-induced reduction of neutrophil brain-invading. Elevated PMNs count and neutrophil-to-lymphocyte ratio in circulation were detected in rat model of ICH, which was reversed by using simvastatin. Simvastatin effectively alleviated PMNs infiltration and proinflammatory factors release in perihematomal area, as well as attenuated ICHinduced brain edema and neurological deficits. Simvastatin significantly downregulated the expression of antiapoptotic protein-Mcl-1 while increased the level of proapoptotic protein-Bax and cleaved caspase 3 in PMNs. Simvastatin treatment significantly alleviated PMNs brain-infiltrating and subsequent neuroinflammatory reaction after ICH, in part by accelerating peripheral PMNs apoptosis through disorganized the expression of apoptotic related proteins. Our data provided new evidence for simvastatin application on patients with ICH.

Keywords: intracerebral hemorrhage, statins, polymorphonuclear neutrophils, apoptosis, inflammation

#### Edited by:

Pietro Giusti, Università degli Studi di Padova, Italy

#### Reviewed by:

Andrea Baragetti, University of Milan, Italy Massimiliano Ruscica, University of Milan, Italy

#### \*Correspondence:

Qianwei Chen qianweichen88@yeah.net Gang Zhu gangzhu6666@yeah.net †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuropharmacology, a section of the journal Frontiers in Neuroscience

Received: 17 August 2018 Accepted: 06 December 2018 Published: 18 December 2018

#### Citation:

Zhang J, Shi X , Hao N, Chen Z, Wei L, Tan L, Chen Y, Feng H, Chen Q and Zhu G (2018) Simvastatin Reduces Neutrophils Infiltration Into Brain Parenchyma After Intracerebral Hemorrhage via Regulating Peripheral Neutrophils Apoptosis. Front. Neurosci. 12:977. doi: 10.3389/fnins.2018.00977

## INTRODUCTION

fnins-12-00977 December 15, 2018 Time: 17:34 # 2

Intracerebral hemorrhage (ICH) accounts for 10–15% of all strokes, and its mortality and morbidity far exceed ischemic strokes (Qureshi et al., 2009). The lack of a specific therapeutic target in the treatment of ICH has increased the need for new treatment options (Xi et al., 2014; Joseph et al., 2016). More and more studies have provided evidence for supporting the key role of neuroinflammation in secondary brain injury following ICH (Zhou et al., 2014).

After ICH, the microglia around the hematoma takes the lead in response to injury, and then releases a variety of chemokines to recruit peripheral inflammatory cells migrate into lesion area of brain (Wan et al., 2016). Among them, the polymorphonuclear neutrophils (PMNs) first arrived perihematomal region. Both previous clinicopathological findings and recent animal experiments suggested that PMNs began to infiltrate into the brain at 6 h after ICH and reached peak on 1˜3 day (Wang, 2010). Subsequently, a large number of proinflammatory cytokines released by PMNs will activate the adjacent microglia/macrophages, leading to the release of more proinflammatory factors and resulting in the "inflammatory cascade effect," which worse the prognosis of patients with ICH (Iadecola and Anrather, 2011; Behrouz, 2016). Taken together, PMNs is considered as the "fuse" of inflammation in the central nervous system (CNS), which could be a promising therapeutic target for ICH. Many years ago, some investigators had tried to deplete peripheral PMNs in a rat model of ICH (Moxon-Emre and Schlichter, 2011; Sansing et al., 2011). They found that PMNs specific antibody injection effectively diminished PMNs and monocyte infiltration into brain, attenuated BBB breakdown and improved functional outcome. However, to some extent, the proper dosage of PMNs antibody is hard to determine among human beings, and rapid loss of PMNs in circulation may breakdown the immune system homeostasis and lead to infection. Therefore, new therapeutic methods were still needed for suppressing PMNs brain-invading post-ICH.

In the past few years, several clinical studies have pointed out that the elevated PMNs count and neutrophil-to-lymphocyte ratio (NLR) in circulation was closely related to the poor prognosis after ICH (Lattanzi et al., 2016; Wang et al., 2016; Gusdon et al., 2017; Tao et al., 2017). However, the mechanisms responsible for this relationship remain poorly characterized. Recently, a randomized, double-blind clinical study reported that systematic administration of simvastatin, an HMG-CoA reductase inhibitor, before cardiopulmonary bypass significantly increased the apoptotic ratio of peripheral PMNs, and ameliorated post-operative inflammation (Chello et al., 2007). Taken together, we hypothesized that simvastatin could be a safe and effective candidate for attenuating PMNs brain-infiltration and subsequent inflammatory reaction following ICH, by reducing the PMNs count and NLR in circulation through regulating the apoptosis of PMNs. The present study was designed to test whether consecutive simvastatin treatment (before and after ICH) could decline the peripheral PMNs count and NLR, by which diminishing PMNs infiltration into brain in a rat model of ICH.

### MATERIALS AND METHODS

### Animals and ICH Model

Two hundred and eighty-five adult male Sprague-Dawley rats (250–350 g; the Third Military Medical University) were used. All animals were housed with a 12 h light/dark cycle and water and food provided ad libitum. A feedback-controlled heating pad was used to maintain rats at 37.0◦C during operation. The animals were kept warm at thermostats and clear respiratory secretions after operation, then return the model rats to cages until its fully awake. Animal use procedures were in compliance with the Guide for the Care and Use of Laboratory Animals and approved by the Laboratory Animal Welfare and Ethics Committee of the Third Military Medical University (SCXK-PLA-20120011). Animals were anesthetized with pentobarbital (40 mg/kg IP), and a feedback-controlled heating pad was used to maintain body temperature at 37.0◦C. The rat model of ICH was established according to our previous published work (Chen et al., 2017; Jiang et al., 2017). For the model of ICH, a cranial burr hole (1 mm) was drilled, and a 29-gauge needle was inserted stereotaxically into the right caudate nucleus (coordinates: 0.2 mm anterior, 5.5 mm ventral, and 3.5 mm lateral to the bregma). Subsequently, 100 µl of autologous arterial blood was infused in 10 min using a microinfusion pump. Then, the hole was sealed using bone wax. The sham groups received only needle injection.

### Experimental Grouping

This study was divided into 3 parts. First, to examine whether simvastatin could decrease peripheral PMNs count and PMNs brain-infiltrating, ninety rats had an intracaudate injection of 100 µl of blood. The sham control received only needle injection. The animals were randomly assigned to three groups. Group 1 received simvastatin (2 mg/kg/d, i.p.) from 5 days before ICH until sacrificed, and the control group received the same volume of vehicle. Some rats (n = 4 per group, each time point) were sacrificed for RT-PCR analysis of TNF-α, IL-6, CCL2 and ICAM-1 at 6, 12, and 24 h after ICH. Other animals (n = 5 for sham group, n = 8 for ICH+simva. group and ICH+Veh. group each time point) were employed for blood cell count and MPO staining at 1, 3, and 7 days after ICH. In the second part, to evaluate the neuroprotective effect of simvastatin treatment on brain injury following ICH, rats were randomly divided into three groups and treated as part I. Neurological function scoring was conducted in some rats one day before ICH and 1, 3, and 7 days following ICH (n = 6 per group). Other animals (n = 6 per group, each time point) were sacrificed for brain water content measurement at 24 and 72 h after ICH. In the third part, to explore the mechanism underlying simvastatin-mediated against peripheral PMNs brain-invading after ICH, rats were randomly divided into three groups and treated as part I. Flow cytometric analysis of peripheral PMNs apoptosis (n = 6 per group, each time point) and lymphocytes apoptosis (n = 6 per group, each time point) were conducted at days 1, 3, and 7 after ICH. Then, the level of apoptotic related proteins was detected using Western blotting analysis at 24 and 72 h after ICH (n = 4 per group for each time point).

### Drug Administration

fnins-12-00977 December 15, 2018 Time: 17:34 # 3

Simvastatin (Sigma, United States) was prepared as a 4 mg/ml stock, as previously described (Leung et al., 2003; Jin et al., 2013). Briefly, 4 mg of simvastatin was dissolved in 100 µl of ethanol and 150 µl of 0.1 N NaOH, incubated at 50◦C for 2 h, then the pH adjusted to 7.0 with 0.1 M HCl, and added water to 1 ml. This simvastatin stock solution was stored at −80◦C and diluted with triple volume of sterile saline immediately before use. Animals were randomized to receive intraperitoneal injection of simvastatin (2 mg/kg/d) from 5 days before ICH until sacrificed. The dose regimen of simvastatin was referred to previous studies in rat model of ICH (Karki et al., 2009; Yang et al., 2013; Chen et al., 2017).

### Routine Blood Counts

Routine blood counts were performed as previously described (Welles et al., 2009). First, animals were anesthetized with pentobarbital, then the heart was exposed and about 4 ml fresh blood samples were collected using the EDTA-anticoagulated tube. After fully shake, 200 µl blood sample was transferred to an eppendorf tube and analyzed on the bench-top analyzer (Hemavet 950, Shandong Excellent Science Instrument Co. Ltd., CHN). Before analyzer running, the quality of machine cleaning liquid, hemolysin liquid and protective agent were checked, and then use double steamed water for calibration.

### PMNs and Lymphocytes Isolation

PMNs isolation and identification were conducted as previously described (Dyugovskaya et al., 2012). The EDTA-anticoagulated whole blood from rats was collected, and it was then mixed with dextran t-500 (1% v/w) for 30 min at 37◦C. The upper leukocyterich layer was transferred to a new tube and centrifuged at 2000 g for 20 min at 20◦C. Pellets were suspended in 2 ml salt solution balanced by D-Hanks; then, they were loaded on the top of the 2 ml Histopaque1119 and Histopaque1083 density gradient carefully and centrifuged at 700 g for 30 min at 20◦C. The cells were collected in the PMN-rich layer between Histopaque1083 and Histopaque1119 and were suspended in D-Hanks. The viability of the cells was determined by Trypan blue dye exclusion. The purity of isolated PMNs was detected by Wright-Giemsa staining. PMNs' viability and purity were more than 95%. Refer to the above methods, we also isolated lymphocytes from circulating blood of ICH rat.

### Wright-Giemsa Staining

Wright-Giemsa staining was performed standard protocols (Qin et al., 2015). Briefly, the smear of peripheral blood or isolation was staining with Wright-Giemsa stains A dye 10–15 s and then staining with Wright-Giemsa stains B dye 2 min, flush slides under running water for at least 30 s after staining.

## PMNs and Lymphocytes Apoptosis Detection

The apoptotic ratio was measured by flow cytometry as previously described (Jin et al., 2013). The D-Hanks-washed cells were incubated on ice with 5 µl Annexin V-fluorescein isothiocyanate (FITC) solution and 10 µl propidium iodide (PI) solution for 15 min in dark. Then, flow cytometry (BD LSRF Ortessa, United States) was used to analyze the apoptosis.

### Brain Water Content

Animals were decapitated under deep anesthesia 24 and 72 h after autologous blood injection for brain water content measurement. Brains were removed quickly, and the frontal poles (4 mm) were cut off. The remaining brains were divided into five parts: cerebellum, ipsilateral cortex and basal ganglia, contralateral cortex and basal ganglia. Brain samples were weighed immediately to record the wet weight and were then dried at 100◦ for 24 h to record the dry weight. Tissue water content was calculated as (wet weight – dry weight)/wet weight. The brain water content was measured by as previously described (Williamson and Colbourne, 2017).

### Assessment of Neurological Abnormalities

Neurological dysfunction of rats was evaluated using a modified Neurological Severity Score (mNSS) method, forelimb placing test and corner test as described previously (Hua et al., 2002; Liew et al., 2012). Briefly, the assessment was performed on day 1 before and on days 1, 3, and 7 after ICH. First, the mNSS is a composite test of motor, sensory, and balance functions. Neurological function was graded on a scale of 0–18 (normal score, 0; maximal deficit score, 18). Then, each rat was tested 10 times for each forelimb, and the percentage of trials in which the rat placed the appropriate forelimb on the edge of the countertop in response to vibrissae stimulation was determined. Next, all rats were allowed to proceed into a corner, whose angle was 30◦ . To exit the corner, the rat could turn to either the left or the right, and that direction was recorded. The test was repeated 10 times, with at least 30 s between trials, and the percentage of right turns was calculated. Testers were highly experienced and blinded to the condition of the animal. The mean neurological score was evaluated by 2 blinded observers.

### Cell Counts

The cell counts were performed at days 1, 3, and 7 after ICH, respectively (Chen et al., 2017). Cell counts analysis was performed as previously described. For quantification of the MPO positive cells in the perihematomal area (0.2 mm anterior to bregma), consecutive slices were made, and two sections per animal (n = 6 per group) with 40 µm space in between were used for cell counts. Three high-power images (40× magnification) were used for cell counting. Cell counts were performed by two researchers in a blinded manner. All measurements were repeated three times, and the mean value was used.


#### TABLE 1 | Primers used for RT-PCR.

fnins-12-00977 December 15, 2018 Time: 17:34 # 4

GAPDH, Glyceraldehyde-3-phosphate dehydrogenase; IL-6, Interleukin-6; TNF-α, Tumor Necrosis Factor-α ; CCL2, C-C motif chemokine ligand 2; ICAM-1, Intercellular Adhesion Molecule 1.

### Real-Time PCR

The PCR was performed and analyzed, as previously described (Tan et al., 2017). For RT-PCR of TNF-α, IL-6, CCL2, and ICAM-1 gene expression analysis, rats (4/group) were sacrificed 6, 12, and 24 h by decapitation after ICH. The brains were then dissected 2 mm anteriorly and 2 mm posteriorly to the needle entry site (easily identifiable on the brain surface) and were divided into separate hemispheres along the midline. Next, perihematomal brain tissue was used for RNA extraction. Primers were designed with the Primer3 Output program (**Table 1**). Total RNA was extracted using TRIzol reagent (Invitrogen). A positive standard curve for each primer was obtained using a serially diluted cDNA (complementary DNA) sample mixture. Gene expression was quantified with standard samples and normalized with glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The data are expressed as normalized mRNA expression. All data analyses were performed in a blinded manner by 2 observers.

### Western Blot Analysis

Western blot analysis was performed as previously described (Zou et al., 2017). Pre-extracted and washed cells were transferred to a sterile EP tube. Then, add the prepared lysis solution to the test tube by a ratio of 1 ml/10<sup>7</sup> cells. Repeatedly pipetting and mixing until all the cellular proteins precipitation. Next, cells were lysed on ice for 5 min, vortexed, and repeat three times. After a centrifugation at 12,000 rpm for 10 min under 4◦ condition, the supernatant was collected and stored at −80◦C for Western blot analysis. Antibodies were used as follows: Rabbit anti-rat Cleaved Caspase-3 (Cell Signaling Technology, United States 1:1000), Rabbit anti-rat Mcl-1 (Cell Signaling Technology, United States 1:1000), Rabbit anti-rat Bcl-2 (Cell Signaling Technology, United States 1:1000), Mice anti-rat Bax (Abcam, United States 1:500). Immunopositive bands of horseradish peroxidase–conjugated secondary antibodies were detected with an ECL system (GE Healthcare). The relative densities of the bands were analyzed using NIH ImageJ software.

### Determination of Myeloperoxidase (MPO) Activity

MPO activity has been used as an index of neutrophil infiltration in brain tissues. In this study, we used MPO activity to indicate neutrophils infiltration in the brain tissue after injection. Immunofluorescence staining of brain tissue was performed on fixed frozen sections as previously described (Feng et al., 2017). Rats were anesthetized with pentobarbital (100 mg/kg intraperitoneal) and perfused with 4% paraformaldehyde in 0.1 mol/L pH 7.4 PBS. The brains were removed and kept in 4% paraformaldehyde for 4–6 h and then immersed in 30% sucrose for 3–4 days at 4◦C. The brains were embedded in an optimal cutting temperature compound (SAKURA, United States), and 18 mm thick slices were cut using a cryostat. The slices were stored at −20◦ before staining. Before being blocked with 10% goat serum for 1 h at room temperature, the sections were washed with PBST (PBS with 0.3% Triton X-100) for 30 min. Then, they were incubated with antibodies against MPO (Abcam, United States 1:100) at 4◦ overnight. The sections were kept at room temperature for 45 min and then washed with PBS. Then, the sections were incubated with an Alexa Fluor 555-conjugated goat anti-rabbit IgG (H+L) (Beyotime, China 1:300) secondary antibody at 37◦C for 3 h. Finally, cell nuclei were stained with 40,6-diamidino-2-phenylindole (DAPI). The stained sections were viewed under identical conditions using a 310 or 320 objective on a confocal microscope (LSM-780; Zeiss).

### Statistical Analyses

The values in this study are presented as mean ± SD. The data in this study are given as the mean ± SD. Data were analyzed by one-way analysis of variance, followed by Scheffe's post hoc test. Differences were considered statistically significant at a P-value of less than 0.05.

## RESULTS

### Elevated Peripheral PMNs Count and Neutrophil-to-Lymphocyte Ratio (NLR) Were Detected in Rat Model of ICH, Which Was Reversed by Using Simvastatin

Many clinical researches have reported that ICH patients with higher peripheral PMNs count and NLR in acute stage predicted more worse outcome (Lattanzi et al., 2016; Wang et al., 2016; Gusdon et al., 2017; Tao et al., 2017). Thus, first, we examined whether this phenomenon also exits in experimental animal model of ICH. As shown in **Figure 1**, compared with the sham group, the ICH rats presented higher PMNs count and higher NLR at 24 h after blood injection. Then, to test the potential effect of simvastatin on modulating leukocyte change following ICH, we conducted the dynamic blood routine analysis. The

PMNs, polymorphonuclear neutrophils; NLR, neutrophil to lymphocyte ratio; PBMCs, peripheral blood mononuclear cells; Lym, lymphocytes.

total leukocyte count in Veh-group has no difference among the Simva-, Veh- and sham- groups on day 1, but markedly increased on day 3, and reached a peak on day 7 after ICH. However, the total leukocyte count in the Simva-group remains stable on the 3 and 7 days post-ICH (**Figure 1A**). The PMNs count in the Simva-group was significantly lower than the control group on days 1 and 7 after ICH. Of note, 24 h after ICH, the Simva-group displayed dramatically decreased NLR compared to Vehicle controls. These results suggest that elevated PMNs count and NLR also presented in the experimental ICH model in a rat, which was effectively reversed after simvastatin treatment (**Figures 1B,C**). Furthermore, we also analyzed the count of peripheral blood mononuclear cells (PBMCs) and monocyte in this rat model of ICH (**Figures 1D,E**). PBMCs elevates on 72 h after ICH and stay high for up to 7 days. Simvastatin effectively reduced the PBMCs count on day 7 post-ICH. Notably, simvastatin significantly decreased the monocyte count from 24 h to day 7 after ICH.

### Simvastatin Effectively Alleviated PMNs Brain-Infiltration and Proinflammatory Mediators Expression in Perihematomal Area

Next, we wondered whether simvastatin also changes the number of PMNs infiltration after ICH. The Laser Confocal Microscope was adopted to trace the infiltrated PMNs around hematoma on diffident time points after ICH. Compared with the Sham-group,

both Simva-group and Veh-group showed a lot of MPO (+) cells in the area around the hematoma, in which the MPO (+) cells count in the Simva-group was less than the control one on the days 1, 3, and 7 post-ICH (**Figures 2A,B**). This data indicate that simvastatin effectively prevented peripheral PMNs infiltrating into brain parenchyma, which may be attributed by cutting down the number of PMNs in the circulation after ICH. In addition, to assess whether the degressive infiltration impacts on the proinflammatory mediators and adhesion molecule expression around the lesion, we determined the transcriptional level of some factors, like TNF-α, IL-6, CCL2, and ICAM-1 in acute stage of ICH. RT-PCR analysis showed a lower expression level of all above chemokines in the Simva-group than in the control at 6, 12, and 24 h after ICH (**Figures 2C–F**), suggesting simvastatin relieved the early neuroinflammatory response to ICH in some extent.

### Simvastatin Significantly Attenuated ICH-Induced Brain Edema and Neurological Deficits

To further investigate the neuroprotective effect of simvastatin on ICH rats, we measured the brain water content of animal models and calculated their neurological scorings on corresponding

using mNSS (C), Forelimb Placement Test (D) and Corner Test (E) at the indicated time points after ICH. Values are expressed as mean ± SD, n = 6 per group; <sup>∗</sup>P < 0.05 ICH-Veh group versus Sham group.

time points post-ICH. The brain water contents in both Vehgroup and Simva-group showed an obvious increase after blood injection when compared with the Sham-group. Notably, the brain water content of ipsilateral hemisphere (basal ganglia, cortex) in the Veh-group was significantly higher than the treatment group at 24 and 72 h after ICH (**Figures 3A,B**). As shown in **Figure 3C,** both Veh-group and Simva-group presented higher neurological score than the Sham-group at days 1, 3, and 7 after ICH, in which the neurological score of treatment group was significantly lower than controls at day 7. Furthermore, higher neurological scores were observed in the Simva-group than the Veh-group during the forelimb placement test (**Figure 3D**) and

corner test (**Figure 3E**) at day 7 after ICH. Our above data indicate that simvastatin administration distinctly ameliorated brain edema and improved neurological function following ICH.

### Simvastatin Broke the Balance Between Apoptotic Related Proteins, Mcl-1/Bax, Which Then Driving PMNs Apoptosis

After isolated from the circulating blood of rat, PMNs were identified by using DAPI and Giemsa staining. As shown in **Figure 4**, at high magnification, PMNs showed the feature typical for PMNs-multilobal nuclei. Compared with the sham control, the other two experimental groups showed higher apoptotic ratio at 24 h after ICH, but obviously declined at days 3 and 7. Thereinto, the simvastatin group presented a more significantly elevated apoptotic ratio at 24 h, whereas a less ratio decreasing than the vehicle control at day 3 (**Figures 5A,C**). Our data indicated that simvastatin markedly enhanced the apoptosis of PMNs in the acute phase of ICH, and maintained PMNs under a higher level of apoptosis than control groups in the subacute stage, which in some extent cut down the lifespan of peripheral PMNs and shortened the time course of neuroinflammatory reaction following ICH. As shown in **Figures 5B,D**, simvastatin also effectively increased the apoptotic ratio of lymphocytes on days 1 and 7 post-ICH, suggesting simvastatin has potential proapoptotic effect on various subtypes of leukocyte not just on PMNs. Previous researches have demonstrated that the balance between antiapoptotic proteins and proapoptotic proteins is closely associated with the fate of PMNs (Dzhagalov et al., 2007; Dyugovskaya et al., 2012). Therefore, to further elucidate mechanisms for simvastatin-induced PMNs apoptosis, we traced the level of apoptotic related proteins expression in the peripheral PMNs isolated from ICH rat. As displayed in **Figures 6A,B**, the expression of antiapoptotic proteins, Bcl-2 and Mcl-1, was markedly inhibited at 24 h after ICH, while the proapoptotic proteins, Bax, was significantly upregulated. In comparision with control group, the ICH rats treated with simvastatin showed less antiapoptotic proteins and more proapoptotic proteins expression in peripheral PMNs. In addition, the level of cleaved caspase 3 in Simva-group was higher than the Veh-group at 24 h after ICH. Quantification of these apoptotic related proteins levels by Western blot indicated that simvastatin broke the balance between Mcl-1 and Bax on mitochondria, leading to more caspase 3 release and activation, which then initiated the apoptosis of PMNs after ICH (**Figures 6C,D**). In addition, we also analyzed the protein level of BCL2, MCL1, Bax and Caspase 3 expression on day 3 after ICH. Notably, as displayed in **Figures 6A–D**, simvastatin presented insignificant effect on the expression of apoptotic related proteins.

### DISCUSSION

In 2016, a research team from Italy reported that in patients with acute ICH, higher PMNs, lower lymphocytes, and higher neutrophil-to-lymphocyte ratio (NLR) in peripheral circulating blood predicted worse 3 months outcome (Lattanzi et al., 2016). More Recently, a research further presented that higher NLR is independently associated with edema growth following ICH (Gusdon et al., 2017). These clinical trials suggest that peripheral PMNs count and NLR may be closely related to the prognosis of ICH. However, the mechanisms responsible for this relationship remain poorly characterized. Thus, we tested it for the first time in an experimental animal model of ICH. Notably, in this study, we observed the similar trend like in ICH patients, all ICH rats presented higher leukocyte count, PMNs count and NLR when compared with the control groups. Our data indicated that this animal model of ICH successfully reproduced the clinical situation in human beings. As PMNs plays a key role in neuroinflammation after ICH, inhibiting the migration of PMNs into brain could be a promising therapy for ICH. Furthermore, we also found that PBMCs elevates on 72 h after ICH and stay high for up to 7 days. These data partially fit with other literature reports, showing that after acute condition (miocardila infarction), PBMCs elevates within hours and stay high for up to 72 h (Nahrendorf et al., 2010). This discrepancy may be attributed to the difference between "ischemic" and "hemorrhagic" pathological nature here.

In normal conditions, PMNs lifespan was about 8–20 h in circulation. But in pathology situation, inflammatory response will delay the apoptosis of PMNs and extends its lifespan, such as arthritis, diabetes, Soehnlein et al. (2017). On this basis, we thought that is there any safe drug could accelerate PMNs apoptosis and promotes inflammation resolution after ICH. Recently, a randomized, double-blind clinical study reported that simvastatin therapy before cardiopulmonary bypass significantly increased peripheral PMNs apoptotic ratio and reduced the postoperative peak values of interleukin (IL)-6 and IL-8 (Chello et al., 2007). In this study, we observed that animals treated with simvastatin displays higher PMNs apoptotic ratio than control groups, as well as lower PMNs count and NLR in circulation. In addition, fewer PMNs invasion and lower transcriptional level of proinflammatory factors were detected in the perihematomal area after simvastatin administration. According to previous studies, CCL2 and ICAM-1 were closed with PMN invasion to lesion site after injury (Yang et al., 2005; Reichel et al., 2009). Thus, in present study, we detected the mRNA level of CCL2 and ICAM-1 in the perihematomal area at 6, 12, and 24 h after ICH. The data shows that ICH significantly upregulated the mRNA level of these chemokines and vascular adhesion molecule, which is reversed by the treatment with simvastatin. Our data indicated that simvastatin-mediated against PMN infiltration into brain may in part by inhibiting RNA transcription of CCL2 and ICAM-1 following ICH. In view of this, we proposed that simvastatin could be a safe and effective mean for attenuating PMNs brain-infiltration and subsequent inflammatory reaction following ICH.

According to previous literature, neutrophils can promote inflammatory mediators release that help recruit monocytes/macrophages, amplifying the inflammation. When neutrophils adhere to the endothelial surface, the contents of secretory granules are released, including cationic antimicrobial protein of 37 kd (CAP37)/azurocidin and proteinase 3, leading to endothelial cell activation, increased cellular adhesion molecule expression (such as ICAM-1), and increased monocyte

FIGURE 4 | PMNs isolation and identification. (A) The schematic diagram of PMNs isolation from peripheral blood of rat by gradient centrifugation method; (B) PMNs well identified by using immunofluorescence and Giemsa staining methods. PMNs presented the feature typical for PMNs-multilobal nuclei at high magnification. The scaling bar represents 5 µm in immunofluorescence image.

adhesion (Lee et al., 2003). Neutrophils also contribute to monocyte chemotaxis by releasing LL-37 (De et al., 2000) and cathepsin G (Sun et al., 2004), and enhancing endothelial cell release of monocyte chemoattractant protein-1 (MCP-1) (Taekema-Roelvink et al., 2001). Consistently, in a rat model of ICH, selective depletion of neutrophils resulted in decreased infiltration of monocytes and macrophages into perihematomal area (Moxon-Emre and Schlichter, 2011; Sansing et al., 2011). In our study, neutrophils infiltrated into brain and promoted the mRNA transcription of CCL2 (also referred to as MCP-1), ICAM-1 and pro-inflammatory factors around hematoma. Notably, all of these effects of neutrophils were prevented by simvastatin. Taken together, our data in some extent suggest that neutrophils mobilization play a role in the process of leukocytes activation following ICH.

Previous researches have demonstrated that the balance between Bax/Mcl-1 has a key role in modulating PMNs apoptosis or survival (Dyugovskaya et al., 2012). PMNs constitutively express the proapoptotic members of the Bcl-2 family, including Bax, Bad, Bak, Bid, and Bik. While Mcl-1 is a member of the antiapoptotic proteins of Bcl-2 family, which plays its role by combining with Bax that expression on the outer membrane of mitochondria, forming heteromeric two dimers and stabilizing mitochondrial membrane potential. Then, preventing mitochondrial release of cytochrome C, and ultimately reducing caspase 3 activation to prevent PMN apoptosis. Recent clinical studies have found that the level of Mcl-1 expression is negatively correlated with the severity of PMNs apoptosis (Moulding et al., 1998). Much more, animal study further confirmed that conditional knockout of Mcl-1 gene significantly accelerated the apoptosis of PMN (Dzhagalov et al., 2007). Therefore, to illuminate how simvastatin promoted PMNs apoptosis, we conducted Western blotting analysis of apoptotic related proteins in present study. Our data show that

the level of antiapoptotic protein, Mcl-1, was down-regulated while the expression of proapoptotic protein, Bax, increased after simvastatin used.

HMG-CoA reductase, as a rate-limiting enzyme, catalyzes the cholesterol synthesis pathway in the liver and other tissues. By inhibiting HMG-CoA reductase, statins lower cholesterol levels and might also reduce intracellular levels of isoprenoids, such as farnesyl pyrophosphate and geranylgeranyl pyrophosphate (Weitz-Schmidt, 2002). Isoprenoids are necessary for the posttranslational lipid modification (prenylation) of a variety of proteins, thus anchoring them to the cell membrane (Zhang and Casey, 1996). Protein targets include the small guanosine triphosphate (GTP)-binding proteins that have a key role in signal transduction pathways that regulate cell proliferation, cell differentiation, vesicular transport and apoptosis. In addition to their cholesterol-lowering activity, statins also have pleiotropic effects, like anti-inflammation, promoting hematoma absorption, neurogenesis and neuroplasticity in some CNS diseases via the non-cholesterol pathway (Karki et al., 2009; van der Most et al., 2009; Yang et al., 2013; Wang et al., 2018). In the past few years, some retrospective clinical studies on ICH have suggested that statin use may has potential to reduce mortality, improve outcomes, with no risk of inducing rehemorrhage (Flint et al., 2014; Pan et al., 2014; Chen et al., 2015). However, to date, no randomized clinical trials have examined the exact effect. Although preclinical studies from our research team and other centers have made some progress during the past decade, it is still a long way for statin use in patients with ICH. In this study, we observed simvastatin effectively prevented PMNs infiltration, suppressed proinflammatory factors release, and attenuated brain edema following ICH through regulating PMNs apoptosis. Thus, our results may provide another evidence for future statin use in ICH patients, and more related researches were needed to understand the potential mechanism. Interestingly, a recent work reported that systemic depletion of PMNs at 24 h after ICH exacerbated neurological deficits, suggesting that PMNs infiltration into the brain also shows some beneficial function at the subacute stage post-ICH (Zhao et al., 2018). After further investigation, they proposed that PMNsmediated protective effect in subacute stage may through the two following ways: (1) PMNs can deliver cytoprotective lactoferrin to the ICH-affected brain, neutralizing iron and blocking its toxicity (Zhao et al., 2018). (2) ICH can also transform the phenotype of PMNs that entering brain at the later stages, which then could enhance the beneficial effects like promoted blood detoxification efficacy (Zhao et al., 2017). From the above results, we can read that peripheral PMNs are like a double-edged sword for ICH- it counteracts the iron toxicity as well banes to exacerbate the secondary neuroinflammation at the different stages following ICH. Therefore, correct therapies should be chosen at correct time when targeting circulating blood PMNs after ICH, so as to maximize the natural processes of brain repair. Contrary to traditional belief, increasing evidence support that neutrophils also play a role in chronic

inflammatory disorders, such as obesity and atherosclerosis (Soehnlein et al., 2017). Soehnlein et al. (2017) highlighted some therapeutic strategies to prevent neutrophil-orchestrated chronic inflammation, such as inhibition of neutrophil extracellular trap (NET)-driven inflammation, dampening neutrophil recruitment, promoting neutrophil "reverse migration" from the tissue into the bloodstream, inhibition of cyclin-dependent kinase 9 (CDK9) promotes neutrophil apoptosis.

### CONCLUSION

Our data revealed that the PMNs count and NLR in circulation significantly increased in experimental ICH, which in consistent with the clinical situation occurred in ICH patients. Simvastatin effectively reduced the elevated PMNs count and NLR in circulation, and significantly inhibited PMNs brain-invading and ameliorated ICH-induced brain edema and neurological deficits in rats. Moreover, simvastatin-mediated against PMNs braininfiltration after ICH may in part by accelerating the apoptosis of peripheral PMNs as well as reducing the mRNA level of proinflammatory mediators (**Figure 7**). Our results not only provide evidence for simvastatin use on patients with ICH, but also

### REFERENCES


suggest that PMNs apoptosis regulation may be a new therapeutic target for ICH.

### ETHICS STATEMENT

All institutional and national guidelines for the care and use of laboratory animals were followed.

### AUTHOR CONTRIBUTIONS

JZ, XS, NH, LW, LT, YC, and QC contributed to the implementation of the experiment. QC, HF, ZC, and GZ contributed to the design and paper writing.

### FUNDING

This work was supported by grant no. 81701147 (QC) from the National Natural Science Foundation of China and grant no. SWH2016JCYB-18 (XS) from Science and Technology Innovation Plan of Southwest Hospital.

in intracerebral hemorrhage. Stroke 48, 2589–2592. doi: 10.1161/STROKEAHA. 117.018120



**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 Zhang, Shi, Hao, Chen, Wei, Tan, Chen, Feng, Chen and Zhu. 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.

# Iso-α-Acids, Bitter Components in Beer, Suppress Inflammatory Responses and Attenuate Neural Hyperactivation in the Hippocampus

Yasuhisa Ano1,2 \*, Misato Yoshikawa<sup>3</sup> , Yuta Takaichi<sup>1</sup> , Makoto Michikawa<sup>4</sup> , Kazuyuki Uchida<sup>1</sup> , Hiroyuki Nakayama<sup>1</sup> and Akihiko Takashima3,5

<sup>1</sup> Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo, Japan, <sup>2</sup> Research Laboratories for Health Science & Food Technologies, Kirin Company Ltd., Kanagawa, Japan, <sup>3</sup> Department of Aging Neurobiology, National Center for Geriatrics and Gerontology, Obu, Japan, <sup>4</sup> Department of Biochemistry, School of Medicine, Nagoya City University, Nagoya, Japan, <sup>5</sup> Faculty of Science, Gakushuin University, Tokyo, Japan

Due to the growth in aging populations worldwide, prevention and therapy for

#### Edited by:

Pietro Giusti, University of Padova, Italy

#### Reviewed by:

Carina Rodrigues Boeck, UFN – Universidade Franciscana, Brazil Xiaohui Wang, Shanxi Medical University, China

> \*Correspondence: Yasuhisa Ano yasuhisa\_ano@kirin.co.jp

#### Specialty section:

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

Received: 06 November 2018 Accepted: 21 January 2019 Published: 11 February 2019

#### Citation:

Ano Y, Yoshikawa M, Takaichi Y, Michikawa M, Uchida K, Nakayama H and Takashima A (2019) Iso-α-Acids, Bitter Components in Beer, Suppress Inflammatory Responses and Attenuate Neural Hyperactivation in the Hippocampus. Front. Pharmacol. 10:81. doi: 10.3389/fphar.2019.00081 age-related cognitive decline and dementia are in great demand. We previously demonstrated that long-term intake of iso-α-acids, which are hop-derived bitter compounds found in beer, prevent Alzheimer's pathology in a rodent model. On the other hand, the effects of iso-α-acids on neural activity in Alzheimer's disease model mice have not been investigated. Here, we demonstrated that short-term intake of iso-α-acids suppresses inflammation in the hippocampus and improves memory impairment even after disease onset. Importantly, we demonstrated that short-term administration of iso-α-acids attenuated the neural hyperactivation in hippocampus. In 6-month-old 5 × FAD mice exhibiting hippocampus inflammation and memory impairment, oral administration of iso-α-acids for 7 days reduced inflammatory cytokines, including MIP-1α and soluble Aβ and improved object memory in the novel object recognition test. In 12-month-old J20 mice, intake of iso-α-acids for 7 days also suppressed inflammatory cytokines and soluble Aβ in the brain. Manganese-enhanced magnetic resonance imaging (MEMRI) of hippocampi of J20 mice showed increased manganese compared with wild type mice, but iso-α-acids canceled this increased MEMRI signal in J20 mice, particularly in the hippocampus CA1 and CA3 region. Taken together, these findings suggest that short-term intake of iso-α-acids can suppress hippocampus inflammation even after disease onset and improve hyper neural activity in Alzheimer's disease model mice.

Keywords: Alzheimer's disease, amyloid β, cognitive decline, hippocampus, inflammation, iso-α-acids, lipopolysaccharide

### INTRODUCTION

The rise in aging populations worldwide is accompanied by increasing rates of dementia and cognitive impairment, which are a burden to national healthcare systems as well as patients and their families. Because of the lack of treatments for dementia, preventive approaches such as diet, exercise, and lifelong learning have received increasing research attention. In etiological studies of lifestyle, low to moderate consumption of alcoholic beverages, such as wine and beer, might

reduce the risk of cognitive decline and the development of dementia. Individuals who consume low to moderate levels of alcohol on a daily basis were shown to have significantly lower risks of neurodegenerative diseases compared with individuals who abstained from alcohol or drank heavily (Matsui et al., 2011; Neafsey and Collins, 2011; Horvat et al., 2015). Apart from the effects of alcohol itself, resveratrol, a polyphenolic compound found in red wine, has been shown to have neuroprotective effects (Vidavalur et al., 2006; Arntzen et al., 2010; Porquet et al., 2014; Witte et al., 2014). Our group previously demonstrated that long-term intake of iso-α-acids for 3 months, which are bitter components in beer, prevented Alzheimer's pathology in a transgenic mouse model. Iso-α-acids are derived from hops, the female inflorescences of the hop plant (Humulus lupulus L.), and have been used in beer production since 822. Hops are used as both a preservative and a flavoring agent in the beer-brewing process. Iso-α-acids were shown to activate the peroxisome proliferator-activated receptor-γ (PPAR-γ) and regulate microglial phagocytosis and inflammation (Ano et al., 2017). Our group has previously demonstrated that iso-α-acids prevented dyslipidemia and type 2 diabetes in a diet-induced obese rodent model (Yajima et al., 2004, 2005), and improved glucose metabolism and decreased body fat in a clinical trial (Obara et al., 2009). Long-term administration of iso-α-acids is applicable for the preventive approaches, but the effects for the therapeutic approaches and for neural activity have not been elucidated. It is reported that hyperactivity in hippocampus is associated with cognitive impairment and improvement of the hyperactivity show therapeutic effects on memory impairment (Bakker et al., 2012). In addition, previous study did not conclude that long-term administration of iso-α-acids suppressed the inflammation in the brain directly or as a result of the improvement of amyloid β (Aβ) deposition. To address these research gaps, in the present study, we examined the effects of short-term intake of iso-α-acids on brain inflammation and neural activity in hippocampus using the manganeseenhanced magnetic resonance imaging (MEMRI) in Alzheimer's model mice.

## MATERIALS AND METHODS

### Animals

Alzheimer's disease model, B6SJL-Tg mice [APPSwFlLon, PSEN1∗M146L∗L286V<sup>1</sup> , (Oakley et al., 2006)], hereafter referred to as 5 × FAD transgenic mice, were purchased from Jackson Laboratory (Sacramento, CA, United States) and maintained by crossing hemizygous transgenic mice with B6SJLF1/J mice at the experimental facility of the University of Tokyo. The 5 × FAD transgenic mice overexpress mutant human APP (695) with the Swedish (K670N, M671L), Florida (I716V), and London (V717I) Familial Alzheimer's Disease (FAD) mutations, along with human PS1 harboring two FAD mutations, namely, M146L and L286V. Non-transgenic wild type (WT) littermates as controls were used in the experiments. All experiments were approved by the Animal Care and Use Committee of the Graduate School of Agricultural and Life Sciences, The University of Tokyo, and conducted in strict accordance with its guidelines. Transgenic (J20) mice express human amyloid precursor protein (hAPP) with the Swedish (K670N, M671L) and Indiana (V717F) mutations under the control of the PDGF β-chain promoter (Mucke et al., 2000). Experiments using J20 mice were approved by the local ethical board and complied with the guidelines for animal experimentation of the National Center for Geriatrics and Gerontology in Japan. Pregnant C57BL/6J mice and 6-weekold CD-1 (ICR) mice were purchased from Charles River Japan (Tokyo, Japan) and maintained at a faculty of Kirin Company Ltd. The experiments were approved by the Animal Experiment Committee of Kirin Company Ltd. and conducted in strict accordance with its guidelines since 2014–2016. All efforts were made to minimize animal suffering. Mice were fed a standard rodent diet (CE-2, CLEA Japan, Tokyo, Japan) and maintained at room temperature (23 ± 1 ◦C) under a constant 12-h light/dark cycle (light period from 8 a.m. to 8 p.m.).

### Preparation of Iso-α-Acids

α-Acids predominantly consist of three congeners: cohumulone, humulone, and adhumulone. During the brewing process, they are each isomerized into two epimeric isomers, namely, cis- and trans-iso-α-acids. Purchased isomerized hop extract (IHE) (Hopsteiner, Mainburg, Germany) with 30.5% (w/v) iso-α-acids, comprising trans-isocohumulone (1.74% w/v), cisisocohumulone (7.61% w/v), trans-isohumulone (3.05% w/v), cis-isohumulone (14.0% w/v), trans-isoadhumulone (0.737% w/v), and cis-isoadhumulone (3.37% w/v) as described previously (Taniguchi et al., 2013).

### Primary Microglia Cell Culture

Microglial cells were isolated from brains of newborn C57BL/6J mice (<7 days old) via magnetic cell sorting (MACS) after conjugation with anti-CD11b antibody, as described previously (Ano et al., 2015). Briefly, isolated CD11b-positive cells (>90% pure, evaluated by flow cytometer) were plated into poly-D-lysine (PDL)–coated 96-well plates (BD Biosciences, Billerica, MA, United States) and cultured in a DMEM/F-12 (Gibco, Carlsbad, CA, United States) medium supplemented with 10% fetal calf serum (Gibco) and 100 U/ml penicillium/streptomycin (Sigma-Aldrich, St. Louis, MO, United States). Microglia at a density of 30,000 were treated with each iso-α-acids for 12 h, and then with lipopolysaccharide (LPS, 5 ng/ml, Sigma-Aldrich, St. Louis, MO, United States) and interferon-γ (IFN-γ, 0.5 ng/ml, R&D systems, Minneapolis, MN, United States) for 12 h. After stimulation, the supernatant was used for the TNF-α production assay. Concentrations of cytokines sand chemokines in the supernatant was measured by the Bio-Plex assay system (Bio-Rad, Hercules, CA, United States).

### Neuronal Inflammation Induced by Lipopolysaccharide

Six-week-old ICR male mice were orally administered 1 mg/kg of iso-α-acids dissolved in distilled water once a day for

<sup>1</sup>http://jaxmice.jax.org/strain/006554.html

3 days. One hour after the final administration, the mice were deeply anesthetized with sodium pentobarbital (Kyoritsu Seiyaku, Tokyo, Japan) and injected intracerebroventricularly with 0.25 mg/kg of LPS (5 µl, L7895, Sigma) or PBS as controls as previously described (Ano et al., 2015). Briefly, a micro syringe with a 27-gauge stainless steel needle, 2 mm in length, was used for micro injection. The needle was inserted unilaterally 1 mm to the right of the midline at the equal distance between eyes and ears, and perpendicular to the plane of the skull (anteroposterior, −0.22 mm from the bregma; lateral, 1 mm from the bregma). LPS was delivered gradually within 30 s. The needle was withdrawn after waiting 30 s. Twenty-four hour later, the mice were subjected to the Y-maze test for 8 min to evaluate spontaneous alternation. After the test, the hippocampus and cerebral cortex were removed and homogenized in TBS buffer containing a protease inhibitor cocktail (BioVision, CA, United States). To quantify cytokine and chemokine production, the homogenate supernatant was measured by the Bio-Plex assay system (Bio-Rad).

### Aβ and Cytokine Measurement in Transgenic Mice

To evaluate the effects of iso-α-acids on Alzheimer's-like disease, 6-month-old transgenic 5 × FAD and WT mice were orally administered daily 1 mg/kg iso-α-acids for 7 days. The mice were subjected to a novel object recognition test after 7 days of administration, and their brains were then removed for quantification of Aβ and cytokines. The hippocampus and cerebral cortex were homogenized in TBS buffer (Wako) with a multibeads shocker (Yasui Kikai, Osaka, Japan). After centrifugation at 50,000 × g for 20 min, the supernatant was collected. Pellets were re-homogenized in TBS containing 1% Triton X-100 (Wako) and the supernatant was collected after centrifugation. The total protein concentration of the supernatant was measured with the BCA protein assay kit (ThermoScientific, Yokohama, Japan). The first supernatant was assayed for quantification of soluble Aβ1-42 (Wako) by enzyme-linked immunosorbent assay (ELISA). To quantify cytokines and chemokines, supernatants were evaluated by the Bio-Plex assay system (Bio-Rad). The second supernatant was used for quantification of insoluble Aβ1-42 (Wako) by ELISA.

### Spontaneous Alternation Test

The spontaneous alternation test was performed as previously described (Ano et al., 2018). The test used a three-arm Y-maze with equal angles between all arms (25 cm long × 5 cm wide × 20 cm high) and with walls constructed from dark black polyvinyl plastic. Each mouse was initially placed in one arm, and the sequence and number of arm entries were counted for 8 min. The alternation score (%) for each mouse was defined as the ratio of the actual number of alternations to the possible number (defined as the total number of arm entries minus two) multiplied by 100, i.e., % Alternation = [(Number of alternations)/(Total arm entries−2)] × 100.

### Novel Object Recognition Test

An object recognition test was performed during the light period in a polyvinyl chloride box (25 cm × 40 cm × 20 cm) without a roof, in accordance with previous work (Ayabe et al., 2018). For the acquisition trial, a pair of wooden triangle poles (4.5 cm × 4.5 cm × 4.5 cm) or wooden pyramids (4.5 cm × 4.5 cm × 4.5 cm) was used; for the retention trial, a pair of poles or pyramids and a golf ball (4.5 cm diameter) were used. In all trials, the objects were placed 7.5 cm from the corner of the box. In the acquisition trial, 1 h after oral administration of the test sample, the mouse was allowed to explore the box containing the two objects for 10 min. Twenty-four hour later, 1 h after administration of the test sample, the mouse was allowed to explore the box containing the novel and familiar objects for 5 min. The discrimination index (DI) was calculated by dividing the difference in time exploring the novel object and familiar object by the total time spent exploring both objects, i.e., (novel object exploration time − familiar object exploration time)/(total exploration time); thus, a DI of 0 indicated equal exploration of both objects.

### Manganese-Enhanced MRI

Manganese-enhanced magnetic resonance imaging was performed in accordance with our previous report (Yoshikawa et al., 2018). Mice were administered with MnCl<sup>2</sup> (20 mg/kg i.p.) and then returned to their home cages after 30 min, they were exposed to a novel environment (clear Perspex cylinder, 30-cm diameter) for 2 h (cylinder moved every 30 min to prevent habituation) and they were video-recorded. The mice were then returned to their home cage for 90 min before MRI scanning.

Anesthesia was induced with 3.0% isoflurane/air and maintained with 0.5–1.5% isoflurane/air; and deep core body temperature and heart rate were monitored (SA Instruments, Inc., United States) throughout the procedure. Scanning was performed 4 h after MnCl<sup>2</sup> injection using a 4.7T AVANCE III PharmaScan (Bruker BioSpin, Germany). Radio frequency transmission and reception were applied with a 23 mm inner diameter birdcage volume coil. Images were acquired with 3D Fast Imaging using a Steady-State Free Precession (FISP) sequence [repetition time (TR) = 8 ms, echo time (TE) = 4 ms, flip angle = 20◦ , number of acquisition = 7, matrix = 160 × 160 × 160, field of view (FOV) = 20 mm × 20 mm × 20 mm, and voxel size = 0.125 mm × 0.125 mm × 0.125 mm]. The total acquisition time was 31 min. MRI data were analyzed as described previously (Kimura et al., 2007), with the aid of a custom-developed MATLAB function (2012a, MathWorks). Brain slices were aligned with reference to the bregma. MR images were realigned and registered non-rigidly to the mouse brain template constructed by aligning and averaging 10 subject images. All voxel data were smoothed using a 3D Gaussian filter (MATLAB Image Processing Toolbox, version 5.02, MathWorks). Image intensities were normalized to the mean signal in the whole brain of each individual mouse. MR images were visualized with Osirix (version 5.0.2), an open-source software for navigating multidimensional DICOM images.

### Statistical Analysis

fphar-10-00081 February 7, 2019 Time: 19:37 # 4

The data represent the mean and the error bars indicate the SEM. Data were analyzed by Student's t-test and one-way analysis of variance (ANOVA) followed by Dunnett's test or the Tukey–Kramer test were performed, as indicated in the Figure legends. All statistical analyses were performed using the Ekuseru-Toukei (2012) software program (Social Survey Research Information, Tokyo, Japan). A p-value <0.05 was considered statistically significant.

### RESULTS

### Effects of Iso-α-Acids on Microglial Cytokine and Chemokine Productions

To evaluate the effects of iso-α-acids on primary microglia culture, cytokines and chemokines in the supernatant of microglial culture treated with LPS were quantified. The concentrations of IL-1β, TNF-α, IL-6, IL-12p40, MIP-1α, and MCP-1 in supernatant were significantly increased after LPS treatment. The concentrations were, however, significantly reduced by iso-α-acids addition concentration-dependently (**Figures 1A–F**, respectively). These results suggest that isoα-acids suppress inflammatory cytokines and chemokines.

### Effects of Iso-α-Acids on Cytokine Productions and Memory Impairment in Mice Treated With LPS

To evaluate the effects of iso-α-acids on cytokine productions and memory impairment in LPS-injected mice, mice inoculated with LPS were subjected to the Y-maze test. LPS treatment significantly reduced spontaneous alternation, whereas oral administration of iso-α-acids significantly improved the reduction (**Figure 2A**). There were no significant changes in arm entries among each group (**Figure 2B**). Next, we measured the levels of cytokines in the hippocampus. Concentrations of TNF-α, IL-12p40, and MIP-1α in the hippocampus of LPS-injected mice were significantly increased compared to those in untreated mice. The concentrations of TNF-α, IL-12p40 and MIP-1α significantly decreased in mice orally administered with iso-α-acids (**Figures 2D,F,G**). However, the concentrations of IL-1β, IL-6 and MCP-1 were unchanged by LPS injection (**Figures 2C,E,H**, respectively). These results showed that oral administration of iso-α-acids improved memory impairment and suppressed cytokine productions in LPS-injected mice.

### Effects of Iso-α-Acids on Inflammation and Memory Impairment in 5 × FAD Mice

To evaluate the effects of iso-α-acids on inflammation and cognitive impairment after disease onset, 5 × FAD mice were administered with 1 mg/kg iso-α-acids for 7 days. In 5 × FAD mice, long-term intake of iso-α-acids for 2.5 months resulted in the reduction of Aβ deposition and in inflammation in the brain (Ano et al., 2017). The amounts of MIP-1α and IL-12p40 in 5 × FAD mice administered with iso-α-acids was significantly lower than those in control 5 × FAD mice, and was significantly increased in control 5 × FAD mice relative to WT mice (**Figures 3B,C**). The amount of TNF-α was also increased in control 5 × FAD mice compared to that in WT mice, while the increase was not observed in 5 × FAD mice treated with iso-α-acids (**Figure 3A**). The amount of TBS-soluble Aβ in the hippocampus of 5 × FAD mice treated with isoα-acids was significantly reduced compared to that in control 5 × FAD mice (**Figure 3D**). The amounts of TBS insoluble and TBS-T soluble Aβ in the hippocampus of 5 × FAD mice treated with iso-α-acids were unchanged (**Figure 3E**). 5 × FAD mice were also subjected to a novel object recognition test. The mice treated with iso-α-acids explored a novel object for a longer time than the control 5 × FAD mice. The amounts of time spent approaching the novel object and DI of 5 × FAD mice treated with iso-α-acids were significantly higher than that of control 5 × FAD mice (**Figures 3F,G**). These results suggest that short-term treatments with iso-α-acids suppress the levels of inflammatory cytokines and soluble Aβ in the hippocampus, which lead to improve memory impairment even after the onset of Alzheimer's pathology.

### Effects of Iso-α-Acids on Inflammation and Neural Activity in J20 Mice

To evaluate the effects of a short-term treatments of iso-α-acids, 12-month-old J20 mice were administered 1 mg/kg iso-α-acids for 7 days and then subjected to MEMRI. The level of MIP-1α in the brain of the mice treated with iso-α-acids was significantly lower than that in control mice, and the level of which was significantly increased compared to that of WT mice (**Figure 4A**). The level of TBS soluble Aβ in the hippocampus of J20 mice treated with iso-α-acids was significantly reduced compared to control J20 mice (**Figure 4B**).

To evaluate the neural activity, mice were injected with MnCl2, allowed to explore a novel space, and then subjected to MRI to detect manganese. Manganese was distributed in the hippocampus and cerebral cortex (**Figure 4C**). The deposition of manganese in some areas of the hippocampus, especially the CA1and CA3 region, of control J20 mice was significantly increased compared to that of WT mice, whereas the increase of deposition was attenuated by iso-α-acids treatment (**Figure 4D**). The normalized intensity of manganese in the whole hippocampus of control J20 mice was significantly increased compared to that of WT mice, whereas an increase in J20 was returned to the level of WT mice by the treatment of iso-α-acids (**Figure 4E**). These results suggest that iso-α-acids suppress inflammation and improve the neural activity in the hippocampus.

## DISCUSSION

The present study demonstrated that short-term treatment of iso-α-acids as 7 days suppressed inflammation in the hippocampus and improved memory impairment in Alzheimer's disease model mice, even after disease onset. Notably, shortterm administration of iso-α-acids attenuate the neural

hyperactivation in hippocampus in Alzheimer's disease model mice detected by MEMRI. Iso-α-acids treatment suppressed the cytokine and chemokine levels in the hippocampus and improved memory impairment induced by LPS injection. The spontaneous alternation test used in this study evaluates spatial working memory, which is a hippocampus-dependent memory function (Lalonde, 2002). We previously reported that iso-α-acids activate the PPAR-γ (Yajima et al., 2004) and that the PPAR-γ activation is involved in the suppression of microglial inflammation (Ano et al., 2017). It has also been reported that PPAR-γ activation changes microglia phenotype to the M2 anti-inflammatory type (Pan et al., 2015; Wen et al., 2018). Pioglitazone, which is a PPAR-γ agonist, also alters microglia to the M2 type in vivo (Mandrekar-Colucci et al., 2012). These results suggest that iso-α-acids suppressed the inflammatory response via activation of PPAR-γ and suppression of microglial inflammation in the hippocampus by iso-α-acids may contribute to the prevention of memory impairment. On the other hand, the phenotype of inflammation induced by LPS, which is toll-like receptor agonist, is different from that in the brain of Alzheimer's disease.

Inflammation in the brain has attracted growing attention to a preventive and therapeutic strategy for Alzheimer's disease (Fernandez et al., 2013). Epidemiological investigation has suggested that an intake of non-steroidal anti-inflammatory drugs (NSAIDs) has a preventive effect against Alzheimer's disease (Breitner et al., 1995; Stewart et al., 1997), suggesting the therapeutic potential of pioglitazone. Microglia are known to play a crucial role in inflammation in the brain, and generally remove old synapses and waste products in the brain to maintain the environment (Kettenmann et al., 2013), yet massively activated microglia produce neurotoxic substances, including reactive oxygen species and inflammatory cytokines (Lull and Block, 2010). It has been suggested that the polarization of microglia between the M1 inflammatory and M2 antiinflammatory types is important for improving neurological pathology and cognitive decline in Alzheimer's disease (Sarlus and Heneka, 2017).

The current study is novel, as we evaluated the effects of iso-α-acids on brain inflammation in Alzheimer's disease after its onset. It was previously demonstrated that a long-term intake of iso-α-acids has a preventive effect against Alzheimer's

alternation test using the Y-maze 1 day after LPS injection. Spontaneous alternations (A) and arm entries (B) in the Y-maze test were recorded to evaluate spatial memory. (C–H) Concentrations of IL-1β, TNF-α, IL-6, IL-12p40, MIP-1α, and MCP-1 in the hippocampus were measured 24 h after LPS injection, respectively. Data are represented as mean ± SE of 10 mice per group. p-Values shown in the graph were calculated by one-way ANOVA followed by the Tukey–Kramer test. <sup>∗</sup>p < 0.05 and ∗∗p < 0.01.

pathology in 5 × FAD mice by suppressing inflammation and Aβ deposition in the brain (Ano et al., 2017). In the study, iso-α-acids were fed as a component (0.05% w/w) of the daily diet, so the daily treatment amount was not strictly controlled. In the present study, we evaluated the therapeutic effects of iso-α-acids on 6-month-old 5 × FAD mice displaying Aβ deposition, inflammation, and cognitive impairment. Treatment with iso-α-acids at 1 mg/kg for 7 days suppressed inflammation in the hippocampus and improved hippocampus-dependent object recognition memory (Cohen and Stackman, 2015). Iso-α-acids

(D) or TBS insoluble and TBS-T soluble (E) Aβ1−<sup>42</sup> in the hippocampus, respectively. (F,G) On days 6 and 7 of administration, 5 × FAD mice were subjected to the novel object recognition test. The time spent exploring novel and familiar objects during 5 min of re-exploration was measured (F). The discrimination index = [time spent with object A – time spent with object B]/[total time exploring both objects] is shown (G). Data are mean ± SE of 10 mice (wild type mice) and 8 mice (5 × FAD mice per group). p-Values were calculated by Student's t-test (D–G) and one-way ANOVA followed by the Tukey–Kramer test (A–C). <sup>∗</sup>p < 0.05 and ∗∗p < 0.01.

treatment also reduced the level of soluble Aβ1−<sup>42</sup> in the hippocampus, which is known to be elevated with inflammation (Lee et al., 2008). Hippocampus-dependent memory is known to decline with inflammation in the hippocampus (Lee et al., 2008), and it has been reported that PPAR-γ activation improves object recognition test performance in 5 × FAD mice (Escribano et al., 2009). These results suggest that suppression of inflammation in the hippocampus by iso-α-acids may contribute to the improvement in memory impairment in Alzheimer's disease model mice.

(A,B) J20 mice were orally administered with 0 or 1 mg/kg iso-α-acids for 7 days. The levels of MIP-1α (A) and TBS soluble Aβ1−<sup>42</sup> in the hippocampus was measured, respectively. (C) Coronal brain sections from wild type mice (left), J20 mice without iso-α-acids (middle), and J20 mice treated with iso-α-acids (right), all of which were 12 months old. Relative MRI signal intensities, after normalization to the mean signal intensity in the whole brain, are depicted (see color spectrum scale bar). (D) Coronal brain sections showing the differences of signal intensities J20 mice between wild type mice and J20 mice without iso-α-acids (left), and between wild type mice and J20 mice with iso-α-acids (right). (E) Normal intensities of hippocampus data are mean ± SE of seven mice (wild type mice), eight mice [IAA(–) J20 mice], and five mice [IAA(+) J20 mice]. p-Values shown in the graph were calculated by Student's t-test (B) and one-way ANOVA followed by the Tukey–Kramer test (A,E). <sup>∗</sup>p < 0.05 and ∗∗p < 0.01.

At last, to evaluate the effects of iso-α-acids on neural activity, J20 Alzheimer's disease model mice treated with isoα-acids at 1 mg/kg for 7 days were subjected to MEMRI analysis after exploration of a novel environment. Our group previously reported that 12-month-old J20 mice displayed hyperactivity detected by MEMRI in hippocampal neurons induced by exploration of a novel environment (Yoshikawa et al., 2018). J20 mice and 5 × FAD mice express human APP with mutations. J20 mice overexpresses human APP with two mutations linked to FAD (the Swedish and Indiana mutations) (Wright et al., 2013), on the other hand, 5 × FAD mice overexpress APP with mutations of Swedish K670N/M671L, Florida I716V and London V717I and PSEN1 with mutation of M146L and L286V (Oakley et al., 2006). 5 × FAD display more severe amyloid pathology including amyloid deposition and inflammation. So in the present study, we evaluate the 6 month-old 5 × FAD and 12-month-old J20 mice. Treatment with iso-α-acids reduced inflammation and soluble Aβ1−<sup>42</sup> in the hippocampi of J20 mice, which is consistent with the results of the experiments using 5 × FAD mice. Manganesedetected neural activation in the hippocampi of J20 mice was significantly higher than that in age-matched WT mice, whereas the increase was not observed in the hippocampi—especially the CA1 and CA3 region—of J20 mice treated with iso-α-acids. It is reported that hyperactivity in the hippocampal region in individuals with Mild Cognitive Impairment (MCI) is observed by functional MRI (Yassa et al., 2010). It is also reported that hyperactivity in hippocampus lead to cognitive impairment and improvement of the hyperactivity show therapeutic effects on memory impairment (Bakker et al., 2012). It has been also reported that oligomeric Aβ induces neural hyperactivity in hippocampus of Alzheimer's disease model mice (Busche et al., 2012). In hAPP transgenic mice, Aβ-induced dysfunction of inhibitory interneurons induces aberrant synchrony in neural networks (Palop and Mucke, 2010; Busche et al., 2012). Taken together, it is suggested that suppression of inflammation and the level of soluble Aβ by iso-α-acids administration in Alzheimer's disease model mice might contribute to the improvement of neural activity in the hippocampus, and resulting memory improvement, especially for the individuals with MCI. On the other hand, the relationship between suppression of hippocampal inflammation and attenuation of neural hyperactivity by iso-α-acids should be investigated in the further study.

Because iso-α-acids are generated from α-acids in hops, which has been used in brewing for more than 1,000 years, they are considered safe for consumption. The amount of iso-α-acids used in the present study was 1 mg/kg, whereas beer generally contains iso-α-acids at 20–50 µg/ml; thus, the amount of isoα-acids in approximately 1 l of beer might be equivalent to an effective dosage in humans. However, as beer also contains alcohol, it is difficult to calculate the reduced volume. Given the mounting evidence that iso-α-acids may be beneficial for cognition, a clinical trial is needed to further assess their effects in human populations.

### AUTHOR CONTRIBUTIONS

YA conducted most of biochemical analysis and wrote most of the paper. MY, MM, and AT conducted the experiment using MRI using transgenic mice and analyzed the data, and AT also wrote the manuscript. YT, KU, and HN conducted the experiment using 5 × FAD model mice, and HN also wrote the manuscript.

### REFERENCES

fphar-10-00081 February 7, 2019 Time: 19:37 # 9


### ACKNOWLEDGMENTS

We express an appreciation to Mr. Tatsuhiro Ayabe for his technical support in the experiments using LPS-inoculated mice.

cognitive impairment through enhancement of beta-amyloid generation. J. Neuroinflamm. 5:37. doi: 10.1186/1742-2094-5-37



Yoshikawa, M., Soeda, Y., Michikawa, M., Almeida, O. F. X., and Takashima, A. (2018). Tau depletion in APP transgenic mice attenuates task-related hyperactivation of the hippocampus and differentially influences locomotor activity and spatial memory. Front. Neurosci. 12:124. doi: 10.3389/fnins.2018. 00124

#### **Conflict of Interest Statement:** YA is employed by Kirin Co., Ltd.

The remaining 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 © 2019 Ano, Yoshikawa, Takaichi, Michikawa, Uchida, Nakayama and Takashima. 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.

# Bicuculline Reduces Neuroinflammation in Hippocampus and Improves Spatial Learning and Anxiety in Hyperammonemic Rats. Role of Glutamate Receptors

Michele Malaguarnera<sup>1</sup> , Marta Llansola<sup>1</sup> , Tiziano Balzano<sup>1</sup> , Belén Gómez-Giménez<sup>1</sup> , Carles Antúnez-Muñoz<sup>1</sup> , Núria Martínez-Alarcón<sup>1</sup> , Rahebeh Mahdinia1,2 and Vicente Felipo<sup>1</sup> \*

<sup>1</sup> Laboratory of Neurobiology, Centro Investigación Príncipe Felipe de Valencia, Valencia, Spain, <sup>2</sup> Faculty of Biology, Damghan University, Damghan, Iran

#### Edited by:

Morena Zusso, University of Padova, Italy

#### Reviewed by:

Georgina Rodríguez De Lores Arnaiz, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina Hiroki Toyoda, Osaka University, Japan

> \*Correspondence: Vicente Felipo vfelipo@cipf.es

#### Specialty section:

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

Received: 18 November 2018 Accepted: 05 February 2019 Published: 25 February 2019

#### Citation:

Malaguarnera M, Llansola M, Balzano T, Gómez-Giménez B, Antúnez-Muñoz C, Martínez-Alarcón N, Mahdinia R and Felipo V (2019) Bicuculline Reduces Neuroinflammation in Hippocampus and Improves Spatial Learning and Anxiety in Hyperammonemic Rats. Role of Glutamate Receptors. Front. Pharmacol. 10:132. doi: 10.3389/fphar.2019.00132 Patients with liver cirrhosis may develop minimal hepatic encephalopathy (MHE) with mild cognitive impairment. Hyperammonemia is a main contributor to cognitive impairment in MHE, which is mediated by neuroinflammation. GABAergic neurotransmission is altered in hyperammonemic rats. We hypothesized that, in hyperammonemic rats, (a) enhanced GABAergic tone would contribute to induce neuroinflammation, which would be improved by reducing GABAergic tone by chronic bicuculline treatment; (b) this would improve spatial learning and memory impairment; and (c) modulation of glutamatergic neurotransmission would mediate this cognitive improvement. The aim of this work was to assess the above hypotheses. Bicuculline was administrated intraperitoneally once a day for 4 weeks to control and hyperammonemic rats. The effects of bicuculline on microglia and astrocyte activation, IL-1β content, on membrane expression of AMPA and NMDA glutamate receptors subunits in the hippocampus and on spatial learning and memory as well as anxiety were assessed. Treatment with bicuculline reduces astrocyte activation and IL-1β but not microglia activation in the hippocampus of hyperammonemic rats. Bicuculline reverses the changes in membrane expression of AMPA receptor subunits GluA1 and GluA2 and of the NR2B (but not NR1 and NR2A) subunit of NMDA receptors. Bicuculline improves spatial learning and working memory and decreases anxiety in hyperammonemic rats. In hyperammonemia, enhanced activation of GABA<sup>A</sup> receptors in the hippocampus contributes to some but not all aspects of neuroinflammation, to altered glutamatergic neurotransmission and to impairment of spatial learning and memory as well as anxiety, all of which are reversed by reducing activation of GABA<sup>A</sup> receptors with bicuculline.

Keywords: hepatic encephalopathy, GABA<sup>A</sup> receptor, hippocampus, spatial learning and memory, anxiety, glutamate receptors, astrocytes activation, IL-1β

**Abbreviations:** AMPA, alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; BS3, Bis(sulfosuccinimidyl) suberate; cGMP, cyclic guanosine monophosphate; CNS, central nervous system; GABA, gamma-aminobutyric acid; GFAP, glial fibrillary acidic protein; GluA, AMPA receptor subunit; IL, interleukin; IL-1R, interleikin-1 receptor; LPS, lipopolysaccharide; MHE, minimal hepatic encephalopathy; NMDA, N-methyl-D-aspartic acid; NR, NMDA receptor subunit; RM, reference memory; ROI, region of interest; TNF, tumor necrosis factor; WM, working memory.

## INTRODUCTION

fphar-10-00132 February 22, 2019 Time: 16:3 # 2

Patients with liver cirrhosis may develop covert or MHE with mild cognitive impairment, attention deficits and psychomotor slowing, which impair quality of life, reduce life span and increase accidents, falls and hospitalizations. MHE affects several million people and is a serious health, social and economic problem (Felipo, 2013).

Hyperammonemia and peripheral inflammation play synergistic roles in inducing the cognitive and motor alterations in MHE (Shawcross et al., 2004; Montoliu et al., 2009; Felipo et al., 2012). These alterations would be mediated by neuroinflammation (Montoliu et al., 2015).

Chronic hyperammonemia per se is enough to induce neuroinflammation with activation of microglia and increased inflammatory markers in the brain associated with impaired cognitive function (Rodrigo et al., 2010). Reducing neuroinflammation with ibuprofen restores learning in a Y maze task in hyperammonemic rats (Rodrigo et al., 2010). Hyperammonemic rats also show neuroinflammation in the hippocampus that is associated with altered membrane expression of glutamate receptors and impaired spatial learning and memory (Cabrera-Pastor et al., 2016). These alterations are reversed by treating the rats with sulforaphane, which reduces neuroinflammation in the hippocampus (Hernández-Rabaza et al., 2016).

GABAergic neurotransmission is altered in hyperammonemic rats, which show increased GABAergic tone in the cerebellum. Chronic treatment with bicuculline, a GABA<sup>A</sup> receptor antagonist, restores GABAergic tone, the function of the glutamate-nitric oxide-cGMP pathway in the cerebellum and learning of a discrimination task modulated by this pathway (Cauli et al., 2009). The same effects are induced by treatment with pregnenolone sulfate, a negative allosteric modulator of the GABA<sup>A</sup> receptor, which also improves motor incoordination caused by increased extracellular GABA in the cerebellum (González-Usano et al., 2013).

Reducing GABAergic tone by treating rats with GR3027, which antagonizes the enhancement of GABA<sup>A</sup> receptor activation by neurosteroids, also restores spatial memory modulated mainly in the hippocampus (Johansson et al., 2015).

These reports show that reducing either GABAergic tone or neuroinflammation in the hippocampus of hyperammonemic rats improve spatial learning. This suggests that there would be a cross-talk between GABAergic tone and neuroinflammation in the modulation of the mechanisms involved in spatial learning and maybe also in other functions modulated in the hippocampus such as short-term memory or anxiety.

Recent studies support this cross-talk between GABAergic neurotransmission and neuroinflammation, which seem to modulate each other (reviewed by Crowley et al., 2016). Different pro-inflammatory cytokines, such as TNFα, IL-1β and IL-6, modulate GABA<sup>A</sup> receptor function in an area- and dose-dependent manner (Stellwagen et al., 2005; García-Oscos et al., 2012; Pribiag and Stellwagen, 2013). IL-1β suppresses GABA-induced currents in the superficial spinal cord (Kawasaki et al., 2008) in hippocampal slices (Nisticò et al., 2013) and in rat hippocampal neurons (Wang et al., 2000). Contrarily, at a different concentration, IL-1β also increases membrane expression of GABA<sup>A</sup> receptor subunits and GABAergic neurotransmission in cultured rat hippocampal neurons (Serantes et al., 2006). Hellstrom et al. (2005) showed that LPS increases GABAergic inhibition in the hippocampus through IL-1β. Additionally, reactive astrocytes release GABA, increasing GABAergic tone in cerebral ischemia (Lin et al., 2018). GABA released from reactive astrocytes impairs learning and memory (Jo et al., 2014). During neuroinflammation, GABAergic tone would increase to reduce excitotoxicity caused by excessive glutamate neurotransmission (Crowley et al., 2016).

In rats with hyperammonemia or hepatic encephalopathy, reducing neuroinflammation reverses the increase of GABAergic tone in the cerebellum and restores impaired motor coordination, suggesting enhancement of GABA neurotransmission by neuroinflammation in the cerebellum in these rats (Rodrigo et al., 2010; Dadsetan et al., 2016a; Hernández-Rabaza et al., 2016; Agusti et al., 2017). These reports show that neuroinflammation modulates GABAergic neurotransmission in different systems, including the cerebellum of hyperammonemic rats.

Gamma-aminobutyric acid neurotransmission also modulates neuroinflammation. Both anti- and pro-inflammatory effects of GABA have been reported. GABA acts as anti-inflammatory in rheumatoid arthritis, downregulating mechanisms that lead to the production of pro-inflammatory agents such as IL-1β (Kelley et al., 2008) and also in neuroinflammation in general (Crowley et al., 2016). GABA acts as anti-inflammatory in microglia through activation of GABA<sup>A</sup> receptors (Lee et al., 2011).

Other studies suggest that GABA can induce proinflammatory cytokines in pathological conditions. Carmans et al. (2013) showed that exogenous GABA increases IL-6 and TNFα mRNA in the CNS. Sallam et al. (2016) reported that intra-cerebral administration of bicuculline inhibited the increase of IL-6 and TNFα induced by LPS in rats. Increased GABA levels and the subsequent activation of GABA<sup>A</sup> receptors induce activation of astrocytes (Runquist and Alonso, 2003).

The mechanisms by which neuroinflammation impairs spatial learning in hyperammonemic rats involve altered membrane expression of AMPA and NMDA receptor subunits. Treatment with sulforaphane reverses changes in membrane expression of the receptors and restores spatial learning (Hernández-Rabaza et al., 2016).

It has not been analyzed whether modulating GABAergic neurotransmission could reduce neuroinflammation in the hippocampus of hyperammonemic rats and restore membrane expression of glutamate receptors and cognitive functions modulated by this area.

We have proposed that there is an interplay between neuroinflammation and GABAergic-glutamatergic neurotransmission in the induction of cognitive and motor alterations in rats with MHE (Agusti et al., 2017). We propose now that a similar interplay in the hippocampus of hyperammonemic rats would induce alterations in spatial learning and is likely in other functions modulated in the hippocampus, such as anxiety or short-term memory.

We hypothesized that, in hyperammonemic rats, (a) enhanced GABAergic tone would contribute to induce neuroinflammation in hippocampus; (b) reducing GABAergic tone by chronic treatment with bicuculline would reduce neuroinflammation; (c) this would be associated with normalization of membrane expression of AMPA receptor subunits and restoration of spatial learning; and (d) bicuculline treatment could also improve other functions modulated in the hippocampus such as short-termmemory and anxiety.

To test these hypotheses, we assessed whether chronic intraperitoneal administration of the GABA<sup>A</sup> receptor antagonist bicuculline in hyperammonemic rats modulates neuroinflammation in the hippocampus by analyzing the activation of microglia, astrocytes, and IL-1β content. We also assessed the effects on membrane expression of AMPA and NMDA receptor subunits and on spatial learning and memory, short-term memory and anxiety.

### MATERIALS AND METHODS

### Study Design, Chronic Hyperammonemia in Rats, and Treatment With Bicuculline

Male Wistar rats (120–140 g, Charles River Laboratories, Barcelona, Spain) were made hyperammonemic by feeding them an ammonium-containing diet as previously described (Felipo et al., 1988). Animals were distributed into four groups: control with vehicle (CV); control treated with bicuculline (CB); hyperammonemic rats (HA); hyperammonemic rats treated with bicuculline (HB). Bicuculline [(+)-Bicuculline, Sigma-Aldrich] was injected intraperitoneally at 0.3 mg/kg once per day. The dose was chosen based on a previous study that shows that hyperammonemia changes GABAergic tone in the CNS (Cauli et al., 2009). This dose is lower than that which induces seizures (>1 mg/kg i.p.) (Meldrum et al., 1987; Giardina, 2000; Martín del Campo et al., 2009). In addition, we did not observe any signs of seizures in any of the injected rats. Bicuculline was dissolved in physiological serum (NaCl 0.9%) with 0.3% DMSO, and this solution was used as vehicle.

The experiment was replicated four times using 36 animals (9 rats per group) each time. A total of 144 rats were used, with 36 rats per group. Not all animals performed behavioral tests because we found that the data obtained with two replicates were enough to reach statistical significance. Concerning the analysis of neuroinflammation and membrane expression of proteins, in each replicate four rats per group were perfused for immunohistochemistry studies while the other five rats per group were used for analysis of membrane surface expression and content of the proteins. The experimental design is summarized in **Figure 1**. The experiments were approved by the Comité de Ética y Bienestar en Experimentación Animal, Prince Felipe Research Center-Consellería de Agricultura, Generalitat Valenciana and carried out in accordance with the Directive of the European Commission (2010/63/EU) for care and management of experimental animals.

### Brain Immunohistochemistry

At week 6 of hyperammonemia, the rats were anesthetized with sodium pentobarbital and transcardially perfused with 0.9% saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were removed and post-fixed in the same fixative solution for 24 h at 4◦C. Paraffin-embedded sections (5 µm) were cut and mounted on coated slide glass. The tissue sections were then processed with the Envision Flex+kit (DAKO) blocking endogenous peroxidase activity for 5 min and then incubated with antibodies. Primary antibodies were against Iba-1 (Wako 019-19741; 1:300 for 30 min), GFAP (Dako IR524; ready for use for 20 min) and IL-1β (Abcam AB9722; 1:100 dilution for 30 min). The reaction was visualized by incubation with Envision Flex + horseradish peroxidase for 20 min and finally diaminobenzidine for 10 min. Sections were counterstained with Mayer's hematoxylin for 5 min.

### Analysis of Microglia Activation

Analysis of Iba-1-stained microglia was performed in the hippocampus using Image J software. Microglia activation was assessed by measuring the cell perimeter in eight randomly selected areas (0.45 mm<sup>2</sup> ) per section according to Vinet et al. (2012). The area of interest was selected. Using Auto Local Threshold and Analyze particle functions in ImageJ, the intensity thresholds and size filter were applied. To measure the perimeter of microglia, the Bernsen method was used and a 2000– 20,000 size filter was applied. For each rat, at least 30–40 cells were quantified, and the results were converted from pixels to micrometers. The perimeter length for each group is expressed as the percentage of values for control rats.

### Analysis of Astrocytes Activation

Astroglial area (µm<sup>2</sup> ) covered by GFAP was measured using Image J software. Using Auto Local Threshold and Analyze Particles functions, the intensity thresholds and size filter were applied. To measure the total cell size, the Bernsen method (Bernsen, 1986) was used and a 1500–7000 size filter was applied. For each rat, at least 20 cells from three different sections were counted.

### Analysis of IL1β Expression

IL1β-positive cells were manually counted by two blinded experimenters and the results (the mean of two blind experimenters) were expressed as a percentage of the total number of cells. For each rat, at least 120–150 cells per section were counted from at least four different sections. The intensity of IL1β in CA1 region was quantified using the ROI manager function in ImageJ. The CA1 region was selected manually. Inverted values of Mean Gray value were recorded and results were expressed as a percentage of control group.

### Analysis of Membrane Surface Expression of Receptors

This analysis was performed by cross-linking with B53 as described by Cabrera-Pastor et al. (2016). Hippocampi were dissected and transversal slices (400 µm) were

obtained using a chopper. Slices were added to tubes containing ice-cold standard buffer with or without 2 mM BS3 (Pierce, Rockford, IL, United States) and incubated for 30 min at 4◦C. Cross-linking was terminated by adding 100 mM glycine (10 min, 4◦C). The slices were homogenized by sonicating for 20 s. Samples treated or not with BS3 were analyzed by western blot. The membrane surface expression of each receptor was calculated as the difference between the intensity of the bands without BS3 (total protein) and with BS3 (non-membrane protein) as described by Cabrera-Pastor et al. (2016).

### Analysis of Protein Content in Hippocampus by Western Blot

Homogenates of the hippocampus were subjected to immunoblotting according to Felipo et al. (1993). Primary antibodies were against IL-1β 1:500 dilution (AF-510-NA) from R&D SYSTEMS (Minneapolis, MN, United States); GluA1, GluA2, NR2A, and NR2B 1:1000 dilution (cat.# 04-855, AB1768, 04-901 and 06-600, respectively) from Merck Millipore (Darmstadt, Germany); and NR1, 1:1000 dilution (cat.# 556308) from BD Biosciences (San Jose, CA, United States). As a control for protein loading, the same membranes were also incubated with anti-actin (1:5000) from Abcam (Cambridge, MA, United States). Secondary antibodies were anti-rabbit, anti-goat or anti-mouse IgG, 1:4000 dilution (cat.# A8025, A7650, A3562, respectively) conjugated with alkaline phosphatase from Sigma (St. Louis, MO, United States). The images were captured using the ScanJet 5300C (Hewlett-Packard, Amsterdam, Netherlands) and band intensities quantified using the A Imager 2200, version 3.1.2 (AInnotech Corporation, San Francisco, CA, United States).

### Spatial Learning in the 8-Arms Radial Maze

It was assessed as described in Hernández-Rabaza et al. (2016). Training was performed over 4 days (three trials per day). The task involved locating four pellets, each placed at the end of a different arm according to a random configuration. Configurations were specific for each rat and were kept invariable throughout training. The number of spatial reference errors (reference memory errors, visits to unbaited arms) and WM errors (WM errors, visits to arms already visited in the same trial) were calculated. Learning index is defined as the difference between the number of right choices and reference errors as in Hernández-Rabaza et al. (2016).

### Short-Term Spatial Recognition Memory

It was analyzed using a Y-maze consisting of three arms made of black metacrilate joined in the middle to form a "Y" shape. This test is based on the rodents' innate curiosity to explore novel areas and presents no negative or positive reinforcement and very little stress for the rats. The protocol is a modification of the test used by Sarnyai et al. (2000) and Sanderson et al. (2009). The rat was handled for 1 min to reduce stress and anxiety, placed into one of the arms of the maze (start arm) and allowed to explore the maze with one of the arms closed for 2 min (training trial) for three times. After 1 min of inter-trial interval, the rat returned to the Y maze by placing it in the start arm. Then, the rat was allowed to explore freely all three arms of the maze for 2 min (test trial). The number of entries into and the time spent in each arm, the first choice of entry and the discrimination ratio [(Time spent in the novel arm – Time spent in the familiar arm)/Total time passed in the two arms] were registered. Because entry into the novel arm could be altered by anxiety, exploration time was also recorded. No significant differences in exploration time were observed between groups (results not shown).

### Analysis of Anxiety Using a Hole-Board Test

This test was performed in a plastic floor with 16 equidistant holes of 4 cm in diameter. The floor was positioned in an open-field activity chamber (43 cm × 43 cm × 30.5 cm) (Med Associates, St Albans, VT, United States), where the animals were positioned. The rat's activity was detected by arrays of infrared motion detection, with two arrays 1 cm above the floor of the chamber and another array 6 cm above the floor, so the upper line of infra-red cells detected the animal movement and the bottom line detected the head-dipping. For the hole-board experiments, each animal was placed in the center of the hole-board and allowed to freely explore the apparatus for 5 min (Kong et al., 2006). Total number of head-dipping and latency to the first head-dipping were recorded by the software Activity Monitor (provided by MED Associates, Inc., St Albans, VT, United States). Anxiety leads to decreased explorative activity that can be quantified by counting the frequency of head-dipping in the holes. The larger is the frequency of head-dipping and the lower is the anxiety of the rat. Novel head dips were also measured to quantify anxiety.

### Statistical Analysis

Data are expressed as mean ± SEM. All statistical analyses were performed using the software program GraphPad Prism

7.0 (GraphPad Prism Software, Inc.). Statistical analysis was carried out using one-way ANOVA or two-way ANOVA with repeated measures, followed by Tukey's post hoc test, as indicated in the figure legends. A confidence level of 95% was accepted as significant.

### RESULTS

Hyperammonemic rats show activated microglia in the hippocampus, with a less ramified and more ameboid morphology reflected in a reduction (p < 0.05) of the perimeter to 280 ± 8 µm compared to 315 ± 9 µm in control rats (**Figure 2**). Activation of microglia is not prevented by bicuculline. The perimeter of microglia in hyperammonemic rats treated with bicuculline was 265 ± 7 µm, similar to untreated hyperammonemic rats. In contrast, bicuculline treatment induced microglia activation in control rats, with reduction of the perimeter to 284 ± 7 µm [F(3,68) = 12,76] (**Figure 2**).

Hyperammonemic rats also showed astrocyte activation in the hippocampus, as indicated by the increased area stained by anti-GFAP (19.9 ± 0.4% of the area, p < 0.01) compared to control rats (17.7 ± 0.2% of the area). Bicuculline decreased astrocyte activation in hyperammonemic rats and the area stained by anti-GFAP returned to normal values (17.8 ± 0.5% of the area, p < 0.01 vs. hyperammonemic rats), indicating de-activation of astrocytes [F(3,56) = 6,068] (**Figure 3**).

Neuroinflammation was also reflected in increased IL1β content in hippocampus as analyzed by immunohistochemistry. Anti-IL1β stained mainly neurons of the CA1 region. The number of cells expressing IL1β increased (p < 0.05) in hyperammonemic rats to 120 ± 8% of controls. Treatment with bicuculline reversed the increase in IL1β to 97 ± 12% of controls (p < 0.05, compared with untreated hyperammonemic rats). In contrast, as occurs for microglial activation, bicuculline increased (p < 0.05) the number of cells expressing IL1β in control rats to 118 ± 3% of untreated controls [F(3,42) = 5,384] (**Figures 4A,B**).

IL1β was also quantified by western blot, which confirmed an increase of IL1β levels in the hippocampus of hyperammonemic rats to 159 ± 2% of control rats. This increase was reversed by bicuculline treatment, which reduces IL1β levels to 97 ± 12% of controls (**Figure 4C**). Western blot analysis also confirmed the increase (p < 0.05) of IL1β in control rats treated with bicuculline, to 189 ± 32% of untreated controls [F(3,30) = 5,427] (**Figure 4C**).

We have proposed that neuroinflammation in the hippocampus leads to impairment of spatial learning and

memory by altering membrane expression of GluA1 and GluA2 subunits of AMPA receptors (Cabrera-Pastor et al., 2016). We therefore assessed if treatment with bicuculline normalizes membrane expression of GluA1 and GluA2 subunits of AMPA receptors and spatial learning and memory in hyperammonemic rats.

At 5 weeks of hyperammonemia, membrane expression of GluA1 was significantly reduced (p < 0.01) to 65 ± 6% of control rats (**Figure 5A**). Treatment with bicuculline reversed the decrease in membrane expression of GluA1 (p < 0.01 compared with hyperammonemic rats), returning to 96 ± 11% of control rats. In control rats, treatment with bicuculline induced a slight reduction (85 ± 6% of control rats) in membrane expression of GluA1 [F(3,47) = 5,667] (**Figure 5A**).

Membrane expression of GluA2 was increased (p < 0.01) in hyperammonemic rats to 188 ± 25% of control rats (**Figure 5B**). Bicuculline completely reversed the increase of GluA2 (p < 0.01, compared with untreated hyperammonemic rats), returning to levels similar to controls (97 ± 12% of controls). In control rats, bicuculline increased membrane expression of GluA2 (208 ± 21% of controls, p < 0.001) [F(3,47) = 10,74] (**Figure 5B**).

We also assessed the effects on the NR1 and NR2 subunits of NMDA receptors. Membrane expression of NR1 was significantly reduced [p < 0.001; F(3,58) = 7,592] in hyperammonemic rats to 50 ± 7% of controls (**Figure 5C**). Membrane expression of NR2A was increased [p < 0.05; F(3,49) = 3,752] to 161 ± 21% of control rats (**Figure 5D**), and that of NR2B to 158 ± 11% of control rats [p < 0.01; F(3,53) = 6,066] (**Figure 5E**). Treatment with bicuculline in hyperammonemic rats did not reverse the decrease of NR1 nor the increase in NR2A, but completely normalized (p < 0.01) the membrane expression of NR2B, returning it to levels similar to control rats (**Figures 5C–E**).

In control rats, bicuculline reduced (p < 0.05) membrane expression of NR1 to 66 ± 11% of controls and increased membrane expression of NR2A (173 ± 21% of controls, p < 0.05) (**Figures 5C,D**).

Spatial learning and memory were assessed at 4– 5 weeks of hyperammonemia in the radial and Y mazes. Hyperammonemia impaired spatial learning in the radial maze. The learning index was lower (p < 0.05) than for control rats. Bicuculline reversed the impairment of spatial learning in hyperammonemic rats. The learning index was higher than in untreated hyperammonemic rats [two-way ANOVA Repeated Measure: group by Learning Index. F(3,44) = 3,684, p < 0.01] and not different from control rats (**Figure 6A**).

The number of working errors was also higher in hyperammonemic than in control rats [two-way ANOVA: group by treatment interaction F(3,33) = 3,706, p < 0.05] (**Figure 6B**). Hyperammonemic rats performed a total of 19 ± 2 working errors [F(3,33) = 3.7], which is more (p < 0.05) than for control rats (10 ± 2 working errors). Treatment with bicuculline reversed the impairment of WM. Hyperammonemic rats treated with bicuculline performed 15 ± 2 working errors, not significantly different from control rats. In control rats, bicuculline increased the number of working errors (15 ± 2), but the difference with untreated controls was not statistically significant (**Figure 6C**).

We also assessed short-term memory in the Y maze. The discrimination ratio was reduced (p < 0.05) in hyperammonemic rats to 0.47 ± 0.05, compared with control rats (0.70 ± 0.05) [F(3,43) = 2,6]. Treatment with bicuculline partially restored short-term memory. The discrimination ratio (0.55 ± 0.08) was higher than in untreated hyperammonemic rats (**Figure 6D**).

As anxiety is also modulated by the hippocampus, we also assessed it in hyperammonemic rats, which show a decreased number of head dips [39 ± 5, p < 0.05, F(3,25) = 4,548] compared to control rats (55 ± 5), indicative of anxiety behavior. Hyperammonemic rats treated with bicuculline showed a tendency to increase the number of head dips, reaching 47 ± 5, which is not significantly different from control rats (**Figure 7A**).

Control rats treated with bicuculline showed a similar number of head dips (37 ± 2, p < 0.05) than hyperammonemic rats (**Figure 7A**), which is not surprising since bicuculline has a well-known anxiogenic effect.

We also analyzed the number of novel head dips, i.e., the number of head dips in unexplored holes. This number was reduced (p < 0.01) in hyperammonemic rats (13.0 ± 0.5) compared to control rats (15.3 ± 0.2) and was restored by treatment with bicuculline [15.1 ± 0.5; F(3,25) = 5,482] (**Figure 7B**). In control rats, treatment with bicuculline reduced the number of novel head dips (14.0 ± 0.3), indicating altered exploratory behavior.

### DISCUSSION

We have analyzed the effects of blocking GABA<sup>A</sup> receptors with bicuculline on three indicators of neuroinflammation in the hippocampus of hyperammonemic rats: activation of microglia and of astrocytes and content of IL1β. Blocking GABA<sup>A</sup> receptors reduces astrocyte activation and IL1β levels, but not microglia activation in the hippocampus of hyperammonemic rats.

GABA<sup>A</sup> receptors are expressed in astrocytes (MacVicar et al., 1989; Bureau et al., 1995) and their expression seems to be increased in reactive astrocytes (Hösli et al., 1997).

Moreover, it has been shown in primary cultures that GABA released by neurons or added exogenously triggers morphological changes in astrocytes through the activation of GABA<sup>A</sup> receptors. The effect was blocked by bicuculline (Matsutani and Yamamoto, 1997). A similar effect of GABA on astrocytes morphology through activation of GABA<sup>A</sup> receptors has been shown in vivo in neonatal rat hypothalamus (Mong et al., 2002).

The increase of GABAergic signaling in vivo, by using an inhibitor of GABA transaminase or intracerebral infusion of the GABA-A agonist muscimol, induced activation of astrocytes, with increased branching and GFAP content (Runquist and Alonso, 2003).

The above reports suggest that increasing GABA levels and/or activation of GABA<sup>A</sup> receptors is enough to induce activation of astrocytes. As we show that bicuculline treatment reverses activation of astrocytes in the hippocampus of hyperammonemic rats, it is likely that this de-activation of astrocytes would be due directly to the block of GABA<sup>A</sup> receptors in astrocytes by bicuculline.

Bicuculline-induced de-activation of astrocytes is associated with a reduction of IL-1β in the hippocampus of hyperammonemic rats. However, IL-1β expression in hyperammonemic rats is increased in neurons of the CA1 region, not in astrocytes, in agreement with previous reports in rats with hyperammonemia and hepatic encephalopathy (Hernandez-Rabaza et al., 2015; Dadsetan et al., 2016b; Balzano et al., 2019). Therefore, the normalization of IL-1β levels by bicuculline would not occur in astrocytes.

It was reported that IL-1β and TNF-a are expressed in hippocampal neurons in vivo in response to lesions (Tchélingérian et al., 1996) or to pneumococcal meningitis (Izadpanah et al., 2014). In situ hybridization studies show that in murine pneumococcal meningitis IL-1β and TNF-a mRNA were first upregulated in astroglial cells but at 18–24 h were strongly increased in hippocampal neurons (Izadpanah et al., 2014). A similar process occurs in the hippocampus of rats with hepatic encephalopathy, leading to increased expression of IL-1β and TNF-a in neurons (Dadsetan et al., 2016b). It seems therefore that neuronal expression of IL-1β occurs in different pathological situations and would be triggered by previous expression in astrocytes.

These reports suggest therefore that increased expression of IL-1β in CA1 neurons of the hippocampus in hyperammonemic rats could be a consequence of previous activation of astrocytes. Bicuculline would reverse activation of astrocytes and this would lead to a reduced expression of IL-1β in neurons in hyperammonemic rats. Alternatively, other direct effects of blocking GABA<sup>A</sup> receptors with bicuculline could be also involved. For example, it has been reported that treatment with bicuculline methiodide, a form of the GABA<sup>A</sup> receptor blocker that does not cross the blood–brain barrier, decreases IL-1β in blood in a rat model of sepsis (Hsu and Liu, 2004).

In contrast to astrocyte activation and over-expression of IL-1β, which are reversed by bicuculline in hyperammonemic rats, microglial activation is not reversed. This suggests that hyperammonemic rats show increased GABAergic tone in the hippocampus, which contributes to activation of astrocytes and to over-expression of IL-1β but not to activation of microglia.

It has already been shown that hyperammonemic rats show increased GABAergic tone in the cerebellum but not in the cerebral cortex (Cauli et al., 2009). Increased GABAergic tone in the hippocampus is supported by a report from Johansson et al. (2015) showing that reducing GABAergic tone with GR3027, which antagonizes GABA<sup>A</sup> receptor potentiating neurosteroids, restores spatial learning, modulated by the hippocampus, in hyperammonemc rats.

In contrast with the effects of bicuculline in hyperammonemic rats, bicuculline increases neuroinflammation in control rats, inducing microglial activation and increasing IL-1β. It has been proposed that GABA acts as an anti-inflammatory in rheumatoid arthritis downregulating mechanisms that lead to the production of pro-inflammatory agents such as IL-1β (Kelley et al., 2008) and also in neuroinflammation in general (Crowley et al., 2016). GABA acts as an anti-inflammatory in microglia through activation of GABA<sup>A</sup> receptors (Lee et al., 2011). In control rats, treatment with bicuculline would prevent this anti-inflammatory effect of GABA resulting in activation of microglia and enhanced production of IL-1β.

microglia activation and increased IL1β in the hippocampus. This is associated with increased membrane expression of GluA2 and NR2A, a tendency to increase for NR2B and reduced membrane expression of NR1. This is associated with increased anxiety and a tendency to impair WM.

In hyperammonemic rats, normalization of IL-1β levels and astrocyte activation in hippocampus by bicuculline is associated with normalization of membrane expression of the GluA1 and GluA2 subunits of AMPA receptors and of the NR2B subunit of NMDA receptors as well as with improvement of spatial learning. These effects would be a direct consequence of the reduction of IL-1β levels. Taoro-Gonzalez et al. (2018) have shown that in the hippocampus of hyperammonemic rats, increased IL-1β levels over-activate the IL-1 receptor, leading to activation of Src and increased membrane expression of NR2B and GluA2 in addition to reduced membrane expression of GluA1. Blocking IL-1 receptor with the endogenous antagonist IL-1Ra reverses all these changes. This indicates that the effects of bicuculline on GluA1, GluA2, and NR2B are due to the reduction of IL-1β. In contrast, this does not normalize membrane expression of NR1 and NR2A subunits of NMDA receptors, which are not modulated by IL-1β under these conditions (Lai et al., 2006).

It has been reported that altered membrane expression of AMPA receptors in the hippocampus is responsible for impairment of spatial learning and WM in hyperammonemic rats and that reversing these changes by different treatments (sulforaphane, extracellular cGMP) restores spatial learning

(Cabrera-Pastor et al., 2016; Hernández-Rabaza et al., 2016). Normalization of membrane expression of AMPA receptors following normalization of IL-1β by bicuculline treatment would be therefore responsible for improvement of spatial learning and WM in hyperammonemic rats.

Our approach does not allow us to discern if membrane expression of AMPA and NMDA receptors is altered in glial cells or in neurons. However, we can assume that the alterations occur in neurons since, in contrast with other cerebral regions, astrocytes in the hippocampus apparently lack AMPA receptors, and expression of NMDA receptors in glial cells in the hippocampus remains controversial (Rose et al., 2018). Moreover, there is clear evidence of the role of neuronal AMPA and NMDA receptors in hippocampal synaptic plasticity and spatial learning (Morris and Frey, 1997; Shapiro, 2001; Baez et al., 2018).

A direct effect of the blockade of GABA<sup>A</sup> receptors by bicuculline on improvement of spatial learning cannot be disregarded. Several studies report beneficial effects on spatial learning and memory of intra-CA1 or intraseptal injection of bicuculline (Roland and Savage, 2009; Torkaman-Boutorabi et al., 2013; Yousefi et al., 2013). Intraperitoneal injection of bicuculline also improves WM in aged ovariectomized rats treated with progesterone (Braden et al., 2015) and in control or chronically stressed rats (Nishimura et al., 2017). On the other hand, bicuculline infusion in the prefrontal cortex impairs working and reference memory (Auger and Floresco, 2017). This, together with increased IL-1β, could be a possible cause of the tendency to increase working errors (not significantly) in control rats treated with bicuculline.

The NR1 subunit of NMDA receptors seems to play a relevant role in short-term memory. The dentate gyrus NR1 knockout mice exhibit a selective impairment in short-term spatial memory (Niewoehner et al., 2007). Treatment with bicuculline did not restore membrane expression of NR1 nor short-term memory in hyperammonemic rats. In both cases, there was a slight tendency to improve that did not reach statistical significance. This suggests that altered membrane expression of NR1 in the hippocampus may contribute to decreased short-term spatial memory in hyperammonemic rats.

Anxiety is common in hyperammonemic cirrhotic patients and decreases quality of life (Nardelli et al., 2013; Malaguarnera et al., 2018). Hyperammonemic rats also show increased anxiety related to neuroinflammation (Luo et al., 2014). Anxiety is modulated by the hippocampus (Nasehi et al., 2011) and by NMDA and GABA receptors (Bina et al., 2014). Injection of NMDA in the ventral hippocampus induces anxiolytic effects, which are prevented by bicuculline, indicating a cross-talk between both neurotransmitter systems in the regulation of anxiety (Bina et al., 2014). In apparent contrast with this report, Barkus et al. (2010) reported that antagonists of NMDA receptors or deletion of the NR1 subunit of NMDA receptors in ventral hippocampus also have an anxiolytic effects.

We found that hyperammonemic rats show increased anxiety and reduced membrane expression of NR1 in the hippocampus. However, treatment with bicuculline reversed anxiety but not membrane expression of NR1, indicating that this would not be directly responsible for anxiety in hyperammonemci rats.

Genetically modified mice lacking the NR2B subunit in hippocampal granule and pyramidal cells in dentate gyrus and CA1, display reduced anxiety (von Engelhardt et al., 2008). As NR2B is increased in membranes in the hippocampus of hyperammonemic rats and treatment with bicuculline reverses this increase, it is possible that the changes in NR2B could contribute to anxiety in hyperammonemic rats and to its reversal by bicuculline.

As summarized in **Figure 8**, this report shows that in hyperammonemic rats, enhanced basal activation of GABA<sup>A</sup> receptors in the hippocampus contributes to activation of astrocytes and increased IL-1β, but not to increased activation of microglia. Blocking GABA<sup>A</sup> receptor with bicuculline reduces astrocyte activation and IL-1β content in hyperammonemic rats. Normalization of IL-1β levels by bicuculline is associated with reversal of the changes in membrane expression of the GluA1 and GluA2 subunits of AMPA receptors and of the NR2B subunit of NMDA receptors, but not of the NR1 and NR2A subunits of NMDA receptors. Normalization of GluA1, GluA2, and NR2B would be a consequence of normalization of IL-1β and would be responsible for restoration of spatial learning and WM. NR2B would also contribute to reverse the increased anxiety in hyperammonemic rats.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

### AUTHOR CONTRIBUTIONS

MM contributed to most of the experiments including treatment of rats and obtaining samples, immunohistochemical studies and analysis, and membrane expression of receptors, performed the radial maze and anxiety tests, analyzed and interpreted the data, and drafted the manuscript. ML supervised the study, analyzed and interpreted the data, and wrote the manuscript. TB, CA-M, and RM contributed to immunohistochemical studies and analysis. BG-G contributed to treatment of rats, obtaining samples, and radial maze. NM-A contributed to membrane expression of receptors, and immunohistochemical studies and analysis. VF conceived, designed, and supervised the study, obtained funding, analyzed and interpreted the data, and wrote the manuscript.

### FUNDING

Supported in part by Ministerio de Ciencia e Innovación of Spain (SAF2014-51851-R and SAF2017-82917-R), Consellería Educación Generalitat Valenciana (PROMETEOII/2014/033), co-funded with European Regional Development Funds (ERDF). MM was supported by the Ministry of Economy and Competitiveness of Spain through a Juan de la Cierva postdoctoral fellowship (FJCI-2015-25825).

### REFERENCES

fphar-10-00132 February 22, 2019 Time: 16:3 # 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 © 2019 Malaguarnera, Llansola, Balzano, Gómez-Giménez, Antúnez-Muñoz, Martínez-Alarcón, Mahdinia and Felipo. 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.

# TRPM2 Channel in Microglia as a New Player in Neuroinflammation Associated With a Spectrum of Central Nervous System Pathologies

Philippa Malko<sup>1</sup> , Sharifah A. Syed Mortadza1,2, Joseph McWilliam<sup>1</sup> and Lin-Hua Jiang<sup>1</sup> \*

<sup>1</sup> School of Biomedical Sciences, Faculty of Biological Sciences, University of Leeds, Leeds, United Kingdom, <sup>2</sup> Department of Biochemistry, Universiti Putra Malaysia, Seri Kembangan, Malaysia

#### Edited by:

Morena Zusso, University of Padova, Italy

#### Reviewed by:

Hisashi Shirakawa, Kyoto University, Japan Marta Llansola, Centro de Investigación Príncipe Felipe, Spain

> \*Correspondence: Lin-Hua Jiang l.h.jiang@leeds.ac.uk

#### Specialty section:

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

Received: 25 January 2019 Accepted: 26 February 2019 Published: 12 March 2019

#### Citation:

Malko P, Syed Mortadza SA, McWilliam J and Jiang L-H (2019) TRPM2 Channel in Microglia as a New Player in Neuroinflammation Associated With a Spectrum of Central Nervous System Pathologies. Front. Pharmacol. 10:239. doi: 10.3389/fphar.2019.00239 Microglial cells in the central nervous system (CNS) are crucial in maintaining a healthy environment for neurons to function properly. However, aberrant microglial cell activation can lead to excessive generation of neurotoxic proinflammatory mediators and neuroinflammation, which represents a contributing factor in a wide spectrum of CNS pathologies, including ischemic stroke, traumatic brain damage, Alzheimer's disease, Parkinson's disease, multiple sclerosis, psychiatric disorders, autism spectrum disorders, and chronic neuropathic pain. Oxidative stress is a salient and common feature of these conditions and has been strongly implicated in microglial cell activation and neuroinflammation. The transient receptor potential melastatin-related 2 (TRPM2) channel, an oxidative stress-sensitive calcium-permeable cationic channel, is highly expressed in microglial cells. In this review, we examine the recent studies that provide evidence to support an important role for the TRPM2 channel, particularly TRPM2-mediated Ca2<sup>+</sup> signaling, in mediating microglial cell activation, generation of proinflammatory mediators and neuroinflammation, which are of relevance to CNS pathologies. These findings lead to a growing interest in the TRPM2 channel, a new player in neuroinflammation, as a novel therapeutic target for CNS diseases.

Keywords: TRPM2 channel, microglial cell activation, CNS pathologies, neuroinflammation, proinflammatory mediators

## INTRODUCTION

The central nervous system (CNS), which is composed of the brain and spinal cord, is a highly integrated and complex network made up principally by neuronal and glial cells. Neuronal cells or neurons as the working unit of the CNS are specialized to transmit information. Glial cells function more in a supportive capacity to surrounding neurons and, nonetheless, as has been

increasingly recognized, also actively participate in many functional aspects of the CNS through bi-directional and dynamic interactions (Jäkel and Dimou, 2017; Allen and Lyons, 2018; Luca et al., 2018). There are several types of glial cells with different embryonic origins (Menassa and Gomez-Nicola, 2018). Astrocytes, oligodendrocytes, and neural-glial antigen 2-positive cells are derived from neuro-ectoderm that also gives rise to neurons, whereas microglial cells are myeloidlineage cells originated from mesoderm that generates cells of the blood and immune system. Therefore, microglial cells are privileged to be the immune-competent cells of the CNS, like macrophages in the systemic immune system, and thus are often referred to as CNS-resident macrophages. Under healthy or steady-state conditions, microglial cells exhibit a distinctive morphology characteristic of high ramification with an extensive network of fine processes stemming from a small cell body and a resting phenotype (Saijo and Glass, 2011). Microglial cells can secret neurotrophic factors [e.g., brain-derived neurotrophic factor (BDNF)] and, using their phagocytic capability, eliminate excessive or dysfunctional synapses and clear apoptotic developing neurons. In this way, microglial cells support neuronal functions, particularly important processes such as neurogenesis and synaptogenesis during brain development and in the adult brain (Marin-Teva et al., 2004; Sierra et al., 2010; Kettenmann et al., 2013; Yirmiya et al., 2015; Kierdorf and Prinz, 2017; Ising and Heneka, 2018; Luca et al., 2018). In addition, microglial cells act as the sentinel of the CNS and unceasingly patrol the surroundings with their fine processes to monitor environmental changes and provide the first defensive mechanism in response to damage and infection. Microglial cells express a repertoire of the so-called pattern recognition receptors (PRRs), with Tolllike receptors (TLRs), and nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs) being two example groups. PRRs detect danger-associated molecular patterns (DAMPs) released from host cells due to damage or stress or pathogenassociated molecular patterns (PAMPs) generated by invading pathogens (Brubaker et al., 2015; Jassam et al., 2017). Upon ligation of PRRs by DAMPs and/or PAMPs, microglial cells become activated and, after retracting their processes and taking on a spherical form, adopt an amoeboid morphology, proliferate and migrate to the site of damage or infection, where they remove damaged cells or pathogens via phagocytosis (Hanisch and Kettenmann, 2007). Microglial cells can generate proinflammatory mediators that are instrumental in heightening acute immune responses, including chemokines [e.g., C-X-C motif ligand 2 (CXCL2)], cytokines [e.g., interleukin (IL)- 1β, tumor necrosis factor (TNF)-α, IL-6], nitric oxide (NO), and reactive oxygen species (ROS). Activated microglial cells can also assume distinctive and anti-inflammatory phenotypes and produce anti-inflammatory cytokines and neurotrophic factors [e.g., IL-10, tissue growth factor (TGF)-β and BDNF] that are important in resolving inflammation and stimulating tissue repair (Wang et al., 2015; Tay et al., 2017; Luca et al., 2018). It is increasingly clear that microglial cells exhibit a high level of heterogeneity in the developing brain and an increase in varied proinflammatory subtypes in the aged, inflamed or neurodegenerative brain (Hammond et al., 2018; Sousa et al., 2018).

It is known that numerous DAMPs are released by cells in the CNS as a result of aging, traumatic damage, chronic psychological stress or neurodegenerative diseases, with ATP being one such well-documented example (Jassam et al., 2017; Wei et al., 2018). It is also well-known that DAMPs are released from degenerating neurons in the brain, such as misfolded amyloid β-peptides (Aβ), α-synuclein, and superoxide dismutase 1 (Glass et al., 2010). These DAMPs are potent inducers of chronical activation or senescence of microglial cells, leading to elevated generation of pro-inflammatory mediators that alters neuronal functions and induces neurotoxicity, a process often referred to as neuroinflammation (Glass et al., 2010; Heneka et al., 2018; Luca et al., 2018). Studies over the past decade have gathered a large body of evidence to support that microglial cells play a key role in mediating neuroinflammation as a significant contributing factor in the progression of aging and a wide spectrum of CNS conditions, including ischemic stroke, traumatic brain damage, Alzheimer's disease (AD), Parkinson's disease (AD), multiple sclerosis (MS), amyotrophic lateral sclerosis (ALS), neuropsychiatric disorders [e.g., depression, bipolar disorder (BD), and schizophrenia], autism spectrum disorders (ASD), and neuropathic pain (Glass et al., 2010; Yirmiya et al., 2015; Du et al., 2017; Inoue, 2017; Jassam et al., 2017; Maiti et al., 2017; Ramirez et al., 2017; Salter and Stevens, 2017; Alibhai et al., 2018; Bodnar et al., 2018; Ising and Heneka, 2018; Luca et al., 2018; Shetty et al., 2018; Szepesi et al., 2018; Voet et al., 2018).

Oxidative stress, resulting from excessive ROS generation, impaired antioxidant capacity, or both, is a common and salient feature in aging and the aforementioned CNS diseases. The transient receptor potential melastatinrelated 2 (TRPM2) channel is a Ca2+-permeable cationic channel with a high sensitivity to oxidative stress or ROS (Hara et al., 2002; Zhang et al., 2003) and is a member of the large transient receptor potential (TRP) superfamily (Clapham, 2003). In the systemic immune system, the TRPM2 channel has been recognized as an important molecular mechanism mediating DAMP/PAMP-induced generation of proinflammatory mediators and innate immune responses (Knowles et al., 2013; Syed Mortadza et al., 2015). Expression of the TRPM2 channel is widely distributed in the CNS with a high level in microglial cells. In this article, we focus on the TRPM2 channel in microglial cells and its role in neuroinflammation. We start with a brief introduction of the TRPM2 channel activation followed by a summary of the evidence supporting TRPM2 channel expression in microglial cells. We proceed to describe the studies that show an important role of the TRPM2 channel in microglial cell activation and generation of proinflammatory mediators in response to various DAMPs and PAMPs, and also the current understanding regarding the molecular mechanisms responsible for DAMP/PAMP-induced TRPM2 channel activation and the downstream TRPM2-dependent signaling pathways engaged in microglial cell activation and generation of proinflammatory mediators. We also discuss

the studies using rodent models that demonstrate the role of the TRPM2 channel in microglial cell activation and neuroinflammation in CNS diseases. Finally, we highlight the gaps in our understanding that require further investigation in order to test whether targeting the TRPM2 channel, a new player in neuroinflammation, could represent a neuroprotective approach to tempering the progression of aging or CNS diseases.

### TRPM2 CHANNEL ACTIVATION

Up to now, it has been established both functionally and structurally that the TRPM2 channel is a ligand-gated Ca2+ permeable cationic channel activated by intracellular ADP-ribose (ADPR), and that ADPR-induced TRPM2 channel activation displays strong dependence of intracellular Ca2<sup>+</sup> (**Figure 1A**) (Perraud et al., 2001; McHugh et al., 2003; Mei et al., 2006; Tong et al., 2006; Xia et al., 2008; Du et al., 2009; Tóth and Csanády, 2010; Huang et al., 2018; Wang et al., 2018; Zhang et al., 2018). Several ADPR analogs, including ADPR-2 0 -phosphate, 2<sup>0</sup> -O-acetyl-ADPR and 2<sup>0</sup> -deoxy-ADPR, have been shown to gate the TRPM2 channel (**Figure 1A**) (Grubisha et al., 2006; Toth et al., 2015; Fliegert et al., 2017). Cyclic ADPR (cADPR), nicotinamide adenine dinucleotide (NAD) and other structurally or metabolically ADPR-related compounds were also reported in earlier studies using whole-cell recording to activate the TRPM2 channel (Sano et al., 2001; Kolisek et al., 2005; Beck et al., 2006; Togashi et al., 2006). This notion however has been challenged by more recent studies using the excised inside-out recording to show that application of these compounds to the intracellular face of the TRPM2 channel failed to induce TRPM2 channel activation (Tóth and Csanády, 2010; Toth et al., 2015).

It is also known that warm temperature (≥35◦C) induces TRPM2 channel activation alone or in synergy with other TRPM2 channel activators (**Figure 1A**), as shown in pancreatic β-cells and macrophages (Togashi et al., 2006; Kashio et al., 2012; Kashio and Tominaga, 2015). In this aspect, recent studies have revealed an important role for the TRPM2 channel in sensory neurons in the peripheral and central nervous systems in detecting non-noxious warmth and regulating body temperature (Song et al., 2016; Tan and McNaughton, 2016).

As introduced above, TRPM2 channels display high sensitivity to activation under oxidative stress or more specifically exposure to elevated levels of ROS, thus gaining increasing recognition for their role in mediating cellular responses to oxidative stress (Jiang et al., 2010; Miller and Zhang, 2011; Takahashi et al., 2011; Knowles et al., 2013; Ru and Yao, 2014; Li et al., 2015, 2017; Syed Mortadza et al., 2015; Yamamoto and Shimizu, 2016). While some earlier studies suggested that ROS such as H2O<sup>2</sup> may directly activate the TRPM2 channel, it is now widely accepted that ROS-induced TRPM2 channel activation is indirect and depends on mechanisms that promote an increase in intracellular ADPR level (Jiang et al., 2010). One widelyemployed mechanism in many types of mammalian cells is generation of ADPR from NAD by poly(ADPR)-polymerase (PARP), particularly PARP-1, and poly(ADPR)-glycohydrolase (PARG) in the nucleus (**Figure 1B**). Some evidence exists to suggest that ADPR generation from NAD catalyzed by NADase in the mitochondria also contributes in ROS-induced TRPM2 channel activation (Perraud et al., 2005).

### TRPM2 CHANNEL EXPRESSION IN MICROGLIAL CELLS

Studies examined TRPM2 channel expression in microglial cells at the mRNA, protein and/or functional levels using reverse transcription-polymerase chain reaction (RT-PCR), immunostaining, western blotting, Ca2<sup>+</sup> imaging and/or patchclamp current recording (Kraft et al., 2004; Fonfria et al., 2006; Lee et al., 2010; Jeong et al., 2017; Syed Mortadza et al., 2017). Kraft et al. (2004) were the first to examine TRPM2 channel expression in cultured rat microglial cells. A high level of TRPM2 mRNA expression was detected, and exposure to H2O<sup>2</sup> induced extracellular Ca2<sup>+</sup> influx, leading to an increase in intracellular Ca2<sup>+</sup> concentration ([Ca2+]i). Furthermore, application of intracellular ADPR opened a cationic conductance with a linear current-voltage (I-V) relationship and a single channel conductance of ∼65 pS (Kraft et al., 2004), the key biophysical characteristics of the TRPM2 channels (Jiang et al., 2010). A recent study shows strong TRPM2 mRNA and protein expression and ADPRinduced cationic currents in cultured mouse microglial cells (Jeong et al., 2017). Consistently, exposure to H2O<sup>2</sup> (10– 300 µM) induced concentration-dependent Ca2<sup>+</sup> influx and increase in [Ca2+]<sup>i</sup> in cultured mouse microglial cells from wild-type (WT) but not TRPM2-knockout (TRPM2-KO) mice (Syed Mortadza et al., 2017). Profiling the TRPM2 mRNA level in numerous human tissues, including the brain and spinal cord, revealed abundant expression and a wide distribution of TRPM2 expression in the CNS (Fonfria et al., 2006). In C13, a human microglial cell line, TRPM2 mRNA transcripts were also readily detected, and exposure to H2O<sup>2</sup> induced a robust increase in [Ca2+]<sup>i</sup> . Both the mRNA expression level and H2O2-induced Ca2<sup>+</sup> responses were reduced in C13 cells after treatment with TRPM2-specific antisense oligomers (Fonfria et al., 2006). Furthermore, application of intracellular ADPR or extracellular H2O<sup>2</sup> elicited cationic currents that exhibited an almost linear I-V relationship and a strong sensitivity to inhibition by flufenamic acid (FFA), a TRPM2 channel inhibitor (**Figure 1A**). In cultured human microglial cells isolated from surgically resected temporal lobe tissues, exposure to H2O<sup>2</sup> elicited a strong increase in [Ca2+]<sup>i</sup> that was inhibited by treatment with clotrimazole (Lee et al., 2010), a TRPM2 channel inhibitor (**Figure 1A**). These studies have gathered compelling evidence to support TRPM2 channel expression in human and rodent microglial cells as a Ca2<sup>+</sup> influx pathway with a significant role in ROS-induced Ca2<sup>+</sup> signaling.

Interestingly, an earlier study noted that there were significantly greater H2O2-induced Ca2<sup>+</sup> responses and

more readily detectable H2O2-induced currents in cultured rat microglial cells after exposure to H2O<sup>2</sup> or treatment with lipopolysaccharide (LPS), an endotoxin found in the outer membrane of Gram-negative bacteria and a widely-used PAMP to induce immune cell activation via TLR4 activation (Kraft et al., 2004). TRPM2 mRNA expression was up-regulated in C13 cells after treatment with IL-1β (Fonfria et al., 2006). As we discuss below, exposure to diverse pathological stimuli or conditions can increase TRPM2 channel expression in microglial cells.

reactive oxygen species; NAD, nicotinamide adenine dinucleotide; pADPR, poly(ADPR); PARP1, poly(ADPR)-polymerase 1; PARG, poly(ADPR)-glycohydrolase.

### TRPM2 CHANNEL IN MICROGLIAL CELL ACTIVATION AND GENERATION OF PROINFLAMMATORY MEDIATORS

An increasing number of studies have, mainly using cultured microglial cells, investigated the role of the TRPM2 channel in microglial cell activation and generation of proinflammatory mediators in response to diverse pathological stimuli. Furthermore, as discussed in detail next, efforts have been made to gain considerable insights into the mechanisms by which the TRPM2 channel is activated by such stimuli and the downstream TRPM2-dependent signaling pathways in microglial cell activation and generation of proinflammatory mediators (**Figure 2**).

### Generation of TNF-α and IL-6 Resulting From Impaired Antioxidant Capacity

Glutathione (GSH) is present as one of the important reducing agents in most mammalian tissues including the CNS that equip cells with a non-enzymatic antioxidant capacity (Meister and Anderson, 1983). Glutamatecysteine ligase (or γ-glutamylcysteine synthase) is a rate-limiting step in GSH synthesis and thus D,L-buthionine-S,R-sulfoximine (BSO), an inhibitor of glutamatecysteine ligase, can cause depletion of intracellular GSH and cellular oxidative stress. It has been proposed that a reduction in intracellular GSH with aging increases age-related susceptibility to oxidative stress, which is worsened in many neurodegenerative conditions (Sohal and Weindruch, 1996). A previous study investigated the role of the TRPM2 channel in generating neurotoxic proinflammatory mediators in cultured human microglial cells under BSO-induced oxidative stress (Lee et al., 2010). Exposure to BSO (1–24 h) induced an exposure durationdependent increase in [Ca2+]<sup>i</sup> . Exposure to BSO for 2 h was sufficient to activate mitogen-activated protein kinases (MAPK), p38, extracellular signal-regulated kinase (ERK) and Jun-Nterminal kinase (JNK), and furthermore downstream nuclear factor NF-κB. BSO-induced increase in [Ca2+]<sup>i</sup> and activation of MAPK and NF-κB signaling pathways were significantly suppressed by supplementation with GSH or treatment with clotrimazole. Exposure to BSO (0.1, 0.5 and 1 mM) also induced concentration-dependent release of TNF-α and IL-6 from microglial cells, which was reduced by treatment with TRPM2-specific small interference RNA (siRNA) (Lee et al., 2010). These results suggest that oxidative stress resulting from GSH depletion activates the TRPM2 channel and TRPM2 mediated Ca2<sup>+</sup> influx in turn initiates downstream MAPK and NF-kB signaling pathways, leading to generation of TNF-α and IL-6 (**Figure 2A**). Human neuroblastoma SH-SY5Y cells cultured in the medium conditioned by BSOtreated microglial cells exhibited substantial cell death (Lee et al., 2010). Such cell death was significantly attenuated in the conditioned culture medium that was prior depleted of TNF-α and IL-6. Moreover, SH-SY5Y cell death in the conditioned culture medium was strongly suppressed by supplementing microglial cell culture medium with GSH or treating microglial cells with clotrimazole or TRPM2-siRNA (Lee et al., 2010). Collectively, these results suggest that TNF-α and IL-6, generated by microglial cells in a TRPM2 dependent manner, under BSO-induced oxidative stress can induce neurotoxicity.

### LPC-Induced Microglial Cell Activation

It is known that lysophosphatidylcholine (LPC), an inflammatory phospholipid endogenously generated under physiological and various pathological conditions, can induce extracellular Ca2<sup>+</sup> influx in microglial cells and microglial cell activation (Schilling et al., 2004; Sheikh et al., 2009). A recent study has investigated the role of the TRPM2 channel in LPC-induced Ca2+-signaling and microglial cell activation in cultured mouse microglial cells (Jeong et al., 2017). Exposure to LPC induced cationic currents as well as an extracellular Ca2+-dependent increase in [Ca2+]<sup>i</sup> . LPC exposure also resulted in phosphorylation of p38 (p-p38), an indicator of microglial cell activation. Consistently, intrathecal injection of LPC enhanced expression of ionized calcium binding adapter molecule 1 (Iba1) and CD11 in spinal microglial cells, suggesting microglial cell activation (Jeong et al., 2017). Such LPC-induced in vitro or in vivo effects in microglial cells were largely prevented by TRPM2-KO (Jeong et al., 2017). These results support a key role for the TRPM2 channel in LPC-induced Ca2<sup>+</sup> signaling and activation of downstream p38 MAPK signaling pathways, leading to microglial cell activation (Jeong et al., 2017) (**Figure 2B**). It remains unclear regarding the mechanisms by which LPC induces TRPM2 channel activation, and the types of proinflammatory mediators that are generated as a result of LPC-induced microglial cell activation. This study has made an interesting observation that the levels of both total and cell surface TRPM2 protein expression was significantly increased in LPC-treated microglial cells but it is not elucidated how such up-regulation of TRPM2 expression and membrane trafficking occurs.

### LPS/IFNγ-Induced Activation of iNOS and Generation of NO

The TRPM2 channel was shown, in an in vivo study discussed below, to play a significant role in mediating spinal microglial cell activation and neuropathic pain (Haraguchi et al., 2012). In this study the authors particularly revealed a role for the TRPM2 channel in cultured microglial cells in the activation of inducible NO synthase (iNOS) and generation of NO after exposure to LPS and IFNγ. A subsequent study by the same group investigated the signaling pathways engaged in LPS/IFNγ-induced TRPM2 channel activation and NO generation (Miyake et al., 2014). LPS/IFNγ exposure evoked extracellular Ca2<sup>+</sup> influx to increase [Ca2+]<sup>i</sup> , which was prevented by TRPM2-KO or treatment with miconazole, a TRPM2 channel inhibitor (**Figure 1A**). Such Ca2<sup>+</sup> response was also efficiently inhibited by treatment with diphenylene iodonium (DPI) and ML-171, inhibitors of nicotinamide adenine dinucleotide phosphate (NADPH)-dependent oxidases (NOXs). LPS/IFNγ-induced NO generation was

protein 3.

fphar-10-00239 March 8, 2019 Time: 17:22 # 6

synthase; PKC, protein kinase C; PARP-1, poly(ADPR) polymerase 1; TLR4; Toll-like receptor 4; NLRP3, nucleotide binding domain-containing leucine-rich repeat

also significantly reduced by TRPM2-KO, or by inclusion of 1,2-bis(o-aminophenoxy)ethane-N,N,N,N-tetraacetic acid (BAPTA), a Ca2<sup>+</sup> chelator, to remove extracellular Ca2+. These results support that LPS/IFNγ induce NOX-mediated ROS generation, TRPM2 channel activation and an increase in [Ca2+]i, leading to NO generation (**Figure 2C**). Moreover, LPS/IFNγ-induced NO generation was attenuated by treatment with AG17, an inhibitor for Ca2+-sensitive proline-rich tyrosine kinase 2 (PYK2), SB203580, a p38 inhibitor, or SP600125, a JNK inhibitor. Inhibition of LPS/IFNγ-induced NO generation by BAPTA, AG17, SB203580 or SP600125 was abolished by TRPM2-KO. LPS/IFNγ-induced NO generation in microglial cells from both WT and TRPM2-KO mice was attenuated by treatment with PD98059, a MEK/ERK inhibitor. Likewise, exposure to LPS/IFNγ induced selective activation of p38 in WT but not TRPM2-KO microglial cells, but indiscriminate activation of ERK in both WT and TRPM2-KO microglial cells. Overall, these results suggest that LPS/IFNγ-induced TRPM2-mediated Ca2<sup>+</sup> signaling initiates activation of PYK2 and downstream p38/JNK MAPK

signaling pathways for activation of iNOS and subsequent NO generation (**Figure 2C**).

### Aβ42-Induced Microglial Cell Activation and Generation of TNF-α

A recent in vivo study using the APP/PS1 mouse model of AD, as discussed further below, has disclosed an important role of the TRPM2 channel in Aβ-induced AD pathologies, including microglial cell activation (Ostapchenko et al., 2015). It is wellestablished that TNF-α contributes to AD and neurodegenerative diseases via direct interaction with its death receptor on neurons as well as induction of microglial cell activation to generate additional neurotoxic mediators (Alam et al., 2016; Jiang et al., 2018). Our recent study has explored the molecular mechanisms responsible for TRPM2 channel activation and TNF-α generation in cultured mouse microglial cells induced by exposure to Aβ42, one of the amyloid-β peptides of high relevance to AD (Syed Mortadza et al., 2018). Exposure to Aβ42 (30–300 nM) induced a concentration-dependent and extracellular Ca2+-dependent increase in [Ca2+]<sup>i</sup> . Aβ42-induced

Ca2<sup>+</sup> response was strongly suppressed by treatment with 2-APB, a TRPM2 channel inhibitor (**Figure 1**), or BAPTA-AM as a membrane-permeable and thus intracellular Ca2<sup>+</sup> chelator, and furthermore by TRPM2-KO. Exposure to Aβ42 induced cellular ROS generation and activation of nuclear PARP-1. Both Aβ42-induced PARP-1 activation and increase in [Ca2+]<sup>i</sup> were suppressed by treatment with PJ34, an inhibitor of PARP enzymes including PARP-1. Furthermore, Aβ42-induced ROS generation, PARP-1 activation and Ca2<sup>+</sup> responses were inhibited by treatment with chelerythrine, a protein kinase C (PKC) inhibitor, GKT137831, a NOX1/4-seletive inhibitor, or Phox-I2, a NOX2 inhibitor as well as the NOX generic inhibitor DPI. These results indicate that Aβ42 activates the TRPM2 channel by inducing PKC/NOX-mediated ROS generation and subsequent PARP-1 activation and generation of ADPR (**Figure 2D**). Aβ42-induced PARP-1 activation and increase in [Ca2+]<sup>i</sup> were also prevented by treatment with PF431396, a PYK2 inhibitor, or U0126, a MEK/ERK inhibitor. Aβ42-induced PARP-1 activation was significantly reduced but incompletely abolished by TRPM2-KO, and the remaining Aβ42-induced PARP-1 activity in TRPM2-KO microglial cells was prevented by treatment with GKT137831 or Phox-I2 and, in striking contrast, not altered by treatment with PF431396 or U0126. Taken together, these results suggest that Aβ42 stimulates PKC/NOX-mediated ROS generation and PARP-1 activation leading to initial TRPM2 channel activation, and that subsequent TRPM2-mediated Ca2<sup>+</sup> flux and activation of PYK2, MEK/ERK, and PARP-1 serves as a positive feedback mechanism for further TRPM2 channel activation (**Figure 2D**). Moreover, exposure to Aβ42 induced noticeable morphological changes in microglial cells and an increase in the expression and release of TNF-α. Aβ42-induced morphological changes and TNF-α generation were prevented by TRPM2-KO and, moreover, by pharmacological inhibition of the aforementioned signaling pathways responsible for TRPM2 channel activation (Syed Mortadza et al., 2018).

### Aβ42-Induced Activation of NLRP3 Inflammasome and Generation of IL-1β

The nucleotide binding domain-containing leucine-rich repeat protein 3 (NLRP3) is a member of the NOD family of PRRs in the cytosol. In response to damage or infection, NLRP3, apoptosis-associated speck-like protein containing a caspase recruitment domain, and procaspase-1, via proteinprotein interactions, assemble a multi-protein complex often termed as the NLRP3 inflammasome. NLRP3 inflammasome activation is required to activate caspase-1, which in turn cleaves pro-IL-1β into IL-1β (Tschopp and Schroder, 2010; Brubaker et al., 2015; Jassam et al., 2017; Song et al., 2017; White et al., 2017). It was shown that genetic inactivation of the NLRP3 inflammasome in APP/PS1 mice reduced IL-1β generation by microglial cells, leading to improved spatial memory and attenuation of other AD-related pathological phenotypes (Heneka et al., 2013). In addition, NLRP3 inflammasome inactivation shifted microglial cells toward an anti-inflammatory phenotype that cleared Aβ peptides, thereby resulting in a reduction in amyloid-β deposition (Heneka et al., 2013). Therefore, neuroinflammation resulting from NLPR3 inflammasome activation and IL-1β generation in microglial cells has emerged as an important factor contributing to AD pathogenesis, inciting an interest in targeting the NLRP3 inflammasome as a therapeutic approach to AD (Heneka et al., 2014; White et al., 2017). It is well-known that NLRP3 inflammasome activation and IL-1β generation in immune cells including microglial cells exhibit a striking convergence on ROS generation (Tschopp and Schroder, 2010; Song et al., 2017). A recent pharmacological study has examined the potential role of the TRPM2 channel in Aβ42-induced NLRP3 inflammasome activation and IL-1β generation in LPS-primed microglial cells (Aminzadeh et al., 2018). Exposure to Aβ42 at a relatively high concentration (10 µM) induced mitochondrial ROS generation and also IL-1β generation, both of which were suppressed by treatment with DPI at a high concentration (20 µM) that presumably targets mitochondrial ROS generation. Aβ42-induced IL-1β generation was inhibited by treatment with VAS2870 or (2R,4R)-4-aminopyrrolidine-2,4-dicarboxylate, NOX inhibitors, indicating engagement of NOX-meditated ROS generation. In addition, Aβ42-induced IL-1β generation was reduced by treatment with N-acetylcysteine, a ROS scavenger, or DPQ, a PARP-1 inhibitor (Aminzadeh et al., 2018). Exposure to Aβ42 resulted in a Ca2<sup>+</sup> influx-dependent increase in [Ca2+]<sup>i</sup> that was also strongly inhibited by treatment with DPI, VAS2870, DPQ, or BAPTA-AM. Finally, Aβ42-induced caspase-1 activation, as shown by western blotting, was inhibited by treatment with DPQ or BAPTA-AM (Aminzadeh et al., 2018). These results are consistent with the notion that Aβ42 induces NLRP3 inflammasome activation and IL-1β generation via stimulating mitochondrial and NOX-mediated ROS generation, activation of PARP-1 and the TRPM2 channel, and subsequent TRPM2-mediated Ca2<sup>+</sup> influx (**Figure 2E**). However, more and definitive evidence is required to corroborate the proposed role of the TRPM2 channel.

### TRPM2 CHANNEL IN NEUROINFLAMMATION AND CNS PATHOLOGIES

It is clear from the above discussion that studies based on cultured microglial cells support an important role for the TRPM2 channel in microglial cell activation and generation of neurotoxic proinflammatory mediators in response to DAMPs/PAMPs of high relevance to various CNS diseases. As discussed next, there is increasing evidence from in vivo studies using rodent models that supports a critical role for the TRPM2 channel in microglial cells in microglial cell activation, generation of proinflammatory mediators and neuroinflammation in the pathogenesis of CNS diseases (**Figure 3**).

### Neuropathic Pain

It is well-recognized that microglial cell activation in the spinal cord, as well as peripheral neuroinflammation, plays

been shown to contribute to the pathogenesis of neuropathic pain, brain damage by chronic cerebral hypo-perfusion and neonatal hypoxia-ischemia, and Aβ-induced AD. CNS, central nervous system; Aβ, amyloid β peptides; AD, Alzheimer's disease.

a significant role in the development of chronic neuropathic pain (Ji and Suter, 2007; Costigan et al., 2009; Tsuda et al., 2013). The role of the TRPM2 channel in mediating spinal microglial cell activation and neuropathic pain was explored in a previous study using two mouse models of neuropathic pain induced by partial sciatic nerve ligation (SNL) and spinal nerve transection (SNT), respectively (Haraguchi et al., 2012). Both mechanical allodynia and thermal hyperalgesia observed in WT mice during the 2 weeks following SNL were largely absent in TRPM2-KO mice. In the sciatic nerves on the ligation site, the TRPM2 mRNA level was markedly elevated. SNL induced a significant increase in the number of neutrophils and also in generation of CXCL2 in WT mice, both of which were mitigated or completely prevented in TRPM2-KO mice. SNL also resulted in a strong increase in the TRPM2 mRNA expression in dorsal microglial cells and in the intensity of immunoreactivity for both Iba1 and CD11b. Moreover, there was an increase in p-p38 and strong co-localization of p-p38 and CD11b, further indicating microglial cell activation. SNLinduced increase in the intensity of immunoreactivity for Iba1, CD11b nd p-p38, and co-localization of p-p38 and CD11b in spinal microglial cells was largely prevented by TRPM2- KO. Similarly, SNT induced mechanical allodynia, increased intensity of immunoreactivity to CD11b and p-p38 and their co-localization in dorsal microglial cells, all of which were significantly subdued in TRPM2-KO mice. The study further examined the role of the TRPM2 channel in the generation of proinflammatory mediators in cultured microglial cells exposed to LPS/IFNγ. LPS/IFNγ stimulated CXCL2 generation and NO release as well as an increase in the mRNA expression of CXCL2, TNF-α, IL-1β, IL-6, and iNOS. LPS/IFNγ-induced generation of CXCL2 and NO, and increased mRNA expression of CXCL2 and iNOS were significantly lowered by TRPM2-KO (Haraguchi et al., 2012). However, the study revealed no significant effect of TRPM2-KO on the mRNA expression of TNF-α, IL-1β and IL-6, indicating engagement of TRPM2-independent mechanisms. These results support the notion that the TRPM2 channel in spinal microglial cells contributes to neuropathic pain by mediating the generation of proinflammatory mediators to aggravate pro-nociceptive inflammatory responses. As discussed above, LPS/IFNγ-induced NO generation depends on TRPM2 mediated Ca2<sup>+</sup> signaling and activation of downstream PYK2 and MAPK p38 and JNK signaling pathways (**Figure 2C**).

### Alzheimer's Disease

Alzheimer's disease is an age-related neurodegenerative disease with increasing prevalence in a rapidly aging society, representing the most common cause of dementia that afflicts tens of millions of older people worldwide. Aβ accumulation is widely thought to be an early and pathogenic event in AD pathogenesis. Oxidative damage is a conspicuous but mechanistically poorly understood feature of AD. As has been recently reviewed (Jiang et al., 2018), studies have shown wide expression of the TRPM2 channel in the brain and strong evidence for the TRPM2 channel as a nexus from Aβ generation and oxidative damage to AD pathologies via multiple cellular and molecular mechanisms, including microglial cell activation. Microglial cells are known to have a dual role in AD (Boche and Nicoll, 2008). They provide a protective role by phagocytic clearance of Aβ, but such a beneficial capacity declines with aging and is overwhelmed by excessive toxic aggregates, becoming inefficient. As introduced above, Aβ can induce chronic activation and senescence of microglial cells leading to excessive generation of ROS and numerous neurotoxic proinflammatory cytokines, such as TNF-α, IL-1β and IL-6, which constitutes a critical component of AD pathogenesis. APP/PS1 mice co-express a chimeric mouse/human amyloid precursor protein (APP) with the Swedish mutations (K670N and M671L) and human presenilin 1 (PS1) with deletion of exon 9 (Jankowsky et al., 2003). A recent study has examined the role of the TRPM2 channel in Aβ-induced AD pathogenesis using this mouse AD model (Ostapchenko et al., 2015). As has been well-documented, the APP/PS1 mice exhibit excessive Aβ generation, amyloid deposits and synaptic loss in the hippocampus and cortex, microglial cell activation, and severe impairment in age-related spatial memory. Genetic deletion of TRPM2 expression in APP/PS1 mice, while resulting in no alteration in amyloid deposition, essentially reversed Aβ-induced synaptic loss, microglial cell activation, and memory impairment (Ostapchenko et al., 2015). These results provide compelling evidence to support a critical role for the TRPM2 channel in Aβ-induced AD-related pathologies. As already discussed above, recent in vitro studies reveal an important role of the TRPM2 channel in mediating Aβ42-induced microglial cell activation and generation of

TNF-α (Syed Mortadza et al., 2018) and possibly IL-1β (Aminzadeh et al., 2018).

### Brain Damage by Deficient Cerebral Blood Circulation

The brain is well-known for its vulnerability to damage by deprivation or restriction of oxygen and/or glucose supply that can occur under conditions such as cerebral ischemic stroke, cardiac arrest, chronic cerebral hypo-perfusion, and neonatal hypoxia-ischemia. Oxidative stress, mainly due to increased ROS generation, is a common characteristic of these conditions. An early study demonstrated elevated TRPM2 mRNA expression in rat brains at 1 and 4 weeks after transient middle cerebral artery occlusion (MCAO), a widely-used rodent model of ischemic stroke (Fonfria et al., 2006). A number of recent studies, using various in vitro and in vivo mouse models of ischemia-reperfusion in conjunction with pharmacological inhibition or genetic deletion of the TRPM2 channel, have supported a critical role of the TRPM2 channel in ischemiareperfusion brain damage and associated cognitive dysfunction (Jia et al., 2011; Alim et al., 2013; Shimizu et al., 2013, 2016; Gelderblom et al., 2014; Ye et al., 2014). There is also emerging evidence to indicate a role for the TRPM2 channel in mediating brain damage due to hypoxia-ischemia in neonates (Huang et al., 2017) and chronic cerebral hypo-perfusion in adults (Miyanohara et al., 2018).

Of notice, much of the research in this area has so far been devoted to the TRPM2 channel in mediating neuronal death. Nonetheless, there is increasing evidence to suggest a significant contribution of TRPM2-mediated neuroinflammation. For example, selective deletion of the TRPM2 expression in peripheral immune cells substantially protected infarction and cognitive impairment in mice after transient MCAO and reperfusion (Gelderblom et al., 2014). However, the role of the TRPM2 channel in microglial cells in ischemia-reperfusion brain damage largely remains unclear. In the case of neonatal hypoxia-ischemia, a recent study shows that infarction in postnatal day 7 pups, induced by ligating the right common carotid artery and reducing oxygen supply and examined 24 h or 7 days afterward, was considerably attenuated in heterozygous and homozygous TRPM2-KO pups. In addition, WT pups exhibited sensorimotor dysfunction at 7 days post hypoxia-ischemia, and such deficits were less noticeable in heterozygous and homozygous TRPM2-KO pups. In WT pups, the TRPM2 mRNA expression was greater in the damaged hemisphere than the healthy hemisphere. Hypoxia-ischemia induced a massive increase in the number of glial fibrillary acidic protein (GFAP) positive cells and Iba1-positive cells in WT pups, but not in heterozygous and homozygous TRPM2-KO pups. These results suggest that the TRPM2 channel plays an important role in mediating activation of glial cells, including microglial cells, thereby inducing neonatal hypoxic-ischemic brain damage. The role of the TRPM2 channel in microglial cells has been best understood in brain damage by chronic cerebral hypoperfusion (Miyanohara et al., 2018). Mice manifested significant white matter damage and cognitive dysfunction 28 days after introduction of bilateral common carotid artery stenosis (BCAS), a model of chronic cerebral hypo-perfusion. At this time point, there was also a significant increase in the TRPM2 mRNA expression and in the level of IL-1β, TNF-α and IL-6 in the corpus callosum. Such BCAS-induced effects, namely, white matter damage, cognitive dysfunction and increased generation of IL-1β, TNF-α and IL-6, were prevented by TRPM2-KO. There was an increase in the number of GFAP positive cells and Iba-1 positive cells in the corpus callosum at 14 and 28 days after BCAS, but only the number of Iba-1 positive cells was strongly suppressed by TRPM2-KO. The increase in the number of Iba1-positive cells and cognitive dysfunction in BCAS-operated mice was effectively prevented by administration of minocycline, an inhibitor of microglial cell and macrophage activation. Further analysis, using WT and TRPM2-KO mice with bone marrow (BM)-derived cells replaced by WT GFP-labeled BM-derived cells, indicates that the Iba-1 positive cells in white matter mainly are largely microglial cells. Collectively, these results therefore support a critical role for the TRPM2 channel in mediating microglial cell activation and generation of proinflammatory cytokines, IL-1β, TNF-α and IL-6, in the aggravation of cognitive impairment by chronic cerebral hypo-perfusion.

### SUMMARY AND PERSPECTIVES

In summary, the TRPM2 channel is highly expressed in microglial cells and mainly functions as a plasma membrane Ca2+-permeable cationic channel with a key role in mediating ROS-induced Ca2<sup>+</sup> signaling (**Figure 1B**). In addition, the TRPM2 channel in microglial cells is potently activated by diverse DAMPs and/or PAMPs that induce mitochondrial and/or NOX-mediated ROS generation, activation of PARP-1 and ADPR generation (**Figure 2**). Studies using rodent models in combination with pharmacological and genetic interventions support a significant role for the TRPM2 channel in microglial cell activation and neuroinflammation in the pathogenesis of various CNS conditions. Currently, this includes neuropathic pain, chronic cerebral hypo-perfusion brain damage, neonatal hypoxia-ischemia and Aβ-induced AD (**Figure 3**). As mentioned in the introduction, microglial cell-mediated neuroinflammation is a well-recognized factor in the pathogenesis of many other CNS conditions besides the aforementioned conditions. Research has also implicated TRPM2 channel in PD (Sun et al., 2018; An et al., 2019; Li and Jiang, 2019), MS (Tsutsui et al., 2018), traumatic brain damage (Cook et al., 2010; Yürüker et al., 2015), and neurodevelopmental disorders such as ASD (Higashida et al., 2018) and depression (Xu et al., 2006; Jang et al., 2015; Zhong et al., 2016; Ko et al., 2019) as well as ischemic stroke brain damage. Evidently, further research is required to investigate whether the TRPM2 channel in microglial cells in mediating neuroinflammation plays a significant role in these CNS conditions.

As discussed above, recent studies have gained significant insights into the molecular mechanisms by which DAMPs and/or PAMPs induce activation of the TRPM2 channel and generation of diverse proinflammatory mediators that are of strong relevance

to various CNS diseases. It is clear from the discussion that the current understanding remains fragmented with better insights in some cases than others (**Figure 2**). Further research is required to provide a coherent understanding of how the TRPM2 channel is activated in response to distinctive stimuli or under different conditions, leading to activation of downstream Ca2<sup>+</sup> signaling pathways, and ultimately how such TRPM2-dependent signaling pathways drive microglial cell activation and generation of proinflammatory mediators.

Given the widespread indication of a significant role for the TRPM2 channel in mediating neuroinflammation and CNS diseases, the TRPM2 channel represents an attractive therapeutic target. The TRPM2 channel also plays important roles in a number of physiological processes, such as insulin release from pancreatic β-cells, regulation of temperature sensation, and peripheral immune responses, which may complicate the concept of targeting TRPM2 as a therapeutic strategy. However, TRPM2 channel expression in the CNS is selectively up-regulated by diverse pathological stimuli or diseased conditions. With continual research into TRPM2 modulation and function in specific cell types, future developments may focus on pharmacological agents that can improve the outcome for patients with

### REFERENCES


CNS diseases while sparing the physiological functions of the channel. Targeting the TRPM2 channel in microglial cells, a newly-emerged player in neuroinflammation, represents an interesting a venue of development of promising therapeutics.

### AUTHOR CONTRIBUTIONS

L-HJ and PM wrote the manuscript. All the authors contributed to literature research and analysis, developed the review topic, and approved the manuscript.

### FUNDING

This research was supported in part by the research grants from the Wellcome Trust (072275/Z/03/Z), National Natural Science Foundation of China (31471118), and Alzheimer's Research Trust (ART/PPG2009A/2) (L-HJ), a Ph.D. studentship from the Faculty of Biological Sciences of University of Leeds (PM), a Ph.D. scholarship from the Government of Malaysia (SSM), and a Laidlaw Scholarship (JM).

experimental traumatic brain injury in rats. J. Mol. Neurosci. 42, 192–199. doi: 10.1007/s12031-010-9347-8


susceptibility to cell death. Mol. Cell 9, 163–173. doi: 10.1016/S1097-2765(01) 00438-5


impairment via activation of microglia in chronic cerebral hypoperfusion. J. Neurosci. 38, 3520–3533. doi: 10.1523/JNEUROSCI.2451-17.2018



Zhong, J., Amina, S., Liang, M., Akther, S., Yuhi, T., Nishimura, T., et al. (2016). Cyclic ADP-ribose and heat tegulate oxytocin release via CD38 and TRPM2 in the hypothalamus during social or psychological stress in mice. Front. Neurosci. 10:304. doi: 10.3389/fnins.2016.00304

**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 © 2019 Malko, Syed Mortadza, McWilliam and Jiang. 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(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.

# Targeting Microglia and Macrophages: A Potential Treatment Strategy for Multiple Sclerosis

Jiaying Wang<sup>1</sup> , Jiajia Wang<sup>1</sup> , Jincheng Wang<sup>1</sup> , Bo Yang<sup>1</sup> , Qinjie Weng1,2 \* and Qiaojun He1,2 \*

<sup>1</sup> Zhejiang Province Key Laboratory of Anti-Cancer Drug Research, College of Pharmaceutical Sciences, Zhejiang University, Hangzhou, China, <sup>2</sup> Center for Drug Safety Evaluation and Research, Zhejiang University, Hangzhou, China

Multiple sclerosis (MS) is a chronic inflammatory neurodegenerative disease of the central nervous system (CNS). The early stage is characterized by relapses and the later stage, by progressive disability. Results from experimental and clinical investigations have demonstrated that microglia and macrophages play a key part in the disease course. These cells actively initiate immune infiltration and the demyelination cascade during the early phase of the disease; however, they promote remyelination and alleviate disease in later stages. This review aims to provide a comprehensive overview of the existing knowledge regarding the neuromodulatory function of macrophages and microglia in the healthy and injured CNS, and it discusses the feasibility of harnessing microglia and macrophage physiology to treat MS. The review encourages further investigations into macrophage-targeted therapy, as well as macrophage-based drug delivery, for realizing efficient treatment strategies for MS.

#### Edited by:

Pietro Giusti, University of Padua, Italy

#### Reviewed by:

Inge Huitinga, Netherlands Institute for Neuroscience (KNAW), Netherlands Irena Lavrnja, University of Belgrade, Serbia

#### \*Correspondence:

Qinjie Weng wengqinjie@zju.edu.cn Qiaojun He qiaojunhe@zju.edu.cn

#### Specialty section:

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

Received: 04 December 2018 Accepted: 08 March 2019 Published: 22 March 2019

#### Citation:

Wang J, Wang J, Wang J, Yang B, Weng Q and He Q (2019) Targeting Microglia and Macrophages: A Potential Treatment Strategy for Multiple Sclerosis. Front. Pharmacol. 10:286. doi: 10.3389/fphar.2019.00286 Keywords: multiple sclerosis, microglia, macrophages, central nervous system, targeted therapy

### INTRODUCTION

Multiple sclerosis, a CNS autoimmune and neurodegenerative disease, affects approximately 2.5 million people worldwide, seriously diminishing their quality of life and causing a significant financial burden (Compston and Coles, 2008; Lemus et al., 2018; Thompson et al., 2018). Studies using EAE, an animal model of MS, have revealed that microglia/macrophages actively participate in the pathogenesis of EAE progression (Jiang et al., 2014). The CNS contains several resident macrophages, such as microglia and non-parenchymal macrophages, in the choroid plexus, perivascular space, and meninges, which function to maintain CNS homeostasis (Prinz et al., 2011). Alterations in CNS homeostasis lead to the recruitment of peripheral blood-derived monocytes (known as monocyte-derived macrophages) into the CNS, the activation of microglia, and the presence of foamy macrophages (Bogie et al., 2014; Kuhlmann et al., 2017; Mrdjen et al., 2018; Zéphir, 2018). This phenomenon appears to be a type of self-protective mechanism designed to eliminate abnormalities and recover a steady state.

**Abbreviations:** APCs, antigen-presenting cells; BBB, blood–brain barrier; BDNF, brain-derived neurotrophic factor; CCL, C-C motif chemokine ligands; CNS, central nervous system; CSF, cerebrospinal fluid; CXCL, C-X-C motif chemokine ligands; EAE, experimental autoimmune encephalomyelitis; IFN, interferon; IL, interleukin; MHC, major histocompatibility complex; MS, multiple sclerosis; TNF, tumor necrosis factor.

Monocyte-derived macrophages and CNS-resident phagocytic cells can promote neuroinflammation, thereby inducing MS, but they can also play neuroprotective and anti-inflammatory roles, depending on many factors. In this review, we discuss the protective and pathogenic mechanism of microglia and macrophages in the development of MS and EAE, and propose the use of macrophages in MS therapy for repairing CNS damage.

### MICROGLIA AND MACROPHAGES IN A HEALTHY CNS

In a healthy CNS, microglia reside in the parenchyma, while non-parenchymal macrophages are present in boundary regions, including perivascular spaces, the meninges, and the choroid plexus. These cells have different developmental origins and serve distinct functions by specific subtypes.

### Origin of CNS Macrophages and Microglia

The earlier view that CNS macrophages are derived from bloodborne myeloid cells during adulthood has now been challenged, because studies using novel transgenic mouse models have shown that macrophages other than choroid plexus macrophages originate exclusively from the yolk sac (**Figure 1**; Gomez Perdiguero et al., 2015; Goldmann et al., 2016). The application of tamoxifen to adult Cx3cr1CreER:R26-YFP mice has further demonstrated that not only microglia but also perivascular and meningeal macrophages retain the yfp label from embryonic day (E)16.0 to 8–9 weeks of age (Goldmann et al., 2016), indicating that non-parenchymal meningeal and perivascular CNS macrophages and microglia arise entirely from embryonic precursor cells. Fate-mapping tools have helped reveal that CNS macrophages arise from the primitive c-kit<sup>+</sup> erythromyeloid precursors located in the yolk sac during E7.5∼8.0. Uncommitted erythromyeloid precursors subsequently disappear at E9.0 and develop into CD45+CX3CR1hiF4/80hi (A2) macrophage progenitors via immature CD45+CX3CR1loF4/80lo (A1) cells. Proliferating A2 cells then become CD11b+F4/80<sup>+</sup> microglia and perivascular, meningeal, and choroid plexus macrophages (Bertrand et al., 2005; Kierdorf et al., 2013; Hoeffel et al., 2015; Matcovitch-Natan et al., 2016). Once established in the CNS, microglia and perivascular and meningeal macrophages persist throughout the life of the organism due to their longevity and their capacity of self-proliferation, rather than the infiltration of peripheral myeloid cells (Ajami et al., 2007, 2011). In contrast, choroid plexus macrophages mainly depend on blood-derived immigrating Ly6Chi monocytes after birth (Goldmann et al., 2016). The astonishing finding of such close ontogeny between microglia and the other non-parenchymal macrophages in the CNS prompts the question of whether they can transform into one another under specific physiological conditions.

### Functions of Microglia

Microglia are widely scattered throughout the brain and spinal cord, and they come into close contact with neurons, astrocytes, and oligodendrocytes (Ransohoff and Cardona, 2010; Szepesi et al., 2018). Studies have demonstrated that microglia mediate synaptic pruning and trophic-factor production through fractalkine-CX3CR1 signaling in neurodevelopment (Tremblay et al., 2010; Ransohoff and El Khoury, 2014; Schafer and Stevens, 2015; Miyamoto et al., 2016). Fractalkine receptordeficient mice showed decreased neuron survival in layer V cortical neurons during the first postnatal week because of the reduced secretion of insulin-like growth factor from microglia (Ueno et al., 2013). Furthermore, microglia depletion by the Cre-Lox technique led to deficiencies in multiple learning tasks, as well as a significant reduction in learningrelated synapse formation (Parkhurst et al., 2014). In addition, microglial-secreted BDNF could increase the activity of neuronal tropomyosin-related kinase receptor B, an important modulator in synaptic plasticity, prompting a crucial role of BDNF signaling in learning and memory-related synapse formation (Parkhurst et al., 2014). Microglia may communicate with neurons; changes in the physiological environment such as chronic stress and light deprivation can cause hyper-ramification of microglia and more frequent neuron–microglia contacts (Torres-Platas et al., 2014; Welberg, 2014; Yirmiya et al., 2015). Moreover, microglia act as scavengers in the adult CNS to remove damaged components, apoptotic cells, and misfolded proteins (Sierra et al., 2010). Once "danger" signals are detected in the surrounding environment, microglia rapidly change into a reactive phenotype, characterized by larger soma, and release various complement factors to facilitate neuron and tissue repair (Hanisch and Kettenmann, 2007; Graeber and Streit, 2010; Kettenmann et al., 2011).

### Functions of Non-parenchymal Macrophages

Non-parenchymal brain macrophages have been largely ignored, but recent studies have refocused on their relevant functions in the perivascular space, meninges, and choroid plexus (Herz et al., 2017). Perivascular macrophages are critical in BBB establishment (Mendes-Jorge et al., 2009; He et al., 2016), and blockage of colony stimulating factor 1 receptor signaling leads to a decreased coverage of pericytes in brain vessels (Yamamoto et al., 2017). Perivascular macrophages continuously retract and protract along blood vessels in the adult brain (Daneman, 2012; Goldmann et al., 2016), suggesting that they may play a role in immune surveillance. Moreover, they can protect against bacterial infection by recruiting circulating leukocytes (Polfliet et al., 2001). The study of meningeal macrophages has been limited; however, it was reported that meninges possess a "glymphatic" system, which guides the clearance of waste products such as β-amyloid in the CNS (Raper et al., 2016). However, whether meningeal macrophages instruct lymphangiogenesis in the "glymphatic" system, similar to their role in the peripheral immune system, remains to be ascertained (Gordon et al., 2011; Louveau et al., 2015). In addition, meningeal macrophages function as APCs, detecting cell debris and antigens to maintain immune homeostasis (Kivisakk et al., 2009). Choroid plexus macrophages are located near the microvilli of the choroid plexus (Liddelow, 2015; Marques et al., 2017), and their main function may support CSF release and flux (Goldmann et al., 2016).

In conclusion, considerable fundamental details about nonparenchymal macrophages remain to be determined.

### Polarization of Microglia and Macrophages

Macrophages (including microglia) are often classified into M1 (inflammatory) and M2 (anti-inflammatory) phenotypes (Mikita et al., 2011; Liu et al., 2013). However, this simple classification does not reveal the full spectrum of macrophages under different stimulation conditions owing to their diversity and plasticity (Prinz and Priller, 2014). Transcriptome profiles have revealed at least nine different subtypes of human macrophages (Veremeyko et al., 2018). In this review, M1 and M2 represent the activation states of macrophages rather than specific cell subtypes.

Interferon-γ and lipopolysaccharide treatment can generate M1 cells, while IL-4, IL-10, or IL-13 are activators of M2 cells. M1 macrophages and microglia are characterized by an amoeboid shape, and they release pro-inflammatory cytokines, such as IL-6, IFN-γ, IL-23, TNF-α; inducible nitric oxide synthase; chemokines, including CCL 4, CCL5, and CCL8; and CXCL 2, CXCL4, and CXCL9 (Juhas et al., 2015; Kapellos and Iqbal, 2016). M1 macrophages and microglia are potent APCs, causing an adaptive immune response to clear foreign substances and abnormal proteins (Das et al., 2015). In contrast, M2 macrophages and microglia have smaller cell bodies and a branched structure (Ransohoff and Perry, 2009), and they express a variety of anti-inflammatory molecules, including IL-4, IL-10, IL-13, and transforming growth factor-β (David and Kroner, 2011). These contribute to immunoregulation by inhibiting inflammation, remodeling tissue, promoting angiogenesis, and clearing parasites (Banerjee et al., 2013). M2 microglia promote oligodendrocyte differentiation, and their depletion inhibits remyelination (Kotter et al., 2005; Miron et al., 2013). M2 polarized macrophages and microglia can be further subdivided into M2a, M2b, and M2c cells. These have different activation and functional mechanisms but have some biochemical overlap (**Table 1**; Murray et al., 2014). Increasing evidence suggests that M1 cells prefer an anaerobic environment, whereas oxygen consumption and oxidative phosphorylation benefit M2 macrophages (Shi et al., 2017; Wang et al., 2018).

In conclusion, microglia and non-parenchymal macrophages are independent immune populations in the CNS. Besides different cell origins, they have their own functions in CNS development, homeostasis, and tissue healing (Lopez-Atalaya et al., 2018). Under steady state, microglia are primarily responsible for supporting the development of the healthy brain, they shape and maintain the neuronal synaptic network, interact with neurons and produce beneficial components (Hagemeyer et al., 2017). And non-parenchymal macrophages strictly limits access of peripheral immune cells (T cells, B cells, and monocytes) reach the CNS parenchyma (Polfliet et al., 2001). Upon detecting "danger" signals, both microglia and nonparenchymal macrophages produce large amounts of chemokines and cytokines to promote leukocyte infiltration into the CNS parenchyma and function as APCs to reactivate T cells in the CNS (Galli et al., 2011; Krausgruber et al., 2011). Moreover, microglia phagocytose damaged cell debris, as well as unnecessary neurons and synapses resulting from defective differentiation and/or migration (Paolicelli et al., 2011; Lopez-Atalaya et al., 2018).

### MICROGLIA AND MACROPHAGES IN MS

After CNS injury and inflammation, while microglia situated in the parenchyma and macrophages in the choroid plexus, perivascular space, and the meninges are activated, monocytes from the blood cross the damaged BBB and become monocytederived macrophages in the CNS (Lassmann et al., 2001). These spatial differences may determine subtype-specific effector functions of these distinctive phagocytes in the diseased CNS. In this section, we compare and elaborate the functions of microglia and macrophages in the progression of MS and EAE.

### Distinguishing Microglia From Other Macrophages

A lack of specific markers that distinguish resident microglia from other macrophages in MS lesions hampers the accurate

evaluation of their contributions to human brain pathology. Originally, the expression profiles of CD11b and CD45 allowed the discrimination of CD11b+CD45med microglia from CD11b+CD45hi monocyte-derived macrophages (Martin et al., 2017). However, such a classification remains controversial (Koeniger and Kuerten, 2017). Studies have suggested TMEM119 (Bennett et al., 2016; Satoh et al., 2016; Li et al., 2019), Sall1 (Buttgereit et al., 2016), Siglec-H (Konishi et al., 2017), and P2Y12 (Mildner et al., 2017) as more valuable markers for distinguishing microglia from other macrophages. For instance, TMEM119 is expressed exclusively on Iba1+CD68<sup>+</sup> microglia and not on infiltrated Iba1+CD68<sup>+</sup> macrophages within demyelinating lesions of MS (Satoh et al., 2016). However, TMEM119 expression is absent in immature microglia (Satoh et al., 2016). Sall1 expression is abundant in neuronal and glial progenitor cells during CNS development (Harrison et al., 2012; Buttgereit et al., 2016), and P2Y12 shows decreased expression in activated microglia (Amadio et al., 2014; Mildner et al., 2017). Almost all microglia in the CNS parenchyma expressed Siglec-H, from developmental to mature stages, and the expression was maintained in activated microglia after CNS injuries (Koso et al., 2018). In contrast, Siglec-H expression was largely absent from other myeloid cells in the CNS (Koso et al., 2018). Overall, these findings suggest that Siglec-H is a promising specific marker for the specific identification of microglia from CNSassociated macrophages. However, these results are all from mice, and human Siglec-L2 is only ∼42% homologous with Siglec-H (Zhang et al., 2006). Whether Siglec-L2 might be a marker for microglia warrants further investigation.

### Microglia

Pronounced microglial activation with high expression levels of pro-inflammatory genes can be found in lesion areas and normal-appearing white matter of MS patients (Lucchinetti et al., 2000; Singh et al., 2013; Zrzavy et al., 2017). These clusters of activated microglia are regarded as "pre-active lesions," with an absence of leukocyte infiltration and demyelination; however, they may eventually develop into active demyelinating MS lesions (Singh et al., 2013). In contrast to healthy controls, the number of P2Y12<sup>+</sup> homeostatic microglia is significantly reduced in the white matter of MS patients, and TMEM119<sup>+</sup> microglia predominantly govern the edge of active lesions in


MS (Zrzavy et al., 2017), indicating that microglial activation is related to disease development.

TSPO tracers have demonstrated inflammatory processes with microglia involvement in MS (Vas et al., 2008; Oh et al., 2011). These microglia are enriched with iron, express proinflammatory cytokines, and cause persistent tissue damage (Wisnieff et al., 2015). Although microglia are important in synaptic plasticity, this function is destroyed under pathological insults, which causes synaptic loss in MS and eventually, cognitive decline (Michailidou et al., 2015; Hong et al., 2016; Salter and Stevens, 2017; Di Filippo et al., 2018). Activated microglia in EAE mice may produce large amounts of TNF-α and thus induce excessive glutamate, leading to spontaneous and miniature excitatory post-synaptic currents (Centonze et al., 2009, 2010). Furthermore, astroglial ATP amplifies microglial signals and subsequently promotes glutamate release from astroglia, which directly inhibits synaptic transmission (Kettenmann et al., 2013). Classically activated neuroinflammatory microglia secrete TNF-α, and C1q, inducing A1 astrocyte dysfunction, thereby resulting in the dysfunction of both oligodendrocytes and neurons (Liddelow et al., 2017; Rothhammer et al., 2018). Importantly, mitochondrial disturbances have been demonstrated in modulating MS lesions, reactive oxygen species and reactive nitrogen species generated by the activated microglia also lead to intra-axonal mitochondrial injury (Balaban et al., 2005; Haile et al., 2017). Kinetics studies from EAE mice indicate that microglia are the first cell line to take up myelin antigens (Sosa et al., 2013), subsequently restimulating encephalitogenic T cells in the CNS through major histocompatibility complex (MHC) molecules and costimulatory molecules (Perry, 1998). Consistent with this, time-course studies in EAE mice showed that the appearance of inflammatory T cells in the CNS is consistent with the activation of CD11b<sup>+</sup> microglia (Murphy et al., 2010). However, specific deletion of MHCII in microglia does not affect disease progression, suggesting that microglia are not important in reactivating efficient T cells in the CNS (Wolf et al., 2018). Microglia also secrete a series of pro-inflammatory cytokines such as IL-18, IL-6, and IL-1b and chemokines like CCL2 and CCL5 to aggravate both MS and EAE (Merson et al., 2010; Jiang et al., 2014). Moreover, activated microglia in white matter lesions may lead to greater brain atrophy in secondary progressive disease (Datta et al., 2017). In conclusion, microglia exert detrimental functions in MS and strongly damage myelin.

Microglial activation may also be beneficial for remyelination in MS due to the phagocytotic ability of microglia, as well as their secretion of neuroprotective molecules and antiinflammatory cytokines (Du et al., 2017). Microglia-specific deletion of A20, the crucial protein in regulating microglial activation, makes microglia lose their ability to regulate neuronal synaptic function, thus exacerbating MS-like disease (Voet et al., 2018). Several receptors, including TREM2, complement receptor 3, and signal regulatory protein-α, invoke microglia to phagocytose myelin debris in lesions (Brendecke and Prinz, 2015). Microglia-specific ablation of TREM2 cannot upregulate genes associated with phagocytosis (Poliani et al., 2015), and the blockage of TREM2 results in increased demyelination in EAE (Piccio et al., 2007). Moreover, microglia secrete

trypsinogen, insulin-like growth factor, and fibroblast growth factor during activation to promote neurogenesis (Pérez-Martín et al., 2010; Voss et al., 2012). Anti-inflammatory molecules from microglia also benefit remyelination in EAE, for example, IL-4 enhances oligodendrogenesis (Butovsky et al., 2006), and activin A drives oligodendrocyte differentiation (Miron et al., 2013). Interestingly, pro-inflammatory cytokine TNF has neuroprotective functions in EAE (Wolf et al., 2017), and deletion of microglial TNFR2 causing early disease onset, with increased leukocyte infiltration, demyelination and T cell activation in CNS (Gao et al., 2017). However, soluble TNF repair remyelination by suppressing clearance of myelin debris (Karamita et al., 2017). Other evidence indicated that microglia can prompt remyelination through P2X4R signaling, and that the blockade of P2X4R results in microglial activation to a pro-inflammatory phenotype and exacerbates the clinical symptoms of EAE mice (Zabala et al., 2018).

### Monocyte-Derived Macrophages

During the effector stage of EAE, Ly6Chi monocytes rapidly infiltrate into the inflamed CNS with the help of matrix metalloproteinases (Ajami et al., 2011; Yamasaki et al., 2014). Monocytes then differentiate into macrophages, express MHC II and costimulatory molecules, and produce pro-inflammatory factors that actively contribute to the demyelination process (Jiang et al., 2014; Yamasaki et al., 2014). Consistently, the abundance of CCR1+/CCR5<sup>+</sup> hematogenous monocytes has been found to appear in demyelination zones compared with the periplaque white matter of MS patients (Trapp et al., 1998; Trebst et al., 2001), and Kim1p staining shows abundant subcortical macrophage infiltration (Tobin et al., 2017).

The activated monocyte-derived macrophages is generally considered to be harmful in MS (Yamasaki et al., 2014). Infiltrating monocytes accumulate at the nodes of Ranvier during early stages of EAE and actively initiate the demyelination of axons (Yamasaki et al., 2014), similar to microglia, but they do not affect CNS-resident microglia (Ajami et al., 2011). Bone marrow chimeric mice show that TRPM2 expressed by CNS-infiltrating macrophages prompts the production of CXCL2, thus increasing neutrophil infiltration and contributing to the progression of EAE (Tsutsui et al., 2018). High F4/80+CD45hi macrophage/Tcell ratios were found in EAE mice (McMahon et al., 2005), indicating that recruited macrophages participate in T-cell proliferation after myelin antigen stimulation. Macrophages also secrete large amounts of pro-inflammatory cytokines, such as IL-6, IL-12, TNF-α, IL-1, and IL-23, to aggravate the inflammatory response (Valentin-Torres et al., 2016). IL-1β expression has been found to be pivotal in the polarization of Th17 cells and therefore for EAE induction (Sutton et al., 2006). IL-6 has also been demonstrated to be critical for CNS autoimmunity, as IL-6-deficient mice demonstrated alleviated EAE (Samoilova et al., 1998). Macrophages also express high levels of CCR4 during the disease, and CCR4 deletion leading to reduced accumulation of macrophages in CNS and attenuated EAE symptoms (Forde et al., 2011). These findings suggest that macrophages are responsible for demyelination, but their direct functions on axons remain to be determined.

Other studies have also defined the neuron-supportive activities of Ly6Chi monocytes in EAE. The activation of invariant natural killer T cells converts inflammatory Ly6Chi monocytes to M2 macrophages in the CNS, with associated improvement in neurological impairment (Denney et al., 2012; Jiang et al., 2017). Recent studies have showed that TG2 mRNA levels are increased in monocytes derived from MS patients and correlates with anti-inflammatory cytokine expression, proposing a more anti-inflammatory monocytes status in MS (Sestito et al., 2017). Monocyte-derived macrophages can also promote CNS repair in the injured spinal cord by clearing myelin debris through scavenger receptors (GrandPré et al., 2000; Kotter et al., 2006). In contrast to microglia, macrophages express TREM-1 rather than TREM-2, and although this isoform plays a role in inflammatory responses (Colonna and Facchetti, 2003), the functions of TREM-1 still need to be further studied in a CNS autoimmunity context. In addition, monocyte-derived macrophages secrete trophic factors and anti-inflammatory factors to alleviate sympathetic neuron dysfunction (Hikawa and Takenaka, 1996; Miron et al., 2013). The transcription factor Nr4a1 represses the autocrine production of norepinephrine in macrophages, and myeloid cells lacking Nr4a1 lead to more leukocyte infiltration within the CNS and more severe EAE clinical symptoms (Shaked et al., 2015). Consistently, expression levels of Nr4a were down-regulated in peripheral blood mononuclear cells from MS patients, and Fingolimod treatment could recovery MS from Nr4a2 deficit (Montarolo et al., 2018). Besides Ly6Chi monocytes, a small number of LyC6lo monocytes also recruited in the injured CNS (Saederup et al., 2010). Although studies have pointed toward a reparative role of infiltrated Ly6Clo monocytes in a spinal cord injury model (Arnold et al., 2007; Nahrendorf et al., 2007); however, more research is needed to determine whether Ly6Clo monocytes do promote repair processes in the pathogenesis of EAE.

### Non-parenchymal Macrophages

Perivascular macrophages, typically CD163+HLADR+, are markedly increased in MS patients and EAE-affected animals (Zhang et al., 2011; Mammana et al., 2018). Substantial evidence indicates that perivascular macrophages have both supportive and neuroprotective roles in MS. Selective elimination of both perivascular and meningeal macrophages by treatment with clodronate-loaded liposomes reduces the progression of EAE in rats (Polfliet et al., 2002), suggesting that perivascular macrophages may act as competent APCs in EAE. EAErelated perivascular macrophages upregulate the expression of MHC II and can present peptides to myelin-specific T cells in the CNS, thereby accelerating disease progression (Perry, 1998; Fabriek et al., 2005; Greter et al., 2005; Vogel et al., 2013). Furthermore, perivascular macrophages express high levels of adhesion molecules such as VCAM-1 and ICAM-1 and chemokines like CCL2 and CCL3 (Hofmann et al., 2002; Vercellino et al., 2017), which may mediated by chondroitin sulfate proteoglycans (Stephenson et al., 2018), to promote leukocyte infiltration into the CNS. However, perivascular macrophages during the early stage of EAE are characterized as mannose receptor-positive M2 macrophages, produce leukemia

inhibitory factors and neuroprotective factors, and secrete antiinflammatory IL-6 and IL-1ra to protect glial and neuronal cells (Boven et al., 2006; Vanderlocht et al., 2006). Moreover, M2 perivascular macrophages can promote the differentiation of Th2 and Treg cells due to their anti-inflammatory characteristic (Weber et al., 2007).

An extensive number of meningeal immune cells in the CNS leads to a more severe clinical course of MS (Howell et al., 2011; Choi et al., 2012; Kilsdonk et al., 2017; Kolber et al., 2017; Bevan et al., 2018). In concordance, meningeal macrophages are equipped to present peptide antigens to T cells, triggering neuroinflammatory events (Bartholomäus et al., 2009; Kivisakk et al., 2009; Ransohoff and Cardona, 2010). In addition, meningeal macrophages can release inflammatory and toxic mediators or even invade the CNS parenchyma to sustain MS pathology (Maxwell et al., 1990; Garabedian et al., 2000; Reuter et al., 2001). In contrast, meningeal macrophages produce neurotrophic and anti-inflammatory factors to inhibit leukocyte activity and their infiltration into the CNS at particular disease stages (Bogie et al., 2014). Leptomeningeal macrophages increase the expression of transforming growth factor-β in glial cells and cortical neurons by releasing prostaglandin E2 during inflammation (Wu et al., 2007). However, other studies demonstrated that prostaglandin E2 could also exacerbate EAE (Lima et al., 2012), depending on the disease stage and microenvironment.

The choroid plexus, the site of CSF production, forms the interface between blood and the CNS (Wolburg and Paulus, 2010; Kaur et al., 2016). It is thus an important element in the communication between the vascular compartment and the CNS during inflammation (Breuer et al., 2018). The severity of MS coincides with enhanced levels of infiltrated leukocytes in the choroid plexus (Engelhardt et al., 2001; Vercellino et al., 2008; Mitchell et al., 2009), strongly indicating that inflammation at the choroid plexus contributes to the disease. Chemokines like CXCL9 and CXCL10 accumulate in large quantities in the CSF during MS progression (Edwards et al., 2013; Puthenparampil et al., 2017), but whether they are secreted by macrophages needs to be confirmed. Choroid plexus macrophages may release neurotoxic mediators and pro-inflammatory factors into the CSF during disease to impact perivascular and meningeal inflammation, as well as the activation and integrity of neurons and glial cells (Garabedian et al., 2000; Bragg et al., 2002a,b). However, remote brain choroid plexus can also recruit M2-like macrophages to an injured spinal cord, which might benefit myelin repair (Shechter et al., 2013).

### Foamy Macrophages

Ample evidence indicates that abundant foamy macrophages are present in MS lesions (Grajchen et al., 2018). After disease initiation, both resident microglia and recruited monocytes ingest and accumulate vast amounts of myelin-derived lipids and thereby acquire their distinctive morphology (Li et al., 1996; Bogie et al., 2017), but how foamy macrophages function in the process of MS is still poorly understood. MHC II and costimulatory molecules expressed on foamy macrophages may present myelin antigens to autoreactive T cells (de Vos et al., 2002; Boven et al., 2006; van Zwam et al., 2011), thereby driving epitope spread and MS progression or even initiation (Stys et al., 2012). In agreement with this, numerous studies have found M1-like foamy macrophages present within MS lesions (Bogie et al., 2014). Interestingly, yet another study showed that microglia show the M1 phenotype after the first 6 h of myelin uptake, and they soon convert to the M2 phenotype after prolonged exposure (Liu et al., 2006), indicating that the polarization of foamy macrophages can be reversed. Both in vivo studies and in vitro cultures demonstrated that foamy macrophages prefer to express anti-inflammatory molecules rather than secrete pro-inflammatory cytokines (van Zwam et al., 2011), indicating that myelin macrophages may resolve the inflammation in MS demyelination lesions. Consistent with this, researchers found that mouse myelin macrophages could inhibit IFN-γ release by Th1 cells and suppress EAE severity (Bogie et al., 2014). Collectively, these studies imply that besides being aggressors in MS pathogenesis, foamy macrophages can also destroy T cell-induced autoimmunity. However, more research is warranted to confirm the specific role of foamy macrophages in MS progression.

In fact, among numerous phagocytic cells, only microglia and infiltrating macrophages play a crucial role in MS progression. Both of them participate in demyelination and remyelination through antigen presentation, phagocytosis, and inflammatory mediator secretion (Shechter et al., 2009; Ruckh et al., 2012; Li and Barres, 2018); however, microglia alone could affect neuron cells directly through fractalkine-CX3CR1 signaling (Eyo et al., 2016). Although microglia and infiltrating macrophages are potent APCs in EAE, the expression levels of MHC II and costimulatory molecules on microglia are lower (Brendecke and Prinz, 2015), indicating that inflammation in the CNS is mainly activated by infiltrating macrophages. The roles of perivascular macrophages, as well as macrophages, in the meninges and choroid plexus in MS are not yet completely understood. In conclusion, whether these macrophages are unique phagocyte subsets in MS, and if so, the underlying mechanism thereof, needs to be determined.

## THERAPEUTIC PERSPECTIVES

As discussed in the previous sections, any change in the physiological functions of macrophages and microglia can impact neuroinflammatory and neurodegenerative events. Therefore, selectively promoting macrophage polarization to the anti-inflammatory subtype is an attractive therapeutic option (**Figure 2**).

### MS Therapy by Regulating Macrophages

Although the crucial role of macrophages and microglia in neurodegenerative events has been proven, therapeutics solely targeting these phagocytes are still lacking. Currently, the clinical treatment of MS includes recombinant IFNβ-1a, glatiramer acetate, and natalizumab to mainly regulate T and B cells (Cheng et al., 2017). Studies have found that the neuroprotective effects of glatiramer acetate are mediated by activated M2

microglia (Ratchford et al., 2012). Furthermore, diverse new compounds that influence macrophage and microglial functions are undergoing non-clinical testing. These agents include ethyl pyruvate (Djedovic et al., 2017 ´ ), spermidine (Yang et al., 2016), bryostatin-1 (Kornberg et al., 2018), forskolin (Veremeyko et al., 2018), and lenalidomide (Weng et al., 2018). Ethyl pyruvate can reduce high-mobility group box 1 expression in activated ED1<sup>+</sup> macrophages and Iba1<sup>+</sup> microglia, thus inhibiting the activation of macrophages/microglia within the CNS to protect against EAE (Djedovic et al., 2017 ´ ). In addition, forskolin has been found to alleviate EAE by suppressing the expression of CD86 while enhancing the expression of ARG1 in macrophages (Veremeyko et al., 2018). Bryostatin-1 provides marked benefits in mice with EAE by acting on APCs, including macrophages, to promote the differentiation of lymphocytes into Th2 cells (Kornberg et al., 2018).

The novel PADRE-Kv1.3 vaccine has been applied to treat EAE. After vaccination, infiltrated microglia/macrophages significantly shift toward the M2 subtype in the CNS (Fan et al., 2018). Mineralocorticoid receptors expressed on macrophages promote their polarization toward M2 cells, and mice selectively lacking mineralocorticoid receptors in their myeloid cells show diminished clinical symptoms of EAE (Montes-Cobos et al., 2017b). Transferring M2 macrophages to eliminate EAE has been widely applied. The long non-coding RNA GAS5, a suppresser of microglial M2 polarization, is expressed at a high level within microglia in MS patients. Transplanting GAS5-downregulated microglia into the lateral ventricles of EAE mice could promote remyelination, similar to a lysolecithin-induced demyelination model (Sun et al., 2017). In addition, Ly6C<sup>+</sup> monocytes could give rise to PD-L1+/PD-L2<sup>+</sup> M2 cells, and transferring PD-L1+/PD-L2<sup>+</sup> M2 macrophages might reduce the incidence of EAE (Terrazas et al., 2017).

### Macrophage-Targeted Therapeutics via Micro-Based Drug Delivery Systems in Neurodegenerative Diseases

Because most drugs do not effectively reach macrophages at therapeutic levels, micro-based drug delivery systems for targeting macrophages have been widely used in treating CNS diseases. For example, glucocorticoids are used extensively to treat acute relapses in MS patients (Pato-Pato et al., 2003; Cohen et al., 2009), but their therapeutic effect is accompanied by serious side effects due to their broad spectrum of immunosuppressive actions (Baschant and Tuckermann, 2010). Interestingly, in vivo, glucocorticoids packaged with specific nanoparticles vs. free glucocorticoids are preferentially taken up by phagocytic cells, including macrophages, rather than T cells. This approach allows their therapeutic efficacy to be retained in EAE mice (Montes-Cobos et al., 2017a). Other research showed that infiltrated CD163<sup>+</sup> macrophages in brains

from a Parkinson's disease model modulate local microglia to promote neuroprotection. Specifically designed liposomes can load dexamethasone solely to CD163<sup>+</sup> macrophages. This modification leads to decreased dopaminergic cell death and better motor performance (Tentillier et al., 2016). However, we cannot ignore that several nanoparticles themselves also induce macrophage polarization toward various phenotypes and modulate macrophage reprogramming (Miao et al., 2017).

### Macrophages as Drug Carriers for MS Therapy

Many CNS diseases are thought to be untreatable because drugs rarely pass through the BBB at therapeutic levels (Fischbach et al., 2013). Chemokine gradients secreted from the CNS parenchyma can induce macrophage migration to the brain, making macrophages potential carriers for drug/nano formulations across the BBB to reach target sites (Ye et al., 2018). Macrophages transporting superparamagnetic iron oxide nanoparticles and exogenous genes were injected intravenously into a lipopolysaccharide-induced acute neuroinflammatory mouse model. The differentiated macrophages demonstrated an excellent ability to spread into the brain, and the number of transported macrophages positively correlated with the number of intravenous cells (Tong et al., 2016). Moreover, macrophage exosomes can interact with brain microvessel endothelial cells that comprise the BBB by binding to carbohydrate-binding C-type lectin receptors. BBB cells can absorb more macrophage exosomes when inflammation occurs due to the upregulated secretion of ICAM-1. After intravenous administration, macrophage exosomes have been shown to cross the BBB and deliver a cargo protein, BDNF, in the presence of CNS inflammation (Yuan et al., 2017). Overall, these results support the concept of using monocytes/macrophages or

### REFERENCES


macrophage exosomes as cellular vehicles for the delivery of small drugs, RNAs, and proteins to the brain parenchyma.

### PERSPECTIVE

It is becoming increasingly clear that microglia and macrophages play critical roles in healthy, inflamed, injured, and recovering CNS due to their dual natures. We have summarized the origin, subtypes, and functions of these phagocytic cells in MS. Therapeutic interventions that block the proinflammatory effects of macrophages/microglia during disease progression, while preserving their anti-inflammatory functions on EAE, have achieved great success. However, as infiltrating inflammatory monocytes and microglia contribute differentially to EAE pathophysiology, strategies that allow the differential manipulation of CNS-resident microglia and infiltrating macrophages to optimally target different mechanisms operating in MS need to be developed. Moreover, extensive additional research is needed before utilizing the physiology of microglia and macrophage subsets for MS therapy.

### AUTHOR CONTRIBUTIONS

QW and QH conceived the review article and made the corrections in the manuscript. BY provided some critical comments. JyW wrote the manuscript. JjW and JcW collected the related research articles.

### FUNDING

This study was supported by the National Natural Science Foundation of China (81872878) and the Zhejiang Provincial Natural Science Foundation of China (LGF18H310001).

structures in nascent autoimmune CNS lesions. Nature 462, 94–98. doi: 10. 1038/nature08478




of p47-PHOX-mediated generation of reactive oxygen species. J. Neurosci. 26, 12904–12913. doi: 10.1523/JNEUROSCI.2531-06.2006



CXCL2 chemokines. J. Neurosci. 38, 8484–8495. doi: 10.1523/JNEUROSCI. 2203-17.2018


lesions: interpreting positive susceptibility and the presence of iron. Magn. Reson. Med. 74, 564–570. doi: 10.1002/mrm.25420


**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 © 2019 Wang, Wang, Wang, Yang, Weng and He. 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.

# Icariin Attenuates M1 Activation of Microglia and Aβ Plaque Accumulation in the Hippocampus and Prefrontal Cortex by Up-Regulating PPARγ in Restraint/Isolation-Stressed APP/PS1 Mice

Yihe Wang<sup>1</sup> , Tianrui Zhu<sup>2</sup> , Min Wang<sup>2</sup> , Feng Zhang<sup>2</sup> , Guitao Zhang<sup>2</sup> , Jing Zhao<sup>2</sup> , Yuanyuan Zhang<sup>2</sup> , Erxi Wu3,4,5 and Xiaohong Li<sup>2</sup> \*

Edited by: Pietro Giusti, University of Padova, Italy

#### Reviewed by:

Shi Jing Shan, Zunyi Medical University, China Liu Dexiang, Shandong University, China

> \*Correspondence: Xiaohong Li xiaohong-li@sdu.edu.cn

#### Specialty section:

This article was submitted to Neuropharmacology, a section of the journal Frontiers in Neuroscience

Received: 11 December 2018 Accepted: 13 March 2019 Published: 28 March 2019

#### Citation:

Wang Y, Zhu T, Wang M, Zhang F, Zhang G, Zhao J, Zhang Y, Wu E and Li X (2019) Icariin Attenuates M1 Activation of Microglia and Aβ Plaque Accumulation in the Hippocampus and Prefrontal Cortex by Up-Regulating PPARγ in Restraint/Isolation-Stressed APP/PS1 Mice. Front. Neurosci. 13:291. doi: 10.3389/fnins.2019.00291 <sup>1</sup> School of Medicine, Shandong University, Jinan, China, <sup>2</sup> Department of Neurology, Jinan Central Hospital, Shandong University, Jinan, China, <sup>3</sup> Department of Neurosurgery and Neuroscience Institute, Baylor Scott & White Health, Temple, TX, United States, <sup>4</sup> Department of Surgery and Department of Pharmaceutical Sciences, Texas A&M University Health Science Center, College Station, TX, United States, <sup>5</sup> LIVESTRONG Cancer Institutes, Dell Medical School, The University of Texas at Austin, Austin, TX, United States

Background: Studies have shown that psychosocial stress is involved in Alzheimer's disease (AD) pathogenesis; it induces M1 microglia polarization and production of proinflammatory cytokines, leading to neurotoxic outcomes and decreased β-amyloid (Aβ) clearance. Icariin has been proven to be an effective anti-inflammatory agent and to activate peroxisome proliferator-activated receptors gamma (PPARγ) which induces the M2 phenotype in the microglia. However, whether restraint/isolation stress reduces the clearance ability of microglia by priming and polarizing microglia to the M1 phenotype, and the effects of icariin in attenuating the inflammatory response and relieving the pathological changes of AD are still unclear.

Methods: APP/PS1 mice (male, aged 3 months) were randomly divided into a control group, a restraint/isolation stress group, and a restraint/isolation stress + icariin group. The restraint/isolation stress group was subjected to a paradigm to build a depressive animal model. Sucrose preference, open field, elevated plus maze, and Y maze test were used to assess the stress paradigm. The Morris water maze test was performed to evaluate spatial reference learning and memory. Enzyme-linked immunosorbent assay and immunohistochemistry were used to identify the microglia phenotype and Aβ accumulation. Western blotting was used to detect the expression of PPARγ in the hippocampus and prefrontal cortex (PFC).

Results: Restraint/isolation stress induced significant depressive-like behaviors in APP/PS1 mice at 4 months of age and memory impairment at 10 months of age,

while 6 months of icariin administration relieved the memory damage. Restraint/isolation stressed mice had elevated pro-inflammatory cytokines, decreased anti-inflammatory cytokines, increased Aβ plaque accumulation and more M1 phenotype microglia in the hippocampus and PFC at 10 months of age, while 6 months of icariin administration relieved these changes. Moreover, restraint/isolation stressed mice had down-regulated PPARγ expression in the hippocampus and PFC at 10 months of age, while 6 months of icariin administration reversed the alteration, especially in the hippocampus.

Conclusion: Restraint/isolation stress induced depressive-like behaviors and spatial memory damage, over-expression of M1 microglia markers and more severe Aβ accumulation by suppressing PPARγ in APP/PS1 mice. Icariin can be considered a new treatment option as it induces the switch of the microglia phenotype by activating PPARγ.

Keywords: stress, Alzheimer's disease, microglia, icariin, PPARγ, cytokine, hippocampus, prefrontal cortex

### INTRODUCTION

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by progressive cognitive impairment (Selkoe and Schenk, 2003). Prominent neuropathological features of AD are β-amyloid (Aβ) plaques and neurofibrillary tauopathy consisting of threads and tangles (NFT), which are observed throughout the brain, including the areas critically involved in memory formation and emotional regulation (Braak and Braak, 1996).

Mounting evidence shows that psychosocial stress is involved in AD pathogenesis, as patients with posttraumatic stress disorder (PTSD) or depression often develop dementia and even clinical AD (Green et al., 2003; Wilson et al., 2005; Qureshi et al., 2010; Yaffe et al., 2010; Gracia-Garcia et al., 2015; Sacuiu et al., 2016). The elevated level of cortisol, a stress hormone, found in AD patients plays a crucial role in contribution to the comorbidity (Hartmann et al., 1997; Elgh et al., 2006). Studies in animal models further demonstrated the association of hypothalamic–pituitary–adrenal axis activation with AD pathogenesis in the frontal cortex and hippocampus (Green et al., 2006; Sotiropoulos et al., 2008; Joshi et al., 2012; Justice et al., 2015). Chronic unpredictable stress (CUS) significantly increased both serum corticosterone levels (Liu et al., 2010; Yang et al., 2012) and amyloid precursor protein (APP) fragments. Aβ infusion triggered APP misprocessing in the hippocampus and prefrontal cortex (PFC) of rats, which was further exacerbated by stress (Catania et al., 2009). Furthermore, chronic stress including restraint/isolation stress in particularly elevated Aβ40 and Aβ42 levels of brain, accelerated amyloid plaque formation and impaired learning and memory in mice (Jeong et al., 2006; Carroll et al., 2011; Lee and Han, 2013). More importantly, a recent study demonstrated that prenatal stress induced a constant neuroinflammatory response which contributes to a more vulnerable profile; the latter can subsequently lead to an aberrant response to accumulating Aβ peptides, and ultimately modify the extent of Aβ neuropathology. The study concluded that early-life stress aggravated plaque pathology in 10-monthold APP/PS1 mice, and was accompanied by reduced microglial accumulation and increased level of pro-inflammatory tumor necrosis factor-α (TNF-α) (Hoeijmakers et al., 2016).

Microglial activation is often classified into classical (M1) and alternative (M2). M1 microglia may contribute to dysfunction of the neurotrophic system by expressing pro-inflammatory cytokines, such as TNF-α, interleukin-1β (IL-1β) and IL-6 (Michelucci et al., 2009). In the M1 phenotype of the microglia, the activation of nuclear factor κB (NF-κB) may play a critical role in the production of pro-inflammatory cytokines, leading to neurotoxic outcomes (Park et al., 2015). The M2 phenotype, also known as the neuroprotective microglial phenotype, releases different mediators including IL-4, IL-10, and transforming growth factor-β (TGF-β) (Michelucci et al., 2009) to antagonize inflammation-induced damage in the central nervous system (CNS) (Kobayashi et al., 2013; Zhao et al., 2015), and is associated with enhanced phagocytosis of deposited amyloid (Mandrekar-Colucci et al., 2012; Heneka et al., 2013; He et al., 2016; Huang et al., 2017). Stress paradigms, including CUS and chronic social defeat stress, are among the factors that could induce M1 microglia polarization (Zhao et al., 2016; Tang et al., 2018).

Peroxisome proliferator-activated receptors (PPARs) are a group of nuclear receptor proteins regulating gene expression as ligand-activated transcription factors (Michalik et al., 2006). Three closely related PPAR isoforms have been identified (alpha, delta, and gamma), transcribed from different genes and characterized by different tissue distribution, ligand specificity, and physiological roles (Berger and Moller, 2002; Breidert et al., 2002; Chang et al., 2007; Tontonoz and Spiegelman, 2008). Among the three isoforms, PPARγ has the highest expression in the CNS, where it has been identified in neurons, astrocytes, and glial cells (Moreno et al., 2004). Specifically, PPARγ appears to be linked to stress modulation (Garcia-Bueno et al., 2005) and to mediate the conversion between microglia phenotypes (Yamanaka et al., 2012; Saijo et al., 2013). PPARγ induced the M2 phenotype microglia under Aβ toxicity via adiponectin treatment through its effects on the anti-inflammatory response (Song et al., 2017). Moreover, pioglitazone which was used to treat depressive-like behaviors in a chronic mild stress mouse model is

associated with PPARγ-mediated M2 activation of the microglia (Zhao et al., 2016).

Icariin, a natural flavonoid compound extracted from Epimedium brevicornum Maxim (a traditional Chinese herb), has been proven to have a wide range of effects including anti-tumor, anti-oxidant, anti-bacterial, and anti-inflammatory properties (Luo et al., 2007; Zhou et al., 2011). Previous studies investigated the pharmacological properties of icariin in the CNS and showed that it can markedly attenuate cognitive deficits in several models of AD (Li et al., 2010; Nie et al., 2010; Jin et al., 2014) and alleviate the neuronal injury induced by ischemia (Li et al., 2005). It was also reported that icariin can attenuate the inflammatory response through PPARγ activation in rats (Xiong et al., 2016).

Previous studies also showed that various stress can activate M1 microglia (Zhao et al., 2016; Tang et al., 2018), which is associated with decreased Aβ clearance. In the present study, we first hypothesized that restraint/isolation stress might prime and polarize microglia to the M1 phenotype and reduce its clearing ability. Second, we aimed to probe the effects of icariin in attenuating the inflammatory response and relieving the pathological changes of AD by targeting PPARγ activation in the hippocampus and PFC of APP/PS1 double-transgenic mice.

### MATERIALS AND METHODS

### Animals and Experimental Procedure

Thirty male APP/PS1 mice, aged 3 months and weighing 24–30 g, obtained from Beijing HFK Bioscience Co., Ltd. (Beijing, China), were housed (five per cage) and maintained under a 12-h lightdark cycle, at 20–24◦C with free access to food and water. After 7 days of adaptation, the mice were randomly divided into three groups (10 mice per group): a control group, a restraint/isolation stress group (RIS), and a restraint/isolation stress + icariin group (RIS+ICA). Mice in both the RIS and RIS+ICA groups were subjected to 28 days of restraint/isolation stress from 3 to 4 months of age. The mice in the RIS+ICA group received daily administration of icariin (60 mg/kg) for 6 months from 4 to 10 months of age. The mice weight was recorded at baseline, every week during the stress procedure, and every month during drug administration. Behavioral tests after stress included the sucrose preference test (SPT), open field test (OFT), elevated plus maze test and Y maze test. The Morris water maze (MWM) test was used to measure learning and memory after drug administration at 10 months of age. Enzyme-linked immunosorbent assay (ELISA) and immunohistochemistry were used to detect Aβ plaque accumulation and microglia phenotypes, whereas western blotting was used to detect the expression of PPARγ in the hippocampus and the PFC. The experimental paradigm is presented in **Figure 1**.

### Restraint/Isolation Stress and Drug Treatments

Restraint/isolation stress is a paradigm used to build depressive animal models (Carroll et al., 2011). Each mouse was individually restrained in restraint tubes for 6 h per day, and the mice were deprived of water and food during the restraint/isolation stress.

Icariin (Cat no. I8760, Beijing Solarbio Science & Technology Co., Ltd.) was orally administered daily to the mice in the RIS+ICA group (60 mg/kg, freshly suspended in ddH2O) from 4 to 10 months of age (Wang et al., 2013; Li et al., 2015; Hoeijmakers et al., 2017). The mice in the control group and restraint/isolation stress group were given ddH2O (vehicle) in the same volume to balance the systematic error.

### Sucrose Preference Test (SPT)

The SPT test was carried out at the end of the exposure to restraint/isolation stress. A decreased sucrose preference is considered to be homologous to anhedonia, the inability to experience pleasure, and thus simulates the defining symptom of major depression. Mice were individually housed during the SPT. Before the test, the mice were trained to adapt to sucrose solution (1%, w/v): Two bottles of sucrose solution were placed in each cage for 24 h, then one bottle of sucrose solution was replaced with pure water for 24 h. After adaptation, mice were deprived of water and food for 24 h. Then, each mouse was given free access to two bottles for 1 h, one with 200 ml of 1% (w/v) sucrose solution and the other with 200 ml of pure water. The drinking bottles were weighed to calculate fluid consumption 1 h later. The sucrose preference percentage was evaluated as the amount of sucrose solution consumed during consumption of all fluids.

### Open Field Test (OFT)

The OFT was used to test the motivation of mice to explore. The open field apparatus is a white wooden box (90 cm in diameter and 45 cm in wall height). The bottom of the apparatus is divided into 25 squares. The mice were individually placed at the center of the apparatus and left free to explore the arena for 5 min. The following indices were recorded: number of grid crossings (horizontal movement), defined as crossing into the nearby grids with more than three paws or half of the body; and number of rears (vertical movement), defined as both forelimbs raised at least 1 cm above the ground. The box was thoroughly cleaned between tests. Horizontal and vertical movements were recorded using a camera linked with a computer fitted with a SMART video tracking system (SMART v3.0, Panlab, Spain).

### Elevated Plus Maze (EPM)

The experimental apparatus consisted of a central part (5 cm× 5 cm), two opposing open arms (30 cm × 5 cm) and two opposing closed arms (30 cm × 5 cm) with non-transparent walls of 15 cm height. The maze was elevated 50 cm above the floor. The mice were individually placed in the center of the maze facing an open arm and allowed free exploration for 5 min. The number of entries into the open and closed arms and the total time spent in the open and closed arms were recorded and measured by the SMART video tracking system. The time and entry ratios were used to obtain the anxiety score. The time ratio was defined as the ratio of total time spent in the open arms of the maze to the total time spent in any arm. The entry ratio was defined as the ratio of the number of entries into the open arms of the maze to the total number of entries into any arm of the maze. An anxiety score

After drug administration, the Morris water maze (MWM) test was performed. Enzyme-linked immunosorbent assay and immunohistochemistry were used to assess the microglia phenotype and western blotting was used to measure the expression of PPARγ in the hippocampus and PFC.

was calculated as 1 − (time ratio + entry ratio/2). Anxious mice were more likely to stay in the closed arms, so that a smaller time ratio or entry ratio indicates more anxious behavior. The maze platforms and walls were thoroughly cleaned with 75% ethanol between test sessions and allowed to dry.

## Y Maze Test

The Y maze test was used to assess various parameters related to spatial memory. The apparatus consisted of three arms (50 cm × 16 cm × 32 cm) made of black non-transparent plastic to form a "Y" shape. Visual cues made of colored paper were placed on the walls of the arms. The floor of the maze was covered with padding. The mice were placed into one of the arms of the maze (start arm) and allowed to explore the maze with one of the arms closed for 15 min (training trial). One hour later, the mice were returned to the start arm, and allowed to explore all three arms of the maze freely for 5 min (test trial). At the end of the training trial, the mice were returned to their home cage and the padding inside the maze was mixed to reduce the possibility of odor interference. The number of entries, the time spent and the distance traveled in each arm were measured by the SMART video tracking system. Ratio time was defined as the total time spent in the novel arm divided by the total time spent in any arm of the maze. Ratio entry was defined as the number of entries into the novel arm divided by the total number of entries into any arm of the maze. Ratio distance was defined as the total distance covered in the novel arm divided by the total distance covered in any arm of the maze. Due to their natural curiosity, mice are more likely to explore the novel arm, so that these parameters reflect spatial recognition memory. More entries or exploration of the novel arm indicate better spatial memory.

## Morris Water Maze (MWM)

The MWM test was performed to evaluate spatial reference learning and memory of APP/PS1 mice at 10 months old. The maze was a blue pool (0.8 m in diameter) filled with water (0.3 m deep, at 25 ± 1 ◦C). Geometric pictures pasted on the surrounding walls were used by the mice for spatial orientation. The maze was divided into four equal quadrants corresponding to four directions: I, II, III, and IV. The mice movements were captured by a CCD camera connected with a computer. All mice were allowed to swim freely for 60 s within 24 h before the formal training. During the following 5 consecutive days, mice were trained to find a platform (12 cm in diameter) hidden under the water surface in quadrant IV four times per day. If a mouse failed to find the platform within 60 s, it was manually guided to the platform and allowed to remain there for 10 s, and the escape latency was scored as 60 s. On the sixth day, all mice were released into the maze, without the hidden platform, from an identical point of quadrant I and allowed to swim freely for 60 s. The escape latency of training days as well as the percentage of time spent in and the number of entries into quadrant IV on the sixth day were analyzed with the SMART video tracking system.

## Protein Isolation

Mice (six per group) were sacrificed under deep anesthesia, and both sides of the frontal and parietal bones were removed to obtain the whole brain from the cranial cavity. The PFC and the hippocampus were collected and immediately immersed in liquid nitrogen, and stored at −80◦C for later protein isolation. The tissue was dissociated using an ultrasonic cell disruptor and lysed in a cold lysis buffer containing 10 mM Tris-HCl, pH 8.0, 240 mM NaCl, 5 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethanesulfonyl fluoride, 1% Triton X-100, 1 mM sodium vanadate and 1 g/ml of leupeptin, pepstatin and aprotinin. Tissue lysates were incubated at 4◦C for 20 min. The samples were centrifuged at 12,000 rpm for 10 min at 4◦C, then for each sample the supernatant was collected and protein content was determined using BCA protein assay reagents (Beyotime Institute of Biotechnology, China).

### Enzyme-Linked Immunosorbent Assay (ELISA)

The levels of IL-1β, IL-6, TNF-α, IL-4, IL-10, TGF-β1, NF-κB component p65, and Aβ 1-42 were measured using commercially available ELISA kits according to the manufacturer's instructions (Beijing Andy Huatai Technology Co., Ltd., China). Briefly, serial dilutions of protein standards and samples were added to 96 well ELISA plates followed by HRP labeled antibodies for IL-1β, IL-6, TNF-α, IL-4, IL-10, TGF-β1, NF-κB p65, and Aβ 1-42 to form antibody-antigen-enzyme labeled antibody complexes. After complete washing with wash solution, TMB substrate solution was added, which under the catalysis of HRP was converted to blue. The reaction was stopped using the stop solution. Optical density at 450 nm was detected using the iMark Microplate Absorbance Reader (Bio-Rad, United States). The concentration of each sample was calculated from the linear equation derived from the standard curve of known concentrations of NF-κB p65, Aβ 1-42, and the cytokines.

### Western Blotting

Brain protein samples containing the same amount of total protein were mixed with a 5x Laemmli loading buffer (protein volume: loading buffer = 4: 1). The mixed protein sample was heated at 99◦C for 5 min to achieve protein denaturation, then 15 µg of protein sample was separated on 10% sodium dodecyl sulfate-polyacrylamide (SDS-PAGE) gel and electro-transferred to polyvinylidene difluoride (PVDF) membranes (Servicebio, China). The membrane was blocked with 5% skim milk in TBS containing 0.1% Tween-20 (TBST) for 1.5 h and incubated in a refrigerator with primary antibodies against PPARγ (1:1000, Abcam, United States) or GAPDH (1:10000, Biogot Technology Co., Ltd., China) at 4◦C overnight. The following day, after washing with TBST three times for 5 min, the PVDF membrane was incubated for 1 h with the secondary antibody (ZSGB-BIO, China). Then, the PVDF membrane was washed again with TBST three times for 15 min; the western blots were visualized after being incubated with ECL solution (Millipore Corp., United States) for 1 min and exposed onto photographic films (Eastman Kodak Company, United States) for 10–90 s. Signal intensities were quantified using the ImageJ 14.0 software, and the intensity value of the protein band of interest was normalized according to that of the GAPDH band of the same sample.

### Immunohistochemistry

The mice (four per group) were anesthetized with pentobarbital and perfused with 50 ml of 0.1 M PBS, followed by 100 ml of ice-cold 4% paraformaldehyde (PFA). Paraffin-embedded PFC and hippocampus tissues were cut into 5 µm thick sections with a microtome. Coronal sections of the hippocampus were selected between bregma −1.22 mm and bregma −3.64 mm, and sections of the PFC were selected between anterior +4.7 mm to bregma posterior +3.7 mm to bregma. After rehydration, the sections were heated at 95–98◦C in 0.01 M citrate buffer (pH 6.0) for 15 min, then cooled to room temperature for 30 min and washed with PBS. Thereafter, the sections were incubated with fresh 3% H2O<sup>2</sup> for 25 min at room temperature to block endogenous peroxidase activity. After washing with PBS, the sections were blocked with 3% BSA for 30 min, incubated with mouse anti-Aβ 1-42 antibody (1:1000, BioLegend, United States), rabbit anti-ionized calciumbinding adapter molecule 1 (Iba-1) antibody (1:2000, Servicebio, China) or rabbit anti-inducible nitric oxide synthase (iNOS) antibody (1:100, Abcam, United States) at 4◦C overnight. After washing with PBS, the sections were incubated with HRP labeled anti-mouse or anti-rabbit antibodies at room temperature for 50 min, rewashed with PBS and incubated in diaminobenzidine (DAB) (Servicebio, China) for chromogen development under a microscope. Finally, the sections were dehydrated, cleared and mounted. After completing the staining, each section was viewed under 400× magnification. The expression of Aβ and the positive reactive cells of Iba-1 and iNOS were observed. Sections from corresponding locations were selected in each group, and five different visual fields (50 µm × 50 µm) of expression areas in the hippocampus [CA1, CA3, and dentate gyrus (DG)] and PFC were selected in each section under a high power lens to count Iba-1- and iNOS-positive cells.

### Statistics

Data are presented as mean ± standard error of mean (SEM). Statistical analysis was carried out by one-way analysis of variance (ANOVA) with Student–Newman–Keuls (SNK) post hoc test. Differences were considered statistically significant if the p-value was less than 0.05.

## RESULTS

### Effect of RIS on Behavioral Changes in SPT, OFT, EPM, and Y Maze Test Sucrose Preference Test

As shown in **Figure 2A**, stress exposure significantly reduced the percentage of sucrose consumption in stressed APP/PS1 mice in both the RIS and the RIS+ICA groups compared with the control animals [F(2,27) = 5.664, p = 0.009; post hoc p < 0.05 for both comparisons]. The results indicated that stress decreased sucrose preference, considered a major symptom of depression, in APP/PS1 mice.

### Open Field Test

As shown in **Figure 2B**, the APP/PS1 mice in the two stress groups showed decreased horizontal and vertical movements compared with the control group [F(2,27) = 10.093, p = 0.001; post hoc p < 0.01 for both comparisons]. The results indicated that stress induced decreased interest in a new environment and behavioral suppression in APP/PS1 mice.

### Elevated Plus Maze

As shown in **Figure 2C**, 4 weeks of exposure to RI stress did not significantly change the anxiety score of mice in the two stress groups [F(2,27) = 0.183, p = 0.834], indicating that the level of stress was too mild to induce anxiety-like behaviors.

mean ± SEM, n = 10. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 vs. the control group.

### Y Maze Test

As shown in **Figure 2D**, 4 weeks of exposure to RI stress did not significantly reduce the Ratio entry [F(2,27) = 0.101, p = 0.904], indicating the absence of memory impairment in stressed animals.

### Effect of Icariin on Learning and Memory in MWM

As shown in **Figure 3A**, mice in the RIS group had longer escape latency on the first, second, and third learning day, but there were no significant differences in escape latency at any day among the three groups [F(2,27) = 1.210, p = 0.314 (first day); F(2,27) = 0.902, p = 0.418 (second day); F(2,27) = 0.728, p = 0.492 (third day); F(2,27) = 0.419, p = 0.662 (fourth day) and F(2,27) = 0.146, p = 0.865 (fifth day)]. However, as shown in **Figures 3B–D**, mice in the RIS group spent less time in the target quadrant in the MWM test than the control group did [F(2,27) = 6.849, p = 0.004; post hoc p < 0.01]. Moreover, RIS mice showed decreased entry ratio in the target quadrant in comparison with the control mice [F(2,27) = 8.148, p = 0.002; post hoc p < 0.01]. Icariin administration significantly relieved these behavioral alterations in the RIS+ICA group as compared with the RIS group (p < 0.01, p < 0.05, respectively). The results indicated that long-term icariin administration effectively relieved the memory impairment of restraint/isolation-stressed mice.

### Effect of Icariin on Cytokines and NF-κB in the Hippocampus and PFC

Cytokines and NF-κB Expression in the Hippocampus Cytokine expression in the hippocampus is shown in **Figures 4A–F**. **Figure 4A** shows that IL-1β levels increased in the RIS group compared with the control group [F(2,15) = 32.507, p < 0.001; post hoc p < 0.01]; however, icariin relieved the alteration (p < 0.01). **Figure 4B** shows that RIS significantly increased IL-6 levels [F(2,15) = 69.126, p < 0.001; post hoc p < 0.01] compared with controls, while icariin decreased them (p < 0.01). **Figure 4C** shows that stress increased TNF-α in APP/PS1 mice [F(2,15) = 10.718, p = 0.001, post hoc p < 0.01], while icariin could reverse this change (p < 0.01).

**Figure 4D** shows that IL-4 levels decreased in the RIS group compared with controls [F(2,15) = 74.971, p < 0.001; post hoc p < 0.01], and icariin relieved such alteration (p < 0.01). Similarly, **Figure 4E** shows that RIS significantly decreased IL-10 levels compared with controls [F(2,15) = 119.031, p < 0.001;

compared with the RIS group. The results are expressed as the mean ± SEM, n = 10. ∗∗p < 0.01 vs. control group; #p < 0.05, ##p < 0.01 vs. RIS group.

post hoc p < 0.01], while icariin increased them (p < 0.01). **Figure 4F** shows that stress decreased TGF-β1 expression in APP/PS1 mice compared with controls [F(2,15) = 82.481, p < 0.001; post hoc p < 0.01], while icariin could reverse such change (p < 0.01).

**Figure 4G** shows that NF-κB levels increased in the RIS group compared with controls [F(2,15) = 16.709, p < 0.001: post hoc p < 0.01]; however, icariin relieved the alteration (p < 0.01).

### Cytokines and NF-κB Expression in the PFC

Cytokine expression in the PFC is shown in **Figures 5A–F**. **Figure 5A** shows that IL-1β levels increased in the RIS group compared with controls [F(2,15) = 17.565, p < 0.001; post hoc

FIGURE 6 | Effect of icariin on Aβ plaques and microglia phenotype in the hippocampus. (A–C) The immunohistochemical staining results of Aβ, Iba-1, and iNOS (400×). (D) RIS significantly increased the levels of Aβ in the RIS group compared with the controls, and administration of icariin decreased its levels in the RIS+ICA group. (E) The number of Iba-1+ microglia positive cells was not significantly different among the three groups. (F) The number of iNOS+ microglia positive cells in the hippocampus of stressed mice was higher than that of control mice, and icariin administration reversed the change in the RIS+ICA group. (G) The RIS group showed higher ratio of iNOS+ to Iba-1+ microglia, while the RIS+ICA group showed a lower ratio. The results of Aβ are expressed as the mean ± SEM, n = 6; the results of Iba-1 and iNOS are expressed as the mean ± SEM, n = 4. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 vs. control group; #p < 0.05, ##p < 0.01 vs. RIS group.

p < 0.01]; however, icariin relieved the alteration (p < 0.01). **Figure 5B** shows that RIS significantly increased the level of IL-6 [F(2,15) = 51.672, p < 0.001; post hoc p < 0.01], which was reversed by treatment of icariin (p < 0.01). **Figure 5C** shows that stress increased TNF-α expression compared with controls [F(2,15) = 16.475, p < 0.001; post hoc p < 0.01], while icariin could reverse this change (p < 0.01).

**Figure 5D** shows that IL-4 levels decreased in the RIS group compared with controls [F(2,15) = 25.263, p < 0.001; post hoc p < 0.01], and icariin failed to relieved this alteration. **Figure 5E** shows that RIS also significantly decreased the level of IL-10 [F(2,15) = 41.906, p < 0.001; post hoc p < 0.01], which was enhanced by the administration of icariin (p < 0.01). **Figure 5F** shows that stress decreased TGF-β1 compared with controls [F(2,15) = 23.709, p < 0.001; post hoc p < 0.01], while icariin could reverse this change (p < 0.01).

**Figure 5G** shows that the level of NF-κB increased in the RIS group compared with controls [F(2,15) = 37.217, p < 0.001; post hoc p < 0.01]; however, icariin relieved the alteration (p < 0.01).

### Effect of Icariin on Aβ Plaques and Microglia Phenotype in the Hippocampus and PFC

#### Aβ Plaques and Microglia Phenotype in the Hippocampus

As shown in **Figure 6A**, the amount and size of Aβ plaques in the hippocampus of the RIS APP/PS1 mice were larger than those in the mice of the control group, while icariin relieved the alteration in the RIS+ICA group. To accurately evaluate the changes in Aβ, we measured the Aβ level in the hippocampus using ELISA, as shown in **Figure 6D**. RIS significantly increased Aβ levels in the RIS group compared with the control group [F(2,15) = 74.140, p < 0.001; post hoc p < 0.01], while administration of icariin decreased the levels of Aβ in the RIS+ICA group (p < 0.01). The counting of Iba-1+ microglia positive cells did not reveal significant differences among the three groups as shown in **Figures 6B,E** [F(2,9) = 0.395, p = 0.685]. However, the number of iNOS+ microglia positive cells in the hippocampus of the stressed mice was higher than that in control mice [F(2,9) = 6.640, p = 0.017; post hoc p < 0.05], and icariin administration reversed the change in the RIS+ICA group (p < 0.05, **Figures 6C,F**). The higher ratio of iNOS+ microglia to Iba-1+ microglia in the stressed group and the lower ratio in the RIS+ICA group further support the promotion by icariin of a microglia phenotype switch in the hippocampus of RIS APP/PS1 mice [F(2,9) = 7.837, p = 0.011; post hoc p < 0.01, control vs. RIS; p < 0.05, RIS vs. RIS+ICA] (**Figure 6G**).

### Aβ Plaques and Microglia Phenotype in the PFC

As shown in **Figure 7A**, the amount and size of Aβ plaques in the PFC of RIS APP/PS1 mice were larger than that in the PFC of control mice, while icariin relieved such alterations. In order to accurately evaluate the change of Aβ, Aβ levels were also measured in the PFC using ELISA as shown in **Figure 7B**. Compared with controls, RIS significantly increased Aβ levels [F(2,15) = 4.368, p = 0.032; post hoc p < 0.05], while icariin decreased them (p < 0.05). The counting of Iba-1+ and iNOS+ microglia cells showed no significant differences among the three groups, as shown in **Figures 7C,D** [F(2,9) = 0.922, p = 0.432; F(2,9) = 1.502, p = 0.274, respectively]. As a result, the ratio of iNOS+ to Iba-1+ microglia also was not significantly different among the three groups [F(2,9) = 0.229, p = 0.800] (**Figure 7E**).

### Effect of Icariin on PPARγ in the Hippocampus and PFC

**Figure 8A** shows PPARγ expression in the hippocampus as assessed using western blotting. The RIS procedure induced a marked decrease in PPARγ levels in APP/PS1 mice compared with controls [F(2,9) = 7.917, p = 0.010; post hoc p < 0.05], and icariin could reverse such alteration (p < 0.05).

**Figure 8B** shows PPARγ expression in the PFC as assessed using western blotting. The RIS procedure induced a marked decrease in PPARγ levels in APP/PS1 mice compared with controls [F(2,12) = 3.916, p = 0.049; post hoc p < 0.05], and icariin could partially reverse the alteration (p > 0.05, control group vs. RIS+ICA).

### DISCUSSION

Mounting evidence has shown that chronic psychosocial stress is a major risk factor for late-onset AD, and is associated with cognitive deficits (Piirainen et al., 2017). Chronic stress could prime the microglia and induce inflammatory responses in the adult brain, thereby compromising the synapse-supportive roles of microglia and contributing to age-related cognitive deterioration. In order to examine the effects of chronic stress on neuroinflammation and neurodegeneration in AD, we used restraint/isolation stress to build a stressed AD model (Jeong et al., 2006; Carroll et al., 2011; Lee and Han, 2013). Consistent with previous studies, 1 month of restraint/isolation stress induced lower sucrose consumption and lower horizontal and vertical movements in the OFT in APP/PS1 mice aged 4 months, compared with unstressed APP/PS1 mice, indicating that stressed animals showed anhedonia and were less interested in a new environment, thus suggesting the validity of the stressed AD model.

The role of chronic neuroinflammation in AD-related dementia has recently gained attention, implicating an early involvement of inflammation in the onset and progression of the disease (Querfurth and LaFerla, 2010). Microglia are the first line of defense against invading pathogens in the CNS (Piirainen et al., 2017). Studies have confirmed that stress could stimulate the microglia and prime them toward the M1 phenotype, which is responsible for neuroinflammation (Zhao et al., 2016; Tang et al., 2018). In the present study, M1 phenotype markers including IL-1β, IL-6, TNF-α, and NF-κB, and M2 markers such as IL-4, IL-10, and TGF-β1 in the hippocampus and the PFC were measured in 10-month-old mice in the control, RIS, and RIS+ICA groups by ELISA (Michelucci et al., 2009; Park et al., 2015). The results showed that stressed APP/PS1 mice had higher levels of M1 markers and lower levels of M2 markers in both

group; #p < 0.05 vs. RIS group.

the hippocampus and the PFC, indicating that stress induced microglia M1 activation in APP/PS1 mice. We also assessed the levels of the M1 phenotype marker iNOS and its ratio to the microglia marker Iba-1. We found that the fraction of M1 microglia was higher in the hippocampus, but not in the PFC of stressed mice than in control mice. A possible explanation is that the hippocampus is the area most sensitive to stress, because it shows the highest expression of corticosterone receptors among brain regions (Anacker et al., 2011; Zhang et al., 2018). Indeed the stress-induced activation of inflammatory pathways in the brain was shown to be region-specific (Ogundele et al., 2017). Given that the microglia M2 activation state was associated with the suppression of inflammation, promotion of phagocytosis and tissue repair (Chawla, 2010; Chinetti-Gbaguidi et al., 2011), as well as characterized by increased Aβ clearance and enhanced tissue remodeling (Heneka et al., 2013), increased microglia M1 activation in stressed APP/PS1 mice resulted in lower clearance ability and higher neurotoxicity of the microglia (Park et al., 2015). Moreover, microglia M1 activation and the release of inflammatory cytokines increased the neuroinflammation of brain, which could result in AD progression (Mandrekar-Colucci et al., 2012; Huang et al., 2017). In the present study, we observed that RIS mice had worse memory impairment in the MWM, together with higher levels of M1 phenotype markers and lower level of M2 phenotype markers in both hippocampus and PFC at 10 months of age, while icariin could attenuate these alterations. The results are consistent with the conclusion that the imbalance of M1/M2 microglia activation is involved in AD pathogenesis (Mandrekar-Colucci et al., 2012; Huang et al., 2017).

Previous studies confirmed that the failure of microglia to clear abnormally accumulating Aβ peptide led to neuroinflammation and neurodegeneration in AD models in which RIS elevated Aβ40 and Aβ42 levels in the brain, accelerated amyloid plaque formation, and impaired learning and memory (Jeong et al., 2006; Carroll et al., 2011; Lee and Han, 2013). In our study, the stressed APP/PS1 mice had higher level of Aβ accumulation in both the hippocampus and the PFC compared with unstressed mice. More importantly, the stressed APP/PS1 mice showed more severe memory impairment in the MWM test at 10 months of age compared with unstressed mice. These findings support the previous conclusion that psychological stress induces lower ability to clear accumulating Aβ peptide, resulting in impaired cognition (Jeong et al., 2006; Hoeijmakers et al., 2017).

The PPARs are a group of nuclear receptor proteins regulating gene expression as ligand-activated transcription factors (Michalik et al., 2006). Studies demonstrated that PPARγ can mediate the conversion of the microglia into the M2 phenotype (Yamanaka et al., 2012; Saijo et al., 2013). Chronic stress induced lower PPARγ expression in the adipose tissue and PPARγ knockout mice displayed more anxiety-like behaviors (Domi et al., 2016; Guo et al., 2017). Moreover, studies have suggested that activation of PPARγ-mediated anti-inflammatory signaling might be a potential therapeutic strategy for AD (Vallee et al., 2017). Our results showed that PPARγ expression levels in the hippocampus and the PFC were significantly lower in stressed than in unstressed mice, which indicates that stress induced microglia M1 activation is via suppression of PPARγ expression.

Epimedium (family Berberidaceae), commonly known as horny goat weed in the West, is known as Yin Yang Huo in Chinese medicine. Icariin is a highly potent active ingredient of this herb, believed to be the source of its many potential health benefits (Lee et al., 1995; Lin et al., 2004). Icariin was proven to possess anti-bacterial and anti-inflammatory efficacy (Luo et al., 2007; Zhou et al., 2011). A previous study indicated that it improved spatial learning and memory abilities in lipopolysaccharide (LPS)-induced rat brain dysfunction through the inhibition of hippocampus IL-1β and cyclooxygenase-2

(COX-2) expression (Guo et al., 2010). Accumulating evidence indicates that, in addition to inhibiting the activation of innate immune cells (including microglia) producing TNF-α and IL-1β, icariin led to stable suppression of NF-κB signaling activation in the hippocampus (Liu et al., 2015). Thus, icariin plays a counter-regulatory role resulting in neuroinflammation suppression. Studies have investigated the pharmacological properties of icariin in the CNS and showed that it can markedly attenuate cognitive deficits in several models of AD (Li et al., 2010; Nie et al., 2010; Jin et al., 2014), and also alleviate neuronal injury induced by ischemia (Li et al., 2005). It was also reported that icariin can attenuate inflammatory response through PPARγ activation in rats (Xiong et al., 2016). However, the role of icariin in attenuating PPARγ alterations in the AD brain has not yet been clarified. The present study compared the expression of PPARγ in the hippocampus and the PFC between controls, RIS, and RIS+ICA mice and found that stress induced lower PPARγ expression in the RIS mice, while icariin relieved such alteration in the RIS+ICA mice. Moreover, the results about the microglia activation phenotype and Aβ accumulation demonstrate that icariin administration normalized the Aβ clearance ability by up-regulating PPARγ in the RIS+ICA mice. As a result, the memory impairment of RIS+ICA mice was milder than that of untreated stressed mice.

### CONCLUSION

Our results demonstrate that RIS induced depressive-like behaviors, spatial memory impairment, over-expression of M1 microglia markers and increased Aβ accumulation by suppressing PPARγ in APP/PS1 mice. Icariin, which induces a microglia phenotype switch through the activation of PPARγ, is a promising candidate for new therapeutic strategies. In conclusion, our study may provide novel insights into the role of chronic psychosocial stress in the pathogenesis and progression of AD and the development of new therapeutic approaches. Icariin affecting the microglia phenotype and cytokine release

### REFERENCES


through other pathways should be further studied both in vitro and in vivo.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

### ETHICS STATEMENT

All procedures used in the study were reviewed and approved by the Ethics Committee of School of Medicine, Shandong University, and comply with the National Institutes of Health's Guide for the Care and Use of Laboratory Animals (NIH publication no. 85-23, revised 1985). In the handling and care of all animals, we followed the international guiding principles for animal research, as stipulated by the World Health Organization (WHO) Chronicle (World Health Organization, 1985), as adopted by the Laboratory Animal Center, Shandong University.

### AUTHOR CONTRIBUTIONS

XL was involved in study design and data interpretation. YW performed the majority of the laboratory work and contributed to the analysis of data and the writing of the manuscript. TZ, MW, FZ, GZ, JZ, and YZ were responsible for the animal model and the behavioral tests. EW revised the manuscript. All authors approved the final version to be submitted.

### FUNDING

This work was supported by the National Natural Science Foundation of China (Grant No. 81373635), the Jinan Municipal Science and Technology Development Program (Grant No. 201212012), and the State Scholarship Fund by China Scholarship Council (Grant No. 201806220218).

and cognitive performance through a corticotropin-releasing factor receptordependent mechanism in a transgenic mouse model of tauopathy. J. Neurosci. 31, 14436–14449. doi: 10.1523/JNEUROSCI.3836-11.2011


response to acute stress and exacerbates anxiety: an effect reversed by rescue of amygdala PPARgamma function. J. Neurosci. 36, 12611–12623. doi: 10.1523/ JNEUROSCI.4127-15.2016


directly activates corticotropin-releasing factor neurons to exacerbate stress responses. J. Neurosci. 35, 2612–2623. doi: 10.1523/JNEUROSCI.3333-14.2015



**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 © 2019 Wang, Zhu, Wang, Zhang, Zhang, Zhao, Zhang, Wu and Li. 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.

# Natural Diterpenoid Oridonin Ameliorates Experimental Autoimmune Neuritis by Promoting Anti-inflammatory Macrophages Through Blocking Notch Pathway

Lu Xu<sup>1</sup> , Lei Li<sup>1</sup> , Chen-Yang Zhang<sup>1</sup> , Hermann Schluesener<sup>2</sup> and Zhi-Yuan Zhang1,2,3 \*

<sup>1</sup> Department of Pathology, Nanjing Medical University, Nanjing, China, <sup>2</sup> Division of Immunopathology of the Nervous System, Institute of Pathology and Neuropathology, University of Tübingen, Tübingen, Germany, <sup>3</sup> Department of Neurology, Sir Run Run Hospital, Nanjing Medical University, Nanjing, China

#### Edited by:

Pietro Giusti, University of Padova, Italy

### Reviewed by:

Harshini Sarojini, University of Louisville, United States Kalliopi Pitarokoili, Ruhr University Bochum, Germany

> \*Correspondence: Zhi-Yuan Zhang zzy@njmu.edu.cn; zhiyuan.zhang@ medizin.uni-tuebingen.de

#### Specialty section:

This article was submitted to Neuropharmacology, a section of the journal Frontiers in Neuroscience

Received: 05 November 2018 Accepted: 07 March 2019 Published: 02 April 2019

#### Citation:

Xu L, Li L, Zhang C-Y, Schluesener H and Zhang Z-Y (2019) Natural Diterpenoid Oridonin Ameliorates Experimental Autoimmune Neuritis by Promoting Anti-inflammatory Macrophages Through Blocking Notch Pathway. Front. Neurosci. 13:272. doi: 10.3389/fnins.2019.00272 The diterpenoid compound, Oridonin, extracted from the Chinese herb, Rabdosia rubescens, possesses multiple biological activities and properties. Oridonin exhibited efficient anti-inflammatory activity by inducing a switch in macrophage polarization to the anti-inflammatory phenotype through inhibition of the Notch pathway in our in vitro study; therefore, its potential therapeutic effects were further investigated in the animal model of human Guillain-Barré syndrome (GBS) and other polyneuropathies – experimental autoimmune neuritis (EAN). Either preventive or therapeutic treatments with Oridonin greatly attenuated disease peak severity, suppressed paraparesis, shortened disease duration, and even delayed EAN onset. Progression of neuropathic pain, demyelination, inflammatory cellular accumulations, and inflammatory cytokines in peripheral nerves were significantly attenuated. Meanwhile, accumulation of immune cells in the spinal roots and microglial activation in the lumbar spinal cord were also reduced. Interestingly, Oridonin treatment significantly increased the proportion of anti-inflammatory macrophages and made them locally dominant among all infiltrated macrophages in the peripheral nerves. The down-regulation of local Notch pathway proteins, together with our in vitro results indicated their possible involvement. Taken together, our results demonstrated that Oridonin effectively suppressed EAN by attenuating local inflammatory reaction and increasing the proportion of immune regulating macrophages in the peripheral nerves, possibly through blockage of the Notch pathway, which suggests Oridonin as a potential therapeutic candidate for human GBS and neuropathies.

Keywords: Guillain-Barré syndrome, experimental autoimmune neuritis, Oridonin, anti-inflammatory macrophage, Notch pathway

**Abbreviations:** CFA, complete Freund's adjuvant; CMC, carboxymethylcellulose; EAN, experimental autoimmune neuritis; FBS, fetal bovine serum; GBS, Guillain-Barré syndrome; GFAP, glial fibrillary acidic protein; HWT, hind-paw withdrawal threshold; IL-1β, interleukin-1β; iNOS, inducible nitric oxide synthase; IR, immunoreactivity; LFB, Luxol fast blue; LPS, lipopolysaccharide; MNCs, mononuclear cells; PNS, peripheral nervous system; SEM, standard errors of means; TBS, Trisbuffered saline; TNF, tumor necrosis factor.

### INTRODUCTION

fnins-13-00272 March 29, 2019 Time: 18:51 # 2

Guillain-Barré syndrome (GBS) is an immune-mediated polyneuropathy; the world's leading cause of acute autoimmune neuromuscular paralysis, caused by an autoimmune attack on the peripheral nervous system (PNS) (Willison, 2005; Kuwabara and Yuki, 2013). Existing treatments, including supportive managements, like advanced intensive care and assisted respiration, as well as some active treatments such as plasma exchange and intravenous immunoglobulin, are not satisfactory (Hughes, 2002). Not all GBS patients respond to these treatments, and even if they do, they still do not recover fully; and continuing weakness, motor units deprivation, and the problem of fatigue continues to be detected (Winer, 2001; Kieseier et al., 2004). Therefore, more effective, and convenient therapeutic approaches for the inflammatory peripheral demyelinating disease are needed. Experimental autoimmune neuritis (EAN), an autoantigen-specific, T cellmediated animal model for demyelinating inflammatory diseases of the PNS, is broadly accepted and used to study pathogenesis, pathological changes, and potential therapeutic approaches for human polyneuropathies like GBS, by mirroring many of their neurological, electrophysiological, immunological, and pathological features (Hughes and Cornblath, 2005).

Experimental autoimmune neuritis is actively induced by immunization with autoantigen-like purified myelin or P2 peptides. Following the breakdown of the blood-nerve barrier, reactivated T cells and macrophages are abundantly infiltrated and accumulated in the PNS, which results in a robust local inflammation, demyelination, and axon degeneration, and all of these are crucial for the progression of EAN (Kieseier et al., 2004; Zhang et al., 2008a). Modulating or controlling the activation of these immune cells is, thus, considered an important therapeutic strategy for EAN and human neuropathies. As the main infiltrating cell population and the major effector cell present during EAN progression, macrophages are responsible for most neuropathological changes due to their antigenpresenting property and direct phagocytosis of axon myelin (Kiefer et al., 2001; Maurer et al., 2002). Infiltrating macrophages also cause tissue damage by secretion of inflammatory cytokines such as interleukin-1β (IL-1β) and tumor necrosis factor (TNF) α, and these are responsible for demyelination and axon loss as well (Kiefer et al., 2001). Inhibition of macrophage activity or their depletion was proved to prevent EAN, as observed through all the clinical, electrophysiological, and histological changes (Craggs et al., 1984). However, recruited macrophages can be broadly divided into two particular phenotypes: classically activated macrophages that promote inflammatory reaction and tissue damage and alternatively activated macrophages with anti-inflammatory activity and tissue-repairing effects (Mosser and Edwards, 2008; Ricardo et al., 2008). Pro-inflammatory macrophages execute host-defense against infection, but also cause inflammatory tissue damage and lead to the high expression of pro-inflammatory cytokines such as IL-12, IL-23, and TNF-α and inducible nitric oxide synthase (iNOS). Anti-inflammatory macrophages are considered to control inflammatory reactions and are responsible for tissue repair; they are characterized by the high expression of anti-inflammatory molecules such as IL-10 and arginase-I (Mosser and Edwards, 2008; Ricardo et al., 2008). It has been previously reported that a phenotypic switch of macrophage polarization to the anti-inflammatory state can attenuate pathological progression and even determine the outcome of EAN (Zhang Z. Y. et al., 2010; Han et al., 2016a,b).

Plant-derived compounds have gained increased attention as nutraceutical treatments for various diseases, including cardiovascular diseases, mental disorders, cancers, and inflammatory diseases (Kumar, 2006). Oridonin, a natural diterpenoid compound extracted from a Chinese herb Rabdosia rubescens (Zhang Z. et al., 2010), exhibits numerous biological activities and effects, including oxygen free-radical clearing, anti-mutagenic, and anti-microbial activities, and is recently being studied intensively for promising anti-neoplastic activity (Ji et al., 2011, Jia et al., 2012). Moreover, its anti-inflammatory activities (He et al., 2018; Zhou et al., 2018), especially antineuroinflammatory and neuroprotective effects, have been reported recently in different in vitro and in vivo studies (Liu et al., 2007; Xu et al., 2009; Chen et al., 2012), which indicates its therapeutic potential in treating inflammatory disorders of the nervous system. As a widely used Traditional Chinese Medicine (TCM) and a main ingredient of many commonly available over-the-counter (OTC) herbal medicines in China, R. rubescens and its aqueous extract have shown excellent tolerance and oral bioavailability; therefore, the proven safety and reliability of the herb further ascertains its promising potential in clinical applications (Chen et al., 2009; Ma et al., 2011; Cheng et al., 2017).

Notch signaling is initiated by the Notch receptor-ligand binding (Mailhos et al., 2001). Typical signaling proteins such as Notch1, Jagged-2, and Hes-1 are usually tested for the activation or inhibition of Notch signaling, and were also investigated in our study. After Oridonin showed efficient anti-inflammatory activity and induced a phenotypic switch of macrophage polarization toward the anti-inflammatory state in our cell culture, the potential therapeutic effects of Oridonin were further investigated in the rat model of EAN, as we hypothesized that the possible effects might be connected to the regulation of immune cells, especially macrophages polarization.

### MATERIALS AND METHODS

### Animals

Male Lewis rats (10–12 weeks, 230–250 g, Charles River, Sulzfeld, Germany) were housed under a 12 h light-12 h dark cycle with free access to food and water. The guideline EU Directive 2010/63/EU for animal experiments was followed and all animal procedures were approved by the local Administration District Official Committee. All efforts were made to minimize the number of animals and their suffering.

### Isolation of Primary Peritoneal Macrophages and Cell Culture

The rats were treated with 4 ml thioglycollate medium intraperitoneally. Seven days later, peritoneal cells were obtained

by peritoneal lavage with 12 ml of ice-cold PBS (phosphatebuffered saline) containing 2% FBS (fetal bovine serum) (Invitrogen, Karlsruhe, Germany). FITC-conjugated anti-CD68 antibodies (Serotec, Oxford, United Kingdom) were used to confirm the purity of the macrophages and more than 90% of the peritoneal cells were macrophages.

Primary peritoneal macrophages and a common murine macrophage cell line, RAW 264.7, (Sigma-Aldrich, Munich, Germany) were both used to evaluate the effects of Oridonin on the inflammatory reaction and polarization of macrophages in vitro. Primary macrophages and RAW cells were maintained in complete RPMI 1640 media, from Gibco (Invitrogen, Karlsruhe, Germany), containing 10% FBS (Invitrogen, Karlsruhe, Germany), penicillin (100 U/ml), and streptomycin (100 U/ml). 10<sup>5</sup> cells were seeded into 12-well plates. After 24 h of culture, lipopolysaccharide (LPS) (Sigma-Aldrich, Munich, Germany) was added in some wells and co-cultured for another 24 h; some groups were treated with Oridonin (0.25, 1, or 4 µM). A previous pharmacokinetic study (Xu et al., 2006) reported that the plasma concentrations of Oridonin after oral dosing of 20 mg/kg or 40 mg/kg in rats were between 0.1 and 0.35 µg/ml for over 10 h, which were equivalent to 0.25 to 0.95 µM. Our in vitro study, therefore, used working concentrations of 0.25, 1, and 4 µM. Further, some groups of cells were incubated for 48 h with 30 µM DAPT (Sigma-Aldrich, Munich, Germany), an inhibitor of the Notch signaling pathway. Thereafter, supernatants from different wells were collected and a standard Griess assay (Sigma-Aldrich, Munich, Germany) was performed to analyze the production of nitric oxide (NO). Total RNA was extracted from cells with the RNeasy Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer's protocol. QuantiTect Reverse Transcription Kit (QIAGEN, Hilden, Germany) was utilized to reverse transcribe RNA (1 µg) into cDNA. Real-time PCR analysis was used to measure the mRNA expression. ELISA kits for IL-1β (Thermo Scientific, Waltham, MA, United States), TNF-α, and IL-10 (Bio legend Inc., San Diego, CA, United States) were used to detect the concentrations of the cytokines in macrophage culture supernatants.

### Transient Transfection

The Notch1-NICD (Notch intracellular domain) was generated using the following primers: sense 5<sup>0</sup> - CGCGGATCCATGCACCTGGATGCCGCTGACCTG-3<sup>0</sup> and antisense 5<sup>0</sup> -ACGTCTAGACTYGAAGGCCTCCGGAATGCG-3 0 . The pair of primers were designed for Notch1-NICD gene sequences from the GenBank database, and the fragment of Notch1-NICD was cloned into the pcDNA3.1 vector (Invitrogen, Karlsruhe, Germany). Lipofectamine 2000 (Invitrogen, Karlsruhe, Germany) was used to conduct transfection based on the manufacturer's protocols. The Real-time-PCR was conducted to monitor the transfection efficiency. RAW cells from the other comparable groups, including the control group, were all transiently transfected with a plasmid vehicle.

### EAN Induction and Oridonin Treatment

After dissolving in PBS at a concentration of 2 mg/ml, the synthetic neurogenic peptide, P2 57-81 (GeneScript Corporation, Scotch Plains, NJ, United States) was emulsified with 2 mg/ml Mycobacterium tuberculosis mixed within Complete Freund's Adjuvant (CFA) (Sigma-Aldrich; Munich, Germany), and the final concentration of peptide in the emulsion was 1 mg/ml. Each rat was injected subcutaneously at the base of the tail with 100 µl emulsion containing 100 µg peptide.

Once a day, neurological signs of EAN were recorded and graded as follows: normal (score = 0), reduced tonus of tail (score = 1), impaired righting, limp tail (score = 2), absent righting (score = 3), gait ataxia (score = 4), mild paresis of the hind limbs (score = 5), moderate paraparesis (score = 6), severe paraparesis or paraplegia of the hind limbs (score = 7), tetraparesis (score = 8), moribund (score = 9), and death (score = 10).

Four rounds of Oridonin treatments were performed: two each of therapeutic and preventive treatments. EAN rats received daily gavages of Oridonin from Day 0 to Day 14 as the preventive treatment or from Day 9 to Day 14 as the therapeutic treatment (six rats/group); rats were prepared for tissue analysis on Day 15. In addition, another three groups of rats administered identical treatments as described were observed until they fully recovered to cover the entire spectrum of disease development and recovery. Oridonin, suspended in 1% carboxymethylcellulose (CMC) (Blanose <sup>R</sup> , Hercules-Aqualon, Düsseldorf, Germany), was fed by gavage once a day at doses of 7 or 20 mg/kg of body weight.

It was reported that 175 mg of Oridonin administered orally to humans, which is a common dosage prescribed for inflammatory conditions in TCM, was comparable to the dose of 13.5 mg/kg administered to rats (Xu et al., 2006). The oral dosing of 20 mg/kg was comparable to the treatments of lupus-like symptoms (Zhou et al., 2013) and neuroinflammation of cerebral amyloidosis (Zhang et al., 2013) in mice. A daily dose of 10 mg/kg of body weight administered by the intraperitoneal (i.p.) route was reported previously to treat animal models of rats (Li et al., 2018). A previous pharmacokinetic study of Oridonin indicated that the oral absolute bioavailability of Oridonin was rather low and it was approximately 1/2–1/3 that of i.p. treatments (Xu et al., 2006). The same pharmacokinetic study further reported that the plasma concentrations of Oridonin after oral dosing of 20 mg/kg or 40 mg/kg in rats were between 0.25 and 0.95 µM, which is comparable to the working concentrations used in our in vitro study. All these reports indicated that the oral dose of 20 mg/kg to EAN rats is comparable to the dosage reported in most published studies and for human treatments in TCM with the herb, R. rubescens. Additionally, we tested the dose of 7 mg/kg (1/3 of the 20 mg/kg dose) in the preventive treatment round to observe the dose-dependent effect.

### Mechanical Allodynia

Mechanical allodynia was defined by the reduction of the hindpaw withdrawal threshold (HWT), which was measured using a mechanical plantar test apparatus, namely, an automatic von Frey system (Ugo Basile, Milan, Italy). All rats were trained every day, starting six days before the first HWT measurement. HWT was examined between 10:00 and 14:00 every day from two days prior to immunization to 14 days post-immunization. Briefly, rats were habituated for 10 min and acclimated for another 10 min. The mechanical force (from 0 to 50 g over a period of 15 s) was exerted onto the middle of the hind-paws using a fine metal filament. Left and right hind-paws were measured 10 times each, one after the other. The forces under which the rats actively lifted their paws on their own initiative were recorded and mean values were calculated. All EAN rats for mechanical allodynia assays were re-grouped before immunization according to the measurement results of the training sessions to stratify the experimental groups.

### Immunohistochemistry

fnins-13-00272 March 29, 2019 Time: 18:51 # 4

On Day 15, after deep anesthesia and intracardial perfusion with 4% paraformaldehyde at 4◦C, the rats were prepared for analysis. To detect the infiltration of inflammatory cells and the levels of pathological changes in the PNS, sciatic nerves from both the legs and the lumbar spinal cord were removed quickly and fixed overnight in 4% paraformaldehyde at 4◦C. Each specimen was cut into two equal long segments, embedded in paraffin, serially sectioned (3 µm), and mounted on slides.

Sections were dewaxed and boiled in a microwave oven (600 W) in citrate buffer (sodium citrate 2.1 g/L, pH 6) for 15 min. Methanol containing 1% H2O<sup>2</sup> was used to inhibit endogenous peroxidase for 15 min. After blockage with 10% pig serum (Biochrom, Berlin, Germany), cross sections were incubated with monoclonal antibodies as follows: CD3 (1:50; Serotec, Oxford, United Kingdom) for T cells, OX22 (1:200; Serotec, Oxford, United Kingdom) for B lymphocytes, glial fibrillary acidic protein (GFAP) (1:500; Chemicon International, Temecula, CA, United States) for astrocytes. Activated macrophages/microglia and anti-inflammatory macrophages were detected by ED1 (a lysosomal membrane protein which recognizes CD68) (1:100; Serotec, Oxford, United Kingdom) and ED2 (1:100; Serotec, Oxford, United Kingdom), respectively. Biotinylated IgG F(ab)2 (DAKO, Hamburg, Germany) was used as secondary antibody fragment to visualize the antibodies binding to sections. A horseradish peroxidase-conjugated streptavidin complex (DAKO, Hamburg, Germany) and diaminobenzidine (DAB) substrate (Fluka, Neu-Ulm, Germany) was subsequently added to sections. Tissues were counterstained with Maier's Hemalum in the end.

The ratio of immunoreactivity (IR) areas to sciatic nerve crosssection areas were calculated as described previously (Zhang et al., 2008a). Briefly, micro-images of staining were captured with a Nikon Cool-scope (Nikon, Düsseldorf, Germany); MetaMorph Offline 7.1 (Molecular Devices, Toronto, ON, Canada) was applied on the micro images to outline and analyze the cross sections of the sciatic nerves. In regions of interest, the percentages of specific IR were selected by color threshold segmentation and analyzed. For each staining, all parameters were fixed for all the images and no automatic adjustments were used. Results are shown as the arithmetic means of percentages of IR area to interest areas on cross-sections and standard errors of means (SEM).

Luxol Fast Blue (LFB) (Sigma-Aldrich; Munich, Germany) staining was used to visualize myelin. For quantification of the demyelination of impacted nerves, the software MetaMorph Offline 7.1 (Molecular Devices, Toronto, ON, Canada) was used to analyze the relative optical density of the nerves from all groups. Raw images of nerve cross sections were acquired at the same exposure level and converted to 8-bit gray scale files. For removing the image background, the mean density calculated from the threshold pixels.

Furthermore, between Oridonin-treated and control EAN rats, histological alterations were analyzed by a semi-quantitative method established previously (Zhang et al., 2008a). In short, cross-sections of middle levels of the left and right sciatic nerves from EAN rats were evaluated. All cross-sections were collected and perivascular areas were evaluated by two pathologists. The levels of pathological changes were semi-quantitatively graded as: normal perivascular area (0); mild inflammatory cellular infiltration in immediate proximity to the blood vessels (1); cellular infiltration and demyelination adjacent to the blood vessels (2); and cellular infiltration plus demyelination throughout the cross-sections (3). Results are shown as mean histological scores (Hartung et al., 1988).

### Tissue Preparation, RNA Isolation, Reverse Transcription, and Semi-Quantitative PCR

Experimental autoimmune neuritis rats were deeply anesthetized and intracardially perfused with 4◦C PBS. The sciatic nerves, lumbar spinal cords, livers and inguinal lymph nodes were removed quickly and stored in liquid nitrogen immediately for RNA isolation. Trizol LS Reagent (Invitrogen, Karlsruhe, Germany) was applied to isolate total RNA and QuantiTect Reverse Transcription Kit (QIAGEN, Hilden, Germany) was used to reverse transcribe RNA to cDNA, The resulting cDNA was used to measure semi-quantitatively the expression of genes using SYBR green qPCR master mix according to the manufacturer's protocol (BioRad, Hercules, CA, United States). Real-time measurements of gene expression were performed with an iCycler thermocycler system and iQ5 optical system software (BioRad, Hercules, CA, United States). All the results are presented as mRNA expression levels relative to the housekeeping gene β-actin.

### Flow Cytometric Analysis

Single-cell suspensions were prepared from the spleens taken from the rats on Day 15. RBCs were lysed and monocytes were suspended in 100 µL PBS containing 1% FBS. All cells were stained, according to the manufacturer's protocol, with FITC-conjugated antibody against CD11b and anti-CD206-PE (eBioscience, San Diego, CA, United States), which is a specific anti-inflammatory marker. Data were analyzed using FlowJo Flow Cytometry Analysis Software<sup>1</sup> .

### Western Blot Analysis

On Day 15, total proteins was extracted from the sciatic nerves of various groups of EAN rats. All samples were adjusted to an equal volume and content and separated electrophoretically on 12% SDS-PAGE gels. Then, proteins were transferred to PVDF membranes (Millipore, Billerica, MA, United States). After being blocked in Tris-buffered saline solution (TBS) containing

<sup>1</sup>flowjo.com

5% BSA for 2 h, membranes were incubated at 4◦C overnight with primary antibodies: Notch1 (1:1000), Jagged-2 (1:500) (Cell Signaling Technology, Beverly, MA, United States), Hes-1 (1:500) (Millipore, Single Oak Drive, Temecula, CA, United States), and β-actin (1:500) (Sigma, St. Louis, MO, United States). The proteins were visualized by secondary antibodies with enhanced chemiluminescence. The signals of specific proteins were detected with a Gel Doc imager (Serial No. 721BR08844; Bio-Rad, Hercules, CA, United States) and reported as the fraction of control β-actin.

### Evaluation and Statistical Analysis

Differences between the neurologic scores and histological scores were evaluated with the Mann–Whitney U test. For paired and multiple group comparisons, Student's t-test and one-way analysis of variance (ANOVA) were applied, respectively. The results are represented as means ± SEM. Statistical significance was defined as p-values < 0.05. Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, United States). For the in vitro studies with cell cultures, graphs show mean ± SEM of triplicate wells and represent data from three independent experiments.

### RESULTS

### Oridonin Inhibited Inflammatory Response and Induced Phenotypic Switch of Macrophage Polarization in Cell Culture

As the anti-inflammatory phenotype of macrophages can attenuate the inflammatory response and promote tissue repair, we first investigated whether Oridonin could inhibit the inflammatory response of the macrophages cell line, RAW, and the primary peritoneal macrophages by favoring a switch of macrophages from classically activated inflammatory into the anti-inflammatory phenotype in vitro.

Upon LPS induction, increased NO production (**Figure 1A**) and mRNA expression of characteristic inflammatory cytokines IL-1β, IL-6, and iNOS (**Figure 1B**) indicated a classically proinflammatory phenotype in both the RAW cells and the primary macrophages. Oridonin treatments significantly reduced the NO production (**Figure 1A**) and reduced mRNA levels of these inflammatory cytokines in dose-dependent patterns, suggesting a suppressed inflammatory reaction (**Figures 1B,C**). Further, mRNA levels of anti-inflammatory phenotype markers IL-10 and CD206 were significantly increased, also in dose-dependent patterns, indicating a switch in the phenotype of the macrophages to the anti-inflammatory phenotype (**Figures 1B,C**). Additional MTT assays excluded possible differences in cell viability among the different treatments (data not shown).

The expression levels of IL-1β, TNF-α, and IL-10 from the primary macrophages were detected with the ELISA assay. The productions of IL-1β and TNF-α were significantly reduced by Oridonin. Further, the macrophages treated with Oridonin released much more IL-10 than LPS induced cells (**Figure 1D**). Therefore, Oridonin showed efficient antiinflammatory activity and induced a switch of macrophage polarization toward the anti-inflammatory phenotype, which suggested its further application in the inflammatory animal model of EAN.

### Oridonin Induced Phenotypic Switch in Macrophage Polarization by Blocking the Notch Signaling

We further investigated the possible mechanism behind the anti-inflammatory effects of Oridonin. The Notch pathway is a direct target of Oridonin, and Notch inhibition was reported to enhance monocyte differentiation into the anti-inflammatory (M2) phenotype (Dong et al., 2014; Singla et al., 2014, 2017; Xia et al., 2017). We, therefore, investigated whether Oridonin exerts these effects by affecting Notch signaling. In LPS-stimulated RAW cells, Oridonin treatment significantly reduced the mRNA levels of Jagged-2, Notch1, and Hes-1, all of which are involved in the Notch pathway (**Figure 1E**). Further, the effects of Oridonin were abolished by the Notch1 transient transfection. Cells from the other comparable groups, especially the control group, were all transient transfected with the plasmid vehicle. Simultaneously, the Notch inhibitor DAPT exhibited anti-inflammatory activity similar to that of Oridonin, the mRNA levels of pro-inflammatory cytokines, including IL-1β, IL-6, and iNOS were decreased, and the anti-inflammatory phenotype markers IL-10 and CD206 were significantly increased (**Figure 1F**). Taken together, our results suggest that Oridonin induced the phenotypic switch of LPS-stimulated macrophages from the pro-inflammatory phenotype to the anti-inflammatory phenotype by blocking the Notch signaling.

### Oridonin Treatment Suppressed EAN Neurological Progression and Accompanying Mechanical Allodynia

Experimental autoimmune neuritis was induced by subcutaneous immunization of Lewis rats with 100 µg of synthetic neurogenic P2 peptide. Oridonin or CMC (the control group) was given by daily gastric gavage starting at Day 0 as a preventive or on Day 9 (when the first neurological sign occurred) as a therapeutic treatment. In the control group, the first neurologic signs were established on Day 9 (0.33 ± 0.17); the neurologic severity represented by the scores peaked on Day 15 (6.3 ± 0.8), receded thereafter, and disappeared by Day 21 (0 ± 0). In comparison, preventive treatment of Oridonin at the dose of 20 mg/kg, initiated since Day 0, effectively delayed EAN onset (the first neurological sign occurred on Day 11), significantly reduced the severity at disease peak (2.7 ± 0.3, p < 0.05, compared to the control group), and shortened disease duration (until Day 18). Moreover, therapeutic treatment with the dose of 20 mg/kg, started after the first neurological sign was seen, also significantly attenuated the disease peak (4.5 ± 0.4, p < 0.05, compared to the control group) and lead to an earlier recovery from EAN (until Day 20) (**Figure 2A**). Both the therapeutic and preventive treatments with Oridonin significantly ameliorated the reduction of body weight, which is also a characteristic sign

of EAN (**Figure 2B**). Additionally, a lower dose of 7 mg/kg also showed therapeutic, but significantly reduced, effects in the preventive treatment. Further, repeated ANOVA assays indicated a significant difference between all the four groups, both in the neurological scores and body weight time courses.

Mechanical allodynia is a characteristic neurological sign of both human neuropathies and animal EAN. For the EAN rats, it can be recorded as a significant reduction of the hind-paw withdrawal threshold (HWT) in comparison to individual baselines (average HWT of the first four days after immunization). Mechanical allodynia was established on Day 7 except in the group treated with Oridonin at the dose of 20 mg/kg from Day 0 (**Figure 2C**). In the control group, HWT decreased persistently and gradually, presenting a steady progression of mechanical allodynia. Meanwhile, in the therapeutic treatment group (since Day 9), reduction of HWT was steady, like in the control group until Day 9, but was significantly slower after that. As shown in **Figure 2C**, HWT values detected between the therapeutic treatment and control groups were significantly different since Day 10, one day after the first treatment, and the differences could be seen till Day 13, the end of the observation period (when the EAN rats in the control groups could not lift their paws anymore). In the group treated prophylactically with Oridonin (20 mg/kg) since Day 0, a significant reduction could not be seen until Day 10; the HWT decreased slowly after that, but the reduction was always a little less than in the other two groups. The additional group administered the preventive treatment with the lower dose of 7 mg/kg also showed protective, but significantly reduced, effects in this test. Our results indicate that Oridonin treatment significantly suppressed and slowed the progression of mechanical allodynia in the EAN rats.

### Oridonin Treatment Ameliorated Pathological Changes of EAN by Reducing Infiltration of Inflammatory Cells and Inflammation-Related Cytokines

Following therapeutic or preventive treatments with CMC or Oridonin, EAN rats were prepared for analysis at the peak of the EAN progression, namely on Day 15, and sciatic nerves were collected for further histological analysis (n = 6). LFB staining and quantification were first performed to investigate nerve demyelination and local infiltration of immune cells. Severe perivascular demyelination and massive infiltration of inflammatory immune cells were observed in the control group. Oridonin treatment significantly attenuated the degree of demyelination and the severity of cellular infiltration (**Figures 3A,E,I**). These pathological changes were evaluated, scored, and compared between treatment and control groups. As shown in the **Figure 3J**, in comparison to the control group (2.23 ± 0.16), both Oridonin preventive- (Oridonin 0) and therapeutic treatments (Oridonin 9) significantly decreased the mean histological scores of the sciatic nerves (1.14 ± 0.08 and 1.68 ± 0.13, respectively, p < 0.05).

Further immunohistochemical analysis showed significant reduction in the local aggregation of various infiltrated cellular populations following the Oridonin treatments. In the control group, massive infiltration of macrophages (ED1+), pan-T cells (W3/13+), and B cells (OX22+) were observed and the macrophages were dominant in the cross-sections (**Figures 3B–D**). These infiltrated cells were not uniformly distributed in the cross-sections but were more concentrated around vessels and in the perineurium; some of them were also seen in the endoneurium. Both preventive and therapeutic Oridonin treatments significantly attenuated infiltration of all these cell types (p < 0.05) but did not alter their distribution pattern (**Figures 3F–H,K**).

In addition, mRNA levels of inflammatory cytokines in sciatic nerves, including those of IL-1β, iNOS, and IL-6, which are crucial for inflammatory progression in EAN, were further analyzed by real-time PCR. As shown in **Figure 3L**, the mRNA levels were significantly lower in the Oridonintreated group than in the control group. However, the IL-10 mRNA level was not significantly changed, probably due to the dramatically reduced macrophage number in the local region.

### Oridonin Treatment Decreased Inflammatory Cell Accumulation in Spinal Roots and Microglial Activation in the Spinal Cord

In EAN rats, significant accumulation of ED1<sup>+</sup> and W3/13<sup>+</sup> cells were observed in the dorsal and ventral roots. Like the sciatic nerves, they mainly accumulated around vessels in the endoneurium and perineurium. In comparison, macrophages and T cells were rarely seen in the dorsal and ventral roots of the spinal cord in naive animals and with a lower cell density than in sciatic nerves, as reported in our previous studies (Zhang et al., 2009). Following the preventive Oridonin treatment, their accumulations were significantly reduced, for ED1: 1.9 ± 0.3% in the

control group and 0.6 ± 0.1% in the treatment group (p < 0.05) (**Figures 4A,D,G**); for W3/13: 0.8 ± 0.1% in the control group and 0.3 ± 0.1% in the treatment group (p < 0.05) (**Figures 4B,E,H**).

ED1 is also considered the microglial activation marker in CNS and its expression in the spinal cord was further analyzed. According to our previously published data (Zhang et al., 2009), only a few microglial cells could be observed in the spinal cord of naive rats. In our EAN rats, however, obvious ED1 IR could be seen, mostly in the gray matter of the lumbar spinal cord, especially in the superficial layers of the dorsal horns. They accumulated near large neurons, particularly in the control group. They exhibited an activated hypertrophic morphology, which was characterized by thicker, shorter, and less branched processes; and enlarged, darkened soma. In the Oridonin group, however, fewer ED1 IR and ED1<sup>+</sup> cells with relatively smaller soma were observed (**Figures 4C,F**). The density of ED1 IR was then analyzed in the dorsal horns, where ED1<sup>+</sup> cells were relatively more abundant. The quantity of ED1 IR was significantly lower in

control group, n = 6.

the Oridonin preventive treatment group than in the control group (0.40 ± 0.04% for control, 0.24 ± 0.03% for treatment, p < 0.05) (**Figure 4I**).

### Oridonin Increased the Proportion of Anti-inflammatory Macrophages in EAN Rats

A significantly increased expression of CD163 is characteristically observed on an anti-inflammatory activated phenotype of macrophages (Moestrup and Moller, 2004; Badylak et al., 2008). ED2 antibody against CD163 was therefore used to detect these macrophages in the sciatic nerves of our EAN rats (**Figures 5A,B**). Even though the total numbers of accumulated macrophages were dramatically reduced (ED1<sup>+</sup> cells, **Figures 3B,F**), CD163/ED2 IR (0.36 ± 0.04%) in the sciatic nerves of the Oridonin group was slightly increased (0.33 ± 0.03%, p > 0.05, compared to the control group, **Figure 5C**). Moreover, the proportion of CD163 IR among total macrophages (ED1<sup>+</sup> cells) was significantly increased following the treatment (36% for the control and 57% for the Oridonin group, **Figure 5D**), indicating a switch of

the local macrophage population from classically activated inflammatory into the anti-inflammatory phenotype, as well as the dominance of the anti-inflammatory phenotype among all infiltrated macrophages.

Further, the effect of Oridonin on the phenotypic polarization of macrophages in the entire body was studied. Mononuclear cells (MNCs) derived from the spleens of EAN rats were determined with flow cytometry and results demonstrated a higher proportion of anti-inflammatory macrophages in spleen MNCs from the Oridonin treated EAN rats than in those of the control group rats (**Figure 5E**).

### Oridonin Inhibits Notch Signaling Protein Expressions in Sciatic Nerves of EAN Rats

As proved by the results from our in vitro study, Oridonin is a potent suppressor of the Notch signaling pathway and may

thereby induce the phenotypic switch in macrophage polarization and suppress the inflammatory reactions. We therefore explored the expressions of the major products of the Notch pathway, including Jagged-2, Notch1, and Hes1 in the EAN sciatic nerves from both groups. Results from Western blotting analysis show potent local suppression of Jagged-2, Notch1, and Hes-1 by Oridonin treatment (**Figure 5F**). The involvement of the Notch pathway has been thereby approved in our EAN rats and all the representative molecules were modified following Oridonin treatments, in accordance with previous reports; therefore, no further investigations were performed to explore the pathway in detail as this has been previously reported in other studies.

### DISCUSSION

As the prime animal model of human GBS and polyneuropathies, EAN is widely accepted and applied in disease mechanism investigation and development of novel therapeutic approaches. After we proved efficient anti-inflammatory activity of Oridonin and ability of inducing an anti-inflammatory phenotypic switch of macrophage polarization in cell culture, potential therapeutic values of Oridonin were further investigated in our animal model of human polyneuropathies and GBS. Our results show that either therapeutic or preventive treatments with Oridonin significantly shortened the disease duration, attenuated the peak severity, and paraparesis, or even delayed EAN onset. Oridonin treatments also significantly attenuated neuropathic pain, inflammatory infiltration of activated macrophages and other immune cells in the peripheral nerves and reduced microglial activation in the lumbar spinal cord. Interestingly, Oridonin treatment increased the systematic and local proportions of anti-inflammatory macrophages over the pro-inflammatory phenotypic population and resulted in the dominance of the anti-inflammatory phenotype among all infiltrated macrophages, which may be crucial for the anti-inflammatory activity, and ameliorated EAN progression. Our results from both in vitro and in vivo experiments suggest that the macrophage phenotypic switching effect of Oridonin may be associated with the inhibition of the Notch pathway.

As an effective diterpenoid isolated from R. rubescens, Oridonin has a variety of physiological and pharmacological properties/activities including anti-bacterial, anti-tumor, and anti-inflammatory. R. rubescens has a long history in TCM for the treatment of tumor and inflammatory diseases (Dong et al., 2014). The herb or its aqueous extract has been successfully applied in TCM for treatment of human inflammation such as gingivitis (Chen et al., 2009) and pharyngitis (Ma et al., 2011) for a long time. Several studies reported that Oridonin inhibited the expressions of COX-2 and iNOS by blocking NF-κB activity (Ikezoe et al., 2005; Leung et al., 2005); inhibited further proinflammatory molecules including IL-6, IL-2, IL-12, TNF-α, and IFN-γ; induced apoptosis of immune cells; as well as affected the anti-inflammatory target HO-1 (Hu et al., 2008); indicating its immunosuppressive and anti-inflammatory properties. However, a significantly modified expression of NF-κB was not observed in our EAN sciatic nerves, which indicated that other mechanisms which could be involved needed further investigation. Moreover, Oridonin has neuro-protective and anti-neuroinflammatory effects by modulating multiple microglial functions (Xu et al., 2009), suggesting its potential therapeutic application against inflammatory disorders of the nervous system.

In this present study, both preventive and therapeutic treatments with Oridonin significantly suppressed disease progression and eventually improved EAN outcome, by attenuating immune cell accumulation and reducing expressions of inflammatory cytokines in peripheral nerves, including the sciatic nerves and spinal roots. The infiltration of reactive leukocytes, mainly macrophages and T cells, into the PNS is the characteristic pathological change of EAN (Schabet et al., 1991); especially the activated macrophages triggered nerve demyelination by secretion of inflammatory mediators and direct phagocytic attack (Kiefer et al., 2001; Maurer et al., 2002). In EAN peripheral nerves, inflammatory cytokines are expressed and secreted by different immune cell types and modulate/regulate inflammatory responses. Pro-inflammatory molecules like IL-1β, IL-6, and iNOS are believed to promote EAN disease progression. IL-1β is considered a participant in the initiation of the autoimmune response of human neuropathies and the animal model EAN (Bettelli et al., 2008), IL-6 is believed to amplify local inflammation and be crucial for the EAN progression (Zhang et al., 2008b), iNOS produces pro-inflammatory NO (Abramson et al., 2001), and iNOS up-regulation in EAN was proved to contribute to the demyelination and even axonal damage in PNS (Conti et al., 2004). Therefore, Oridonin attenuated local inflammatory reactions and demyelination in the PNS and finally favored EAN outcome, by reducing local infiltration of reactivated immune cells and expression of inflammatory cytokines/molecules. In addition, the working concentrations used in our macrophage cell culture were comparable to the oral doses used in a previous study and presented significant dose-dependent anti-inflammatory patterns. The dosage was also comparable to orally administered Oridonin and the standard treatment followed in TCM (Xu et al., 2006).

Besides macrophages, we also observed a microglial accumulation in the lumbar spinal cord, together with pain hypersensitivity. While the EAN pathological changes were mainly studied and observed in the PNS, pathological changes in EAN spinal cord have drawn increased attention, especially focusing on spinal microglia (Piehl and Lidman, 2001; Nakamura, 2002), as they were considered to play a key role in inducing neuropathic pain in a variety of models (Moalem and Tracey, 2006). In the spinal cord of EAN animals, the activation of microglia is seen (Beiter et al., 2005), which is considered crucial in the induction of neuropathic pain (Raghavendra et al., 2003). Microglia are very sensitive to micro-environmental changes and spinal microglia are stimulated by various factors ranging from peripheral injury to central nervous system (CNS) inflammation. Reactive microglia in the spinal cord express various receptors and release inflammatory molecules acting either directly on pain transmitting dorsal horn neurons (nociceptive neurons) or on primary afferents, inducing raised sensitivity of the nociceptive neurons and contributing indirectly to central sensitization of neuropathic pain (Beggs et al., 2012; Ji et al., 2016). Therefore,

activation of spinal microglia is essential and sufficient to initiate neuropathic pain. Suppression of spinal microglia can significantly relieve or even block the process of neuropathic pain in various animal models (Ji and Strichartz, 2004). Our study showed significantly reduced microglial accumulation, which might be due to attenuated stimuli from the periphery or direct inhibiting effects on spinal microglia. Our results indicate that Oridonin could suppress neuropathic pain, exert direct or indirect neuro-protective effects in the animal model, and possibly even in human patients, which makes the therapeutic and preventive effects of Oridonin more comprehensive.

As the major effector cells responsible for the main EAN pathological changes, macrophages' activation states/status and phenotype may determine the disease progression. Based on our in vitro and in vivo results, the potent anti-inflammatory activity of Oridonin during EAN progression is, at least partially, due to the phenotypic switch toward anti-inflammatory macrophages and eventually their local dominance in peripheral nerves. This is also in accordance with previous studies demonstrating that anti-inflammatory activated macrophages played a protective and immune-modulatory role in EAN progression; as well as a switch of macrophage polarization from the pro-inflammatory phenotype toward the anti-inflammatory could resolve inflammation and thereby favor the outcome of different inflammatory disorders (Li et al., 2016; Klinkert et al., 2017); including animal models of EAN (Zhang Z. Y. et al., 2010; Han et al., 2016b). In our EAN rats, Oridonin raised the proportion of anti-inflammatory macrophages (ED2<sup>+</sup> cells) substantially over the pro-inflammatory phenotype population not only locally in the peripheral nerves but also systemically as determined in the spleen, the most important storage organ of the peripheral immune cells. The latter may also suggest an improved environmental milieu in immune organs. Moreover, Oridonin efficiently reduced local expression of pro-inflammatory representative cytokines/molecules. Although the anti-inflammatory representative cytokine IL-10 was not significantly changed, considering the dramatically reduced number of total immune cells, a slight increase of IL-10 could also support the dominance of anti-inflammatory over proinflammatory cells.

Although the phenotypic switch of macrophage polarization has not been associated directly with Oridonin treatment, particularly in an inflammatory condition, previous data suggest possible involvement of the Notch pathway: the Notch pathway is a direct target of Oridonin and Notch inhibition could promote monocyte differentiation into anti-inflammatory macrophages (Singla et al., 2017; Xia et al., 2017). Dong et al. (2014) demonstrated that Oridonin could suppress the Notch activity, down-regulate Notch pathway proteins such as Jagged-2, Notch-1, and their downstream molecules, in the same manner as a Notch inhibitor could. Singla et al. reported that suppression of the Notch-1 pathway or of several further Notch signaling proteins decreased the inflammatory macrophages and expression of their cytokines, enhanced

#### FIGURE 6 | Schematic diagram showing how Oridonin ameliorates experimental autoimmune neuritis by promoting anti-inflammatory macrophages through blocking Notch pathway. Oridonin exhibits efficient anti-inflammatory activity and induced a switch in macrophage polarization to the anti-inflammatory phenotype through inhibition of the Notch pathway in our in vitro study. Further in the EAN rats, Oridonin significantly reduced neuropathic pain, demyelination, inflammatory cellular accumulations and inflammatory cytokines in peripheral nerves, effectively suppressed disease progression by attenuating local inflammatory reaction and increasing the proportion of immune regulating macrophages, possibly through blockage of the Notch pathway.

secretion of anti-inflammatory molecules, and subsequently promoted anti-inflammatory macrophage polarization (Singla et al., 2014), even for human primary monocytes (Singla et al., 2017). In the present study, we demonstrated that the antiinflammatory switch of macrophage polarization in LPS-induced cell culture was promoted by Oridonin or the Notch inhibitor DAPT, but the transfection of Notch-1 could abolish this effect by Oridonin. This proves the effects of Oridonin on inhibition of the Notch-1 pathway. Further, expression levels of the Notch pathway proteins were also decreased locally in EAN sciatic nerves following Oridonin treatment, meanwhile proportions of anti-inflammatory macrophages were increased locally or systemically in the spleen. All these suggest that inhibition of the Notch pathway by Oridonin may be, at least partially, involved in the anti-inflammatory phenotypic switch of macrophage polarization.

### CONCLUSION

Oridonin ameliorated inflammatory EAN progression and finally favored the disease outcome through inducing the

### REFERENCES


phenotypic switch of macrophage polarization toward the antiinflammatory state, probably by the blockage of the Notch pathway (**Figure 6**). Therefore, with its proven safety and biocompatibility, Oridonin could be considered a potential antiinflammatory therapeutic candidate of human GBS and other inflammatory neuropathies.

### AUTHOR CONTRIBUTIONS

HS and Z-YZ designed the experiments and obtained resources and funding acquisition. LX and Z-YZ conducted the experiments with assistance from LL and C-YZ. LX, LL, C-YZ, and Z-YZ collected the data and contributed to the statistical analysis. LX, Z-YZ, and HS analyzed the data and wrote the manuscript. All authors read and approved the final manuscript.

### FUNDING

This work was supported by the National Natural Science Foundation of China (Grant Nos. 81571240 and 81771171) and the Jiangsu Provincial Innovation Team Program Foundation.

mTOR pathway, elevating Akt activity and polarizing M2 macrophages. Exp. Neurol. 280, 106–114. doi: 10.1016/j.expneurol.2016.04.005


**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 © 2019 Xu, Li, Zhang, Schluesener and Zhang. 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.

fnins-13-00272 March 29, 2019 Time: 18:51 # 14

# Connecting Metainflammation and Neuroinflammation Through the PTN-MK-RPTPβ/ζ Axis: Relevance in Therapeutic Development

Gonzalo Herradon<sup>1</sup> \*, M. Pilar Ramos-Alvarez<sup>2</sup> and Esther Gramage<sup>1</sup>

<sup>1</sup> Departamento de Ciencias Farmacéuticas y de la Salud, Facultad de Farmacia, Universidad San Pablo-CEU, CEU Universities, Madrid, Spain, <sup>2</sup> Departmento de Química y Bioquímica, Facultad de Farmacia, Universidad San Pablo-CEU, CEU Universities, Madrid, Spain

Inflammation is a common factor of pathologies such as obesity, type 2 diabetes or neurodegenerative diseases. Chronic inflammation is considered part of the pathogenic mechanisms of different disorders associated with aging. Interestingly, peripheral inflammation and the associated metabolic alterations not only facilitate insulin resistance and diabetes but also neurodegenerative disorders. Therefore, the identification of novel pathways, common to the development of these diseases, which modulate the immune response and signaling is key. It will provide highly relevant information to advance our knowledge of the multifactorial process of aging, and to establish new biomarkers and/or therapeutic targets to counteract the underlying chronic inflammatory processes. One novel pathway that regulates peripheral and central immune responses is triggered by the cytokines pleiotrophin (PTN) and midkine (MK), which bind its receptor, Receptor Protein Tyrosine Phosphatase (RPTP) β/ζ, and inactivate its phosphatase activity. In this review, we compile a growing body of knowledge suggesting that PTN and MK modulate the immune response and/or inflammation in different pathologies characterized by peripheral inflammation associated with insulin resistance, such as aging, and in central disorders characterized by overt neuroinflammation, such as neurodegenerative diseases and endotoxemia. Evidence strongly suggests that regulation of the PTN and MK signaling pathways may provide new therapeutic opportunities particularly in those neurological disorders characterized by increased PTN and/or MK cerebral levels and neuroinflammation. Importantly, we discuss existing therapeutics, and others being developed, that modulate these signaling pathways, and their potential use in pathologies characterized by overt neuroinflammation.

#### Edited by:

Morena Zusso, University of Padova, Italy

### Reviewed by:

Dimitris Beis, Biomedical Research Foundation of the Academy of Athens, Greece Madhuvika Murugan, New Jersey Institute of Technology, United States

> \*Correspondence: Gonzalo Herradon herradon@ceu.es

#### Specialty section:

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

Received: 23 January 2019 Accepted: 26 March 2019 Published: 12 April 2019

#### Citation:

Herradon G, Ramos-Alvarez MP and Gramage E (2019) Connecting Metainflammation and Neuroinflammation Through the PTN-MK-RPTPβ/ζ Axis: Relevance in Therapeutic Development. Front. Pharmacol. 10:377. doi: 10.3389/fphar.2019.00377 Keywords: pleiotrophin, midkine, PTPRZ, neuroinflammation, inflammation, aging, obesity, neurodegeneration

### NEUROINFLAMMATION AND CNS DISORDERS

Activation of the innate immunity in the Central Nervous System (CNS) is a critical step in the healing process after brain injury. However, cumulative evidence points to deleterious effects of chronic neuroinflammation (Szepesi et al., 2018). Activation of astrocytes and microglia is a hallmark of neuroinflammation. Activated glial cells release free radicals, cytokines and other

Herradon et al. PTN and MK in Neuroinflammation

pro-inflammatory factors. Activated microglia can be prophylactic and protect the brain from a wide variety of events including traumatic injury, stroke or neurodegenerative diseases, through modulation of neuronal synapses, by promoting neurogenesis, clearing debris, and suppressing inflammation (Chen and Trapp, 2016). However, exacerbated or persistent neuroinflammation can potentially exert negative effects on CNS integrity and function accompanying a wide variety of CNS diseases (Kielian, 2016). Sustained neuroinflammation is implicated in the progressive nature of neurodegenerative diseases (Faden et al., 2016; Gasiorowski et al., 2017). In addition to the peripheral stimuli that can cause chronic neuroinflammation, inflammatory processes in the CNS are also observed in response to different central events such as contact with neurotoxins or traumatic brain injury (TBI) (Spielman et al., 2014) and have been related to vascular injury and angiogenesis (Popp et al., 2017). The perpetuation of these inflammatory processes can subsequently cause acute secondary injury, leading to chronic neurodegenerative diseases (Simon et al., 2017). Supporting this role of neuroinflammation in neurodegenerative diseases, genome-wide association studies have shown associations with neurodegenerative disorders susceptibility, such as Parkinson's disease (PD), in the human leukocyte antigen region and other polymorphisms related to the immune system including inflammatory cytokines and their receptors (Nalls et al., 2014). In addition, chronic neuroinflammation has been shown to contribute to neuronal loss by apoptosis in different pathologies, including multiple sclerosis (Mori F. et al., 2014; Mandolesi et al., 2017), drug addiction and alcoholism, and psychiatric disorders such as major depression (Montesinos et al., 2016; Neupane, 2016). In the light of this dual role of neuroinflammation and microglial activation in health and disease, the identification of the mechanisms responsible for maintenance of microglial activation and associated continuous neuron damage is critical to understand the etiology and pathology of diseases characterized by overt neuroinflammation.

### Pleiotrophin and Midkine: Novel Modulators of Neuroinflammation

In neuroinflammatory processes, the activation of innate immunity, contribution of Toll-like receptors (TLRs) and induction of expression of inflammation-related molecules such as cytokines, have been described in detail. Although the discovery of new modulators of neuroinflammation can potentially translate in novel targets and biomarkers for different diseases, the genetic bases of neuroinflammation are still far from being fully understood. Recently, we identified Pleiotrophin (PTN) and Midkine (MK) as two novel neurotrophic factors that modulate neuroinflammation in different contexts (**Figure 1**; Vicente-Rodriguez et al., 2016a,b; Fernandez-Calle et al., 2017). Pleiotrophin and MK are the only cytokines that constitute the Ptn/Mk developmental gene family (Kadomatsu et al., 1988; Milner et al., 1989). The pattern of expression of these cytokines in developing and adult nervous system has been extensively described in different species and led to study the importance

astrocytosis. (1) Vicente-Rodriguez et al. (2016a), (2) Vicente-Rodriguez et al. (2016b), (3) Fernandez-Calle et al. (2017).

of these cytokines in neural-glial interactions (Vanderwinden et al., 1992; Silos-Santiago et al., 1996; Xu et al., 2014). They are widely expressed during development but, in adults, their pattern of expression is restricted to a few cell types in different organs, including the brain, where they are mainly expressed in neurons. Importantly, the levels of expression of MK and PTN are highly upregulated after injury in different cells including microglia and inflammatory macrophages (Jin et al., 2009; Martin et al., 2011; Muramatsu, 2011; Gonzalez-Castillo et al., 2014). As suggested by their similar pattern of expression, PTN and MK overlap many functions including wound repair and survival of neurons (Gramage and Herradon, 2011; Herradon and Perez-Garcia, 2014). Accordingly, PTN and MK are upregulated in senile plaques and sera of patients with Alzheimer's disease (AD) and in the substantia nigra of patients with PD among other neurodegenerative disorders (Herradon and Perez-Garcia, 2014). Midkine is also found in glial cytoplasmic inclusions of multiple system atrophy brains (Kato et al., 2000). In addition, both cytokines are also upregulated in the brain in pathological conditions characterized by neuroinflammation such as different types of brain injury, including ischemia, and after administration of drugs of abuse like amphetamine, alcohol and opioids (**Table 1**; Herradon and Perez-Garcia, 2014). It has been demonstrated that Ptn−/<sup>−</sup> mice exhibit exacerbated amphetamine-induced dopaminergic injury in the nigrostriatal pathway (Gramage et al., 2010), which correlated with changes in the striatal phosphoproteome similar to those found in PD (Gramage et al., 2013). Other authors have demonstrated that Ptn overexpression in the brain exerts neurotrophic effects in rodent models of PD (Gombash et al., 2012). Taken together, these findings suggest that PTN protects against dopaminergic neurodegeneration in different pathological contexts.

Midkine exerts relevant functions in peripheral inflammatory processes in different pathological conditions (Muramatsu, 2014; Weckbach et al., 2014), facilitating the migration of macrophages



The upregulation of these cytokines in the CNS has been described after different stimuli or in pathologies related to neuroinflammatory processes.

and neutrophils (Takada et al., 1997; Horiba et al., 2000; Sato et al., 2001) and preventing differentiation of regulatory T-cells (Wang et al., 2008; Sonobe et al., 2012). In the CNS, we found that amphetamine-induced microglial response and astrocytosis are enhanced in the striatum, but not in the hippocampus, of Mk−/<sup>−</sup> mice. In contrast, lipopolysaccharide (LPS)-induced striatal astrogliosis was blocked by genetic inactivation of Mk, suggesting a differential regulation of astrocytosis by MK depending on the inflammatory stimulus (**Figure 1**; Vicente-Rodriguez et al., 2016a). On the other hand, Ptn overexpression in the brain (Ptn-Tg mice) potentiates the striatal astrocytosis induced by acute administrations of amphetamine (Vicente-Rodriguez et al., 2016b). Pleiotrophin overexpression also potentiates microglial activation after an acute systemic administration of LPS (Fernandez-Calle et al., 2017). Importantly, LPS-induced increases of pro-inflammatory cytokines in the brain, including TNF-α and IL-1β, were significantly enhanced in Ptn-Tg mice (Fernandez-Calle et al., 2017). The data suggest that PTN potentiates the acute neuroimmune response induced by different stimuli, including microglial activation (**Figure 1**), which is necessary and critical for host defense (Chen and Trapp, 2016) and could contribute to the known neurotrophic effects of this cytokine. However, although upregulation of PTN levels in the brain has been described in different chronic diseases such as AD and PD (Herradon and Perez-Garcia, 2014), in which longterm neuroinflammation is a pathological hallmark, the possible modulatory role of PTN in persistent neuroinflammation remains to be elucidated. If chronic upregulation of PTN levels in these diseases contributes to prolonged neuroinflammation, PTN would parallel the dual role of glial activation in health and disease depending on the duration of its actions. Although further studies are needed to better understand the roles of PTN and MK in the regulation of neuroinflammation in chronic CNS disorders, evidence strongly suggests that pharmacological modulation of the actions of PTN and MK is a novel strategy to treat disorders characterized by neuroinflammation.

It has to be noted that aging-associated neuroinflammatory conditions (e.g., AD) have been related with vascular injury and angiogenesis (Popp et al., 2017) and PTN expression levels are associated strongly with aging in human cerebrospinal fluid (Wyss-Coray, 2016). Both PTN and MK induce tumor growth mainly by promoting angiogenesis (Perez-Pinera et al., 2008; Weckbach et al., 2012; Papadimitriou et al., 2016). Pleiotrophin exerts mitogenic activity in fibroblasts (Milner et al., 1989), brain capillary endothelial cells (Courty et al., 1991), and in adrenal carcinoma cells (Fang et al., 1992). Importantly, Ptn is upregulated in vascular endothelial cells in sites of neovascularization after focal cerebral ischemia (Yeh et al., 1998). The angiogenic potential of PTN has been shown in studies in ischemic myocardium in rats, in which it was demonstrated that PTN stimulates significant increases in normal appearing new capillaries, arterioles, and newly formed blood vessels that interconnect with the existent coronary vascular system (Christman et al., 2005a). On the other hand, MK promotes endothelial cell proliferation and the recruitment of inflammatory cells to lesions (Kadomatsu et al., 2014). As in the case of PTN, an important proangiogenic role under hypoxia conditions has been described for MK (Weckbach et al., 2012). Thus, both cytokines have been proposed as important factors for inflammation, tissue repair and angiogenesis. Accordingly, it seems reasonable to propose the modulation of PTN/MK signaling pathways for the treatment of the above mentioned disorders characterized by neuroinflammation and vascular injury.

### The Mechanism of Action of Pleiotrophin and Midkine in Relation to Neuroinflammation

Midkine and PTN bind different receptors such as Receptor Protein Tyrosine Phosphatase (RPTP) β/ζ (Maeda et al., 1996, 1999; Meng et al., 2000), syndecan-3 (Rauvala et al., 2000), and anaplastic lymphoma kinase (ALK) (Stoica et al., 2001). Midkine also binds other proteoglycans (Muramatsu, 2011; Kadomatsu et al., 2013), including Neuroglycan C, through which MK has been shown to promote neurites in oligodendrocyte precursor-like cells (Ichihara-Tanaka et al., 2006). Also, MK has been shown to bind low density lipoprotein receptor-related protein (LRP), integrin α4β1 and integrin α6β1 in the brain (Muramatsu et al., 2000, 2004).

RPTPβ/ζ (a.k.a. RPTPβ, PTPRZ, and PTPζ) (Krueger and Saito, 1992), is abundantly expressed in the CNS as a chondroitin sulfate proteoglycan (Krueger and Saito, 1992; Maeda et al., 1994). Pleiotrophin and RPTPβ/ζ are highly expressed in human white matter oligodendroglial precursor cells (OPCs), and PTN-PTPRZ signaling promotes postnatal OPC differentiation during developmental myelination and remyelination after injury (Harroch et al., 2002; Sim et al., 2006; Kuboyama et al., 2015; Tanga et al., 2019). RPTPβ/ζ is composed of an N-terminal carbonic anhydrase-like domain, a fibronectin type III domain, a serine, glycine-rich domain that is thought to be chondroitin

sulfate attachment region, a transmembrane segment, and two tyrosine phosphatase domains (Krueger and Saito, 1992). There are four splice variants of this molecule including a fulllength (PTPζ-A), a short form (PTPζ-B), the secreted form (PTPζ-S or 6B4 proteoglycan/phosphacan), which corresponds to the extracellular region of PTPζ-A (Maeda et al., 1994; Maurel et al., 1994) and the PSI isoform, expressed only in neurons (Heck et al., 2005). Midkine binds mostly to the chondroitin sulfate portion of RPTPβ/ζ (Maeda et al., 1999) which has been proven to be essential for MK-induced neuron survival effects. The interaction of RPTPβ/ζ with PTN inactivates the intrinsic tyrosine phosphatase activity of RPTPβ/ζ, presumably by enforcing a conformational change in RPTPβ/ζ that denies substrates access to its active site in the D1 domain (Meng et al., 2000).

These signaling pathways triggered by PTN and MK through RPTPβ/ζ support important roles of these cytokines in neuroinflammation (**Figure 1**) since they modulate the tyrosine phosphorylation of substrates of RPTPβ/ζ that are known regulators of neuroinflammation such as TrkA (Shintani and Noda, 2008) and Fyn kinase (Pariser et al., 2005; Panicker et al., 2015). After LPS administration, Fyn is activated in microglia (Panicker et al., 2015). Activated Fyn phosphorylates PKCδ at Y311, contributing to an increase in PKCδ kinase activity and activation of the NFκB pathway (Panicker et al., 2015), suggesting that the PTN/RPTPβ/ζ is a major regulatory pathway of this pro-inflammatory signaling cascade. These signaling events are also observed in animal models of PD (Panicker et al., 2015).

The wide pattern of expression of Ptn in developing and adult tissues, including nervous system and peripheral organs, has been described in detail (Vanderwinden et al., 1992; Xu et al., 2014). The sites of expression of RPTPβ/ζ in peripheral organs are highly relevant for its potential role in inflammation. RPTPβ/ζ is expressed in the intestine, mononuclear cells, monocytes, macrophages (Zwicker et al., 2016), in hematopoietic stem cells (HSC) and B cells, in which RPTPβ/ζ promotes HSC maintenance and B cell survival (Sorrelle et al., 2017). RPTPβ/ζ expression is increased in inflammatory processes associated with kidney injury, in which interleukin 34 (IL-34) increases its phosphatase activity and promotes monocyte and macrophage infiltration in the kidney, contributing to chronic renal damage (Sanchez-Nino et al., 2016). As in the case of PTN and MK, these findings point to an important role of RPTPβ/ζ in the regulation of central and peripheral inflammatory processes. However, the relevance of the changes in the phosphatase activity of RPTPβ/ζ may differ in acute and chronic inflammatory processes and has not been studied before, which is due, in part, to the lack of proper ligands of this receptor. Recently, Fujikawa et al. (2016) identified SCB4380 as a potent inhibitor of RPTPβ/ζ in a rat allograft model of glioblastoma. Unfortunately, the low lipophilicity of this inhibitor, which requires liposome carriers for intracellular delivery in vitro, is a critical limitation for its use in studies in the CNS. To fill this gap, we recently developed BBB permeable inhibitors of RPTPβ/ζ, selecting the lead compound (MY10) after in vitro and in vivo validation (Fernandez-Calle et al., 2018; Pastor et al., 2018).

### METAINFLAMMATION AND NEUROINFLAMMATION

Inflammatory responses in the brain in response to metabolic stress have been observed in both mice and humans, but little is known about the mechanisms that activate them (Thaler et al., 2012). For instance, low level inflammation in response to an excess of nutrients or energy is not necessarily a pathological process but reflects on a series of potentially harmful alterations in metabolic homeostasis, which has been called metainflammation (Gregor and Hotamisligil, 2011). Chronic overfeeding is associated with excess peripheral proinflammatory mediators that contribute to neuroinflammation, which subsequently exacerbates neurodegeneration (Spielman et al., 2014). Accordingly, epidemiological studies indicate that AD and PD risk positively correlate with pro-inflammatory conditions such as diabetes mellitus or metabolic syndrome (Spielman et al., 2014). Thus, the discovery of new pharmacological targets involved in the preservation of a healthy basal metabolism would potentially translate in novel therapeutic interventions not only in metabolic diseases but in neuroinflammatory CNS disorders. In this context, we will now summarize the evidence demonstrating that PTN is essential to preserve the appropriate basal metabolism.

### Pleiotrophin: A Novel Modulator of Metainflammation, Insulin Resistance, and Adipose Tissue Plasticity

Although Ptn is expressed in adults (Vanderwinden et al., 1992), its role in adipose tissue is not fully understood. Ptn expression is found in human adipose tissue, being higher in visceral white adipose tissue (WAT) than in subcutaneous WAT (Hoggard et al., 2012). In addition, the expression of Ptn in mice also depends on the type of visceral deposit, being about 30 times higher in the mesenteric than in the perigonadal fat. Pleiotrophin deletion is associated with a lipodystrophic phenotype (Sevillano et al., 2012), altered energy metabolism and insulin resistance (Sevillano et al., 2019). These results, together with previous reports from our group showing that PTN is increased in the visceral WAT of obese pregnant women, point to a novel role of PTN in metabolic homeostasis and metainflammation.

The association of insulin resistance with inflammation has been studied in depth during the last few decades (Shoelson et al., 2006; Olefsky and Glass, 2010; Hotamisligil, 2017), including the role of macrophages (Olefsky and Glass, 2010). It has been shown that inflammatory cytokines activate kinases that promote Ser phosphorylation of insulin receptor substrate 1 (pSerIRS1) (Bluher et al., 2009; Copps and White, 2012), which is considered a marker of insulin resistance in adipose tissue, not only in pathological (Schmitz-Peiffer and Whitehead, 2003) but also physiological situations, as shown previously (Sevillano et al., 2005, 2007; de Castro et al., 2011). An additional factor is the heterogeneity of macrophages; M1 are pro-inflammatory while M2 are anti-inflammatory and participate in homeostasis and tissue remodeling (Stienstra et al., 2008). Peroxisome proliferator-activated receptor gamma (PPAR-γ) is key in

metainflammation, since it inhibits the inflammatory response in macrophages (Welch et al., 2003). Thus, activators of PPAR-γ, such as thiazolidinediones (TZD), improve the insulin response of adipose tissue by inhibiting the production of inflammatory mediators, including TNF-α (Sharma and Staels, 2007), and by increasing the levels of adiponectin (Combs et al., 2002), an antiinflammatory adipokine that is downregulated in states of insulin resistance such as obesity or type 2 diabetes (T2D) (Statnick et al., 2000). In the same manner, treatment with TZDs reverses the polarization toward the pro-inflammatory phenotype M1 in diet-induced obesity models (Stienstra et al., 2008; Fujisaka et al., 2009). Precisely, we have proposed very recently that defective PPAR-γ activation may underlie insulin resistance of Ptn deficient mice by promoting an inflammatory condition that impairs lipid and glucose homeostasis (Sevillano et al., 2019).

The great plasticity of WAT enables this tissue to contract or expand in response to alterations in the energy balance. It has been proposed that WAT plasticity is a key factor in metabolic alterations such as obesity (Carobbio et al., 2017). Expansion of adipose tissue can be caused by hypertrophy, due to accumulation of triacylglycerides and by hyperplasia characterized by adipogenesis. Numerous molecules participate in this process, such as WNT and its downstream effectors, transcription factors, such as PPAR-γ (Spiegelman, 1998), or oxidative stress (Lowe et al., 2011). Interestingly, it has been suggested that PTN could inhibit the expression of PPAR-γ, which would block white adipocyte differentiation (Gu et al., 2007). The role of PTN in this context, however, remains unclear, since rPTN has been found to decrease Pparγ2 expression in vitro, but the same study shows in vivo that injection of a PTN-neutralizing antibody slightly decreased Pparγ2 in adipose tissue (Wong et al., 2016). Beige adipocytes can develop in WAT, a process known as browning, in response to certain stimuli, such as treatment with TZDs or cold exposure (Wang and Seale, 2016). WAT browning and brown adipose tissue (BAT) (Bartelt and Heeren, 2014; Lee et al., 2014) may represent important therapeutic targets because they facilitate the oxidation of excess lipids (Kajimura et al., 2015), thus preventing hyperlipidaemia and potentially also lipotoxicity. In this context, it has been recently demonstrated that PTN impairs brown adipocyte differentiation (Sevillano et al., 2019). In fact, the lipodystrophic phenotype of Ptn−/<sup>−</sup> mice is related to an enhanced thermogenesis in BAT (Sevillano et al., 2019). The molecular events responsible for these newly discovered functions of PTN seems to involve the extracellular matrix (ECM).

The plasticity of WAT also requires vascular remodeling and ECM (Pellegrinelli et al., 2016). Obesity is usually accompanied by infiltration of macrophages in the tissue, and it has been proposed that the matrix metalloproteinases (MMPs) secreted by resident macrophages in WAT participate in the degradation of ECM proteins and promote adipocyte-induced secretion of MMPs (Mariman and Wang, 2010), thus supporting WAT remodeling. In this context, it is important the phosphatase ADAMTS1, an MMP that contributes to the degradation of highly expressed collagens in WAT, such as Col1α1 (Mori S. et al., 2014). It has been proposed that ADAMTS1 plays a modulatory role in ECM and adipogenesis by inhibiting the differentiation of mesenchymal cells to preadipocytes. In fact, in a diet-induced obesity mouse model, expression of Adamts1 is decreased in the perigonadal WAT, which is consequently expanded by an activated adipogenesis. Accordingly, an inverse correlation between Adamts1 expression and body mass index (BMI) has been found in humans (Chen et al., 2016). It is important to note that ADAMTS1 induces Ptn expression through the WNT/β-catenin signaling pathway, suggesting a role of this cytokine in the effects of MMPs on adipocyte differentiation. This is supported by studies in humans with dietinduced weight gain that show concomitant increases in the levels of expression of ADAMTS1, PTN and different mediators of WNT signaling (Wong et al., 2016). Interestingly, preliminary results from our group showed that, when brown adipocytes differentiate, Adamts1 expression is blunted, which correlates with a decreased expression of Ptn (Sevillano et al., 2019).

### The Metainflammation-Neuroinflammation Connection: Role of Pleiotrophin

Although metainflammation was first described in WAT, it may also affect other tissues such as the liver, pancreas, or hypothalamus. In fact, to maintain organism homeostasis, the brain must control the energy state of the periphery, as the CNS depends on the continuous supply of nutrients from the general circulation. Adipose tissue is innervated by the sympathetic endings of the autonomic nervous system, with BAT being much more innervated than WAT (**Figure 2**). It has been shown that CNS neurons are involved in the multisynaptic pathways to both liver and WAT, which allows a coordinated control of peripheral metabolism (Stanley et al., 2010). The alteration of any of the key components of this system, as it occurs in inflammation, or the failure of its integration could be involved in the etiopathogenesis of metabolic disorders such as obesity and diabetes, but also in CNS disorders such as neurodegenerative diseases (**Figure 2**; Santiago and Potashkin, 2014). As mentioned before, peripheral inflammatory diseases such as T2D and obesity confers an increased risk of developing AD and PD (Spielman et al., 2014), presumably through their capacity to cause prolonged low-grade neuroinflammation. It is conceivable, therefore, that bidirectional communication and cross-talk between the CNS and the periphery may be relevant to these pathological conditions (**Figure 2**). In fact, the CNS seems to be involved in the control of proliferation and differentiation of white adipocytes. Accordingly, in vivo surgical or pharmacological denervation of WAT triggers an increase in the number of white pre-adipocytes and adipocytes (Cousin et al., 1993; Bowers et al., 2004), whereas noradrenaline inhibits the proliferation of adipocyte precursor cells in vitro (Jones et al., 1992). On the other hand, when BBB homeostasis is altered, as it occurs in some metabolic disorders, the development of central pathological events such as neurodegeneration are facilitated (Mauro et al., 2014), probably by increasing brain sensitivity to peripheral stimuli that eventually end up causing chronic neuroinflammation. For instance, systemic

administration of LPS is an animal model of peripherallyinduced neuroinflammation and neurodegeneration (Qin et al., 2007, 2013), and there is evidence that peripheral infections accompanied by inflammation represent major risk factors for the development of sporadic AD and PD (Holmes, 2013; Su and Federoff, 2014). Other alterations of BBB such as pericyte dysfunction at the neurovascular unit have been linked with aberrant angiogenesis (Sweeney et al., 2016). As PTN is known to modulate the astrocytic response in different contexts and to play a role in vascular formation (Martin et al., 2013; Zhang et al., 2014; Wang et al., 2017; Fernandez-Calle et al., 2018), a dual function of this cytokine in both angiogenesis and neuroinflammation in these conditions should not be ruled out.

In addition, it has to be noted that PPAR-γ is known to be involved in degenerative processes in the brain and peripheral inflammation in pathologies such as DM or obesity (Villapol, 2018). Importantly, PPAR-γ is also associated with the control of angiogenesis in the brain (Arany et al., 2008). As explained before, PTN is involved in the regulation of PPAR-γ expression (Gu et al., 2007), suggesting the interesting possibility that this mechanism may underlie the common role of this cytokine in diabetes, neurodegeneration and angiogenesis.

It is also important to identify modulatory factors regarding the communication between the CNS and the immune system involved in the control of immune responses both centrally and peripherally. Extracellular vesicles (ExV), including exosomes, are released by all cells, including those of adipose tissue and nervous system. Their capacity to deliver a variety of bioactive molecules (e.g., protein, mRNA, miRNA, DNA, and lipids) to both nearby and distal cells suggests that exosomes and specific miRNAs play a role in cancer, diabetes, obesity, cardiovascular disease (Trajkovski et al., 2011) and CNS disorders, including neuroinflammation and PD (Leggio et al., 2017). In this context, it has to be noted that PTN is secreted by murine and human neural precursor cells (NPCs) (Qin et al., 2017), which is consistent with reports identifying PTN in the secretomes of various neural stem cell populations (Furuta et al., 2004; Lee et al., 2012).

Since adipocytes secrete large proportions of exosomes, it has been proposed that exosomes and miRNAs derived from fat play a role in pathologies related with insulin resistance (Thomou et al., 2017). Furthermore, there is evidence that exosomes can penetrate the BBB, moving from the peripheral circulation to the CNS (Matsumoto et al., 2017). It is also interesting to note that brain endothelial exosomes may be used as serological biomarkers for BBB status during neuroinflammation (Ramirez et al., 2018), pointing to a key role of exosomes in CNSperiphery cross-talk. The possible involvement of exosomes in PTN-mediated effects needs to be elucidated.

### THERAPEUTIC PERSPECTIVES

Pleiotrophin and MK are novel modulators of neuroinflammation that play different roles depending on the inflammatory stimulus and the duration of the neuroinflammatory processes. Further studies are needed to fully understand the role of these cytokines on neuroinflammation, particularly in chronic neuroinflammation. However, in the light of evidence summarized here, it is reasonable to hypothesize that both potentiation and inhibition of the PTN/MK signaling pathways are novel therapeutic strategies to modulate neuroinflammation in acute and chronic pathological states. For instance, potentiation of the acute neuroinflammatory actions of these cytokines may exert therapeutic effects in different types of brain injury through limitation of the damage and improvement of brain repair. In terms of metainflammation, potentiation of PTN/MK signaling pathways may contribute to control the inflammatory condition associated to metabolic disorders. On the contrary, inhibiting the actions of these cytokines may be beneficial in chronic neuroinflammatory states that promotes exacerbation of neurodegeneration.

The administration of rMK and rPTN has been proposed in a number of CNS and peripheral injuries, rending already significant effects in preclinical models, particularly in the case of myocardial ischemia (Christman et al., 2005b; Fukui et al., 2008) and bone repair (Lamprou et al., 2014; Liedert et al., 2014). Gene therapy or stem cells overexpressing these genes have also been proposed as delivery systems of these cytokines in different animal models of peripheral and central disorders and have been reviewed elsewhere (Muramatsu, 2011; Herradon and Perez-Garcia, 2014; Kadomatsu et al., 2014).

The potential of therapeutic inhibition of MK and PTN has been studied in depth for the treatment of malignant diseases in different experimental models, including siRNA, shRNA, antibodies and RNA aptamers (Muramatsu, 2011; Kishida and Kadomatsu, 2014; Shi et al., 2017). Interestingly, an aptamer to MK has already been used to treat experimental autoimmune encephalitis (EAE) in mice (Wang et al., 2008), which is an animal model of multiple sclerosis.

Although promising, the above mentioned strategies to potentiate or inhibit the actions of PTN and MK in neuroinflammation in CNS disorders are commonly restricted by the limitation of the route of administration (e.g., intracranial). Thus, these effective therapeutic strategies in preclinical models of different diseases are not likely to progress through clinical development in the midterm. We will now summarize the existing modulators of PTN/MK signaling with approved therapeutic uses in humans and other promising strategies at the preclinical level that are more likely to progress through clinical development in the midterm.

### Potentiation of MK/PTN Actions RPTPβ/ζ Inhibitors

Recently, Fujikawa et al. (2016) described SCB4380 as a potent inhibitor of PTPRZ (RPTPβ/ζ) as a potential new drug for glioma therapy. Unfortunately, its physicochemical properties make very difficult for this molecule to go through biological barriers. Thus, SCB4380 required liposome carriers for intracellular delivery (Fujikawa et al., 2016). To overcome this limitation in the RPTPβ/ζ inhibition strategy, novel BBB permeable selective small-molecule inhibitors of RPTPβ/ζ were recently designed and synthesized through rational drug design (Pastor et al., 2018). The lead compound, MY10, interacts with the intracellular domain PD1 of RPTPβ/ζ and inhibits its tyrosine phosphatase activity. MY10 has been validated in vitro and in vivo. Interestingly, systemic administration of MY10 reduces alcohol consumption and blocks the rewarding effects of alcohol in mice (Fernandez-Calle et al., 2018), replicating the effects obtained before with the transgenic mouse model (Ptn-Tg) overexpressing in the brain Ptn, the endogenous inhibitor of RPTPβ/ζ (Vicente-Rodriguez et al., 2014). Using these Ptn-Tg mice, it was demonstrated that PTN overexpression in the brain potentiates LPS-induced microglial activation and neuroinflammation (Fernandez-Calle et al., 2017), suggesting that acute treatment with MY10 would cause similar effects and could be used in brain repair after injury by promoting acute neuroimmune responses. However, persistent and/or over-activation of microglia is deleterious. Thus, more knowledge regarding the role of PTN in chronic neuroinflammation is needed to substantiate the pharmacological use of RPTPβ/ζ inhibitors as a potential therapeutic strategy in CNS disorders related with chronic neuroinflammation.

### Inhibition of MK/PTN Actions ALK Inhibitors

ALK has been identified as a receptor for PTN (Stoica et al., 2001), but is also a substrate of RPTPβ/ζ (Perez-Pinera et al., 2007). It has been shown to mediate many of the central actions of PTN and MK (Herradon and Perez-Garcia, 2014; Dutton et al., 2017) and has been recently identified as a potent regulator of NLRP3 inflammasome activation in macrophages through its capacity to activate NF-kB (Zhang et al., 2018). Interestingly, it has been recently demonstrated that PTN activates AKT signaling through ALK to promote the morphological maturation and synaptic integration of newborn neurons (Tang et al., 2019). Pleiotrophin or MK-induced inhibition of RPTPβ/ζ causes an increase in tyrosine phosphorylation in ALK and subsequent activation of its tyrosine kinase activity (Perez-Pinera et al., 2007; Herradon et al., 2009).

Three generations of ALK inhibitors have been developed for the treatment of non–small cell lung cancer with certain mutations in Alk (Fan et al., 2018; Rothenstein and Chooback, 2018). Interestingly, these inhibitors are also effective in preclinical models of non-malignant completely unrelated pathological conditions. For instance, inhibition of ALK with alectinib, a second generation ALK inhibitor, significantly reduces alcohol binge drinking in rodents (Dutton et al., 2017). Inhibition of ALK with ceritinib and lorlatinib blocks NLRP3 inflammasome activation in macrophages (Zhang et al., 2018). Taking together, evidence suggests that the effects of the PTN-MK-RPTPβ/ζ axis on neuroinflammation could be

modulated through pharmacologic regulation of one of its main effectors: ALK.

### Fyn Inhibitors

Fyn kinase is a substrate of RPTPβ/ζ (Pariser et al., 2005). Fyn has been found activated in microglia in animal models of PD and after treatment with LPS and triggers the activation of the pro-inflammatory NFκB pathway in these models (Panicker et al., 2015). These data suggest that Fyn is a mediator of the actions of the PTN-MK-RPTPβ/ζ axis on neuroinflammation and, thus, pharmacologic inhibition of Fyn could be considered as a potential strategy to modulate the effects of this axis in diseases characterized by neuroinflammatory processes. In this context, it is interesting to note that Poli et al. (2018) have recently developed through rational drug design new derivatives of previous hit compounds Fyn inhibitors with considerable potency (IC<sup>50</sup> = 0.76 µM).

### MK Inhibitors

Small-molecule midkine inhibitors have been investigated mainly in the field of cancer therapy (Matsui et al., 2010). A promising compound (iMDK) has been recently proven to be effective in preclinical models of non–small cell lung cancer (Hao et al., 2013), oral squamous cell carcinoma (Masui et al., 2016) and prostate cancer (Erdogan et al., 2017). By inhibiting the actions of MK on RPTPβ/ζ, iMDK would leave unchecked the intrinsic phosphatase activity of this receptor. As a result, decreased phosphorylation and activation of its substrates (e.g., Fyn, ALK) would contribute to decreased neuroimmune responses.

### CONCLUSION

The expression of the components of the PTN-MK-RPTPβ/ζ axis in immune cells and in inflammatory diseases

### REFERENCES


suggests important roles for this axis in inflammation. Pleiotrophin has been recently identified as a limiting factor of metainflammation, a chronic pathological state that contributes to neuroinflammation and neurodegeneration. Pleiotrophin also seems to potentiate acute neuroinflammation independently of the inflammatory stimulus while MK seems to play different -even opposite- roles in acute neuroinflammation depending on the stimulus. Which are the functions of MK and PTN in chronic neuroinflammation is still a question of great biologic interest.

For its pattern of expression and its known signaling cascades involving important regulators of inflammation as Fyn kinase and ALK, RPTPβ/ζ is a target receptor for PTN and MK in neuroinflammation. Pharmacologic modulation of the PTN, MK, RPTPβ/ζ and/or its downstream effectors, Fyn and ALK, is a novel therapeutic strategy to modulate neuroinflammation, from central or peripheral origin, in different pathological contexts.

### AUTHOR CONTRIBUTIONS

GH and EG wrote the manuscript and built **Figure 1** and **Table 1**. MPR-A wrote the manuscript and built **Figure 2**.

### FUNDING

This work has been supported by grants SAF2014-56671- R from Ministerio de Economía y Competitividad of Spain, PNSD001I2015 from National Plan on Drug abuse, Ministerio de Sanidad, Servicios Sociales e Igualdad of Spain, B2017/BMD-3684 from Comunidad de Madrid and MBS18PP2 from FUSP-CEU and Banco de Santander.


resistance of adipose tissue at late pregnancy. Endocrinology 152, 4094–4105. doi: 10.1210/en.2011-0068






and chemokine responses by interleukin-34 and colony stimulating factor-1. PLoS One 11:e0167324. doi: 10.1371/journal.pone.0167324

**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 © 2019 Herradon, Ramos-Alvarez and Gramage. 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.

## Intravenously Injected Mesenchymal Stem Cells Penetrate the Brain and Treat Inflammation-Induced Brain Damage and Memory Impairment in Mice

#### *Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Kalliopi Pitarokoili, Ruhr University Bochum, Germany Alexander A. Sosunov, Columbia University, United States*

*\*Correspondence: Maryna Skok skok@biochem.kiev.ua*

#### *Specialty section:*

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

*Received: 14 January 2019 Accepted: 21 March 2019 Published: 17 April 2019*

#### *Citation:*

*Lykhmus O, Koval L, Voytenko L, Uspenska K, Komisarenko S, Deryabina O, Shuvalova N, Kordium V, Ustymenko A, Kyryk V and Skok M (2019) Intravenously Injected Mesenchymal Stem Cells Penetrate the Brain and Treat Inflammation-Induced Brain Damage and Memory Impairment in Mice. Front. Pharmacol. 10:355. doi: 10.3389/fphar.2019.00355*

*Olena Lykhmus1 , Lyudmyla Koval1 , Larysa Voytenko1 , Kateryna Uspenska1 , Serhiy Komisarenko1 , Olena Deryabina2 , Nadia Shuvalova2 , Vitalii Kordium 2,3 , Alina Ustymenko 2 , Vitalii Kyryk 2 and Maryna Skok1 \**

*1 Laboratory of Cell Receptors Immunology, Palladin Institute of Biochemistry NAS, Kyiv, Ukraine, 2 Department of Gene Technologies, State Institute of Genetic and Regenerative Medicine NAMS, Kyiv, Ukraine, 3 Department of Cell Regulatory Mechanisms, Institute of Molecular Biology and Genetics NAS, Kyiv, Ukraine*

Neuroinflammation is regarded as one of the pathogenic factors of Alzheimer disease (AD). Previously, we showed that mice regularly injected with bacterial lipopolysaccharide (LPS) possessed the AD-like symptoms like episodic memory decline, elevated amounts of amyloid beta (Aβ) peptide (1–42), and decreased levels of nicotinic acetylcholine receptors (nAChRs) in the brain. The use of mesenchymal stem cells (MSCs), which can differentiate into multiple cell types, including neurons, is an attractive idea of regenerative medicine, in particular, for neurodegenerative disorders like AD. In the present study, we aimed to investigate whether pathogenic effect of LPS on the brain and behavior of mice can be prevented or treated by injection of MSCs or MSC-produced soluble factors. Fluorescently-labeled MSCs, injected intravenously, were found in the brain blood vessels of LPS-treated mice. Mice co-injected with LPS and MSCs did not demonstrate episodic memory impairment, Aβ (1–42) accumulation, and nAChR decrease in the brain and brain mitochondria. Their mitochondria released less cytochrome *c* under the effect of Ca2+ compared to mitochondria of LPS-only-treated mice. Moreover, MSCs could reverse the pathogenic symptoms developed 3 weeks after LPS injection. Cultured MSCs produced IL-6 in response to LPS and MSCs effect *in vivo* was accompanied by additional stimulation of both micro- and macroglia. Xenogeneic (human) MSCs were almost as efficient as allogeneic (mouse) ones and regular injections of human MSC-conditioned medium also produced positive effect. These data allow suggesting MSCs as a potential therapeutic tool to cure neuroinflammation-related cognitive pathology.

Keywords: neuroinflammation, mesenchymal stem cells, nicotinic acetylcholine receptor, episodic memory, mitochondria, Alzheimer disease

### INTRODUCTION

Alzheimer disease (AD) is an age-dependent neurodegenerative disorder resulting in impairment of memory, speech, and practical habits. The brains of AD patients are characterized by cholinergic deficiency and accumulation of extracellular senile plaques formed by oligomerized amyloid beta (Aβ) peptides. AD is accompanied by inflammatory reactions; moreover, neuroinflammation often precedes the development of cognitive symptoms and may be regarded as one of the pathogenic factors causing neurodegeneration (reviewed in Skok and Lykhmus, 2016).

Cholinergic deficiency is manifested as the decrease of acetylcholine content in the brain due to decreased activity of choline acetyltransferase and the loss of nicotinic acetylcholine receptors (nAChRs) in the cognitively important brain areas leading to degeneration of cholinergic neurons (Keverne and Ray, 2005). Consequently, the use of either acetylcholine esterase (AChE) inhibitors or selective agonists of certain nAChR subtypes aimed to enhance cholinergic signaling has been suggested as a current medication for symptomatic treatment of AD patients (Russo et al., 2014).

The nAChRs are ligand-gated ion channels composed of various combinations of α (α1–α10) and β (β1–β4) subunits (reviewed in Changeux, 2012). The two main nAChR subtypes found in the brain are α7 and α4β2 (Champtiaux and Changeux, 2002; Dineley et al., 2015) and both of them are related to AD pathogenesis (Wevers et al., 1999; Guan et al., 2000; Gotti et al., 2006; Posadas et al., 2013). The α4β2 nAChR signaling underlies the pro-cognitive effects of nicotine (Picciotto et al., 2001), the absence of α4β2 nAChRs in knockout mice favors neurodegeneration upon ageing (Zoli et al., 1999), and the density of α4β2 nAChRs is decreased in people with neurodegenerative diseases including the AD (Wevers et al., 1999; Guan et al., 2000). The α7 nAChRs directly interact with Aβ to affect its proper metabolism (Wang et al., 2000; Parri and Dineley, 2010). This nAChR subtype is expressed in both the plasma membrane and mitochondria of the brain cells to support their viability (Parada et al., 2010; Gergalova et al., 2012; Lykhmus et al., 2014). In addition, the α7 nAChRs expressed in the glial cells regulate inflammatory reactions in the brain (Suzuki et al., 2006; Tyagi et al., 2010; Thomsen and Mikkelsen, 2012).

In our previous studies, we showed that mice regularly injected with bacterial lipopolysaccharide (LPS) possessed decreased levels of α7 and α4β2 nAChRs, elevated amounts of Aβ (1–42) in the brain, and demonstrated episodic memory decline (Lykhmus et al., 2015b, 2017). The LPS treatment stimulated astrocytosis in the cortex and striatum and evidently decreased the number of cells in the hippocampus and striatum. The brain mitochondria of LPS-treated mice also contained less α7 nAChRs, more Aβ peptides (1–40) and (1–42) and released more cytochrome *c* (cyt *c*) in response to apoptogenic doses of Ca2+. These data allowed us to suggest that neuroinflammation caused by external inflammatory stimuli led to α7 nAChR down-regulation, accumulation of Aβ (1–42), and mitochondria impairment resulting in memory decline and finally neurodegeneration.

Mesenchymal stem cells (MSCs) are self-renewing multipotent cells able to differentiate into multiple cell types including neurons (NIH Stem Cell Information Home Page, 2016). In addition, MSCs produce numerous trophic and growth factors affecting neurogenesis, synaptogenesis, astrocytosis and cell survival (Konala et al., 2016). The efficiency of regenerative MSC therapy has been studied in many experimental models (Kariminekoo et al., 2016) including transgenic mice bearing AD-related mutations (Yang et al., 2013; Chang et al., 2014; Shin et al., 2014) and their role in modulating inflammation (Klinker and Wei, 2015; Zachar et al., 2016) have been demonstrated. In addition, the regenerative potential of MSC-secreted factors vs. the cells *per se* is being widely discussed (Konala et al., 2016).

In the present study, we put an aim to investigate whether pathogenic effect of LPS on the brain and behavior of mice can be prevented or reversed by MSCs, and if yes, whether the effect can be reproduced by MSC-produced soluble factors.

### MATERIALS AND METHODS

### Materials

All reagents were of chemical grade and purchased from Sigma-Aldrich (Saint Louis, USA), unless specially indicated. Antibodies against α3(181–192), α4(181–192), α7(179–190), α9(11–23), β2(190–200) or β4(190–200) nAChR fragments and rabbit cyt *c*-specific antibodies were generated using methods previously developed in our lab (Skok et al., 1999; Koval et al., 2004, 2011; Lykhmus et al., 2010; Gergalova et al., 2014). The antibodies were biotinylated according to standard procedures (Harlow and Lane, 1988). Antibodies against Aβ (1–42), ionized calcium binding adaptor molecule 1 [(Iba-1) or allograft inflammatory factor (AIF-1)] and neutravidin-peroxidase conjugate were purchased from ALT Ukraine Ltd (representing Thermo Fisher Scientific in Ukraine). Rabbit polyclonal antibody against glial fibrillary acidic protein (GFAP) was from Dako (Agilent Technologies); goat antirabbit IgG Alexa 488-labeled and IL-6-specific antibody pair were from Invitrogen.

### Animals

As MSCs recipients, we used C57BL/6 J female mice 3–5 months of age. Placental MSCs were obtained from 6 week-old female FVB wild-type pregnant mice, and GFP-labeled MSCs were obtained from FVB-Cg-Tg (GFPU) 5Nagy/J mice, transgenic by green fluorescent protein (GFP) gene. Mice were kept in the animal facilities of either Palladin Institute of Biochemistry NAS of Ukraine or Dmitry F. Chebotarev Institute of Gerontology NAMS of Ukraine in Kyiv. Mice were housed in quiet, temperature-controlled rooms, and provided with water and food pellets *ad libitum*. Before removing the brains, mice were

**Abbreviations:** AD, Alzheimer disease; cyt *c*, cytochrome *c*; GFP, green fluorescence protein; LPS, lipopolysaccharide; MSCs, mesenchymal stem cells; nAChR, nicotinic acetylcholine receptor.

sacrificed by cervical dislocation. All procedures complied with the ARRIVE guidelines, were carried out in accordance with the Directive 2010/63/EU for animal experiments and were approved by the Animal Care and Use Committee of Palladin Institute of Biochemistry.

### Mesenchymal Stem Cells Isolation, Propagation, and Characterization

Human MSCs (hMSCs) were obtained from Wharton jelly (WJ) using the explant method (Maslova et al., 2013; Shuvalova and Kordium, 2016). Umbilical cords were collected from healthy donors (39–40 weeks of gestation) after their consent. The umbilical cord fragment (5–10 cm) was washed with PBS, the vessels were mechanically removed. WJ was cut into 0.4–0.5 mm pieces that were placed in the 75 cm2 cultural flasks containing complete growth medium α-MEM (BioWest, Austria) supplemented with 10% fetal bovine serum (Invitogen), penicillin 100 U/ml (Arterium, Ukraine), and streptomycin 100 μg/ml (Arterium, Ukraine). Cultivation was performed under conditions of humidified air with 5% CO2 at 37°C. The medium was changed every 4–5 days. The first attached cells were visible on the 7–10th day. After 14 days the clones reached the size and confluence (70–80%) sufficient for passing, which was performed by standard method with the use of trypsin-EDTA mixture (Shuvalova et al., 2013). The surface marker proteins CD34, CD45, CD90, CD73, CD105 expression was determined at the second passage by flow cytometry (BD FACS Aria) with FITC- and PE- conjugated antibodies (UsBiological, USA) according to minimal criteria for defining multipotent mesenchymal stromal cells (Dominici et al., 2006). The cells of the second passage were used for both *in vitro* assays and transplantation into LPS-treated mice.

Murine placental multipotent mesenchymal stem cells (mMMSCs, further mMSCs) were obtained from FVB-Cg-Tg (GFPU) 5Nagy/J mice 19th day of pregnancy. Under sterile conditions, placentae were transferred into a Petri dish with cold PBS, containing antibiotics. Fetal membranes were minced and incubated with 0.1% collagenase type I (Sigma-Aldrich, USA) for 90 min at 37°C. Cell pellet obtained after digestion and filtration was washed and seeded in 75 cm2 flasks containing culture medium DMEM-LG (Low Glucose, 1 g/L) supplemented with 10% fetal bovine serum, penicillin 100 U/ml, streptomycin 100 μg/ml and 1:100 nonessential amino acids (Sigma-Aldrich, USA). Cultivation was carried out in CO2-incubator under conditions of humidified air with 5% CO2 at 37°C. The medium was changed every 3–4 days. After approximately 14 days, the cells were rinsed with Dulbecco's Phosphate Buffered Saline (Sigma-Aldrich, USA), and then exposed to pre-warmed trypsin– EDTA (0.25% trypsin, 4 mM EDTA, Invitrogen) for 2 min. The resulting detached cells were resuspended in serumsupplemented medium, counted and seeded as first passage cultures at 4,000 cells per cm2 . Subcultivation was performed at 80% confluence of the monolayer. Cells of the second passage were used in the experiment (Fhilho and Oliveira, 2012).

Phenotyping of cells for markers CD34, CD44, CD45, CD73, CD90, CD105 was performed using fluorochrome-labeled monoclonal antibodies to mouse membrane antigens by flow cytometry. Obtained cell cultures satisfied criteria of MMSCs by phenotype and ability to directed multilinear differentiation.

### *In vitro* Assays

Mouse MSCs (4 × 104 cells per well) were seeded in 96-well tissue culture plates containing complete growth medium DMEM/F12 supplemented with 10% fetal bovine serum, penicillin 100 U/ml, streptomycin 100 μg/ml (all–Sigma-Aldrich, USA) and were cultured in the presence of different doses of LPS at 37°C and 5% CO2 during 72 h. Then, the cell supernatant was collected and the cells' quantity/viability was studied in MTT test (Carmichael et al., 1987). The supernatants were tested for the presence of IL-6 using the Murine IL-6 ELI-Pair kit from Diaclone (Gen-Probe, France), according to manufacturer's instructions.

### Animal Treatment and Brain Preparations

In the first set of experiments, three groups of C57Bl/6 mice, eight animals per group, were intraperitoneally injected with 2 mg kg−1 LPS (*E. coli* strain 055:B5) in 0.1 ml of saline. Two of these groups, in addition, obtained intravenously, in the tail vein, 106 mMSCs or hMSCs in 0.1 ml of incubation medium. Three weeks thereafter, mice were examined in behavioral novel object recognition test, sacrificed and their brains were removed for examination.

In the second set of experiments, three groups of mice, five animals in each, were injected with LPS as described above. After 3 weeks, the mice were examined in behavioral test and one group obtained hMSCs (106 in the tail vein), while another group was injected intraperitoneally with 0.3 ml of hMSCconditioned medium obtained after 2 days of cells incubation in serum-free medium. Injections of conditioned medium were repeated every 7 days for 3 weeks more and mice were examined in behavioral test every week. In a month, a week after the last conditioned medium injection, mice were sacrificed and their brains were removed for examination.

To study if intravenously introduced MSC penetrate into the brain parenchyma, GFP-labeled mMSCs (106 per animal) were injected into two mice pre-injected with LPS a day before. Mice were sacrificed 24 and 72 h thereafter and their brains were removed for examination.

For sandwich ELISA experiments and mitochondria examination, the mouse brains were homogenized in a glass homogenizer. The homogenate was fractionated into mitochondria and non-mitochondria by standard procedure of differential centrifugation (Gergalova et al., 2012; Lykhmus et al., 2015a). The purity of fractions obtained was evaluated by ELISA using the antibodies against different cellular compartments, as described previously (Uspenska et al., 2017). Live mitochondria were further examined in functional test of cytochrome c (cyt *c*) release (see below), while the pellets of both mitochondria and non-mitochondria fractions were used to prepare the detergent lysates, as described previously (Lykhmus et al., 2015b). Protein concentration was measured with the BCA kit (Thermo Scientific, France).

For immunohistochemistry studies, the brains were fixed in 4% paraformaldehyde for 48 h, washed in PBS, dehydrated with increasing concentrations of alcohol and embedded in ParaplastX-TRA (McCormick Scientific LLC). The paraplastembedded specimens were cut into serial frontal 5 μM sections with rotational microtome (HM 325, MICROM International GmbH). The sections were placed onto adhesive microscopic slides to be further examined by immunohistochemistry.

To study MSC penetration into the brain, the brains were fixed in 4% paraformaldehyde and cut by vibratome into coronal 40 μm sections. The floated sections were placed onto microscopic slides to be examined by confocal microscopy.

### ELISA Assays

To determine the level of various nAChR subunits within the brain or mitochondria preparations, the immunoplates (Nunc, Maxisorp) were coated with rabbit α7(1–208)-specific antibody (20 μg/ml), blocked with 1% BSA, and the detergent lysates of brain tissue or mitochondria were applied into the wells (1 μg of protein per 0.05 ml per well) for 2 h at 37°C. The plates were washed with water and the second biotinylated α3(181–192)-, α4(181–192)-, α7(179–190)-, α9(11–23)-, β2(190–200)- or β4(190–200)-specific antibody was applied for additional 2 h being revealed with Neutravidin-peroxidase conjugate and *o*-phenylendiaminecontaining substrate solution.

To determine the level of Aβ (1–42) bound to α7 nAChR, the plates were coated with α7(1-208)-specific antibody, and the α7-Aβ complex from the brain or mitochondria preparation applied as described above was revealed with biotinylated Aβ (1–42)-specific antibody, Neutravidin-peroxidase conjugate and *o*-phenylendiamine-containing substrate solution. The optical density was read at 490 nm using Stat-Fax 2000 ELISA Reader (Awareness Technologies, USA).

### Immunohistochemistry and Confocal Microscopy

Before immunohistochemical staining, the brain sections were de-paraffinated by standard procedure; the nonspecific binding was blocked with 1% BSA in PBS (30 min, RT). For staining the Aβ (1–42), the slides were incubated with biotinylated mouse Aβ (1–42)-specific antibody (1:200) overnight at room temperature, washed with PBS and incubated with Extravidin-Cy3 (1,200) in 1% BSA-containing PBS for 1 h at RT. Nonspecific binding was blocked with goat anti-mouse IgG.

For staining the astrocytes, the slides were incubated with rabbit anti-GFAP antibody (1:100) followed by goat anti-rabbit-Alexa 488.

For staining the microglia, the slides were incubated with biotinylated anti-Iba-1 antibody (1:100) followed by Streptavidin-Cy3.

As negative controls, the incubations without primary antibody were performed in both control and experimental sections.

All slides with paraplast sections were embedded in MOWIOL-DABCO, while slides with floated brain sections were embedded in Vectashield and examined under Zeiss LSM 510 Meta confocal lasers canning microscope (Zeiss, Germany). The brain regions were identified according to Paxinos and Franklin, 2001.

### Mitochondria Functional Assay

The purified live mitochondria (120 μg of protein per ml) were incubated with either 0.1 μM CaCl2 or 0.9 μM CaCl2 for 5 min at room temperature and were immediately pelleted by centrifugation (10 min, 7,000 g) at 4°C. The incubation medium contained 10 mM HEPES, 125 mM KCl, 25 mM NaCl, 5 mM sodium succinate and 0.1 mM Pi(K), pH 7.4. The mitochondria supernatants were collected and tested for the presence of cyt *c* by sandwich assay as described previously (Gergalova et al., 2012, 2014).

### Behavioral Experiments

Mice of all groups were tested in the "Novel Object Recognition" (NOR) behavioral test (Antunes and Biala, 2012; Lykhmus et al., 2015a,b) prior and post-treatments. Briefly, the animals were individually placed in a rectangular novel open field containing two identical objects with distinctive features (shape and texture). The animals were subjected to a 10-min session of exploration of the objects followed by 10 min in a waiting cage. During the second 10-min session, one of the objects was replaced by a novel one, and we scored the time spent in contact with each object. It is widely acknowledged that rodents spontaneously explore novel objects by touching the objects with their nose and prefer novel objects to familiar ones that reflect their episodic memory (Save et al., 1992; Thinus-Blanc, 1996). Therefore, the time spent in contact with each object reflects the time of exploration for the object. The results of NOR test are presented as discrimination index (DI) calculated as the difference in the number of "novel" and "famous" object explorations divided by the total number of explorations of two identical objects.

### Statistical Analysis

ELISA experiments have been performed in triplicates and mean values for individual mice were used for statistical analysis assessed using one-way ANOVA test. Behavioral tests were also performed in triplicate for each mouse and mean values for individual mice were taken for statistical analysis. The data are presented as mean ± SD; \**p* < 0.05; \*\* *p* < 0.005; \*\*\* *p* < 0.0005.

### RESULTS

### Both Mouse and Human MSCs Prevent Pathogenic Effect of LPS on Mouse Brains and Behavior

We used MSCs from two sources: either human umbilical cord (hMSCs) or mouse placenta (mMSCs). Both types of cells satisfied minimal criteria for defining multipotent mesenchymal stromal cells according to CD34, CD44, CD45, CD90, CD73, CD105 surface markers expression (Dominici et al., 2006) and were used for transplantation after two passages *in vitro*. Morphology of mMSCs in primary culture and at the second passage is shown in **Figures 1A,B**. Both mMSCs and hMSCs were shown to proliferate and to produce IL-6 in response to LPS stimulation (data for mMSCs are shown in **Figures 1C,D**).

In the first set of experiments, mice were injected intraperitoneally with LPS and intravenously with either hMSCs or mMSCs (106 cells per mouse) at the same time. Three weeks later, they were examined in memory test and their brains and brain mitochondria were studied for the levels of α7 nAChR, Aβ (1–42) and cyt *c* released upon Ca2+ stimulation.

It was found that MSCs injection prevented α7 nAChR decrease, Aβ (1–42) accumulation and increased cyt *c* release from mitochondria, as well as episodic memory decline caused by LPS. Human MSCs were almost as efficient as mouse ones in a majority of tests (**Figure 2**).

The use of GFP-labeled mouse MSCs demonstrated that green fluorescent signal could be found along/around the brain blood vessels 3 days after intravenous injection (**Figure 3**).

### Both Human MSCs and Their Conditioned Media Reverse the Pathogenic Effects Developed 3 Weeks After LPS Injection

The second set of experiments was undertaken to find out whether the MSCs effect is due to cellular or humoral influence and whether it is only prophylactic or can also be therapeutic, i.e., cure the already developed pathogenic symptoms caused by LPS.

For this purpose, mice were injected intravenously with xenogeneic (human) MSCs 3 weeks after LPS injection when memory decline was already observed. Another group of LPS-treated mice was injected once per week intraperitoneally with the conditioned culture medium in which hMSCs were grown to 80% confluency and were maintained for 2 days without serum. Mice were examined in memory test every week thereafter, then sacrificed and their brains and brain mitochondria were studied as in the previous set of experiments. In this case, we studied a broader range of nAChR subunits in both the mitochondrial and nonmitochondrial brain fractions. It was found that either hMSCs or their supernatants up-regulated α4, α9 and β2 nAChR subunits in their brains (**Figure 4A**) and brain mitochondria (**Figure 4C**); MSCs additionally up-regulated β4 subunits and decreased the level of Aβ (1–42) (**Figures 4B,D**). Either hMSCs or their supernatants also restored memory of LPS-treated mice and significantly improved mitochondria sustainability to Ca2+ (**Figures 5A,B**). In contrast to MSCs, which supported memory of LPS-treated mice for at least 3 weeks, the effect of a single injection of conditioned medium was transient and disappeared after 2 weeks (**Figure 5A**).

The visible decrease of Aβ (1–42) accumulated in the frontal cortex, striatum and hippocampus of LPS injected mice under the effect of either MSC or their conditioned medium was found by means of immunohistochemistry (**Figure 6**).

Finally, MSC visibly increased the green signal for GFAP (**Figure 7**), while their conditioned medium increased Iba-1-specific staining (**Figure 8**) in the brains of LPS-treated mice. Although no quantitative analysis has been performed in this experiment, the data allow suggesting that the treatments increased the number of activated astrocytes (GFAP) or microglial cells (Iba-1).

### DISCUSSION

The data obtained in the first set of experiments indicated that MSCs injected intravenously prevent the pathogenic effect of LPS on the brain, brain mitochondria and behavior of LPS-treated mice. In particular, they prevented the α7 nAChR decrease, Aβ (1–42) accumulation, mitochondria and episodic memory impairment caused by LPS injection. Xenogeneic (human) MSCs were almost as efficient as allogeneic (mouse) ones suggesting that their effect was mostly due to soluble growth/trophic factors produced. This suggestion was approved in the second set of experiments where regular injections of hMSC-conditioned medium appeared to be almost as efficient as MSC themselves. In addition, it was found that MSCs can not only prevent the pathogenic effect of LPS, but also reverse the already developed nAChR decrease, Aβ (1–42) accumulation, mitochondria, and memory impairment. The anti-LPS effect of MSCs was accompanied by additional stimulation of both micro- and macroglia suggesting that these cells are the targets of MSC-produced soluble factors. We can speculate that positive effect of MSCs is, at least partly, due to activation of trophic functions of glial cells.

The use of multipotent stem cells is an attractive idea of regenerative medicine, in particular, for neurodegenerative disorders like AD. The therapeutic effect of MSCs has been already studied in several AD models. The first studies were performed with the bone-marrow-derived MSCs transplanted intracerebrally (Lee et al., 2009, 2010a,b, 2012; Zhang et al., 2012; Bae et al., 2013) in mice either pre-injected with Aβ (Lee et al., 2009, 2010a) or transgenic for APP/PS1 (Lee et al., 2010b, 2012; Bae et al., 2013). A big piece of studies was performed with human MSCs from adipose tissue, a good source of autologous MSCs (Ma et al., 2013; Chang et al., 2014; Yan et al., 2014). Human umbilical cord MSCs were also transplanted intracerebrally in APP/PS1 mice (Lee et al., 2012; Yang et al., 2013). In all cases, the improvement of

cognitive (memory) and synaptic functions, as well as reduced Aβ deposition has been observed that is in accord with our data. Advantage of our model is that the Aβ accumulation in the brain occurred in a natural way in genetically non-modified mice. The wild-type mice do not form senile plaques; however, significant accumulation of soluble Aβ could be observed and it was significantly decreased upon MSC injection. Moreover, we show that MSCs can not only prevent the pathogenic effect of LPS, but cure the already developed pathogenic symptoms including memory impairment.

In contrast to the studies cited above, we injected MSCs intravenously. The brain is protected from penetration of peripheral cells by the blood-brain barrier. However, the integrity of the cerebral vasculature is compromised following inflammation (Zlokovic, 2008). In particular, a diminished function of the blood-brain barrier is an early event in multiple sclerosis when inflammation facilitates the massive influx of leukocytes into the brain parenchyma inducing demyelination, tissue damage and axonal disfunction (Kamphuis et al., 2015). Previously, we reported that the blood-brain barrier in

LPS-treated mice had been disturbed to allow the nAChRspecific antibody penetration (Lykhmus et al., 2015b). Here, we show that MSCs can be found along the brain vessels of LPS-pre-treated mice 3 days after intravenous injection. This is in accord with the data of Lee et al., who found intraarterially injected MSCs in the brain vasculature (Lee et al., 2016). It was also shown that, at least *in vitro*, MSCs possess leukocytelike molecular mechanisms enabling interaction with vascular endothelial cells (Rüster et al., 2006). Therefore, it is reasonable to expect that MSCs can penetrate the brain parenchyma *via* transmigration. Indeed, intravenously injected mouse bone marrow cells or human adipose-derived MSCs did cross the blood-brain barrier and migrated into the brain in a rat (Salem et al., 2014) or mouse models for AD (Kim et al., 2012). The AD pathology in humans is accompanied by elevated pro-inflammatory cytokines, which can affect the integrity of the blood-brain barrier (Zenaro et al., 2017; Toropova et al., 2018). However, whether its damage is comparable to that found upon multiple sclerosis and other autoimmune pathologies of the central nervous system and whether it is sufficient to allow intravenously injected MSC to penetrate the brain parenchyma is a reasonable question which needs to be addressed.

The important fact, not examined in previous studies, is that MSC injection restores/elevates the level of nAChRs in the brain and brain mitochondria decreased as a result of LPS treatment. The MSCs up-regulated α4, α7, α9, β2 and β4 subunits in the brain and brain mitochondria, while the conditioned medium did not affect α7 subunits decreased by LPS. Both α7 and α4β2 nAChRs expressed in the brain are involved in regulating cognition and memory (Gotti et al., 2006) and mitochondriaexpressed α7β2, α4β2, and α9 nAChRs are involved in the anti-apoptotic pathways (Gergalova et al., 2012, 2014; Lykhmus et al., 2014; Uspenska et al., 2018); therefore, their increased levels make mitochondria more resistant to apoptogenic influence and support the viability of brain cells. In one of the published papers, MSCs were introduced together with galantaminecontaining nanoparticles; that was expected to inhibit AChE and increase cholinergic signaling (Misra et al., 2016). Our data indicate that MSCs themselves contribute to cholinergic signaling in the brain by increasing the level of nAChRs.

Another important question arising from positive effects of MSC in AD models is whether they are due to direct involvement of differentiated MSC into neuronal networks in the brain (the true regenerative medicine) or are mediated by numerous trophic and growth factors produced by MSCs to stimulate the host cells. The latter possibility is therapeutically important because it allows avoiding cell transplantation by substituting it with the use of MSC-produced substances. Previous studies demonstrated that human MSCs stimulated neurogenesis both *in vitro* (Park et al., 2016) and *in vivo* (Yan et al., 2014; Kim et al., 2015; Oh et al., 2015) by producing soluble factors like activin A, growth differentiation factor-15 and activating Wnt signaling pathway in neuronal progenitor cells. Moreover, it was found that MSCs produce extracellular vesicles (exosomes), which contain neprilysin, enzyme involved in Aβ degradation (Katsuda et al., 2013), and these exosomes can be suggested for AD therapy (Katsuda et al., 2015). We show here that supernatants of hMSC culture, applied intraperitoneally, improved memory of LPS-treated mice and affected their brains almost similarly to MSCs. However, in contrast to MSCs, which improved the state of mice for at least 3 weeks after a single injection, the effect of MSC supernatant was transient and regular injections were required to maintain the improvement achieved. Additional studies are required to establish whether MSC-conditioned medium can provide a stable therapeutic effect.

Finally, it is not definitely clear which cells in the brain are the targets for MSC-produced factors. The effects of MSC co-culture with neuronal progenitor cells suggested the direct influence on the brain neurons (Oh et al., 2015; Park et al., 2016). However, other studies demonstrated that intracerebrally transplanted human adipose MSCs activate microglia around senile plaques in the brain of APP/PS1 transgenic mice (Ma et al., 2013). We observed a visible activation of both microglia and astrocytes in the brains of MSC-injected mice, additional to that induced by LPS, and showed that MSCs produce IL-6 in response to LPS stimulation *in vitro*. Therefore, injecting MSCs simultaneously with LPS could stimulate MSCs for IL-6 production. IL-6 is a pro-inflammatory cytokine, but is also known as a neurotrophic factor (Hirota et al., 1996; Wagner, 1996). Recently, it was reported that mesenchymal progenitor cells derived from induced pluripotent stem cells enhance neuritogenesis *via* neurotrophin and cytokine (including IL-6) production (Brick et al., 2018). Therefore, IL-6 may be one of neurotrophic soluble factors produced by MSCs penetrating the brain that can affect both neurons and glial cells.

The data obtained put a wider question on what happens in the mouse brain under the effect of injected MSCs. The ability of MSCs to prevent LPS pathogenic effect (simultaneous injection of LPS and MSCs) indicates that MSCs do not allow such effect to develop. However, positive MSCs effect when pathological symptoms have already developed (MSCs injection 3 weeks after LPS) allows suggesting that MSCs, or their soluble factors, directly or indirectly (by activating glial cells) restore the activity of damaged brain neurons. Further experiments

### REFERENCES


are needed to reveal whether the recovery observed under the MSCs effect is long-lasting or just temporal.

### CONCLUSIONS


### ETHICS STATEMENT

This study was carried out in accordance with the recommendation of the guidelines of the Animal Care and Use Committee of Palladin Institute of Biochemistry, Kiev. The protocol was approved by the IACUC of Palladin Institute of Biochemistry.

### AUTHOR CONTRIBUTIONS

MS, OD, SK, OL, and VKo made substantial contributions to the conception or design of the work. LK, OL, LV, KU, MS, NS, and AU contributed to acquisition, analysis, and interpretation of data for the work. MS drafted the work. OD, VKy, and AU revised it critically for important intellectual content. LK, OL, LV, KU, SK, VKo, OD, NS, VKy, and AU finally approved the version to be published. LK, OL, LV, KU, SK, VKo, OD, NS, VKy, and AU agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.


through kinase-mediated pathways. *Int. J. Biochem. Cell Biol.* 49, 26–31. doi: 10.1016/j.biocel.2014.01.001


Paxinos, G., and Franklin, K. B. J. (2001). *The mouse brain in stereotaxic coordinates*. (New York: Academic Press).


**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 © 2019 Lykhmus, Koval, Voytenko, Uspenska, Komisarenko, Deryabina, Shuvalova, Kordium, Ustymenko, Kyryk and Skok. 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.*

# A Bcr-Abl Inhibitor GNF-2 Attenuates Inflammatory Activation of Glia and Chronic Pain

Gyun Jee Song1,2† , Md Habibur Rahman<sup>3</sup>† , Mithilesh Kumar Jha3,4 , Deepak Prasad Gupta1,3, Sung Hee Park<sup>1</sup> , Jae-Hong Kim<sup>3</sup> , Sun-Hwa Lee<sup>5</sup> , In-Kyu Lee<sup>6</sup> , Taebo Sim7,8, Yong Chul Bae<sup>9</sup> , Won-Ha Lee<sup>10</sup> and Kyoungho Suk<sup>3</sup> \*

<sup>1</sup> Department of Medical Science, College of Medicine, Catholic Kwandong University, Gangneung-si, South Korea, <sup>2</sup> Translational Brain Research Center, International St. Mary's Hospital, Catholic Kwandong University, Incheon, South Korea, <sup>3</sup> Department of Pharmacology, Brain Science and Engineering Institute, BK21 Plus KNU Biomedical Convergence Program, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>4</sup> Department of Neurology, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, <sup>5</sup> New Drug Development Center, Daegu Gyeongbuk Medical Innovation Foundation, Daegu, and VORONOI Inc., Incheon, South Korea, <sup>6</sup> Department of Internal Medicine, Division of Endocrinology and Metabolism, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>7</sup> KU-KIST Graduate School of Converging Science and Technology, Korea University, Seoul, South Korea, <sup>8</sup> Chemical Kinomics Research Center, Korea Institute of Science and Technology, Seoul, South Korea, <sup>9</sup> Department of Anatomy and Neurobiology, School of Dentistry, Kyungpook National University, Daegu, South Korea, <sup>10</sup> BK21 Plus KNU Creative BioResearch Group, School of Life Sciences, Kyungpook National University, Daegu, South Korea

Edited by: Morena Zusso, University of Padua, Italy

### Reviewed by:

Zhigang Liu, Northwest A&F University, China Katsura Takano, Osaka Prefecture University, Japan

> \*Correspondence: Kyoungho Suk ksuk@knu.ac.kr

†These authors have contributed equally to this work

#### Specialty section:

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

Received: 25 January 2019 Accepted: 30 April 2019 Published: 20 May 2019

#### Citation:

Song GJ, Rahman MH, Jha MK, Gupta DP, Park SH, Kim J-H, Lee S-H, Lee I-K, Sim T, Bae YC, Lee W-H and Suk K (2019) A Bcr-Abl Inhibitor GNF-2 Attenuates Inflammatory Activation of Glia and Chronic Pain. Front. Pharmacol. 10:543. doi: 10.3389/fphar.2019.00543 GNF-2 is an allosteric inhibitor of Bcr-Abl. It was developed as a new class of anti-cancer drug to treat resistant chronic myelogenous leukemia. Recent studies suggest that c-Abl inhibition would provide a neuroprotective effect in animal models of Parkinson's disease as well as in clinical trials. However, the role of c-Abl and effects of GNF-2 in glia-mediated neuroinflammation or pain hypersensitivity has not been investigated. Thus, in the present study, we tested the hypothesis that c-Abl inhibition by GNF-2 may attenuate the inflammatory activation of glia and the ensuing pain behaviors in animal models. Our results show that GNF-2 reduced lipopolysaccharide (LPS)-induced nitric oxide and pro-inflammatory cytokine production in cultured glial cells in a c-Abldependent manner. The small interfering ribonucleic acid (siRNA)-mediated knockdown of c-Abl attenuated LPS-induced nuclear factor kappa light chain enhancer of activated B cell (NF-κB) activation and the production of pro-inflammatory mediators in glial cell cultures. Moreover, GNF-2 administration significantly attenuated mechanical and thermal hypersensitivities in experimental models of diabetic and inflammatory pain. Together, our findings suggest the involvement of c-Abl in neuroinflammation and pain pathogenesis and that GNF-2 can be used for the management of chronic pain.

Keywords: GNF-2, c-Abl, glia, neuroinflammation, pain

## INTRODUCTION

Neuroinflammation is highly associated with several neurodegenerative diseases including Alzheimer's disease (AD), Parkinson's disease (PD), and chronic pain (Katsulov and Mazneikova, 1987; Mosley et al., 2006; Tansey et al., 2007; Calsolaro and Edison, 2016; Chen et al., 2018). The chronic pain pathophysiology is complex and includes peripheral and central neuronal alterations and neuroinflammation. The neuroinflammatory process is characterized by the activation of microglia and astrocytes, macrophage infiltration, release of diverse proinflammatory mediators

[e.g., nitric oxide (NO), cytokines, and chemokines]. This process leads to neuronal death or neurodegeneration (Frank-Cannon et al., 2009; Song and Suk, 2017). In this regard, important clues to the molecular mechanisms of neuropathic pain may be found by closely examining the microglial inflammatory activation and neuroinflammation (Carniglia et al., 2017; Kiguchi et al., 2017; Chen et al., 2018). Therefore, studies which target inflammatory mediators may provide novel therapeutic approaches for chronic pain management.

Non-receptor tyrosine kinase c-Abl activation is associated with AD and PD pathogenesis in human and animal models. c-Abl phosphorylation is robustly increased in brain samples from AD and PD patients as well as in animal models of AD, PD, and synucleinopathies (Ko et al., 2010; Imam et al., 2011; Vargas et al., 2018). Furthermore, imatinib or nilotinib, which are FDA-approved c-Abl inhibitors, showed neuroprotective effects when administered in animal models of PD and AD (Cancino et al., 2008; Hebron et al., 2013). More recently, activated c-Abl was observed in the spinal cord of G93A-SOD1 transgenic mice, a widely-used model of amyotrophic lateral sclerosis (ALS). This study revealed that the administration of dasatinib (a c-Abl inhibitor) improved the innervation status of neuromuscular junctions (Katsumata et al., 2012). It is quite well known that oxidative stress-induced c-Abl activation leads to nuclear factor kappa light chain enhancer of activated B cell (NF-κB) activation and neuronal death (Xiao et al., 2011). However, most studies have focused on the neuroprotective effects of c-Abl inhibitors and related molecular mechanisms in neurons. Recent findings on pain pathogenesis demonstrate that glial cells, particularly microglia and astrocytes, are an important source of inflammatory mediators fundamentally involved in the pathogenesis of inflammatory and neuropathic pain (Carniglia et al., 2017; Chen et al., 2018). Therefore, it is necessary to study the function of glial c-Abl in the pathogenesis of both inflammatory and neuropathic pain.

GNF-2 is a selective allosteric inhibitor of Bcr-Abl (the oncogenic fusion protein of Bcr and c-Abl caused by reciprocal chromosomal translocations), which was developed as an anticancer drug (Zhang et al., 2010; Rossari et al., 2018). GNF-2 binds to the myristate-binding site of c-Abl, leading to improved pharmacokinetic properties (Fabbro et al., 2010; Zhang et al., 2010). GNF-2 is a very selective non-ATP competitive inhibitor of Bcr-Abl and c-Abl. Unlike other inhibitors, it does not show activity against many other kinases such as fms-like tyrosine kinase 3, platelet-derived growth factor receptor, Janus kinase-1, tyrosine-protein kinase Met. Therefore, in this study, we used GNF-2 to assess the effect of c-Abl on neuroinflammation and associated pain pathogenesis using multiple pain models. It has been reported that reactive microglia release a various array of toxic molecules including pro-inflammatory cytokines, NO, and superoxide, which have been shown to play a complex role in the pathogenesis of neuropathic pain. However, the effects of c-Abl inhibition by GNF-2 on neuroinflammation and associated chronic pain pathogenesis remain elusive. Thus, in the present study, we investigated the role of c-Abl in the inflammatory activation of glia and their contribution to the pathogenesis of inflammatory and neuropathic pain by the in vitro and in vivo application of the Bcr-Abl inhibitor GNF-2.

### MATERIALS AND METHODS

### Materials

GNF-2 and methylated GNF-2 compounds were prepared as described previously (Adrian et al., 2006). Lipopolysaccharide (LPS) was purchased from Sigma-Aldrich. It was obtained from Escherichia coli 0111:B4 prepared by phenolic extraction and gel filtration chromatography. Recombinant mouse interferon-γ (IFN-γ) protein was purchased from R&D Systems. The c-Abl siRNA (1:1 mix of siRNA #2 and #3) and control siRNA were purchased from Genolution Pharmaceuticals (Seoul, South Korea); siCont- 5<sup>0</sup> -CCUCGUGCCGUUCCAUCAGG UAGUU-3<sup>0</sup> , siAbl-#2, 5<sup>0</sup> -GCAACAAGCCCACUAUCUAUU-3<sup>0</sup> , siAbl-#3, 5<sup>0</sup> -UGAUGAAGGAGAUCAAACAUU-3<sup>0</sup> .

### Cell Culture

BV-2 immortalized murine microglial cell line was maintained in Dulbecco's modified Eagle's medium (DMEM) containing 5% heat-inactivated fetal bovine serum (FBS) and 50 mg/ml gentamicin at 37◦C. For mouse primary mixed glial cells (MGCs) culture, the brains of 3-day old C57BL6 mice were isolated and homogenized and mechanically disrupted by a nylon mesh. The MGCs were seeded in poly-L-lysine-coated culture flasks with DMEM containing 10% FBS, 100 U/ml of penicillin, and 100 µg/ml of streptomycin (Gibco, Grand Island, NY, United States) and allowed to grow at 37◦C in a humidified atmosphere with 5% CO2. Culture medium was changed initially after 5 days and then changed every 3 days. After 14 days of culture, MGCs (mixed microglial and astrocytes) were prepared by trypsinization, as previously described (Song et al., 2016). The collected cells were further plated using the same media condition and used for experiments.

### Nitric Oxide Production

The BV-2 cells (4 × 10<sup>4</sup> cells/well in 96-well plates) were treated with 100 ng/ml of LPS and the level of NO production was assessed by measuring the amount of nitrite as previously described (Lee et al., 2009). Briefly, After 24-h of incubation, 50 µl of the cell culture media was mixed with an equal volume of a Griess reagent (0.1% naphthylethylenediamine dihydrochloride and 1% sulfanilamide in 5% phosphoric acid) in a 96-well microtiter plate. Absorbance at 540 nm was measured on a microplate reader. Sodium nitrite was used as the standard curve to calculate NO concentration.

### Assessment of Cell Viability

Both BV-2 microglia and primary MGCs (4 × 10<sup>4</sup> cells/well in 96-well plates) were used to measure cell viability using 3-(4, 5 dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT; Sigma-Aldrich) assay, as previously described (Song et al., 2016). After 24 h of LPS treatment, the culture media was removed and MTT (0.5 mg/ml in PBS) was added to the cells, which were then incubated at 37◦C for 2 h in a 5% CO<sup>2</sup> incubator. The insoluble

formazan crystals were completely dissolved in DMSO. The absorbance at 570 nm was measured using a microplate reader.

### Enzyme-Linked Immunosorbent Assay (ELISA) for TNF-α

The BV-2 cells or primary cells were treated with LPS either in the presence or absence of GNF-2 for 24 h. The concentration of TNF-α protein in the culture media was assessed using a rat monoclonal anti-mouse TNF-α antibody (capture antibody), and a goat biotinylated polyclonal antimouse TNF-α antibody (detection antibody), as described in the product manual (ELISA development reagent; R&D systems, Minneapolis, MN, United States). The recombinant TNF-α protein was used as a standard.

### Small Interfering Ribonucleic Acid (siRNA)-Mediated Knockdown of the c-Abl Gene

Cells were transfected with siRNAs using LipofectamineTM iMAX (Invitrogen, Carlsbad, CA, United States), based on the manufacturer's instructions. The cells were used after 48 h of transfection.

### Traditional and Real-Time Reverse Transcription Polymerase Chain Reaction (RT-PCR)

Total ribonucleic acid (RNA) was extracted from the treated cells or tissues (spinal cord and brain) using TRIZOL reagent (Invitrogen, Carlsbad, CA, United States). Reverse transcription (RT) was conducted using the Superscript II reverse transcriptase (Invitrogen) and an oligo (dT) primer. Traditional PCR amplification was done using specific primer sets at 55–60◦C as annealing temperature and 25–32 cycles in a C1000 Touch Thermal Cycler (Bio-Rad, Richmond, CA, United States). PCR products with ethidium bromide were electrophoresed on a 1% agarose gel, and bands were observed under ultraviolet light for analysis. Real-time PCR was performed using One Step SYBR PrimeScript RT-PCR Kit (Takara Bio, Otsu, Shiga, Japan), according to the manufacturer's instructions, followed by detection using the ABI Prism 7000 Sequence Detection System (Applied Biosystems, California, CA, United States). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an internal control. The primer sequences were designed based on published complementary deoxyribonucleic acid (cDNA) sequences (**Table 1**).



### Western Blotting Analysis

Cells or brain tissues were lysed in 300 µl of lysis buffer [150 mM sodium chloride, 1% Triton X-100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulphate (SDS), 50 mM Tris-HCl (pH 7.5), 2 mM EDTA] containing mixture of HaltTM protease and phosphatase inhibitors (1×) (Thermo Fisher Scientific). The brain tissues were individually homogenized and then centrifuged at 13,400× g at 4◦C for 15 min. Protein concentration was determined using the Pierce BCA protein assay kit (Thermo Fisher Scientific). Bovine serum albumin was used as the standard. Proteins (20–30 µg) for each sample were separated using 12% sodium dodecyl sulfate-PAGE and transferred to polyvinylidene fluoride filter membranes (Bio-Rad) by the semi-dry electroblotting method. The membranes were blocked with 5% skim milk and incubated sequentially with the following primary antibodies against either c-Abl (rabbit monoclonal antibody, 1:1000; Santa Cruz), p-p65, p-65, p-IκB, IκB (rabbit monoclonal antibody, 1:1000; Cell Signaling), TNFα (rat anti-mouse monoclonal antibody, 1:500; Millipore) or α-tubulin (mouse monoclonal antibody, 1:2000; Sigma-Aldrich) and horseradish peroxidase-conjugated secondary antibodies (anti-rabbit or mouse IgG antibody; Cell Signaling), followed by chemiluminescence detection (Thermo Fisher Scientific).

### Animals and Maintenance

All experiments were conducted in accordance with approved animal protocols and guidelines established by the Animal Care Committee of Kyungpook National University. All efforts were made to reduce the number of animals and their sufferings. Agematched male C57BL/6 mice (8–10 weeks old) were supplied by Samtako Bio (South Korea). Mice were housed in the groups of three to five per cage under standard condition using a 12 h light/dark cycle (lights on 07:00–19:00) at a constant ambient temperature of 23 ± 2 ◦C. Each individual animal was used for a single experiment.

### Neuroinflammation Model Based on Intraperitoneal LPS Injection

Lipopolysaccharide was administered to evoke neuroinflammation in mice as described previously (Jo et al., 2017). Mice were injected a single dose of vehicle or LPS (5 mg/kg) intraperitoneally. Phosphate buffered saline (PBS) was used as vehicle and administered the same volume. Mice were sacrificed 48 h after injection and brain tissues collected for further analysis.

### The Complete Freund's Adjuvant (CFA)-Induced Chronic Inflammatory Pain Model

Chronic inflammation in mice was induced by a single dose of CFA injection, as described previously (Jha et al., 2015). Briefly, mice were gently anesthetized with 5% of isoflurane for induction and 2% for maintenance. They received CFA (30 µl, 0.5 mg/ml; Sigma-Aldrich) unilaterally in their left hind paws (ipsilateral paws) by intraplantar injections. Mice in the control group received an equal amount of saline in their left hind

paws. Pain behaviors were assessed before and up to 5 days post-CFA administration.

### Measurement of CFA-Induced Paw Edema

Complete Freund's adjuvant-induced paw edema was assessed by the measurement of paw thickness. One experimenter, who was blinded to the treatment conditions, handled and tested all the animals. The dorsoventral thickness of the middle portions of the hind paws were measured using a caliper, as described previously (Jha et al., 2015).

### Streptozotocin (STZ)-Induced Diabetes Model

The mouse model of diabetes was generated as described previously (Rahman et al., 2016). Briefly, type-1 diabetes was induced by an intraperitoneal administration of STZ (150 mg/kg body weight; Sigma-Aldrich), prepared in 0.1 M citrate buffer (pH 4.5). An equal amount of citrate buffer was injected into control animals. Glycaemia level was tested in blood samples collected from the tail vein 3 days post-STZ injection by using an SD CodeFreeTM glucometer (SD Biosensor Inc., Suwonsi, South Korea). Mice with fasting blood glucose levels over 260 mg/dl were considered diabetic and used for further study.

### Behavioral Test for Pain

Before the actual test, mice were allowed to familiarize the experimenter, testing room, and equipment for at least 1 week. Paw withdrawal thresholds (PWTs) in response to mechanical stimulations were measured at different time points following CFA and STZ injection. The mechanical sensitivity was examined using calibrated Von Frey filaments (BiosebTM, Chaville, France), as described previously (Rahman et al., 2016). PWT was calculated from five consecutive withdrawal responses using Dixon's up-down method. Thermal hyperalgesia is defined as a decrease in paw withdrawal latencies (PWL) in response to a noxious thermal stimulus. The thermal sensitivity was tested using the Hargreaves' Plantar Test Analgesy-Meter (Ugo Basile), as previously described (Jha et al., 2015). To obtain the mean PWL values, tests were repeated at least three times and averaged with 5 min intervals between tests to avoid heatinduced sensitization. One experimenter, who was unaware of the experimental conditions, handled and examined all the animals.

### Immunohistochemistry and Histopathology

Mice were deeply anesthetized and then subjected to intracardiac perfusion-fixation through the aorta with 0.1 M PBS followed by 4% paraformaldehyde dissolved in 0.1 M PBS. The tissues were further post-fixed in the same paraformaldehyde overnight. Tissues were washed with 0.1 M PBS and cryoprotected in 30% sucrose in 0.1 M PBS overnight at 4◦C. Tissues were embedded in frozen section compound (FSC 22 Clear; Leica), and a cryostat was used to prepare 20 µm-thick cross-sections for the spinal cord tissues and 30 µm-thick coronal sections for the brain tissues. Tissue sections or fixed cells were then blocked with 1% bovine serum albumin or normal serum in 0.3% Triton X-100 for 60 min at room temperature. For immunofluorescence staining, tissue sections were incubated with the following primary antibodies against c-Abl (rabbit, 1:100; Santa Cruz), Iba-1 (goat, 1:200; Novus Biologicals, Littleton, CO, United States), GFAP (mouse, 1:500; BD Biosciences), inducible nitric oxide synthase (iNOS) antibody (mouse, 1:200 dilution; BD Transduction Laboratories) or NF-κB p65 (rabbit, 1:500) overnight at 4◦C, and then incubated with Cy3- or FITC-conjugated secondary antibodies (1:200; Jackson ImmunoResearch, West Grove, PA, United States). Slides were washed three times with 0.1 M PBS in 0.3% Triton X-100, and mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA, United States) using glass cover-slips, and visualized under a fluorescence microscope (Leica Microsystems, DM2500, Wetzlar, Germany).

### Quantification and Statistical Analysis

Statistical analysis was performed using either a Student's t test or a one/two-way ANOVA with Dunnett's multiple-comparisons test using GraphPad Prism (version 5.01). Differences with p-values <0.05 were considered to be statistically significant. For the immunohistochemical analysis, 5–6 microscopic images were chosen randomly for statistical analysis. For the measurement of either immunofluorescence or western blot band intensities, the area of the whole image or each band was selected, and the mean intensity was measured using the ImageJ software (National Institutes of Health, Bethesda, MD, United States). The background intensity of the band was also measured and deducted from the values obtained.

## RESULTS

### GNF-2 Inhibits LPS-Induced Inflammatory Activation of Glial Cells in Culture

c-Abl is activated by oxidative stress and its activation in neurons increases NF-κB activation leading to neuronal death (Xiao et al., 2011). In this study, we examined whether c-Abl is involved in the process of inflammatory microglial activation. To investigate the role of c-Abl in inflammatory microglial activation, BV-2 immortalized mouse microglial cell line was stimulated with LPS after GNF-2 pre-treatment (**Figure 1A**). GNF-2 significantly inhibited LPS-induced NO (**Figure 1B**) and TNF-α production (**Figure 1D**) in a dose-dependent manner and GNF-2 did not show any apparent cytotoxicity in the microglia (**Figure 1C**). Similarly, exposure of BV2 microglia to LPS significantly increased the expression of IL-1β mRNA, whereas GNF-2 treatment markedly attenuated LPS-induced upregulation of IL-1β (**Figure 1E**). GNF-2 treatment significantly reduced LPS-induced NF-κB activation in BV-2 microglial cells. Notably, LPS-induced NF-κB activation (phosphorylation of IκB and p65) was decreased in microglia following GNF-2 treatment (**Figures 1F–I**). NF-κB activation is associated with nuclear translocation of the p65, a component of the NF-κB complex (Tanaka and Iino, 2016). LPS-induced nuclear translocation

of p65 was measured in BV-2 cells after pretreatment with GNF-2. GNF-2 significantly reduced nuclear p65 expression (**Figures 1J,K**). These findings suggest that GNF-2 attenuates inflammatory activation of microglia induced by LPS through inhibition of c-Abl activity.

### Knockdown of c-Abl Expression Inhibits LPS-Induced Glial Activation

It has been documented that the expression of active c-Abl in adult mouse forebrain neurons induces severe and progressive neurodegeneration in the Cornu Ammonis 1 (CA1) region of the hippocampus and reactive gliosis (Schlatterer et al., 2011). However, the function of c-Abl in microglia upon inflammatory stimulation has not been studied yet. To examine the role of c-Abl expression in microglia following inflammatory stimulation, BV-2 cells were transfected with siRNA for c-Abl knockdown and the inflammatory activation of microglia was examined (**Figures 2A,B**). The knockdown of the c-Abl gene expression (more than 50%) in BV-2 cells significantly attenuated LPS-induced nitric oxide production (**Figure 2C**) as well as the expression of pro-inflammatory TNF-α mRNA (**Figure 2D**).

### Verification of the Anti-inflammatory Effect of GNF-2 in Primary Microglia and Astrocytes

To determine whether the anti-inflammatory effect of GNF-2 is also observed in primary microglia and astrocytes, primary MGCs were treated with various inflammatory stimuli including LPS, TNF-α, or combination of LPS and interferon-gamma (IFN-γ). As in the BV-2 cells, GNF-2 significantly inhibited LPS-induced NO release. In addition, the anti-inflammatory effect of GNF-2 in MGC upon stimulation with recombinant TNF-α protein, an alternative potent inflammatory stimulus, was also assessed. Similarly, GNF-2 significantly inhibited TNFα-induced NO production in primary glial cells (**Figure 3A**). Further, these findings led us to investigate whether GNF-2 can

inhibit LPS and IFN-γ-induced production of pro-inflammatory mediators in primary glial cells or not. We found that GNF-2 pre-treatment dramatically suppressed such upregulation of IL-1β mRNA expression (**Figure 3B**). However, pre-treatment with a methylated GNF-2 analog, mGNF-2 (methylation of the aniline nitrogen at the C4 position of the pyrimidine) (Choi et al., 2009), did not suppress LPS-induced NO production. This is because methylation of GNF-2 abolished the binding specificity of GNF-2 for c-Abl (Choi et al., 2009), suggesting that the effect of GNF-2 is highly specific for c-Abl (**Figure 3C**). Subsequently, the effects of GNF-2 on NF-κB activation after LPS and IFN-γ stimulation was examined in MGCs pretreated with GNF-2. The pre-exposure of MGCs to GNF-2 strongly inhibited LPS and IFN-γ–induced NF-κB activation. This was assessed by western blot analyses of phosphorylatedp65 and -IκB protein (**Figures 3D–F**). Furthermore, knocking down the c-Abl gene by siRNA also significantly inhibited LPS/IFN-γ-induced NO production (**Figure 3G**), TNF-α mRNA expression (**Figures 3H,I**), and NF-κB activation (**Figures 3J–L**) in the primary MGC. Taken together, these findings suggest that GNF-2 may have a potent anti-inflammatory role in gliamediated neuroinflammation.

### Anti-inflammatory Effects of GNF-2 in a Neuroinflammation Model

To investigate the role of c-Abl in neuroinflammation, a mouse model of neuroinflammation induced by the intraperitoneal (IP) administration of LPS was used. First, the expression of c-Abl at the level of the mRNA and protein was examined by using RT-PCR and immunostaining in brain tissues isolated from LPSinjected mice. RT-PCR analysis revealed enhanced expression of c-Abl mRNA after 2 days post-LPS injection (**Figures 4A,B**). Similarly, immunofluorescence analysis showed a substantial upregulation of c-Abl protein in the brain cortex following LPS administration (**Figure 4C**). Co-immunostaining analysis revealed that c-Abl is found to be expressed in Iba-1-positive microglial cells, particularly in hyperactivated microglia with amoeboid shape (indicated by arrows in **Figure 4C**). The effect of GNF-2 on microglial activation was further tested in vivo using a mouse model of LPS-induced neuroinflammation. As shown in **Figure 5A**, GNF-2 treatment in the pre-treatment group started 24 h before LPS injection. It was administered daily for three more days. The mice were then sacrificed and examined for neuroinflammation. To evaluate the anti-inflammatory effect of GNF-2 in vivo, the expression of pro-inflammatory cytokines was measured in brain tissues at the levels of both mRNA and protein. The expression levels of TNF-α and IL-1β mRNA were significantly diminished following GNF-2 pre-treatment (**Figures 5B,G**). Similarly, GNF-2 pre-treatment significantly reduced the expression of TNF-α protein induced by LPS (**Figures 5C,H**). Our in vitro data showed a potent inhibitory effect of GNF-2 on NO production by glial cells upon inflammatory stimulation. To confirm whether GNF-2 can reduce the expression level of iNOS in the mouse brain following LPS administration, immunostaining of brain tissue sections isolated from mice treated with LPS and GNF-2 was performed. The immunostaining analyses revealed an upregulation of iNOS immunoreactivity in the brain cortex after 48 h of LPS injection when compared with vehicleinjected control animals; the immunoreactivity was significantly attenuated in GNF-2-injected mice (**Figures 5D,I**). Further, the levels of Iba-1 and GFAP immunoreactivity were assessed, since they are molecular markers of microglia and astrocyte activation (Rahman et al., 2016). Upon LPS injection, a significant increase in the number of Iba-1-positive microglial cells was observed in the cortex of mice brain, where microglia displayed enhanced Iba-1 immunoreactivity with short and thick processes when compared to control mice (**Figures 5E,J**). These morphological features of the microglia and the increased Iba-1 immunoreactivity in the cortex were attenuated in mice treated with GNF-2. Similarly, the GFAP-positive astrocytes in the cortex of LPS-injected mice showed enhanced immunoreactivity and hypertrophic morphology in comparison to that in the cortex of the vehicle-injected control animals; the immunoreactivity and hypertrophic morphology were both significantly attenuated in GNF-2-treated mice (**Figures 5F,K**). These findings suggest that the intraperitoneal administration of GNF-2 has a potent antiinflammatory effect on LPS-induced neuroinflammation. The potential brain uptake mechanism of small-molecule compounds is based on physiochemical and molecular properties (Mikitsh and Chacko, 2014). In line with these parameters, GNF-2 is a small-molecule compound with comparatively low molecular weight (374.323) and has a high lipophilicity (Adrian et al., 2006),

(A) Nitric oxide (NO) production induced by LPS (1 µg/ml) or TNF-α (10 µg/ml) was significantly reduced by GNF-2 treatment. (B) IL-1β mRNA expression induced by LPS (1 µg/ml) and IFN-γ (50 U/ml) was significantly reduced by GNF-2 treatment. (C) NO production induced by LPS (1 µg/ml) and IFN-γ (50 U/ml) was significantly reduced by GNF-2 treatment but not by mGNF-2, the methylated version of GNF-2. (D) NF-κB activation was reduced by GNF-2 pre-treatment for 1 h. Western blotting was performed to measure phosphorylation of p65 and IκB in MGC after treatment with the indicated reagents. (E,F) Quantification for the relative % change of p-p65 (E) and p-IkB (F) at 24 h after LPS/IFN-γ or GNF-2 treatment. (G–L) MGC were transfected with siRNA for the c-Abl gene and incubated for 2 days. Cells were then treated with LPS and IFN-γ for 24 h. siRNA-mediated knockdown of the c-Abl gene significantly reduced LPS/IFN-γ-induced NO production (G), TNF-α mRNA expression (H). (I) Quantification for the relative % change of TNF-α mRNA following LPS/IFN-γ treatment presented in the adjacent graph. (J) c-Abl expression and phosphorylation of IκB, measured by Western blot analysis. (K) Quantification for the relative protein expression of c-Abl after c-Abl gene knockdown following LPS/IFN-γ treatment. (L) Quantification for the relative protein expression for p-IkB after c-Abl gene knockdown following LPS/IFN-γ treatment. si-C, control siRNA; si-A, siRNA for c-Abl gene knockdown. Alpha-tubulin was used as the protein loading control. Data are presented as mean ± SEM. <sup>∗</sup>p < 0.05 from ANOVA and unpaired two-tailed Student's t test; n = 3 for each group. ns, not significant.

which are important criteria for BBB permeability (Mikitsh and Chacko, 2014). In addition, several studies have demonstrated that mice injected with LPS intraperitoneally show blood-brain barrier (BBB) disruption and increased permeability (Jangula and Murphy, 2013; Banks et al., 2015; Varatharaj and Galea, 2017). Thus, BBB disruption and increased permeability in the LPS model may facilitate GNF-2 transportation from circulation to the central nervous system (CNS), thereby enabling the anti-inflammatory effects of peripherally administered GNF-2 on CNS events.

### GNF-2 Ameliorates Inflammatory Pain Hypersensitivity

Our in vitro and in vivo studies strongly indicated that treatment with GNF-2 suppresses neuroinflammation. Based on these observations, the pharmacological efficacy of GNF-2 was validated in a mouse model of CFA-induced chronic inflammatory pain. This model recapitulates several key inflammatory phenotypes including paw edema and proinflammatory cytokine release in the hind paw and spinal cord tissues, which has been suggested as a mechanistic consequence of peripheral and CNS pathology of CFA-induced pain hypersensitivity (Jha et al., 2014, 2016). To investigate whether intraperitoneally administering GNF-2 in this mouse model could inhibit CFA-induced pain hypersensitivity (**Figure 6A**), a single injection of GNF-2 (10 mg/kg body weight) was administered 30 min before CFA-administration. GNF-2-injected mice showed significantly diminished CFA (10 mg/kg)-induced paw edema formation (**Figure 6B**), the development of thermal hyperalgesia (**Figure 6C**), and mechanical allodynia (**Figure 6D**). However, the analgesic effect of 1 mg/kg GNF-2

was partial (**Figures 6C,D**). The withdrawal latency in response to thermal stimuli and the withdrawal threshold in response to mechanical stimuli were unchanged in the contralateral hind paws following CFA and GNF-2 treatment (data not shown). These results demonstrate the crucial role of c-Abl in chronic inflammatory pain.

### GNF-2 Attenuates Diabetes-Induced Neuroinflammation and Pain Hypersensitivity

To investigate the role of GNF-2 in neuroinflammation and pain hypersensitivity, an STZ-induced painful diabetic mouse model was also used. The mRNA and protein levels of c-Abl were first examined in spinal cord tissues isolated from mice 2 weeks after STZ injection using RT-PCR and immunostaining (**Figure 7A**). RT-PCR analysis revealed that the induction of diabetes significantly increases the expression of c-Abl mRNA in the spinal cord (**Figures 7B,C**). Similarly, immunofluorescence analysis showed a substantial upregulation of the c-Abl protein in the dorsal horn of the lumbar segment of the spinal cord (**Figure 7D**). In addition, c-Abl was expressed in GFAP-positive astrocytes, particularly those located in the lamina I region of the diabetic spinal cord. Subsequently, the expression levels of pro-inflammatory cytokines and glial activation were examined in the spinal cord of diabetic mice. The expression of TNF-α and IL-1β mRNAs in the lumbar segment of the spinal cord of mice with diabetes was significantly increased 2 weeks after STZ injection (**Figures 8A,B**). The intraperitoneal administration of GNF-2 significantly decreased the diabetes-induced increase in the expression of pro-inflammatory cytokines such as TNFα and IL-1β mRNAs in the spinal cord tissues (**Figure 8B**). To evaluate the effects of GNF-2 on neuroinflammation, we also assessed diabetes-induced changes in glial activation and proliferation in the spinal cord using immunostaining. In diabetic mice, there was a significant increase in the number of Iba-1-positive microglial cells in the dorsal horn of the spinal cord tissues isolated from the lumbar segment, where the microglia displayed enhanced Iba-1 immunoreactivity with reactive morphological changes (**Figure 8C**). Similarly, the number of GFAP-positive astrocytes was markedly increased in the spinal cord dorsal horn of STZ-induced diabetic mice. This increase was accompanied by an increase in GFAP immunoreactivity and hypertrophic morphology with

attenuated this increase in IR. The nuclei were stained with DAPI. (G) Quantification for the TNF-α and IL-1β mRNA expression is displayed as the fold change of gene expression normalized to GAPDH. (H) Quantification for the TNF-α protein from the western blot. (I–K) Quantification for the relative intensities of iNOS, Iba-1, and GFAP IR is presented in the graph. <sup>∗</sup>p < 0.05 vs. vehicle-treated control animals; #p < 0.05 between the indicated groups; unpaired two-tailed Student's t test; n = 3 for each group; data are presented as mean ± SEM. Scale bar 400 µm (D), 200 µm (E,F).

thick processes (**Figure 8D**). Notably, GNF-2 administration significantly downregulated Iba-1 and GFAP immunoreactivity in the spinal cord of STZ-injected mice (**Figures 8E,F**). These results obtained through the pharmacological inhibition of c-Abl demonstrate that c-Abl plays a crucial role in diabetesassociated neuroinflammation.

Furthermore, the contribution of c-Abl activation to the pathogenesis of diabetic pain was assessed through the pharmacological inhibition of c-Abl. A single intraperitoneal injection of mice with GNF-2 (10 mg/kg) significantly attenuated diabetes-induced thermal hyperalgesia (**Figure 9A**) as well as mechanical allodynia (**Figure 9B**). However, the vehicle alone did not alter the withdrawal latency in response to thermal stimuli or withdrawal threshold in response to mechanical stimuli. These findings suggest that c-Abl plays a critical role in diabetes-induced neuroinflammation and associated pain hypersensitivity. Thus, GNF-2 might be a potent therapeutic agent for the treatment of chronic pain which results from diabetic peripheral neuropathy.

### DISCUSSION

In this study, we show that the upregulation of c-Abl expression promotes the classical pro-inflammatory activation of microglia and that GNF-2, a potent c-Abl inhibitor, attenuates neuroinflammation and pain hypersensitivities in CFA- and STZ- induced pain models. Our findings demonstrate that c-Abl contributes to the pathogenesis of chronic pain by regulating microglial activation and neuroinflammation.

Chronic pain is caused by nerve damage which occurs during nerve compression, diabetes, inflammation, and shingles virus infection (Campbell and Meyer, 2006). Specifically, cytokines, chemokines, prostaglandins, and NO released from activated microglia and astrocytes in the dorsal horn of the spinal cord are known to play important roles in the pathogenesis of chronic pain (Skaper et al., 2012; Vega-Avelaira et al., 2013). Therefore, several studies have targeted activated microglia in order to reduce pain hypersensitivity (Hsieh et al., 2018). For example, treatment with minocycline, a microglial inhibitor,

FIGURE 6 | Effects of GNF-2 on CFA-induced paw edema and inflammatory pain behaviors. (A) To investigate the role of c-Abl in CFA-induced inflammatory pain phenotypes, GNF-2 were administered into hind paw 30 min before the intraplantar injection of CFA at the indicated time points as shown in the experimental outline. Paw edema and pain responses were measured at 6 h, 1, 3, and 5 d post-CFA injection. In the ipsilateral sides, CFA injection increased paw thickness (B) and reduced PWL to heat (C) as well as PWT to force (D) when compared with vehicle-injected control animals. The CFA-induced paw edema and pain hypersensitivity were attenuated in the GNF-2-injected mice (1 and 10 mg/kg) in a dose-dependent manner. No significant change in paw edema or pain-related behavior was observed in the contralateral sides and vehicle-injected animals. <sup>∗</sup>p < 0.05 vs. vehicle + CFA injected animals Student's t test; n = 3 for each group; data are represented as mean ± SEM. d, day (s).

prevents pain hypersensitivity caused by systemic LPS exposure in neonates, and its protective effect may be related to its ability to attenuate LPS-induced microglial activation, proinflammatory cytokine IL-1β, and pain mediator PGE2. In addition, microglial activation is pivotal to the development and maintenance of allodynia after spinal cord injury through mechanisms involving both TNF-α and IL-1β and in chronic states, IL-6 (Detloff et al., 2008). Previous studies have shown that inflammation and oxidative stress are associated with the overexpression and activation of c-Abl (Schlatterer et al., 2011; Lawana et al., 2017). In fact, in our study, an increase in c-Abl expression was observed in the inflammatory pain models and glial activation and pain hypersensitivity were reduced by the c-Abl inhibitor.

The activation of c-Abl has been reported to play a role in neurodegenerative diseases. For example, in AD,

beta-amyloid (Aβ) activates c-Abl in hippocampal neurons (Alvarez et al., 2004) and c-Abl levels are increased in pre-tangle neurons in AD (Derkinderen et al., 2005). The inhibition of c-Abl activity by Imatinib (STI-571) protects hippocampal neurons from Aβ-induced apoptosis and the reduction of c-Abl mRNA levels protects neuronal cells from Aβ-induced toxicity (Alvarez et al., 2004). Recently, it has been reported that the tyrosine phosphorylation of parkin by the oxidative stress-induced c-Abl is part of a regulatory mechanism in parkin function (Imam et al., 2011). Tyrosine phosphorylation of parkin results in impaired E3-ubiquitin ligase activity and auto-ubiquitination of parkin. c-Abl activation also plays a key role in α-synuclein neurodegeneration. c-Abl overexpression in mice leads to dopaminergic neuron degeneration and α-synuclein pathologies, while c-Abl deletion reduces α-synuclein pathologies. Activation of c-Abl leads to tyrosine 39 phosphorylation of α-synuclein, which strongly correlates with disease progression in hA53Tα-syn transgenic mice, suggesting a critical role of c-Abl in neuronal function and survival (Brahmachari et al., 2016).

The use of c-Abl inhibitors such as imatinib and nilotinib have been proposed for the treatment of AD and PD. Indeed, there have been a few reports which indicated that c-Abl inhibition might be beneficial in PD and α-synucleinopathies (Ko et al., 2010; Hebron et al., 2013). The c-Abl inhibitor, STI-571 (Imatinib), restores the E3 ligase activity of parkin and reduces the accumulation of parkin substrates, thereby protecting against 1-methyl-4-phenylpyridinium (MPP+)-induced neurotoxicity in vitro (Ko et al., 2010; Imam et al., 2011). However, in the aforementioned studies, it was difficult to conclude whether c-Abl inhibition could be an effective neuroprotective strategy because of the lack of selectivity of the c-Abl inhibitors they used. Imatinib and nilotinib are potent inhibitors of tyrosine kinases which include c-Abl, Src families, c-Kit, and PDGFR. However, GNF-2, a third-generation c-Abl inhibitor, had no detectable inhibitory effect on the Src family kinases Hck, Lyn, Lck, and c-Src (Choi et al., 2009). Therefore, in the current study, we used GNF-2, which is a more specific c-Abl inhibitor, and siRNA to knockdown c-Abl gene expression in order to investigate the role of c-Abl in neuroinflammation and related pathology.

### REFERENCES


### CONCLUSION

In conclusion, GNF-2 significantly inhibits NF-κB activation and LPS-induced pro-inflammatory molecules including TNF-α and NO in microglia and in vivo models of chronic inflammatory and neuropathic pain. Furthermore, we show that GNF-2 very efficiently prevents inflammatory and diabetic pain in animal models. Our results buttress the role of c-Abl in the pathogenesis of neuroinflammatory diseases. These findings indicate that c-Abl can be therapeutically targeted for both the prevention and reversal of chronic pathological pain.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Animal Care Committee of Kyungpook National University.

### AUTHOR CONTRIBUTIONS

KS and GS: conceptualization. GS, MR, MJ, DG, SP, and J-HK: data acquisition and methodology. S-HL, TS, I-KL, YB, W-HL, GS, and KS: investigation. KS: project administration, resources, and supervision. GS and KS: validation. All authors wrote and edited the manuscript.

### FUNDING

This work was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry of Health & Welfare, South Korea (HI16C1501) and the Basic Science Research Program through the National Research Foundation (NRF), which is funded by the Korean government (MSIT) (2018R1A2A1A05077118, 2016M3C7A1904148, and NRF-2017R1A5A2015391). GS was supported by the NRF grant (Grant No. 2016R1D1A1B01009186). TS was supported by the KU-KIST Graduate School of Converging Science and Technology Program.

alpha-synuclein-induced neurodegeneration. J. Clin. Invest. 126, 2970–2988. doi: 10.1172/JCI85456



**Conflict of Interest Statement:** S-HL was employed by company VORONOI Inc.

The remaining 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 © 2019 Song, Rahman, Jha, Gupta, Park, Kim, Lee, Lee, Sim, Bae, Lee and Suk. 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.

# Protective Effects of *Phyllanthus amarus* Against Lipopolysaccharide-Induced Neuroinflammation and Cognitive Impairment in Rats

*Akilandeshwari Alagan1, Ibrahim Jantan2, Endang Kumolosasi1, Satoshi Ogawa3, Maizaton Atmadini Abdullah4 and Norazrina Azmi1\**

*1 Drug and Herbal Research Centre, Faculty of Pharmacy, Universiti Kebangsaan Malaysia, Kuala Lumpur, Malaysia, 2 School of Pharmacy-SRI, Faculty of Health & Medical Sciences, Taylor's University, Subang Jaya, Malaysia, 3 Brain Research Institute, Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia, Bandar Sunway, Malaysia, 4 Department of Pathology, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Serdang, Malaysia*

*Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by: Yashar Yousefzadeh Fard,*

*Stony Brook Medicine, United States Denis Melo Soares, Federal University of Bahia, Brazil*

*\*Correspondence: Norazrina Azmi azrina.azmi@ukm.edu.my*

#### *Specialty section:*

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

*Recieved: 17 January 2019 Accepted: 17 May 2019 Published: 04 June 2019*

#### *Citation:*

*Alagan A, Jantan I, Kumolosasi E, Ogawa S, Abdullah MA and Azmi N (2019) Protective Effects of Phyllanthus amarus Against Lipopolysaccharide-Induced Neuroinflammation and Cognitive Impairment in Rats. Front. Pharmacol. 10:632. doi: 10.3389/fphar.2019.00632*

Background: *Phyllanthus amarus* (PA) is widely studied for its hepatoprotective properties but has recently received increasing attention due to its diverse anti-inflammatory effects. However, the effects of PA in modulating immune responses in the central nervous system leading to protection against functional changes remain unexplored. Therefore, we sought to examine the protective effects of 80% v/v ethanol extract of PA on lipopolysaccharide (LPS)-induced non-spatial memory impairment and neuroinflammation.

Methods: Selected major phytoconstituents of PA extract were identified and quantified using high-performance liquid chromatography. Subchronic neurotoxicity was performed in male Wistar rats given daily oral administration of 100, 200, and 400 mg/kg of the PA extract. Their neurobehavioral activities (functional observation battery and locomotor activity) were scored, and the extracted brains were examined for neuropathological changes. Rats were treated orally with vehicle (5% Tween 20), PA extract (100, 200, and 400 mg/kg), or ibuprofen (IBF; 40 mg/kg) for 14 and 28 days before being subjected to novel object discrimination test. All groups were challenged with LPS (1 mg/kg) given intraperitoneally a day prior to the behavioral tests except for the negative control group. At the end of the behavioral tests, the levels of tumor necrosis factor-α (TNF-α), interleukin (IL)-1β, nitric oxide (NO), inducible nitric oxide synthase (iNOS), CD11b/c integrin expression, and synaptophysin immunoreactivity were determined in the brain tissues.

Results: Gallic acid, ellagic acid, corilagin, geraniin, niranthin, phyllanthin, hypophyllanthin, phyltetralin, and isonirtetralin were identified in the PA extract. Subchronic administration of PA extract (100, 200, and 400 mg/kg) showed no abnormalities in neurobehavior and brain histology. PA extract administered at 200 and 400 mg/kg for 14 and 28 days effectively protected the rodents from LPS-induced memory impairment. Similar doses significantly (*p* < 0.05) decreased the release of proteins like TNF-α, IL-1β, and iNOS in the brain tissue. NO levels, CD11b/c integrin expression, and synaptophysin immunoreactivity were also reduced as compared with those in the LPS-challenged group.

Conclusion: Pre-treatment with PA extract for 14 and 28 days was comparable with pre-treatment with IBF in prevention of memory impairment and alleviation of neuroinflammatory responses induced by LPS. Further studies are essential to identify the bioactive phytochemicals and the precise underlying mechanisms.

Keywords: *Phyllanthus amarus*, neuroinflammation, neuroprotection, non-spatial memory, pro-inflammatory markers

### INTRODUCTION

*Phyllanthus amarus* Schumah & Thonn. (PA) belongs to the Euphorbiaceae family and is traditionally used for kidney ailments, diabetes, pain, jaundice, gonorrhea, chronic dysentery, skin ulcer, and hepatitis B. Recently, the plant has received increasing attention and has been studied for various pharmacological properties such as immunomodulatory, antinociceptive, antiinflammatory, antioxidant, antibacterial, anticancer, antiulcer, gastroprotective, antifungal, antiplasmodic, antiviral, aphrodisiac, contraceptive, hepatoprotective, antihyperglycemic, antilipidemic, nephroprotective, and anti-amnesic activities (Joshi and Parle, 2007; Patel et al., 2011). Although it demonstrates a wide spectrum of pharmacological actions, the unifying features of all these actions are directed towards the anti-inflammatory and antioxidant properties of the plant. PA contains various phytoconstituents such as lignans, alkaloids, phenolics, terpenes, tannins, flavonoids, sterols, and volatile oils (Patel et al., 2011). Of all these phytochemicals, phyllanthin, hypophyllanthin, corilagin, and geraniin are found in abundance and potentially responsible for the reported anti-inflammatory actions of PA (Patel et al., 2011; Jantan et al., 2014). Most of the anti-inflammatory studies were performed in models of inflammation either *in vitro* or *in vivo*. However, limited available data to substantiate the effects of PA in neuroinflammation have warranted a study to explore the antiinflammatory actions of PA in the central nervous system (CNS).

Neuroinflammation has been implicated in the pathogenesis of neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD), and multiple sclerosis (MS) and of neuropsychiatric disorders such as major depressive disorder, schizophrenia, and bipolar disorder (Radtke et al., 2017). Unlike other systemic immune responses, microglia are the resident macrophage in the brain. Microglial activation forms the basis of many immune-mediated responses of the CNS, which enables it to cope with pathogens, toxins, trauma, and degeneration. However, it is the hyperactivation of microglial cells that initiates excessive production of inflammatory mediators leading to destructive inflammatory responses in the brain (Kempuraj et al., 2016). Additionally, the brain is also exposed to constitutive defense responses, such as systemic inflammation. The systemic

inflammation may also lead to the initiation of circulating cytokines, thereby impacting CNS and triggering neuroinflammation. Over the decades, it was evident that the immune system plays a central role in learning, memory modulation, and neural plasticity. Under normal quiescent conditions, immune mechanisms positively regulate the neural circuits remodeling, promotion of memory functions, and neurogenesis (Thomson and Sutherland, 2005). In conditions under which the immune system is strongly activated by infection, injury, or severe or chronic stressful conditions, the inflammatory mediators disrupt the delicate balance needed for neurophysiological actions and produces detrimental effects on memory, neural plasticity, and neurogenesis (Okun et al., 2012).

Bacterial infections initiate innate immune responses including activation of toll-like receptor 4 (TLR4) and the transcription factor nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) in microglia and macrophages, which subsequently provoke the expression of cytokines and generation of nitric oxide (NO) (Parajuli et al., 2012). These pathways play a basic role in the degradation of bacteria by immune cells, which unfavorably influence neuronal death. Lipopolysaccharide (LPS) is a cell wall component of gram-negative bacteria and a ligand of TLR4 that initiates immune responses to infections. As compared with other glial cells, microglial cells are markedly responsive to LPS leading to learning and memory impairment. Intraperitoneal injection of LPS induces secretion of pro-inflammatory cytokines resulting in neuroinflammation, hippocampal apoptosis, cognitive impairment, learning deficits, and even beta-amyloid plaques generation in the hippocampus (Rock et al., 2004; Zarifkar et al., 2010).

Previous studies in our laboratory identified the presence of phyllanthin, hypophyllanthin, gallic acid, geraniin, corilagin, ellagic acid, and niranthin in 80% ethanol extract of PA (Jantan et al., 2014). It was also found that *P. amarus* at doses of 100 to 500 mg/kg for 14 days revealed non-toxic effect with no abnormalities in general behavior and physiology of rats (Ilangkovan et al., 2015). Additionally, a single or daily repeated doses administration of PA for 28 days revealed no morphological changes in histopathological observation of the kidney, liver, and pancreas (Lawson-Evi et al., 2008; Kushwaha et al., 2013). Lack of study for assessment of neurotoxicity of PA has led us to examine the effects of this plant extract on neurobehavior and brain histopathological changes in rats.

Although the anti-inflammatory activities of PA have been documented (Ilangkovan et al., 2015; Harikrishnan et al., 2018), there is a lack of evidence to substantiate similar effects in the CNS. Treatment with PA extract and phyllanthin was found to improve memory impairment and exhibited anticholinesterase activity in young and older mice (Joshi and Parle, 2006; Joshi and Parle, 2007). These are important early findings that demonstrated the plant activity in the brain suggestive of its potential value in

**Abbreviations:** PA, *Phyllanthus amarus*; LPS, lipopolysaccharide; IBF, ibuprofen; i.p., intraperitoneal injection; CNS, central nervous system; AD, Alzheimer's disease; PD, Parkinson's disease; MS, multiple sclerosis; TLR4, tolllike receptor 4; NF-κB, nuclear transcription factor; HPLC, high-performance liquid chromatography; LOD, limit of detection; LOQ, limit of quantification; NOD, novel object discrimination test, TNF-α, tumor necrosis factor-α; IL-1β, interleukin-1β; NO, nitric oxide; iNOS, inducible nitric oxide synthase; CD11b/c, cluster differentiation 11b/c.

the prevention and treatment of neurodegenerative diseases. Similarly, other *Phyllanthus* species such as *Phyllanthus niruri* (Ambali et al., 2012) and *Phyllanthus emblica* (Ashwlayan and Singh, 2011) have also been reported to reverse memory deficits induced by scopolamine, sodium nitrite, or chlorpyrifos in different animal models of cognitive behavior, which further support a notion of their neuroprotective role. Therefore, the present study sought to examine the neuroprotective effects of PA extract as compared with IBF, a widely studied nonsteroidal anti-inflammatory drug, for its neuroprotective effects against LPS-induced memory impairment and inflammation in rodents.

### MATERIALS AND METHODS

### Animals

Adult male Wistar rats weighing 190–200 g (5 weeks old) were obtained from the Laboratory Animal Resource Unit (LARU), Universiti Kebangsaan Malaysia (UKM), Malaysia. The rats were housed in a temperature-controlled room (22–25°C) and exposed to 12 h dark/light cycles. Experiments were carried out on the basis of procedures approved by UKM Animal Ethics Committee. Animals were allowed to acclimatize for 7 days before the initiation of treatment. The animal laboratory was maintained under standard conditions. The studies were performed according to procedures for the use of animals in research as approved by the UKM Animal Ethics Committee with the approval number FF/2017/NORAZRINA/ 24-MAY/850-JUNE-2017-JULY-2018 for the toxicity assessment in rats and FF/2015/NORAZRINA/20-MAY/683-MAY-2015- MAY-2016 for the efficacy and molecular study done in rats.

### Chemicals

LPS from *Escherichia coli* (055:B5) and ibuprofen (IBF; >98% purity) were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Radioimmunoprecipitation assay (RIPA) buffer, enzyme-linked immunosorbent assay (ELISA) kits for tumor necrosis factor-α (TNF-α), and interleukin (IL)-1β were obtained from R&D Technology (Minneapolis, MN, USA). Inducible nitric oxide synthase (iNOS), anti-CD11b/c, and synaptophysin were from Abcam (Cambridge, MA, USA). β-Actin was from Cell Signaling Technology (Beverly, MA, USA). Methanol and acetonitrile [high-performance liquid chromatography (HPLC) grade] were brought from Fisher Scientific (Loughborough, UK). Phyllanthin, hypophyllanthin, gallic acid, geraniin, corilagin, ellagic acid, niranthin, phyltetralin, and isonirtetralin were purchased from ChromaDex (CA, USA) with purity > 98%.

### Preparation of Extract

The whole plant of PA Schumah & Thonn. was obtained from Marang, Kuala Terengganu, Malaysia, in the month of February 2016. Dr. Abdul Latif Mohamad from the Faculty of Science and Technology, Universiti Kebangsaan Malaysia (UKM), identified the plant, and a voucher specimen (*P. amarus*  UKMB 30078) was deposited at the Herbarium of UKM, Bangi, Malaysia. After collection, the plant was allowed to dry for a week and then ground to form the coarse powder. A total of 1 kg of the coarse powder was soaked in 80% ethanol for 9 days. The extract was filtered, and the solvent was changed every 72 h. The solvent was removed using a rotary evaporator, and the extract was subjected to freeze-drying and stored in −20°C.

### High-Performance Liquid Chromatography Analysis of 80% Ethanol Extract of *Phyllanthus amarus*

Qualitative and quantitative HPLC analysis was performed according to the method of Jantan and colleagues, with a slight modification (Jantan et al., 2014). Twenty milligrams of PA extract and 1 mg of reference standards (gallic acid, ellagic acid, corilagin, geraniin, niranthin, phyllanthin, hypophyllanthin, phyltetralin, and isonirtetralin) each in 1 mL of methanol were ultra-sonicated for 10 min. Then they were filtered through 0.45-µm Millipore Millex polytetrafluoroethylene membrane (Maidstone, Kent, UK). The diluted extract and the reference standards were analyzed using the following conditions: reversed-phase column, C-18 (100 mm × 4.6 mm inner diameter, 5 µm, XBridge, Waters, Ireland); detector, photodiode array (Waters 2998); wavelength, 220 nm; and flow rate, 1.0 mL/min. Phyllanthin and hypophyllanthin were eluted isocratically (mobile phase A, acetonitrile; B, acidified water with 0.1% orthophosphoric acid) with 5% B, increased to 95% over 20 min, and then followed by 95% for 15 min. In addition to that, gallic acid, ellagic acid, corilagin, geraniin, niranthin, phyltetralin, and isonirtetralin were eluted by gradient method (mobile phase A, acetonitrile; B, acidified water with 0.2% orthophosphoric acid) with 5–70% A (0 to 15 min) and 70% to 95% A (15 to 30 min) and then followed by 95% A for 20 min. Compound identification was done by comparing the retention time and spectra peaks with the reference standards. A calibration curve was plotted with three different concentrations (250, 500, and 1,000 µg/mL) for each standard versus the area under the peaks. The standard curve equations obtained from each bioactive compound were used to quantify the compounds in the extract (Jantan et al., 2014; Yuandani et al., 2016).

### High-Performance Liquid Chromatography Validation

Validation was performed by determining the linearity, precision, limit of detection (LOD), and limit of quantification (LOQ). Linear calibration analysis was evaluated to determine the linearity by calculating the correlation coefficient (*r*<sup>2</sup> ) from the calibration curve. A calibration curve was plotted with three different concentrations (250, 500, and 1,000 µg/mL) for each standard. The method precision was determined by intra-day and inter-day assays, and they are performed by injecting the extract (20 mg/mL) and standards (1 mg/mL) separately three times in a day and on three different days. The mean values of the area under the peak and retention time obtained from the intra-day and inter-day assay were used to confirm the reproducibility of the results. LOD and LOQ were calculated using the formula LOD = 3.3 × (σ/*S*) and LOQ = 10 × (σ/*S*), where σ = standard deviation of the response and *S* = slope.

### Subchronic Neurotoxicity Study

Toxicity tests were conducted according to Organisation for Economic Co-operation and Development 423 guidelines with modified procedure (Organisation for Economic Co-operation and Development, 2002). Twenty adult male Wistar rats were divided into four groups randomly. Group 1 was considered as vehicle control with 5% Tween 20; groups 2, 3, and 4 were administered orally with 100, 200, and 400 mg/kg of PA extract for 28 days. The general behaviors of rats were observed by functional observational battery (FOB) test and followed by histopathological evaluation.

### Functional Observational Battery

Neurobehavioral changes were evaluated by FOB, a widely used screening method to identify potential neurotoxicity of new and existing chemicals. The procedure was based on the McDaniel and Moser (1993) protocol with slight modifications. Home-cage movement and hand-held observations were performed in the FOB. The number of line crossings, rearing against a wall, center square entry, grooming, defecation, and urination were observed in the open field activity test for 28 days (McDaniel and Moser, 1993; Brown et al., 1999).

### Histopathological Evaluation

After behavioral analysis, rats were sacrificed, and their brains were dissected and weighed. They were immersed in 10% neutral buffered formalin and stored until further analysis. Tissues were processed, embedded in paraffin, and sectioned at 5-μm thickness. The slides were stained with hematoxylin and eosin (H&E) and viewed under a light microscope for semiquantitative histological evaluation (Clausen et al., 2009) and morphological observation of the neurons and pyramidal cells. Responses to insult were evaluated as edema of the brain parenchyma and infiltration of the cortical lesion and hippocampus ipsilaterally by macrophages, neutrophils, and lymphocytes as the signs of inflammation. Evaluation of morphological and pathological changes was performed by an investigator (MAM) blinded to the treatment status of each animal.

### Experimental Design for 14 and 28 Days of *Phyllanthus amarus* Extract Treatment

The animals were divided into six groups of eight and received daily oral administration of 5% Tween 20 (negative and vehicle control groups), IBF 40 mg/kg (positive control), and PA extract at different doses (100, 200, and 400 mg/kg). The treatment was scheduled for 14 and 28 days. A day prior to the cognitive behavioral tests, all groups were challenged with 1 mg/kg of LPS given intraperitoneally except the negative control group, which received an intraperitoneal injection of the vehicle. The animals were decapitated at the end of the experiment, and their brains were collected for further analysis.

## Cognitive Behavioral Studies

### Novel Object Discrimination Task

Novel object discrimination test (NOD) is utilized to assess the non-spatial memory in rodents. This test examines the time spent in exploring the novel object and a familiar object. The method was adopted from previous studies with minor modifications (Ennaceur, 2010; Azmi et al., 2011). NOD was performed in an open box (width × length × height = 40 cm × 40 cm × 40 cm) made of Perspex. The lighting setup was fixed above the ceiling inside the behavior room with constant illumination. The object was a glass bottle made with a distinct color. The floor was divided into nine equal squares by black lines for scoring of locomotor activity. The procedure consisted of three phases: habituation, familiarization, and test phases. Each rat was pre-habituated for 1 h per day before the experiment. On the experiment day, animals were allowed to explore the empty arena for 3 min individually, followed by the familiarization phase, where an animal was allowed to explore two identical objects (A1 and A2) for another 3 min. During the test phase, animals were allowed to explore one identical object that has been explored before (A3) and one novel object (B) inside the arena for 3 min. In between phases, the open field and objects were cleaned with 70% v/v ethanol to eliminate olfactory cues while the rat was returned to its home cage. The location and combination of objects used were alternated between rats to decrease bias. Scoring of exploration was carried out when the rat nose was directed towards the object at a distance of ≤2 cm or touching the object. Turning around, sitting, or climbing on the object was not scored as an exploratory behavior. Their exploratory behavior was scored during a playback of the closedcircuit television (CCTV) footage. The non-spatial memory was determined using the values of discrimination index (DI). DI was calculated for every rat using the formula DI = (B − A3)/(B + A3).

### Locomotor Activity

The locomotor activity was measured for each rat by scoring the number of line crossings inside the arena during familiarization and test phases. The activity was scored as the frequency of line crossings with all four paws across the line during a playback of the CCTV footage.

### Brain Sample Collection

After the behavioral study, animals were decapitated and their brains dissected into two hemispheres to be stored in −80°C until further analysis. The left and right hemispheres were collected alternately from each group and homogenized in an equal amount of RIPA (R&D Technology, Minneapolis, MN, USA) buffer with protease inhibitor followed by centrifugation at 12,000 rpm for 20 min at 4°C. The supernatant was collected and molecular analysis performed.

### Measurement of Cytokines

The concentrations of pro-inflammatory cytokines TNF-α and IL-1β were estimated using ELISA kits (R&D Systems, Minneapolis, MN, USA) as per manufacturer protocol. Cytokine levels were expressed as percentage of inhibition (Achoui et al., 2010; Harikrishnan et al., 2018). The percentage inhibition was calculated as

% [( ) ( ) *Inhibition cytokine control sa* = × <sup>−</sup> <sup>100</sup> *cytokine mple control* ] ( ) *cytokine* 

### Griess Assay

Nitrite levels were measured in the brain tissue using the Griess assay. The Griess reagent consisted of 0.1% *N*-(1-naphthyl) ethylenediamine dihydrochloride, 1% sulfanilamide, and 2.5% phosphoric acid. Equal volume of Griess reagent and brain sample was mixed and incubated at 37°C for 10 min. The absorbance was read at 542-nm wavelength spectrophotometrically in ELISA plate reader. Nitrite levels were expressed as percentage of inhibition (Green et al., 1982; Achoui et al., 2010). The percentage inhibition was calculated as

$$\%Inabilities = 100 \times \left\{ \frac{[\text{(NO}\_2\text{)}\_{\text{control}} - (\text{NO}\_2\text{)}\_{\text{sample}}]}{\text{(NO}\_2\text{)}\_{\text{control}}} \right\}$$

### Western Blot

Total protein concentrations in the left and right hemispheres were quantified using NanoQuant. Equal amount of protein (60 µg) was separated by 8% and 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membrane. Blots were blocked for 2 h at room temperature with 1% bovine serum albumin (BSA) in tris-buffered saline with Tween 20 (TBST) and washed. The blots were incubated with primary antibody iNOS (1:1,000, Abcam, Cambridge, MA, USA) and synaptophysin (1:500, Abcam, Cambridge, MA, USA) overnight. After being washed, blots were incubated with horseradish peroxidase-conjugated secondary antibody for 2 h. The membrane was then developed for visualizing the protein band in gel doc image analyzer using enhanced chemiluminescence kit (R&D Systems) (Zhu et al., 2016; Harikrishnan et al., 2018).

### Immunohistochemistry of Brain Slices

After the behavioral study, two rats from each group were anesthetized with a combination of ketamine, xylazine, tiletamine, and zolazepam. Transcardiac perfusion with normal saline (0.9%) and buffered 4% paraformaldehyde was performed to fix the brain. The whole brain was collected and cryoprotected with 30% sucrose. The hippocampus area was sectioned coronally in the cryostat (30 µm in thickness) and stored in anti-freezing solution. Then the sections were washed twice with 0.1M phosphate-buffered saline (PBS). The sections were blocked with 3% BSA, 0.75% Triton X, and 0.1M PBS for 1 h in 60 rpm at room temperature. They were incubated overnight with mouse monoclonal anti-CD11b/c (1:500, Abcam, Cambridge, MA, USA). The next day, the sections were washed twice and kept for 1-h incubation with Alexa Fluor 488 secondary antibody. Finally, the sections were mounted on slides after being washed and viewed under a confocal microscope (Nikon Microscope ECLIPSE TE 2000-E) with a 20× objective. The immunostained area in the hippocampus for CD11b/c was measured using ImageJ software referred to a particular area (4 × 104 μm2 ) (Hernangomez et al., 2016; Shen et al., 2016).

### Statistical Analysis

Data obtained were analyzed by one-way analysis of variance (ANOVA) followed by Dunnett's *post hoc* test using GraphPad Prism 5. All experimental data were expressed as mean ± standard error of mean (SEM), where "*n*" denotes the number of rats involved in the test. *p* < 0.05 value was set as statistically significant.

### RESULTS

### Qualitative and Quantitative Analysis of 80% Ethanolic Extract of *Phyllanthus amarus*

Identification of major phytochemicals in 80% ethanol extract of PA was carried out using HPLC. Separation of phyllanthin and hypophyllanthin was identified at 25.423 and 25.617 min, respectively (**Figure 1A**). Gallic acid, geraniin, corilagin, ellagic acid, niranthin, phyltetralin, and isonirtetralin appeared at 11.156, 14.204, 15.273, 16.283, 23.933, 32.157, and 33.628 min, respectively (**Figure 1B**). Compound identification was accomplished by comparing retention time with respective standard compounds (**Table 1**). The LOD and LOQ of the identified compounds are listed in **Table 1**. Among the identified compounds in PA extract, ellagic acid (218.06 µg/mL) content was found to be the highest followed by phyllanthin (162.69 µg/mL), corilagin (158.68 µg/mL), phyltetralin (145.83 µg/mL), niranthin (102.97 µg/mL), hypophyllanthin (95.37 µg/mL), isonirtetralin (71.80 µg/mL), geraniin (67.47 µg/mL), and gallic acid (18.32 µg/mL).

### Functional Observational Battery

Rats treated with PA extract for 28 days did not show any abnormal behavioral changes while handling and home-cage movement. No significant changes were observed in body weight, and there was no mortality rate. In locomotor activity, PA extract administered rats did not show any significant changes in the frequency of line crossing, rearing, central square entry, defecation, and urination (**Supplementary Figure S1**).

### Histopathological Evaluation of *Phyllanthus amarus*-Treated Brain Tissues

The cerebral cortex and hippocampus were sectioned and stained to identify whether PA extract has shown any morphological changes in rats. The vehicle control group and PA extracttreated group at three different doses did not show any edema or inflammatory changes in the cerebral cortex and hippocampus. In addition, no neuronal morphological changes were observed in the neurons and pyramidal cells (**Supplementary Figure S2**).

### Effects of *Phyllanthus amarus* Extract on Cognitive Behavioral Studies

The DI was calculated as mentioned earlier and shown in **Figure 2**. **Figure 2A**,**B** reveals that the group that received LPS alone showed negative DI values, indicating non-spatial memory impairment where the rats spent more time exploring the familiar object than the novel object. Pre-treatment of PA extract (100, 200, and 400 mg/kg) for 14 days demonstrated positive DI,

(16.283 min), niranthin (23.933 min), phyltetralin (32.157 min), and isonirtetralin (33.628 min) in the extract.

indicating that the rats were able to discriminate the novel from familiar objects significantly (\**p* < 0.05, \*\**p* < 0.01). However, after 28 days of treatment, only groups treated with 200 (\**p* < 0.05) and 400 mg/kg of PA extract showed a positive DI. Taken together, it was evident that PA extract at 200 and 400 mg/kg protected against LPS-induced memory impairment. Similarly, the positive control (\*\**p* < 0.01) group also demonstrated a positive DI value. Locomotor activity was evaluated by the number of line crossings in the test arena. **Figure 2C**,**D** indicates that all groups demonstrated significantly higher locomotor activity (\*\*\**p* < 0.001, \*\**p* < 0.01) than did the group that received LPS alone.

### Effects of *Phyllanthus amarus* Extract on Lipopolysaccharide-Induced Pro-Inflammatory Cytokines

The concentrations of TNF-α and IL-1β proteins were measured in rat brains. Elevation of pro-inflammatory cytokines in the LPS-induced group was observed as an evidence for



*RT, retention time; RSD, relative standard deviation.*

neuroinflammation through systemic inflammation. PA extracttreated groups showed a significant decrease of TNF-α (\*\*\**p* < 0.001, \*\**p* < 0.01, respectively) and IL-1β (\**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, respectively) levels, than did vehicle control in 14- and 28-day treatments. The percentage of inhibition of TNF-α and IL-1β is shown in **Table 2**.

### Determination of Nitric Oxide Levels

Administration of PA extract for 14 and 28 days at three different doses showed significant reduction of NO production (\*\**p* < 0.01, \*\*\**p* < 0.001). NO level in the LPS-induced group was greatly increased than in other groups (**Table 3**).

### Effects of *Phyllanthus amarus* Extract in Lipopolysaccharide-Induced Changes in Pro-Inflammatory Enzymes and Synaptic Marker

**Figure 3** shows the expression of iNOS protein (A and B) and synaptophysin immunoreactivity (C and D) following 14 and 28 days' treatment of PA extract in the brain tissue homogenate. The doses of 100 and 200 mg/kg of PA extract significantly decreased the LPS-induced iNOS protein expression (\**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001), respectively, after 14 and 28 days of pre-treatment. The levels of presynaptic marker synaptophysin


*TNF-*α*, tumor necrosis factor-*α*; IL-1*β*, interleukin-1-beta; IBF, ibuprofen; PA, Phyllanthus amarus; LPS, lipopolysaccharide. Data expressed as mean ± standard error of the mean (n = 3); \*\*\*p < 0.001 and \*\*p < 0.01 compared with vehicle control (vehicle + LPS).*

#### TABLE 3 | Percentage inhibition of nitric oxide (NO) production.


*TNF-*α*, tumor necrosis factor-*α*; IL-1*β*, interleukin-1-beta; IBF, ibuprofen; PA, Phyllanthus amarus; LP, lipopolysaccharide. Data expressed as mean ± standard error of the mean (n = 3); \*\*\*p < 0.001 and \*\*p < 0.01 compared with vehicle control (vehicle + LPS).*

were significantly decreased by LPS (\*\*\**p* < 0.001), whereas in the PA-treated groups, the level of synaptophysin significantly increased (\*\*\**p* < 0.001).

### Protective Effects of *Phyllanthus amarus* Extract Against Glial Cell Activation

The CD11b/c integrin is expressed in activated microglial as a surface marker. It is an effective marker to recognize microglial activation at the time of neurodegeneration. **Figure 4A**,**B** shows the immunohistochemistry staining and intensity of CD11b/c integrin expression in the hippocampal region of the rat brain after 14 and 28 days' administration of PA extract. The LPStreated group, as compared with other treated groups, showed small and numerous slender cell morphological features, which clearly explains the presence of activated microglia. Pre-treatment of PA extract of 200 and 400 mg/kg in 14 (\**p* < 0.05) and 28 days (\*\**p* < 0.01) showed a significant reduction of CD11b/c integrin expression comparable with that of the positive control.

### DISCUSSION

PA has been widely studied for its anti-inflammatory activity in *in vitro* and *in vivo* models. A previous study demonstrated that PA prevented memory impairment and possessed anticholinergic and anti-inflammatory properties (Joshi and Parle, 2006). These findings suggested that PA is neuroactive and can alter the brain functions. However, to the best of our knowledge, no studies have been done to determine the effects of PA on neuroinflammation. Therefore, the present study sought to investigate the protective effects of PA extract against memory impairment and immune responses in LPS-induced neuroinflammation in rodents. The biological activities of medicinal plants are usually due to the presence of bioactive phytoconstituents. Similar to other studies, our results showed the presence of major phytoconstituents such as gallic acid, geraniin, corilagin, ellagic acid, niranthin, phyllanthin, hypophyllanthin, phyltetralin, and isonirtetralin in PA. The present finding is in accordance with previous studies reported for the presence of lignans and polyphenols in this species (Jantan et al., 2014). We have also demonstrated that subchronic treatment of PA extract did not show any signs of toxicity in the animal behavior. In addition, no pathological changes were detected in pyramidal cells of the hippocampus and cerebral cortex. A similar study in our laboratory showed that 80% ethanol extract of PA at doses ranging from 100 to 500 mg/kg given orally did not cause any abnormal behavior and mortality in rats (Ilangkovan et al., 2015). Taken together, it indicates that the concentrations of the extract used for this study were within the safe limit.

Novel object discrimination task was performed to assess the rodents' working memory and preferences towards novelty (Ennaceur, 2010). After 14 days' administration of PA extract (100, 200, and 400 mg/kg), rats showed positive preference towards the novel object than did the vehicle control group. However, 28 days' administration of PA extract at 200 mg/kg showed a significant positive preference for the novel object but not with the other doses. It is suggested that pre-treatment with PA protected against LPS-induced impairment in non-spatial memory at an optimum dose of 200 mg/kg and prevented hypolocomotion induced by LPS (Custódio et al., 2013; Tortorelli et al., 2015), which were comparable with the effects of IBF. Cognitive studies done with other *Phyllanthus* species, namely, *P. emblica*, *P. niruri*, and *P. reticulates* have also demonstrated reversal of memory impairment possibly due to the anti-cholinesterase activity of this genus (Ambali et al., 2012; Malve et al., 2014; Uddin et al., 2016).

Cognitive impairment is noticeable in Alzheimer's disease (AD) and also in other disorders where neuroinflammation is proposed to play a prominent role (Ownby, 2010). Initiation of TLR4, which is expressed in astrocytes, microglia, and neuron can antagonistically

influence neuronal plasticity and survival in brain injury (Okun et al., 2012). Additionally, TLR4 signaling reduces neurogenesis and results in cognitive impairment. As a consequence, microglial cells are activated and release inflammatory mediators like TNFα, IL-1β, IL-6, iNOS, NO, reactive nitrogen species, and reactive oxygen species which mediates oxidative stress (Zhang et al., 2018). Indeed, an inflammatory response is a common observation in the brain tissue of patients with dementia (Zhao et al., 2018). Likewise, our present study showed that LPS increased the levels of proinflammatory cytokines (TNF-α and IL-1β), which act as a central part in the onset and maintenance of inflammation. Related proteins such as iNOS and NO were significantly increased, and a decrease in the level of synaptic marker was noted after exposure to LPS. Indeed, it is known that high concentrations of NO but not the lower concentrations may advance excitotoxicity and result in cognitive impairment. It was suggested that LPS might stimulate the inflammatory cytokines levels in the brain *via* TLR4 activation and induces neuroinflammation (Zhang et al., 2018).

The immune-mediated changes seen with LPS were effectively prevented by pre-treatment of PA extract at 200 and 400 mg/kg for 14 and 28 days. In fact, previous studies have reported that corilagin from PA, ellagic acid, and geraniin were found to decrease the release of TNF-α and IL-β in the brain (Farbood et al., 2015; Tong et al., 2016). Additionally, there is a correlation between pro-inflammatory cytokine release with the decreased level of synaptophysin, which affects memory status (Strużynska et al., 2006). In the present study, we demonstrated that 200 and 400 mg/kg of PA extract given for 14 and 28 days attenuated the LPS-induced pro-inflammatory cytokine release with resulting synaptic loss and memory impairment suggestive of a protective effect against LPS-induced neuroinflammation. Neuroinflammatory conditions, for example, traumatic brain damage, AD, Down syndrome, and

microglial activation in hippocampus after 14 (A) and 28 (B) days' treatment. Graphical representation (A, B) refers to CD11b/c immunoreactivity area in rats' hippocampus. Data expressed as mean ± standard error of the mean (\**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001 vs. LPS, respectively). LPS, lipopolysaccharide; IBF, ibuprofen (positive control); and PA, *Phyllanthus amarus* extract.

aging are often presented with memory impairment. A relationship between cytokine expression in the brain and memory deficits has also been established. For example, it has been demonstrated that the release of TNF-α and IL-1β is a sign of neuroinflammatory-induced memory impairment. Indeed, acute inflammation induced by LPS or IL-1β infusion led to memory impairment (Shaftel et al., 2007). Similarly, previous findings demonstrated that hippocampal IL-1β overexpression hinders contextual and spatial long-term memory (Thomson and Sutherland, 2005). It becomes increasingly evident that prevention of memory impairment observed in the PA-treated groups could be related to immunological changes in the brain.

Astrocytes, microglia, and neurons respond to the acute or chronic stimulation and can aggressively influence neuronal plasticity and surface marker (Okun et al., 2012) resulting in neuroinflammation. CD11b/c is considered as an active integral marker of the innate immune response in microglial cells. Researches revealed that activation of microglia is often noted by expressing the surface marker CD11b/c in either the administration of endotoxins (for example LPS) or injury (Liu et al., 2016). It is believed that microglial activation expresses various proteins and surface markers. Among them, CD11b/c is more sensitive and an intense marker, which plays a role as a binding protein for intracellular adhesion molecule-1 and complement receptor type 3. Therefore, CD11b/c was examined in the hippocampal region of the brain tissue. The outcome of the present study showed that PA extract and IBF attenuated CD11b/c integrin expression in LPS-induced rat brain, which points to a notion that PA inhibits microglial activation similar to IBF. Studies have shown that increased microglial marker CD11b/c are influenced by NO production. However, their complete signaling mechanism is uncertain to researchers (Nillert et al., 2017).

Most of the changes in cognitive function and immunological markers were optimum at 200 mg/kg of PA extract, and increasing the dose by two-fold did not cause any further increase in the effects. It is highly unlikely that the lack of dose–response effects between 200 and 400 mg/kg in many of the parameters measured was due to toxicity. This is evident from our toxicity study that did not demonstrate any neurotoxicity at the highest dose of 400 mg/ kg. The lack of dose–response effects could also be due to saturation of receptors at higher concentrations where a similar observation has been reported in other studies (Lima et al., 2016; Manalo et al., 2017). In the present study, it is unclear as to whether the observed effects were produced by the major phytoconstituents of the plant or as a result of a combined effects of various phytochemicals present in the extract. Indeed, this study sought to determine the efficacy of PA crude extract in modulating inflammatory responses in the brain by measuring only selected markers of neuroinflammation. Although limited, the present findings provide an early indication of the protective actions of PA against non-spatial memory impairment, which may result from inflammatory processes in the brain. It is also unclear if PA alleviated inflammation in the brain by inhibiting TLR4 activation induced by LPS. A different group from our laboratory has similarly demonstrated an antiinflammatory action of PA through inhibition of LPS-induced responses in human macrophages (Harikrishnan et al., 2018). The study revealed that PA targeted the NF-κB, mitogen-activated protein kinase, and phosphatidylinositol 3-kinase/protein kinase B signaling pathways to exert its anti-inflammatory effects. Therefore, on the basis of our present findings, we suggest that PA exhibited anti-inflammatory actions *via* LPS-induced signaling pathway.

### CONCLUSION

Pre-treatment with PA extract for 14 and 28 days was comparable with that of IBF in the prevention of non-spatial memory impairment and alleviation of neuroinflammatory responses induced by LPS. However, further investigations are warranted to support this notion and to better understand the exact protective mechanisms of PA in altering immune and inflammatory responses in the brain. Identifying the bioactive compounds in the plant that are responsible for these effects is also essential in future studies.

### ETHICS STATEMENT

The studies were performed according to procedures for the use of animals in research as approved by the UKM Animal Ethics Committee with the approval number FF/2017/

### REFERENCES


NORAZRINA/24-MAY/850-JUNE-2017-JULY-2018 for the toxicity assessment in rats and FF/2015/NORAZRINA/20- MAY/683-MAY-2015-MAY-2016 for the efficacy and molecular study done in rats.

### AUTHOR CONTRIBUTIONS

NA, IJ, and EK were involved in the study design, analysis, and interpretation of the data. AA performed the experiments and data analysis. SO participated in the immunohistochemistry work, and MA analyzed the brain histology. The manuscript draft was written by AA and finalized by NA, IJ, EK, SO, and MA.

### FUNDING

This study was supported by the Ministry of Agriculture & Agro-Based Industry, Malaysia under the NKEA Research Grant Scheme (NRGS) with grant no. NH1014D023.

### ACKNOWLEDGEMENTS

Immunohistochemistry work was conducted with a technical support provided by Ms. Rachel Shalini from the Brain Research Institute, Jeffrey Cheah School of Medicine and Health Sciences, and the Bioimaging Infrastructure Platform, Monash University Malaysia.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019.00632/ full#supplementary-material


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

*Copyright © 2019 Alagan, Jantan, Kumolosasi, Ogawa, Abdullah and Azmi. 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.*

## Antidepressant-Like Activity of Myelophil *via* Attenuation of Microglial-Mediated Neuroinflammation in Mice Undergoing Unpredictable Chronic Mild Stress

*Jin-Seok Lee1, Won-Young Kim1, Yoo-Jin Jeon1, Sung-Bae Lee1, Dong-Soo Lee2 and Chang-Gue Son1\**

*1 Institute of Traditional Medicine and Bioscience, Dunsan Hospital of Daejeon University, Daejeon, South Korea, 2 Department of Internal Medicine, Daejeon St. Mary's Hospital, The Catholic University of Korea, Daejeon, South Korea* 

#### *Edited by:*

*Leanne Stokes, University of East Anglia, United Kingdom*

#### *Reviewed by:*

*Trisha Anne Jenkins, RMIT University, Australia Karolina Pytka, Jagiellonian University, Poland*

> *\*Correspondence: Chang-Gue Son ckson@dju.ac.kr*

#### *Specialty section:*

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

*Received: 07 February 2019 Accepted: 27 May 2019 Published: 13 June 2019*

#### *Citation:*

*Lee J-S, Kim W-Y, Jeon Y-J, Lee S-B, Lee D-S and Son C-G (2019) Antidepressant-Like Activity of Myelophil via Attenuation of Microglial-Mediated Neuroinflammation in Mice Undergoing Unpredictable Chronic Mild Stress. Front. Pharmacol. 10:683. doi: 10.3389/fphar.2019.00683*

Myelophil, a 30% ethanol extract that has an equal rate in both *Astragali Radix* and *Salviae Radix*, is a remedy for the treatment of fatigue-linked disorders in traditional Oriental medicine. The majority of patients with chronic fatigue have a risk of comorbidity with depression symptoms. To evaluate the anti-depressant activity of Myelophil, mice were subjected to unpredictable chronic mild stress (UCMS, eight different stresses) for 3 weeks with daily administration of distilled water, Myelophil (25, 50, or 100 mg/ kg), or *n*-acetyl-*l*-cysteine (NAC) (100 mg/kg). After the final stress exposure, three behavioral tests, including the open field test (OFT), forced swimming test (FST), and tail suspension test (TST), and stress-derived alterations of the serotonergic signal and inflammatory response in the hippocampus were measured. UCMS notably induced depressive behaviors, whereas these behavioral alterations were significantly reversed by the administration of Myelophil in regard to the OFT, FST, and TST results. Myelophil also significantly attenuated the over-activation of microglial cells and the inflammatory response in the hippocampal region (TNF-α, tumor necrosis factor-alpha; IL-1β, interleukin-1beta; and caspase-1). Furthermore, Myelophil significantly restored the distortions of serotonergic function in the dorsal raphe nuclei and neurogenesis in the subgranular zone of the hippocampus. These results support the clinical relevance of the anti-depressant activity of Myelophil, specifically by modulating serotonergic function and the neuroinflammatory response.

Keywords: myelophil, depression, anxiety, antidepressants, microglia, serotonin

### INTRODUCTION

Depression, a pervasive emotional disorder, is the single largest contributor to the global burden of disease (WHO, 2017). The number of patients who suffer from depressive disorder is estimated to be 322 million worldwide (WHO, 2017). Its major symptoms, such as low mood, a feeling of sadness, and a loss of interest in things, lead to low socio-economic activity (Nestler et al., 2002). Annually, 0.8 million patients with depressive disorder commit suicide (Klonsky et al., 2016). However, the pathophysiological mechanisms have been unclear to date, and no curative therapeutics exists yet.

As a credible etiology, the hypothalamic–pituitary–adrenal (HPA) axis hypothesis is suggested (Pariante and Lightman, 2008). Accumulating evidence suggests that HPA axis hyperactivity induces the overproduction of brain pro-inflammatory cytokines, such as tumor necrosis factor-alpha (TNF-α) and interleukin-1beta (IL-1β), through microglial activation (Brites and Fernandes, 2015; Yirmiya et al., 2015), and this phenomenon has consistently been observed in subjects with depressive disorders (Zou et al., 2018). Recently, the role of Nucleotide-binding oligomerization domain (NACHT), Leucine-rich repeat (LRR), and Pyrin domation (PYD) domains-containing protein 3 (NLRP3) inflammasome-dependent IL-1β has emerged as a novel contributor to depressive disorders (Alcocer-Gomez and Cordero, 2014; Kaufmann et al., 2017). These molecular alterations eventually led to impaired serotonergic synthesis and neurotransmission, which are characteristics of depression (Velasquez and Rappaport, 2016). Therefore, maintenance of serotonergic homeostasis is considered as a main strategy for psychiatric disorders (Nordquist and Oreland, 2010).

Antidepressants have been developed on the basis of serotoninergic signal modulation, while selective serotonin reuptake inhibitors (SSRIs) and serotonin–norepinephrine reuptake inhibitors (SNRIs) represent the majority of the antidepressant market (Artigas, 2013). The global antidepressant market was estimated to be 11.6 billion dollars in 2017 (Mordor-Intelligence, 2017), and serotoninergic modulators account for 90% of market (Artigas, 2013). Nevertheless, current antidepressants have critical limitations, such as extensive adverse effects, poor treatment compliance and remission rate, and a high risk of relapse following drug withdrawal, for example, 61.8% in the case of fluoxetine (Gaynes et al., 2009; Andrews et al., 2012; Berwian et al., 2017).

Medicinal herbs have been attractive as candidates in antidepressant drug development (Lee and Bae, 2017). Myelophil, a 30% ethanol extract of *Astragali Radix* and *Salviae Radix*, is used to treat fatigue-associated disorders, including idiopathic chronic fatigue and chronic fatigue syndrome, in clinics of traditional Korean medicine. We previously presented that Myelophil exerted anti-fatigue in both clinical and experimental studies (Cho et al., 2009; Lee et al., 2015). Besides, our previous findings showed the neuropharmacological actions of Myelophil against hippocampal memory dysfunction, brain oxidative damage, and endocrine abnormality of HPA axis (Kim et al., 2013; Kim et al., 2014; Lee et al., 2014). However, there is no scientific evidence for the pharmacological property of Myelophil in depressive disorder. We hypothesized that Myelophil might have antidepressant-like effects because of the high relevance and comorbidity between chronic fatigue and depression (Demyttenaere et al., 2005; Penner and Paul, 2017).

The present study aimed to investigate the antidepressant-like properties of Myelophil and its underlying mechanisms using an unpredictable chronic mild stress (UCMS)-induced depression mouse model. In order to compare the relative pharmacological potential, NAC was adapted as a positive control.

### MATERIALS AND METHODS

### Myelophil Preparation and Standardization

Myelophil is composed of 30% ethanol extract in equal amounts of *Astragali Radix* (*Astragalus membranaceus*) and *Salviae Radix* (*Salvia miltiorrhizae*). Myelophil was manufactured by Kyung-Bang Pharmacy (Incheon, Republic of Korea, lot. no. KB-Myelo-1801) according to the approved good manufacturing practice (GMP) guidelines of the Korean Ministry of Food and Drug Safety (MFDS). Fingerprinting analysis of Myelophil was performed to confirm the reproducibility as previously described (Lee et al., 2018). Briefly, ultra-high-performance liquid chromatography (UHPLC, Thermo Scientific, San Jose, CA, USA) coupled with a high-resolution LTQ Orbitrap mass spectrometry (MS) system (Thermo Scientific Co., San Jose, CA, USA) was used, and identifying analysis was conducted with each relative reference compound (rosmarinic acid; salvianolic acid A, B, C, and D; and formononetin).

### Animals and Stress Procedure

Forty-eight specific pathogen-free BALB/c male mice (8 weeks old, 22–24 g) were purchased from Dae Han Biolink Co., Ltd. (Eumseong, Republic of Korea). They were housed in plastic cages maintained at 23 ± 1°C with a 12-h light–dark cycle and freely fed food pellets (Cargill Agri Purina, Gyeonggido, Republic of Korea) and water. After acclimation for 1 week, the mice were randomly divided into six groups (*n* = 8): vehicle, UCMS, Myelophil (25, 50, or 100 mg/kg), and NAC (100 mg/kg, as a positive control) groups.

Animal care and experiments were conducted in accordance with the guidelines issued by the Institutional Animal Care and Use Committee of Daejeon University (Daejeon, Republic of Korea; Approval No. DJUARB 2017-017) and the Guide for the Care and Use of Laboratory Animals published by the United States National Institutes of Health.

The UCMS procedure was conducted as previously described (Nollet et al., 2013) with slight modifications. Briefly, except vehicle group, mice were subjected to a stress paradigm once per day over a period of 3 weeks: continuous illumination during the dark cycle, wet bedding for 24 h, isolation stress for 24 h, 45° tilting for 12 h, food and/or water deprivation for 12 h, restraint stress for 3 h, 4°C cold stress for 1 h, and swimming in cold water for 15 min. Detailed information for stress procedure is indicated in **Supplementary Figure 2**. After the final day of stress, the mice were sequentially subjected to behavioral tests to assess the depression/ anxiety-like behavior [open field test (OFT) on day 22, a forced swimming test (FST) on day 23, and a tail suspension test (TST) on day 24]. During the entire experiment period, the mice were orally administered with distilled water (vehicle and UCMS), Myelophil, or NAC, respectively, at 11:00 am every day. The experimental scheme is summarized in **Supplementary Figure 2**.

### Open Field Test, Forced Swimming Test, and Tail Suspension Test

The OFT was performed as previously described (Ieraci and Herrera, 2006) with slight modification. The plastic enclosure box for the open field apparatus was contained in the black square side (40 × 40 × 30 cm), and the center of the field was Lee et al. Myelophil Exerts Antidepressant Effects

distinguishable in the recording software. To evaluate the depressive and anxious conditions, each mouse was placed in the center of the field, and their spent time in the center zone and the total distance were subsequently recorded for 5 min at 25-lux illumination using a video camera connected to the corresponding software (Smart Junior).

The FST was performed as previously described (Porsolt et al., 1977; Chatterjee et al., 2012) with slight modification. The apparatus was contained in the plastic cylinder (30 × 30 × 50 cm), and it was filled with tap water at 25 ± 1°C up to 23 cm in height. Individual mouse was allowed to swim for 1 min (pre-test), and the immobility time (passive floating with no additional activity) and global activity (swimming duration) were recorded for 5 min at 30-lux illumination. The behavior between immobility and activity was judged by corresponding software (Smart Junior).

The TST was conducted as previously described (Steru et al., 1985; Chatterjee et al., 2012) with slight modification. The apparatus for the test consisted of a rectangular box (30 × 30 × 50 cm) with a rack on the top. Each mouse was individually suspended for 1 min for pre-test, and behaviors such as immobility and activity were recorded for 5 min at 25-lux illumination by software (Smart Junior).

Animal behavior tracking software (Smart Junior, Panlab SL; Barcelona, Spain) was used for recording distances, speeds, trajectories, and global activity (activity duration). Immobility and activity duration were judged by designated threshold (immobility, 0 to 120; low and high activities, 121 to 300). Behavioral tests were performed by researchers as a blind manner to experimental conditions.

### Sample Preparation

After the final behavioral tests, the mice were sacrificed under CO2 anesthesia on day 25. The serum was collected by centrifugation at 3,000 × *g* for 15 min. The brains of the five mice for each group were immediately removed and dissected to isolate the hippocampal tissue. The sera and hippocampi were stored at −80°C or RNAlater (Ambion, TX, USA) until use. The hippocampal tissue was homogenized in a radioimmunoprecipitation assay (RIPA) buffer, which was used for the biochemical analysis. For the immunohistological analysis, the remaining three mice for each group were subjected to transcardial perfusion with heparin (10 units/mL) and paraformaldehyde (PFA) solution, and the brains were maintained in 4% PFA solution. The total protein concentrations were measured using a bicinchoninic acid protein assay kit (Sigma). The absorbance at 560 nm was measured using a UV spectrophotometer (Molecular Devices Corp., Sunnyvale, CA, USA).

### Immunohistological Analysis of DCX, Iba-1, and 5-HT

The brains were gradually cryoprotected in 10%, 20%, and 30% sucrose for each 24-h interval and were subsequently embedded in tissue-freezing medium (Leica Microsystems, Bensheim, Germany) with liquid nitrogen. They were cut into frozen coronal sections (35 μm) using a Leica CM3050 cryostat. The sections were stored in free-floating buffer. For the immunohistological analysis, after washing with ice-cold PBS, the sections were blocked in 5% normal chicken serum (which contained 0.3% Triton X-100 in PBS) for 1 h. After being washed, the sections were incubated with primary antibodies against doublecortin (DCX, 1:200, sc-8066, Santa Cruz Biotechnology), ionized calcium binding adaptor molecule 1 (Iba-1, 1:200, 019-19741, Wako), or 5-hydroxytryptamine (5-HT, 1:200, ab66047, Abcam) overnight at 4°C. The sections were incubated with donkey anti-goat IgG H&L (1:400, Alexa Fluor® 488, ab150129, Abcam) or goat anti-rabbit IgG Horseradish peroxidase (HRP) (1:400, ab6722, Abcam) secondary antibodies for 2 h at room temperature. For the Iba-1-positive signal, the sections were subsequently exposed to an avidin–biotin peroxidase complex (Vectastain ABC kit, Vector Laboratories) for 2 h. The peroxidase activity was visualized using a stable diaminobenzidine solution. For the DCX- and 5-HT-positive signal, the sections were subsequently exposed to 4', 6-Diamidino-2-Phenylindole, Dihydrochloride (DAPI) (1:1,000, D9542, Sigma) to stain cell nuclei. Immunoreactions were observed under an Axio-phot microscope (Carl Zeiss, Germany), and the signals were quantified using ImageJ 1.46 software (NIH, Bethesda, MD, USA).

### Determination of Corticosterone and Pro-Inflammatory Cytokines

The serum corticosterone level was determined using a DetectX® corticosterone Arbor assay kit (Ann Arbor, MI, USA). The absorbance at 450 nm was measured using a UV spectrophotometer. The levels of pro-inflammatory cytokines in the hippocampal homogenates were determined using commercially available enzyme immunoassay (EIA) kits for TNF-α (BD Biosciences, San Diego, CA, USA) and IL-1β (R&D Systems Inc., Minneapolis, MN, USA), and the absorbance was read within 10 min at 450 and 570 nm using a UV spectrophotometer (Molecular Devices).

### Determination of Caspase-1 Activity

The caspase-1 activity in the hippocampal homogenates was measured using a mouse caspase-1 enzyme-linked immunosorbent assay kit (Novus Biologicals, Littleton, CO, USA). The absorbance at 450 nm was measured using a UV spectrophotometer (Molecular Devices).

### Western Blotting Analysis

The protein expressions of NACHT, LRR, and PYD Domains-Containing Protein 3 (NLRP3), apoptosis-associated Speck-like protein containing a CARD (ASC), pro-IL-1β and mature IL-1β, and β-actin in the hippocampal homogenates were evaluated using a western blotting method. The protein concentration of the homogenates was equalized, and the samples were separated by 10% polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride (PVDF) membranes. To minimize the non-specific binding, the membrane was blocked in 5% bovine serum albumin for 1 h. The membranes were incubated with primary antibodies, such as NLRP3 (1:500, ab214185, Abcam), ASC (1:500, sc-271054, Santa Cruz), pro-IL-1β and mature IL-1β (1:1,000, ab9722, Abcam), or β-actin (1:2500, PA1-183, Thermo-Fisher Scientific), overnight at 4°C. After being washed, the membranes were incubated with an HRP-conjugated anti-rabbit or anti-mouse antibody (GeneTex, Inc., Irvine, CA) for 1 h. The western blotting results were visualized with an enhanced chemiluminescence (ECL) advanced kit. The intensity was analyzed with ImageJ version 1.46 (NIH, Bethesda, MD, USA).

### Determination of Nitric Oxide in BV2 Microglial Cells

Mouse microglia cells (BV2 cell line) were cultured in DMEM supplemented with 10% Fetal Bovine Serum (FBS) and 1% penicillin–streptomycin. BV2 cells were incubated at 37°C under 5% CO2. The BV2 cells were seeded at 2 × 104 cells/well into 96-well microplates. After incubation for 12 h, the BV2 cells were pre-treated with Myelophil (5, 10, or 20 μg/mL) or NAC (100 μM) for 2 h. After exposure to the gram-negative lipopolysaccharide bacteria (LPS, 1 μg/mL) for 24 h, the nitric oxide levels of the cell supernatants were determined using the previously described method (Green et al., 1982). Briefly, the cell supernatant was responded by the Griess reagent [1% sulfanilamide, 0.1% *N*-(1-naphthyl) ethylenediamine hydrochloride, and 2.5% H3PO4] at 37°C for 20 min, and the resulting purple azo dye product was subsequently measured at 540 nm using a UV spectrophotometer (Molecular Devices).

### Statistical Analysis

Data are expressed as the mean ± standard error of the mean (SEM). The statistically significant differences between the groups were evaluated by one-way analysis of variance (ANOVA) followed by *post hoc* multiple comparisons with Tukey's honestly significant difference (HSD) test using IBM SPSS statistics software, ver. 25.0 (SPSS Inc., Chicago, IL, USA). Differences at *p* < 0.05 indicate statistical significance.

### RESULTS

### Compounds Present in Myelophil

The major peaks and their retention times, formulae, and molecular weight were consistent with previous analysis (Lee et al., 2018). Six major peaks were detected at 7.47, 8.14, 8.27, 8.71, 9.38, and 12.65 min of retention time under the UV wavelength of 254 nm. Their mass to charge ratios (*m*/*z*) were analyzed in 419.0969 (salvianolic acid D), 361.0917 (rosmarinic acid), 493.1125 (salvianolic acid C), 719.1604 (salvianolic acid B), 495.1276 (salvianolic acid A), and 269.0806 (formononetin) *m*/*z* as displayed in **Supplementary Figure 1**.

### Effects on Depressive and Anxious-Like Behaviors

As an anxiety-based test, the OFT was conducted, and the UCMS significantly decreased the total distance [*F*(5, 42) = 12.68; *p* < 0.001] (*p* < 0.01) and time spent in the center zone [*F*(5, 42) = 5.69; *p* < 0.001] (*p* < 0.05) compared with those in the vehicle-treated group. These reductions were completely reversed by the treatment of Myelophil, particularly at the dose of 100 mg/kg, compared with those in the UCMS group (*p* < 0.05 and *p* < 0.01, respectively, **Figure 1A** and **B**).

To evaluate depressive-like behaviors, the forced swimming test (FST) and tail suspension test (TST) were performed. Regarding the FST results, the UCMS significantly reduced the global activity [*F*(5, 42) = 18.59; *p* < 0.001] and elevated immobility time [*F*(5, 42) = 19.96; *p* < 0.001] compared with those in the vehicle-treated group (*p* < 0.01 for both parameters). Myelophil treatment, however, showed noteworthy antidepressant-like activity particularly for the 100 mg/kg dose (*p* < 0.01, **Figure 1C** and **D**).

A low global activity [*F*(5, 42) = 42.55; *p* < 0.001] and prolonged immobility time [*F*(5, 42) = 55.84; *p* < 0.001] in the TST were induced by the UCMS, and they were significantly different compared with those in the vehicle-treated group (*p* < 0.01 for both parameters). Myelophil treatment (100 mg/kg) significantly attenuated these behavioral alterations compared with those in the UCMS group (*p* < 0.01, **Figure 1E** and **F**). NAC also showed positive effects in the depressive and anxious-like behaviors compared with those in the vehicle-treated group; however, these effects were not significant in all tests.

### Effects on Microglial Activation in Hippocampus

The UCMS remarkably increased the microglial activation in cornus ammonis (CA)1 [*F*(5, 12) = 20.52; *p* < 0.001], dentate gyrus [*F*(5, 12) = 23.88; *p* < 0.001], and CA3 region [*F*(5, 12) = 14.28; *p* < 0.001] of the hippocampus (*p* < 0.01 for all regions). Hyper-activation of microglia cell was significantly attenuated by the Myelophil treatment especially in doses of 50 and 100 mg/kg (*p* < 0.05 or *p* < 0.01, **Figure 2A** and **B**). NAC also showed similar effects against microglial activation.

### Effects on Serum Corticosterone

The UCMS significantly increased the serum corticosterone (2.9-fold) compared with that in the vehicle-treated group [*F*(5, 24) = 4.69; *p* = 0.004] (*p* < 0.01), while the Myelophiltreated group showed significant reductions for the high corticosterone level (*p* < 0.05 for both 50 and 100 mg/kg, **Figure 3A**). Treatment with NAC also showed similar effects as Myelophil.

### Effects on Nitric Oxide in BV2 Microglia Cells

The LPS induced an inflammatory response in BV2 cells, as evidenced by the increase of nitric oxide [*F*(5, 24) = 99.67; *p* < 0.001] (4.3-fold, *p* < 0.01) in the cell supernatant. Pretreatment with Myelophil inhibited the highly increased nitric oxide level than did the LPS-treated cells (*p* < 0.01 for all doses, **Figure 3B**). Nitric oxide scavenging activity was also present in NACpretreated cells similar to Myelophil.

### Effects on Pro-Inflammatory Cytokines in Hippocampus

The UCMS-subjected group showed significant elevations of TNF-α [*F*(5, 24) = 9.29; *p* < 0.001] (1.9-fold) and IL-1β

immobility time (D) in the forced swimming test (on the 23rd day), and the global activity (E) and immobility time (F) in the tail suspension test (on the 24th day) were assessed. The data are expressed as the mean ± standard error of the mean (SEM) (*n* = 8). Significant differences were evaluated by one-way analysis of variance (ANOVA) with Tukey's honestly significant difference (HSD) *post hoc* test. #*p* < 0.05 and ##*p* < 0.01 compared with the vehicle-treated group; \**p* < 0.05 and \*\**p* < 0.01 compared with the UCMS-subjected group.

[*F*(5, 24) = 2.67; *p* = 0.047] (1.8-fold) in the hippocampal tissue than did the vehicle-treated group (*p* < 0.05 and *p* < 0.01, respectively). These elevations were significantly attenuated by Myelophil treatment compared with the UCMS group: TNF-α level (*p* < 0.05 for 25 and 100 mg/kg, *p* < 0.01 for 50 mg/kg) and IL-1β level (*p* < 0.05 for 50 and 100 mg/kg; **Figure 3C** and **D**). The anti-inflammatory effects of NAC were only present in the TNF-α result.

### Effects on Hippocampal NLRP3 Inflammasome

The UCMS significantly activated the NLRP3 inflammasome, as evidenced by NLRP3 [*F*(5, 24) = 143.56; *p* < 0.001] (2-fold, *p* < 0.01), ASC [*F*(5, 24) = 113.80; *p* < 0.001] (1.5-fold, *p* < 0.01), pro-IL-1β [*F*(5, 24) = 50.25; *p* < 0.001] (*p* < 0.05), and mature IL-1β [*F*(5, 24) = 220.91; *p* < 0.001] (1.9-fold, *p* < 0.01) compared with those in the vehicle-treated group. Myelophil treatment

notably inhibited against NLRP3 inflammasome activation; the differences in NLRP3, ASC, pro-IL-1β, and mature IL-1β were statistically significant compared with those in the UCMS group (*p* < 0.05 or *p* < 0.01 for 100 mg/kg, **Figure 4A** and **C**).

### Effects on Caspase-1 Activity in Hippocampus

The caspase-1 activity in the hippocampal tissue of the UCMSsubjected group was higher than that in the vehicle-treated group [*F*(5, 24) = 5.17; *p* = 0.002] (approximately 2.1-fold, *p* < 0.01), whereas the hyper-activity of caspase-1 was significantly inhibited by Myelophil treatment compared with the UCMS group (*p* < 0.05 for 100 mg/kg, **Figure 4B**). NAC treatment also had a similar effect on the caspase-1 hyper-activation.

### Effects on Hippocampal Neurogenesis and Serotonergic Function in Dorsal Raphe Nuclei

The UCMS exerted a predominant reduction of the 5-HT activity in the region of the dorsal raphe nuclei compared with those in the vehicle-treated group [*F*(5, 12) = 33.42; *p* < 0.001] (*p* < 0.01), whereas treatment with Myelophil notably recovered the low 5-HT activity (*p* < 0.01 for both 50 and 100 mg/kg). The DCX-positive signal in the subgranular zone of the hippocampus was markedly decreased by UCMS compared with that in the vehicle-treated group [*F*(5, 12) = 14.13; *p* < 0.001] (*p* < 0.01). Moreover, an increase of DCX-positive neuronal dendrites was observed in the Myelophil-treated group (*p* < 0.01 for 50 mg/kg, *p* < 0.05 for 100 mg/kg). NAC also showed similar effects as Myelophil. The immunofluorescence staining results were quantified, and there was a significant difference between the UCMS-subjected and Myelophil-treated groups (*p* < 0.01, **Figures 5** and **6**).

## DISCUSSION

In present study, we found that Myelophil could alleviate the depressive-like behaviors *via* modulation of microglial-mediated neuroinflammation. Our findings are the first evidence of the antidepressant-like effects of Myelophil, which indicates the therapeutic possibilities on the mood disorders.

To verify the hypothesis that Myelophil exerts antidepressantlike effect, we chose the BALB/c mouse strain to induce depressive behaviors because it is known to be more susceptible to UCMS relative to other rodent strains (Farley et al., 2012), and the UCMS method has been commonly used as a depressive and anxious animal model (Nollet et al., 2013). The reliability of the

UCMS animal model for inducing depression has been validated by numerous studies, and the unpredictable stress model can mimic the human psychopathology (Willner, 2017). We also adopted the FST, TST, and OFT as the behavioral tests, which are representative assessments for evaluating antidepressant interventions (Cryan et al., 2005; Slattery and Cryan, 2012; Daniel et al., 2017). As expected, chronic exposure to unpredictable stress led to predominant behavioral alterations in the FST, TST, and OFT, while these depression-related behaviors were significantly ameliorated by administration of Myelophil (**Figure 1A** to **F**). These Myelophil-derived results were similar with those of another study using hesperidin (a kind of citrus bioflavonoid) under UCMS-induced depressive behavioral tests (Fu et al., 2019).

To explain the underlying mechanisms of Myelophil, we subsequently examined the stress-responsive system, particularly focusing on the HPA axis hyper-activation. The HPA axis abnormality has been implicated in the pathophysiology of major depressive disorder. Hypercortisolemia is commonly observed in 40% to 60% of depressed patients (Murphy, 1991). The glucocorticoid receptor-mediated negative regulation of cortisol release is impaired in conditions of psychiatric pathology (Pariante and Lightman, 2008). In a previous animal study, chronically dexamethasone-injected mice have shown depressive-like behaviors and a decrease of hippocampal gene expression for glucocorticoid receptor (Skupio et al., 2015). Consistently with previous reports, we verified the increase of serum corticosterone by UCMS. However, this over-release of corticosterone was significantly attenuated by Myelophil administration, which proposed the antidepressant-like activity of Myelophil *via* balancing the endogenous glucocorticoid system (**Figure 3A**).

High glucocorticoids lead to morphological and functional changes of microglial cells in the brain, for example, from resting state into reactive phenotype as a neuroinflammatory response (Nair and Bonneau, 2006). Microglia, resident immune cells in the central nervous system, plays a pivotal role in the pathogenesis of depression (Yirmiya et al., 2015). Antidepressants, such as imipramine, suppressed the M1 phenotype microglia in the hippocampus of mice exposed to chronic mild stress (Zhao et al., 2016). As expected, the administration of Myelophil predominantly attenuated the over-activation of microglia in the CA1, dentate gyrus, and CA3 of the hippocampus (**Figure 2A** and **B**). Microglial activation in the hippocampal granule cell layer, hilus, CA1, and CA3 regions was known to trigger the onset of depression (Iwata et al., 2016). We further confirmed the inhibitory effects of Myelophil against the LPS-induced production of nitric oxide in the BV2 murine microglia cell line (**Figure 3B**). Researchers in pharmacology are attempting to identify the antidepressant agents that involve the regulation of microglia-derived neuroinflammation (Chen et al., 2017). In our study, the hippocampal pro-inflammatory cytokines, including TNF-α and IL-1β, were also completely normalized by Myelophil treatment (**Figure 3C** and **D**).

Emerging evidence reinforces the importance of NLRP3 inflammasome in neuropsychiatric disorders, particularly in depression and anxiety (Velasquez and Rappaport, 2016; Song et al., 2017). One preclinical study showed that 4 weeks of UCMS in mice resulted in depressive-like behaviors, and its contributing factor was NLRP3 inflammasome in the hippocampal region (Zhang et al., 2015). As expected, we found that Myelophil (100 mg/kg) significantly inhibited the hippocampal protein expressions of NLRP3, ASC, and caspase-1 activity (**Figure 4A**–**C**). These results implied the possibility of an NLRP3-dependent antidepressant-like action of Myelophil. In a case–control study, 20 patients with major depressive disorder exhibited high anxiety score and increased levels of caspase-1 and NLRP3 expression in peripheral blood mononuclear cells than did the healthy group (Alcocer-Gomez et al., 2016). The glucocorticoidinduced NLRP3 inflammasome formation was observed in a primary microglia cell isolated from mouse hippocampus (Frank et al., 2014). Furthermore, one clinical study identified a high correlation between the glial activation level and depressive score in patients with CFS using positron emission

tomography (PET) (Nakatomi et al., 2014). These facts support that antidepressant-like effects of Myelophil are linked to anti-fatigue activity.

Excessive microglial activation causes an impairment of the hippocampal neurogenesis under conditions of stress and inflammation (Sierra et al., 2014). In addition, the reduced hippocampal neurogenesis is closely involved in the pathophysiology of major depression (Snyder et al., 2011). Our data also showed a notable suppression of the hippocampal neurogenesis, which was significantly ameliorated by Myelophil treatment (**Figure 5**). It is known that chronic antidepressant treatment (serotonergic modulator) increases adult hippocampal neurogenesis in both non-human primates and humans (Perera et al., 2011; Boldrini et al., 2012). Serotonin, 5-HT, as a monoamine neurotransmitter, is the most important target molecule in depressive disorder. Therefore, serotonin-target medications, such as SSRIs and SNRIs, are prescribed most frequently in approximately 90% of cases (Artigas, 2013). It is particularly interesting that dysregulation of the HPA axis attenuates 5-HT neurotransmission (Mahar et al., 2014). In our results, the

altered 5-HT signals in the dorsal raphe nuclei were restored by Myelophil treatment (**Figure 6A** and **B**).

Myelophil was developed for the treatment of fatiguerelated disorders on the basis of the theory of traditional Chinese/Korean medicine. *Astragali Radix* and *Salviae miltiorrhizae Radix* are the main materials in Myelophil, and these two herbal plants are commonly used to maintain qi and bloodstream homeostasis in the human body (Cho et al., 2009). From our previous studies, Myelophil showed pharmacological actions against fatigue-related pathology such as brain oxidative stress and memory deficit (Lee et al., 2012; Lee et al., 2015). Our present data supported the applicability of Myelophil to depressive symptoms, which clearly accompany chronic fatigue disorders. However, the present study has a limitation of unknown information regarding the active compounds that correspond to the antidepressive property. A study found that salvianolic acid B, a major compound in *Salviae Radix*, exhibits anti-inflammatory effects by modulating NLRP3 inflammasome (Jiang et al., 2017). Salvianolic acid B promotes microglial M2-polarization against UCMS-induced M1 phenotype (Zhang et al., 2017). Astragaloside IV, a main *Astragali Radix* compound, inhibits depressive-like behaviors by regulating the nuclear factor-κB/ NLRP3 axis (Song et al., 2018). Further studies are required to identify the major active compound in the future. We used NAC as a positive control because the antidepressant-like effects of NAC have been explored in several studies (Ferreira et al., 2008; Costa-Campos et al., 2013). One study group is planning to evaluate the inflammation-inhibiting NAC effects on depressive disorder in a randomized placebo-controlled trial (Yang et al., 2018).

Taken together, our findings comprise the first evidence for the antidepressant-like effects of Myelophil, and its underlying mechanism may involve the regulation of NLRP3-dependent

neuroinflammation and the serotonergic signal. This study would support the pharmacological applicability of Myelophil in chronic fatigue, as well as neuropsychiatric disorders.

### ETHICS STATEMENT

Animal care and experiments were conducted in accordance with the guidelines issued by the Institutional Animal Care and Use Committee of Daejeon University (Daejeon, Republic of Korea; Approval No. DJUARB 2017-017) and the Guide for the Care and Use of Laboratory Animals published by the United States National Institutes of Health.

### AUTHOR CONTRIBUTIONS

JL wrote the main manuscript text and conducted the experiments. WK supported the behavioral test for depression and anxiety. YJ prepared **Figure 2** and other immunofluorescence staining results. SL performed a statistical analysis. DL advised on the mechanism of NLRP3 inflammasome activation. CS supervised the preparation of manuscript and directed the final version of all contents. All authors reviewed and approved this manuscript.

## FUNDING

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Oriental Medicine R&D Project, Ministry of Health & Welfare, and Republic of Korea (HI15C0112), as well as the Ministry of Education, Science and Technology (NRF-2018R1A6A1A03025221).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019.00683/ full#supplementary-material

### REFERENCES


signaling, and the kynurenine pathway. *Biol Psychiatry* 80, 4–5. doi: 10.1016/j. biopsych.2016.04.019


**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 © 2019 Lee, Kim, Jeon, Lee, Lee and Son. 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.*

# Annexin A1 Bioactive Peptide Promotes Resolution of Neuroinflammation in a Rat Model of Exsanguinating Cardiac Arrest Treated by Emergency Preservation and Resuscitation

#### Edited by:

Pietro Giusti, University of Padua, Italy

#### Reviewed by:

Harshini Sarojini, University of Louisville, United States Tomas Drabek, University of Pittsburgh, United States

#### \*Correspondence:

Qing Ma qing.ma@duke.edu Zhiquan Zhang zhiquan.zhang@duke.edu

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuropharmacology, a section of the journal Frontiers in Neuroscience

Received: 01 March 2019 Accepted: 28 May 2019 Published: 14 June 2019

#### Citation:

Ma Q, Zhang Z, Shim J-K, Venkatraman TN, Lascola CD, Quinones QJ, Mathew JP, Terrando N and Podgoreanu MV (2019) Annexin A1 Bioactive Peptide Promotes Resolution of Neuroinflammation in a Rat Model of Exsanguinating Cardiac Arrest Treated by Emergency Preservation and Resuscitation. Front. Neurosci. 13:608. doi: 10.3389/fnins.2019.00608 Qing Ma<sup>1</sup> \* † , Zhiquan Zhang2,3 \* † , Jae-Kwang Shim<sup>4</sup> , Talaignair N. Venkatraman<sup>5</sup> , Christopher D. Lascola5,6, Quintin J. Quinones<sup>1</sup> , Joseph P. Mathew<sup>7</sup> , Niccolò Terrando2,3 and Mihai V. Podgoreanu<sup>1</sup>

<sup>1</sup> Systems Modeling of Perioperative Organ Injury Laboratory, Department of Anesthesiology, Duke University, Durham, NC, United States, <sup>2</sup> Neuroinflammation and Cognitive Outcomes Laboratory, Department of Anesthesiology, Duke University, Durham, NC, United States, <sup>3</sup> Center for Translational Pain Medicine, Duke University, Durham, NC, United States, <sup>4</sup> Department of Anesthesiology and Pain Medicine, Yonsei University College of Medicine, Seoul, South Korea, <sup>5</sup> Departments of Radiology and Neurobiology, Duke University, Durham, NC, United States, <sup>6</sup> Duke-UNC Brain Imaging and Analysis Center, Duke University, Durham, NC, United States, <sup>7</sup> Department of Anesthesiology, Duke University, Durham, NC, United States

Neuroinflammation initiated by damage-associated molecular patterns, including high mobility group box 1 protein (HMGB1), has been implicated in adverse neurological outcomes following lethal hemorrhagic shock and polytrauma. Emergency preservation and resuscitation (EPR) is a novel method of resuscitation for victims of exsanguinating cardiac arrest, shown in preclinical studies to improve survival with acceptable neurological recovery. Sirtuin 3 (SIRT3), the primary mitochondrial deacetylase, has emerged as a key regulator of metabolic and energy stress response pathways in the brain and a pharmacological target to induce a neuronal pro-survival phenotype. This study aims to examine whether systemic administration of an Annexin-A1 bioactive peptide (ANXA1sp) could resolve neuroinflammation and induce sirtuin-3 regulated cytoprotective pathways in a novel rat model of exsanguinating cardiac arrest and EPR. Adult male rats underwent hemorrhagic shock and ventricular fibrillation, induction of profound hypothermia, followed by resuscitation and rewarming using cardiopulmonary bypass (EPR). Animals randomly received ANXA1sp (3 mg/kg, in divided doses) or vehicle. Neuroinflammation (HMGB1, TNFα, IL-6, and IL-10 levels), cerebral cell death (TUNEL, caspase-3, pro and antiapoptotic protein levels), and neurologic scores were assessed to evaluate the inflammation resolving effects of ANXA1sp following EPR. Furthermore, western blot analysis and immunohistochemistry were used to interrogate the mechanisms involved. Compared to vehicle controls, ANXA1sp effectively reduced expression of cerebral HMGB1, IL-6, and TNFα and increased IL-10 expression, which were associated with improved neurological scores.

ANXA1sp reversed EPR-induced increases in expression of proapoptotic protein Bax and reduction in antiapoptotic protein Bcl-2, with a corresponding decrease in cerebral levels of cleaved caspase-3. Furthermore, ANXA1sp induced autophagic flux (increased LC3II and reduced p62 expression) in the brain. Mechanistically, these findings were accompanied by upregulation of the mitochondrial protein deacetylase Sirtuin-3, and its downstream targets FOXO3a and MnSOD in ANXA1sp-treated animals. Our data provide new evidence that engaging pro-resolving pharmacological strategies such as Annexin-A1 biomimetic peptides can effectively attenuate neuroinflammation and enhance the neuroprotective effects of EPR after exsanguinating cardiac arrest.

Keywords: sirtuins, autophagy, apoptosis, neuroprotection, HMGB1

### INTRODUCTION

Exsanguinating hemorrhage, leading to cardiac arrest and multiple organ failure, remains the most common cause of death among trauma patients without traumatic brain injury, and neurologic outcomes in survivors are poor. An new therapeutic paradigm – termed EPR – involves rapidly cooling victims of exsanguinating cardiac arrest to deep or profound hypothermia levels (≤20◦C), in an effort to extend ischemic time, maintain organ viability during severe shock, allow operative repair of injuries and resuscitation, and ultimately improve survival and preserve neurological function (Tisherman et al., 2017). Following initial characterization of EPR, a growing body of proof-of-concept large animal studies have investigated the optimal depth and rate of cooling, duration of EPR, and physiological conditions prior to induction of EPR (Rhee et al., 2000; Alam et al., 2002, 2004, 2005, 2006a,b; Wu et al., 2006). Rapid cooling and subsequent controlled resuscitation and rewarming is accomplished through the use of CPB (Sailhamer et al., 2007; Alam et al., 2008). The development of rodent models of HS and EPR enabled characterization of key cellular and molecular changes involved in exsanguination cardiac arrest (Drabek et al., 2007a,b) – chief among those are neuroinflammation and apoptosis – which constitute targets for adjunctive therapies (Han et al., 2008; Alam et al., 2010).

Neuroinflammation is a complex immune response commonly observed following acute brain insults such as HS and associated warm ischemia, ischemia-reperfusion injury (I/RI), therapeutic hypothermia, exposure to CPB, and post-cardiac arrest syndrome. Multiple rodent models of surgical trauma, cardiac arrest and resuscitation have identified upregulation of pro-inflammatory cytokines and inflammatory mediators in both peripheral tissues and the central nervous system (CNS) (Terrando et al., 2010a, 2011). Acute neuroinflammation is characterized by increased cytokines, such as interleukin-1β (IL-1β), tumor necrosis factor-α (TNF-α), HMGB1 and their cognate receptors ultimately leading to neuron-glia dysfunction and blood–brain barrier impairments (Cao et al., 2010; Terrando et al., 2010b, 2011; He et al., 2012; Liu and McCullough, 2013; Wohleb et al., 2014; Xiang et al., 2016; Skvarc et al., 2018). Oxidative stress and mitochondrial dysfunction converge with the neuroinflammatory pathway, creating a positive feedback loop (Netto et al., 2018). Collectively, these mechanisms actively contribute to neuronal death and cognitive impairment. There is evidence that the neuroinflammatory milieu persists following surgical trauma, HS, I/RI, and resuscitation from cardiac arrest because is not efficiently controlled by endogenous anti-inflammatory mechanisms, thereby contributing to secondary injury, CNS dysfunction, and neuronal network hyperexcitability (Tahsili-Fahadan et al., 2018). Despite improvements in resuscitation science for victims of HS cardiac arrest, including EPR, total body cooling is technically and logistically challenging and the development of effective therapies to decrease morbidity and improve long-term neurologic outcomes following exsanguinating cardiac arrest remains a critical need.

Dysfunction of inflammation-resolving pathways have been described in experimental rodent trauma models, leading to exaggerated post-injury cognitive decline (Su et al., 2013). In recent years, a number of lipid mediators such as resolvins, lipoxins, and maresins have begun to receive attention as possible resolvers of neuroinflammation (Terrando et al., 2013; Vacas et al., 2013). Resolvins act to block neutrophil and monocyte migration and reduce the oxidative burst of neutrophils (Serhan et al., 2007; Perretti et al., 2015). Similarly, the glucocorticoid-regulated protein Annexin A1 and its peptide mimetics display pro-resolution effects, and have been shown to mitigate cerebral I/RI by attenuating neuroinflammation in many experimental models (Ansari et al., 2018) including cardiac surgery with deep hypothermic circulatory arrest (Zhang et al., 2017). The aim of this study was to determine the neuroprotective efficacy of a bioactive Annexin A1 short peptide (ANXA1sp) in a small animal model of exsanguination cardiac arrest and EPR. We hypothesized that ANXA1sp treatment would attenuate neuroinflammation and neuronal cell death, in part through upregulation of the mitochondrial protein deacetylase SIRT3 and downstream cytoprotective pathways.

**Abbreviations:** ANXA1sp, Annexin A1 short peptide; CPB, cardiopulmonary bypass; EPR, emergency preservation and resuscitation; HMGB1, high mobility group box 1; HS, hemorrhagic shock; LC3-II, microtubule-associated protein 1A/1B-light chain 3-phosphatidylethanolamine conjugate; MnSOD, manganese superoxide dismutase; p62/SQSTM1, sequestosome-1; ROS, reactive oxygen species; SIRT3, sirtuin-3.

### MATERIALS AND METHODS

fnins-13-00608 June 13, 2019 Time: 17:39 # 3

### Animals

The experimental protocol was approved by the Duke University Animal Care and Use Committee. All procedures met the guidelines of the National Institutes of Health for animal care (Guide for the Care and Use of Laboratory Animals, Health and Human Services, National Institute of Health Publication No. 86-23, revised 1996). Adult male Sprague–Dawley rats (age 13–15 weeks; weight 400–450 g; Charles River Laboratories, Wilmington, MA, United States) were housed (two animals per cage) in a 12-h light-dark cycle environment with free access to food and water. Rats were acclimated for at least 1 week before starting the experiment.

### Surgical Preparation for Experimental EPR

Experimental procedures are summarized in **Figure 1**. Fasted rats were anesthetized with isoflurane, intubated, and mechanically ventilated to maintain the arterial PaCO<sup>2</sup> between 35 and 45 mmHg. Anesthesia was provided with isoflurane (1.5–2%) with an oxygen and air mixture at a FiO<sup>2</sup> of 0.5. Bupivacaine 0.25% was injected at the operative sites to achieve local anesthesia. Routine physiologic parameters, and rectal and pericranial [hypodermic needle probe (Omega <sup>R</sup> , Stamford, CT, United States) placed beneath the temporalis muscle] temperatures were monitored. The right caudal epigastric artery was cannulated with a polyethylene catheter (PE10, Clay Adams, Sparks, MD, United States) and used for systemic mean arterial pressure (MAP) monitoring. The tail artery was cannulated with a 20-gauge catheter, which served as blood sampling and inflow CPB cannula. The right internal jugular vein was cannulated with a customized 4.5-F multi-orifice catheter advanced into the right atrium and used for withdrawing blood and venous out-flow CPB cannula. The external carotid artery was cannulated with a polyethylene catheter (PE50; Clay Adams, Sparks, MD, United States) advanced into the aortic arch and used for induction of profound hypothermia by flushing ice-cold normal saline. During surgical preparation, normothermia was maintained using a heat lamp.

### Induction of Hemorrhagic Shock and Ventricular Fibrillation

After a 30-min equilibration period, rapid exsanguination (50% blood volume, 0.034 ml/g body weight, 12–15 ml of blood) over 10 min was performed via the jugular venous cannula and MAP was allowed to drop below 20 mmHg. For the next 35 min (simulating pre-hospital transport time), the MAP remained <20 mm Hg without any resuscitative efforts HS. The shed blood was salvaged in a 20 ml syringe prefilled with 2 ml citrate phosphate dextrose solution (Fenwal Inc., Lake Zurich, IL, United States) and stored at 4◦C for later transfusion/resuscitation. At the end of HS, ventricular fibrillation (VF) was induced by transesophageal burst pacing at 50 V (S48 stimulator, RI, United States) for 90 s, and confirmed electrocardiographically and echocardiographically (Phillips Sonos7500 system, Andover, Mass 01810). During HS, anesthesia was maintained using 0.4% isoflurane. No neuromuscular blocker was administered until just prior to VF. A dose of vecuronium bromide (0.1 mg/kg) was given intravenously to prevent muscle contraction during VF and cooling.

### Induction of Profound Hypothermia

After the induction of VF, the animals were rapidly cooled to a core temperature of 10–15◦C for 45 min by flushing ice-cold solution [total volume 45 ml = 40 ml normal saline + 2 ml 5% human plasma protein fraction (Plasmanate <sup>R</sup> , Grifols Therapeutics, Inc., NC, United States) + 3 ml 8.5% Sodium Bicarbonate] into the ascending aorta/aortic arch and returning from right jugular vein using a peristaltic pump (Masterflex, model 770201-60, Cole-Parmer, Vernon Hills, IL, United States) at rate of 20 ml/kg/min (T208 transonic volume flow meter, Transonic Systems Inc. Ithaca, NY, United States). A water bath with ice water as well as topical cooling with ice bags was also applied.

### Resuscitation and Rewarming With Cardiopulmonary Bypass

Following 45 min of profound hypothermia, rats were rewarmed and resuscitated using CPB. The CPB circuit consisted of a peristaltic pump (MasterflexC, Cole-Parmer, Vernon Hills, IL, United States), a custom-designed oxygenator, and a venous reservoir, as previously described (Bartels et al., 2014; Shim et al., 2014; Zhang et al., 2017). The reservoir was primed with 3 ml of 6% hydroxyethylstarch 130/0.4 and 2 ml of previously shed and salvaged blood. Heparin (200 IU) and vecuronium bromide (0.1 mg/kg) were added to the venous reservoir. CPB was initiated at a flow rate of 20–30 ml/min, which was gradually increased to 50–60 ml/min upon reaching 33–34◦C. The gradient between the water bath and core body temperatures was not allowed to exceed 10◦C. Once core temperature increased from 10 to 34◦C, all salvaged blood was gradually re-transfused to keep up with the increased oxygen demand. The animals were rewarmed for 80 min until core temperatures of 34◦C were achieved; subsequently, CPB was terminated. Rewarming rate was maintained at approximately 0.4◦C/min. Acid-base abnormalities were corrected as needed. During rewarming, MAP was kept above 50 mmHg once core temperature reached >30◦C using intermittent administration of epinephrine.

After decannulation, animals were kept ventilated for 2 h under anesthesia with 0.5–1% isoflurane (core temperature 36–37◦C). To increase the hematocrit value to greater than 30%, the remaining autologous blood in the CPB circuit was collected, concentrated by centrifugation (3000 rpm for 5 min), and re-transfused. Heparin-induced anticoagulation was not reversed and allowed to dissipate spontaneously. Routine blood gas analysis was conducted serially (GEM Premier 3000 analyzer, Instrumentation Laboratory, Bedford, MA, United States). After spontaneous ventilation had resumed, animals were extubated and allowed to recover in an oxygen-enriched and humidified environment for 24 h, with free access to water and food.

To harvest the brain, 24 h after EPR rats were re-anesthetized, intubated, and mechanically ventilated. One sample of brain tissue was immediately fixed in 10% buffered formalin and paraffin-embedded for immunostaining. The remaining brain tissue was frozen in liquid nitrogen and stored at −80◦C until further use. Blood samples were also collected before and after induction of HS and at 0–120 min and 24 h after CPB and stored at −80◦C until analysis. Naïve rats were sacrificed under 5% isoflurane.

### Drug Treatments

fnins-13-00608 June 13, 2019 Time: 17:39 # 4

Annexin A1 biomimetic tripeptide (ANXA1sp or Ac-QAW, Ac = acetyl, MW = 445.47 Da) was synthesized and purified (>98% purity) by GenScript (Piscataway, NJ, United States). The peptide was suspended in 100% DMSO. For experiments, this stock solution was diluted in saline to a final dose of 1 mg/kg ANXA1sp and a concentration of 1% DMSO in saline as vehicle (control). ANXA1sp treatment solutions were freshly prepared immediately before use.

Rats were randomly assigned to three groups (EPR + ANXA1sp, n = 14; EPR+ vehicle, n = 15; naïve controls, n = 3) and terminated for histologic and biochemical analyses at 24 h after EPR. Rats received ANXA1sp (1 mg/kg iv) or vehicle (1% DMSO iv) in 1 ml saline immediately after induction of VF, then again at 1 h after CPB. Additionally, blood salvaged during exsanguination HS was also treated ex vivo with ANXA1sp 1 mg/kg or 1% DMSO and re-transfused during CPB rewarming and resuscitation (**Figure 1**). All treatments were administered in a blinded manner.

### Cell Death Assessment

Apoptosis was determined by terminal deoxynucleotidyl nick-end labeling (TUNEL) per assay manufacturer's protocol

(Roche Diagnostics, Indianapolis, IN, United States). Briefly, sections of paraffin-embedded brain tissue samples (5 µm thick) were deparaffinized using xylene and descending grades of ethanol, and pretreated with microwave radiation (350 W, in 200 mL of 0.1 M Citrate buffer, pH 6.0) for 5 min. Tissue sections were then incubated with terminal deoxynucleotidyl transferase (TdT) for 1.5 h at 37◦C and then rinsed with PBS. Slides of five representative areas of the retrosplenial and posterior parietal cortex and CA1 area of the hippocampus were mounted using UltraCruzTMMounting Medium with DAPI (Santa Cruz Biotechnology, Santa Cruz, CA, United States). Negative controls were incubated in label solution without TdT. A separate set of brain tissue sections was stained with acid fuchsin-celestine blue to identify possible necrotic cells. Cell counting was performed in a blinded manner across five representative areas of the cerebral cortex and CA1 areas using fluorescence microscopy (Leica DM IRB, Germany) with a 20×/0.4 PH objective at 1.5-fold magnification. Data obtained in every field were added together to make a final data count for each slide and expressed as percentage of total cell number within the relevant fields.

### Western Blots

Frozen brain samples were homogenized and protein quantified by BCA assay (Thermo Fisher Scientific). Western blotting was performed using SDS–PAGE 4–20 and 8–16% gels gradient gels (Bio-Rad) with the following antibodies: rabbit anti-HMGB1 (Bioss Antibodies Inc., Woburn, MA, United States), rabbit polyclonal antibodies against SIRT3 (28-kDa isoform, Cell Signaling and Abcam), FOXO3a (Cell Signaling), Mn-SOD (Santa Cruz), cleaved caspase-3 (Cell Signaling), Bax and Bcl-2 (Santa Cruz), rabbit anti-LC3B (cell signaling), mouse anti-p62/SQSTM1 (R&D Systems, Minneapolis, MN, United States), and rabbit monoclonal antibody against GAPDH (Cell Signaling Technology, Boston, MA, United States). The bands were detected by Super-signal West Dura Extended Duration Substrate (Thermo Scientific Fisher, Rockford, IL, United States). Band intensities of HMGB1, SIRT3, FOXO3a, Mn-SOD, cleaved caspase-3, Bax, Bcl-2, LC3II, p62/SQSTM1 were all normalized with a GAPDH loading control.

### Cytokine Measurements

Concentrations of IL-10, IL-6, and TNFα in brain homogenates were measured using rat-specific ELISA kits per manufacturer's protocol (Thermo Fisher Scientific, Grand Island, NY, United States). Brain homogenates were separated by centrifugation at 14,000 g for 10 min at 4◦C to remove cellular debris. In addition, left ventricular myocardial concentrations of IL-6 and TNFα were analyzed by ELISA. Change in absorbance in every well was detected at 450 nm on a microplate reader. All measurements were performed in triplicate.

### Confocal and Fluorescence Microscopy

After deparaffinization, brain tissue sections were treated with 10 mM citrate buffer (pH 6.0) for antigen retrieval. After blocking with 10% normal goat serum at RT for 1 h, the sections were incubated with rabbit anti-SIRT3 antibody (1:300) and mouse anti-COXIV (1:500, Santa Cruz Biotechnology, Santa Cruz, CA, United States), or rabbit anti-LC3B (1:400) and mouse anti-p62/SQSTM1 (1:400), respectively, at 4◦C overnight. The sections were then incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:500; Invitrogen, Carlsbad, CA, United States) and Alexa Fluor 550-conjugated goat anti-mouse IgG (1:500; Invitrogen, Carlsbad, CA, United States) at RT for 1 h. After washing with PBS, slides were prepared and mounted using UltraCruzTM Mounting Medium with DAPI (Santa Cruz Biotechnology, Santa Cruz, CA, United States) to detect nuclei. Images were captured on a Leica SP5 confocal microscope (Leica Microsystems, Germany) using a 40×/1.25–0.75 Plan APO oil objective, and the images were analyzed by NIH ImageJ software (version 1.51).

To assess biodistribution of systemically administered ANXA1sp, rats were anesthetized, intubated and mechanically ventilated, and the tail vein cannulated as described above. Fluorescein isothiocyanate (FITC)-conjugated ANXA1sp (FITC-QAW, MW = 905.48, >98% purity) was synthesized and purified by GenScript (Piscataway, NJ, United States) and injected via the tail vein (0.5 mg/kg). Brain tissue was harvested at 1 h after injection, fixed with 10% neutrally buffered formalin and embedded in paraffin. The 5 µm sections were deparaffinized and evaluated by fluorescence microscopy as above.

### Neurologic Evaluation

On postoperative day 1, rats underwent standardized functional neurologic testing by an observer blinded to group assignment, using an established neurologic scoring system that evaluates motor deficit (Homi et al., 2010). Briefly, rats were first placed on a 35 × 31 cm screen (grid size 0.6 × 0.6 cm) that could be rotated from horizontal (0◦ ) to vertical (90◦ ). The length of time that the rat could hold onto the screen after being rotated from 0 to 90◦ was recorded to a maximum of 15 s (0–3). Rats were then tested for balance on a horizontal wooden rod, and the time lapse before falling off the rod was recorded to a maximum of 30 s (0–3). Finally, rats underwent a prehensile traction test, and the length of time that the rat could cling to a horizontal rope was recorded to a maximum of 5 s (0–3). Animals received a score for each of the three tests. The final score was the sum of the individual test scores, with 0 the best score, and 9 the worst score.

### Statistical Analysis

Statistical analysis was performed using Prism 8 (GraphPad Software, San Diego, CA, United States). Results are expressed as mean ± standard deviation (SD). Parametric values, including physiologic values, data from western blots, ELISA were compared between groups using one-way analysis of variance (ANOVA) with post hoc Tukey's multiple comparison test, or a Student's t-test (equal variance not assumed), according to the characteristics of each experiment. Statistical significance was defined as p < 0.05.

## RESULTS

The 24 h survival rates were 75.0% (9/12) and 61.5% (8/13) for the ANXA1sp and vehicle treated groups, respectively. One rat

in vehicle group and three in ANXAsp1 group died from acute cardiac failure during the early postoperative phase. Additionally, three rats died of severe spinal cord injury and one of severe brain injury, all in the vehicle treated group. Two rats in each group died of technical failures and were excluded from analyses.

Baseline and intraoperative physiological parameters were similar between groups and are shown in **Figures 2A,B** and **Table 1**. Profound systemic hypotension (MAP < 20 mmHg) occurred during rapid exsanguination and cardiac arrest (**Figure 2A**) and was associated with severe metabolic acidosis (**Table 1**). All animals underwent the same cooling and rewarming protocol (**Figure 2B**). Rats exhibited spontaneous hypothermia (from 37◦C to 32◦C) before the onset of systemic cooling (shock-induced hypothermia). Following induction of hypothermia, pericranial temperature rapidly dropped to 23◦C at 5 min (rate of cooling 1.8◦C/min) and to 15◦C at 20 min (rate of cooling 0.85◦C/min). After 45 min of profound hypothermia (10–15◦C), animals were resuscitated and slowly rewarmed

using CPB to 34–35◦C over 80 min (rate of rewarming 0.4◦C/min).

### Systemically Administered ANXA1sp Traverses the Blood-Brain Barrier and Attenuates Neuroinflammation After EPR

Relative abundance of HMBG1, a key initiator of neuroinflammation, was assessed by Western analysis in brain homogenates. We found a 10% (p < 0.05) attenuation in cerebral expression of HMGB1 in ANXA1sp compared to HS-EPR vehicle treated animals (**Figures 3A,B**). Moreover, brain levels of the pro-inflammatory cytokines IL-6 and TNFα were reduced by 55% (p < 0.01) and 27% (p = 0.059), respectively, with ANXA1sp treatment (**Figures 3C,D**). Conversely, brain levels of the anti-inflammatory cytokine IL-10 were increased by 25% (p < 0.05) in ANXA1sp-treated compared to vehicle controls at 24 h after EPR (**Figure 3E**).

Fluorescence-conjugated peptide (FITC-QAW) was detectable in all areas of brain parenchyma at 1 h following intravenous injection (**Supplementary Figure 1**).

### Regulation of Cell Death by ANXA1sp After EPR

Acid fuchsin-celestine blue staining revealed acidophilic neurons and possible necrosis in the cortex at 24 h after EPR in vehicle-treated animals, which were reduced by 72% (p < 0.05) following ANXA1sp treatment (**Figures 4A,B**). ANXA1sp treatment was also associated with a 62% (p < 0.05) reduction in TUNEL-positive cells in the cerebral cortex, but not in the hippocampus, at 24 h after EPR (**Figures 4C,D**). This was corroborated by a 56% (p < 0.05) reduction in cerebral expression of cleaved caspase-3 in ANXA1sp-treated animals (**Figures 4E,F**).

We used Western analysis of brain homogenates to assess changes in ratios of death and survival factors. At 24 h, a modest but significant (10%, p < 0.05) increase in the relative abundance of the anti-cell death protein Bcl-2 was observed in ANXA1sp compared to vehicle-treated animals after HS-EPR (**Figures 5A,B**). Conversely, a marked 5-fold (p < 0.01) increase in expression of the pro-cell death protein Bax was seen after HS-EPR in vehicle-treated animals, with a significant 33% (p < 0.05) reduction following ANXA1sp treatment (**Figures 5A,C**). Consequently, the near 5-fold surge in Bax:Bcl-2 ratio detected after HS-EPR in vehicle-treated compared to naïve animals (p < 0.001) was attenuated by 38% (p < 0.01) with ANXA1sp treatment (**Figure 5D**).

### Modulation of SIRT3 Expression and SIRT3 Pathway by ANXA1sp

Western blot analysis in brain homogenates revealed a 33% (p < 0.01) reduction in relative abundance of the primary mitochondrial deacetylase SIRT3 (28 kDa isoform) in vehicle-treated HS-EPR compared to naïve animals, which was partially restored with ANXA1sp treatment (increased by 25%, p < 0.05) (**Figures 6A,B**). Increased SIRT3 expression with ANXA1sp treatment in both cortical and hippocampal


Results are shown as Mean ± SD; <sup>∗</sup>p < 0.05 between groups. ANXA1sp, Annexin A1 short peptide; Hct, hematocrit; BE, base excess; End of HS, at the end of 45 min of hemorrhagic shock; End of cooling, at the end of 45 min cooling; End of CPB, at the end of cardiopulmonary bypass.

cells was further confirmed by confocal microscopy, with SIRT3 immunoreactivity colocalizing with the mitochondrial protein COXIV (**Figure 6E**). We further measured expression of FOXO3a and MnSOD, two downstream targets of SIRT3. EPR resulted in an 87% (p < 0.01) reduction in relative abundance of the transcription factor FOXO3a and a 42% (p < 0.001) reduction its regulated antioxidant enzyme MnSOD compared to naïve animals, and were both partially restored with ANXA1sp treatment (a 4-fold increase in FOXO3a expression, p < 0.05; and a 18% increase in MnSOD expression, p < 0.05) (**Figures 6A,C,D**).

### ANXA1sp Promotes Autophagy Following EPR

Western blot was performed to detect the expression of autophagy associated proteins in brain homogenates. Compared to vehicle, ANXA1sp treatment was associated with a 35% (p < 0.01) increased expression of LC3II and a 25% (p < 0.05) reduced expression of p62/SQSTM1 (**Figures 7A,B,C**). In addition, we detected expression of LC3B by immunohistochemistry, with LC3B and p62 colocalizing in cortical cells (**Figure 7D**).

### Neurologic Outcome After EPR and ANXA1sp Treatment

Finally, we evaluated neurologic changes at 24 h after EPR and ANXA1sp treatment. Compared to vehicle controls, ANXA1sp-treated animals had significantly lower neurologic severity scores, showing improved sensory-motor functions (including processing involving retrosplenial and posterior parietal cortex, **Figure 4G**).

### Systemically Administered ANXA1sp Also Attenuates Myocardial Inflammation

To evaluate the effects of ANXA1sp in organs that are not shielded by the blood-brain barrier, we assessed pro-inflammatory cytokines in left ventricular myocardial homogenates between groups. At 24 h after EPR, heart levels of IL-6 and TNFα were reduced in ANXA1sp-treated animals by 10% (p < 0.05) and 17% (p = 0.054), respectively (**Supplementary Figure 2**).

### DISCUSSION

In this study, we used a novel experimental model of acute neuroinflammation triggered by exsanguinating cardiac arrest and treated by emergent preservation and resuscitation (EPR) to test the neuroprotective effects and mechanisms of ANXA1sp administration. Together with our previous findings that ANXA1sp exerts pro-resolving effects in a rat model of CPB with global ischemia-reperfusion via deep hypothermic circulatory arrest (Zhang et al., 2017), we continue to illustrate how this small peptide impacts neuroinflammation. Here, we show that ANXA1sp treatment is superior to induction of profound hypothermia (EPR) alone by increasing cortical cell viability, promoting a favorable expression of pro-survival (Bcl-2) versus pro-death (Bax) factors in the brain, reducing caspase-3 activation, and overall improving neurological performance at 24 h after EPR.

High mobility group box 1 is a key endogenous danger associated molecular pattern which acts as a mediator of neuroinflammation resulting from a variety of conditions such as cerebral ischemia-reperfusion (Kim et al., 2006; Yang et al., 2011), septic shock, and traumatic brain injury (Parker et al., 2017). HMGB1 is actively released by neurons and glial cells upon inflammasome activation, and in turn activates two pattern recognition receptors on target cells (TLR4 and RAGE) (Gao et al., 2012), leading to NF-κB mediated production of pro-inflammatory cytokines (Paudel et al., 2018). HMGB1 also plays a pivotal role in BBB disruption (He et al., 2012; Festoff et al., 2016; Yang et al., 2018), either directly via cytokine mediated activation of metalloproteinase or via disruption of tight junctions (Gloor et al., 2001; Utech et al., 2010), although the precise mechanism remains elusive. Nonetheless, altered BBB permeability is known to then amplify neuroinflammation and neuronal excitation (Frank et al., 2015, 2016). We provide primary evidence that HS and exsanguinating cardiac arrest treated with induction of profound hypothermia and EPR

elicit neuroinflammation. Further, we show ANXA1sp treatment to favorably modulate the neuroinflammatory response, by attenuating the HS-EPR induced increase in cerebral levels of the proinflammatory alarmin HMGB1 and altering the balance of pro-inflammatory (IL-6 and TNFα) and anti-inflammatory cytokines (IL-10) (**Figures 3A–E**). This is consistent with the pleiotropic immunomodulatory functions of IL-10, to polarize the inflammatory system toward an anti-inflammatory phenotype (Ip et al., 2017), aiding in the resolution of neuroinflammation (Garcia et al., 2017). The pro-resolving effects were accompanied by improved cell survival (**Figures 4A–E**) and neurological function (**Figure 4G**). This study builds on our previous findings demonstrating beneficial effects of ANXA1sp on neuroinflammation, microglial activation, NF-κB activation

arrest/EPR and ANXA1sp or vehicle treatment. (A,B) Necrosis, detected by acid fuchsin-celestine blue, was reduced in the cerebral cortex of ANXA1sp treated rats. Arrowheads in panel A identify acidophilic neurons in the cortex and hippocampus (Hpc). (C,D) Terminal deoxynucleotidyl nick-end labeling (TUNEL) staining and quantification of apoptotic cells. Apoptosis was significantly reduced in the cerebral cortex of ANXA1sp-treated rats on postoperative day 1. Arrowheads in panel (C) identify pyknotic positive cells. Scale bar: 20 µm. (E,F) Western blot analysis of cleaved caspase-3 revealed significantly reduced levels in brain homogenates from ANXA1sp-treated animals compared to vehicle (controls). (G) Neurologic score (sensory-motor function) was significantly improved at postoperative day 1 in ANXA1sp rats. Data presented as mean ± SD (n = 5–7/group), <sup>∗</sup>P < 0.05 and ∗∗P < 0.01 compared to vehicle (controls) analyzed by unpaired t-test (B,D,F) or Mann–Whitney U-test (G).

and postoperative neurocognitive performance following global I/RI associated with deep hypothermic circulatory arrest (Zhang et al., 2017). While our experimental model is complex, sequentially involving HS, warm cardiac arrest, induction of profound hypothermia followed by rewarming and resuscitation, the protective effects of ANXA1sp recapitulate those seen in other experimental models of cerebral I/RI (Relton et al., 1991; Gavins et al., 2007; Smith et al., 2015; Vital et al., 2016).

Systemically administered ANXA1sp was also associated with anti-inflammatory effects in the heart (reduced levels of IL-6 and TNFα) following HS-EPR. Consistent with previous reports in experimental models of regional myocardial I/RI (La et al., 2001; Qin et al., 2015), these results support the broader therapeutic roles attributed to ANXA1 and its peptide mimetics in reducing systemic inflammation and conferring organ protection from a variety of insults involving IR/I.

To better understand the added neuroprotective effects and mechanisms of action of ANXA1sp in the setting of HS-EPR with

profound hypothermia, we focused on SIRT3, a mitochondrial protein deacetylase known to be widely and abundantly expressed in most cell types in the CNS (Kim et al., 2011) and differentially regulated in brain regions and across stages of development in the rat (Sidorova-Darmos et al., 2014). SIRT3 is part of the silent information regulator of transcription (SIRT) family of NAD+-dependent protein deacetylases, which have emerged as key regulators of metabolic and energy stress-response pathways including neuroinflammation (Someya et al., 2010; Lee et al., 2013; Hattori et al., 2015; Liu L. et al., 2015; Yuan et al., 2016), but also systemic inflammation (Preyat and Leo, 2013; Vachharajani et al., 2014; Liu T. F. et al., 2015; Traba et al., 2015). Notably, SIRT3 exerts neuroprotective effects. For instance, SIRT3 overexpression rescues mutant SOD1-induced neuronal cell death (Song et al., 2013), increases neuronal lifespan under mitochondrial oxidative stress (Weir et al., 2012), protects against excitotoxic injury (Kim et al., 2011), mediates adaptive neuronal responses (resistance to oxidative stress, apoptotic cell death) to in vitro bioenergetic, oxidative and excitatory stress (Cheng et al., 2016), and protects mice against noise-induced hearing loss in vivo (Someya et al., 2010). Specifically, SIRT3 has been shown to have a major involvement in CNS ROS metabolism by regulating FOXO3a transcription factor and contributing to transactivation of key antioxidant enzymes including MnSOD and catalase (Rangarajan et al., 2015). Together, even though results from rodent studies are not completely consistent (Novgorodov et al., 2016), most data

regulated proteins after exsanguination cardiac arrest/EPR and ANXA1sp or vehicle treatment. (A,B) A significant reduction in SIRT3 (28 kDa short, processed, active form) was seen after EPR + vehicle treatment compared to naïve animals, and was rescued in ANXA1sp-treated animals. This was accompanied by similar expression changes for the forkhead box O3a (FOXO3a) transcription factor (C), known to transactivate the antioxidant enzyme manganese superoxide dismutase (MnSOD) (D). GAPDH was used as loading control (E). At 24 h postoperatively, colocalization of sirtuin-3 (SIRT3) with the mitochondrial protein cytochrome c oxidase complex IV (COXIV) in cortical and hippocampal slides from ANXA1sp-treated animals was visualized via double immunofluorescence staining and confocal microscopy. Data presented as mean ± SD (n = 3–7/group), <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 compared to either naïve or vehicle controls analyzed by one-way ANOVA with Tukey's multiple comparisons test.

suggest that SIRT3 expression is vital for neuronal survival and suppression of neuroinflammation following physical stressors. In this study, we noted that relative abundance of SIRT3 – the 28 kDa processed active isoform, known to possess deacetylase activity (Schwer et al., 2002; Sundaresan et al., 2008) – as well as FOXO3a and MnSOD, were all reduced in brain homogenates following HS-EPR, consistent with previous reports in experimental models of cerebral I/R (Zhao et al., 2018), but were rescued with systemic administration of ANXA1sp (**Figure 6**). HS, ischemia-reperfusion, and EPR incite neuroinflammation, and SIRT3-mediated upregulation of FOXO3a-dependent antioxidant gene expression could be

implicated in the attenuation of ROS neurotoxicity associated with acute neuroinflammation and activation of microglia. This may translate into attenuation of secondary brain injury and decreased cerebral edema. Along with downregulation of pro-apoptotic proteins, these findings suggest that ANXA1sp treatment offers an element of neuroprotection.

Modulation of protein acetylation is emerging as a therapeutic strategy to create a pro-survival and anti-inflammatory phenotype in various shock states, including lethal HS (Li and Alam, 2011). Intriguingly, promoting histone acetylation via administration of the histone deacetylase (HDAC) inhibitor valproic acid has been shown to decrease neuronal apoptosis, neuroinflammation, and brain lesion size while improving neural plasticity and resulting in faster neurocognitive and neurologic recovery in pre-clinical models of traumatic brain injury, HS, and polytrauma (Nikolian et al., 2017, 2018; Chang et al., 2019). An imbalance in histone acetyltransferase/deacetylase activity has been reported in both HS (Lin et al., 2006) and neurodegeneration (Rouaux et al., 2003). Valproic acid administration results in epigenetic and posttranslational modifications that induce differential gene expression (Dekker et al., 2014; Bambakidis et al., 2016), metabolic changes (Hwabejire et al., 2013) and alterations of the proteome. It appears that promoting acetylation of histones and other target transcription factors through the use of HDAC inhibitors (like valproic acid), as well as promoting mitochondrial protein deacetylation via induction of SIRT3 (in this study using ANXA1sp) can both exert neuroprotective effects in the setting of lethal HS. The precise mechanisms by which modulation of protein acetylation results in attenuation or resolution of neuroinflammation remain to be elucidated.

We found that ANXA1sp treatment promoted cerebral autophagy following EPR. This was accompanied by increased SIRT3 protein expression, reduced cell apoptosis and improved neuro severity scores. Importantly, upregulation of SIRT3 in response to ANXA1sp treatment was associated with increased expression of LC3II and a reduced level of p62/SQSTM1, indicating increased autophagic flux (**Figure 7**). These results are consistent with previous reports implicating SIRT3 as an activator of autophagy in neuronal cells (Dai et al., 2017; Yan et al., 2018) and macrophages (Liu et al., 2018) under various stress conditions in vitro, and in vivo studies showing sirtuin activation to be protective in spinal cord injury, brain trauma, and cerebral ischemia (He et al., 2017; Zhao et al., 2017). A schematic representation of the relationships between ANXA1sp, the SIRT3 pathway and neuroinflammation in the context of cerebral I/RI is shown in **Figure 8**.

Although the therapeutic effects of hypothermia in HS in general, and those of EPR for exsanguinating cardiac arrest in particular, have been previously reported in large animal experiments (Alam et al., 2002, 2004, 2005, 2006a,b; Chen et al., 2005; Wu et al., 2006), benefits of the clinically realistic rat model described in this study include lower costs to test efficacy of selective pharmacological neuroprotective interventions (such as the pro-resolving ANXA1sp) in enhancing clinical outcomes after EPR, as well as availability of molecular biology tools to understand the mechanisms at play. To increase clinical relevance, we modified previously described rodent models of EPR (Drabek et al., 2007a,b) in several ways: first, we introduced a prolonged period of HS followed by induction of fibrillatory cardiac arrest; second, we rapidly induced profound hypothermia via aortic flushing of a relatively low volume (1.5–2 fold the estimated blood volume) of ice-cold saline-based solution using a light weight cycling system at 20 ml/min; finally, we re-transfused salvaged autologous blood during resuscitation, thereby avoiding additional confounding effects from exposure to allogeneic blood transfusions. In this severe model, administration of a pro-resolving ANXA1 bioactive peptide during EPR was superior to EPR alone in attenuating neuroinflammation, increasing pro-survival protein expression and autophagic flux.

Several limitations of this study follow. HS was achieved by closely controlled exsanguination of blood, which was salvaged and re-transfused. Although we designed the model to closely mimic the clinical scenario, HS inevitably requires surgical bleeding control and transfusion of allogeneic blood products, which would significantly amplify the inflammatory response. Rats serve as imperfect surrogates for human subjects, and only male young adult animals were used. An injured but un-resuscitated control group was not used, as the survival without resuscitation is extremely poor. Instead, we used a naïve uninjured control group, to allow assessing whether ANXA1sp effects are restorative or de novo. We have not measured directly markers of resolution of inflammation (neutrophil apoptosis/efferocytosis, functional macrophage switches), or microglial activation. However, microglial activation was attenuated by ANXA1sp treatment in our previously reported rat deep hypothermic circulatory arrest study (Zhang et al., 2017). This study focuses on ANXA1sp-induced upregulation of the SIRT3 pathway, and thus we have not explored the traditional mechanisms implicated in the pro-resolving effects of Annexin A1 and its bioactive peptides (via the FPR2/ALX receptor pathway). The neuroinflammatory response to hypothermic cardiac arrest is not uniform, displaying important spatio-temporal differences between brain regions (Drabek et al., 2015). Our investigation focused on the cortex and hippocampus as two vulnerable brain regions, but other areas displaying robust early neuroinflammatory responses (e.g., the striatum) were not studied. Temporally, our observation period was limited to 24 h, and longer-term evaluations are required to fully assess ANXA1sp's protective effects, including neurocognitive tests. Although previous reports from experimental focal cerebral I/RI in the rat suggest neutrophil accumulation in the infarct area (which correlates closely with cortical lesion size) to be maximal at 24 h post-reperfusion (Zhang et al., 1994), it should be noted that our results may represent a delay rather than elimination of neuronal death in the selectively vulnerable brain regions studied.

## CONCLUSION

Using a clinically relevant rodent model of lethal HS and EPR-induced acute neuroinflammation, we show that administration of a pro-resolving ANXA1 peptide mimetic results in a pro-survival phenotype with attenuated cortical cell death, reduction of several common functional biomarkers of neuroinflammation (HMGB1, IL-6, and TNFα), and improved early neurological outcomes, compared to induction of profound hypothermia (EPR) alone. The protective effects were associated with increased expression of SIRT3 and corresponding upregulation of the FOXO3a-MnSOD antioxidant pathway. In addition, we note increased cerebral autophagy in ANXA1sp treated animals. Further studies need to refine the mechanisms involved and evaluate the use of SIRT3-activating drugs to resolve acute neuroinflammation and attenuate the uncontrolled production of ROS by activated microglia, as well as the long-term benefits of the proposed therapeutics.

### DATA AVAILABILITY

fnins-13-00608 June 13, 2019 Time: 17:39 # 12

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the National Institutes of Health for animal care (Guide for the Care and Use of Laboratory Animals, Health and Human Services, National Institute of Health Publication No. 86-23, revised 1996). The protocol was approved by the Duke University Animal Care and Use Committee.

### AUTHOR CONTRIBUTIONS

ZZ, QM, J-KS, and MP designed and performed the research. TV, CL, JM, and NT contributed new reagents and analytic tools. ZZ,

### REFERENCES


QM, QQ, and MP analyzed the data. QM, ZZ, and MP wrote the manuscript. QQ, JM, and NT provided the critical edits to the manuscript. All authors read and approved the final draft of the manuscript.

### FUNDING

This work was supported by grants from the American Heart Association (AHA)-11BGIA (ZZ), 2009 SCA-IARS Mid-Career Grant (QM), NIH R01 HL092071 (MP), and the Department of Anesthesiology, Duke University Medical Center.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00608/full#supplementary-material



preservation and resuscitation for cardiac arrest from trauma clinical trial. J. Trauma Acute Care Surg. 83, 803–809. doi: 10.1097/TA.000000000000 1585


fnins-13-00608 June 13, 2019 Time: 17:39 # 14

pathway. Neuroscience 348, 241–251. doi: 10.1016/j.neuroscience.2017. 02.027

Zhao, H., Luo, Y., Chen, L., Zhang, Z., Shen, C., Li, Y., et al. (2018). Sirt3 inhibits cerebral ischemia-reperfusion injury through normalizing Wnt/β-catenin pathway and blocking mitochondrial fission. Cell Stress Chaperones 23, 1079–1092. doi: 10.1007/s12192-018- 0917-y

**Conflict of Interest Statement:** ZZ, QM, NT, and MP are co-inventors on patents for the use of Annexin A1 peptides for activation of sirtuins and to attenuate neuroinflammation. NT is Associate Editor for Frontiers in Immunology.

The remaining 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 © 2019 Ma, Zhang, Shim, Venkatraman, Lascola, Quinones, Mathew, Terrando and Podgoreanu. 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.

# Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-**κ**B Signaling

*Fei Zhang†, Yuan Zhang†, Ting Yang†, Ze-Qing Ye, Jing Tian, Hai-Rong Fang, Juan-Juan Han, Zhe-Zhi Wang\* and Xing Li\**

### *Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Jose Martinez-Navio, University of Miami Health System, United States Narendra Prasad Singh, University of South Carolina, United States*

#### *\*Correspondence:*

*Zhe-Zhi Wang zzwang@snnu.edu.cn Xing Li xingli\_xian@126.com*

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

#### *Specialty section:*

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

*Received: 27 February 2019 Accepted: 08 July 2019 Published: 02 August 2019*

#### *Citation:*

*Zhang F, Zhang Y, Yang T, Ye Z-Q, Tian J, Fang H-R, Han J-J, Wang Z-Z and Li X (2019) Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-κB Signaling. Front. Pharmacol. 10:863. doi: 10.3389/fphar.2019.00863*

*National Engineering Laboratory for Resource Development of Endangered Crude Drugs in Northwest China, The Key Laboratory of Medicinal Resources and Natural Pharmaceutical Chemistry, The Ministry of Education, College of Life Sciences, Shaanxi Normal University, Xi'an, China*

Scopoletin, a phenolic coumarin derived from many medical or edible plants, is involved in various pharmacological functions. In the present study, we showed that Scopoletin effectively ameliorated experimental autoimmune encephalomyelitis (EAE), an animal model of multiple sclerosis (MS), through novel regulatory mechanisms involving inhibition of NF-κB activity in dendritic cells (DCs). Scopoletin treatment significantly improved the severity of the disease and prominently decreased inflammation and demyelination of central nervous system (CNS) in EAE mice. Disease alleviation correlated with the downregulation of major histocompatibility complex (MHC) class II, CD80 and CD86, expressed on DCs of CNS or spleens, and the infiltration and polarization of encephalitogenic Th1/Th17 cells. Consistent with the *in vivo* data, Scopoletin-treated, bone marrow-derived dendritic cells (BM-DCs) exhibited reduced expression of MHC class II and costimulatory molecules (e.g., CD80 and CD86) and reduced NF-κB phosphorylation. These findings, for the first time, demonstrated the ability of Scopoletin to impair DC activation, downregulating pathogenic Th1/Th17 inflammatory cell responses and, eventually, reducing EAE severity. Our study demonstrates new evidence that natural products derived from medical or edible plants, such as Scopoletin, will be valuable in developing a novel therapeutic agent for MS in the future.

Keywords: scopoletin, experimental autoimmune encephalomyelitis, multiple sclerosis, dendritic cells, NF-**κ**B signaling

### INTRODUCTION

Multiple sclerosis (MS) is a T-cell-mediated chronic autoimmune disease featured by neuroinflammation and demyelination in the central nervous system (CNS), affecting 2.5 million people worldwide, and the incidence continues to increase (Hemmer et al., 2015). Experimental autoimmune encephalomyelitis (EAE) is a well-known animal model of MS that is used to study the underlying mechanism and that provides the theoretical basis for developing new therapies of MS (Procaccini et al., 2015; Lassmann and Bradl, 2016).

Dendritic cells (DCs) are antigen-presenting cells that connect the innate and adaptive immune response. Depending on the variety activation status, DCs can either activate or modulate naive T lymphocytes. When DCs are activated, costimulatory molecules are upregulated and produce cytokines that drive T-cell priming and effector differentiation (Amodio and Gregori, 2012). In the absence of activation, antigen presentation by steady-state DCs can result to T-cell unresponsiveness and tolerance (Besusso et al., 2015). Therefore, modulation of DCs has been shown to be a promising area for the treatment of autoimmune disease. For instance, it was proved that DCs treated with natural compounds or synthetic drugs blocked the inflammatory signal and halted disease development (Ginwala et al., 2016; Kornberg et al., 2018). DC-induced suppression is mainly dependent on inhibiting Th17 cell differentiation or promoting Treg cell generation (Thome et al., 2014a; Thome et al., 2018). It is known that naive CD4+ T cells can develop into various effector subsets containing Th1, Th2, Th17, and Treg cells, etc.; among these, Th17 cells are crucially related to the pathogenesis of MS/EAE (He et al., 2017; Malik et al., 2017). Suppression of the cellular response, Th17 cells in particular, can therefore act as a therapeutic target for controlling autoimmune and inflammatory diseases.

NF-κB is a broad transcription factor which can regulate various biological functions containing inflammation, immunity, cell growth, and survival, and its signaling pathway plays a role in DC activation and maturation (Ardeshna et al., 2000). Phosphorylation of NF-κB in innate immune cells induces inflammation reaction, cytokine secretion, and tissue damage, thus promoting to the progress of autoimmune diseases including EAE (Lu et al., 2013). In addition, immature DCs respond to danger signals or antigen and, when mature, produce abundance of proinflammatory cytokines, such as type I interferons (IFN). Subsequently, mature DCs present antigens to T lymphocytes to stimulate a sustained immune response, given the interferon regulatory factor (IRF) family's important role in DC development and maturation. For example, IRF-7 is necessary for producing IFN-α and IFN-β, and it serves as the prime regulator of type-I interferon-dependent immune responses (Honda et al., 2005; Gabriele and Ozato, 2007; Yasuda et al., 2007).

Scopoletin is a phenolic coumarin extracted from many medical plants, including *Erycibe obtusifolia*, *Aster tataricus*, *Foeniculum vulgare*, and *Artemisia iwayomogi*, as well as some edible plants, such as *Lycium barbarum* and *Morinda citrifolia* (Dou et al., 2013; Lee et al., 2014; Forino et al., 2016; Shalan et al., 2016)*.*  Scopoletin has been shown to possess anti-inflammatory, antioxidant, antidepressant, hypouricemic, and neuroprotective effects (Ding et al., 2005; Basu et al., 2016). Moreover, a recent report suggests that Scopoletin is effective in an adjuvant-induced arthritis rat model, increasing the possibility of its application as a therapeutic agent for autoimmune diseases (Pan et al., 2010). However, little is known about the regulatory properties and underlying mechanism of Scopoletin relative to innate immune cells such as DCs and its effect on the pathogenic progression of MS and EAE. In light of the pharmacological profile of Scopoletin and its reported immunoregulatory properties, we hypothesized that Scopoletin might have a therapeutic effect on the animal model of MS and EAE. The purpose of this study was, therefore, to test the therapeutic activity of Scopoletin on EAE and to elucidate the underlying therapeutic mechanisms of its action, as part of our ongoing search for development of immunomodulation medicine extracted from medical plants or edible plants (Zhao et al., 2018).

### MATERIALS AND METHODS

### Reagents and Animals

Scopoletin was purchased from Sigma (Sigma-Aldrich, St. Louis, MO). C57BL/6 mice, female, 8 weeks old, were obtained from the Fourth Military University (Xi'an, China). Studies were carried out by a protocol approved by the Animal Ethics Experimental Committee at of Shaanxi Normal University, and all experiments were operated strictly according to approved institutional guidelines and regulations.

### EAE Induction and Treatment

EAE was induced by subcutaneously injecting an emulsion containing 200 μg Myelin Oligodendrocyte Glycoprotein35–55 (MOG35–55) (Genescript, Piscataway, NJ), complete Freund's adjuvant (CFA) (Sigma-Aldrich), and 5 mg/ml *Mycobacterium tuberculosis* H37Ra (Difco Laboratories, Lawrence, KS), and intraperitoneally (i.p.) injected 200 ng pertussis toxin (Sigma-Aldrich) in PBS on day 0 and 2 postimmunization (p.i.) (Zhang et al., 2015). Disease progression of EAE mouse was evaluated daily in a blind way by two researchers, scoring standard for EAE disease progression as follows: 0, no symptoms; 0.5, stiff tail; 1, limp tail; 1.5, limp tail and waddle; 2, paralysis of one limb; 2.5, paralysis of one limb and weakness of one other limb; 3, complete paralysis of two hind limbs; 4, moribund; and 5, death. (Li et al., 2016). Scopoletin was prepared in dimethylsulfoxide (DMSO) at 100 mg/ml for stock, which was further diluted 20 times with PBS as a working solution. DMSO (5%) was dissolved in PBS as vehicle. Vehicle or Scopoletin (50 mg/kg) was injected i.p. daily from the day 0 p.i.

### Histopathology

Mice were killed at day 20 postimmunization (p.i.). To assess CNS histopathological, mice were perfused through the heart with cold PBS; spinal cord was then fixed with 4% paraformaldehyde, cut into 7-μm sections, and stained with hematoxylin and eosin (H&E) or Luxol fast blue (LFB) for evaluation of inflammation or demyelination, respectively. Slides were evaluated in a blinded fashion for inflammation using a 0–3 scale as description, and two investigators selected 10 areas in the white matter of the spinal cord and scored for slides, evaluation methods, and criteria as description followed our previous studies (Li et al., 2016). The white matter area was manually outlined, and the Image-Pro Plus software was used to calculated areas (%) of demyelination for evaluation of demyelination (Li et al., 2017). Demyelination area percentage was quantified as the area without LFB staining (white) in white matter divided by the area white matter of spinal cord slices. Fixed spinal cords were embedded in optimum cutting temperature (OCT) solution (Tissue-Tek, Sakura Finetek, Japan) and then sectioned coronally cut into 12 μm. Finally, sections stained with primary and secondary antibodies. Results were acquired by Nikon Eclipse E600 fluorescent microscopy (Melville, NY).

### Mononuclear Cells (MNCs) and BMDCs Preparation

Spleens were mechanically grinded and filtrated with a 100 µm cell strainer (Falcon, Tewksbury, MA) to harvest splenic cells, subsequently incubated with red blood cell lysis buffer (Biolegend, San Diego, CA) for 60 s. After washing cells with cold PBS, cells were stimulated with MOG35–55 *in vitro*. To acquire CNS infiltrating cells, the protocol was followed as previously described (Li et al., 2017). To acquire the bonemarrow-derived dendritic cells (BM-DCs), femurs and tibias were separated from the naive adult C57BL/6 mice, and the cells were blown out of the bone marrow with a syringe full of the precooled PBS. After filtering with 100-µm cell strainer, the harvested cells were subsequently cultured with PRMI1640 containing granulocyte colony-stimulating factor (GM-CSF) cytokine (10 ng/ml) until the ninth day to obtain mature BM-DCs. Dendritic cells are cultured in medium without GM-CSF cytokine after ninth day.

### Drug Preparation for Cell Experiments

Stock solution of Scopoletin was dissolved into the concentration of 30 mM for cell experiments with DMSO.

### T-cell Proliferation Assay

Splenic CD4+ T cells were isolated from the naive adult C57BL/6 mice using a CD4+ T Cell Isolation Kit (Miltenyi Biotec) and labeled with CFSE. Both BM-DCs and T cells were cocultured at 1:10 ratio in the presence of 10 μg/ml of MOG35–55 (Genescript, Piscataway, NJ) and 100 ng/ml of lipopolysaccharide (LPS) (Sigma-Aldrich). T-cell proliferation was measured by analyzing CFSE intensity by flow cytometry after 72 h.

### ENZYME LINKED IMMUNOSORBENT ASSAY (ELISA)

DCs culture medium or supernatants from splenocytes were harvested at 18 or 72 h, and sandwich ELISA was performed to determine the concentrations of cytokine production by ELISA kits (R&D Systems, Minneapolis, MN). according to the manufacturer's instructions, including IFN-γ, IL-17, GM-CSF, IL-1β, IL-6, and IL-23. Briefly, the capture antibody was used to precoat overnight. Then, the plate was washed with wash buffer (R&D Systems, Minneapolis, MN). for three times. Blocking with 1% BSA in PBS for 1 h, gradient standard, and samples were added to the plate for 2 h incubation. Next, the samples and standard were removed clearly, and the detection antibody was incubated for 2 h. The wells were again washed, a TMB substrate solution was added to the wells, and color developed in proportion to the amounts of cytokines. The Stop Solution changed color from blue to yellow, and the intensity of the color was measured at 450 nm. A standard curve was run for each microwell plate. According to the standard curve, the concentration of cytokine was determined.

### Flow Cytometry Analysis

For surface-marker staining, cells were washed with PBS and incubated 30 min at 4°C with anti-CD4, anti-CD8, anti-CD11b, anti-CD11c, anti-CD80, and anti-CD86 (BD Biosciences, San Jose, CA) or isotype control Abs according to the manufacturer's instructions to dilute. With 25 or 10 μg/ml MOG35–55 to stimulate splenocytes or infiltrating MNCs in CNS for 72 h or overnight to analyze Th1 and Th17 response to MOG35–55. For intracellular staining, cells were stimulated with 50 ng/ml PMA, 500 ng/ml ionomycin, and GolgiPlug for 5 h. For blocking the nonspecific staining, 1 μg Fc Block reagent for the FcγII/III receptor (BD Biosciences, San Jose, CA) was addded to 1 × 106 cells in 100 μl staining buffer (BD Biosciences) and incubated for 15 min at 4°C. The cells are then washed. For surface staining, cells were incubated with anti-CD4, anti-CD8, anti-CD11b, anti-CD11c, anti-MHC class II, anti-CD80, and anti-CD86. Then, cells were fixed and permeabilized by Fix & Perm Medium (Invitrogen, Waltham, MA), using Abs anti-IL-17, anti-IFN-γ, or anti-GM-CSF (BD Biosciences, San Jose, CA) and were incubated overnight at 4°C to stain intracellular cytokines. FACS Aria was used for flow cytometric analysis (BD Biosciences, San Jose, CA). The results were evaluated by FlowJo (Treestar, Ashland, OR).

### Western Blot

Mature BM-DCs were placed in six-well plates with a density of 1.5 × 106 cells/ml and stimulated with 100 ng/ml LPS to induce DC activation. Meanwhile, 100 μM Scopoletin was added to DCs. After 18 h, proteins were extracted from BM-DCs as described. The total protein contents were measured by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, Rockford, MA). DC proteins were equally loaded onto SDS-PAGE, electrophoresed, transferred onto PVDF membranes, and after blocking with 5% nonfat milk in TBS, the PVDF membranes were incubated with primary antibody over 12 h at 4°C (Li et al., 2018), including JNK, p-JNK, NF-κB, p-NF-κB, p38, p-p38 (cell signal technology), IRF-7, ERK1/2, p-ERK1/2 (Abcam, Cambridge, UK), and p-IRF-7 (Signalway Antibody, College Park, MD). After washing, the membranes were incubated with 0.5% horseradish peroxidase-labeled IgG. Temembrane-bound antibodies were detected with an Immolilon™ Western chemiluminescent HRP Substrate (Millipore, Billerica, MA) and analyzed with an ImageJ (ChemiDoc XRS System; Bio-Rad).

### Quantitative RT-PCR

Total RNA was extracted by RNeasy Mini Kit (Qiagen, Valencia, CA). cDNA was subsequently synthesized by QuantiTect® Reverse Transcription Kit (QIAGEN). Quantitative RT-PCR was performed with the QuantiFast™ SYBR® Green PCR Kit (QIAGEN) under standard thermocycler conditions (Applied Biosystems, Foster City, CA). Quantification was performed by normalization with housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and relative expression level as fold changes *via* the 2−ΔΔCt method. Primers are summarized in **Supplementary Table 1.**

### Statistical Analysis

Results are described as mean ± SD. Statistical differences between two or multiple groups were implemented using Student's *t*-test or the ANOVA. *P* < 0.05 was considered statistically significant. All statistical analyses were performed by GraphPad Prism 6 (GraphPad, La Jolla, CA).

### RESULTS

### Scopoletin Suppressed the Development of EAE

Previous studies have shown that Scopoletin is safe at the dosage of 10–100 mg/kg daily with no other side effects in *in vivo*  experiment (Jamuna et al., 2015; Basu et al., 2016; Zeng et al., 2017). To assess the effect of Scopoletin on the disease course of EAE, EAE mice were i.p. injected Scopoletin (50 mg/kg) from day 0 p.i. Vehicle-treated EAE mice demonstrated a progressive increase in disease severity after day 10 p.i., while the Scopoletintreated group showed a marked decrease in mean disease course from day 14 to 20 p.i. (**Figure 1A**). Consistent with the disease development, mice treated with vehicle demonstrated abundant inflammatory cell infiltrated in the white matter compared with the Scopoletin-treated EAE mice (**Figures 1B, C**). At the same time, there was a dramatic decrease in demyelination areas in the spinal cord of Scopoletin-treated EAE mice compared to control (**Figures 1D, E**). In the therapeutic regimen, Scopoletin administration starting from disease onset (day 11 p.i.) effectively ameliorated EAE progression (**Supplementary Figure 1**). Together, these results indicate that Scopoletin effectively suppressed EAE pathogenesis.

### Scopoletin Treatment Inhibited DC Activation *In Vivo*

To assess the impact of Scopoletin on peripheral immune responses in EAE, splenocytes were harvested from EAE mice treated with Scopoletin or vehicle at day 20 p.i. and cultured in the presence of MOG35–55 (25 μg/ml) for 72 h. Expression of costimulatory molecules on CD11c+ DCs was examined by flow cytometry, which showed significantly decreased levels of MHC class II, CD80 and CD86 in Scopoletin-treated EAE mice compared to the vehicle group (**Figures 2A–D**). These results indicate that Scopoletin inhibited DC activation during the development of EAE.

Consistent with the observed effect of Scopoletin in EAE and DC activation, there was a remarkable diminishment in the percentages of CD4+ and CD8+ T cells in the periphery (**Figures 2E, F**). The percentages of Th1 and Th17 cells were also determined by intracellular staining and showed an obvious inhibition under Scopoletin treatment (**Figures 2G–I**). When stimulated with MOG35–55 *ex vivo*, splenocytes of Scopoletin-treated EAE mice secreted significantly lower levels of MOG-induced IFN-γ, IL-17, and GM-CSF (**Figure 2J**). When EAE symptoms is at its peak, spleen cells were harvested at 15 days p.i. and stimulated with MOG35–55 for 3 days. The concentration of IFN-γ, IL-17, and GM-CSF in cells culture supernatants were significantly decreased in Scopoletin-treated group (**Supplementary Figure 3A**). Furthermore, the inflammatory cytokine expression levels from sera were also detected by ELISA and RT-PCR. Although we did not detected protein expression by ELISA due to limited samples from sera, RT-PCR results showed that IL-17A, GM-CSF, and IL-1β expression were significantly decreased after Scopoletin treatment (**Supplementary Figure 3B**). These results indicated that Scopoletin alleviated the disease development of EAE by altering expression of costimulatory molecules and reduced activation of DCs, inhibiting Th1 and Th17 cells development, as well as suppressing proinflammatory cytokine secretion.

### Scopoletin Treatment Alleviated CNS Inflammation

In the CNS, effector T cells produce a number of molecules that recruit inflammatory leukocytes, ultimately leading to demyelination (Thome et al., 2014b). To evaluate the effect of Scopoletin on EAE-related CNS pathology, thoracic spinal cords were harvested from Scopoletin- or vehicle-treated EAE mice. Immunofluorescence evaluation of CNS tissues showed that Scopoletin markedly decreased infiltration of CD45+ cells compared with control mice (**Figures 3A, B**). The number of MNCs in the CNS of Scopoletin-treated mice was dramatically decreased compared to the vehicle group (**Figure 3C**). Furthermore, flow cytometry was used to analyze CD45+ (leukocyte), CD11c+ (DC), and CD11b+ (macrophage/ microglia) cell populations. Results showed that the percentages and absolute numbers of these cells were significantly reduced in the CNS of Scopoletin-treated mice compared with vehicletreated EAE mice. In addition, Scopoletin-treated mice have lower percentages and absolute numbers of CD4+ and CD8+ T cells in the CNS (**Figures 3D–F**).

Emerging data suggest that both interferon-γ-producing (Th1) and interlukin-17-producing Th17 (Th17) cells contribute to CNS autoimmunity and mediate disease pathogenesis in EAE (El-Behi et al., 2010). In addition, although various cell types produce GM-CSF, myelin-specific CD4+ T cells are essential to EAE development (Rasouli et al., 2015). To explore potential mechanisms underlying the therapeutic effect of Scopoletin in EAE, we measured the numbers of Th1, Th17, and pathogenic CD4+GM-CSF+ (Rasouli et al., 2015) cells in the CNS of EAE mice. As shown in **Figures 4A–C**, the percentages and absolute numbers of Th17 cells were reduced in the Scopoletin-treated group in comparison to the vehicle-treated mice. While the percentages of CD4+IFN-γ+ and CD4+IL-17+ T cells were not significantly changed, the absolute numbers of these cells were decreased in the Scopoletin-treated group (**Figures 4A–C**) due to the reduced total numbers of CD4+ T cells in the CNS after Scopoletin treatment

FIGURE 1 | Scopoletin suppressed the development of experimental autoimmune encephalomyelitis (EAE). (A) Mice were injected i.p. with vehicle or Scopoletin (50 mg/kg) daily from day 0 p.i. to day 20 p.i. experimental autoimmune encephalomyelitis (EAE) development were evaluated and recorded daily following a 0–5 scale. (B) Hematoxylin and eosin (H&E), below is high magnification of H&E analysis of the spinal cord sections, and (D) Luxol fast blue staining, below is high magnification of Luxol fast blue (LFB) analysis of the spinal cord sections. (C) Mean score of inflammation in H&E staining. (E) Quantification of demyelination area was analyzed by Image-Pro Plus software. Scale bar = 100 µm. *n* = 8 mice each group. Symbols represent mean ± SD, \*\* *P* < 0.01 and \*\*\* *P* < 0.001, determined by two-way ANOVA (A), or Student's *t*-test (B, D, F). Data are combined from three independent experiments.

(**Figures 3D–F**). CD4+GM-CSF+ T cells infiltration in the CNS was found in these groups. Our data therefore suggest that the effect of Scopoletin in alleviating EAE development may be due to an alteration in the differentiation of Th1, Th17, and CD4+GM-CSF+ T cells (**Figures 4A–C**).

### Scopoletin Inhibited the Activation of Murine BM-DCs *In Vitro*

We have shown the inhibitory effect of Scopoletin on DC activation *in vivo* in EAE mice (**Figures 2A–D**); here, we evaluated its direct effects on DCs *in vitro*. To that end, BM-DCs were generated. Upon

DCs, CD4+, and CD8+ T cells were measured by flow cytometry. (F) Absolute numbers of different subtypes of CNS infiltrates were calculated by multiplying the percentages of these cells with total numbers of MNCs in each spinal cord and brain tissue. Figures (A, D) are one representative of three independent experiments. Data were determined by Student's *t*-test, and symbols represent mean ± SD, \**P* < 0.05, \*\**P* < 0.01, and \*\*\**P* < 0.001.

LPS stimulation, DCs expressed relatively high levels of MHC class II, CD80, and CD86, the markers of DCs activation, and their expression was significantly inhibited by Scopoletin treatment (**Figures 5A–F**). Scopoletin, thus, showed a potential for restricting DCs activation and inhibiting expression of costimulatory molecules in BM-DCs cultures.

by flow cytometry. (C) Absolute numbers of these cells were calculated by multiplying their percentages of cytokine-producing cells and total CD4+ cell numbers in each CNS. Data are representative of three independent experiments. Symbols represent mean ± SD (n = 5 each group), \*\**P* < 0.01 and \*\*\**P* < 0.001, determined by Student's t-test.

To further examine the inhibitory effect of Scopoletin on BM-DCs activation, we then determined the mRNA levels of multiple inflammatory-associated genes expressed by BM-DCs, such as IL-6, IL-12p35, IL-12/IL23p40, IL-23p19, IL-1β, and TNF-α. Previous studies have shown that IL-12 is probably related to Th1 cells differentiation, and IL-23 and IL-6 are crucial to the Th17 cells differentiation (Luckheeram et al., 2012). Importantly, Scopoletin treatment caused a remarkably inhibition in the expression of these genes except for IL-1β (**Figure 6A**). ELISA analysis indicated the suppressive effects of Scopoletin on the secretion of IL-6, IL-23, and TNF-α by BM-DCs (**Figure 6B**). These findings suggest that the effect of Scopoletin treatment may be through inhibiting production of proinflammatory cytokines and expression of costimulatory molecules in DCs.

To test the effects of Scopoletin in DC-induced T-cell proliferation assays, we performed the DC and T-cell coculture and quantified the T-cell proliferation with CFSE; the results showed that Scopoletin reduced the T-cell proliferation *via* DCs induction (**Supplementary Figure 3C**).

### Scopoletin Inhibited NF-**κ**B Signaling During LPS-Stimulated DC Activation *In Vitro*

It is well known that NF-κB and MAPK signaling pathways play a central role in the coordinated regulation of gene expression during DC activation (Lu et al., 2014; Baratin et al., 2015). Thus, expression of the total and phosphorylation forms of IRF-7, p38, JNK, and NF-κB in BM-DCs was determined by Western blot. The level of phosphorylation of NF-κB was significantly lower in Scopoletin-treated BM-DCs when compared to DCs stimulated with LPS alone. In contrast, expression of IRF-7, p38, and JNK were not markedly affected by Scopoletin treatment (**Figures 7A**, **B**). In addition, BM-DCs were pretreated with SN50 (20 μM) for 1 h, a specific NF-κB peptide inhibitor (Lin et al., 1995), before LPS (100 ng/ml) stimulation and Scopoletin (100 μM) treatment, after 4 h cells were collected, and then, expression level of cytokines were determined by quantitative real time PCR (qRT-PCR). The expression of NF-κB downstream genes such as IL-1β and TNF-α was significantly decreased by Scopoletin and SN50 treatment. When combined with SN-50 and Scopoletin, TNF-α expression

Statistical data are expressed as mean ± SD of three independent experiments. \**P* < 0.05 by Student's *t*-test.

level was significantly enhanced compared with Scopoletin only (**Supplementary Figure 3D**). These data suggest that Scopoletin may exert its action and ameliorate the progression of EAE specifically through inhibiting NF-κB signaling.

### DISCUSSION

In this study, we provide evidence, for the first time, of the antiinflammatory activities and protective effects of Scopoletin in the mouse model of MS, EAE. Our data present a promising bioactive drug, derived from a natural source, for the treatment of MS and possibly other autoimmune diseases.

In addition, Scopoletin is also the main coumarin constituent occurring in the stems of *Erycibe obtusifolia*, a classical medicinal plant that has demonstrated various biological activities mainly used for rheumatoid arthritis with a long history (Pan et al., 2010). Scopoletin has been reported to have anti-inflammatory properties, e.g., inhibition of lymphocyte proliferation and reduction of cytokine production (Arcos et al., 2006). *In vitro* studies showed that Scopoletin inhibited proinflammatory cytokine secretion from RAW 246.7 and HMC-1 cell lines (Moon et al., 2007; Connell et al., 2017). Scopoletin also exhibited antiinflammatory, antioxidant, and antiacetylcholinesterase potential against Alzheimer's disease (Garcia-Morales et al., 2015). However, its role is unknown in the regulation of APC functions in an autoimmune disease such as MS.

In autoimmune diseases, DCs as professional APCs have a potent ability to trigger naive T-cell reaction and activate an

autoreactive response (Thompson et al., 2018), and decreasing CD80/86 prevented T-cell activation and Th-cell differentiation by inhibiting the costimulatory second signal (Grimbert et al., 2011; Yamamoto et al., 2012). Our results showed that the therapeutic effects of Scopoletin were largely dependent on the suppression of DC activation. Scopoletin reduced the expression of costimulatory molecules CD80 and CD86 as well as MHC class II *in vivo* and *in vitro.* Furthermore, in MS patients and EAE mice, there is a marked increase in the levels of IFN-γ and IL-17, signature cytokines for Th1 and Th17 cells, respectively (Arellano et al., 2017). As drivers of autoimmune response, pathogenic Th1 and Th17 cells are therefore considered to be the main culprits in MS and EAE (Grigoriadis et al., 2015). Our data demonstrated that Scopoletin treatment remarkably decreased the percentages of Th1 and Th17 cells. On the other hand, GM-CSF, which can be produced by both Th1 and Th17 cells, is required for encephalitogenicity of these pathogenic T-cell subsets (Codarri et al., 2011; El-Behi et al., 2011). Reduced numbers of CD4+GM-CSF+ T cells in the spleen and CNS following Scopoletin treatment would, therefore, be an important mechanism underlying the effect of Scopoletin on EAE. When the direct effect of Scopoletin on Th1 and Th17 cell differentiation was further investigated *in vitro*, our results showed that differentiation of Th1 and Th17 was not significantly altered by Scopoletin treatment (**Supplementary Figure 2**). Furthermore, the inhibitory effect of Scopoletin treatment on Th1/Th17 cells *in vivo* in EAE mice must therefore be through an indirect mechanism, i.e., by inhibition of DC activation during EAE progression.

It has been well documented that IL-6 plays a key role in determination of Th17 polarization, and knockout of these cytokines completely inhibits EAE development (Giralt et al., 2013). In addition, IL-23 is important for expanding and maintaining the developing Th17 population (Smith and Colbert, 2014) and is required for differentiation of Th17 cells and the pathogenic role of these T cells (El-Behi et al., 2011). Given that DCs are major producers of these cytokines, differentiation of naive T cells into Th17 phenotype is dependent on activated DCs (Agalioti et al., 2018). IL-6 signaling plays a key role in the differentiation of CD4 + T cells, in combination with TGF-β, IL-1β, IL-23, and IL-6, promotes the differentiation of Th17 cells by activating transcription factors containing signal transducers and activators of transcription 3 (STAT3), retinoic acid-related orphan receptors gamma t (RORγt) (Yao et al., 2014). Here, we found that Scopoletin treatment significantly decreased expression of activation and costimulation markers of DCs and their IL-6 and IL-23 production, further

100 μM. After 18 h, cells were harvested and homogenized for Western blot. (A) Protein expression levels of nuclear transcription factors and protein kinase were measured by Western blot. (B) Densitometry was determined by ImageJ, and data were analyzed by GraphPad Prism 6. Data are expressed as mean ± SD of three biological replicates. \*\**P* < 0.01 by Student's *t*-test. Figure A is one representative of three independent experiments.

indicating that Scopoletin blocked DC secretion of Th17-inducing cytokines, thus resulting in inhibited Th17 phenotype formation.

When antigen presentation occurs, epitope-major histocompatibility complexes identified by T-cell receptors and costimulatory signals are sent by interaction of CD80, CD86 of DCs and CD28 expressing on T cells, as well as the cytokine milieu formed by DCs and other inflammatory cells in the microenvironment (Thome et al., 2014b). These signals working together results in T cell polarization into Th1, Th17, and other effector CD4+ T cells. Studies have shown that the activation of DCs is possibly regulated by three intracellular signaling pathways, including NF-κB, MAPKs, and PI3K (Chesi et al., 2016; Rescigno et al., 1998). To further explore the mechanism of Scopoletin-induced amelioration of EAE, we investigated the effects of Scopoletin on BM-DCs *in vitro,* which indicated decreased expression of a series of inflammation-related genes. While the activation of MAPKs containing ERK, p38, and JNK was not prominently influenced in LPS-stimulated DCs, Scopoletin treatment exerted a profound inhibitory effect on NF-κB activity during DC activation. It was shown that Scopoletin markedly suppressed the phosphorylation of NF-κB, which plays a significant role in the activation and maturation of DCs. Previous studies showed that IL-6 gene expression was reduced through downregulating of NF-κB signaling (Chiang et al., 2014), and the promoter activity of IL-23 was modulated by the NF-κB pathway. Our results indicated that the therapeutic effect of Scopoletin may be mediated by inhibiting the NF-κB signaling pathway, which is vitally involved in producing proinflammatory cytokines.

In summary, we reported, for the first time, that natural compound Scopoletin ameliorated the severity of EAE by regulating DC activation and reducing CNS inflammation *via* suppression of NF-κB signaling. Scopoletin treatment could therefore be a potential candidate for the modulation of inflammatory conditions of autoimmune diseases.

### CONCLUSION

Our study demonstrates that natural products Scopoletin derived from medical or edible plants will be valuable in developing a novel therapeutic agent for MS in the future.

### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

### REFERENCES


### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of 'institutional guidelines and regulations approved by the Animal Ethics Experimental Committee at of Shaanxi Normal University', 'the Animal Ethics Experimental Committee at of Shaanxi Normal University'. The protocol was approved by the 'Animal Ethics Experimental Committee at of Shaanxi Normal University'.

### AUTHOR CONTRIBUTIONS

FZ, YZ, TY, and XL conceived and designed the experiments. FZ, Z-QY, TY, H-RF, and J-JH carried out the experiments. FZ, YZ, TY, JT, and XL analyzed data and wrote the manuscript. Z-ZW and XL cosupervised the study and revised the paper. All authors read and approved the final manuscript.

### FUNDING

This study was supported by the Chinese National Natural Science Foundation (Grant No. 81771345, 81501062, 31670299), the Natural Science Foundation of Shaanxi Province, China (Grant No. 2018JZ3001, 2018JQ8033, 2019KJXX-022), and the Fundamental Research Funds for the Central Universities (Grant No. GK201903062, GK20182010, GK201701009, 2018CSLZ018, 2018CSLZ019, 2018CSLZ020, 201810718052).

### ACKNOWLEDGMENT

We gratefully thank Katherine Regan for editorial assistance.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019.00863/ full#supplementary-material


experimental autoimmune encephalomyelitis. *Immunol. Cell Biol.* 92, 124–132. doi: 10.1038/icb.2013.73


**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 © 2019 Zhang, Zhang, Yang, Ye, Tian, Fang, Han, Wang and Li. 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.*

# Corrigendum: Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-**κ**B Signaling

#### *Approved by:*

*Frontiers Editorial Office, Frontiers Media SA, Switzerland*

#### *\*Correspondence:*

*Zhe Zhi Wang zzwang@snnu.edu.cn Xing Li xingli\_xian@126.com*

*†These authors have contributed equally to this work*

#### *Specialty section:*

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

*Received: 13 August 2019 Accepted: 14 August 2019 Published: 13 September 2019*

#### *Citation:*

*Zhang F, Zhang Y, Yang T, Ye Z-Q, Tian J, Fang H-R, Han J-J, Wang Z-Z and Li X (2019) Corrigendum: Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-κB Signaling. Front. Pharmacol. 10:1037. doi: 10.3389/fphar.2019.01037*

*Fei Zhang †, Yuan Zhang †, Ting Yang †, Ze-Qing Ye, Jing Tian, Hai-Rong Fang, Juan-Juan Han, Zhe-Zhi Wang\* and Xing Li\**

*National Engineering Laboratory for Resource Development of Endangered Crude Drugs in Northwest China, The Key Laboratory of Medicinal Resources and Natural Pharmaceutical Chemistry, The Ministry of Education, College of Life Sciences, Shaanxi Normal University, Xi'an, China*

Keywords: scopoletin, experimental autoimmune encephalomyelitis, multiple sclerosis, dendritic cells, NF-**κ**B signaling

### **A Corrigendum on**

### **Scopoletin Suppresses Activation of Dendritic Cells and Pathogenesis of Experimental Autoimmune Encephalomyelitis by Inhibiting NF-κB Signaling**

*by Zhang F, Zhang Y, Yang T, Ye Z-Q, Tian J, Fang H-R, Han J-J, Wang Z-Z and Li X (2019). Front. Pharmacol.10:863. doi: 10.3389/fphar.2019.00863*

In the published article, author Zhe-Zhi Wang's identity was incorrect. Instead of "co-author," it should be "co-corresponding author." The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

*Copyright © 2019 Zhang, Zhang, Yang, Ye, Tian, Fang, Han, Wang and Li. 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.*

# Microglial Drug Targets in AD: Opportunities and Challenges in Drug Discovery and Development

*Knut Biber1, Anindya Bhattacharya2, Brian M. Campbell3, Justin R. Piro4, Michael Rohe1, Roland G.W. Staal5, Robert V. Talanian4 and Thomas Möller4\**

*1 AbbVie Deutschland GmbH & Co. KG, Neuroscience Research, Ludwigshafen, Germany, 2 Janssen Research & Development LLC, San Diego, CA, United States, 3 Sage Therapeutics, Cambridge, MA, United States, 4 AbbVie Foundational Neuroscience Center, Cambridge, MA, United States, 5 Paracelsus Neuroscience, Metuchen, NJ, United States*

#### *Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Claes Wahlestedt, Leonard M. Miller School of Medicine, United States Anthony John Hannan, Florey Institute of Neuroscience and Mental Health, Australia*

*\*Correspondence: Thomas Möller thomas.moeller@abbvie.com*

#### *Specialty section:*

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

*Received: 14 March 2019 Accepted: 01 July 2019 Published: 23 August 2019*

#### *Citation:*

*Biber K, Bhattacharya A, Campbell BM, Piro JR, Rohe M, Staal RGW, Talanian RV and Möller T (2019) Microglial Drug Targets in AD: Opportunities and Challenges in Drug Discovery and Development. Front. Pharmacol. 10:840. doi: 10.3389/fphar.2019.00840*

Alzheimer's disease (AD) is a large and increasing unmet medical need with no diseasemodifying treatment currently available. Genetic evidence from genome-wide association studies (GWASs) and gene network analysis has clearly revealed a key role of the innate immune system in the brain, of which microglia are the most important element. Singlenucleotide polymorphisms (SNPs) in genes predominantly expressed in microglia have been associated with altered risk of developing AD. Furthermore, microglia-specific pathways are affected on the messenger RNA (mRNA) expression level in post-mortem AD tissue and in mouse models of AD. Together these findings have increased the interest in microglia biology, and numerous scientific reports have proposed microglial molecules and pathways as drug targets for AD. Target identification and validation are generally the first steps in drug discovery. Both target validation and drug lead identification for central nervous system (CNS) targets and diseases entail additional significant obstacles compared to peripheral targets and diseases. This makes CNS drug discovery, even with well-validated targets, challenging. In this article, we will illustrate the special challenges of AD drug discovery by discussing the viability/practicality of possible microglia drug targets including cluster of differentiation 33 (CD33), KCa3.1, kynurenines, ionotropic P2 receptor 7 (P2X7), programmed death-1 (PD-1), Toll-like receptors (TLRs), and triggering receptor expressed in myeloid cells 2 (TREM2).

Keywords: drug target, microglia, target identification, target validation, screening cascade

### HIGHLIGHTS:


### INTRODUCTION

Microglial reactivity has long been recognized as a pathological hallmark of a wide variety of neurological diseases. While morphological changes of microglia were initially interpreted as a reactive response to neuronal damage, growing evidence suggests that these changes are not merely "reactive" but indicate that glial pathology contributes to disease progression (Raj et al., 2014a; Verkhratsky et al., 2014; Ben Haim et al., 2015; Bergles and Richardson, 2015; Biber et al., 2015; Heneka et al., 2015; Heppner et al., 2015; Jain et al., 2015; Ransohoff and El Khoury, 2015; Robel and Sontheimer, 2015; Zuchero and Barres, 2015; Osborn et al., 2016). Especially with regard to Alzheimer's disease (AD), numerous reviews cover various aspects of microglial biology in the context of disease (see for example: (Guedes et al., 2018; Hansen et al., 2018; Kinney et al., 2018; Shi and Holtzman, 2018; Thei et al., 2018; Ulland and Colonna, 2018; Henstridge et al., 2019). Using the search term "microglia Alzheimer," PubMed lists close to 100 reviews that were published in the last 2 years (Jan 2017–Jan 2019). It is not the purpose of the current review to discuss the potential role of microglia in AD, but to focus on the challenges and practical aspects of microglial proteins or pathways as potential drug targets.

A considerable set of genetic risk factors for AD are predominantly expressed in microglia (Wes et al., 2016; Henstridge et al., 2019). Thus, it is very clear from human genetics that regardless of being "reactive," "active," "dystrophic," "senescent," "dysfunctional," or "disease associated," microglial involvement in AD offers a hitherto unexplored intervention point. Many publications concerning microglia in AD include conclusions like: "The presented data here suggest that protein 'XYZ' can be considered as a drug target for AD." As there are some common misunderstandings between colleagues focused on biology vs. drug discovery regarding the definition of a target, we first aim to clarify what it takes to consider a microglial molecule suitable for a drug discovery program. We will then review some commonly suggested microglial targets and pathways, i.e., cluster of differentiation 33 (CD33), KCa3.1, kynurenines (KYN), ionotropic P2 receptor 7 (P2X7), programmed death-1 (PD-1), Toll-like receptors (TLRs), and triggering receptor expressed in myeloid cells 2 (TREM2), discussing their potential role in AD and their respective challenges in regard to drug discovery.

### TARGET IDENTIFICATION

In general, preclinical drug discovery can be staged as follows: target identification, in which the validity of a target for modulating a disease process and the initial chemical matter along with the means of characterizing them are established; lead optimization, in which target validity is further strengthened while identifying candidate molecules with the full range of necessary drug-like properties; and finally, preclinical testing for safety and efficacy. Apart from target identification, we will not discuss the other stages of preclinical drug development (**Figure 1**). The interested reader is referred to (Möller and Boddeke, 2016) for a detailed description of the whole process.

For potential targets, it is of utmost importance to demonstrate relevancy to the disease and that modulating the target would result in therapeutic benefit with an acceptable safety margin. Evidence for a role of a candidate target in disease might include i) altered expression in disease, ii) integral relationship to disease pathophysiology, iii) genetic association with disease, iv) having a mechanistic link to disease etiology, and v) demonstration of a beneficial effect when modulated in a disease-relevant *in vitro* or (preferably) *in vivo* model.

Unfortunately, generating expression data for the human central nervous system (CNS) is extremely difficult. While for peripheral diseases biopsies, or blood draws are routine, for the CNS, most parameters can only be derived from post-mortem brain tissue that at best represents end-stage disease, or by limited imaging methods, making it difficult to observe changes in earlier and therapeutically more relevant disease states. Non-human models can also provide insight, but with significant translational uncertainties.

A strong genetic association of a microglial molecule with AD, for example, CD33, complement receptor 1 (CR1), phospholipase C gamma 2 (PLCγ2), and TREM2 (Wes et al., 2016; Henstridge et al., 2019), also suggests promise as a drug target. If such evidence can further be strengthened by a mechanistic link to a disease model by *in vitro* or *in vivo* genetic or pharmacologic intervention studies [e.g., knocking out CD33 (Bradshaw et al., 2013)], that would serve to increase confidence in the target.

### WHAT MAKES A GOOD (MICROGLIAL) TARGET?

### Drugability

Besides linking a target to disease, there are other more pragmatic considerations for selecting a target. Admittedly, "drugable" is not a well-defined term; it usually includes chemical tractability with a ligand (small molecule or antibody) but can also refer more broadly to the properties of such ligands, e.g., pharmacokinetics, distribution, metabolism (and the pharmacologic properties of

significant metabolites), pharmaceutical properties including chemical stability, and manufacturability.

Traditionally, a target would be considered chemically tractable if a small molecule or antibody can bind with high affinity and specificity and induce the desired biological effect. For example, G-protein coupled receptors (GPCRs) are generally considered chemically tractable; more than 30% of currently marketed drugs target GPCRs (Ma and Zemmel, 2002; Hauser et al., 2017). Ion channels or enzymes are also considered tractable as those targets often include structural features that permit potent and specific ligand binding, with consequent direct inhibition of conductance or catalytic activity. Blocking protein–protein interactions is generally more difficult; however, recent advances have challenged this view, and modulation of protein–protein interactions by small molecules is now an area of active research and will increase the number of targets considered drugable (Higueruelo et al., 2013). Furthermore, blocking protein–protein interactions can often be accomplished with therapeutic antibodies, and several innovative approaches have been devised to improve antibody penetration into the brain (Watts and Dennis, 2013; Salameh and Banks, 2014; Pardridge, 2015). Despite these recent advancements, developing therapeutics for the CNS remains challenging, and while there are many biologically interesting molecules related to microglia, only a small subset of them might be considered drugable.

### Specificity

It is not merely enough to generate a potent pharmacological agent against a drug target; it is also of utmost importance that the molecule is selectively modulating only this one target to avoid off-target effects that may hamper the interpretation of experimental data. Unfortunately, many so-called "reference compounds," such as commercially available inhibitors, are not as specific as advertised by the suppliers (Frye, 2010). A group of 50 scientists from different life science disciplines have covered this topic in detail in Commentary in *Nature Chemical Biology* (Arrowsmith et al., 2015). This lack of specificity makes interpretation of results gained with many compounds virtually impossible. A prominent example in microglia biology (although not directly related to a single target) is minocycline, which often is referred to as a "microglia inhibitor" (Möller et al., 2016).

### Safety

Arguably the most important criterion for target selection, however, is safety. During clinical development, many drugs fail, not due to the lack of efficacy but due to adverse safety events inadequately predicted by preclinical studies. Toxicity might not only arise from modulating the target itself, meaning that the therapeutic window is small, but that there could be offtarget effects of a drug, or that toxicity may also arise from a metabolite of the drug that itself is toxic. It is therefore important to select targets with the highest likelihood of being safe as early as possible. A common question is how broadly a target is expressed. Localized expression preferentially in the target cell and/or tissue is considered advantageous as there is less likelihood for adverse effects unrelated to the desired mechanism of action. Unfortunately, there are very few potential targets restricted to microglia as most potential targets are also expressed in peripheral myeloid cells, so effects on the peripheral immune system are a persistent concern. Safety concerns, whether based on side effects of a development compound, a target knockout phenotype, or suspected essential physiological roles of the target, need to be addressed as early as possible by experiments.

Taken together, there is a whole range of drugability considerations that must be met to make a target practical for drug discovery. There are few targets in microglia that are often discussed in the literature to be of value in AD. Here we will discuss the potential and challenges around these targets when it comes down to drugability aspects.

### CURRENTLY PROPOSED MICROGLIAL DRUG TARGETS

### Proposed Role of CD33 in AD

In 2011, large-scale genome-wide association studies (GWASs) identified CD33, a member of the sialic acid–binding immunoglobulin-like lectins (siglecs) family, as a genetic locus associated with the risk to develop AD (Hollingworth et al., 2011; Naj et al., 2011). This association with AD risk has been confirmed in a variety of subsequent genetic studies in different ethnic groups, with odds ratios between 0.7 and 0.9; thus, the CD33 rs3865444A allele protects from the development of AD [for recent metaanalysis, see (Jiang et al., 2018)]. It was furthermore described that the protective single-nucleotide polymorphism (SNP) rs3865444A is in linkage disequilibrium with rs12459419T, which causes increased splicing of exon2, and therefore deletes the sialic acid–binding site of CD33. The protective SNP therefore leads to increased expression of a truncated, non-functional version of the receptor, called d2-CD33 or CD33m (Bradshaw et al., 2013; Malik et al., 2013; Raj et al., 2014b). A reduction in full-length CD33 expression was observed in myeloid cells of CD33 rs3865444A allele carriers, resulting in an overall reduced CD33 function (Bradshaw et al., 2013; Raj et al., 2014b).

CD33 signaling in humans depends on two intracellular inhibitory domains, one immunoreceptor tyrosine-based inhibitory motif (ITIM), and one ITIM-like domain. CD33 is therefore considered to be an inhibitory immune receptor. Indeed, when activated CD33 becomes tyrosine-phosphorylated by Src kinases, which subsequently recruit Src homology-2 domain (SH2)–containing tyrosine phosphatases (SHP1/2), this results in the dephosphorylation and inhibition of signaling cascades (see for review Macauley et al., 2014). CD33 is expressed in myeloid cells, and its signaling negatively regulates the function of these cells. Accordingly, knockdown of CD33 in human monocytes leads to spontaneous release of interleukin 1 beta (IL-1β), Tumor necrosis factor alpha (TNFα), and IL-8 (Lajaunias et al., 2005). In monocytes from carriers of the CD33 rs3865444A allele (decreased CD33 function), an increased uptake capacity of amyloid beta(1–42) (Aβ1–42) was described (Bradshaw et al., 2013).

In the brain, CD33 is specifically expressed in microglia (Bradshaw et al., 2013; Griciuc et al., 2013; Malik et al., 2013; Walker et al., 2015), where an inverse relationship between CD33 expression and Aβ uptake capacity was observed (Griciuc et al., 2013). Accordingly, in the cortex of AD patients, a positive correlation of microglial CD33 expression and amyloid pathology was described, indicating that increased CD33 expression in microglia promotes plaque pathology (Griciuc et al., 2013). CD33-deficient Amyloid precursor protein (APP)/ Presenilin 1 (PS1) mice showed reduced amyloid pathology compared to age-matched wildtype (WT) controls, supporting the idea that reduced microglial CD33 function is beneficial with respect to amyloid pathology (Griciuc et al., 2013). Given this variety of data concerning the role of CD33 as regulator of microglial Aβ uptake capacity, this siglec receptor is currently considered as drug target in AD. It is considered that inhibition of CD33 function in microglia would increase their amyloid uptake capacity, which in turn would be beneficial in AD **Figure 2** (Heneka et al., 2015).

### Target Challenges Around CD33

There are many challenges with respect to CD33 as a drug target in AD. In general, finding ligands for siglecs is difficult, as glycan binding often is of relatively low affinity and relies on multivalency to achieve adequate affinity (Varki, 2009). There is evidence that CD33 binds α2.3 sialic acid and α2.6 sialic acid; the corresponding glycosylated proteins that are decorated with these sialic acids, however, have not been identified (Linnartz-Gerlach et al., 2014; Macauley et al., 2014). As CD33 is the shortest of all known siglecs, it most likely is activated in a cis-manner (O'Reilly and Paulson, 2009; Linnartz-Gerlach et al., 2014; Macauley et al., 2014),

activators of microglial phagocytosis, respectively. For TREM2, activating antibodies, and for CD33, inhibitory antibodies, have been proposed as therapeutic interventions. Figure created with Biorender.com

meaning that the sialylated glycoproteins that activate CD33 would also be expressed in microglia. One report provides data that soluble sialyllactosamine in the mM range might activate CD33 in de-sialylated monocytes (Lajaunias et al., 2005). This finding awaits independent confirmation. Apart from our very limited understanding around natural ligands for CD33, functional tool compounds for CD33 have not yet been described. The only published molecule displaying a reasonable binding at CD33 is cpd22, which displaces bead-coupled α2.6 sialic acid with an half maximal effective concentration (EC50) of about 10 µM (Rillahan et al., 2014). It therefore is suggested that cpd22 binds CD33 at the sialic acid–binding domain; whether this binding affects CD33 signaling is not understood. Thus, it is not clear whether cpd22 might function as an agonist or antagonist for CD33. The lack of ligands or tool compounds seriously hampers the development of functional analysis of CD33 in cellular assays, making the development of functional inhibitors for CD33 difficult.

Therapeutic inhibitors of CD33 are hypothesized to increase phagocytic properties in microglia; from a drug development perspective, they might also be envisaged to inhibit the CD33 signaling cascade. It is not yet understood how recruitment of SHP1 leads to inhibition of cellular phagocytosis in CD33 expressing myeloid cells (Linnartz-Gerlach et al., 2014).

Another challenge concerns the large species difference of CD33. Mouse and human CD33 differ not only in the extracellular domain but also in the transmembrane domain as well as intracellularly. In contrast to human CD33, mouse CD33 only has one ITIM-like domain but lacks the ITIM domain (Cao and Crocker, 2011). Moreover, in the transmembrane region of mouse CD33, a positively charged residue is present, which might enable mouse CD33 to recruit immunoreceptor tyrosine-based activation motif (ITAM)-motif–containing adapter proteins like DNAX activating protein of 12 kDa (DAP12) (Linnartz-Gerlach et al., 2014). These data suggest that mouse CD33 can act as an activating receptor instead of being inhibitory, a hypothesis that awaits experimental evidence. In cultured microglia from CD33KO mice, it was shown that CD33 negatively couples to phagocytosis (Griciuc et al., 2013). Whether this is due to ITIM-like signaling or recruitment of ITAM-motif–containing adapter proteins is currently an open question. Potential drugs that inhibit CD33 function, however, need to be tested in animal models. Whether or not the mouse would be an adequate model system to do so is questionable given the species differences and the limited knowledge on CD33 function in mouse cells.

Taken together, despite the genetic association between CD33 and AD development and the straightforward hypothesis of how reduced CD33 function might protect from AD, CD33 remains a target with its unique set of challenges.

## Proposed Role of KCa3.1 in AD

KCa3.1 is an intermediate-conductance potassium channel that is activated when intracellular calcium is increased. The Ca2+ binds to calmodulin, which is constitutively bound to the calmodulin binding domain of the channel, resulting in the opening of the K+ channel (Kd ~ 300 nM) (Xia et al., 1998; Fanger et al., 1999; Bouhy et al., 2011). The KCa3.1-mediated K+ efflux augments the driving force Biber et al. Microglia as Drug Targets in AD

for further Ca2+ influx *via* Ca2+ release-activated calcium (CRAC) channels and subsequent refilling of intracellular calcium stores. Thus, one of the primary functions of KCa3.1 is the augmentation of Ca2+ signaling (Feske et al., 2012). Expression of KCa3.1 in the CNS and its controversies have been reviewed (Dale et al., 2016). Here we will focus on microglia, as the expression of KCa3.1 was established using well-validated antibodies and demonstration of voltage-independent KCa3.1-like Ca2+-activated K+ currents.

Voltage-independent KCa3.1-like Ca2+-activated K+ currents in cultured murine microglia were first reported by Eder et al. (Eder et al., 1997). Khanna et al. subsequently demonstrated the presence of KCa3.1 mRNA in cultured rat and mouse microglia (Khanna et al., 2001). The conclusion that microglia express KCa3.1 was supported by pharmacological studies that demonstrated that charybdotoxin and clotrimazole could inhibit the voltageindependent KCa3.1-like Ca2+-activated K+ currents (Schilling et al., 2002; Schilling and Eder, 2004). There is further evidence that charybdotoxin and clotrimazole prevent the induced microglial oxidative burst (Khanna et al., 2001) and microglial migration (Schilling and Eder, 2004). The KCa3.1 inhibitor TRAM-34 reduced lipopolysaccharide (LPS)-induced microglial neurotoxicity (Kaushal et al., 2007), and inhibition of KCa3.1 using senicapoc (see below) blocked the LPS-induced IL-1β and Nitric oxide (NO) production by microglia (Staal et al., 2017a).

Early *in vivo* studies using antibodies validated in KCa3.1 WT and knock out (KO) mice did not provide any evidence of KCa3.1 expression on any cells in the mouse or human brain (including juvenile, adult, and aged mice) (Schilling and Eder, 2004; Schilling and Eder, 2007). However, studies using antibodies validated in this manner demonstrated KCa3.1 expression on CD68-positive cells (microglia or macrophages) upon injury to the CNS as induced by ischemic injury (Chen et al., 2011), suggesting that KCa3.1 is expressed by microglia only in response to injury. The study by Chen et al. also demonstrated the therapeutic efficacy of inhibiting KCa3.1 (using TRAM-34) in ischemic injury, a finding they validated using KCa3.1-null mice (Chen et al., 2011; Chen et al., 2015). Many other studies have since demonstrated the therapeutic efficacy of KCa3.1 inhibition in other models of neurological injury and disease in which microglia are believed to play a significant role, including experimental autoimmune encephalomyelitis, traumatic brain injury, spinal cord injury, optic nerve transection, glioblastoma multiforme, neuropathic pain, and AD (Dale et al., 2016; Staal et al., 2017b).

Taken together, there is convincing evidence for the expression of KCa3.1 in cultured rodent microglia. While the details of regulation of KCa3.1 await confirmation with more specific tool compounds, the available data suggest a role for KCa3.1 at the very least in microglia reactive oxygen species (ROS) production and signaling.

### Challenges in KCa3.1 Inhibition as Therapeutic Intervention in AD

The presence of "activated" microglia identified by expression of markers of microglial activation [major histocompatibility complex II (MHCII), Iba1] or an activated morphology (enlarged cell bodies and shorter, thicker processes) in brains of AD is well established. The observation that microglial "activation" can occur prior to plaque deposition and neurofibrillary tangle formation, and correlates with cognitive deficits, is consistent with the theory that microglial activation occurs at the earliest stages of disease (perhaps even before disease onset) and is somehow tied to its progression (Okello et al., 2009a; Okello et al., 2009b; Sastre et al., 2011). Maezawa et al. demonstrated that TRAM-34 blocks Aβ-induced proliferation of microglia, p38 Mitogen-activated protein kinase (MAPK) phosphorylation, nuclear factor kappa b (NF-κb) activation, and NO generation in primary microglia cultures (Maezawa et al., 2012). Furthermore, TRAM-34 inhibited neurotoxic effects of Aβ oligomers in mixed microglia–neuron cultures and in organotypic hippocampal slices by decreasing microglial activation and partially preventing synaptic loss (Maezawa et al., 2011).

Senicapoc is a potent and selective inhibitor of KCa3.1 developed for the treatment of sickle cell anemia [half maximal inhibitory concentration (IC50) = 11 nM; screening of 57 targets yielded no hits below 1 µM other than KCa3.1] (Ataga et al., 2006; Ataga et al., 2008; Ataga and Stocker, 2009; Dale et al., 2016; Staal et al., 2017a). In a recent and comprehensive study, Maezawa and colleagues demonstrated that KCa3.1 is functionally upregulated in microglia in brains from 5xFAD mice as well as AD patients (Jin et al., In Press). The authors demonstrate that Aβ oligomers impair long-term potentiation (LTP) and that this effect is reversed by the KCa3.1 inhibitor TRAM-34 as well as senicapoc in hippocampal slices or *in vivo*. It was furthermore demonstrated in this paper that senicapoc can reduce the pro-inflammatory effects as well as rescue the hippocampal LTP affected by Aβ oligomer (Jin et al., In Press). Senicapoc was then tested in the 5xFAD mice, starting at 6 months of age and continuing for three months. Results show that treatment with senicapoc reduced the amyloid load as well as neuroinflammation while enhancing neuronal plasticity compared to vehicle-treated mice (Jin et al., In Press). Thus, the study provides a strong rationale for repurposing senicapoc for the treatment of AD.

In fact, the Alzheimer's Drug Discovery Foundation has provided funding to manufacture clinical-grade senicapoc and conduct the required stability testing in support of an investigational new drug (IND) filing. Upon approval of the IND filing, senicapoc will be tested in a phase II clinical trial in prodromal and mild AD (https://www.alzdiscovery.org/newsroom/announcements/1.8-million-in-new-funding-supportsclinical-stage-treatments).

Senicapoc has been through extensive preclinical safety testing as well as phase I, II, and III clinical trials with few significant side effects noted. Given our knowledge of the pharmacokinetics of senicapoc, including its CNS penetrance (Staal et al., 2017b), it is likely that a dose similar to that used in the previous clinical trials will be used. AD patients, however, are very different from patients with sickle cell disease. It is likely that the AD patients will be older and have more comorbidities than those with sickle cell anemia. The age and comorbidities could reveal underlying susceptibilities to KCa3.1 inhibition that may not have been problematic in younger patients with sickle cell disease (e.g., susceptibility to infections). On the other hand, many diseases or conditions afflicting the elderly have inflammatory components, and a mild immunosuppressant such as senicapoc may benefit those conditions as well (e.g., stroke, chronic and neuropathic pain, atherosclerosis).

Yet, the current AD study is still incredibly important to test the hypothesis that targeting neuroinflammation, specifically KCa3.1, can slow or halt the progression of AD. If the trial results demonstrated significant and convincing slowing of disease progression, pharmaceutical companies could initiate drug discovery projects to develop improved CNS penetrant KCa3.1 inhibitors.

### Targeting the Kynurenine Pathway to Treat AD

Development of drugs that target products of tryptophan metabolism has transformed the treatment of psychiatric disorders, particularly with the advent of serotonin (5-HT) reuptake inhibitors (Wong et al., 1995). However, it is noteworthy that metabolism to 5-HT represents only a small fraction of the fate of free tryptophan, and in fact, the great majority is converted to KYN (Leklem, 1971) and its subsequent products by an array of enzymes located throughout the body and brain. Yet, surprisingly, there are no marketed medications that directly target metabolism of the KYN pathway.

Entry into the KYN pathway begins with metabolism of tryptophan into N-formylkynurenine by indole-2,3-dioxygenase (IDO) or tryptophan-2,3-dioxygenase (TDO), which is then rapidly converted to L-kynurenine (Takikawa et al., 1986; Zhang et al., 2007). Metabolism of KYN produces a host of neuro-active products in the brain that may contribute to CNS diseases (Török et al., 2016; Lovelace et al., 2017; Schwarcz and Stone, 2017). The KYN pathway bifurcates into neurotoxic and neuroprotective metabolic routes. As an example, astrocytes largely express kynurenine aminotransferases (KATs) and so are the primary source of the neuroprotective metabolite, kynurenic acid (KYNA), in the brain (Guidetti et al., 2007). Alternatively, microglia express kynurenine 3-monooxygenase (KMO), which converts KYN to neurotoxic metabolites such as 3-hyroxykynurenine (3-HK) and subsequently quinolinic acid (Quin) (Heyes et al., 1996; Guillemin et al., 2003a) (**Figure 3**). While the topic of this discussion is targeting the disruption in microglial KYN metabolism, one should understand that the various branches of KYN metabolism do not act in isolation, and so a change in one may cause a change in the others by altering the availability of substrate.

Inflammation triggers activation of microglia and dysregulation of KYN metabolism, leading to production of neurotoxic metabolites from the KMO branch, a process that is predicted to be involved in the development of AD pathology (Campbell et al., 2014; Zádori et al., 2018). However, direct evidence of microglial KYN metabolism from human AD subjects suggests that this effect may be limited to localized pathological events rather than global changes in CNS KYN metabolism. As an example, levels of KYN and 3-HK were both surprisingly reduced in CSF of AD patients with dementia (Tohgi et al., 1992). However, tryptophan, 5-HT, and melatonin were all also significantly reduced in the same study. In addition, the same group reported that KYN was reduced in a second cohort of patients with AD dementia but not in ones with vascular dementia (Tohgi et al., 1995). Furthermore, Quin (Mourdian et al., 1989; Heyes et al., 1992) and KYNA (Wennström et al., 2014) were reportedly unchanged in recent studies, though earlier evidence indicated that KYNA could also be decreased (Heyes et al., 1992; Hartai et al., 2007; Gulaj et al., 2010) in AD patients. Overall, there is limited evidence that CSF tryptophan/ KYN metabolites are reproducibly altered in AD, though where it has been assessed, most studies point to either no consistent changes in metabolites or a reduction in KYN metabolism (**Table 1**).

In post-mortem AD brain tissue, early studies reported a lack of change in KYN metabolism. Indeed, Quin levels were shown to be unchanged (Moroni et al., 1986; Mourdian et al., 1989; Sofic et al., 1989; Pearson and Reynolds, 1992), as were KYNA concentrations (Flint Beal et al., 1992). However, more recent analysis revealed


regional changes that indicated greater production of KYN as well as shunting of KYN metabolism toward production of neurotoxic microglial metabolites. Immunohistochemical analysis showed that both TDO and IDO were elevated in the hippocampus of AD patients (Guillemin et al., 2005; Wu et al., 2013). Importantly, the increase in IDO expression was found in microglia, astrocytes, and neurons, with the primary product found in proximity to senile plaques and neurofibrillary tangles in postmortem AD brain being Quin, the microglial product (Guillemin et al., 2005; Bonda et al., 2010). These data are in line with reports that treatment of microglia with Aβ1–42 induces IDO expression (Guillemin et al., 2003a; Guillemin et al., 2003b) and that Quin may seed α-synuclein aggregation (Tavassoly et al., 2018) and induce tau hyperphosphorylation (Rahman et al., 2009). Given the welldocumented evidence that tryptophan and KYN metabolism is sensitive to inflammatory stimuli (Campbell et al., 2014; Strasser et al., 2017), it is plausible to predict that neuroinflammation centered around plaques and tangles in the brains of AD patients produces localized pro-inflammatory activation of microglia, resulting in dysregulation of KYN metabolism, which favors production of neurotoxic metabolites such as 3-HK and Quin. A deleterious impact resulting in local brain circuit disruption and longer-term neurodegeneration is hypothesized, but further analysis in longitudinal studies is needed to confirm or disprove this.

### Challenges of Developing Kynurenine Pathway Drugs for AD

The main challenge for developing KYN metabolite inhibitors for AD is a clear understanding of the role of KYN metabolic products in the development and/or progression of AD, as well as the availability of agents to test the efficacy of CNS inhibition of neurotoxic KYN enzymes, specifically KMO. Peripheral measures of KYN metabolism in serum and plasma support the hypothesis that the KYN pathway is induced in AD, leading to increased production of KYN and products associated with the microglial metabolic branch. However, it is unclear whether this peripheral pattern of metabolism reflects CNS KYN metabolism or contributes to neural pathology. Indeed, distribution of metabolites in the CNS (brain, CSF) suggest a more nuanced pattern of regulation that also supports a shunting of tryptophan metabolism toward neurotoxic KYN metabolites, but only in proximity to plaques and tangles, indicating a need for brainpenetrable KMO inhibitors to act locally at the site of pathology. As discussed above, there is also a lack of longitudinal data on KYN metabolism in AD patients and an analysis of how these data correlate with disease progression or severity state. The limited correlation between blood and CSF metabolite profiles, as well as with post-mortem brain patterns, highlights the need for methodologies that correlate disease progression with direct measurements of brain KYN metabolites and/or enzymes in living AD patients. A concerted effort to develop positron-emission tomography (PET) ligands and/or other imaging agents targeting the KYN pathway would be a great benefit toward understanding the role of this pathway in AD and improving the rationale for the substantial investment required to develop brain-penetrant drugs.

While KMO has been targeted by the pharmaceutical industry for at least 3 decades, there are no published agents with good CNS drug exposure. Importantly, targeting KMO in the periphery creates a substantial increase in substrate, KYN, which is actively transported into the brain, where it is converted to substances such as KYNA, which could be neuroprotective, but also into 3-HK and Quin (Beaumont et al., 2016), which may increase oxidative stress and neuronal damage. This could be especially problematic in neurodegenerative disorders such as AD where neuroinflammation associated with plaques and tangles is predicted to increase KMO expression, thus more readily converting KYN to 3-HK and Quin (Guillemin et al., 2003a; Guillemin et al., 2005; Bonda et al., 2010). Therefore, it will be necessary to focus drug discovery efforts on identifying KMO inhibitor chemotypes with a high degree of CNS penetrance to test their therapeutic benefit in AD. Indeed, it may even be advantageous to attempt developing non-competitive KMO inhibitors along the lines of recent advances in KAT II inhibitors (Dounay et al., 2012) since they would not compete with increased KYN pumped into the brain from the periphery that could otherwise limit their efficacy. Finally, while the paucity of brain-penetrant drugs targeting the KMO branch of the KYN pathway is discouraging, advances have been made in developing agents that stimulate the neuroprotective KAT branch, such as 4-Chlorokynurenine (4CL-KYN; AV-101). Peripheral administration of 4Cl-KYN is transported into the brain, where it is converted to 7-chloro-KYNA, thereby acting as a locally synthesized KYNA mimetic (Hokari et al., 1996), where it appears to have promising neuroprotective effects (Wu et al., 2000; Wu et al., 2002). These data are encouraging and support the hypothesis that development of centrally acting KMO inhibitors could be neuroprotective in AD since blocking microglial KYN metabolism would not only prevent production of neurotoxic metabolites such as 3-HK and Quin but also indirectly increase endogenously produced KYNA. Thus, a greater understanding of the impact of rebalancing the neurotoxic and neuroprotective branches of the KYN pathway in AD may provide additional mechanistic evidence to support development of a new generation of neuroprotective brain-penetrant KMO inhibitors.

### The P2X7–NLRP3 Axis in Alzheimer's Disease

The P2X7–NLRP3 (nucleotide-binding oligomerization domain, leucine-rich repeat, and pyrin domain–containing inflammasome complex 3) pathway has received extensive attention, with several publications pointing to its potential role in neuroinflammatory disorders, from psychiatry to neurodegenerative diseases. Microglia express several cell-surface and cytoplasmic sensors of danger signals that react to invading "dangers" and cause neuroinflammation. The ATP-gated ion channel P2X7 is expressed abundantly in microglia and responds to extracellular ATP, which often occurs during tissue distress or damage and can lead to activation of the NLRP3 inflammasome complex, producing proinflammatory cytokines, IL-1β and IL-18 (Bhattacharya and Biber, 2016). As such, the P2X7–NLRP3 pathway has been the focus of intense drug discovery efforts: while brain-penetrant P2X7 antagonists have made progress and are in the clinic (**Table 2**), brain-penetrant, drug-like NLRP3 inhibitors are the focus of medicinal chemistry discovery programs.

P2X7 belongs to the P2X family of trimeric ligand–gated cation channels. Its activation by ATP allows for the influx of several cations, including Ca2+, Na+, and K+ (Bartlett et al., 2014). Even though P2X7 expression is abundant in microglia (and in peripheral immune cells), it is "silent" under normal physiology, where ATP concentrations do not reach the high micromolar levels required to activate its ion channel (Bhattacharya and Biber, 2016). As such, P2X7 is an attractive drug target as antagonism of a silent channel by true neutral antagonists would not cause any serious target mediated (adverse) effects: antagonism will only be evident when the channel is activated by high ATP concentrations, which is believed to occur during neuroinflammatory disorders of the CNS. NLRP3, on the other hand, is a convergent point of inflammasome activation, both dependent and independent of P2X7 activation (Bhattacharya and Jones, 2018), and as such is an attractive drug target. The advantages and disadvantages of targeting P2X7 vs. NLRP3 in neuroinflammatory disorders remain to be determined.

The P2X7–NLRP3 pathway is a case in point where microgliadriven neuroinflammation may play a role in AD causality (Heneka et al., 2013), and there has been growing interest in probing the role of pathogenic Aβ in neuroinflammation, although the exact mechanism of its involvement is still unknown (Illes et al., 2019; Martin et al., 2019; Thawkar and Kaur, 2019). P2X7 activation and consequent release of IL-1β is one such mechanism of interest in the AD field in the quest to understand the role of neuroinflammation in AD (Sanz et al., 2009). P2X7 upregulation has been reported in post-mortem AD brains and in animal models of tauopathy and amyloid deposition, mostly around the areas of pathology (Parvathenani et al., 2003; McLarnon et al., 2006; Lopez-Gonzalez et al., 2015). For example, P2X7 upregulation was seen in cortical samples from AD patients compared to healthy volunteers (Martin et al., 2019). With the recent discoveries of clinically available P2X7 PET ligands (Bhattacharya, 2018), the detection of P2X7 upregulation in humans becomes possible. In a mouse model of AD/ frontotemporal lobar degeneration (FTLD) (P301S), enhanced uptake of P2X7 PET signal was detected in brain regions of tauopathy (Jin et al., 2018). P2X7 activation has also been linked to neuronal damage and synaptotoxicity in transgenic AD models (Lee et al., 2011). There are signals of an active P2X7–NLRP3 axis in animal models of AD. For example, in the J20 mouse model of AD, which overexpresses the human APP gene with two familial AD-linked mutations, the authors noted that P2X7 activation prevents APP processing by α-secretase, facilitating the formation of toxic Aβ *via* P2X7-mediated activation of glycogen synthase kinase 3 (GSK-3) (Diaz-Hernandez et al., 2012). Supporting these data, inhibition of P2X7 with Brilliant Blue G (BBG) in the same mouse model increased α-secretase activity and reduced the formation of Aβ plaques, supporting a beneficial role of P2X7 antagonism in AD (Chen et al., 2014). Recently, it was reported that mice lacking P2X7 have reduced Aβ load and cognitive impairment in another AD model overexpressing both human mutated APP and mutated presenilin-1 (APP–PSEN1) (Martin et al., 2018). In addition to P2X7, the downstream intracellular NLRP3 inflammasome has been receiving special attention through several publications, and there is a growing interest to bring forward NLRP3 inhibitors for clinical testing in AD (Heneka, 2017; Heneka et al., 2018). For example, MCC950 (NLRP3 inhibitor) has been reported to prevent cognitive decline in APP/PS1 mice (Dempsey et al., 2017). A comprehensive study by Venegas et al. demonstrated mechanistically that ASC [adapter protein apoptosis-associated speck-like protein containing a caspase activation and recruitment domain (CARD)] specks (involved in NLRP3 activation) contributed to the spreading of Aβ plaques in APP/PSEN1 double-mutant mice (Venegas et al., 2017). The study by Martin et al. points to the unique role of P2X7, outside the NLRP3–IL-1β signaling cascade. While support for the NLRP3 pathway being involved in disease is gaining momentum, there are subtleties of targeting P2X7 vs. NLRP3 inflammasome with therapeutics. P2X7 antagonism may bring additional benefits as highlighted in the study by Martin et al. (2018). While NLRP3 is downstream of P2X7 and may be a critical mediator of IL-1β release through the inflammasome, P2X7 antagonism may provide additional benefits such as modulation of synaptic plasticity, modulation of chemokine release (e.g., CCL3), and consequent effects on CD8+ T cell recruitment to focal regions of the diseased brain.

#### Challenges Around P2X7/NLRP3

While the scientific community continues to generate data supporting the role of P2X7–NLRP3 in models of AD, there must be a healthy balance of caution and opportunistic investments in clinical trials to test these new mechanisms. The biggest challenge of NLRP3 for AD is a lack thus far of a suitable clinical compound with acceptable brain penetration. For P2X7, several brain-penetrant antagonists have been disclosed, some of which are in clinical development (**Table 2**). With the availability of brain-penetrant P2X7 antagonists and P2X7 PET ligands, these clinical tools could be tested in a patient population, perhaps with the help of PET and associated biomarkers (CSF, blood, genomics) that would enrich AD patients with an active P2X7– NLRP3 axis. In addition to using P2X7 PET ligands as imaging tools to enrich patient clusters for proof-of-concept testing, one must also demonstrate target engagement at clinically achievable doses. Target engagement in patients receiving a clinical candidate for proof-of-concept testing is critical to the evaluation of primary end points. Emerging data supporting the role of microglia-driven processes (neuroinflammation, phagocytosis), and description of different phenotypes of microglia in normal and disease states has prompted renewed thinking regarding their role in AD. To that end, it is of some comfort to witness that the role of microglia and microglia-driven neuroinflammation in AD has been the subject of some high-profile publications in the last few years, including significant progress with clinical compounds. While brain-penetrant P2X7 compounds are clinically available (**Table 2**), NLRP3 inhibitors that adequately demonstrate central target engagement (with NLRP3 PET ligands) are currently lacking. Demonstration of central target engagement is fundamental to drug development; once a doseoccupancy is established, moving to AD patients is the next step, with the challenge of selecting patients who can be intervened in early on in the disease process.

### Proposed Role of PD-1 Immune Checkpoint Blockade in AD

PD-1 (or CD279) is a member of the CD28 family of T cell regulators (Riley and June, 2005). It is weakly expressed on TABLE 2 | Clinical strategies used to dampen IL-1β signaling.


naive T cells but can be induced upon activation in several types of immune cells, including B cells, T cells, natural killer cells, dendritic cells, and monocytes. PD-1 plays an important role in immune responses related to B cell activity and/or T cell activity, such as antibody production, immune tolerance, and autoimmunity (Okazaki et al., 2013). PD-1 function is important to preserve immune tolerance. Taken together, PD-1 signaling ameliorates immune overreaction.

Recent studies have indicated that boosting the immune response in the brain might have therapeutic potential. It was shown that inhibition of forkhead box protein P3 (FOXP3) positive regulatory T cells increased interferon-γ (IFNγ)– dependent leukocyte trafficking to the brain, promoting clearance of Aβ and improving cognitive function in a mouse model of amyloidosis (Baruch et al., 2015). As PD-1 checkpoint blockade was previously reported to stimulate IFNγ-dependent immune responses in cancer immunotherapy, the therapeutic potential of PD-1 blockade in AD models was explored (Baruch et al., 2016).

In their studies, the authors treated the 5XFAD mouse model of amyloidosis at 10 months of age (an age with advanced pathology) with two intraperitoneal injections of a PD-1– specific antibody with 3-day intervals. One week after the first treatment, the mice were reported to exhibit a systemic IFNγ immune response. Further analysis of the myeloid cell populations in the brain upon treatment documented IFNγdependent recruitment of monocyte-derived macrophages as characterized by high expression of lymphocyte antigen 6c and expression of the chemokine receptor CCR2. Intriguingly, treatment also resulted in reduced cognitive deficits. In detail, spatial learning and memory in the radial-arm water-maze task was improved 1 month after treatment. Furthermore, 2 months after initial treatment, 5XFAD mice that had received 2 rounds of the anti–PD-1 treatment at a 1-month interval performed indistinguishably from non-transgenic mice. In comparison, mice with a single round of PD-1 blockade that were examined 2 months later were not different from placebo control and hardly showed improved memory. These results indicated that repeated treatment was needed to maintain improved cognition. Treatment with the PD-1–specific antibodies also caused a reduction in plaque load in the hippocampus and in the cortex. This treatment effect was more pronounced after two rounds of anti–PD-1 treatment. In addition, astrogliosis was reduced in the hippocampus of 5XFAD mice with either one or two rounds of anti–PD-1 treatment. The beneficial effects described in 5xFAD mice after PD-1 blockade were largely confirmed in a second mouse model of amyloidosis (APP/PS1APPswe, PSEN1dE9), in which treatment reduced plaque deposition in hippocampus.

Given the fact that various antibodies or small molecules targeting PD-1/PD-1L are already marketed or under development (Kobold et al., 2018; Shaabani et al., 2018), the findings by Baruch et al. triggered considerable interest in immune checkpoint blockade as a novel therapeutic strategy for AD. Accordingly, this study stimulated evaluation of the PD-1 immune checkpoint blockade for AD by three independent pharmaceutical companies.

### Challenges of PD-1 Immune Checkpoint Blockade as Therapeutic Intervention in AD

In subsequent studies jointly published in *Glia* (Latta-Mahieu et al., 2018) by Sanofi, Janssen, and Eli Lilly, the consequences of PD-1 immunotherapy in a range of different APP transgenic models (ThyAPP/PS1M146L, ThyAPP/PS1A246E, PD-APPAPPV717F) with the same anti–PD-1 antibody were documented. In addition to the antibody used by Baruch et al., two mouse chimeric variants and corresponding Immunglobulin G (IgG) controls were tested. In accordance with Baruch et al. (2016), PD-1 immunotherapy caused a systemic IFNγ-dependent immune response (Latta-Mahieu et al., 2018). However, in contrast to Baruch et al. (2016) there was no effect of anti–PD-1 treatment either on amyloid pathology or on monocyte-derived macrophage infiltration into the brain. This result point is consistent with data from another laboratory that explored the consequences of PD-1 deficiency in a murine prion disease model of neurodegeneration (ME7). In this study, there was no detectable infiltration of peripheral myeloid cells into the brain (Obst et al., 2018).

Latta-Mahieu and coworkers used, in addition to the rat mouse antibody (mAb) used by Baruch et al., two mouse chimeric variants with mouse IgG1 and IgG2a Fc domains as additional controls to evaluate whether observed effects were directly related to PD-1 target engagement. The *in vivo* studies were designed to ensure high statistical power (n = 9/group at Sanofi, n = 20/group at Janssen, and n = 25/group at Eli Lilly) and carried out in a blinded fashion. Furthermore, additional cohorts of animals for baseline measurement of pathology at the start of the anti–PD-1 treatment have been included. This way, age-dependent increases in amyloid pathology in each model of amyloidosis could be compared to anti–PD-1 treatment effects at the end of the study (Latta-Mahieu et al., 2018).

The reason for the different findings between Baruch et al. and Latta-Mathieu et al. is unclear, and more research is needed to understand the potential of PD-1 as a target for AD. It should be noted that in the study by Baruch et al., a xenogeneic rat antibody was used that potentially made the interpretation of the results difficult. Thus, it remained uncertain whether PD-1 target engagement or stimulation of the immune system, independent of PD-1, was causing the reduction in pathology and the functional improvements.

### The Proposed Role of Toll-Like Receptors in AD

The innate immune system is the first line of defense against pathogens and tissue damage and elicits a rapid and robust response. Within the CNS, the innate immune system performs a similar role that is mediated by microglial and astroglial cells (Ransohoff and Brown, 2012). In the case of CNS infection or injury, microglial cells detect pathogen-associated molecular patterns (PAMPs) or danger-associated molecular patterns (DAMPs) *via* specialized cell-surface pattern recognition receptors (PRRs). PRRs transmit information about the CNS microenvironment to the glial cells, which then orchestrate an appropriate immune response (Kigerl et al., 2014). Glia direct the neuroimmune response by migrating toward lesions or pathology, producing pro- and anti-inflammatory cytokines and chemokines, releasing growth factors and up regulating phagocytic mechanisms (Hansen et al., 2018). After a period, there is a resolution signal, which returns glia to their homeostatic state (Chavan et al., 2017). However, in AD, where Aβ, hyperphosphorylated tau, and neuronal damage persist for decades, there is no resolution, resulting in chronic activation of the innate immune system. This persistent alteration of glial cell phenotype is believed to contribute to AD progression and severity, and therefore, the identification of drugable molecular targets within the innate immune system of the CNS is of paramount interest.

TLRs are a major class of PRRs characterized by an extracellular leucine-rich repeat domain and an intracellular TIR domain (Kielian, 2006). Ten mammalian TLRs have been identified, each with a particular ligand specificity. Complicating the picture, TLRs can hetero-dimerize with additional co-receptors, resulting in an expansion of their ligand repertoire (Kielian, 2006). TLRs are expressed in microglia (Laflamme and Rivest, 2001; Laflamme Biber et al. Microglia as Drug Targets in AD

et al., 2001; Bsibsi et al., 2002; Dalpke et al., 2002; Olson and Miller, 2004; Zhang et al., 2012), and several have been shown to have increased expression both in the brains of AD transgenic animals and in human AD patients (Liu et al., 2005; Walter et al., 2007; Frank et al., 2009; Letiembre et al., 2009; Wirths et al., 2010; Rangasamy et al., 2018). The primary consensus among researchers is that persistent TLR signaling in AD models is detrimental; however, conflicting evidence also suggests beneficial roles of TLR signaling. The most extensively studied TLRs, in the context of AD, are TLR2 and TLR4 along with their co-receptor, CD14.

TLR2 and its co-receptor CD14 are upregulated in AD transgenic mouse models and in human AD brain sections (Letiembre et al., 2009). Functional characterization of TLR2 in primary mouse microglia and the mouse BV-2 microglia cell line demonstrated that fibrillar preparations of Aβ peptide can increase expression of NO synthase, pro-inflammatory cytokines, and integrin markers in a TLR2-dependent manner (Jana et al., 2008). Consistent with this, primary microglia and bone marrow–derived macrophages from TLR2 knockout mice show a reduced pro-inflammatory reaction and an enhanced phagocytic response when stimulated with aggregated Aβ1–42 peptide (Liu et al., 2012). Anti-TLR2 antibodies, administered over seven months, improved spatial learning, reduced gliosis [as assessed by immunoreactivity of CD68 and glial fibrillary acidic protein (GFAP)], and decreased Aβ plaque burden (McDonald et al., 2016). Anti-TLR2 antibodies can prevent a metabolic switch in microglia toward glycolysis, which results in an increase in phagocytosis (Rubio-Araiz et al., 2018). A peptide corresponding to the TLR2-interacting domain of myeloid differentiation primary response 88 (MyD88) (TIDM) disrupted the interaction of TLR2 with MyD88, resulting in a blockade of TLR2 signaling (Rangasamy et al., 2018). When administered intranasally to 5xFAD transgenic mice, the peptide could be detected in the hippocampus, where it reduced glial activation and Aβ plaque burden (Rangasamy et al., 2018). In contrast, several studies have postulated that TLR2 can promote beneficial responses in AD models. TLR2 deficiency accelerated spatial and contextual memory deficits while increasing levels of Aβ1–42 in the brains of PS1/APP transgenic mice (Richard et al., 2008). In addition, the TLR2 agonist peptidoglycan (PGN) enhanced uptake of Aβ1–42 in primary mouse microglia (Chen et al., 2006).

Like TLR2, TLR4 is expressed in microglial cells and can mediate binding and internalization of Aβ1–42 *in vitro* (Tahara et al., 2006). TLR4 was originally identified as an endotoxin receptor, mediating immune response to stimuli such as LPSs (Janssens and Beyaert, 2003). Studies using LPS as a tool to activate TLR4 signaling in AD transgenic mice have produced conflicting results. Intra-hippocampal administration of LPS to Tg2576 or PS1/APP mice activated microglia (as assessed by CD45 and CR3 or MHCII immunoreactivity) and reduced diffuse Aβ staining but not compact plaque (DiCarlo et al., 2001; Herber et al., 2007). Multiple doses of peripheral LPS improved cognitive function and reduced tau pathology in the P301S model (Qin et al., 2016). In contrast, repeated peripheral administration of LPS to WT mice was shown to increase Aβ levels, activate microglia, and induce cognitive impairment (Lee et al., 2008; Lee et al., 2012). Studies with in APPswe PS1dE9 and PS1/APP TLR4 knockout mice confirm an increase in plaque pathology (Tahara et al., 2006; Song et al., 2011), resulting in reduced glial activation and improved cognitive function (Song et al., 2011). Therapeutic blockade of TLR4 signaling using the cyanobacterial product CyP (Macagno et al., 2006) injected intracerebroventricularly (ICV) was able to reduce memory impairment and glial cell activation induced by ICV administration of Aβ oligomers (Balducci et al., 2017).

### Challenges Targeting TLRs

Over the last two decades, considerable attention has been paid to understanding the role of TLRs in innate immunity, but their role in AD initiation and progression remains elusive. As highlighted above, the transgenic models used, the timing of the intervention, and the end points measured can impact the conclusions made. AD is a complex disease, and the effects of modulating TLRs may differ depending on the stage of the disease during intervention. For instance, memory impairment was reduced after ICV injection of Aβ oligomers by pre-treatment with TLR2 and TLR4 agonists (Pourbadie et al., 2018). In contrast co-administration of TLR4 antagonist produced similar results (Balducci et al., 2017). Additional complexities in targeting TLRs arise from their ability to downregulate the immune response following repeated exposures to TLR agonists like LPS. This "endotoxin tolerance" has been described in animals and a variety of human diseases (Lopez-Collazo and del Fresno, 2013): this might confound interpretation of experimental data using TLR agonists. This is highlighted by recent work that demonstrated that the microglia response to LPS increases in aged WT animals compared to young WT mice. However, in the PS1/APP model, the microglia response to LPS is decreased in aged animals compared to WT (Go et al., 2016). This altered TLR signaling (i.e., tolerance) could affect the efficacy of TLR agonists or antagonists especially where chronic administration of the drug will be needed, such as in AD.

TLRs are expressed on a wide variety of cells, including peripheral immune cells and non-immune cells within the CNS such a neurons (Kielian, 2009). With traditional small molecule approaches, selectively modulating CNS and not peripheral TLRs is not feasible. Therefore, a suitable therapeutic window will need to be established for any TLR therapeutic developed. This will require a delicate balance between achieving the brain exposures needed to drive a therapeutically beneficial effect and minimizing peripheral exposures to reduce the risk for potential toxicities. The typically lower brain exposure compared to the periphery that is observed for many drugs adds to this challenge.

Several small molecule TLR agonists and antagonists have been described (Zhu et al., 2018). Discovery of TLR8 inhibitors was greatly facilitated by structure-based design (Hu et al., 2018). Similar approaches may be useful in designing small molecules for additional TLRs. Rather than drugging TLRs directly, others have sought to interfere with key protein– protein interactions required for TLR signaling. The TIDM peptide serves as a MyD88 decoy, essentially capping TLR2, preventing assembly of the myddosome and subsequent TLR2 signaling (Rangasamy et al., 2018). TAK-242 is a small molecule that interacts with cysteine-747 in TLR4, preventing interaction with TIR domain–containing adaptor protein (TIRAP) or TRAM resulting in blockade of TLR4 signaling (Matsunaga et al., 2011). Degrader approaches may also be useful to selectively shut down TLR signaling, but the use of such technology has not been demonstrated on TLRs. The therapeutic potential of TLRs in AD remains compelling, and the recent development of selective TLR ligands will support investigations of the role of TLRs in AD while helping to frame a robust therapeutic hypothesis.

### Proposed Role of TREM2 in AD

TREM2 is an immunoreceptor of partially understood function, expressed in myeloid cells in the periphery including dendritic cells, tissue macrophages and osteoclasts, and brain microglia (Colonna and Wang, 2016). TREM2 is a prominent AD risk gene (Pimenova et al., 2018) and is remarkable in the context of AD for at least two reasons: 1) its selective brain expression in microglia clearly links microglia specifically to neurodegenerative disease, and 2) there is the possibility of pharmacologic modulation of TREM2 to treat neurodegenerative disease.

In 2013, two landmark papers (Guerreiro et al., 2013; Jonsson et al., 2013) identified, by exome sequencing, rare mutants in TREM2 associated with an AD risk similar to that of a single allele of the strongest known risk factor, the E4 isoform of Apolipoprotein E (ApoE). Specifically, they showed that the mutation R47H, present in about 0.2% of the human population, elevated the AD odds ratio to at least 3.0. Several subsequent studies found slightly lower odds ratios but have confirmed the TREM2 R47H observations and identified additional TREM2 risk alleles and variants of unknown consequence (Bellenguez et al., 2017; Sims et al., 2017) that, in total, are present in about 1.5% of the human population (http://exac.broadinstitute.org/). Many other genes of modest AD risk have been identified before and since (Pimenova et al., 2018). But TREM2 is the strongest risk factor expressed in the brain only in microglia, unambiguously linking innate immune mechanisms in general, and microglial biology in particular, to AD. This observation has focused a great deal of attention on the role of microglia in AD and encouraged exploration of novel microglia-specific approaches to AD therapy (Hansen et al., 2018; Song and Colonna, 2018).

Functions of TREM2 in health and disease are summarized in several excellent reviews (for example, Hansen et al., 2018). AD-associated TREM2 mutants confer at least partially reduced ligand affinity, signaling, phagocytosis, and microgliosis (Ulrich et al., 2017). Significant evidence suggests that TREM2 is a direct phagocytic receptor (Takahashi et al., 2005; N'Diaye et al., 2009), which implies that failure to phagocytose aggregated proteins is a key driver of AD and related diseases. Note, however, that TREM2 is but one of many phagocytic receptors in microglia (Sierra et al., 2013), and it is unclear why the contribution of TREM2 in particular to phagocytosis would so strongly affect AD risk. Perhaps TREM2 has a prominent role in the phagocytosis of substrates specific to AD, e.g., Aβ or tau, or perhaps compromised TREM2 function has other pathological consequences (**Figure 4)**. TREM2 also seems necessary for microglial plaque association in mouse models and in human disease (Wang et al., 2016; Yuan et al., 2016). Functional studies have suggested that TREM2 is a lipid receptor, and that the binding of some but not all lipids is compromised by AD-associated mutations (Wang et al., 2015). Evidence for other ligands including ApoE (Atagi et al., 2015; Bailey et al., 2015; Yeh et al., 2016) and Aβ peptide (Lessard et al., 2018; Zhao et al., 2018; Zhong et al., 2018) has been presented. These results should be interpreted cautiously. Aβ peptide is markedly hydrophobic, making specific interactions difficult to interpret. Also, depending on the presence and nature of bound lipids, ApoE is likely to have different interaction properties. The key ligand(s) for TREM2 in AD may not yet be known, or TREM2 may simply be a promiscuous receptor (Kober and Brett, 2017). The phenotype of mouse ApoE knockouts mimics TREM2 knockouts with respect to microglial plaque association (Shi and Holtzman, 2018; Ulrich et al., 2018), supporting ApoE as a TREM2 ligand and suggesting a model in which TREM2-mediated effects on plaque may require or be enhanced by a ternary complex with ApoE. Structural studies provide some insight into the specific effects of AD risk of TREM2 mutations. R47, for example, is a surface residue that may constitute part of the TREM2 ligand binding surface (Kober et al., 2016). In mutant form (R47H), histidine substantially alters the conformation of a surface loop, which may change the energetics of ligand binding (Sudom et al., 2018). It should be noted that some TREM2 mutants, including ones that eliminate expression by chain termination, are associated with a different neurologic condition called Nasu–Hakola disease (NHD). AD-associated mutations change amino acids on the outer surface of the protein and are predicted to have subtle structural effects. In contrast, NHD sequence variants are buried in the core of the protein and are predicted to have profound effects on protein structure and expression (Kober et al., 2016). Thus, relatively conservative mutations increase the risk of AD, while protein elimination is associated with NHD.

Some challenges in the TREM2 field have slowed progress in our understanding of its role in AD. Many studies are in cells other than microglia—often, peripheral macrophages. Many studies are done with TREM2 knockouts. As noted above, such knockouts are associated with NHD rather than AD, and the mechanistic relationships between the two are essentially not known. Some studies rely on overexpression of TREM2 (e.g., Jiang et al., 2016; Lee et al., 2018), which risks overwhelming TREM2 signaling partners in these cells, with uncertain consequences. Much of this is understandable. A major problem is that the phenotype of the AD risk sequence variants is quite subtle, making differential readouts and clear conclusions difficult. Some investigators use heterozygous as well as homozygous knockouts to reveal trends that are complex to interpret. Likewise, the relationship between a heterozygous knockout and, for example, the R47H sequence variant is not

clear, and the latter is the preferable experimental target (for example, Cheng et al., 2018; Song et al., 2018). Compounding the TREM2 phenotype problem are difficulties common to all studies of microglia; these cells are extremely sensitive to their environment, rapidly changing their gene expression profile *ex vivo* (Bohlen et al., 2017). Human primary cells are available, of course, only from cadavers, and even then, only after several hours of hypoxia, with uncertain impact. Inducible pluripotent stem cell (iPSC)-derived "microglia-like" cells have, at best, an uncertain relationship with human microglia in their natural setting. Finally, due largely to the lack of reagents, intervention studies with pharmacologic modulators are rare: one such study using an antibody reported to activate TREM2 signaling caused significant increases in microglial survival and migration (Cheng et al., 2018). More studies with bettercharacterized reagents are needed.

### TREM2 as Drug Target: Challenges

Besides directing attention to microglia as an approach to AD drug discovery, TREM2 itself may have promise as a target if the appropriate pharmacologic modulation can be identified. As mutants that increase AD risk seem to have a partial loss of TREM2 function, and considering that overexpression studies with TREM2 seem to enhance its function in processes

associated with AD (Takahashi et al., 2005; Lee et al., 2018), one hypothesis is that augmenting the amount or function of TREM2 might oppose AD progression. However, increasing the function of a protein is challenging. Antibodies stabilizing TREM2 on the cell surface might be another possibility, including those that increase cell-surface TREM2 by blocking its proteolytic cleavage and release (Schlepckow et al., 2017; Thornton et al., 2017). Small molecule binders that stabilize TREM2 by blocking proteolytic cleavage or other mechanisms are conceivable but likely quite difficult to find. More promising, but further off, would be modulation of TREM2 regulatory factors. Such signaling or regulatory mechanisms specific enough to TREM2 to expect avoidance of undesired pleiotropic effects have not been reported. These may be revealed by ongoing research, including phenotypic screens for genes or compounds that boost TREM2 function. But as a caveat to this whole strategy, recall that TREM2 mutants that increase AD risk are quite rare. The vast majority of AD patients carry the common variant sequence of TREM2. There is no direct evidence in humans, and only limited evidence in model systems, that boosting the function of "normal" TREM2 would have beneficial effects for most AD patients. There is much to be learned about TREM2 and its promise as a drug target.

### CONCLUSIONS

To date, no disease-modifying treatment for AD has achieved Food and Drug Administration (FDA) approval despite many studies showing their efficacy in mouse models. Aβ and tau are key features of AD, and either could be a "disease-modifying" target in principle. Multiple failures to stop AD by facilitating Aβ clearance or inhibition of production have considerably increased the interest in other modalities to target this devastating disease, such as neuroinflammation.

Microglia were first recognized as key players in neurodegenerative diseases more than two decades ago. More recently, unambiguous genetic evidence clearly links microglia function to AD pathogenesis. A large percentage of diseaseassociated SNPs are present in genes that are specifically expressed in microglia, a finding that has boosted the interest in understanding how microglia could be targeted to treat AD. Accordingly, numerous microglia proteins or pathways have

### REFERENCES


been suggested to be drug targets. Here we reviewed several target candidates and their drugability challenges. Ideal drug targets should have a strong biological case, ideally combining genetic evidence with a well-understood mechanism of action that is supported by *in vitro* and *in vivo* data. However, even targets which fullfill this criteria need to be evaluated on whether they can be drugged safely. For the targets that have been briefly discussed in this review, these high hurdles have not yet been overcome. We remain optimistic that microglia-driven mechanisms will eventually bear clinical success and that one of the many discussed targets in this article will find a drug in the market helping AD patients.

### AUTHOR CONTRIBUTIONS

KB, AB, BC, JP, MR, RS, RT, TM wrote the manuscript. KB, RT, and TM edited the manuscript.

mitigates Alzheimer's disease pathology. *Nat. Commun.* 6, 7967. doi: 10.1038/ ncomms8967


Biber, K., Moller, T., Boddeke, E., and Prinz, M. (2015). Central nervous system myeloid cells as drug targets: current status and translational challenges. *Nat. Rev. Drug Discov.* 15 (2), 110–124. doi: 10.1038/nrd.2015.14


intermediate conductance KCa channel, IKCa1. *J. Biol. Chem.* 274 (9), 5746– 5754. doi: 10.1074/jbc.274.9.5746


synthetic serotonin uptake inhibitor. *J. Neural Transm. Park. Dis. Dement. Sect.* 9 (2-3), 121–131. doi: 10.1007/BF02259654


**Conflict of Interest Statement:** JP, RT, and TM are full-time employees of the Abbvie Foundational Neuroscience Center, Cambridge, MA, US. KB and MR are or have been full-time employees of AbbVie Deutschland GmbH & Co. KG, Ludwigshafen, Germany at the time of writing this article,. AB is a full-time employee of Janssen Research & Development LLC, San Diego, CA, US. BC is a full-time employee of Sage Therapeutics, Cambridge, MA, US. RS is a fulltime employee of Paracelsus Neuroscience, Metuchen, NJ, US.

*Copyright © 2019 Biber, Bhattacharya, Campbell, Piro, Rohe, Staal, Talanian and Möller. 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.*

# Neuroinflammation as a Common Feature of Neurodegenerative Disorders

*Leonardo Guzman-Martinez1, Ricardo B. Maccioni1,2\*, Víctor Andrade1, Leonardo Patricio Navarrete1, María Gabriela Pastor1,3 and Nicolas Ramos-Escobar1*

*1 Laboratory of Neuroscience, Faculty of Sciences, University of Chile & International Center for Biomedicine (ICC), Santiago, Chile, 2 Department of Neurological Sciences, Faculty of Medicine, University of Chile, Santiago, Chile, 3 Departamento de Biología, Facultad de Química y Biología, Universidad de Santiago de Chile, Santiago, Chile*

Neurodegenerative diseases share the fact that they derive from altered proteins that undergo an unfolding process followed by formation of β-structures and a pathological tendency to self-aggregate in neuronal cells. This is a characteristic of tau protein in Alzheimer's disease and several tauopathies associated with tau unfolding, α-synuclein in Parkinson's disease, and huntingtin in Huntington disease. Usually, the self-aggregation products are toxic to these cells, and toxicity spreads all over different brain areas. We have postulated that these protein unfolding events are the molecular alterations that trigger several neurodegenerative disorders. Most interestingly, these events occur as a result of neuroinflammatory cascades involving alterations in the cross-talks between glial cells and neurons as a consequence of the activation of microglia and astrocytes. The model we have hypothesized for Alzheimer's disease involves damage signals that promote glial activation, followed by nuclear factor NF-kβ activation, synthesis, and release of proinflammatory cytokines such as tumor necrosis factor (TNF)-α, interleukin (IL)-1, IL-6, and IL-12 that affect neuronal receptors with an overactivation of protein kinases. These patterns of pathological events can be applied to several neurodegenerative disorders. In this context, the involvement of innate immunity seems to be a major paradigm in the pathogenesis of these diseases. This is an important element for the search for potential therapeutic approaches for all these brain disorders.

Keywords: Alzheimer's disease, Parkinson's disease, tauopathies, neuroinflammation, microglia, astrocytes, proinflammatory cytokines

### CROSS-TALKS BETWEEN GLIAL CELLS AND NEURONS AND ORIGINS OF ALZHEIMER'S DISEASE

The German physician Alois Alzheimer discovered, in the beginning of the past century, a neuropsychiatric disorder, with clinical features of a dementia, called Alzheimer's disease (AD) after him. He analyzed the *postmortem* brain of an elderly woman with cognitive impairment and found anomalous structures which correspond to the intracellular neurofibrillary tangles (NFTs) formed by aggregates of hyperphosphorylated tau protein. These along with the oligomers of β-amyloid (Aβ) peptide became the major hallmarks of this disease. Along with these hallmarks, during many years of research, several factors have been elucidated, neuroinflammation being a key element in the development of the disease. In dementia, one of the most frequent is AD that affects mainly people

#### *Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Kyoungho Suk, Kyungpook National University, South Korea Roberta Ward, Imperial College, United Kingdom*

*\*Correspondence: Ricardo B. Maccioni rmaccion@manquehue.net*

#### *Specialty section:*

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

*Received: 24 February 2019 Accepted: 08 August 2019 Published: 12 September 2019*

#### *Citation:*

*Guzman-Martinez L, Maccioni RB, Andrade V, Navarrete LP, Pastor MG and Ramos-Escobar N (2019) Neuroinflammation as a Common Feature of Neurodegenerative Disorders. Front. Pharmacol. 10:1008. doi: 10.3389/fphar.2019.01008*

**214**

over 65 years old. Because of the expansion of life expectancy, AD has become a major health problem, with an estimated 50 million people all over the world having it (Bettens et al., 2010). According to the World Health Organization (WHO), AD progressively affects learning and memory as well as mood and behavior, displaying a constantly increasing prevalence and impact (Maccioni, 2012; Guzman-Martinez et al., 2013).

A major constituent of NFTs is a hyperphosphorylated form of the axonal protein tau, whereas a major constituent of senile plaques (SPs) is Aβ protein. SPs are extracellular deposits and correspond to deposition of Aβ peptides, derived from the amyloid precursor protein (AβPP) (Chapman et al., 2002). Aβ is generated by a sequential processing of the AβPP by two proteases and usually exported from the brain to the cerebrospinal fluid (CSF) and local degradation by microglia, the major constituent of the brain's innate immune system. In principle, microglia can engulf Aβ by phagocytosis (Heneka et al., 2015).

Hyperphosphorylated tau protein originally forms oligomeric structures called paired helical filaments (PHFs); then it turns into NFTs. The deposition of these structures causes loss of synaptic function and finally neuronal death (Giannakopoulos et al., 2003). Evidence supports the toxicity of tau aggregates when they are exported into the extracellular environment, along with being spread all over the brain (Neumann et al., 2011; Andrade et al., 2017). Studies of cell morphology and organelle distribution under tau overexpression show alterations in transport through the axis by motor axonal microtubuleassociated proteins (MAPs) (Cambiazo et al., 1995).

On the other hand, in AD pathophysiology, a key event is neuroinflammation in the central nervous system (CNS). Thus, in this review, we will focus on how neuroinflammatory processes are directly related to cognitive impairment and to the neurodegenerative processes, describing yet the implications of the involvement of both astrocytes and microglia in the inflammatory and neuro-immunomodulatory processes (Fernandez et al., 2008; Morales et al., 2010; Maccioni, 2011; Neumann et al., 2011). The microglial cells regulate the innate immune functions of astrocytes, under both physiological and pathological conditions; the inflammatory factors released by activated microglia can induce transduction of intracellular signals in astrocytes. On the other hand, the reactive astrocytes release factors that favor changes in the permeability of the blood– brain barrier (BBB), resulting in the recruitment of immune cells in the brain parenchyma. This leads to an amplification of the initial innate immune response. In turn, these reactive astrocytes secrete a wide range of factors, such as neurotrophic factors, growth factors, and cytokines, promoting neuronal survival, neurite growth, and neurogenesis. Both the microglia and the astrocytes release various signaling molecules, establishing an autocrine feedback. The feedback between both types of glial cells generates a close reciprocal modulation for various lesions in the CNS (Jha et al., 2019).

There are several neuroinflammatory factors that are involved in both the onset and the progression of AD. This process depends on the innate immune system which includes microglia and astrocytes (Maccioni et al., 2009). Residues from bacteria, viruses, fungi, abnormal endogenous proteins, iron overload,

complement factors, antibodies, cytokines, and chemokines, including toll-like receptors (TLRs) and receptor for advanced glycation end products (RAGE), comprise a large number of damage signals, which represent a danger for homeostasis of the CNS, and participate in microglial action and its activation (Shastri et al., 2013). Under these conditions, microglial cells regulate the expression of different surface markers, such as the major histocompatibility complex II (MHC-II) molecular pattern recognition receptors (PPRs), which produce cytokine proinflammatory drugs such as interleukin (IL) 1 beta (IL-1β), IL-6, IL-12, interferon (IFN) gamma (IFN-γ), and tumor necrosis factor (TNF) alpha (TNF-α). They also synthesize and release short-lived cytotoxic factors, such as superoxide radicals (O2 −), nitric oxide (NO), and reactive oxygen species (ROS) (Meda et al., 2001; Colton and Wilcock, 2010). Therefore, and in relation to the above, microglial cells have an important role in innate immunity and are the main source of proinflammatory factors in the human brain. The microglial activation process depends on phenotypic characteristics and is functionally diverse, because the response depends on the type, intensity, and context of the stimulus that generates it. The factors that affect microglia can also generate neuroprotection. Under pathological conditions, neurotoxicity will be expressed, due to the breakdown of the delicate balance between neurotoxic and neuroprotective effects.

Microglial cells exhibit ramified processes having high motility and allowing a dynamic and continual survey of the healthy brain as observed by using *in vivo* two-photon imaging (Nimmerjahn et al., 2005). They sample, detect, and eliminate debris or apoptotic neurons by phagocytosis, but this ability is considerably decreased in a proinflammatory context (Koenigsknecht-Talboo and Landreth, 2005). Microglia is involved in multiple processes such as neurogenesis, synapse elimination in a complementdependent manner, or synapse plasticity (Paolicelli et al., 2014). The involvement of microglia in AD pathogenesis was studied in the light of the Aβ (Guillot-Sestier et al., 2015; Heneka et al., 2015) and also in the context of tau oligomerization (Maccioni et al., 2009; Morales et al., 2010; Maccioni, 2012; Morales et al., 2014).

Another key factor is the accumulation of monocytes and microglia around blood vessels, due to the CCL2 chemoattractant protein and its affinity receptor CCR2. Studies showed that removal of the receptor increases the microglia accumulation phenomena, possibly through recruitment of mononuclear phagocytes and bone marrow, which promotes the deposition of perivascular Aβ (Ransohoff, 2016b). Care should be taken, since most experiments of circulating monocytes include conditions in which the BBB is open by irradiation procedures in AD, leading to controversy. Interestingly, the reduction of monocyte infiltration following ccr2 deficiency has been involved in tau hyperphosphorylation in traumatic brain injury (TBI) (Ransohoff, 2016b).

It is known that Aβ oligomers induce the activation of microglia through oligomers-surface receptors such as TLRs, being part of a physiological duty to eliminate them *via* phagocytosis (Walter et al., 2007). Certain receptors are associated with the reduction of microglial Aβ phagocytic capacity like triggering receptors expressed in myeloid cells 2 (*TREM*2), whose specific missense mutations increase the risk of AD (Jonsson et al., 2013; Parhizkar et al., 2019). Deficiencies in receptor CX3CR1, a chemokine CX3CL1 microglial receptor which partly mediates the infiltration of monocytes, induce overexpression or activation of microglia and tau hyperphosphorylation (Maphis et al., 2015), increased CD33 with a specific single-nucleotide polymorphism (SNP) in the promoter which inhibits immune response promoting Aβ1–42 accumulation (Griciuc et al., 2013), B3 domain-containing transcription factor ABI3 (*ABI3*) (Sims et al., 2017), and several other factors.

On the other hand, some variants in phospholipase C gamma 2 (*PLCG2*) have a positive outcome for AD, reducing the late onset of the disease (Sims et al., 2017). Friedman et al. (2018) determined the gene expression profile associated with neurodegeneration, where 75% of these genes are linked with gene ontology (GO) related to plasmatic membrane. Altogether, this information and the effect of mutations in several receptors and other plasma membrane proteins suggest that changes depend on the interaction with the environment. The paper of Keren-Shaul et al. (2017) describes a new kind of microglia, the disease-associated microglia (DAM) that only gets expressed in AD. The study determined several gene modules. In DAM, a neurodegeneration gene core is expressed. Other modules include the IFN gene. In AD animal models, there are abundant cells that express the IFN module. The DAM gene expression changes as follows: there is downregulation of homeostatic genes like CX3CR1 and upregulation of genes associated with the disease like Apoε and phagocytic genes for plaque clearance. Every microglia has promoters and enhancers associated with DAMs, indicating that these stage changes might pass through an epigenomic change. They also discovered the three stages of microglia: homeostatic, intermediated, and finally DAM stage through an unknown Trem2-independent mechanism. In microglia gene expression modules, lipopolysaccharide (LPS) and neutrophil/monocyte are exacerbated, suggesting that inflammation and infiltration elements are involved in the neurodegenerative disease (ND).

Along with microglia, astrocytes are involved in the neuroinflammation process. Astrocytes have roles in metabolic regulation, neuronal scaffold, and synaptogenesis. In addition, there is a close contact with microglia and blood vessels in BBB (Morales et al., 2014). It also participates in the clearance of Aβ, by enzyme secretion (Mulder et al., 2012), and APOε from the ε2 allele is considered a protective factor. (Koistinaho et al., 2004). Like microglia, astrocytes also surround Aβ plaques (Medeiros and LaFerla, 2013), turning into an activated phase. Calcium deregulation, expression of the APOε4 allele, gives rise to APOε4 activity, which does not affect the synthesis of Aβ but does increase the deposition of the same, meaning a defect in the Aβ clearance (Holtzman et al., 2000). The astrocytes can be activated through a pathway involving NF-κB, to release a C3 complement which binds to the C3aR receptor, inducing neuronal damage (Lian et al., 2015), along with soluble CD40, which binds to microglia and induces the release of TNF-α and other proinflammatory cytokines (Frankola et al., 2011). In neuroinflammation, astrocytes also contribute to NO toxicity, by expressing inducible NO synthase (iNOS) (Phillips et al., 2014). Besides, there is overexpression of the glial fibrillary acidic protein (GFAP), a protein essential in the astrocyte cytoskeleton, related to astrocyte activation (Hol and Pekny, 2015). In the tauopathy context, Aβ can bind to the calcium sensing receptor (CaSR) in astrocytes, which triggers signaling pathways involved in the production and release of phosphorylated tau (Chiarini et al., 2017).

### NEUROINFLAMMATION IN AD

Neuroinflammation is a process related with the onset of several neurodegenerative disorders and it is an important contributor to AD pathogenesis and progression. Several damage signals appear to induce neuroinflammation, such as trauma, infection, oxidative agents, redox iron, oligomers of tau, and Aβ. In effect, neuroinflammation is responsible for an abnormal secretion of proinflammatory cytokines that trigger signaling pathways that activate brain tau hyperphosphorylation in residues that are not modified under normal physiological conditions. Indeed, evidence exists that AD pathogenesis is not restricted to the neuronal compartment but includes strong interactions with immunological cells in the brain such as astrocytes, microglia, and infiltrating immune cells from the periphery, which could contribute to the modification of the process of neuroinflammation and neurodegeneration in AD brains. In this context, this is where our theory of neuroimmunomodulation plays an important role and focuses on the link between neuronal damage and brain inflammatory process, mediated by the progressive activation of astrocytes and microglial cells with the consequent overproduction of proinflammatory agents (Maccioni et al., 2009). Despite clinical and pathological differences, increasing experimental evidence indicates that neuroinflammatory events lead to tau protein misfolding (Cortes et al., 2018).

The participation of the innate immune system in disease progression has shown a harmful bidirectional connection with regard to tau pathology. It is known that the tau protein belongs to the family of MAPs and is expressed mainly by neurons with preferential axonal localization. It has been observed that tau *in vitro* promotes the polymerization of tubulin and decreases the transition rate between the phases of growth and contraction, generating a stable but dynamic state in microtubules (Weingarten et al., 1975; Drechsel et al., 1992).

Tau is found mainly in axons, but a small amount is distributed physiologically in dendrites. The postsynaptic function of tau is not yet well defined, but it may be involved in synaptic plasticity. On the other hand, in addition to axons and dendrites, a nuclear function of tau (Citron, 2010) has been discovered, which could be regulating transcriptional activity and maintaining DNA/RNA integrity under physiological and stress conditions (Weingarten et al., 1975; Violet et al., 2014).

The tau structure corresponds to a hierarchical phosphorylation process in which different sites modulate the conformation of the protein, promoting the action of secondary kinases. In AD, different sites are phosphorylated earlier than others, leading to the creation of new epitopes. This sequential process has been studied by the use of antibodies such us AT100, whose epitopes in PHFs only appear after successive phosphorylation of residues Thr212 and Ser214, by glycogen synthase kinase (GSK)-3β and protein kinase A (PKA) along with Ser199, Ser202, Ser208, and Thr205 (Bussiere et al., 1999; Malia et al., 2016).

It was also shown that the expression of tau by microglial cells promotes its activation (Wang et al., 2013). Overall, the exact pathway leading to phosphorylation of tau remains poorly defined, but subsequent structural changes induce its detachment from the microtubules and produce higher levels of soluble free tau. Before the formation of NFTs, the hyperphosphorylation of tau favors a dynamic and progressive self-assembly of tau in oligomeric forms and insoluble materials such as PHFs throughout the disease with different degrees of neurotoxicity (Braak and Braak, 1991).

### NEUROINFLAMMATION IN SEVERAL TAUOPATHIES

### Neuroinflammation in the Context of Tau and Tauopathies

These neurodegenerative disorders (tauopathies) do not have a defined clinical, biochemical, and morphological characteristic, like other diseases. Neurodegenerative disorders are distinguished by accumulation of misfolded proteins, such as α-synuclein (α-syn) protein in Parkinson's disease (PD) and tau protein in AD (Cortes et al., 2018). An important group of degenerative diseases are the so-called tauopathies, which consist in the pathological accumulation of tau protein in intracellular fibrillary aggregates. The spectrum of tauopathies covers a large number of disorders such as progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), frontotemporal dementia (FTD), FTD and parkinsonism linked to chromosome 17, chronic traumatic encephalopathy, and argyrophilic grain disease (Spillantini and Goedert, 2013; Arendt et al., 2016). On the other hand, this last pathology is identified by atrophy of the ambient gyrus and presence of argyrophilic and 4R-tau immunoreactive grains in medial temporal-lobe structures (Tolnay and Clavaguera, 2004). On the other hand, AD is considered a secondary tauopathy because it also presents aggregates of Aβ (SPs) (Musiek and Holtzman, 2015). On the other hand, the pathological aggregates of tau protein, cells of the cerebral immune system, such as activated astrocytes and microglia, are other common pathological features of tauopathies (Wyss-Coray and Mucke, 2002; Ransohoff, 2016a). The existence of neuroinflammatory processes exacerbated in various tau pathologies, known as the *theory of neuroimmunomodulation*, was initially described in AD by Dr. Maccioni's group, in which the bases of the molecular cascades associated with these events were laid (Maccioni et al., 2009). Recent studies have discovered multiple mechanisms by which an overstimulation of glial cells causes a harmful neuroinflammation that would influence the tau pathology and accelerate the neurodegenerative processes. The chronic activation of glial cells alters the physiological function of the tau protein, inducing the activation of enzymes that phosphorylate tau, such as the enzymes CDK5 and GSK-3β, giving way to the formation of NFTs, thus decreasing the neuronal capacity (Wyss-Coray and Mucke, 2002; Lull and Block, 2010; Ransohoff, 2016a). In addition, glial cells can also contribute physically to the spread of tau pathology (Asai et al., 2015). In turn, the glial cells are also positively fed back by the tau pathology, since the degenerating neurons and their axons and dendrites release aggregated and toxic tau species, generating a constant neuroinflammatory cycle (Morales et al., 2010; Morales et al., 2013; Cortes et al., 2018).

As indicated, activated microglia release proinflammatory cytokines to their cell environment, among which we can highlight IL-1β, IL-6, IL-12, IFN-γ, and TNF-α. In turn, they can also produce ROS and NO, among others that may be characteristic of neurodegenerative disorders (Wang et al., 2015). On the other hand, the astrocytes are the main and most numerous of the glial cells; they are fundamental to supporting the function and health of the neuronal cells. In turn, astrogliosis can also be considered as an important factor in chronic neuroinflammation, affecting considerably the neuron and its integrity (Sofroniew and Vinters, 2010). As microglia, astrocytes also synthesize and secrete proinflammatory cytokines. In addition, experimental evidence has indicated that ILs such as IL-1β, TNF-α, IL-6, and C1q (secreted by the microglia) coactivate the astrocytes, resulting in neuronal dysfunction and ultimately death (Jacobs and Tavitian, 2012). In contrast, in PSP, only IL-1β increased significantly in the *substantia nigra* and in the *subthalamic nucleus* (Fernandez-Botran et al., 2011; Lopez Gonzalez et al., 2016).

Currently, there is an increase in the evidence that the pathological activation of both microglia and astrocytes causes chronic neuroinflammation in patients with tauopathies, negatively affecting the progression of the disease, although the first signs of neuroinflammation considered reactive gliosis in tauopathies and other NDs (Leyns and Holtzman, 2017).

### Frontotemporal Dementia

More than a new pathological entity is the redefinition of the classic Pick's disease. The term FTD is used to mean all those primary degenerative processes of the anterior portion of the brain, characterized by their clinical manifestations, neuroimaging findings, and histopathological elements, which are of particular importance for psychiatry, due to the tendency of patients to present behavioural disorders, being a frequent cause of dementia (Neary et al., 1988). FTD is a heterogeneous syndrome that involves several disorders that originate mainly in the frontal lobe and temporal areas of the human brain. These alterations can cause problems in language and motor and behavioural disorders (Neary et al., 1998; Olney et al., 2017). In terms of prevalence, after AD, FTD is considered the second most important type of neurodegenerative disorder (Knopman and Roberts, 2011; Hughes et al., 2015). The onset of FTD is between 45 and 65 years of age, with a survival range of between 2 and 20 years, averaging 8 years. In turn, demographic data indicate that the distribution of this disease is similar among men and women (Hodges et al., 2003; Onyike and Diehl-Schmid, 2013; Coyle-Gilchrist et al., 2016), with a prevalence between 15 and 22 people per 100,000 inhabitants, varying according to the age of onset (Irwin et al., 2015). On the other hand, FTD associated with pathological tau represents 36–50% of all cases of FTD (Bang et al., 2015). There are different subtypes of FTD, which are directly related to an existing clinical classification, establishing the following main forms of FTD: the behavior variant of FTD (bvFTD), the nonfluid variant (nfFTD), and the semantic variant (svFTD).

Both nfFTD and svFTD are classified as primary progressive aphasia, since they mainly affect language functions, according to the criteria of clinical diagnosis. The type of FTD that presents the greatest insight is bvFTD, which covers about 60% of cases (Rascovsky et al., 2011; Onyike and Diehl-Schmid, 2013; Olney et al., 2017). Likewise, other alterations in the FTD category are related to motor neuron disorder (MNFTD), PSP-FTD, and corticobasal syndrome (CBS) (Wang and Mandelkow, 2015; Olney et al., 2017). In the same context, frontotemporal lobar degeneration (FTLD) is based on the neuropathological alterations that occur in the frontal and temporal lobes. Different types of FTD are determined by protein aggregates. In this way, the protein aggregates associated with FTD are the following: (i) tau protein (FTLD-tau); (ii) ubiquitin proteasome system (FTLD-UPS); (iii) transactive response of the DNA binding protein (FTLD-TDP); and (iv) fused in sarcoma, Ewing sarcoma, and TAF15 protein family (FTLD-FET) (Mackenzie et al., 2010; Irwin et al., 2015; Olney et al., 2017). For the purpose of this review, we focus on FTDs caused by pathological inclusions of tau, which correspond to approximately half of FTDs, whose histopathological hallmark corresponds to aggregates of tau protein in both neurons and glial cells (Sieben et al., 2012). These include Pick's disease (PiD-FTD), CB, PSP-FTD, and other rare FTDs, described as globular glial tauopathies and argyrophilic grain disease, which are included in former classifications (Mackenzie et al., 2010; Ghetti et al., 2015; Olney et al., 2017). The pathologies described previously are principally characterized by repetitions of 4R-tau sequences; on the other hand, PiD-FTD are related to 3R-tau aggregates (Ghetti et al., 2015).

Previously, a study reported an increase in the levels of cytokines associated to inflammation, specifically TNF-α and transforming growth factor (TGF)-β in subjects affected by an unspecified form of FTD as compared to healthy people, suggesting a likely interference of inflammatory proteins in the pathogenesis of FTD that was immediately verified by Bellucci et al. (2004); this group reported a considerable rise of cytokines as IL-1 and cyclooxygenase-2 (COX2), both being involved in proinflammatory response; also, studies in transgenic mice with mutation of tau show activation of microglia cells with tau inclusions in the brainstem and spinal cord. Moreover, studies conducted in transgenic mouse with mutation of human tau associated with FTD model reported microgliosis and synaptic disruption, prior to formation of NFT in the hippocampus, suggesting that inflammatory response can carry over to the formation of NFT in FTD (Yoshiyama et al., 2007), in accordance with our theory of neuroimmunomodulation (Fernandez et al., 2008; Rojo et al., 2008; Maccioni et al., 2009). In turn, several trans-models with the FTD tau mutation model have also reported a microglial activation with the subsequent inflammatory process, which further emphasizes that these alterations directly depended on the expression of tau (Wes et al., 2014). In this way, the neuroinflammatory process has been proposed as possible diagnostic tools, through *in vivo* uptake of the microglia, using positron emission tomography (PET) images with the translocator protein (TSPO) ligand [11C]-PK11195 in the FTD and other tauopathies (Cagnin et al., 2004; Venneti et al., 2009; Zhang, 2015).

### Progressive Supranuclear Palsy

It is a rare neurodegenerative disorder that is increasing over time. It affects movement, walking, balance, speech, swallowing, vision, mood, behavior, and thinking. One of the classic signs of the disease is the inability to focus and move the eyes correctly, which people can manifest as blurred vision. The prevalence of PSP is 5.8 to 6.5 per 100,000 (Ling, 2016). Like CBD, PSP presents hyperphosphorylated 4R-tau in neurons and glial cells. PSP is defined primarily by tau-positive NFTs, coiled bodies, threads, and tufted astrocytes, in contrast to the ballooned neurons, pre-tangles, threads, and astrocytic plaques that are characteristic of CBD (Yoshida, 2014).

Imaging studies used conventional magnetic resonance; atrophy can be seen at the level of the midbrain and superior cerebellar peduncle. In turn, when using diffusion tensor, the white matter of degeneration can be appreciated, especially in the superior cerebellar peduncles and the superior longitudinal fasciculus in the case of Richardson's syndrome; dopamine transporter singlephoton emission computed tomography imaging shows reduced tracer uptake in the striatum; finally, fluorodeoxyglucose PET may identify focal midbrain hypometabolism (Yoshida, 2014; Ling, 2016; Kovacs, 2017).

The NFTs of tau in subcortical structures are a characteristic sign of PSP. These pathological aggregates of tau are located especially in the subthalamic nucleus, the basal ganglia, and the brainstem. The subcortical NFTs are associated in a variable but characteristic way, with astrocytes in tufts and spiral oligodendroglial bodies, as well as with threads, which present immunoreactivity for the isoform of tau 4 repetitions (4R-tau). Studies conducted by Williams et al. (2007) identified that certain brain regions are affected by these pathological tau aggregates. In this context, the pallido-luyso-nigral system is affected early, followed by the basal ganglia, the pontine nuclei, and the dentate nucleus; then the frontal and parietal lobes; and finally other neocortical areas and cerebellar structures (Williams et al., 2007; Kovacs, 2017). When doing comparative studies between CBD and PSP, a greater amount of pathological tau is observed in neurons in the anterior brain in CBD, whereas in PSP the structures of the posterior brain are mainly affected (Dickson, 1999). In addition, the different astroglial pathology associated with the pathological tau protein, associated with the presence of subcortical NFTs in the PSP, facilitates its neuropathological differentiation (Clavaguera et al., 2013).

### Corticobasal Degeneration

This disease, a rare and progressive neurodegenerative disorder that affects about 4.9 to 7.3 per 100,000 of the population (Mahapatra et al., 2004), is characterized pathophysiologically by neuronal loss, asymmetric frontoparietal cortical atrophy, gliosis, and swollen achromatic neuronal cell bodies (Rebeiz et al., 1968; Gibb et al., 1989). However, the exact cause of CBD still remains unknown, and although CBD is considered a sporadic disease, there have been some reports of families with pathology similar to CBD (Uchihara and Nakayama, 2006) or mutations in tau protein, a gene linked to pathological findings similar to CBD (Mirra et al., 1999; Spillantini et al., 2000). The average age at which the symptoms begin to manifest in the affected subjects, due to this disease, is close to 62 ± 7 years. The average survival can vary from 2 to 13 years (Rinne et al., 1994c; Schneider et al., 1997; Wenning et al., 1998; Josephs et al., 2006a). The youngest case, according to the pathological confirmation, began at age 45 (Wenning et al., 1998). Demographic studies indicate a higher incidence of this disease in women compared to men of the same age (Rinne et al., 1994c).

CBD can be characterized by the following signs and symptoms: action and postural tremor, resting tremor, bradykinesia, myoclonus, ideomotor apraxia, exotic extremity phenomena, extremity dystonia, gait deterioration, dysarthria, aphasia, speech apraxia, and/or dementia; these symptoms can be presented, in turn, in combination (Wenning et al., 1998). In this disease, cognitive deterioration is more affected in the area of speech and language (Josephs, 2010) and, in a lesser extent, visuospatial and perceptual deficits (Tang-Wai et al., 2003a; Bak et al., 2006). The CBD is molecularly characterized by deposits of hyperphosphorylated 4R-tau protein; this makes it possible to differentiate it from other tauopathies, which consist mainly of 3R-tau or a mixture of 3R- and 4R-tau (Trojanowski and Dickson, 2001). Brain neurons with CBD react positively with antibodies generated against ubiquitin, hyperphosphorylated tau, phosphorylated neurofilament protein, alpha-B crystal, and synaptophysin. In addition, aggregates of Aβ have been found in some cases of CBD, which are similar to those found in AD (Armstrong, 2015).

Neuropathological CBD alters the anatomical pathways associated with movement control: (i) striatum and (ii) *substantia nigra*, which usually presents with loss of pigmented neurons. In turn, it also presents at the neuronal level cytoplasmic inclusions and neuronal bodies with greater volume, compared with normal neuronal cells, also presenting glial pathology that affects both oligodendroglia and astrocytes (Armstrong et al., 2000). These damages are especially complex in the posterior frontal area anterior to the precentral gyrus, but they are lesser in the primary motor area (Tsuchiya et al., 1997). These neuronal hallmarks of CBD are found in several brain regions, including the superior temporal gyrus, the frontal cortex, the motor cortex, the brainstem tegmentum, the basal ganglia, the thalamus, and the amygdala (Halliday et al., 1995; Matsumoto et al., 1996).

The experimental evidence shows that the dysregulation of proinflammatory cytokines is one of the most harmful factors in tauopathies. Several investigations have been carried out, using neuroimaging technologies by means of PET, in order to examine more deeply the neuroinflammatory events in the neurodegenerative process. These studies have been used with markers that bind to the TSPO, which is expressed by glial cells (microglia and astrocytes) and other cells of the immune system infiltrating the brain. These findings have shown that the TSPO signal increases proportionally with the activation of microglia in various tauopathies, which include AD, PSP, FTD, and FTDP-17 (Maeda et al., 2011; Zhang, 2015), as well as other NDs, and in models of brain injuries such as FTD, PD, stroke, and traumatic brain injuries (Wu et al., 2013).

The overexpression of IL-1β, TNF-α, and IL-6 has one positive feedback, which generates a cascade that leads to an increase in the hyperphosphorylation of the tau protein, reducing the markers of synapses and finally leading to degeneration and neuronal death (Morales et al., 2014). Currently, less information is available about the levels of cytokine transcription in other less common tauopathies; However, the histological characteristics of activated glial cells are a common characteristic of tau protein aggregates (Morales et al., 2010; Leyns and Holtzman, 2017; Cortes et al., 2018).

In summary, tauopathies including AD involve the neuroinflammatory cascade that leads to tau modifications and subsequent oligomerization. However, in PD, the neuroinflammatory cascade seems to trigger α-syn oligomerization, an event that is critical in PD (**Figure 1**).

### FUTURE DIRECTIONS: SEARCH FOR THERAPEUTIC TARGETS BASED ON THE NEUROIMMUNOLOGICAL MODEL

### Alzheimer Disease

According to the information collected, the microglial cells become one of the points of convergence in the development of the neuroinflammatory process, since not only did the activation process and the changes that occur determine this condition, but it was also shown that a permanent activation can increase the permeability of the BBB and promote an increase in macrophage infiltration peripherals These alterations can complicate the vicinity of the damaged area and contribute to neuronal dysfunction, thus accelerating the neurodegenerative process (Zarrouk et al., 2018).

Within the various factors that modulate the phenotypes activated by the glia, the APO immunomodulatory genotype and the recently identified AD genes can be included. Another important factor to mention is normal aging, which is also associated with the chronic activation of the glia (Zverova et al., 2018) and the factors induced by lesions depending on the focal stage. Context-dependent responses must be expected for nonsteroidal anti-inflammatory drugs (NSAIDs), which act as inhibitors of cyclooxygenase by reducing the concentrations of prostaglandin products, in particular PGE2, by tapping PTGER1–4 receptors and producing very different results. One can highlight the activation of PTGER2, which is predominantly dedicated to proinflammatory neurotoxic pathways, whereas the ligands of PTGER4 have been found to produce anti-inflammatory and neuroprotective effects (Salmon et al., 2015; Alzheimer's Association, 2018). Thus, NSAIDs that inhibit conventional cyclooxygenase could cause a blockage of the incipient pathogenesis of AD driven by inflammation in the early stages. In addition, these NSAIDs may have adverse effects in advanced disease, mainly by restriction of resolution and interference with phagocytic clearance of Aβ and extracellular tau aggregates (McKee et al., 2008).

Recent findings suggest for anti-Aβ immunotherapy that stage-dependent efficacy also stimulates microglial phagocytosis of Aβ, and the potential benefits could be seen only with early intervention. A remarkable argument is that aging or inflammation induced by Aβ or by lesions initiates tauopathy, leading to neurodegeneration and subsequent clinical decline. Therefore, possible explanations for the failure of immunotherapy to treat established dementia include an inability to stop the spread of fully established and planted tauopathy and to rescue deficits caused by the loss of neurons. Whatever the explanations for the failures of previous NSAID trials, based on new and convincing genetic evidence of the causal role of innate immunity in AD risk, new trials with longer and earlier interventions and alternative approaches to treatment are warranted.

There is evidence that neuroinflammation could drive the pathogenic process in AD. In this context, it should be noted that the brain can no longer be seen as an organ with immune privileges, and advances in immunology must be integrated into the known pathological pathways of various neurodegenerative disorders. Ligand–receptor interactions in the CNS microenvironment, which keep the microglia under strict control in the healthy brain, may be disturbed in chronic ND, although it is not yet clear when and how this occurs in AD. While it is true that the simple idea of activated microglia has been useful, it has undoubtedly prevented the understanding and recognition of the diversity of microglial phenotypes and the extraordinary plasticity of these cells. An important approach that future studies should have should be to better understand the individual contributions of microglia and other cell types to the neuroinflammatory response during the course of AD (Maccioni, 2012; Guzman-Martinez et al., 2013; Morales et al., 2013; Morales et al., 2014).

Innate immune cells of the brain can respond rapidly to systemic events, and these responses are further accentuated in the stages of aging and the diseased brain. In future studies, the effect of systemic comorbidities of AD (such as diabetes and hypertension), associated systemic inflammation, and aging as an important risk factor for AD should be considered in order to understand and exploit the immunological processes associated with AD. The recognition that the modification of the immune system contributes to the pathogenesis of chronic NDs could open several pathways in potential treatments to delay its onset and progression (Morales et al., 2014; Andrade et al., 2017; Cortes et al., 2018).

On the other hand, to date, treatments for AD have been have established based on the nature of the symptomatic AD and, in most cases, are used to counteract the disturbance produced in the level of the neurotransmitters involved in this disease. Therapies based on tau protein suggest that this protein is an interesting target, because the formation of PHFs constitute a critical event in the neurodegenerative process.

An important research suggests that the anti-inflammatory activity may be controlled with a natural formula which contains the Andean compound, a natural product endemic to the north of Chile, and vitamins of the B complex (i.e., B6, B9, and B12), called BrainUp-10®. Interestingly, according to studies done by Cornejo et al. (2011), the active principal of this compound, fulvic acid, is able to block auto-aggregation of tau *in vitro* that inhibits the length and morphology of the PHFs generated. Indeed, this compound can disassemble the preformed PHFs and oligomers and tau species released into the extracellular environment. This natural compound is a potent anti-inflammatory substance and a biologically safe nutraceutical. In a clinical trial pilot, patients with AD who underwent treatment with this compound showed less tendency towards cognitive deterioration, in addition to a reduction in neuropsychological symptoms and less neuropsychiatric stress for caregivers of patients. The Andean compound is a complex mixture of humic substances generated by decomposition of plant material through thousands of years (Carrasco-Gallardo et al., 2012). Currently, there are therapeutic approaches based on the use of antioxidant and antiinflammatory nutraceutical compounds, a multi-target therapy which appears to have benefits as compared with the monotarget approach using drugs, thus contributing to the health and quality of life of AD patients. Studies based on this type of natural compounds have been growing, as well as the search for natural antioxidant compounds with a strong anti-inflammatory activity and the ability to cross the BBB. Therefore, an important strategy to prevent brain damage is based on changes in lifestyles, diet, and science-based nutraceuticals, among other multiple factors. However, there is a need for further investigations, in order to medically validate this natural approach (Andrade et al., 2017; Cortes et al., 2018).

### Treatments for Tauopathies

Drugs such as cholinesterase inhibitors and *N*-methyl-daspartate receptor antagonist are being used to treat cognitive problems. On the other hand, speech and physical therapies have shown positive effects on patients with aphasia syndromes and alterations in motor function, respectively. Serotonin reuptake inhibitors (SSRIs) have been used as a therapy in cases of apathy and depression (Halliday et al., 1995). In addition, specifically, medicaments such as sertraline, paroxetine, and fluvoxamine, which may improve behavioral and psychiatric symptoms but not cognitive impairment, have been selected to treat FTD due to the serotonergic deficit involved (Yoshida, 2014). Citalopram has been reported to lead to an improvement in behavioral symptoms after treatment (Irwin et al., 2015). Acetylcholinesterase inhibitors (AChEIs) are well-established treatments of AD, but until now, there is no report of their efficacy in patients affected by FTD (Kempster et al., 2007). Memantine, which has some effects in advanced AD and neuroprotective characteristics (Ling, 2016), could improve behavioral disorders and enhance metabolic rate in certain brain regions (Williams et al., 2007).

Treatment of cognitive and other symptoms of PSP based on serotonergic drugs has shown to be ineffective, despite its positive effect on depression (Dickson, 1999). Rivastigmine trials in a group of PSP patients reported a mild improvement in cognitive disorders (Garcia-Reitboeck et al., 2013) and had also reported improvement in neuropsychiatric symptoms, but they did not show an effect in cognition (Goedert et al., 2013). Levodopa trials have shown only 20–40% of response in patients affected by PSP, and studies of CBD patients have reported its ineffectiveness because of unresponsiveness (Harper et al., 2008). Actually, some therapeutic strategies are linked to inhibition of tau posttranslational modifications, proteolytic activity, and self-aggregation. The decrease of tau mRNA with antisense oligonucleotides has been reported to decrease harmful tau aggregates, disrupt neuronal loss, and increase life duration of transgenic mice expressing human tau harboring the diseaseassociated P301S mutation (PS19 mouse model). Furthermore, a study of monkeys has reported a drastic decrease of tau mRNA and protein in the CNS (King et al., 2013). Also, the Food and Drug Administration (FDA) has approved an antisense oligonucleotide directed against mutant survival motor neuron gene 1 (*SMN1*) for treatment of spinal muscular atrophy (Bezard et al., 2013), characterized by alteration of lower motor neurons. Studies of brain injury in mice treated with an antibody directed against phospho-tau (p-tau) reported that it halted formation of tau oligomers and complexes, prevented expansion of harmful tau in neighbor cells, decreased brain atrophy, and regained longterm potentiation (Arendt et al., 2016).

According to the existing relation among tauopathies and insulin resistance, therapies previously utilized for treatment of diabetes approved by the FDA are being tested in preclinical and clinical trials. Insulin supplied intranasally in subjects presenting amnestic mild cognitive disorder or AD has been reported to have cognitive benefits (Jochum et al., 2004), (Musiek and Holtzman, 2015), and thus, this opened the way to a new clinical trial: the Study of Nasal Insulin in the Fight Against Forgetfulness (SNIFF, NCT01767909). This trial is studying the effects of insulin supplied intranasally on cognition and brain atrophy; it is not actually certain how insulin administered by this way affects tau protein (Orr et al., 2017). Liraglutide, a drug which stimulates insulin production, has been shown to prevent and decrease the phosphorylation of tau in a mouse model of type II diabetes (Wyss-Coray and Mucke, 2002) and to decrease phosphorylated tau and improve motor function in a mouse model of tauopathy (hTauP301L); also, there is a report about improvement in motor function on this (Ransohoff, 2016a). Studies on 3xTg-AD mice treated with linagliptin has shown that there is a decrease of tau phosphorylation and improvement in cognition (Asai et al., 2015). Metformin, a suppressor of hepatic glucose production, tested in a neuronal cell model of insulin resistance has been reported to prevent tau phosphorylation (Sarazin et al., 2003). A decrease in pathological tau phosphorylation has been reported in trials with tau transgenic mice, along with an increase in tau cleavage, aggregation, synaptic disruption, and hind limb atrophy (Maccioni et al., 2009).

Immunotherapy-based antibodies against tau and active immunizations targeting pathogenic tau oligomers are actually in clinical trials (Riederer et al., 2003). These would prevent intercellular tau spread. For instance, tau vaccine (AADvac1) has been reported to have favorable safety and immunogenicity outcomes (Bussiere et al., 2003), and antibodies against tau have been shown to be assimilated by neurons in *ex vivo* cultured brain slices *in vivo* (Grant et al., 1997; Kovari et al., 2003; Josephs et al., 2006b; Yu et al., 2016), and tau linked to an antibody may be removed by a pathway implying lysosome (Krishnamurthy et al., 2011; Collin et al., 2014). Rapamycin, a mammalian target of rapamycin (mTOR) inhibitor, upregulates autophagy, thus decreasing aberrant tau and improving cognition in multiple mouse models (Laihinen et al., 1994; Lee et al., 1994; Rinne et al., 1994a; Rinne et al., 1994c), and prevents neuronal death in tau transgenic *Drosophila* (Rinne et al., 1994b). Studies conducted by Salonen et al (1994) with transgenic mice of human tau P301S, observed that rapamycin decreases both phosphorylation of tau and its aggregate state when it is delivered at different stages of the disease. In turn, in trials with 3xTg-AD mice, the intake of prophylactic rapamycin notably prevents the formation of aberrant tau deposits (Salonen et al., 1994). Analogs of rapamycin have been reported to increase autophagy, decrease p-tau and NFTs, and improve cognition in studies with tau transgenic mice (Tang-Wai et al., 2003b; Bak et al., 2006). The natural compound wogonin that inhibits mTOR demonstrated a reduction of p-tau in cultured cells (Tang-Wai et al., 2003a). In *Drosophila*, a decrease of tau-induced neurotoxicity has been achieved by multiple strategies such as heterochromatin loosening, genetic reversal of filamentous actin, nuclear envelope disorder, and decline of oxidative stress (Schmidt et al., 2001; Trojanowski and Dickson, 2001; Fulga et al., 2007; Kandimalla et al., 2014; Armstrong, 2015).

### THE NEUROIMMUNE CONTEXT OF PD

### Neuroimmunology of PD

After AD, the second most common neurodegenerative disorder is PD. This pathology affects the normal movement of the subject, as a consequence of multifactorial factors, such as environmental and genetic factors. The molecular basis is still unclear. The triggering causes have not yet been determined, but it is known that factors such as oligomerization of α-syn, mitochondrial dysfunction, oxidative stress, inflammation, and aging have pathogenic roles in the disease Its main neuropathological marker is the degeneration of neurons which contain neuromelanin in *substantia nigra pars compacta*; this results in loss of dopamine and cytoplasmic protein aggregates, called Lewy bodies (LBs), mainly composed by α-syn filaments (Forno, 1996). α-Syn has 140 amino acids and three regions, a carboxyl end which is negatively charged, an amino terminal end which is charged positively, and a hydrophobic segment at the center, between residues 61 and 90 (considered as the non-amyloid component or NAC). The protein has four tyrosine residues, Tyr39 next to the amino region and Tyr125, Tyr133, and Tyr136 close to the C-terminus. It is also able to bind lipids (Uversky, 2007).

Among people over 50 years, ~2.0% is affected with PD (Maarouf et al., 2012). The clinical signature consists of motor, cognitive (dementia), neuropsychiatric (depression and anxiety), and autonomic dysfunctions (hypotension and constipation). The motor affections usually present (i) rest tremor, (ii) bradykinesia (impairment in the normal movement, mainly of complex voluntary movements), (iii) postural instability, and (iv) rigidity (Thomas and Beal, 2007). There are a diminished number of dopaminergic neurons in the *substantia nigra*, which consequently ameliorates dopamine in the striatum, promoting dysregulation in the basal ganglia. The previous effects trigger the motor symptoms observed. Briefly, this pathology is considered a ND characterized among the synucleinopathies, which also considers PD with dementia (PDD), LB dementia (DLB), and multiple-system atrophy (MSA). DLB is also able to present parkinsonism, but also hallucinations (mainly visual) and dementia (McKeith et al., 2005). These symptoms make the diagnosis for this disease difficult (McKeith et al., 2005).

### Relationships Between Tauopathies and Synucleinopathies

α-Syn hyperphosphorylation promotes misfolding and oligomerization. α-Syn deposits are ubiquitous in the CNS, commonly at presynaptic neuron terminals. These molecular effects are categorized among synucleinopathies (Golde and Miller, 2009; Uversky, 2009). Synucleinopathies usually share their occurrence with tauopathies and different diseases commonly associated with protein misfolding. Although the etiology of most of the processes involved in these pathologies suggest the effect of misfolded proteins in pathways, also affecting them, they are still unknown (Jellinger, 2010a; Jellinger, 2010b; Kovacs et al., 2010). It seems there is an overlap in both diseases, but they also present different genetic, clinical and pathological characteristics. It is common to find NFTs and LB presence in the brain or in a cell (Arai et al., 2001; Iseki et al., 2003). Their co-occurrence has been previously reported several times by (i) Schneider et al. (2006), who found NFT in the substantia nigra of PD patients with displacement damage, and (ii) Joachim et al. (1987), who found them in the same region in AD patients, Down syndrome, and PD. Furthermore, the presence of p-tau has been noticed in dopaminergic neurons of PD and PDD subjects (Wills et al., 2010). p-Tau has also been seen in striatal neurons of a transgenic model that overexpress human α-syn (Haggerty et al., 2011). Moreover, it was noticed that the phosphorylation of the GSK-3β protein does not happen if the expression of α-syn is silenced (Duka et al., 2006). Finally, these mechanisms considering tau and α-syn can be discussed by this manner: the promotion of α-syn expression leads to its accumulation in the brain, so on, GSK-3β gets phosphorylated, which in turn phosphorylate tau (Duka et al., 2006). NFT start forming because of the increase in p-tau. It is also interesting that almost 60% of AD subjects present LB, in familial and sporadic cases (Arai et al., 2001; Jellinger, 2011). We can consider an aggressive progression of these pathologies and an accelerated cognitive dysfunction because of the presence of synucleinopathies and tauopathies at the same time (Langlais et al., 1993; Olichney et al., 1998; Kraybill et al., 2005). It is suggested that the synergistic interaction between tau, Aβ, α-syn and the activate form of GSK-3β could trigger their misfolding, oligomerization and accumulation (Giasson et al., 2003; Lee et al., 2004).

There has been studies that shown the tau/α-syn binding *in vitro*, which in turn, promote their phosphorylations (Jensen et al., 1999). Tau fibril generation could be induced by α-syn, but also they induce pathological filaments formation between each other when they are co-incubated (Giasson et al., 2003). The tau and α-syn interaction have also been observed *in vivo*, using mice that overexpress Ala53Thr α-syn (A53T SNCA), evidencing aggregation in both proteins (Giasson et al., 2003). Besides, p-tau in Ser396 and Ser202, 396/404, were found in PD cortex synapses (Muntane et al., 2008) and brainstem samples, respectively, in mice models overexpressing A-309P α-syn (Frasier et al., 2005). Moreover, the direct relation of tau and α-syn in these diseases is supported by: (i) hyperphosphorylation of tau as consequence of α-syn effect in the mice model for PD, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (Duka et al., 2006), (ii) α-syn and p-tau presence in NFT and LBs (Shao et al., 2006); (iii) the proteasome promotes the oxidation of α-syn, which in turn triggers recruitment of tau inside oligodendroglial cells in synucleinopathies (Riedel et al., 2009). In addition, it was also observed tau phosphorylation by α-syn in an *in vitro* study, in the residues Ser262 and Ser356 by the PKA (Jensen et al., 1999). As an interesting fact, GSK-3β does not phosphorylate tau in Ser262, meanwhile PKA does not in residues Ser396 and Ser404, suggesting their synergic role in the development of tauopathies mediated by α-syn.

It was reported that α-syn could promote activation of GSK-3β, hyperphosphorylation tau in Thr181, Ser396, and Ser404 (Duka et al., 2006; Duka et al., 2009; Kawakami et al., 2011; Ciaccioli et al., 2013). It seems that this is a consequence of an augmented activity in GSK-3β (Duka et al., 2009; Wills et al., 2011) but also of the generation of a tau, α-syn, and GSK-3β complexes. Although there are more kinases binding to hyperphosphorylated tau and α-syn. Actually, c-Jun N-terminal kinase (JNK) and extracellular signalregulated kinase (ERK), which also phosphorylate tau in Ser404 and Ser396, were shown by fluorescence intensity distribution analysis (FIDA) to have correlation with the p-tau presence in mice models of PD (α-syn overexpression) (Frasier et al., 2005; Kaul et al., 2011; Oaks et al., 2013). It was also showed that in the presence of inductors for cationic aggregation (Fe3+, Al3+, or DMSO) co-oligomers of α-syn and tau, and their co-aggregated forms are present at nanomolar concentrations (Nubling et al., 2012). Nonetheless, tau phosphorylation by GSK-3β strongly promotes mixed oligomers appearance (Nubling et al., 2012). Finally, we are demonstrating that α-syn polymerization is promoted by tau and that α-syn induce the same to tau, through the NAC hydrophobic region. In line with this, the main difference of both, is that α-syn aggregates itself, meanwhile tau cannot and needs an inductor (Goedert et al., 1996). Considering all the previous data, we are able to integrate the molecular mechanisms discussed into the neuroimmunomodulation theory (Maccioni et al., 2009) described above, since there are common effectors with the capacity to trigger neuronal damage and death by inflammatory processes.

### Neuroinflammation and Parkinsonism

Following the inflammatory cascade triggered by damage signals, there is synapse impair by several molecular mechanisms. The previous process also generates a positive feedback loop promoting even more damage, mainly mediated by microglial cells (Rojo et al., 2008; Andrade et al., 2017).

Increased permeability of the BBB and neurovascular dysfunction have been associated with severe conditions in PD. This effect could be linked to infiltration of inflammation molecules to the middle brain, microglia activation and dopaminergic neurons death (Collins et al., 2012). The systemic inflammatory response in PD seems to be promoted by peripheral lymphocytes activation and augmented levels of serum cytokines, such as IL-2, IL-6 and TNF-α in PD patients (Collins et al., 2012). However, there is not a general confirmation on the release of proinflammatory cytokines associated with PD. The adaptive immune response could be explained by elevated levels of MHC II in astrocytes of the ventral midbrain and microglia in a mice PD model, after inflammatory processes induction (Martin et al., 2016). Among the different mechanisms which direct microorganisms or noncerebral immune cells into the brain, one of the recently described are direct vascular channels. They connect the skull bone marrow to the brain surface through the meninges, enabling other cells to travel into this region, which is commonly considered as "aseptic" (Kandimalla et al., 2011b). Leukocytes derived from the bone marrow can trigger inflammatory processes in the tissue they exert their protective function. In this context, it is well established that several brain pathologies, which involved neurodegeneration, commonly present neuroimmune dynamics triggering neuroinflammation, among their paramount action mechanisms (Rojo et al., 2008; Andrade et al., 2017). Since the immune response is also triggered by pathogens, inflammatory processes are elicited as well. This could explain part of the onset and progression of not just PD, but also, several neuronal illness conditions, after their invasion in the brain.

As widely discussed, microglia activation can be triggered by different damage signals including pathogens, toxins, endogenous proteins, products generated by dying neurons, and other toxic agents. The constitutive expression of proinflammatory cytokines, such as IL-2, IL-6, IL-1β, TNF-α, and IFN-γ, the presence of ROS and eicosanoids, were also noticed in *postmortem* patients with PD by cerebral analysis. Moreover, it was seen in serum and CSF *in vivo*, and in animal models for PD. The microglial activation product of neuronal death may lead to a vicious circle of neuroinflammation and neurodegeneration (Collins et al., 2012). Some of these substances released upon degeneration of neurons include aggregates of α-syn, neuromelanin, adenosine triphosphate, and metalloproteinase-3 (MMP-3) (Collins et al., 2012). The activation of microglial cells by the pathological forms of α-syn in PD, DLB and multisystemic atrophy, results in a balance between the elimination of α-syn-mediated phagocytosis through neuronal dysfunction, TLR4 microglia and degeneration of neurons because of proinflammatory cytokines and ROS presence (Bruck et al., 2016). Recently, astrocytes have been also implied in brain degradation of α-syn fibrils, instead of spreading. This mechanism is triggered by the transfer of α-syn to neighbor cells, which is more efficient into astrocytes, which then localize inside lysosomes, where they seem to be degraded (Kandimalla et al., 2011a). Later, since lysosomal dysfunction is a common hallmark of NDs, it was shown that α-syn fibrils are transferred through tunneling nanotubes (TNTs) between neurons inside lysosomes and induce the misfolding/aggregation of the normal soluble protein (Wani et al., 2011). Finally, the stressed lysosomes that function as the vehicle of transfer α-syn fibrils inside TNTs allows, after fibrils escape, the seeding of cytosolic protein, thus explaining the progression of the pathology and highlighting an unsuspected role of lysosomes in this process. (Kandimalla et al., 2013). Thus, and considering also the presence p-tau and NFT in regions affected in PD as discussed before, we are again, able to consider the neuroimmunomodulation theory in the context of inflammation for PD.

The new insights are the gut-brain axis regulation, and its role in the pathogenesis of PD. Gut disorders present another way to induce an immune response and inflammatory processes which could contribute to PD pathology. The most common early manifestations in PD that don't affect movement or motor skills, are olfactory impairment and constipation. This is shared with the Braak staging system: the enteric nervous system (ENS), α-syn in the dorsal motor nucleus of the vagus nerve, the olfactory bulb and the submandibular gland, since all of them are a gateway to the environment. The neuropathological process leading to the pathology seems to start in the ENS or the olfactory bulb, spreading by the rostrocranial transmission to the *substantia Nigra* and finally to the CNS, suggesting that the environment could be part of the disease onset and development.

For instance, it was shown that the pesticide rotenone can almost completely reproduce the common clinical and pathogenic features of PD after intragastric administration (Kandimalla et al., 2011b). It has been reported that there is an altered intestinal microbiome in patients with PD and that there is an influence of gut microbiota in enteric neuron activity (Chang et al., 2017; Maldonado-Lasuncion et al., 2018; Wei et al., 2018). Alselmi and co-workers evaluated that the gastric coadministration of subthreshold doses of lectins and paraquat reproduces the disease symptoms in rats and its behavioral affections. They used a solution containing paraquat+lectin for administration *via* gastric gavage and then evaluated behavior in the context of PD and gastric dysmotility. Pathological α-syn in the dorsal motor nucleus of the vagus (DMV) and in the *substantia nigra pars compacta* of SNpc neurons was also observed. Besides, nigrovagally evoked gastric motility was affected in the rats which underwent the treatment. This was seen before the onset of PD manifestations, which were improved by l-dopa treatment. They also made a vagotomy, preventing the progression of PD effects and constraining the appearance of pathological α-syn only to enteric neurons. These reports demonstrated that coadministration of these molecules induces a progressive, l-dopa-responsive PD which is preceded by gastric dysmotility (Anselmi et al., 2018). Finally, it is important to notice the role of the microbiota in the CNS, where the host is constantly controlling the maturation and function of the microglial cells (Cosin-Tomas et al., 2018), a reason to consider their effects and imbalances in the subject, which, in the future, could be key in the control, prevention, or even treatment of PD.

Another factor for evaluation is sleep disorders that are common in these diseases, the most common being called rapid eye movement (REM) sleep behavior disorder (RBD), representing between 30% and 46% of the cases for PD (Gagnon et al., 2002) and 50–80% of the cases for DLB (Boeve et al., 2007). Even Donaghy and McKeith (2014) suggest that this symptom could be used as a previous diagnosis for DLB. This is interesting to consider, since previous research has pointed out that REM sleep disturbances could be associated with the formation of aggregated forms of the proteins already discussed, promoting the mechanisms underlying the inflammatory context (Hebert et al., 2008; Leidinger et al., 2013; Nagaraj et al., 2017). As a conclusion, the new pharmacotherapies proposed and their further research for these pathologies, and possible sleep-disruptive effects in PD understanding, are crucial in order to improve the quality of life of the patients.

### Novel Therapies for Parkinson

The pharmacological approaches in PD are the usual modality to treat the pathology. Oral levodopa and a dopamine decarboxylase inhibitor, such as carbidopa, are considered as the best therapeutic agents (Geekiyanage and Chan, 2011). In the immunomodulatory context, several approaches target to diminish the inflammatory response. Recently, the focus has been mainly aimed to the immune signaling from the periphery of the CNS. Williams et al. proposed to target the chemokine receptor type 2 (CCR2) and the MHC OO (MHCII), since this has been previously demonstrated as neuroprotective in rodent models of PD (Martinez and Peplow, 2019). The group evaluated the genetic knockout and RNA silencing of the class II transactivator (CIITA), which coactivates transcriptionally MHCII. Their results provided evidence that CIITA is needed for the induction and infiltration of MHCII in peripheral immune cells by α-syn, in a mice model for PD, presenting it as a novel promising therapeutic target (Benussi et al., 2017). In this context, it is also important in a preventive manner, since it has also been reported that peripheral immune cell recruitment occurs prior to neurodegeneration and microglia; monocytes and macrophages all contribute to MHCII expression in PD (Du et al., 2018). Harms et al. also demonstrate that extravasation of proinflammatory peripheral monocytes into the CNS has a paramount role in neurodegeneration. Using a PD synucleinopathy model, they ended up proposing that peripheral monocytes are targets for PD,

### REFERENCES

Andrade, V., Guzmán-Martínez, L., Pulgar, K. V., and Maccioni, R. B. (2017). "Neuroimmune dynamics in Alzheimer's disease progression," in *Mechanisms of neuroinflammation*. Ed. G. E. A. Abreu (Rijeka: InTech). doi: 10.5772/ intechopen.68941

as a neuroprotective therapy. They observed that the expression of the full-length human α-syn *in vivo* promotes the infiltration of proinflammatory CCR2+ peripheral monocytes into the *substantia nigra*. Moreover, they found that α-syn-induced monocyte entry can be prevented by the genetic deletion of CCR2, attenuating MHCII expression and ending the degeneration of dopaminergic neurons (Sjodin et al., 2017).

Lots of immunotherapies for the treatment of PD use vaccines with AS or antibodies against it. A variety of procedures for vaccination have evidenced that the induction of regulatory T cells in the periphery protects the animal in PD models. In this context, the formulation glatiramer acetate (Copaxone®), which is commonly used for treatment of multiple sclerosis, is presented as a possible candidate because of its capacity to increase the number and the action as suppressors of regulatory T cells (Wani et al., 2014). Several reviewers also considered other approaches in order to diminish the inflammatory context; for instance, transcription factors were proposed in order to address this issue. There are alternatives like this promoting inflammation by effectors such as STAT 3, AP1, NF-κB, and TLRs which are constitutively upregulated in PD, while pathways considered as neuroprotective such as TGF-β, YY1, and mTOR are significantly downregulated in the microglia of patients with PD. Finally, it seems that their regulation could contribute to novel agent generation in order to treat PD, improve patient condition, or prevent the development of the pathology (Pal et al., 2016).

### AUTHOR CONTRIBUTIONS

LG-M and NR-E prepared the section on tauopathies and their therapy; LN and MP wrote the Alzheimer's disease section; VA prepared the Parkinson's disease section; and RM organized the strategy and goals of the paper and coordinated the different sections in the final writing of the text.

### FUNDING

This study was supported by the CORFO Innova projects (Grant 17ITE2-87685), the International Center for Biomedicine (ICC), and "the Ricardo Benjamin Maccioni Foundation."

### ACKNOWLEDGMENTS

We acknowledge the CORFO Innova projects as well as the International Center for Biomedicine and the Ricardo Benjamin Maccioni Foundation for the support of this investigation.

Anselmi, L., Bove, C., Coleman, F. H., Le, K., Subramanian, M. P., Venkiteswaran, K., et al. (2018). Ingestion of subthreshold doses of environmental toxins induces ascending Parkinsonism in the rat. *NPJ Parkinsons Dis.* 4, 30. doi: 10.1038/ s41531-018-0066-0

Arai, Y., Yamazaki, M., Mori, O., Muramatsu, H., Asano, G., and Katayama, Y. (2001). Alpha-synuclein-positive structures in cases with sporadic Alzheimer's disease: morphology and its relationship to tau aggregation. *Brain Res.* 888 (2), 287–296. doi: 10.1016/S0006-8993(00)03082-1


it within exosomes, effects suppressed by calcilytic NPS 2143—further implications for Alzheimer's therapy. *Front. Neurosci.* 11, 217. doi: 10.3389/ fnins.2017.00217


corticobasal degenerations and PSP. *Neurology* 66 (1), 41–48. doi: 10.1212/01. wnl.0000191307.69661.c3


human brain with carbon-11-SCH 39166 and carbon-11-NNC 756. *J. Nucl. Med.* 35 (12), 1916–1920.


pathogenesis of the MPTP mouse model of Parkinson's disease. *Glia* 64 (3), 386–395. doi: 10.1002/glia.22935


between 2002 and 2013. *Biomed. Environ. Sci.* 29 (3), 165–176. doi: 10.3967/ bes2016.021


**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 © 2019 Guzman-Martinez, Maccioni, Andrade, Navarrete, Pastor and Ramos-Escobar. 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.*

# N-3 Polyunsaturated Fatty Acids and the Resolution of Neuroinflammation

*Corinne Joffre1,2\*, Charlotte Rey1,2,3 and Sophie Layé1,2*

*1 INRA, Nutrition et Neurobiologie Intégrée, UMR 1286, Bordeaux, France, 2 Université de Bordeaux 2, Bordeaux, France, 3 ITERG, Nutrition Health and Lipid Biochemistry Department, Canéjan, France*

In the past few decades, as a result of their anti-inflammatory properties, n-3 long chain polyunsaturated fatty acids (n-3 LC-PUFAs), have gained greater importance in the regulation of inflammation, especially in the central nervous system (in this case known as neuroinflammation). If sustained, neuroinflammation is a common denominator of neurological disorders, including Alzheimer's disease and major depression, and of aging. Hence, limiting neuroinflammation is a real strategy for neuroinflammatory disease therapy and treatment. Recent data show that n-3 LC-PUFAs exert anti-inflammatory properties in part through the synthesis of specialized pro-resolving mediators (SPMs) such as resolvins, maresins and protectins. These SPMs are crucially involved in the resolution of inflammation. They could be good candidates to resolve brain inflammation and to contribute to neuroprotective functions and could lead to novel therapeutics for brain inflammatory diseases. This review presents an overview 1) of brain n-3 LC-PUFAs as precursors of SPMs with an emphasis on the effect of n-3 PUFAs on neuroinflammation, 2) of the formation and action of SPMs in the brain and their biological roles, and the possible regulation of their synthesis by environmental factors such as inflammation and nutrition and, in particular, PUFA consumption.

Keywords: n-3 long-chain PUFAs, docosahexaenoic acid, eicosapentaenoic acid, specialized pro-resolving mediators, nutrition, neuroinflammation, resolvins

### INTRODUCTION

Inflammation is a critical process in host defense, facilitating tissue repair, regeneration and maintenance of homeostasis. However, if uncontrolled, it becomes a chronic low-grade inflammation that is characterized by the production of pro-inflammatory cytokines and adipokines leading to tissue damage and loss of function (López-Vicario et al., 2016). This chronic low-grade inflammation is observed in many chronic pathologies including neurodegenerative diseases and also in aging. Hence it becomes a significant public health concern and constitutes a financial burden that impacts millions worldwide (Serhan, 2017b). It is thus of importance to find treatments that ensure resolution of inflammation in a specific time-limited manner. Nutrition has gained importance in recent years since nutriments such as n-3 polyunsaturated fatty acids (PUFAs) have anti-inflammatory properties (Calder, 2016; Calder et al., 2017). They exert their effect in part through their conversion into bioactive lipid mediators called specialized pro-resolving mediators (SPMs) that underlie most of the beneficial effects attributed to their precursors (Serhan, 2014; López-Vicario et al., 2016). Recently, attention has been paid to these derivatives in the regulation of neuroinflammation. In this review, we present an overview of brain n-3 PUFAs and their effect on neuroinflammation and of the formation and mechanisms of the action of SPMs in the brain.

*Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Luigia Trabace, University of Foggia, Italy Antonio Carlos Pinheiro De Oliveira, Federal University of Minas Gerais, Brazil*

> *\*Correspondence: Corinne Joffre corinne.joffre@inra.fr*

#### *Specialty section:*

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

*Received: 21 March 2019 Accepted: 12 August 2019 Published: 13 September 2019*

#### *Citation:*

*Joffre C, Rey C and Layé S (2019) N-3 Polyunsaturated Fatty Acids and the Resolution of Neuroinflammation. Front. Pharmacol. 10:1022. doi: 10.3389/fphar.2019.01022*

### BRAIN N-3 PUFAS AS PRECURSORS OF SPMs

### Brain n-3 PUFAs

The brain contains high levels of PUFAs (25–30%) that are mainly docosahexaenoic acid (DHA, n-3 PUFA) (12–14% of total fatty acids) and arachidonic acid (AA, n-6 PUFA) (8–10% of total fatty acids) (Carrie et al., 2000; Xiao et al., 2005; McNamara and Carlson, 2006; Little et al., 2007; Chung et al., 2008; Joffre et al., 2016). There are regional differences: the hippocampus and prefrontal cortex contain the highest DHA content whereas the hypothalamus has the lowest (Carrie et al., 2000; Xiao et al., 2005; Joffre et al., 2016). There are also cellular differences: astrocytes, oligodendrocytes, and microglial cells (representing respectively 70%, 10–15% and 10–15% of brain glial cells; Renaud et al., 2015) contain DHA in different proportions. DHA represents 10–12% in astrocytes, 5% in the oligodendrocytes and up to 2% in microglial cells (Bourre et al., 1984; Rey et al., 2018). Neurons cannot synthesize long-chain PUFAs (LC-PUFAs) but can incorporate them in their membranes: DHA represents 8.3% of the total fatty acids in neurons (Bourre et al., 1984).

Numerous studies have discussed the transport of DHA through the blood–brain barrier (BBB). DHA enters the brain as unesterified DHA that is the major pool supplying the brain with DHA. However, the precise mechanisms of entry are still not fully described. Some transporters facilitate the uptake of DHA into the brain: fatty acid transport proteins (FATPs), fatty acid translocase (CD36) and major facilitator superfamily domain containing 2A (MFSD2A) (Chouinard-Watkins and Bazinet, 2018). Fernandez et al. (2018) recently reported that a member of the long-chain acyl-CoA synthetase (ACSL) family, ACSL6, is also involved in brain DHA uptake but its role has yet to be determined (Fernandez et al., 2018).

This brain fatty acid composition can be affected by environmental factors such as nutrition, something to which individuals are continuously exposed. Indeed, the PUFA content in all brain structures is strongly impacted by the PUFAs present in the diet (Alashmali et al., 2016; Joffre et al., 2016). A diet rich in n-3 PUFAs (DHA found in fish or its precursor, alpha-linolenic acid, found in vegetable oil) increases brain DHA in rodents (Hiratsuka et al., 2009; de Theije et al., 2015; Skorve et al., 2015; Kitson et al., 2016). However, DHA supplementation is more effective than alpha-linolenic supplementation in increasing the DHA content in the brain (Lacombe et al., 2017; Rey et al., 2019). In rodents, DHA supplementation from 16 weeks to 16 months or from 20 to 22 months of age compensates a DHA decrease due to aging (Little et al., 2007; Labrousse et al., 2012; Bascoul- Colombo et al., 2016). On the other hand, a diet deficient in n-3 PUFAs decreases brain DHA in all brain structures, with the hippocampus, containing most DHA, being the most affected and the hypothalamus the least affected (Delpech et al., 2015b; Joffre et al., 2016; Manduca et al., 2017).

Studies from Broadhurst and Crawford suggest that the amount of DHA incorporated into the brain depends on the complexity of the brain structure and on behavior development (Crawford et al., 1999; Broadhurst et al., 2002). A decrease in brain DHA induced by a deficient diet during gestation and lactation can be reversed by 2-month DHA supplementation at weaning (Orr et al., 2013). The organism also adapts its metabolism to an n-3 PUFA deficiency condition: the half-life of DHA increases in the brain to reduce its loss (Rapoport et al., 2007) and the activity of the enzymes responsible for DHA conversion, the ∆6 desaturase and elongase, is increased in the liver (Cho et al., 1999; Wang et al., 2005; Igarashi et al., 2007). Brain cells are also impacted by dietary PUFA supply. An n-3 PUFA deficient diet decreases DHA in astrocytes, neurons and oligodendrocytes (Bourre et al., 1984) whereas n-3 PUFA supplementation increases DHA levels in glial cells (Bowen and Clandinin, 2005; Destaillats et al., 2010). Moreover, we recently showed that the fatty acid composition of the microglial cells is also modulated by n-3 PUFA dietary intake during the gestation/lactation period. Maternal n-3 LC-PUFA dietary supplementation during gestation and lactation increases the DHA level in the offspring's microglia at P21 as compared with a maternal diet that contains equilibrated levels of n-6 and n-3 PUFA precursors (Rey et al., 2018). Interestingly, maternal n-3 PUFA deficiency does not impact the DHA level suggesting that microglial cells are protected from n-3 PUFA deficiency (Rey et al., 2018).

### n-3 PUFAs as Regulators of Neuroinflammation

n-3 PUFAs have powerful anti-inflammatory properties (Calder, 2005). They play an important role in the regulation of the synthesis and release of pro-inflammatory mediators (Delpech et al., 2015b; Hanisch and Kettenmann, 2007; Cunningham and Sanderson, 2008; Yirmiya and Goshen, 2011; Pascual et al., 2012). Pro-inflammatory factors include interleukin-1 beta (IL-1β), interleukin-6 (IL-6) and tumor necrosis factor alpha (TNF-α) and play a role in neuronal plasticity (Delpech et al., 2015b; Yirmiya and Goshen, 2011). If sustained, uncontrolled inflammation can lead to neuronal damage that is involved in many neuronal pathologies (Blais and Rivest, 2003; Layé, 2010; Solito and Sastre, 2012). Hence, limiting the inflammation and enhancing the resolution of inflammation is of great interest.

### Evidence in Humans

In humans, the anti-inflammatory properties of n-3 PUFAs were first identified in epidemiological studies in Eskimos that consume a lot of n-3 LC-PUFAs from eating fish (Dyerberg and Bang, 1979; Kromann and Green, 1980; Simopoulos, 2008). Clinical studies have highlighted the beneficial effect of n-3 LC-PUFAs in chronic inflammatory and autoimmune diseases. Indeed, fish oil supplementation decreases pro-inflammatory cytokine expression, such as IL-1β in blood monocytes, and improves the symptoms of patients suffering from rheumatoid arthritis (Kremer et al., 1990; James et al., 1997; Kremer, 2000) or multiple sclerosis (Stewart and Bowling, 2005; Weinstock-Guttman et al., 2005). Moreover, DHA supply significantly decreases the circulating inflammatory markers and the oxidative stress (Freund-Levi et al., 2006; Kiecolt-Glaser et al., 2012). DHA supply for several months also improves the working and long-term memories in patients with moderate cognitive alterations (Freund-Levi et al., 2006; Lee et al., 2013).

#### Evidence in Animals

In animals, numerous studies have demonstrated the antiinflammatory properties of n-3 PUFAs in the brain. In lipopolysaccharide (LPS)-, or IL-1β-, induced inflammation models, dietary n-3 LC-PUFA supplementation in adulthood prevents LPS-induced hippocampal increase of pro-inflammatory cytokines IL-1β and TNF-α in rats and mice (Orr et al., 2013; Dehkordi et al., 2015; Rey et al., 2019). A dietary supply in eicosapentaenoic acid (EPA) decreases the production of LPSinduced pro-inflammatory cytokines IL-1β, IL-6 and TNF-α in the hippocampus and increases the production of antiinflammatory cytokines IL-10 and IL-4 in both rats (Kavanagh et al., 2004; Lonergan et al., 2004; Dong et al., 2017) and mice (Shi et al., 2016). This modification is associated with a decrease in the phosphorylation of c-Jun and c-Jun N-terminal kinase proteins and in nuclear factor-kappa B (NFκB) that regulate inflammation (Lonergan et al., 2004; Shi et al., 2016). EPA dietary supply attenuates the activation of microglial cells and astrocytes triggered by an intracerebral IL-1β administration and increases the production of IL-10 in the hippocampus in rats (Song and Horrobin, 2004; Song et al., 2008; Dong et al., 2017). Following dietary DHA supplementation, there is an increase of DHA in the phospholipid and free fatty acid fractions. However, only the unesterified DHA is necessary to attenuate neuroinflammation. Nevertheless, the DHA phospholipid pool is an important source of unesterified DHA (Orr et al., 2013). On the contrary, in mice, an n-3 PUFA deficiency from the first day of gestation to weaning increases the expression of pro-inflammatory markers in the hippocampus and an alteration in the motility and the phenotype of microglial cells and alters synaptic plasticity (Lafourcade et al., 2011; Madore et al., 2014; Thomazeau et al., 2017). Using the model of intraperitoneal (ip) injection of LPS, we showed that IL-6 expression is strongly induced in n-3 PUFA-deficient mice whereas sickness behavior is down-regulated *via* the impairment of IL-6 signaling in the brain (Mingam et al., 2008). Moreover, when the deficiency continues in adulthood, it alters GABAergic, dopaminergic and cholinergic neurotransmission (Zimmer et al., 2000; Aid et al., 2003; Chalon, 2006; Larrieu et al., 2014) and anxiety-related, depressive-like behaviors that can be compared with autistic behavior in adults (Mingam et al., 2008; Larrieu et al., 2012; Larrieu et al., 2014; Hughes et al., 2016; Manduca et al., 2017). Following ip injection of LPS, n-3 PUFAdeficient mice display altered hippocampal synaptic plasticity that likely contributes to spatial memory impairment and higher glucocorticoid levels (Delpech et al., 2015b). On the contrary, an n-3 PUFA dietary supply improves emotional behavior alteration and memory deficit in rats (Song and Horrobin, 2004; Song et al., 2008; Dong et al., 2017), restores social and memory performance altered in autism models and improves depressive symptoms in both mice (Pietropaolo et al., 2014; Madore et al., 2016; Weiser et al., 2016) and rats (Bove et al., 2018; Morgese et al., 2018). Of note, n-3 LC-PUFA dietary supplementation attenuates depressive symptoms in depressed patients presenting inflammation (McNamara, 2015; Rapaport et al., 2016; Larrieu and Layé, 2018; Layé et al., 2018). This protective effect seems to be linked to EPA which could target microglia to reduce neuroinflammation (Bazinet et al., 2019). n-3 LC-PUFAs could also modulate neuroinflammation through their effect on the HPA axis (Larrieu et al., 2014; Hughes et al., 2016; Larrieu and Layé, 2018). With Alzheimer's disease, β-amyloid triggers depressive symptoms through neuroinflammatory processes that could be improved by dietary n-3 PUFAs (Bove et al., 2018; Morgese et al., 2018). Altogether these data reinforce the importance of n-3 PUFAs as regulators of inflammation and associated depressive symptoms.

Another way to modulate neuroinflammation is to administer n-3 PUFAs directly into the brain or peripherally. Indeed, a 24-hour intracerebroventriculary (icv) DHA brain infusion attenuates hippocampal neuroinflammation initiated by icv LPS in mice (Orr et al., 2013). In a mouse model of cerebral ischemia, DHA icv administration inhibits NFκB activation and cyclooxygenase-2 (COX-2) expression in the hippocampus (Marcheselli et al., 2003). Moreover, intrathecal injection of DHA decreases microglial activation, mitogen-activated protein kinase (MAPK) phosphorylation and the production of proinflammatory cytokines in the spinal cord of mice (Lu et al., 2013). In addition, in a model of traumatic brain injury in rats, n-3 PUFA ip injection attenuates microglial-induced inflammation by inhibiting the NFκB pathway (Chen et al., 2017).

Dietary n-3 LC-PUFA supplementation requires the use of fish oil. However, fish oil may provide confusing factors such as vitamins, for example. Thus, the use of Fat-1 transgenic mice that convert n-6 to n-3 PUFAs through a desaturase from *C. elegans* is a model that enables the elimination of confusing factors provided by the diet (Kang et al., 2004). Such mice have higher DHA levels in the hippocampus and cortex (Delpech et al., 2015a; Boudrault et al., 2010; Orr et al., 2013). They express less COX-2 (involved in the production of lipid mediators) in the cortex than wild type mice (Boudrault et al., 2010). They are protected against cognitive deficits induced by ip LPS injection through a decrease in neuroinflammation (Orr et al., 2013; Delpech et al., 2015a). This is associated with a decrease in microglial activation (Orr et al., 2013).

### Evidence in *In Vitro* Studies

*In vitro* studies have shown that n-3 PUFAs have an inhibitory effect on the production of pro-inflammatory cytokines in microglial cells. Indeed, in the microglial cell line or primary culture microglial cells, DHA prevents LPS-induced NKκB activation and then cytokine production by inhibiting LPS receptor presentation and decreases the oxidative stress, nitric oxide (NO) production and inducible nitric oxide synthase (iNOS) expression (De Smedt-Peyrusse et al., 2008; Lu et al., 2010; Antonietta Ajmone-Cat et al., 2012; Pettit et al., 2013; Chen et al., 2014; Corsi et al., 2015; Fourrier et al., 2017; Inoue et al., 2017;). Moreover, DHA prevents LPS-induced MAPK phosphorylation (MAPK pathway playing essential roles in the expression of inflammatory molecules) and induces peroxisome proliferatoractivated receptor (PPAR)-γ nuclear translocation that exerts anti-inflammatory effects (Antonietta Ajmone-Cat et al., 2012). Of interest, EPA also affects the production of TNF-α, IL-6 and

NO by inhibiting NFκB phosphorylation *via* sirtuin-1 (SIRT-1) (Moon et al., 2007; Chen et al., 2014; Inoue et al., 2017). It is likely that DHA attenuates the inflammatory response in LPS-activated microglia by remodeling lipid bodies that are dynamic organelles in which DHA is incorporated and by altering their interplay with mitochondria and other associated organelles (Tremblay et al., 2016). In addition, DHA and EPA are able to enhance myelin or amyloid β (Aβ) peptide phagocytosis that is associated with a shift in microglial polarization toward the beneficial M2 phenotype and to a decrease in pro-inflammatory cytokine production (Hjorth et al., 2013; Chen et al., 2014).

EPA, eicosapentaenoic acid; IL, interleukin; TGF, transforming growth factor; TNF, tumor necrosis factor.

This modulation of neuroinflammation induced by n-3 LC-PUFA supply is attributed, in part, to SPM synthesis (Barden et al., 2016). In the various phases of inflammatory response, prostaglandins, leukotrienes and thromboxane are synthesized first (**Figure 1**). They permit the propagation of inflammation. They also stimulate the synthesis of SPMs with pro-resolutive properties (Levy et al., 2001; Serhan and Savill, 2005).

### n-3 PUFA-Derived SPMs and Their Synthesis Pathway

Many of the n-3 PUFA-derived immunomodulators that orchestrate the inflammatory response are lipids (Doyle et al., 2018). Some of them are involved in the regulation of inflammation (prostaglandins, thromboxanes, leukotrienes, etc.) and others are implicated in the resolution of inflammation (resolvins, protectins and maresins). There is a temporal shift in the lipid mediator synthesis from the initiation of inflammation to the resolution that allows the formation of different lipid mediators at different times (**Figure 1**). Here we have focused on SPMs, which are of great interest since they permit a return to homeostasis (Serhan et al., 2015; Sugimoto et al., 2016). Charles N. Serhan first identified these SPMs at the periphery and characterized their anti-inflammatory and pro-resolutive properties (Serhan et al., 2000). SPMs actively orchestrate and finely tune the inflammatory response. They decrease pro-inflammatory cytokines and increase anti-inflammatory cytokines, accelerate the phagocytosis of cellular debris and dead cells without immune suppression. A failure in the resolution of inflammation is detrimental for tissue and contributes to a chronic inflammatory response.

### BIOSYNTHESIS AND BIOLOGICAL ROLES OF n-3 PUFA-DERIVED SPMs

### Biosynthesis of n-3 PUFA-Derived SPMs

Free (unesterified) n-3 LC-PUFAs are released from membrane phospholipids through the action of phospholipases A2 (PLA2) in response to stimulation. DHA is hydrolyzed by calcium independent PLA2 (iPLA2) from phospholipids and plasmenylethanolamine-PLA2 from plasmalogens (Farooqui and Horrocks, 2006). After this step, n-3 LC-PUFAs undergo an enzymatic conversion to generate SPMs (**Figures 2** and **3**).

phospholipase A2.

n-3 PUFA-derived SPMs are synthesized mainly from DHA and EPA *via* COX-2, lipoxygenases (LOX) and CYP450 monoxygenases (CYP450). In the brain, 15-LOX, 12/15- LOX and 5-LOX are the most abundant LOX and are widely distributed, suggesting the potential production of SPMs (Shalini et al., 2017). They are expressed in the second step of inflammation in the hippocampus (Czapski et al., 2010; Birnie et al., 2013). 15-LOX is involved in neurodegeneration and neurotoxicity due to the increased oxidative stress it generates in models of Alzheimer's disease (Pratico et al., 2004; Wang et al., 2015) and brain ischemia (Yigitkanli et al., 2017). However, it is also implicated in neuroprotection (Sun et al., 2015). Indeed, it increases the production of 12-HETE and 15-HETE that promote the activation of PPARγ that is neuroprotective through its anti-inflammatory properties. The inhibition of 15-LOX induces hippocampus-dependent cognitive alterations (Shalini et al., 2017). Indeed, it prevented long term potentiation, a major molecular mechanism that describes the strengthening of synapses and underlies learning and memory. Moreover, 15-LOX deletion drastically decreases SPM production in the brain while SPMs also modulate synaptic plasticity (Park et al., 2011; Shalini et al., 2017). In the brain, the inducible COX-2 is activated *via* an NFκB pathway (Nadjar et al., 2005). COX-2 catalyzes the first step of the synthesis of prostaglandins and thromboxanes derived from n-6 PUFAs that contribute to the initiation of inflammation (Davidson et al., 2001; Salinas et al., 2007; Engstrom et al., 2012). COX-2 also produces hydroxyl fatty acids from DHA in both an aspirin-dependent and aspirinindependent manner (Serhan et al., 2002; Oh et al., 2011; Gabbs et al., 2015; Nebert, 2017). In response to LPS, COX-2

is rapidly expressed in the hippocampus (Czapski et al., 2010; Rey et al., 2019). Inhibition of COX-2 delays resolution of acute inflammation (Schwab et al., 2007). CYP450s play a complex role in inflammation by producing n-6 PUFAderived bioactive mediators such as epoxides that are antiinflammatory in monocytes and macrophages (Bystrom et al., 2011; Fleming, 2011; Nebert et al., 2013; Gilroy et al., 2016).

DHA is the precursor of resolvins D1-6 (RvD1-6), neuroprotectin D1 (NPD1) and maresins 1–2 (Mar1-2) which all have pro-resolutive and anti-inflammatory properties (Spite and Serhan, 2010; Halade et al., 2018). RvD1-6 are synthesized from DHA but RvD1 is the most studied because it has powerful anti-inflammatory and proresolutive properties. DHA is converted into monohydroxy DHA, 17-hydroxy docosahexaenoic acid (17-HDHA) by acetylated COX-2, CYP450 and 15-LOX (Barden et al., 2016; Halade et al., 2018) and then into RvD1 by 5-LOX (Sun et al., 2007; Recchiuti, 2013). DHA is also converted into di-hydroxy-DHA, termed protectin D1 (PD1) or neuroprotectin D1 (NPD1) when produced in the central nervous system by 5- and 15-LOX (Hong et al., 2003; Aursnes et al., 2014; Kuda, 2017; Doyle et al., 2018). Acetylated COX-2 permits the synthesis of aspirin-triggered PD1 (AT-PD1) which has powerful protective effects (Bazan et al., 2012). DHA is transformed into Mar 1-2 by 12/15-LOX *via* the synthesis of 14-HDHA (Serhan et al., 2009; Barden et al., 2016; Halade et al., 2018).

EPA is the precursor of resolvins E1 (RvE1), E2 and E3 that have many biological roles (Serhan et al., 2000; Rey et al., 2016; Halade et al., 2018). It is converted by aspirin-triggered acetylated COX-2 or CYP450 into 18R-hydroxyeicosapentaenoic acid (18R-HEPE), that is transformed into RvE1 or E2 by 5-LOX (Ohira et al., 2010; Barden et al., 2016) or into RvE3 by 15-LOX (Isobe et al., 2012).

Little is known concerning the pharmacokinetics and dynamics of oxylipins. They are synthesized *in situ*, reinforcing the importance of the fatty acid composition of the brain. They are not stored, but produced on demand and they are unstable, being rapidly metabolized.

The structure of all derivatives is highly preserved in the evolution from fish to humans suggesting their great bioactive role in all organ systems. Dysfunction of SPM production can be due to insufficient EPA and DHA supply leading to inadequate production of SPMs or to the polymorphism of the enzymes involved in their synthesis or to a defect in the binding of SPMs to their receptors (Sugimoto et al., 2016). In humans, reduced SPM production is associated with chronicity and the magnitude of persistent inflammation. SPMs have multiple biological roles in the maintenance of homeostasis.

### Biological Roles of DHA-Derived SPMs

In human serum, the DHA-derivatives represent 30.7% of the identified SPMs (Colas et al., 2014; Serhan et al., 2018). RvD1, PD1 and Mar1 were detected at 30.9 pg/mL, 5.6 pg/mL and 21.2 pg/mL, respectively. They act in the nanomolar or picomolar range as compared with DHA and EPA acting in the micromolar range (López-Vicario et al., 2016; Clària et al., 2017; Serhan, 2017a; Wang and Colgan, 2017; Rosenthal et al., 2018).

### Resolvins

RvDs have been identified in mice peritoneal exudates (Serhan et al., 2002). They have many properties at the periphery: protection against bacterial infection, prevention of leucocyte infiltration, inhibition of the production of cytokines, etc. (Spite et al., 2009; Serhan and Petasis, 2011; Chiang et al., 2012; Dalli et al., 2013a; Winkler et al., 2016). RvD1 and its precursor metabolites have been detected in the brain.

*In vitro*, 17-HDHA has been found in glial cells after DHA and inflammatory stimulus (Hong et al., 2003). *In vivo*, RvD1 has been identified in mouse brain following cerebral ischemia (Marcheselli et al., 2003). Intravenous (iv) DHA injection increases RvD1 levels in rats suggesting the conversion of DHA into RvD1 (Mulik et al., 2016). RvD1 acts through the lipoxin A4 receptor/formyl peptide receptor 2 (ALX/Fpr2) in rodents and the G protein coupling receptor 32 (GPR32) in humans (Recchiuti, 2013) in the picomolar range but induces biological effects in the nanomolar range (Sun et al., 2007; Krishnamoorthy et al., 2010). An overexpression of ALX/Fpr2 or GPR32 increases in phagocytosis by macrophages whereas their deletion decreases the phagocytosis response (Perretti et al., 2002; Krishnamoorthy et al., 2010; Cooray et al., 2013). RvD1 regulates the expression of specific micro-RNA (miR) to control the intensity and the length of inflammation *via* the regulation of target genes such as inflammatory cytokine genes (Fredman and Serhan, 2011; Recchiuti et al., 2011; Rey et al., 2016; Bisicchia et al., 2018). RvD1 increases the expression of miR-21, miR-146b and miR-219 and decreases the expression of miR-208a in macrophages and peritoneal exudates in mice (Recchiuti et al., 2011; Krishnamoorthy et al., 2012). RvD1 modulates miR-155, miR-146, miR-21 and miR-219 in microglial cells (Rey et al., 2016). These miRs have different biological roles: miR-21 is essential to the production of anti-inflammatory cytokine IL-10, miR-146 regulates the transcription of cytokines, chemokines and their receptor, miR-219 decreases the transcription of TNF-α and miR-208a regulates the activation of NFkB (Recchiuti et al., 2011).

RvD1 controls the inflammatory response in many animal models *via* its anti-inflammatory and pro-resolutive properties. In rats, endogenous RvD1 levels decrease at the beginning of inflammation and then increase during the resolution phase (Sun et al., 2014). Recent studies describe the anti-inflammatory properties of RvD1 in microglia and astrocyte cell cultures (Abdelmoaty et al., 2013; Li et al., 2014; Rey et al., 2016). Indeed, in BV2 microglia cell culture, RvD1 promotes the IL-4-induced M2 phenotype (Li et al., 2014) and inhibits LPS-induced proinflammatory cytokines (Rey et al., 2016). In astrocyte cell culture, RvD1 attenuates LPS-induced TNF-α (Abdelmoaty et al., 2013). *In vivo*, in a model of remote damage, Bisicchia et al. recently show in rats that RvD1 promotes functional recovery and reduces neuroinflammation *via* miRs (Bisicchia et al., 2018).

### *Biological Roles of RvD1 in Humans*

The effect of RvD1 has been studied in patients suffering from Alzheimer's disease. This pathology is characterized by an increase in microglial activation and in pro-inflammatory cytokine production in the brain (Griffin et al., 1989; Cagnin et al., 2001). Interestingly, RvD1 levels in the cerebrospinal fluid are positively correlated with the enhancement of cognitive functions of patients with dementia (Wang et al., 2015). Indeed, RvD1 may be involved in Aβ phagocytosis. This has been shown *in vitro* in macrophages isolated from Alzheimer's patients (Mizwicki et al., 2013; Famenini et al., 2017). Thus the decrease in RvD1 levels in Alzheimer patients' brains could contribute to the evolution of the disease.

### *Biological Roles of RvD1 in Rodents*

RvD1 attenuates the pro-inflammatory status in the central nervous system. Indeed, an intrathecal injection of 17R-HDHA decreases TNF-α release in the spinal cord in rats (Abdelmoaty et al., 2013). Orr et al. show that icv injection of 17S-HDHA into mice decreases the expression of hippocampal pro-inflammatory cytokines IL-1β and TNF-α induced by LPS acute icv injection (Orr et al., 2013). However, these authors detected NPD1 but did not detect RvD1, suggesting that the effect of 17S-HDHA was rather due to the conversion of 17S-HDHA into NPD1. RvD1 is also able to stimulate phagocytosis in macrophages (Rossi et al., 2015). Indeed, DHA and RvD1 induce the polarization of macrophages toward an M2 phagocytic phenotype in mice (Titos et al., 2011).

Studies have highlighted the protective role of RvD1 in the occurrence of cognitive deficits. Terrando et al. showed that an ip injection of 17-HDHA restores transmission and synaptic plasticity and prevents astrogliosis and cognitive decline in a systemic inflammation model in mice (Terrando et al., 2013). Conversely, an inhibition of 15-LOX, associated with a decrease in RvD1, alters synaptic plasticity and working memory as demonstrated in rats (Shalini et al., 2017). RvD1 plays also a role during the recovery phase following cerebral ischemia. Indeed, the precursor of 17-HDHA, 17-hydroperoxy docosahexaenoic acid (17-HpDHA), accumulates in the hippocampus of mice (Marcheselli et al., 2003). An ip chronic administration of 17R-HDHA synthesized by acetylated COX-2, prevents cognitive deficits and attenuates motor disorders but doesn't ameliorate microglial activation and sleep quality in mice (Harrison et al., 2015). Moreover, Fat-1 mice that have more brain n-3 LC-PUFAs, have higher hippocampus RvD1 levels, associated with less cognitive deficits, a better neuronal survival, a decrease in astrocyte and microglial activation and a reduction in pro-inflammatory status following brain ischemia (Luo et al., 2014; Delpech et al., 2015a).

Studies have also highlighted the protective role of resolvins in the depressive-like behavior in rodents. Some of them have been recently reported by Furuyashiki et al. (Furuyashiki et al., 2019). An icv injection of RvD1, D2, E1, E2, or E3 significantly decreases LPS-induced depressive-like behavior in mice (Deyama et al., 2017; Deyama et al., 2018a; Deyama et al., 2018b). Moreover, the occurrence of depressive-like behavior associated with pain can also be prevented by an intrathecal injection of 17R-HDHA that is associated with the decrease of pain perception and a restoration of dopamine and glutamate levels in the brain both in rats (Abdelmoaty et al., 2013) and mice (Klein et al., 2014). RvD1 and D2 have also positive effects in chronic mild stressinduced depression and in post-myocardial infarct depression in rats (Gilbert et al., 2014) and mice (Ishikawa et al., 2017).

### *Biological Roles of RvD1 in In Vitro Models*

The effects of RvD1 were tested on different brain cells. In microglial cells, RvD1 potentiates the effect of the antiinflammatory cytokine IL-4 in the activation of M2 phenotype of microglia (Li et al., 2014). Moreover, we showed that RvD1 decreases LPS-induced pro-inflammatory cytokines (TNF-α, IL-6 and IL-1β) gene expression in microglial BV2 cells *via* the modulation of miRs (Rey et al., 2016). This suggests its proresolutive activity in microglia. In astrocytes, RvD1 decreases TNF-α release induced by LPS injection (Abdelmoaty et al., 2013). In neurons from spinal nods, RvD1 increases neurite outgrowth (Shevalye et al., 2015). All these studies suggest that RvD1 can play a central role in the regulation of neuroinflammatory pathologies.

Other D resolvins have been identified in rodent brain: RvD2, RvD4 and RvD5 (Orr et al., 2013; Hashimoto et al., 2015; Winkler et al., 2016). RvD2 limits the activation of microglial cells and inhibits the TLR4/NFκB pathway (Tian et al., 2015). Moreover, chronic intrathecal RvD2 injection prevents the behavioral alterations induced by LPS central injection (Tian et al., 2015). It also prevents depressive-like behavior induced by ip LPS injection by regulating mammalian target of rapamycin complex 1 (mTORC1) complex (Deyama et al., 2017). RvD5 level is decreased in the brain of Alzheimer's disease patients but its role has not yet been identified (Zhu et al., 2016).

### Neuroprotectin

Di-hydroxy-DHA termed protection D1 (PD1) has been identified in blood, peritoneal neutrophils and neuroprotectin D1 (NPD1) in the brain in response to zymosan in mice (Hong et al., 2003). Marcheselli et al. have measured NPD1 production in mice hippocampus following brain ischemia (Marcheselli et al., 2003). One receptor for NPD1 has recently been identified as GPR37 (Qu and Caterina, 2018). NPD1 inhibits the oxidative stress in retinal epithelial cells and stimulates their proliferation (Bazan, 2009; Calandria and Bazan, 2010). NPD1 is also able to inhibit neovascularization *via* microglial cell ramification in mice (Sheets et al., 2013). In the central nervous system of mice, the NPD1 level greatly increases in the hippocampus following brain ischemia or acute central LPS injection (Marcheselli et al., 2003; Orr et al., 2013). Hence, NPD1 limits neutrophil infiltration, inhibits NFκB and then decreases pro-inflammatory gene expression (Marcheselli et al., 2003; Bazan et al., 2012; Yao et al., 2013). NPD1 levels decrease in the hippocampus of Alzheimer's disease patients (Lukiw et al., 2005). It plays a role in cellular survival *via* anti-apoptotic protein induction and attenuates pro-inflammatory responses following Aβ exposure *via* NFκB regulation (Lukiw et al., 2005; Bazan, 2008; Asatryan and Bazan, 2017).

### Maresins

Mar1 has been identified in mice peritoneal macrophages (Dalli et al., 2013b; Serhan et al., 2009). Its receptor has not been identified yet (Zhu et al., 2016). Mar1 is involved in the resolution of inflammation, prevents neutrophil infiltration, and increases the phagocytosis of apoptotic neutrophils by macrophages in a peritonitis murine model (Serhan et al., 2009). In a murine model of colitis, Mar1 decreases the expression of pro-inflammatory cytokines by inhibiting the NFκB pathway and activating the M2 phenotype in macrophages (Marcon et al., 2013). Mar1 and its precursor 14-HDHA have recently been identified in the hippocampus of mice (Orr et al., 2013). In *post-mortem* brain of Alzheimer's disease patients, the Mar1 level is decreased (Zhu et al., 2016). In this pathology, its role is to stimulate Aβ plaque phagocytosis by microglial cells and to decrease inflammatory marker levels. Hence, Mar1 may play an important role in the pathogenesis of Alzheimer's disease (Zhu et al., 2016). In cerebral ischemia in mice, Mar1 icv injection decreases inflammation and mitochondrial damage and also reduces neurological deficits *via* activation of SIRT-1 signaling (Xian et al., 2016; Xian et al., 2019). After spinal cord injury in mice, iv injection of Mar1 promotes resolution of inflammation (reducing pro-inflammatory cytokines, silencing pro-inflammatory signaling cascades and enhancing the M2 repair macrophage phenotype) and functional recovery (Francos-Quijorna et al., 2017). Mar1 also decreases *in vitro* neuronal death (Zhu et al., 2016).

### Biological Roles of EPA-Derived SPMs

RvE1, and its precursor 18-HEPE, have been detected in the hippocampus of rats (Chen et al., 2011) and mice (Orr et al., 2013; Siegert et al., 2017). In human serum, the EPA-derivatives represent 25.9% of the identified SPMs (Colas et al., 2014; Serhan et al., 2018).

RvE1 has been initially identified in mouse exudates (Serhan et al., 2000). RvE1 directly binds to its receptor ChemR23 or CMKLR1 (chemokine-like receptor 1) (Ohira et al., 2010). It is also a partial agonist of LTB4 receptor (BLT1) (Arita et al., 2007). In the central nervous system, ChemR23 has been identified in the prefrontal cortex, hippocampus and brainstem (Guo et al., 2012), in microglial cells (Graham et al., 2009; Rey et al., 2016) and in neurons (Xu et al., 2010). It is highly expressed in neurons, in microglial cells and in astrocytes in the *post-mortem* hippocampus of Alzheimer's patient brain (Wang et al., 2015). This increase could be due to a compensatory mechanism to counter-balance the decrease in RvE1 in such patients (Wang et al., 2015). *In vitro*, RvE1 plays also a direct role in microglial cells by inhibiting microglial activation and pro-inflammatory cytokine release (Xu et al., 2013; Rey et al., 2016). RvE1 and its precursor, 18R-HEPE, exert anti-inflammatory and anti-apoptotic properties (Hecker et al., 2018). Indeed, they restore mitochondrial dysfunction induced by inflammation in mononuclear blood cells. *In vivo*, RvE1 modulates the inflammatory profile and microglial activation in mice (Xu et al., 2013; Harrison et al., 2015). Intraperitoneal injection of RvE1 also modulates inflammation (by reducing Il-1β, Il-6 and IL-10 levels in the prefrontal cortex) and decreases the measures of Aβ pathology in a murine model of Alzheimer's disease (Kantarci et al., 2018). RvE1 and RvE2 centrally administered also reduce the LPS-induced depressive-like behavior through ChemR23 in mice (Deyama et al., 2018b). RvE1, a total agonist of ChemR23, is more effective than RvE2, which is only a partial agonist of this receptor (Serhan and Chiang, 2013).

### REGULATION OF THE n-3 PUFA-DERIVED SPMs

Resolution of inflammation is an active process involving the regulation of the synthesis of numerous mediators in a tightly coordinated manner. The balance between n-3 PUFA-derived SPMs and the pro-inflammatory mediators determines the duration of the inflammatory response and the timing of resolution (Fredman et al., 2017).

### Regulation of the n-3 PUFA-Derived SPMs by PUFA Consumption

We have previously shown that PUFA consumption leads to modifications in PUFA levels in the brain. The PUFA derivative levels and their biosynthetic enzyme expression also depend on dietary PUFAs.

SPM levels in peripheral organs and the brain are modulated by different dietary supplies of PUFAs (**Figure 4**). In humans supplemented with 2.4g/day of n-3 LC-PUFAs for 3 weeks (EPA+DHA), SPM levels are generally found to be between 20 and 200 pg/mL (Mas et al., 2012). A shorter supplementation period (5 days) also increased DHA- and EPA- derived SPMs in plasma, especially 18-HEPE, RvE1, 17- and 14-HDHA (Barden et al., 2014). In another study, supplementing patients with a higher fish oil supplement (17.6g/day EPA+DHA) for 24h, only PD1 is detected in the plasma and this increases from 1.0–1.2 pg/mL to 2.7–4.1 pg/mL (Skarke et al., 2015; Barden

PUFA, polyunsaturated fatty acids; SPM, specialized pro-resolving mediators; TNF, tumor necrosis factor.

et al., 2016). These differences in SPM levels in human plasma could be due to differences in study design, dose, duration of n-3 fatty acid supplementation and patient characteristics, etc. (Barden et al., 2016; Chiang and Serhan, 2017). Moreover, dietary supplementation in DHA (1.7g/j) and EPA (0.6g/j) for 6 months increased the RvD1 released by blood mononuclear cells in Alzheimer's disease patients as compared with controls (Wang et al., 2015). Results from intervention studies show that EPA and DHA dietary supplementation increases EPA- and DHAderived oxylipins although with high inter-individual variances (Ostermann et al., 2017).

In animals, a 3-week fish oil supplementation in arthritic mice increases synthesis of n-3 PUFA-derived SPMs associated with a diminished production of pro-inflammatory mediators (Norling et al., 2016). n-3 PUFA dietary supplementation (140 mg EPA+110 mg DHA/kg animal) in aged rats for 10 weeks enhances the production of lipid mediators derived from DHA and EPA in the cerebral cortex (Hashimoto et al., 2015). DHA and EPA supply also decreases the n-6 PUFA-derived metabolites (which are mainly pro-inflammatory) and increases the n-3 PUFAderived metabolites in the plasma and at the periphery both in mice (Balvers et al., 2012; Lazic et al., 2014) and rats (Hashimoto et al., 2015). A diet enriched with fish oil (containing 9% EPA + 9% DHA) for 8 weeks in mice increases plasmatic RvD1 (Shevalye et al., 2015). In Fat-1 mice, brain EPA increase is accompanied by an increase in EPA-derived HEPEs (Siegert et al., 2017). Conversely, an n-6 PUFA dietary supply from weaning to 15 weeks of age increases the n-6 PUFA metabolites and decreases the EPA-derived metabolites in the cortex and the plasma of rats (Taha et al., 2016). Recently, we showed that, in mice, n-3 LC-PUFA supplementation from weaning for 2 months induces n-3 LC-PUFA enrichment in the hippocampus and subsequently an increase in n-3 PUFA-derived SPMs and a decrease in n-6 PUFA-derived mediators (Rey et al., 2019). Moreover, in response to LPS, n-3 LC-PUFA-deficient mice present a pro-inflammatory lipid mediator profile whereas n-3 LC-PUFA-supplemented mice display an anti-inflammatory profile in the hippocampus (Rey et al., 2019). Indeed, the consumption of n-3 LC-PUFAs increases the EPA and DHA derivatives. Hashimoto et al. found the same in the cortex of aged rats (Hashimoto et al., 2015). These results differ from those of Trépanier et al. who found that, in a model of Fat-1 mice, a greater increase in brain DHA induced by dietary supply has no effect on the resolution of inflammation after an icv LPS injection (Trépanier et al., 2018).

SPM production is also finely tuned by the regulation of the enzymes in their biosynthesis pathway. A 15-week n-3 PUFA-deficient diet increases the COX-2 expression in the prefrontal cortex of rats, suggesting an increase in AA-derived pro-inflammatory mediator levels (Rao et al., 2007). Conversely, a 15-week n-6 PUFA-deficient diet decreases brain COX-2 expression in rats (Kim et al., 2011). Interestingly, Taha et al. showed that rats fed on an n-6 PUFA-deficient diet for 15 weeks are protected from an increase in COX-2 observed in LPS-treated n-3 PUFA-deficient animals (Taha et al., 2017). However, we showed that 2 months' n-3 LC-PUFA dietary supplementation was not sufficient to modulate COX-2 expression in the brain of mice (Rey et al., 2019). An n-3 LC-PUFA diet increases 15-LOX expression in the hippocampus without changes in the 5-LOX expression (Gabbs et al., 2015). The increase in 15-LOX in the brain has also been described following a 15-week n-6 PUFA-deficient diet in rats (Kim et al., 2011). 15-LOX is responsible for many n-6 and n-3 PUFA derivatives. Those synthesized from AA drastically decrease, and those produced from DHA increase, following a diet enriched in n-3 PUFAs in both mice and rats (Ostermann et al., 2017; Shalini et al., 2017; Rey et al., 2019).

Results on 15-LOX functions on inflammation regulation are conflicting. 15-LOX was initially described as deleterious in neurodegenerative pathologies because it increased the oxidative stress and neuronal degeneration (Pratico et al., 2004; Chinnici et al., 2007; Di Meco et al., 2014; Wang et al., 2015; Yigitkanli et al., 2017). It alters the mitochondrial function of neurons and then induces neuronal death (Pallast et al., 2009). However, recent studies demonstrate the protective role of 15-LOX, particularly in cognition. An increase in its expression is associated with a better working memory (Sun et al., 2015; Shalini et al., 2017). Moreover, an increase in 15-LOX expression during cerebral ischemia is essential for the recovery of neurological functions after the ischemic event (Wang et al., 2017). These results suggest that 15-LOX has a beneficial role in acute inflammation (ischemia, LPS) and is deleterious in chronic inflammation (neurodegenerative pathologies). SPM production is also partly regulated by the availability of free PUFAs that are their precursors. Indeed, PLA2 activity also depends on PUFA dietary supply. An n-6 PUFA-deficient diet increases the iPLA2 that is responsible for the release of DHA from the membranes and a decrease in calcium sensitive cytosolic PLA2 (cPLA2) associated with the hydrolysis of AA from the membranes (Kim et al., 2011).

Lipid nutrition, to which people are exposed throughout their lives, seems to play a major role in the synthesis of bioactive SPMs.

### Regulation of the n-3 PUFA-Derived SPMs by Inflammation

Numerous studies have highlighted that inflammation modulates lipid mediator synthesis at the periphery and in the brain (**Figure 4**).

In humans, Wang et al. showed that the RvD1 level in the cerebrospinal fluid (CSF) of Alzheimer's disease patients is positively correlated with cognitive function (Wang et al., 2015). Moreover, 15-LOX was decreased in the CSF suggesting an alteration of the resolution of inflammation in Alzheimer's patients.

In rats, brain ischemia increases the production of 5 of mono-, di- and tri-hydroxy-DHA derivatives (Farias et al., 2008; Hennebelle et al., 2017). However, LPS icv injection does not impact the RvD1 level (Rosenberger et al., 2004; Taha et al., 2017). We have recently shown that, in mice, ip LPS injection modifies the n-6 PUFA derivative profile but not n-3 LC-PUFA derivatives (Rey et al., 2019).

*In vitro,* most studies have been performed on peripheral immune cells. The lipid mediator profile in rodent neutrophils and macrophages changes with macrophage phenotype M1 or M2, or with the inflammatory stimulus in neutrophils (Dieter et al., 2002). Indeed, M2 macrophages produce more D resolvins, protectins and maresins and less AA-derived pro-inflammatory mediators than M1 macrophages (Dalli and Serhan, 2012).

Inflammation more drastically alters the expression of biosynthetic enzymes than an n-3 LC-PUFA dietary supply. In rats, traumatic brain injury increases the COX-2, 5-LOX, 15-LOX and CYP450 expression in the hippocampus and cortex, suggesting an alteration of all lipid mediator biosynthesis pathways (Birnie et al., 2013). LPS also regulates the expression of these enzymes. An icv LPS injection increases COX-2 expression and activity in the brain in mice (Rey et al., 2019) and rats. Moreover, in *in vitro* human monocytes and dendritic cells, antiinflammatory cytokines IL-4 or IL-13 increase the production of 15-LOX and decrease the production of 5-LOX (Nassar et al., 1994; Spanbroek et al., 2001).

Neuroinflammation also modifies SPM receptor expression. Indeed, we show that LPS increases significantly the expression of RvD1 and RvE1 receptors (ALX/Fpr2 and ChemR23, respectively) in BV-2 microglial cells (Rey et al., 2016). The increase in ALX/Fpr2 expression was also detected in monocytes, in the hippocampus and in the cortex and in microglial cells in response to an inflammatory stimulus (Krishnamoorthy et al., 2010; Wang et al., 2011; Wang et al., 2015; Guo et al., 2016). Then, inflammation activates the SPM signaling pathway in the brain to regulate the inflammatory response.

### HUMAN TRANSLATION

Several possibilities can be considered to translate the findings described above and then attenuate the inflammatory tone, amplitude and duration of inflammation. The first one is to increase the local production of n-3 LC-PUFA-derived SPMs. We see that dietary means is a good way to modulate the level of the fatty acids from which they are synthesized and then to modify their synthesis. The SPM profile synthesized by each individual could be responsible for the differences in the effects of n-3 PUFA obtained in humans. SPM profiles should be established in patients with different acute and chronic inflammatory pathologies, and in mice under the same conditions to find markers of neuroinflammation in the plasma that can be transposed to humans. Thanks to new technologies in liquid chromatography mass spectrometry (LC-MS/ MS), specific mediators produced during physiological and non-physiological conditions should be identified, allowing patient stratification according to disease severity. It could be interesting to determine an individual metabolomic profile to define personalized nutrition (n-3 PUFAs and doses) allowing an increase of n-3 PUFA-derived SPMs in the target tissue. Indeed, there are individual differences in diets and in n-3 PUFA supplementation and also in nutrient metabolism and biological responses to food/nutrients. The aim of personalized nutrition is to increase health using nutrition by delivering specific personalized intervention suited to each individual based on the individual's nutritional phenotype, metabolic profile, and environmental factors in order to prevent and treat chronic disease. Personalized nutrition can also be applied to healthy people. It is nowadays accessible because of a better understanding of the mechanisms of the effect of nutrition on health and also because of the progress in technologies enabling the identification of specific markers. Personalized nutrition has already shown its efficacy, especially in the Food4Me study involving >1600 participants from 7 European countries and in a systematic review and meta-analysis showing a greater efficacy of personalized nutrition in changing diet than a conventional approach (Celis-Morales et al., 2018).

The second possibility for taking advantage of research on the resolution of inflammation is to administer exogenous SPMs. Serhan defines a new concept of resolutive pharmacology targeting the development of SPM analogs, resistant to local

### REFERENCES


inactivation, to stimulate natural circuits of resolution (Chiang and Serhan, 2017). The objective of this new therapeutic pathway is to administer these analogs in association with classical therapy in order to decrease the doses, thus limiting the secondary effects. A clinical trial has reported for the first time the efficacy of an RvE1 analog in patients with dry eye symptoms (Resolvyx Pharmaceuticals). These encouraging results should be extended to the use of such molecules to treat other inflammatory diseases.

### CONCLUSION

In the investigation of new anti-inflammatory treatments without the secondary effects of traditional therapy, SPMs are promising therapeutic compounds: they are of natural origin and are active at low concentrations (nM) as compared with their precursor (µM) (Ariel and Serhan, 2007; Bannenberg and Serhan, 2010). SPMs are detectable in the brain and their level can be modulated by dietary supplementation. They have potent anti-inflammatory and pro-resolutive properties and we confirm the main role of nutrition as an environmental factor that greatly influences the inflammatory response. It is important to determine if inflammatory pathologies are due to unresolved inflammation, attributed to a decrease in n-3 LC-PUFA dietary intake leading to a decrease in SPM levels, to an enzyme or receptor polymorphism, to a de-regulation of SPM receptors or to a decrease in their expression. We should also determine if these mechanisms could be restored by n-3 LC-PUFA or SPM analog administration.

### AUTHOR CONTRIBUTIONS

All authors (CJ, CR, SL) contributed to the writing of the manuscript.

ChemR23 to regulate inflammation. *J. Immunol.* 178, 3912–3917. doi: 10.4049/ jimmunol.178.6.3912


ageing (inflammageing) and the role of nutrition. *Ageing Res. Rev.* 40, 95–119. doi: 10.1016/j.arr.2017.09.001


polyunsaturated fatty acids, exerts protective actions in murine models of colitis. *J. Immunol.* 191, 4288–4298. doi: 10.4049/jimmunol.1202743


V1/A1, inflammatory pain, and spinal cord synaptic plasticity in mice: distinct roles of resolvin D1, D2, and E1. *J. Neurosci.* 31, 18433–18438. doi: 10.1523/ JNEUROSCI.4192-11.2011


**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 © 2019 Joffre, Rey and Layé. 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.*

# NVP-BEZ235 (Dactolisib) Has Protective Effects in a Transgenic Mouse Model of Alzheimer's Disease

*Paula Maria Quaglio Bellozi1†, Giovanni Freitas Gomes1†, Leonardo Rossi de Oliveira1, Isabella Guimarães Olmo2, Érica Leandro Marciano Vieira3, Fabíola Mara Ribeiro2, Bernd L. Fiebich4\* and Antônio Carlos Pinheiro de Oliveira1\**

1Department of Pharmacology, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil, 2Department of Biochemistry and Immunology, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil, 3Department of Internal Medicine, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil, 4Neuroimmunology and Neurochemistry Research Group, Department of Psychiatry and Psychotherapy, Medical Center—University of Freiburg, Faculty of Medicine, Freiburg im Breisgau, Germany

#### Edited by:

Pietro Giusti, University of Padova, Italy

#### Reviewed by:

Esther Gramage, Universidad San Pablo CEU, Spain Jorge Valero, Achucarro Basque Center for Neuroscience, Spain

#### \*Correspondence:

Bernd L. Fiebich bernd.fiebich@uniklinik-freiburg.de Antônio Carlos Pinheiro de Oliveira antoniooliveira@icb.ufmg.br

†These authors have contributed equally to this work

#### Specialty section:

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

Received: 08 April 2019 Accepted: 24 October 2019 Published: 13 November 2019

#### Citation:

Bellozi PMQ, Gomes GF, de Oliveira LR, Olmo IG, Vieira ÉLM, Ribeiro FM, Fiebich BL and de Oliveira ACP (2019) NVP-BEZ235 (Dactolisib) Has Protective Effects in a Transgenic Mouse Model of Alzheimer's Disease. Front. Pharmacol. 10:1345. doi: 10.3389/fphar.2019.01345

Alzheimer's disease (AD) is a neurodegenerative disease and the main cause of dementia. Its major symptom is memory loss, which is a result of neuronal cell death, which is accompanied by neuroinflammation. Some studies indicate the overactivation of the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (Akt)/mechanistic target of rapamycin (mTOR) pathway in this disease, being, thus, a potential target for pharmacological treatment. Here, we used a transgenic mouse model of AD that expresses a mutant amyloid-β precursor protein (T41 mice) to investigate the effects of dactolisib (alternative name: NVP-BEZ235, abbreviation BEZ), a dual PI3K/mTOR inhibitor. Ten-months-old T41 animals were treated for 14 days with BEZ or vehicle via oral gavage and then submitted to social memory, open field and contextual conditioned fear tests. Hippocampal slices were prepared and Aβ1-42 content, NeuN, Iba-1, CD68 and GFAP were evaluated. Tissues were further processed to evaluate cytokines levels through cytometric bead array. The treatment with BEZ (5 mg/kg) reduced social memory impairment in T41 mice. However, BEZ did not have any effect on altered Aβ levels, NeuN, or GFAP staining. The drug reduced the CD68/Iba-1 ratio in CA3 region of hippocampus. Finally, BEZ diminished IL-10 levels in T41 mice. Thus, although its mechanisms are not clear, BEZ protects against memory impairment, reduces microglial activation and reestablishes IL-10 levels, revealing beneficial effects, which should be further investigated for the treatment of AD.

Keywords: Alzheimer's disease, dactolisib, NVP-BEZ235, neurodegeneration, neuroinflammation, PI3K, mechanistic target of rapamycin

## INTRODUCTION

Alzheimer's disease (AD) is the leading cause of dementia and is characterized as a progressive neurodegenerative disease (Romberg et al., 2012), whose main risk factor is aging (O'neill, 2013). The main sign of AD is memory loss (El Haj and Kessels, 2013), which is directly related to hippocampal and cortical dysfunctions (Wirth et al., 2013).

**Abbreviations:** AD, Alzheimer's disease; Aβ, amyloid-β; APP, amyloid precursor protein; BEZ, NVP-BEZ235; BSA, bovine serum albumin; CBA, cytometric bead array; CA, cornus ammonis; DG, dentate gyrus; i.p., intraperitoneal; PBS, phosphate buffered saline; SEM, standard error of the mean; TBS, tris buffered saline; TBST, TBS-Tween-20; T41, Tg(Thy1-APPSweLon)41Ema.

Neuronal toxicity, which occurs as a consequence of amyloid-β (Aβ) accumulation and tau hyperphosphorylation, results in synaptic loss, neuronal death, and brain atrophy (Lazzari et al., 2015; Kocahan and Dogan, 2017), especially in the entorhinal cortex and hippocampus (Hampel et al., 2002; Holtzman et al., 2011). Synaptic density decrease observed in mesial temporal regions in the early stages of AD correlates with cognitive deficits, since these regions are responsible for the formation and storage of new information (Brouillette, 2014).

Another important phenomenon that occurs in AD is neuroinflammation, which can be mediated by glial cells, especially microglia (Hurley and Tizabi, 2013). Despite evidences that microglia initially removes and degrades Aβ (Krabbe et al., 2013), they lose this ability with the progression of the disease, but they are still capable of producing proinflammatory cytokines (Hickman et al., 2008). Hypertrophic reactive astrocytes are also a hallmark of AD (Frost and Li, 2017), and they are found associated with Aβ in the brain (Chun and Lee, 2018). Alongside with microglia, astrocytes can mediate the clearance of Aβ (Cai et al., 2017; Garwood et al., 2017). The increase of several inflammatory cytokines induce activation of these glial cells, which can contribute to Aβ overgeneration, metabolic misbalance, and problems associated with glutamatergic dysfunction and excitotoxicity (Ourdev et al., 2015; Chen et al., 2016; Cai et al., 2017; Gonzalez-Reyes et al., 2017).

Different pathways contribute to the maintenance and progression of AD. The phosphatidylinositol 3-kinase (PI3K)/ protein kinase B (Akt)/mechanistic target of rapamycin (mTOR) signaling pathway regulates cell metabolism, growth, and survival (Kitagishi et al., 2012), being also fundamental for healthy aging. Several studies have shown that in AD early stages, abnormal and continuous activation of PI3K/Akt/mTOR signaling occurs, with increased phosphorylation of mTOR, contributing to disease progression and cognitive decline (Bhaskar et al., 2009; O'neill, 2013). PI3Kγ inhibition and the double inhibition of PI3K and mTOR after Aβ intracerebral injection reduced pathological changes associated with AD (Passos et al., 2010; Bellozi et al., 2016). Importantly, inhibitors of PI3K and mTOR enzymes have been developed in order to elucidate their participation in several diseases and as a possible treatment strategy (Mukherjee et al., 2012). Both enzymes can be inhibited by dactolisib (alternative name: NVP-BEZ235, abbreviation BEZ), a drug that has undergone several clinical trials for the treatment of different tumor types (Maira et al., 2008; Clinicaltrials.Gov, 2015).

Considering that the PI3K/Akt/mTOR pathway is involved in neuroinflammation and neurodegeneration, it is important to establish its roles in AD. Although we have recently demonstrated that the dual inhibition of PI3K and mTOR by BEZ reverses neuropathological changes induced by Aβ (Bellozi et al., 2016), it is important to investigate the effects of the drug in other models, such as transgenic animals, which better resemble the pathological condition (Bilkei-Gorzo, 2014). Therefore, we studied the potential neuroprotective effects of BEZ in a transgenic mouse model that overexpresses APP.

## MATERIAL AND METHODS

### Drugs

The following substances were used in this study: NVP-BEZ235 (BEZ; LC Laboratories, Woburn, EUA), 10% ketamine hydrochloride (Syntec, Brazil), and 2% xylazine hydrochloride (Syntec, Brazil).

## Animals

All procedures were approved by the institutional Ethic Committee on Animal Use (protocol 159/2012) and followed the NIH guide for the care and use of laboratory animals. Experiments were conducted using 10-months-old male transgenic Tg (Thy1- APPSweLon) 41Ema (T41) mice and their wild-type (WT) littermates (Faizi et al., 2012). Male T41 mice were kindly donated by Prof. Tony Wyss-Coray (Glenn Center for aging animal facility, Stanford, USA), and mated with C57Bl/6 female, from Animal Care Facilities of Federal University of Minas Gerais. Animals were kept under controlled room temperature (24°C), under 12 h: 12 h light-dark cycle, with free access to food and water.

### Animals Genotyping

We performed DNA extraction from the tail of the animals followed by polymerase chain reaction (PCR) in the presence of primers specific for the mutant APP and electrophoresis in agarose gel to identify the mutant sequences in the region of 364 base pairs.

### Experimental Protocol

Animals were treated by oral gavage with 5or 25 mg/kg of BEZ, diluted in 1-metyl 2-pirrolidone 10% in PEG 300, or vehicle, once a day for 14 days. According to the genotype (WT or T41) and the treatment (BEZ or vehicle), the animals were divided into five groups: WT+Vehicle, WT + BEZ 25 mg/kg (WT + BEZ 25), T41+Vehicle, T41+BEZ 5mg/kg (T41 + BEZ 5), and T41+BEZ 25mg/kg (T41 + BEZ 25). The doses of BEZ were chosen based on previous published data (Bellozi et al., 2016). Animals were weighed every day before drug administration and the volume of vehicle and drug solution administered was 4 ml/kg of animal weight. Behavioral tests started 30 min after the administration of the drug or vehicle.

### Social Memory Test

Social memory test was performed on the 11th day of treatment, in an acrylic box measuring 60 cm × 40 cm × 23 cm, subdivided in three chambers of equal size, with communication to each other through passages. For habituation, WT+Vehicle (n = 11), WT + BEZ 25 (n = 10), T41+Vehicle (n = 6), T41 + BEZ 5 (n = 6), and T41 + BEZ 25 (n = 6) animals were introduced for 5 min into the central chamber, with both passages closed. To perform the test, two juveniles C57Bl/6 animals, obtained from a different source of the test animals, were used. One min after the habituation phase, the first juvenile was introduced into one of the lateral chambers, inside a compartment. The passages were opened for 10 min. Then, the test animal was driven to the central compartment, the passages were closed, and a new C57Bl/6 juvenile animal was introduced into the other side chamber, inside another compartment. One min after the second phase, the passages were opened and the interaction of the test animal with both juveniles was evaluated for further 10 min. Each trial (the three phases of the test) was performed for each animal before starting the next. The percentage of time spent in the compartment with the new juvenile animals was calculated using the formula: 100\*[time exploring compartment with new juvenile animal/(time exploring compartment with old animal + time exploring compartment with new juvenile animal)] (Nadler et al., 2004; Kobayashi and Chen, 2005). Tests were recorded and videos were analyzed using ANY-maze software version 4.99.

### Open Field Test

Open field test was performed on the 12th day of treatment. Animals were introduced in the center of a 30cm diameter open field during 10 min, and the total traveled distance was assessed. Experimental groups and sample sizes were the same used for social memory test.

### Contextual Conditioned Fear Test

On the 13th day of treatment, animals were habituated for 3 min in a conditioning fear chamber containing a grid bottom, dimensions 23 cm × 20 cm × 21 cm. After that, a cycle of 3 shocks of 600 mA for 2 s was started, with intervals of 30, 60, and 40 s, respectively. After the last shock, animals were left into the chamber for 1 min more. On the following day, they were reintroduced into the chamber, and the freezing time was measured during 5 min (Roy et al., 2016). Experimental groups and sample sizes were the same used for social memory test.

### Intracardiac Perfusion, Brain Slice Preparation, and Tissue Dissection

On the last day of treatment, 1 hour after drug administration, a subgroup of animals [WT+Vehicle (n = 7), T41+Vehicle (n = 6) and T41+BEZ 5 (n = 6)] was intraperitoneally anesthetized with ketamine (80 mg/kg) and xylazine (8 mg/kg) and perfused with PBS. Then, animals were decapitated, brains were removed, stored in PFA 4% overnight, and subsequently moved to a 30% sucrose solution until complete saturation. Brains were frozen, stored at −80°C (Gage et al., 2012), and posteriorly sliced into 30-µm-thick sections at −20° C with the aid of a cryostat.

Another subgroup of animals [WT+Vehicle (n = 11), T41+Vehicle (n = 6) and T41 + BEZ 5 (n = 5)] had their hippocampus carefully dissected 1 h after drug administration, and stored at −80°C, until the day of analysis.

### Histological Analysis

Free-floating slices from perfused animals were incubated with citrate buffer at 70°C for 1 h for antigen retrieval, followed by blocking solution [BSA (4%), Triton X (0.5%) in TBS] for 1 h. Then, the primary antibodies rabbit anti-Fox3/NeuN (1:800; EnCor, USA); rabbit anti-Aβ 1-42 (1:1,600; 1-42 specific; D9A3A, Cell Signaling, USA), mouse anti-GFAP (1:800; Cell Signaling, USA), rabbit anti-Iba-1 (1:500; Wako, Japan), or rat anti-CD68 (1:500; Bio-Rad, USA) were added and incubated overnight. On the next day, the secondary antibodies donkey antirabbit (1:1,000; Alexa Fluor 594, Invitrogen, USA), goat antimouse (1:1,000, Alexa Fluor 488, Life Technologies, USA)or goat antirat (1:1,000, Alexa Fluor 488, Life Technologies, USA) were added for 1h. Slices were mounted in gelatinized slides and coverslipped with Fluoromount media (Sigma-Aldrich, USA).

Immunostaining was analyzed under a Zeiss fluorescence microscope in 20X/0.4NA magnification lens. In order to obtain a broad and representative perspective of the whole region that was being evaluated, separated slices ranging from −2.06 to −2.54 mm relative to bregma were used. In order to make a proper comparison, equivalent regions containing similar portions were chosen for all the groups. Three slices per animal were used. One picture per region (CA1, CA3, or DG) of each of the three slices were taken. Each picture contained the oriens, pyramidal, and lacunous-molecular layers in the case of CA1 and CA3 regions. In the case of DG, the molecular, granular and polymorphic layers were included in the images and all of them were analyzed together. Each picture had dimension of 710 µm × 530 µm (1388 × 1040 pixels) and resolution of 1.96 pixels/µm. The boundaries of the CA1, CA3, and DG layers of hippocampus were determined by anatomical delimitation as previously established (George Paxinos, 2001). Photomicrographs of stained fluorescence were quantified with the aid of ImageJ software (NIH), and the whole picture was used for quantification. To perform immunostaining analysis, images were converted to 8-bits type, then the threshold method with default algorithm was applied, followed by the percentage of area occupied, mean intensity or integrated density quantification by the analyze particle method, with size (pixels^2) of 0-infinity range and circularity of 0.0–1.0 range. For GFAP, NeuN, Iba-1, and CD68 staining, threshold level was selected as automatically provided by the software. For Aβ1-42 staining, it was established a minimum size of 20 pixels (9 µm) with the analyze particle tool in order to avoid the detection of unspecific objects. The integrated density (IntDen) reflects the product between the area and the mean gray value and it was used in addition to mean intensity. In order to evaluate the percentage of Iba-1-stained area counterstained by CD68, CD68/Iba-1 ratio was obtained by the ratio between CD68 and Iba-1 stained areas. Finally, the number of Iba-1+ cell clusters was counted using 10X (0.25 of numerical aperture) magnification photomicrographs.

### Cytokine Analysis

Hippocampi from dissected animals were homogenized in 200 µl of a buffer containing protease inhibitors (0.4 M NaCl; 0.05% Tween 20; 0.5% BSA; 0.1 mM phenylmethylsulfonyl fluoride; 0.1 mM benzethonium chloride; 10 mM EDTA; 20 IU aprotinin in PBS). Total proteins were measured by Bradford method (Bradford, 1976) and analyzed by cytometric bead array (CBA) with the Th1/Th2 kit (BD, USA) to detect IL-2, IL-4, IL-5 IL-6, IFN-γ, TNF-α, IL-10, and IL-17A. All the procedures followed manufacturer's instructions.

### Statistical Analysis

Statistical analysis was performed using the statistical software graphpad prism 6.0 and statistica 7.0. Quartile extreme test for identification of outlier values was applied on the results, and extreme values were removed before analysis with the aid of the interquartile range method. body weight data were analyzed by two-way Analysis of Variance (ANOVA), followed by Newman-Keuls Posttest. Behavioral, biochemical, and histological data were analyzed by one-way ANOVA followed by Newman-Keuls post-test. the data were presented as mean±standard error of the mean (SEM). The level of significance was set At P < 0.05.

### RESULTS

### BEZ Does Not Alter Body Weight

We first tested whether BEZ would alter body weight, in the same groups of animals submitted to behavioral tests. We did not observe differences between experimental groups (F(4,12) = 0.804, p = 0.545) and there was no interaction between variables (F(44,132) = 1.290, p = 0.137). However, there was effect of the variable time, with reduction of the mean body weight in all experimental groups (F(11,132) = 2.123; p = 0.022; **Figure 1A**).

### BEZ Rescues Social Memory in T41 Mice

Memory loss is the main sign of AD. Therefore, we aimed to evaluate T41 animal's cognition in the social memory test. T41 mice explored for less time the new juveniles compared to WT + Vehicle and WT + BEZ 25 groups. Importantly, BEZ 5 mg/ kg significantly reversed this memory impairment in T41 mice [F(4,33) = 3.596; p = 0.014; **Figure 1B**]. We did not observe any difference in total distance traveled in the apparatus between

groups [F(4,32) = 1.308; ns; **Figure 1C**]. Mean ± SEM of the total time of exploration of each compartment in 1st and 2nd sessions and the total distance traveled in the 1st session can be found in **Supplementary Figure 1**.

### BEZ Does Not Change Locomotor Activity in T41 Mice

Memory loss may be accompanied by locomotor changes. Thus, we also evaluated animal's locomotor activity in an open field. Treated and nontreated T41 mice did not exhibited any locomotor changes when compared with WT treated or nontreated groups [F(4,31) = 0.577; ns; **Figure 1D**].

### T41 Mice Do Not Reveal Altered Behavior in Contextual Conditioned Fear Test

Contextual conditioned fear test was used to assess memory through another behavioral paradigm. However, T41 mice did not present memory impairment in this task and both BEZ 5 and 25 mg/kg did not change the behavior [F(4,31) = 0.494; ns; **Figure 1E**].

### BEZ Does Not Change the Increase in Aβ1-42 Plaque Load in the Hippocampus of T41 Mice

Considering that the higher dose of BEZ did not alter the behavior tasks, we further evaluated the effect of the lower dose in different parameters associated with AD. Since we observed that BEZ rescued social memory in T41 mice, we investigated whether this effect was related to changes in Aβ plaque load. Aβ1- 42 was increased in CA1 (F(2,15) = 5.496; p = 0.0162), in CA3 (F(2,15) = 16.47; p = 0.0002) and in the DG (F(2,15) = 5.606; p = 0.0152) of the hippocampus of T41 mice treated with vehicle. BEZ treatment did not reverse the increased Aβ1-42 load observed in T41 mice (**Figures 2A, B**).

### BEZ Does Not Alter the Immunostaining of NeuN in the Hippocampus of T41 Mice

In order to further investigate the possible neuroprotective effect of BEZ, we evaluated its effect in NeuN immunostaining. There was no alteration in NeuN stained area in CA1 (F(2,16) = 1.798; ns), CA3 (F(2,16) = 2.252; ns), and DG (F(2,16) = 0.2116; ns) regions of the hippocampus, when comparing WT and T41 nontreated or treated mice (**Figures 2C, D**).

### T41 Mice Present Astrocytosis in the Hippocampus, Which Is Not Altered by BEZ

Astrocytosis is a common event observed in AD that may also contribute to the progression of the disease. Therefore, we evaluated whether T41 mice presented changes in GFAP intensity and whether BEZ would alter this scenario. There was an increase in the GFAP mean intensity (F(2,14) = 5.132; p = 0.021) and GFAP IntDen (F(2,14) = 4.077; p = 0.040) in CA1 in nontreated T41 mice, as compared with their WT littermates. In CA3 region, we observed only a trend towards an increase in GFAP mean intensity (F(2,15) = 2.757; ns) but an increase in GFAP IntDen (F(2,14) = 5.468; p = 0.017) in nontreated T41 group, as compared to nontreated WT group. In the DG, both GFAP mean intensity (F(2,15) = 4.044; p = 0.039) and IntDen (F(2,14) = 4.762; p = 0.026) were significantly increased in nontreated T41 mice, in comparison with WT. We also noticed that the treatment did not reverse the increase in neither GFAP mean intensity nor GFAP IntDen found in nontreated T41 mice. In addition, we observed an increase in GFAP IntDen in CA3 of treated T41 mice, as compared with nontreated WT mice (**Figures 2E–G**).

### BEZ Partially Reduces the Increased Microglial Activation in the Hippocampus of T41 Mice

Microglia activation may be a result of neurodegeneration, as well as it could also contribute to the progression of the disease. Since CD68 has been suggested as a lysosomal marker highly expressed in activated macrophages/microglia, and Iba-1 is a marker of microglia, we evaluated whether BEZ would also reduce the ratio between CD68 and Iba-1 immunostaining. Importantly, we did not notice CD68 staining out of Iba-1+ cells. There was an increase in CD68/Iba-1 ratio in CA1 (F(2,12) = 7.796; p = 0.0068), CA3 (F(2,10) = 55.14; p < 0.0001), and DG (F(2,12) = 4.775; p = 0.0298) layers in both T41 nontreated and treated mice. Importantly, BEZ reduced the increase in CD68/Iba-1 ratio in CA3 layer of T41 mice. In addition to the CD68/Iba-1 ratio, we also counted the number of Iba-1+ cell clusters in each layer. We considered as a cluster every Iba-1+ cell agglomerate, which probably occurs around the plaques. The number of cell clusters was increased in CA1 (F(2,12) = 7.267; p = 0.0086) and in DG (F(2,12) = 6.222; p = 0.014) layers in both T41 nontreated and T41 treated mice in comparison with the WT littermates. In CA3 layer, there was an increase in the number of cell clusters in T41 nontreated mice, as compared to WT nontreated mice (F(2,12) = 8.00; p = 0.0062; p < 0.01). However, there was no post-test difference between T41 treated mice and WT nontreated mice or between T41 treated and T41 nontreated mice (**Figure 3**).

### BEZ Reduces Hippocampal IL-10 in T41 Mice, But No Other Cytokines Are Changed

Finally, we evaluated the effects of BEZ on cytokines levels. Among the cytokines evaluated (**Supplementary Table 1**), only IL-10 was found to be increased in T41 mice, which was reversed by treatment with BEZ [F(2,13) = 4.079; p < 0.05; **Figure 4**].

## DISCUSSION

In the present study, we demonstrated that BEZ reversed memory impairment and levels of hippocampal IL-10 in 10-months-old T41 mice. Besides, it reduced CD68/Iba-1 immunostaining ratio in CA3 region of these transgenic animals.

Increased Akt activation and mTOR phosphorylation have been reported in brains of AD patients, which is associated with a disrupted clearance of Aβ and tau, synaptic loss, and cognitive decline (Heras-Sandoval et al., 2014). Thus, the adequate control of PI3K/Akt/mTOR pathway activation might have a potential to ameliorate AD associated features. The dual inhibition of PI3K and mTOR with BEZ 5mg/kg reduced memory loss in T41 mice, without, however, changing the locomotor parameter. The higher BEZ dose did not induce significant memory improvement in this mice strain aged 10 months old. The ineffectiveness of this dose should be evaluated in future studies. It was previously demonstrated that the inhibition of PI3K with LX2343 ameliorates memory loss in APP/PS1 transgenic mouse (Guo et al., 2016). Moreover, the inhibition of mTOR with rapamycin rescues memory impairment (Spilman et al., 2010), since it regulates autophagy leading to a reduction of APP levels, processing and metabolites production, and oxidative stress (Tramutola et al., 2018). Although further studies are necessary, these mechanisms may be related to the BEZ effects in the present study.

(ANOVA followed by Newman-Keuls test). Scale bar for 20× objective = 125 µm. Scale bar for 40× objective = 10 µm.

PI3K and mTOR inhibition have neuroprotective effects in different mice models of neurodegeneration (Malagelada et al., 2010; Bellozi et al., 2016; Saliba et al., 2017). In animal models of AD, administration of rapamycin reduces the accumulation of Aβ, leading to the reduction of synaptic neurotransmission dysfunction (Spilman et al., 2010; Singh et al., 2017). Furthermore, dual inhibition of PI3K and mTOR with BEZ reduces memory impairment induced by intrahippocampal injection of Aβ, which is associated with reduced neurodegeneration and reduced microglial activation (Bellozi et al., 2016). In the present investigation, 14 days of treatment with BEZ did not evidence changes in Aβ1-42 immunostaining, when compared with T41 animals, suggesting that the cognitive improvement mediated by BEZ is related to another mechanism, instead of the modulation of Aβ processing. Since it was previously demonstrated that BEZ reduces Aβ-induced neurodegeneration (Bellozi et al., 2016), we evaluated NeuN immunostaining in T41 mice to assess its potential protective effect. We evaluated CA1, CA3, and DG, since these regions are involved in memory and the neuronal circuits involving them may be affected in AD (Brouillette, 2014; Moorthi et al., 2015). Nevertheless, we did not observe changes in the density of neurons, as indicated by the percentage of stained area with anti-NeuN antibody, nor did BEZ alter this parameter. However, further estimation of the total NeuN+ cells could contribute to the understanding of the possible neuroprotective effect of the tested drug.

Aβ accumulation also leads to neuroinflammation in AD, which is primarily mediated by microglia and astrocytes (Giovannini et al., 2002; Hurley and Tizabi, 2013; Janota et al., 2015). Herein, we observed astrocytosis in hippocampi of T41 animals. It is important to note that GFAP is also expressed by radial glia-like stem cells in the DG. However, BEZ did not induce significant effects in GFAP expressing cells in T41 mice. In addition, we also evaluated the microglia profile. It is known that microglia response can significantly contribute to the chronic neuroinflammation, which has strong relationship with the progression of AD (Li et al., 2018; Nordengen et al., 2019). BEZ reduced the increased CD68/Iba-1 ratio in CA3 region of transgenic animals, suggesting an interference in microglial activation, since CD68 has been suggested as a lysosomal marker highly expressed in activated macrophages/ microglia (Graeber and Streit, 2010; Lehmann et al., 2016; Navarro et al., 2018). Moreover, despite treatment with BEZ did not significantly reduce the number of Iba-1 cell clusters in CA3 of T41 mice, there were no differences between the treated T41 group and WT animals. Despite the decrease in the CD68/Iba-1 ratio induced by BEZ was restrained to CA3 region of T41 mice, it is worth highlighting that the treatment was performed for only 14 days. Thus, although it is possible to speculate that the drug may interfere with microglia activity

FIGURE 3 | Effects of BEZ treatment on CD68 and Iba-1 immunostaining ratio. Bar graphs represent the quantification of BEZ treatment effect on ratio between CD68 and Iba-1 immunostaining (A) or the number of Iba-1+ cell clusters (B) in CA1, CA3, and DG. Representative images are shown in 20× and 40× magnitude (C). For CD68 and Iba-1 immunofluorescence, it was used WT + Vehicle (n = 5), T41 + vehicle (n = 5), and T41 + BEZ 5 (n = 5). \*p < 0.05, \*\*p < 0.01, and \*\*\*p < 0.001 (ANOVA followed by Newman-Keuls test). Scale bar for 20× objective = 125 µm. Scale bar for 40× objective = 10 µm.

and neuroinflammatory process, this pharmacological effect should be further investigated. It is equally relevant to point out the limitations of bidimensional (2D) evaluations, such as the lack of volume data and the reduction of information regarding objects in the three-dimensional (3D) structure (Boyce et al., 2010; Fujisawa et al., 2015), which could be further used to corroborate the present histological findings.

Altered cytokine gene expression and protein levels are also associated with the neuroinflammatory process of AD (Strauss et al., 1992; Janelsins et al., 2008; Morimoto et al., 2011), having protective or nonprotective roles (Su et al., 2016; Zheng et al., 2016). Interestingly, we observed elevated levels of IL-10 in the hippocampus of T41 mice, which was reversed by BEZ. Our finding corroborates previous reports of elevated IL-10 levels in patients (Asselineau et al., 2015) and AD mouse models (Jin et al., 2008). However, in our previous study BEZ increased IL-10 hippocampal levels (Bellozi et al., 2016), a difference that may be related to the age of the animal and stimuli used. Although it is assumed that IL-10 has an antiinflammatory role, its function in AD is still controversial, since it could be either protective or deleterious (Bellozi et al., 2016; Lobo-Silva et al., 2016). We did not find alterations in the other evaluated cytokines in the hippocampus of neither nontreated nor treated T41 animals, as compared to WT mice, which might probably be due to the age.

The effects observed in the present study were not due to body weight change, since the alteration of this parameter along the time was less than 5% and there was no difference between the groups of animals. When comparing the present study results with the model using a hippocampal injection of Aβ, we must consider that the transgenic model better mimics AD, leading to Aβ accumulation in the whole brain (Rockenstein et al., 2001). Herein, a 14-day treatment in a transgenic mouse model expressing a mutant APP was already enough to detect an important cognitive improvement and some benefits related to neuroinflammation, albeit it was not able to change other pathological features. Thus, other mouse models, longer treatment duration or an earlier intervention should be considered to fully understand the potential of BEZ in the treatment of AD.

### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

### ETHICS STATEMENT

All procedures were approved by the institutional Ethic Committee on Animal Experimentation from Federal University of Minas Gerais (CEUA/UFMG) under the protocol number 159/2012. Procedures are in agreement with the Ethical Principles in Animal Experimentation, adopted by CEUA/UFMG, and followed the National Institutes of Health guide for the care and use of Laboratory animals.

### AUTHOR CONTRIBUTIONS

PB, BF, and AO designed the study. IO did the animal genotyping. PB was responsible for animal breeding, performing the treatments, behavioral tasks, removal of tissues, intracardiac perfusion, and slices and tissues preparation. LO helped in behavioral tasks and intracardiac perfusion. GG did histological staining. Cytokines dosages were done by ÉV and PB. Results were analyzed by PB and GG. Article was written by PB, GG, BF, and AO. Authors PB, GG, FR, BF, and AO revised the data and discussed and corrected the manuscript.

### FUNDING

Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG; process numbers CBB-APQ-02044-15 and APQ-02559-17), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq - 424588/2016-1) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). The project was in part supported by the AIF project GmbH (BMWi) (AGEsense).

### ACKNOWLEDGMENTS

We would like to thank Prof. Wyss-Coray for gently donating the T41 mice. We acknowledge CAPI for the use of its infrastructure. AO and FR acknowledge CNPq for the productivity fellowships (process numbers 310347/2018-1 and 304877/2017-4, respectively).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019.01345/ full#supplementary-material

### REFERENCES


mice by targeting both amyloid beta production and clearance. *Acta Pharmacol. Sin.* 37, 1281–1297. doi: 10.1038/aps.2016.80


10

**256**

Frontiers in Pharmacology | www.frontiersin.org


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

*Copyright © 2019 Bellozi, Gomes, de Oliveira, Olmo, Vieira, Ribeiro, Fiebich and de Oliveira. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The Role of Intercellular Adhesion Molecule-1 in the Pathogenesis of Psychiatric Disorders

#### *Norbert Müller\**

*Department of Psychiatry and Psychotherapy, Ludwig-Maximilians-Universität Munich, Munich, Germany*

Intercellular adhesion molecule-1 (ICAM-1) is a transmembrane glycoprotein that is overexpressed in many pathological states. Although, like many other immune molecules, ICAM-1 plays only a limited role in the abundant concert of the immune response, it may be more important than we realize. In the central nervous system (CNS), ICAM-1 is expressed in microglial cells and astrocytes and in endothelial cells in the white and gray matter of the human forebrain. It is of particular interest in psychiatric disorders for two reasons: It has a key function for the blood–brain barrier, which plays an important role in the biology of psychiatric disorders, and it is a marker for inflammation. Although the blood level of soluble ICAM-1 (sICAM-1) might be lower in acute unmedicated schizophrenia, it has been reported to be increased in many other psychiatric conditions, such as major depression, bipolar disorder, and dementia. In bipolar disorder, high sICAM levels were found during both the depressed and the manic states and also during the euthymic phase (the free interval), possibly indicating that sICAM is a trait marker. High sICAM-1 blood levels have also been found in depression comorbid to a somatic disease state. Interestingly, sICAM-1 levels also increase during aging. Some studies investigated sICAM-1 levels in the cerebrospinal fluid of psychiatric disorders and ICAM-1 expression in postmortem CNS tissue of psychiatric patients and found that the overall duration and duration of the chronic phase of the psychiatric disorder seem to play a role in both. Moreover, confounders, such as antipsychotic and antidepressive medication, have to be considered. sICAM-1 levels seem to be associated with hypopermeability or hyperpermeability of the blood-brain barrier and thus to influence the communication between the CNS immune system, represented by glia cells, and the peripheral immune system. The balance between the influx and efflux of immune molecules into and out of the CNS may be one of the pinpoints in psychiatric disorders, in particular in the chronic phase, e.g., in schizophrenia. This aspect, however, needs further intense research, in particular to enable researchers to develop therapeutic principles based on an immune/ inflammatory approach.

Keywords: intercellular adhesion molecule-1, adhesion molecule, schizophrenia, depression, bipolar disorder, immunity, psychoimmunology

### *Edited by:*

*Pietro Giusti, University of Padova, Italy*

#### *Reviewed by:*

*Dietmar Fuchs, Innsbruck Medical University, Austria Bradley D. Pearce, Emory University, United States*

> *\*Correspondence: Norbert Müller norbert.mueller@med. uni-muenchen.de*

#### *Specialty section:*

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

*Received: 28 March 2019 Accepted: 27 September 2019 Published: 22 November 2019*

#### *Citation:*

*Müller N (2019) The Role of Intercellular Adhesion Molecule-1 in the Pathogenesis of Psychiatric Disorders. Front. Pharmacol. 10:1251. doi: 10.3389/fphar.2019.01251*

### INTRODUCTION

Psychotropic drugs such as antidepressants and antipsychotics act on serotonergic, noradrenergic, dopaminergic, and glutamatergic neurotransmission and are effective and well tolerated in the majority of patients. However, around one third of patients with depression are resistant to treatment and even fewer recover. In patients with schizophrenia, the effects of antipsychotics are often unsatisfactory because these drugs cannot prevent the chronic phase. These examples indicate that pathological mechanisms beyond those involving neurotransmitters may play an important role in psychiatric disorders and that research focused on neurotransmitters has reached its limits.

One important argument for the necessity of widening the view on the role of inflammation and the immune system in psychiatric disorders is that no pathogenetic cause has been found for the known dysfunction of monoaminergic neurotransmission, despite intense genetic and biochemical research over the last 40 years. Inflammation and immune dysfunction have direct and indirect influences on neurotransmission and thus may at least partly explain the biological dysfunctions in psychiatric disorders. For example, an activation of the inflammatory response system is well documented in schizophrenia and major depressive disorder (MDD) (Maes et al., 1992; Müller et al., 1993; Maes, 1994; Rothermundt et al., 2001; Myint et al., 2005; Müller and Bechter, 2013). Two meta-analyses on the role of cytokines clearly revealed similar changes in pro- and anti-inflammatory cytokines in various psychiatric disorders, such as MDD, schizophrenia, and bipolar disorder, in both the blood and the cerebrospinal fluid (CSF) (Goldsmith et al., 2016; Wang and Miller, 2018), although they also showed differences between disorders in the levels of the inflammatory markers C-reactive protein, interleukin (IL)-1, and tumor necrosis factor-α (Goldsmith et al., 2016; Wang and Miller, 2018). Other examples are that levels of the pro-inflammatory cytokine IL-6 during childhood predict the risk for later psychosis or depression (Khandaker et al., 2014) and that inflammatory or autoimmune diseases predict the risk for later schizophrenia or mood disorders (Benros et al., 2011; Benros et al., 2013).

In general, the inflammatory response system appears to be activated in psychiatric disorders, but the levels of the different markers vary across studies. The immune system is a highly divergent and differentiated system composed of numerous interconnected molecules. Therefore, at first it may seem unwarranted to discuss the role of just one molecule in psychiatric disorders. Nevertheless, this review will take a closer look at intercellular adhesion molecule-1 (ICAM-1), one of five intercellular adhesion molecules, because it is a well-characterized molecule that has important functions in the process of inflammation and has been studied in various psychiatric disorders as a representative of a pro-inflammatory immune response. It is of special interest because studies indicate that it is differentially expressed in different psychiatric disorders. This review will focus on the role of ICAM-1 in schizophrenia, depression, bipolar disorder, and dementia. It will not discuss ICAM-1 in anxiety disorders because research on immune parameters in this disorder is rare and no studies have been published on ICAM-1 in anxiety disorders.

### FUNCTION OF ICAM-1 AND SOLUBLE ICAM-1

ICAM-1 is an immunoglobulin (Ig)-like transmembrane glycoprotein that is overexpressed on the endothelial lumen in many pathological states (Springer, 1990; Muro, 2007; Lawson and Wolf, 2009). It is ~100 kDa in size and belongs to the immunoglobulin supergene family. The membranebound form of ICAM-1 serves as a counter receptor for the β2-integrins, CD11a/CD18 (LFA-1) and CD11b/CD18 (Mac-1), found on leukocytes. Interactions with membranebound ICAM-1 facilitate leukocyte transmigration across the endothelium of many cell types. A soluble form of the molecule, soluble intercellular adhesion molecule-1 (sICAM-1), is found in serum and other body fluids, including CSF (Schwarz et al., 1998). sICAM-1 may be generated by proteolytic cleavage and/ or alternative splicing of mICAM-1 messenger RNA (Ramos et al., 2014). Like membrane-bound ICAM-1, sICAM-1 interacts with LFA-1/Mac-1 to compete with leukocyte binding to membrane-bound ICAM-1 (Tsakadze et al., 2006) and to stimulate leukocytes (Schmal et al., 1998). Besides endothelial cells, membrane-bound ICAM-1 is primarily expressed not only in microglia cells but also in astrocytes of the central nervous system (CNS), e.g., in the white and gray matter of the human forebrain (Lee and Benveniste, 1999). ICAM-1 expression in the CNS is often associated with glial fibrillary acidic protein (GFAP)-immunoreactive astrocytes (Miguel-Hidalgo et al., 2007); i.e., there is a co-localization of ICAM-1 and GFAP. ICAM-1 is of interest in psychiatric disorders for two reasons: 1) It has a key function in the blood–brain barrier (BBB), which plays an important role in the biology of psychiatric disorders by regulating the movement of molecules from the peripheral body, in particular components of the immune system, into and out of the CNS (see 2.1), and 2) it is a marker for inflammation.

Soluble ICAM-1, a circulating form of ICAM-1, arises from alternative splicing and proteolytic cleavage of membrane-bound ICAM-1 (Ramos et al., 2014). An interesting biological process associated with ICAM-1 overexpression during inflammation and its engagement by leukocytes is that of endothelial release of the ectodomain of ICAM-1, which can then circulate as sICAM-1 (Witkowska and Borawska, 2004; Lawson and Wolf, 2009). The literature describes sICAM-1 as an inflammatory marker and regulator that can promote the inflammatory response (Witkowska and Borawska, 2004; Lawson and Wolf, 2009). Serum levels of sICAM-1 appear to be low in healthy individuals, but they increase in many pathologies and are associated with disease progression and severity, for example, in cancer, cardiovascular disease, immune syndromes, and maladies involving chronic inflammation (Witkowska and Borawska, 2004; Lawson and Wolf, 2009). This association is not merely circumstantial, and sICAM-1 appears to be functionally involved in these diseases (Witkowska and Borawska, 2004). *In vitro* studies have shown a direct correlation between levels of shedded sICAM-1 and cellsurface ICAM-1, suggesting that increased levels of sICAM-1 in the CSF—as in the blood—are indicative of upregulation of surface-bound ICAM-1 in the brain (Leeuwenberg et al., 1992).

When discussing the role of ICAM-1 in psychiatric disorders, it must be noted that increased expression and shedding of ICAM-1 is observed in a wide range of diseases, only a few of which are neuropsychiatric. Inflammation, however, is a general condition of the human body, and the mechanisms of inflammation in psychiatric conditions parallel those of inflammatory disorders in non-psychiatric ones. A specific, important aspect of ICAM-1 in psychiatric disorders might be its role at the BBB.

### ICAM-1 and the BBB

In healthy states, there is an ongoing but limited communication between the CNS and peripheral immune system. In some CNS diseases, however, a disturbance or breakdown of the BBB leads to an uncontrolled invasion of components of the peripheral immune system into the CNS, where they are primed, e.g., by microglial cells; after re-entering the periphery, they act to activate or inhibit the peripheral immune system. Proinflammatory molecules lead to an activation of brain-derived cells, including microglia and astrocytes; these two cell types almost completely surround the capillary endothelial cells that make up the BBB and modulate the barrier *via* effects on these endothelial cells (Benveniste, 1992). During acute inflammation, immune cells invade the CNS parenchyma through the disturbed BBB, i.e., through the endothelium of the small vessels and the tight junctions of astrocytes around the vessels. Activation of endothelial cells and increased expression of ICAM-1 result in breakdown of the BBB and increase leukocyte recruitment, adhesion, and infiltration. This immune cell invasion is mediated by cytokines, chemokines, adhesion molecules, and other mediators of inflammation, and ICAM-1 is a key molecule in this process. The physiological purpose of this process is to identify and eliminate antigens in the CNS. For example, infections with a virus that does not activate glial cells have a considerably more unfavorable course than do infections with a virus that does stimulate an immune response that involves astrocytes (Lewandowski et al., 1994).

Abnormalities in the CSF that indicate a disturbance of the BBB are described in at least 20% to 30% of psychiatric patients. One study found that the BBB was disturbed in 27% of patients with schizophrenia, for example, and IgG was produced intrathecally in 15% of these patients (IgG was measured by the relatively "rough" method of analyzing albumin and IgG in the CSF) (Müller and Ackenheil, 1995). Interestingly, the amount of IgG in the CSF correlated significantly with the psychopathology of schizophrenia, in particular with negative symptoms. In MDD, pathological changes in the CSF were observed in about 25% to 30% of patients (Hampel et al., 1995). Much more severe disturbances of the BBB are seen in acute, fulminant inflammatory processes, e.g., bacterial or viral meningitis or encephalitis. The signs of a mild inflammatory process in a range of mental disorders prompted Müller and Bechter to formulate the "mild encephalitis hypothesis" (Müller and Bechter, 2013). The significant correlation between psychopathology and IgG levels indicates close links between immune and disease processes, at least in schizophrenia.

### ICAM-1 IN PSYCHIATRIC DISORDERS

As described above, sICAM-1 can be determined in body fluids, whereas ICAM-1 expression is analyzed in tissue, including CNS tissue. The results of sICAM-1 and ICAM-1 evaluations in different psychiatric disorders are described below.

### Schizophrenia

A study of the *peripheral blood* in schizophrenia showed lower sICAM-1 levels in the blood of unmedicated patients than in healthy controls (Schwarz et al., 2000). During antipsychotic treatment lasting about 12 weeks, the study found a trend toward an increase in sICAM-1 and detected a relationship between sICAM-1 blood levels and schizophrenic psychopathology: Negative symptoms correlated positively with sICAM-1 levels, indicating that higher levels of sICAM-1 are associated with a more severe schizophrenic negative syndrome (Schwarz et al., 2000). The view that sICAM-1 levels are influenced by antipsychotic medication is supported by a study on the ICAM-1 ligand leucocyte function antigen-1 (LFA-1), which found that LFA-1 expression on leucocytes increased significantly during antipsychotic therapy (Müller et al., 1999). A recent study assessed sICAM-1 (and additional inflammatory adhesion molecules) in the plasma of 78 patients with schizophrenia or schizoaffective disorder and 73 healthy controls (Cai et al., 2018); interestingly, the authors found that sICAM-1 was significantly elevated in schizophrenia. However, the paper gives no information on antipsychotic treatment, so that a role of medication in the increase in sICAM-1 levels cannot be excluded. Another study described an association of sICAM-1 plasma levels with the stage of schizophrenia: In the early stage of schizophrenia, levels of sICAM-1 did not differ from those in healthy controls, but in the late stage of the disease, levels were higher than those in controls (Stefanovic et al., 2016). Of interest is that there was a significant relationship between sICAM-1 levels (but not levels of vascular cell adhesion molecule, VCAM-1, a cell adhesion molecule expressed on blood vessels) with clinical features; i.e., in the early stage of schizophrenia, sICAM-1 was related to the severity and type of current psychopathology, whereas in the late stage, it was associated with the progression of the disorder (Stefanovic et al., 2016). In the early stage of schizophrenia, patients with higher acute sICAM-1 levels had a less favorable treatment response. Similarly, in the late stage of schizophrenia, patients with a longer duration of untreated psychosis and shorter lifetime exposure to treatment had higher sICAM-1 levels compared with the early episode. In addition, higher daily chlorpromazine-equivalent doses applied at the initiation of the psychotic episode and maintained to the remission phase resulted in significantly lower sICAM-1 levels in patients in the late stage of the disease (Stefanovic et al., 2016). These findings are only partly in line with those of Schwarz and colleagues reported above (Schwarz et al., 2000). Many studies have described an increase of pro-inflammatory markers in patients with a longer duration of disease and longer treatment period. A possible explanation why Stefanovic and colleagues did not find decreased sICAM-1 in unmedicated patients with schizophrenia may be that the patients in their study were in different clinical and psychopathological states.

Levels of **sICAM-1 in the CSF** better reflect the environmental condition of the CNS than do blood levels. However, valid CSF studies are lacking. A small study of sICAM-1 levels in the CSF of schizophrenia patients found significantly lower levels than those in healthy controls and in patients with a non-inflammatory neurological disorder (Schwarz MJ, unpublished results). Another study showed that sICAM-1 levels in the CSF are related to a disturbance of the BBB (Schwarz et al., 1998), a finding that supports the role of ICAM-1 in the BBB.

A recent study in schizophrenia patients estimated **ICAM-1 expression in the brain**. ICAM-1 mRNA expression in the prefrontal cortex was compared in a "high inflammation" schizophrenia subgroup—defined *a priori* acording to a cluster of cortical pro-inflammatory cytokine mRNAs in the patients—a low inflammation subgroup, and healthy controls. The study found that mRNA expression was higher in the "high inflammation" subgroup than in the "low inflammation" subgroup and controls (Cai et al., 2018). The role of antipsychotic medication and other confounding factors, however, needs to be evaluated in further studies.

An immunogenetic influence of the ICAM **G241A** or **A469G polymorphisms** in schizophrenia could not be detected, and no difference in sICAM-1 levels was found (Riedel et al., 2003). In healthy control persons carrying the polymorphic A allele (G241A), however, markedly lower sICAM-1 serum levels were found than in carriers of the homozygous GG wild type (G241G) (p < 0.004) (Kronig et al., 2005).

An overview of studies on ICAM-1 in schizophrenia is provided in **Table 1**.

### MDD

In the past decade, a huge number of studies have been published on pro- and anti-inflammatory markers, and a role for inflammation in at least a subgroup of depressed patients is well established. Interestingly, the role of ICAM-1 has been poorly evaluated, in particular in MDD patients without co-morbidities. Because accompanying depressive states often occur in somatic disorders and especially in inflammatory disorders—one of the starting points for research on the role of inflammation in depression—the function of ICAM-1 has often been studied in depressed states comorbid to somatic disorders. Moreover, because aging is associated with an increase in the pro-inflammatory immune state in healthy people (immunosenescence) (Fulop et al., 2017), research has focused more on the role of ICAM-1 in elderly people.

Various research groups have studied *blood levels of sICAM-1*, as well as of other adhesion molecules, in late-life depression. A recently published meta-analysis of data from 43,600 participants aged >40 years, which included 9,203 people with depression, found an association between higher sICAM-1 levels and depression (van Agtmaal et al., 2017). Another study examined the relationship between sICAM-1 and depression and found significantly higher sICAM-1 levels in MDD patients after a 3-day wash-out of antidepressants than in age- and sexmatched healthy controls (Baghai et al., 2018). The authors took into account a medication and treatment effect because they described a trend toward an increase in sICAM-1 levels during treatment before discharge. Nevertheless, a 3-day wash-out may be too short to exclude an effect of medication on sICAM-1. So far, six different meta-analyses have presented pooled odd ratios. These analyses showed a significant association between high


*ICAM-1, intercellular adhesion molecule-1; CSF, cerebrospinal fluid; sICAM-1, soluble ICAM-1; FACS, fluorescence-activated cell sorting.*

levels of sICAM-1 (and some other adhesion molecules) and symptoms of depression in different samples of elderly people (Lesperance et al., 2004; Dimopoulos et al., 2006; Thomas et al., 2007; van Sloten et al., 2014; Tchalla et al., 2015; van Dooren et al., 2016). Most of the participants were not diagnosed with depression, but depression scores were higher in people with higher sICAM-1 levels (odds ratio 1.58, 95% confidence interval 1.28–1.96). Functionally, the authors interpreted the results as showing that generalized microvascular dysfunction is associated with depression. Multiple markers of microvascular dysfunction, including endothelial plasma markers and markers of cerebral small vessel disease, were cross-sectionally associated with a higher level of depressive symptoms and depressive disorder. These meta-analyses support the view that higher sICAM levels are associated with more severe psychopathology, here with symptoms of depression.

*sICAM-1 blood levels* in depressive disorder comorbid with a somatic disease, e.g., diabetes, have been studied in different conditions. An often-studied psychoneuroimmunological model is treatment with interferon-alpha. In interferon-alpha–treated patients, an increase in sICAM-1 levels was observed in parallel to an increase in depression scores, and significantly higher sICAM levels were found in more severely depressed patients (Schaefer et al., 2004). sICAM-1 blood levels were also examined in a study that aimed to investigate whether biomarkers of inflammation were associated with symptoms of depression in individuals with recently diagnosed diabetes (Herder et al., 2017). The authors were also interested to investigate whether such associations may differ by diabetes type. In contrast to the findings in people with type 2 diabetes, in people with type 1 diabetes, serum levels of sICAM-1 were positively associated with the score on the German version of the Center for Epidemiological Studies Depression Scale (Allgemeine Depressionsskala, Langversion).

*CSF levels of sICAM-1* (and other pro-inflammatory markers) were assessed in patients after CNS trauma and found to predict the risk for developing a posttraumatic depressive syndrome (Juengst et al., 2015), providing further evidence of the close relationship between sICAM-1 levels and the psychopathology of depression.

The*CNS expression of sICAM-1*may bestreflect the relationship between the psychopathological state of depression and the role of ICAM-1. One interesting study compared the expression of ICAM-1 in CNS tissue in patients with unipolar and bipolar depression, patients with schizophrenia, and healthy controls (Thomas et al., 2004). The study found significant increases in the expression of ICAM-1 (and additional CAMs) in the gray matter of the dorsolateral prefrontal cortex (DLPFC) in the unipolar depression group but no comparable differences between groups in the anterior cingulate cortex (ACC) or occipital cortex (Thomas et al., 2004). It also found a non-significant increase in ICAM-1 in the white matter of the DLPFC in the unipolar depression group (but no increase in the other areas of the brain or of VCAM-1 in any area). Paired tests showed the specificity for the DLPFC in the unipolar depression group only (Thomas et al., 2004). Another group of researchers estimated the expression of ICAM-1 in the choroid plexus (ChP), a highly vascularized tissue in the CNS that produces CSF, does not have a BBB and is an interface between the peripheral and central immune responses (Devorak et al., 2015). The authors found lower ICAM-1 expression in suicide victims with depression than in controls. Because ICAM-1 is expressed by choroid epithelial cells and is thought to support immune cell trafficking in the ChP, this finding might be an indication of decreased immune cell trafficking through the BBB in depression and suicide. ICAM-1 in the choroidal epithelium is also known to be upregulated in response to exposure to acute pro-inflammatory molecules *in vitro*. The observed downregulation of ICAM-1 expression in the ChP may represent a compensatory mechanism that functions to counteract (chronically) increased activation of proinflammatory signaling pathways elicited by chronically elevated cytokine levels, which are frequently observed in depressed patients. A decrease in immune cell trafficking through the BBB from the periphery into the CNS or vice versa could consequently limit central and/or peripheral pro-inflammatory signaling (Devorak et al., 2015). A converse interpretation of the finding might be that under certain psychopathological conditions the BBB shows hypopermeability and thus does not allow adequate molecular interchange between the CNS and peripheral immune system, so that immune activation in the CNS cannot be cleared.

### Bipolar Disorder

As mentioned in the above discussion of the role of ICAM-1 in depression, the majority of studies to date did not examine individuals with a diagnosis of depression, but with a depressed state of varying severity. Accordingly, they did not differentiate between different types of depression, e.g., recurrent depression or bipolar depression. However, some studies have evaluated individuals with bipolar disorder, during either a manic or a euthymic state.

One study found a weak positive correlation between sICAM-1 levels and the scores on the Yale Mania Rating Scale in patients in a manic state (Turan et al., 2014). In their study in patients with bipolar disorder (n = 83), another group of researchers described higher blood levels of sICAM-1 (and higher IL-6, but lower tumor necrosis factor-α and lower sVCAM) in both acute and remission phases of the disorder than in age-, sex-, and bodymass–matched healthy controls (n = 73) (Pantovic-Stefanovic et al., 2016). Similarly, another study found significantly higher blood levels of sICAM-1 in euthymic bipolar patients than in healthy controls (Reininghaus et al., 2016). This study did not find an association with smoking, sex, or the duration of disease, but patients in a progressive stage of the disease had higher levels of sICAM-1 than did those in an earlier stage. Higher sICAM-1 levels were also found in patients on atypical antipsychotics, whereas prophylactic treatment with lithium or antiepileptics showed no association with sICAM-1 levels (Reininghaus et al., 2016). This finding is in line with earlier results showing signs of an immune activation during the euthymic phase (or free interval) of bipolar patients (Müller et al., 1993). It also corresponds with an interesting finding in a sample of healthy people (without a psychiatric diagnosis) who were grouped according to their affective temperament: In a multiple linear regression model, sICAM-1 blood levels were not related to affective states per se but to the state severity of manic symptoms measured with the Yale Mania Rating Scale (Ivkovic et al., 2017).

*In CNS tissue*, increased expression of ICAM-1 was described in both the gray and white matter of the ACC in bipolar patients compared with controls (gray: *p* = .001; white: *p* < .001) and schizophrenia patients (gray: *p* = .016; white: *p* = .025), and modestly increased expression was described in the white matter of the ACC in bipolar patients compared with patients with unipolar depression (*p* = .049). No differences were found in the DLPFC (Thomas et al., 2004).

### Dementia

Aging is the main risk factor for Alzheimer's disease (AD), and a dramatic increase in extravascular ICAM-1 (associated with GFAP-immunoreactive astrocytes) is seen in the orbitofrontal cortex in normal aging. This increase may contribute to an enhanced risk for CNS inflammatory processes during aging (Miguel-Hidalgo et al., 2007). Decades ago, ICAM-1 was found to be increased in age-related neurodegenerative diseases (Akiyama et al., 1993). Accordingly, high ICAM-1 levels can be expected in AD. Besides localizing to the brain vasculature, in AD ICAM-1 also localizes to amyloid plaques, the biological hallmark of the disease, forming extravascular aggregates (Verbeek et al., 1994). It has been shown that these ICAM-1 immunoreactive aggregates are mostly absent in non-demented control individuals, even in the presence of normal vascular ICAM-1 immunoreactivity (Lee and Benveniste, 1999). However, it is still unclear whether age-related changes in extravascular ICAM-1 immunoreactivity occur in the CNS of normal, non-demented individuals, even though inflammatory processes in the CNS are well known to increase with aging (Bodles and Barger, 2004) and increases observed in neurodegenerative disorders could be age related. In animal models of brain injury, levels of mRNA for ICAM-1 and inflammatory cytokines increase more in old than in young rats (Kyrkanides et al., 2001), in parallel with increased immunostaining of GFAP in astrocytes. Furthermore, in humans, there seems to be an age-related increase in astrocytic GFAP immunoreactivity in the cerebral cortex (Prolla and Mattson, 2001) that might be paralleled by changes in key inflammatory molecules, particularly ICAM-1. In fact, astrocytes express ICAM-1 immunoreactivity *in vitro* and *in vivo* after brain damage (Lee and Benveniste, 1999).

The direct association between ICAM-1 and neurodegeneration has not yet been intensely studied. An *analysis of plasma sICAM-1* in patients with AD or Lewy body dementia (LBD) revealed significantly higher sICAM-1 levels in both patient groups than in healthy controls (p < 0.001). *In the CSF*, however, only LBD patients showed higher sICAM-1 levels (and significantly lower sVCAM-1 levels) than AD patients and healthy controls (p < 0.001) (Nielsen et al., 2007). As expected, the study found a strong correlation between sICAM-1 levels and BBB permeability in the AD patients and controls, but not in the LBD patients. A further evaluation of the data revealed a significant correlation between CSF sICAM-1 and kynurenic acid, a metabolite of tryptophan/kynurenine metabolism that has N-methyl-d-aspartate antagonistic effects and is driven by pro-inflammatory cytokines. Interestingly, the levels of kynurenic acid significantly correlated with the AD biomarker phosphorylated tau, supporting the association between sICAM-1 CSF levels and neurodegeneration markers in AD (Wennstrom et al., 2014).

A large study in nearly 800 individuals showed that CSF levels of sICAM-1 (and other markers of neuroinflammation) were increased not only during the dementia stages of AD, but already during the preclinical and prodromal stages of the disease. High levels of sICAM-1 were associated with increased CSF levels of total tau and phosphorylated tau. These associations were found in both Aß-positive and Aß-negative individuals, although the association was stronger in the former. The results were similar for associations between phosphorylated tau and ICAM-1. High levels of sICAM-1 were also associated with cortical thinning (primarily in the precuneus and superior parietal regions) and with subsequent cognitive deterioration in patients without dementia, as measured with the Clinical Dementia Rating Scale Sum of Boxes. Finally, higher levels of CSF ICAM-1 increased the risk of developing AD dementia in patients without dementia (Janelidze et al., 2018). The authors concluded that neuroinflammation and cerebrovascular dysfunction are early events that occur at presymptomatic stages of AD and contribute to disease progression.

*In postmortem tissue* of demented patients, in AD an increased expression of ICAM-1 was detected in amyloid plaques and astrocytes around plaques (Akiyama et al., 1993). A recent study of 143 cases of LBD determined levels of ICAM-1 (and the antiinflammatory molecule CD200) in postmortem CNS tissue, in particular in the temporal and cingulate cortex, and found that high levels of ICAM-1 expression correlated with the density of amyloid plaques and neurofibrillary tangles; moreover, there was a strong positive correlation between phosphorylated tau levels and ICAM-1 (Walker et al., 2017). In this study, ICAM-1 expression correlated more with the typical AD pathology (and phosphorylated tau as a general marker of neurodestruction) than with the typical LBD hallmarks.

### DISCUSSION/CONCLUSION

In recent decades, psychoneuroimmunological research has come into the focus of biological psychiatry, and it is now widely accepted that mild inflammation plays a role in many different psychiatric disorders. The data indicate that ICAM-1 may be a key molecule in various psychiatric disorders. This view is supported by findings that ICAM-1 is related to clinical features of schizophrenia, depression, and bipolar disorder and to specific neurodegeneration markers in dementia. The results of studies on sICAM-1 are more consistent than those of studies on several other adhesion molecules, such as VCAM-1, ICAM-3, platelet endothelial cell adhesion molecule-1, and others, possibly because sICAM-1 is biologically more stable and easier to measure. ICAM-1 has more diverse functions than sICAM-1. On the one hand, it is a marker for inflammation in the periphery and CNS, and on the other, it reflects the permeability of the BBB and the communication between the CNS and peripheral immune system.

Decreased levels of sICAM-1, as partly described in schizophrenia, may result from dysfunctional neuroendocrineimmune communication, wherein an adequate immune response is not mounted or, alternatively, neuroinflammation is prolonged. A decrease in immune cell trafficking through the BBB from the periphery into the CNS or vice versa could consequently limit central or peripheral pro-inflammatory signaling or both (Devorak et al., 2015). Another interpretation of this finding might be that under certain psychopathological conditions the BBB is hypopermeable, which prevents adequate molecular interchanges between the CNS and peripheral immune systems and does not allow immune activation in the CNS to be cleared. Increased levels of sICAM-1 reflect an at least partial breakdown of the BBB, with the invasion of pro-inflammatory molecules into the CNS. This invasion is necessary to clear an infectious or inflammatory process in the CNS; however, once the process is complete, the inflammatory process must be downregulated; i.e., the number of antiinflammatory molecules in the CNS and at the BBB must be increased. A disturbance of this regulatory process might be the key factor in diverse psychopathological states.

As a final point, one has to ask whether the increase in sICAM-1 levels and expression of ICAM on tissue reflect an unspecific process and why this increase in sICAM-1 is specifically associated with different disorders, including depression, schizophrenia, dementia, or bipolar disorder. The answer is provided by postmortem studies of CNS tissue, which reveal that different CNS regions are involved in the neuroinflammatory process in different psychopathological states, e.g., the DLPFC

### REFERENCES


in depression and the ACC in bipolar disorder. *Ex vivo* neuroimaging, e.g., positron emission tomography, may be able to provide additional information in answer to these questions, but to date, no ligand for ICAM has been developed. The differences in the neuroinflammatory process between disorders may also explain why anti-inflammatory treatment has beneficial therapeutic effects in some psychiatric disorders, such as MDD and schizophrenia (Müller, 2015), and prophylactic effects in dementia (Müller et al., 2015).

### AUTHOR CONTRIBUTIONS

NM, the sole author, was responsible for all aspects of the manuscript.

### FUNDING

The work was supported by the Foundation "Immunität und Seele.

### ACKNOWLEDGMENTS

The author thanks Jacquie Klesing, Board-certified Editor in the Life Sciences, for editing assistance with the manuscript.


for an active immune process in schizophrenia. *Neuroimmunomodulation* 12 (1), 54–59. doi: 10.1159/000082364


**Conflict of Interest:** The author declares 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 © 2019 Müller. 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.*

,

# Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats via Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC

Ke Fang1,2,3† , Hua-Rong Li <sup>1</sup>† , Xing-Xing Chen1,2,3, Xin-Ran Gao1,2,3, Ling-Ling Huang1 An-Qi Du<sup>1</sup> , Chuan Jiang<sup>1</sup> , Hua Li 1,4 and Jin-Fang Ge1,2,3\*

<sup>1</sup> School of Pharmacy, Anhui Medical University, Hefei, China, <sup>2</sup> Anhui Province Key Laboratory of Major Autoimmune Diseases, Anhui Institute of Innovative Drugs, Hefei, China, <sup>3</sup> The Key Laboratory of Anti-inflammatory and Immune Medicine, Ministry of Education, Anhui Medical University, Hefei, China, <sup>4</sup> The First Clinical College, Anhui Medical University, Hefei, China

Quercetin is a polyphenol with multiple biological activities, and results of our preliminary study showed that it could shorten the immobility time of mice in the forced swimming test and tail suspending test. The aim of this study was to investigate its effects on the behavioral performance of lipopolysaccharide (LPS)-challenged rats and explore the potential mechanism. The results showed that intragastrical administration of quercetin (40 mg/kg) could improve the bodyweight gain of LPS-challenged rats, increase the saccharin preference index in the saccharin preference test and the novel arm preference index in the Y-maze, and decrease the immobility time in the FST. However, it showed no significant effect on the performance of LPS-challenged rats in the Morris water maze and the plasma concentrations of nesfatin-1, C-reactive protein (CRP), and IL-6. Results of western blot showed that the expression levels of BDNF, Copine 6, p-TrkB, and the triggering receptors expressed on myeloid cells (TREM) 1 were decreased in both the hippocampus and the prefrontal cortex (PFC) of LPS-challenged rats, while the expression of TREM2 was increased. The protein expression of synapsin-1 was decreased in the hippocampus without significant changes in the PFC. These imbalance protein expressions could be balanced by treatment with quercetin. The results suggested that quercetin could alleviate LPS-induced depression-like behaviors and impairment of learning and memory in rats, the mechanism of which might be involved with regulating the BDNF-related imbalance expression of Copine 6 and TREM1/2 in the hippocampus and the PFC.

Keywords: quercetin, nesfatin-1, brain derived neurotrophic factor (BDNF), Copine 6, the triggering receptors expressed on myeloid cells (TREMs), synapsin-1

#### Edited by:

Pietro Giusti, University of Padova, Italy

#### Reviewed by:

Jean-Philippe Guilloux, Université Paris-Sud, France Qian Ren, Hebei Medical University, China

#### \*Correspondence:

Jin-Fang Ge gejinfang@ahmu.edu.cn † These authors have contributed equally to this work

#### Specialty section:

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

Received: 06 April 2019 Accepted: 28 November 2019 Published: 17 January 2020

#### Citation:

Fang K, Li H-R, Chen X-X, Gao X-R, Huang L-L, Du A-Q, Jiang C, Li H and Ge J-F (2020) Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats via Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC. Front. Pharmacol. 10:1544. doi: 10.3389/fphar.2019.01544

## INTRODUCTION

Depression is a prevalent and recurrent mental abnormality that affects human health and economical development worldwide. It has been reported that during 2013–2016, 8.1% of the American adults aging 20 and above suffer from depression in a given 2 week period (Brody et al., 2018). Since first discovered in 1950s, most of the antidepressants are designed to correct the imbalances of neurotransmitters, which are taken as the key factor of regulating mind and emotion. Unfortunately, they are not effective for all the depression patients, and almost all the classical antidepressants share the same defect of long latency. Although the finding of the robust and rapid-onset antidepressant effects of ketamine has shifted the direction of antidepressants research and development (Dhir, 2017), and been taken as the single most important advance in the treatment of depression in over 50 years (Singh et al., 2017), it should be taken into account of the adverse effects including not only the cardiovascular symptoms, but also hallucinations, confusion, and irrational behaviors. Thus, it is urgent to further investigate the pathogenesis of depression and explore new effective antidepressants with less side effects and shorter acting period.

Inflammation has been long taken as one of the cornerstones in the development of depression (Jeon and Kim, 2017). Higher incidence of depression has been found in patients with infection, and depression patients present increased levels of pro-inflammatory cytokines and decreased level of brain-derived neurotrophic factor (BDNF) (Felger and Lotrich, 2013). Creactive protein (CRP) and Interleukin 6 (IL-6) are common inflammatory markers, and elevated levels of CRP and IL-6 are associated with increased risk for psychological distress and depression (Wium-Andersen et al., 2013; Khandaker et al., 2014). Consistently, results of the animal studies reported that administration of lipopolysaccharide (LPS) could significantly elevate the IL-6 mRNA expression in both the brain and the spleen (Szot et al., 2017), resulting in depressive-like behavior and hippocampal microglial activation (Yamawaki et al., 2018). Moreover, the elevated plasma concentrations of CRP and IL-6 in patients with major depressive disorder are reported to be positively related to the increased plasma nesfatin-1 level (Xia et al., 2018), which is an anorexigenic molecule localized widely in the brain and peripheral tissues with multiple biological activities including regulation of feeding and mood (Weibert et al., 2019). Consistently, in our previous study (Ge et al., 2015a), administration with nesfatin-1 could induce depression-like behaviors, accompanied with the increased plasma concentrations of IL-6 and CRP. Moreover, the antidepressant-like effect of ibuprofen, which is one of the common used nonsteroidal anti-inflammatory drugs (NSAIDs), has demonstrated in animal studies (Mesripour et al., 2019; Seo et al., 2019). These findings support an overlapping pathobiology between inflammation and depression (Merali et al., 2003).

It is important to notice that the disruption of BDNF/TrkB signaling pathway and impairment of neuronal plasticity have been implicated in the pathogenesis of depressive disorders (Erickson et al., 2012). Hippocampus and prefrontal cortex (PFC) are key brain regions involved in the pathogenesis of depression and the antidepressant effects. Reduced serum BDNF levels, together with altered expression of BDNF and its highaffinity receptor tropomyosin receptor kinase B (TrkB) in the hippocampus and PFC, have been found in depression state and could be reversed by effective treatment with antidepressants (Gawali et al., 2016; Song et al., 2019; Seo et al., 2019). Results of our previous studies have suggested a positive relationship between the hippocampal expression of BDNF and Copine 6, synaptotagmin I, and synapsin I in stressed rats (Han et al., 2018). Copine 6 is another important molecular that plays a vital role in regulation of synaptic plasticity (Reinhard et al., 2016), via alternating the neurotransmission process (Liu et al., 2018) or morphology (Burk et al., 2018). Knockout of Copine 6 could induce a deficiency of hippocampal long time potentiation (LTP) and learning and memory in mice (Xu et al., 2015). However, little is known about their alternations in LPS-challenged rats.

The triggering receptors expressed on myeloid cells (TREMs) are a family of activating receptors, potentially may be manipulated to alter the inflammatory response (Bouchon et al., 2001). Inflammatory conditions that alter the balance in TREM expression could exert an important influence on homeostatic activity (Owens et al., 2017), resulting in different outcomes in different diseases (Konishi and Kiyama, 2018). Recent years, much attention has been paid to the role of TREM1/2 in the neuropsychiatric diseases including Alzheimer's disease (AD), although the results are not always consistent (Jiang et al., 2016; Suarez-Calvet et al., 2019; Jiang et al., 2017; Sayed et al., 2018). In our previous research, the protein expressions of TREM2 in the hippocampus are different between rats aging 2 months and 6 months, accompanied with a different performance in behavior tasks. These results suggest that the function of TREMs could be affected by a variety of factors, such as disease duration and stage (Jay et al., 2017).

Querctin is a polyphenol derived from many kinds of plants with multiple activities including antiinflammation and neuroprotective effects. Our previous results demonstrated that it could bind to beta amyloid (Ab)1-40 and Ab1-42 in monomer or fibril state. Recently, it has been reported that quercetin could protect against stress-induced anxiety- and depression-like behavior and improve memory in male mice (Samad et al., 2018). Consistently, results of our preliminary study showed that it could shorten the immobility time of mice in the forced swimming test (FST) and tail suspending test, though the mechanism remains unclear.

The aim of the present study is to investigate the effects of quercetin on LPS-induced depression-like behavior and explore its potential mechanism, especially its effect on the BDNF-related protein expression of the key molecules in the hippocampus and PFC.

### MATERIALS AND METHODS

### Drugs

Quercetin was provided by Haoyang Biotechnology Co, LTD. (Xi'an, Shanxi, China) with the purity 95%, and the structure of quercetin is shown in Figure 1. Fluoxetine Hydrochloride (Prozac) was purchased from Eli Lilly Pharmaceuticals. Ibuprofen was the production of Renfu pharmaceutical co., LTD (Yichang, Hubei, China).

### Animals

Fourty-five male Sprague-Dawley rats, aged 2 months, were purchased from Anhui Experimental Anhui Center of China. The rats were assigned randomly to five groups including an unhandled control group (Control), a LPS-injection group (Model), a ibuprofen treatment group (240 mg/kg), a fluoxetine treatment group (5 mg/kg), and a quercetin treatment group (40 mg/kg), with nine rats in each group. After a 7 days' adaptive breeding, rats were administered intragastrically the drugs daily with the dose mentioned above, while equal volume of 0.5% carboxymethyl cellulose (CMC) were given to rats in control group and model groups. During the 3th– 5th day, 0.5 mg/Kg LPS (from Escherichia coli, serotype 055:B5, Sigma-Aldrich) was administered through intraperitoneal injection to rats in model and drug-treated groups, and equivalent sterile normal saline solution was administered to rats in control group. The behavioral experiments were put into operation from the 5th day. The schedule of the experimental design is shown in Figure 2.

Rats were raised and maintained under a 12-h light-dark cycle and housed 4–5 per cage with free access to food and water. The ambient temperature was maintained at 21°C–22°C with 50%– 60% relative humidity. The experimental procedures were approved by the animal Care and Use committee at Anhui Medical University and complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. (NIH publication No.85-23, revised 1985)

### Behavioral Tests

Behavioral tests were performed in a soundproof room with a neutral environment. All the rats in this study including the control ones received all the behavior tasks one by one. As the following description order, the first one was the saccharin preference test (SPT), followed by the open field test (OFT), then Y-maze and Morris water maze (MWM), and the last one

FIGURE 2 | Schedule of the experimental design.

was the FST. All the tests were carried out between 08:30 and 12:30, one hour after intragastric administration of solvent or drug, with matching between the groups. The observers were blind to the treatment. The behavioral tests were monitored and recorded by a digital camera interfaced to a computer running the ANY-maze video imaging software (Stoleting Co.,Wood Dale, IL, USA).

### Saccharin Preference Test

The SPT is one of the routine methods to evaluate the anhedonic behavior in rodents. After a 12-h period of food and water deprivation, all rats were housed solely and had free access to two bottles containing either plain water or 2% saccharin solution. After 6 h, the volumes of water and saccharin consumed were measured. The percentage of saccharin solution ingested, which is also named as saccharin preference index (SPI), is a measure to assess the sensitivity of rats to hedonia.

### Open Field Test

The OFT provides synchronous measures of locomotion, exploration, and anxiety. The apparatus of the test consisted of a black square arena with a length of 100 cm and a height of 30 cm. The floor was marked with white grids dividing it into 16 equally sized squares. During a 5-min observation period, rats were placed at one corner of the apparatus facing the wall. The total distance, duration in the center, and the frequencies of rearing and grooming were recorded. Rats were returned to their housed cages after test, and the open field was cleaned with 75% alcohol and permitted to dry between tests.

### Y-Maze

Y-maze consists of three identical arms with an angle of 120° between each other. Each arm was 10 cm × 48 cm × 20 cm (length× width× height). They were randomly designated as the start arm, the novel arm, and the familiar arm. The novel arm was blocked during the 1st trail but open during the 2nd trail, while the other two arms were always open. The maze was cleaned with 75% alcohol and permitted to dry between tests. The Y-maze test consisted of two trials separated by an intertrail interval to assess spatial recognition memory. The first trail allowed the rats 10 min to explore the start arm and the other arm. Twenty-four hours later, the second trail was conducted. The rat was placed back in the same starting arm, with free access to all the three arms for 5 min, and the percentage of the duration in the novel arm to that in both the novel and familiar arms was taken as the novel arm preference index.

### Morris Water Maze

MWM was used to test the spatial learning and memory. The maze was consisted of black circular pool (diameter 160 cm, height 50 cm, filled with water to 30 cm high containing black edible pigment at 21°C–22°C). A black circular platform (9 cm in diameter) was 2.0 cm below the water line in the center of one quadrant, and remained in the same position. Several constant and large visual cues surrounded the tank at a height of 60 cm to facilitate orientation.

The task consisted of a 3-day acquisition phase with four trails administered each day and a 1-day memory retention test phase. During the acquisition phase, the rat was placed in the water facing the wall at one random start location of four (north, south, east, and west, locating at equal distances from each other on the pool rim). Each rat was allowed to find the submerged platform within 60 s, and rest on it for 30s. If the rat failed to find the hidden platform within 60 s, it was guided to the platform and allowed to remain there for 30 s. The procedure was repeated for all the four start locations.

On the 4th day (the test phase), memory retention was determined in a single 60 s probe trial. The underwater platform was removed. The rats were placed into water from the opposite quadrant of the platform, facing the wall, and were permitted to explore the environment for 60 s ad libitum. Performance parameters of each rat including total swim distance, mean swim velocity, and the duration in each quadrant were monitored and recorded.

### Forced Swimming Test

The behavioral cylinder was 60 cm high with the diameter 25cm, filled with 30 cm of water and maintained at 24°C–25°C. The immobility time was recorded and rats were considered immobile when they did not make any active movements.

### Measurement of the Plasma Concentration of Nesfatin-1, IL-6, and CRP

Twenty-four hours after the last behavioral test, the rats were deeply anesthetized with chloral hydrate, and the blood was drawn from abdominal aorta. The plasma concentrations of nesfatin-1, IL-6, and CRP were measured using commercially available enzyme-linked immunosorbent assay (ELISA) kits (Nesfatin-1: Huamei Biotech. Co., LTD., Wuhan, China; IL-6 and CRP: Yuanye Biotech. Co., LTD., Shanghai, China) based on the manufactures' instructions.

### Western Blot Assays

The hippocampus and the PFC of three rats from each group were promptly dissected, frozen in liquid nitrogen, and stored at −80°C. The tissue was homogenized in radio immuniprecipitation assay (RIPA) buffer (50mM Tris-HCl, pH 7.4, 0.1% SDS, 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, and 1 mM Na3Vo4). A protease inhibitor cocktail (Roche, IN, USA) and the phosphatase inhibitor PhosSTOP (Roche, IN, USA) were added before homogenization. The protein content was measured using a Lowry Protein Assay Kit (Meiji Biotech Co., Ltd., Shanghai, China). The same quantity (50 µɡ) of protein from each sample was loaded and separated by 15% SDS-PAGE and then transferred onto a polyvinylidene difluoride membrane (Amersham Biosciences, UK). The membrane was blocked with 5% skim milk for 1 h; incubated with antibodies targeting BDNF (1:1000; Abcam, Cambridge, UK), TrkB (1:10000; Abcam, Cambridge, UK), phosphorylated TrkB (p-TrkB) (1:10000; Abcam, Cambridge, UK), Copine 6 (1:1,000; Santa Cruz Biotechnology, Inc., Delaware, USA), synapsin-1 (1:1000; ImmunoWay, Newark, Delaware, USA), TREM-1 (1:1000; ImmunoWay, Newark, Delaware, USA), TREM-2 (1:1000; ImmunoWay, Newark, Delaware, USA), or b-actin (1:1000; Zhongshan Biotechnology, INC, Beijing, China) at 4°C overnight, and then incubated with a horseradish peroxidaseconjugated secondary antibody (1:2000) at 37°C for 2 h. The blots were developed with the Easy Enhanced Chemiluminescence Western Blot Kit (Pierce Biotechnology, Rockford, IL, USA). The protein bands were scanned and analyzed using Image J software (NIH). The TrkB/TrkB ratio was observed, and the expression of other proteins was normalized to b-actin and analyzed.

### Statistical Analysis

All statistical analyses were performed using SPSS (version 12.0.1, SPSS Inc., Chicago, IL, United States). Data are expressed as the means ± SEM and P < 0.05 was considered statistically significant. Between-group effects on the escaping latency in the MWM task were analyzed by repeated measures ANOVA with group and time as the factors. Statistical analyses of the effect of quercetin on other parameters were carried out using ANOVA followed by LSD posthoc tests. The correlation analysis was performed by Pearson's correlation test.

## RESULTS

### Quercetin Administration Improved the Decreased Bodyweight Gain in LPS-Challenged Rats

Figure 3 shows the bodyweight of the rats in different group before and after the test. As shown in Figure 3A, the bodyweight increased in all the groups, indicating that all the rats were growing. However, the bodyweight gain of the model ones were less than that of the control rats (F[4, 40] = 4.795, P = 0.031), and this change could be reversed by treatment with quercetin (F[4, 40] = 4.795, P = 0.005) or fluoxtine (F[4, 40] = 4.795, P = 0.001) (Figure 3B).

FIGURE 3 | Effect of quercetin on the bodyweight and bodyweight gain in lipopolysaccharide (LPS)-challenged rats.The data are presented as the mean ± SEM, with nine rats in each group.There was no significant difference among groups with regard to the bodyweight (A), but the bodyweight gain of the model rats in LPS-challenged group were less than that of the control ones, and this change could be reversed in the quercetin- and fluoxetinetreated group (B). \*P < 0.05, compared with the control group; ##P < 0.01, compared with the model group.

### Quercetin Administration Alleviated the Depression-Like Behavior in LPS-Challenged Rats

Figure 4 shows the performance of rats in the OFT. Although there was no significant difference among groups with regards to the total distance (F[4, 40] = 0.377, <sup>P</sup> = 0.317; Figure 4A) and the grooming frequency (F[4, 40] = 0.918, <sup>P</sup> = 0.264; Figure 4D), the LPS-challenged rats spent less time in the center (F[4, 40] = 1.601, <sup>P</sup> = 0.030; Figure 4B) and showed less rearing frequency (F[4, 40] = 3.869, <sup>P</sup> = 0.031; Figure 4C), and treatment with ibuprofen (F[4, 40] = 3.869, <sup>P</sup> < 0.001; Figure 4C) or fluoxetine (F[4, 40] = 3.869, <sup>P</sup> = 0.042; Figure 4C) could significantly increase the rearing frequency. Compared with that in the control group, the model rats showed less SPI in the SPT (F[4, 40] = 4.809, <sup>P</sup> = 0.041; Figure 5A) and longer immobility time in

FIGURE 4 | Effect of quercetin on the performance of lipopolysaccharide (LPS)-challenged rats in the open field test (OFT).The data are presented as the mean ± SEM, with nine rats in each group.In the OFT, there was no significant difference among groups with respect to the total moving distance (A) and grooming frequency (D), but the duration in the center (B) and the rearing frequency (C) were significant decreased in the LPS-challenged group. Treatment with quercetin could increase the rearing frequency (C).\*P < 0.05, compared with the control group; # P < 0.05 and ##P < 0.01, compared with the model group.

the FST (F[4, 40] = 4.829, <sup>P</sup> = 0.039; Figure 5B). However, treatment with quercetin, fluoxetine, and ibuprofen could reverse the changes of SPI and immobility time (P < 0.05 or P < 0.01).

Results of the Pearson's correlation test showed that the immobility time in the FST was negatively correlated to the sucrose preference index (r = <sup>−</sup>0.375, P = 0.012; Figure 5C) and the bodyweight gain (r = <sup>−</sup>0.322, P = 0.031; Figure 5D).

FIGURE 5 | Effect of quercetin on the performance of lipopolysaccharide (LPS)-challenged rats in the saccharin preference test (SPT) and forced swimming test (FST). The data are presented as the mean ± SEM, with nine rats in each group. Compared with that in the control group, the sucrose preference index of rats in LPS-challenged group was lower (A) and the immobility time in the FST (B). These changes could be reversed by treatment with quercetin, fluoxetine, and ibuprofen. The immobility time in the FST was negatively correlated to the sucrose preference index (C) and the bodyweight gain (D). \*P < 0.05, compared with the control group; # P < 0.05 and ##P < 0.01, compared with the model group.

### Quercetin Administration Improved the Performance of LPS-Challenged Rats in Y-Maze Without Significant Influence in the MWM

For all the rats studied in the experiment, the escape latency declined with each day during the acquisition phase, as is shown in Figure 6A. Analyzing the latency using repeated measures ANOVA with experimental treatment as the between-subject factor and training days as the within-subject factor, the results showed that only the training days (F[2, 39] = 65.426, P < 0.01), but not the experimental treatment (F[4, 40] = 2.077, P = 0. 102) had a significant effect on the escape latency, without a significant interaction between each other (F[8, 80] = 1.888, P = 0. 073). In the probe trail, there was no significant difference among groups with regard to the duration in the target quadrant (F[4, 40] = 0.136, <sup>P</sup> = 0.968; Figure 6B).

In the Y-maze task, the LPS-challenged rats spent less time in the novel arm than the controls (F[4, 40] = 2.099, P = 0.011; Figure 6C), together with a decreased novel arm preference (F[4, 40] = 2.669, <sup>P</sup> = 0.008; Figure 6D). These changes could be significantly reversed by treatment of fluoxetine or ibuprofen (P < 0.05). Treated with quercetin could increase the decreased novel arm preference index induced by LPS-challenge (F[4, 40] = 2.669, <sup>P</sup> = 0.028; Figure 6D).

### Quercetin Administration Did Not Change the Plasma Concentration of Nesfatin-1, CRP, and IL-6 in LPS-Challenged Rats

As shown in Figure 7, the plasma concentrations of nesfatin-1 (F[4, 40] = 3.170, <sup>P</sup> = 0.003; Figure 7A), CRP (F[4, 40] = 2.610, <sup>P</sup> = 0.039; Figure 7B), and IL-6 (F[4, 40] = 2.442, <sup>P</sup> = 0.005; Figure 7B) were all remarkably increased in the model group, and there was no significant change after treatment with ibuprofen, fluxetine, or quercetin.

Results of the Pearson's correlation test showed that the plasma CRP concentration was negatively correlated to the novel arm preference index (r = <sup>−</sup>0.315, P = 0.042; Figure 7C), and the plasma IL-6 concentration was positively correlated to the immobility time in the FST (r = 0.365, P = 0.016; Figure 7D).

### Quercetin Administration Balanced the Protein Expression of BDNF, P-TrkB/TrkB, Copine 6, Synapsin-1, TREM1, and TREM2 in the Hippocampus and PFC of LPS-Challenged Rats

Figure 8 shows the key protein expression in the hippocampus and the PFC of rats in all the groups. As compared with that in the control group, the protein expression of BDNF, p-TrkB/ TrkB, Copine 6, and TREM1 were decreased in both the hippocampus and the PFC of LPS-challenged rats, while the expression of TREM2 were increased, and the protein expression of synapsin-1 were decreased in the hippocampus without significant changes in the PFC. These changes could be balanced by treatment of quercetin.

Apart from the negative correlation between the protein expression of TREM1 and TREM2 (Hippocampus: r = −0.635,

FIGURE 6 | Effect of quercetin on the performance of lipopolysaccharide (LPS)-challenged rats in the Morris water maze (MWM) and Y-maze. The data are presented as the mean ± SEM, with nine rats in each group. The escape latency of all the rats was declined with each day during the acquisition phase (A). Results of repeated measures showed that the training days, but not the experimental treatment had a significant effect on the escape latency. In the probe trail, there was no significant difference among groups of the duration in the target quadrant (B). In the Y-maze, the duration or rats in the novel arm was decreased in the model group but significantly increase in the ibuprofenand fluoxetine- treated group (C). Treatment with quercetin could reverse the decreased novel arm preference index of LPS-challenged rats (D). \*P < 0.05 and \*\*P < 0.001, compared with the control group; # P < 0.05, compared with the model group.

FIGURE 7 | Effect of quercetin on the plasma concentrations of nesfatin-1, C-reactive protein (CRP), and IL-6 in lipopolysaccharide (LPS)-challenged rats. The data are presented as the mean ± SEM, with nine rats in each group. Compared with that of the control rats, the plasma concentrations of nesfatin-1, CRP, and IL-6 were all remarkably increased in the model group. There was no significant difference between the model group and the three groups with treatment of ibuprofen, fluoxetine, or quercetin (A, B). Results of the Pearson's correlation test showed that the plasma CRP concentration was negatively correlated to the novel arm preference index (C), and the plasma IL-6 concentration was positively correlated to the immobility time in the forced swimming test (FST) (D). \*P < 0.05 and \*\*P < 0.001 compared with the control group.

FIGURE 8 | Effect of quercetin on the protein expression of brain derived neurotrophic factor (BDNF), Copine 6, TREM1, TREM2, and Synapsin-1 in the hippocampus and prefrontal cortex (PFC) of lipopolysaccharide (LPS) challenged rats (A) shows a typical graph, and (B) and (C) show the statistical analysis of western blotting results. The data in (B) and (C) are presented as the mean ± SEM, with n = 3 for each group. The protein expression of BDNF, Copine 6, and TREM1 were decreased in both the hippocampus and the PFC of LPS-challenged rats, while the expression of TREM2 were increased, and the protein expression of synapsin-1 were decreased in the hippocampus without significant changes in the PFC. These changes could be balanced by treatment of ibuprofen, fluoxetine, and quercetin. \*P < 0.05 and \*\*P < 0.001, compared with the control group; # P < 0.05 and ##P < 0.001, compared with the NAFLD model group. 1 control; 2 Model; 3 Ibuprofen; 4 Fluoxetine; 5 Quercetin.

P = 0.011, Figure 9A; PFC: r = <sup>−</sup>0.758, P = 0.001, Figure 9B), results of the Pearson's correlation test showed that the protein expression of BDNF was positively correlated to that of Copine 6 (Hippocampus: r = 0.923, <sup>P</sup> < 0.001, Figure 9C; PFC: r = 0.603, P = 0.017, Figure 9D) and TREM1 (Hippocampus: r = 0.592, P = 0.020, Figure 9E; PFC: r = 0.547, P = 0.035, Figure 9F), while negatively correlated to that of TREM2 (Hippocampus: r = <sup>−</sup>0.516, P = 0.049, Figure 9G; PFC: r = <sup>−</sup>0.547, P = 0.035, Figure 9H). A positive correlation was found between the expression of BDNF to that of synapsin-1 in the hippocampus (r = 0.613, P = 0.015), but not in the PFC (r = −0.060, P = 0.832).

### DISCUSSION

In the present study, our results showed that quercetin could alleviate the depression-like behaviors induced by intraperitoneal injection of LPS in rats, as indicated by the increased SPI in the SPT and decreased immobility time in the FST. Moreover, the mechanism of its antidepressant-like effects might be associated with its ability in regulating the BDNF-related imbalance of Copine 6 and other synaptic plasticity-related protein expression in the hippocampus and PFC. Additionally, although administration of quercetin could not improve the performance of LPS-challenged rats in the MWM task, it increased the novel arm preference index in the Y-maze.

Inflammation is an important pathogenic factor involved with the development of depression. LPS is a common agent used as inflammatory stimuli. In spite of the different experimental protocol with different dose, frequency or administrating route, it has been reported that LPS-challenged rodents present depression-like behavior (Szot et al., 2017) (Wickens et al., 2017; Yamawaki et al., 2018) with a sexspecific Effect (Millett et al., 2019). In line with these findings, in our present study, LPS-challenged rats showed less bodyweight gain, less exploratory behavior, anhedonia, and despair behaviors. Using directional heading error as the evaluation index, Kupferschmid and his colleagues reported a selective, age-dependent spatial learning impairment in LPSadministered rats (Kupferschmid and Therrien, 2017). In the present study, although there was no significant difference among groups in the MWM tasks, the LPS-challenged rats showed a decline of novel arm preference index in the Y-maze. Additionally, results of the Pearson's correlation test showed a close relationship among the different behavioral indicators. These results indicated again that LPS could be widely used to investigate the relationship between inflammation challenge and the impairment of neuropsychiatric behaviors including anxiety, depression, and memory deficit. As a common used NASID, the antidepressant effect of ibuprofen has been demonstrated in several animal models and human studies (Saleh et al., 2014; Kohler et al., 2015; Mesripour et al., 2019; Seo et al., 2019). Consistently, our results showed that treatment with ibuprofen could increase the rearing frequency and SPI of LPS-challenged rats, together with the decrease of immobility time in the FST. The results suggested again the close relationship between inflammation and depression.

As an established antidepressant, fluoxetine was selected as a positive control in the present study. Consistent with its clinical application, fluoxetine has also been demonstrated to prevent LPS-induced increase in the immobility time in the FST and tail suspension tests (Taniguti et al., 2019). Unsurprisingly, in the present study, rats in fluoxetine-administered group showed improvement in both bodyweight gain and depression-like behaviors. Similar with the results of fluoxetine, administration of quercetin could also improve the bodyweight gain, increase the SPI in the SPT, and decrease the immobility time in the FST, although it had no effect on the performance in the OFT. Together with the findings that quercetin could reduce the immobility time of mice in the FST and tail suspend test in our preliminary study (attached in the Supplementary Materials) and relieve the depression-like behavior of stressed mice (Samad et al., 2018), our results suggested again the antidepressant effect of quercetin. Moreover, our results showed a task-related improvement effect of quercetin on the learning and memory.

Raised inflammatory markers are reported in depressed patients and taken as an important mediating factor for behavior, neural plasticity and brain structure (Valkanova et al., 2013; Lotrich, 2015). There is increasing evidence that inflammatory cytokines can induce or worsen depressive symptoms (Horowitz and Zunszain, 2015). Consistently, it has been reported in animal studies that LPS challenge could induce an increase of pro-inflammatory cytokine response (Ge et al., 2015b; Szot et al., 2017). In line with these findings, our results proved that plasma concentrations of CRP and IL-6 were remarkably increased in LPS-challenged rats. Moreover, our results showed a close relationship between the plasma inflammatory indicators and the behavioral performance. Furthermore, in line with the report that LPS could induce an elevation of plasma nesfatin-1 concentration in rats (Stengel et al., 2011), the plasma concentration of nesfatin-1 of LPS-challenged rats was also increased. These results suggested again the relationship between inflammation and depression. However, administration of quercetin could not reverse the increased plasma concentrations of nesfatin-1, CRP, and IL-6. Although we did not detect their alternation in the brain, our results indicated the hypothesis that there would be other mechanisms underlying the antidepressant-like effect of quercetin.

BDNF is well recognized for its neuroprotective functions, via binding and activating its high affinity receptor TrkB. The alternation of BDNF/TrkB signaling pathway has been demonstrated to play a key role in the pathophysiology of depression and in the therapeutic mechanisms of antidepressant (Zhang et al., 2016). In the present study, the protein expressions of BDNF, p-TrkB/TrkB, Copine 6, and synapsin-1were all reduced in the hippocampus and PFC of LPS-challenged rats, and a positive relationship was found between the protein expression of BDNF and Copine 6 or synapsin-1. Taking together with their important role in linking neuropsychiatric behaviors to synaptic plasticity (Jangra et al., 2014; Reinhard et al., 2016), these results indicated that BDNFrelated imbalance expression of Copine-6 and synapse plasticityassociated proteins play a vital role in the depression-like behavior induced by LPS challenge. However, treatment with fluoxetine, ibuprofen, or quercetin could reverse these alternations. Therefore, it is plausible to infer that, apart from the reported mechanisms including modulating cytokines production and inhibiting oxidative stress (Sah et al., 2011), the antidepressant-like effect of quercetin might be associated with its ability in regulating the BDNF-related imbalance of key proteins expression involved in neuroinflammation and neuroplasticity.

The TREM family proteins are cell surface receptors with important roles in regulation of myeloid cell inflammatory activity, and different TREM receptors appear to have contrasting roles in controlling myeloid immune activity (Owens et al., 2017). Therefore the relative and coordinated regulation of their expression is important to understand, though the inconsistent reports made it more complicated. Although TREM-1 are demonstrated to amplify inflammation in sepsis (Bouchon et al., 2001), soluble TREM-1 has been taken as an antiinflammatory mediator in sepsis (Giamarellos-Bourboulis et al., 2006; Gibot and Massin, 2006). It has been reported that TREM2 can promote microglial cell survival by activating the Wnt/ b-catenin signaling pathway (Zheng et al., 2017), and accelerate the removal of Ab to reduce oxidative stress damage in hippocampal neurons (Fan et al., 2019). Upregulating the expression of TREM-2 could inhibit the apoptosis of hippocampal neurons (Liu et al., 2019), ameliorate neuropathology, and rescue spatial cognitive impairment (Jiang et al., 2014). Loss of TREM2 function increases amyloid seeding but reduces plaque-associated ApoE (Parhizkar et al., 2019), and aggravates spatial learning impairment in P301S transgenic mice (Jiang et al., 2015). However, it has also been reported that the ameliorative effect of TREM2 overexpression in microglia on the behaviors and neuropathological injuries including Ab aggregation, neuroinflammation, and loss of neurons and synapses (Jiang et al., 2014). Similarly, Inconsistent or contrary effects of TREM2 on the microglial injury response and tau pathology were also reported (Bemiller et al., 2017) (Leyns et al., 2017), and even the effect would be varied by partial or complete loss of TREM2 function (Sayed et al., 2018). Additionally, different expression of TREM2 has also been found in the hippocampus between rats aging 2 months and 6 months in our previous study. In the present study, the LPS-challenged rats presented not only depression-like behaviors and impairment of learning and memory, but also the decreased expression of TREM1 and increased expression of TREM2 in the hippocampus and PFC. Despite the small sample size, our results suggested a negative relationship between the protein expression of TREM1 and that of TREM2. Moreover, these alternated expressions of TREM1 and TREM2 could be reversed by treatment of quercetin. These results suggested again the important role of TREMs in regulating the neurological behaviors and the potential effect of quercetin, although more details should be investigated thoroughly in the future.

There are several limitations in our study. Firstly, the multiple behavioral tasks were carried out in the same cohort of rats. Although the confounding factors including time, place, order have been taken into account and controlled, we cannot exclude absolutely the interactions among the behavioral tasks. Secondly, due to the limited data of continuous variables, the Pearson correlation test were carried out based on all the data together. Although it could increase the sample size and enlarge the variation of the observation indicators, making it more convenient and reliable to observe the relationship between the changes of the indicators. However, we cannot exclude the limitation by the influence of different treatments. More detailed investigation should be taken in our further study.

Taken together, our study revealed that quercetin could alleviate LPS-induced depression like behaviors and impairment of learning and memory in rats, the mechanism of which might be partly associated with its ability in regulating the BDNF-related imbalance expression of Copine 6 and TREM1/2 in the hippocampus and PFC.

### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

### ETHICS STATEMENT

The experimental procedures were approved by the animal Care and Use committee at Anhui Medical University and complied

### REFERENCES


with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH publication No. 85-23, revised 1985).

### AUTHOR CONTRIBUTIONS

J-FG designed the study and wrote the protocol. H-RL wrote the first draft of the manuscript. KF and X-XC managed the literature searches and the statistical analyses. KF, L-LH, CJ, A-QD, and HL performed animal model experiments. X-XC and X-RG performed the gene expression experiments and wrote parts of the manuscript. All authors contributed to and have approved the final manuscript.

### ACKNOWLEDGMENTS

This study was provided by the National Natural Science Foundation of China (81401122, 81870403), Foundation of Innovation and entrepreneurship education and training for colleges by Anhui Province (201810366116).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2019. 01544/full#supplementary-material

and oxido-nitrosative mediators. Pharmacol. Biochem. Behav. 149, 1–8. doi: 10.1016/j.pbb.2016.07.004


in a mouse model of Alzheimer's disease. J. Neurosci. Off. J. Soc Neurosci. 37, 637–647. doi: 10.1523/JNEUROSCI.2110-16.2016


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

Copyright © 2020 Fang, Li, Chen, Gao, Huang, Du, Jiang, Li and Ge. 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(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.

# Corrigendum: Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats via Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC

#### Approved by:

Frontiers Editorial Office, Frontiers Media SA, Switzerland

#### \*Correspondence:

Jin-Fang Ge gejinfang@ahmu.edu.cn † These authors have contributed equally to this work

#### Specialty section:

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

Received: 30 March 2020 Accepted: 02 April 2020 Published: 15 April 2020

#### Citation:

Fang K, Li H-R, Chen X-X, Gao X-R, Huang L-L, Du A-Q, Jiang C, Li H and Ge J-F (2020) Corrigendum: Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats via Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC. Front. Pharmacol. 11:518. doi: 10.3389/fphar.2020.00518 Ke Fang1,2,3† , Hua-Rong Li <sup>1</sup>† , Xing-Xing Chen1,2,3, Xin-Ran Gao1,2,3, Ling-Ling Huang<sup>1</sup> , An-Qi Du<sup>1</sup> , Chuan Jiang<sup>1</sup> , Hua Li 1,4 and Jin-Fang Ge1,2,3\*

<sup>1</sup> School of Pharmacy, Anhui Medical University, Hefei, China, <sup>2</sup> Anhui Province Key Laboratory of Major Autoimmune Diseases, Anhui Institute of Innovative Drugs, Hefei, China, <sup>3</sup> The Key Laboratory of Anti-inflammatory and Immune Medicine, Ministry of Education, Anhui Medical University, Hefei, China, <sup>4</sup> The First Clinical College, Anhui Medical University, Hefei, China

Keywords: quercetin, nesfatin-1, brain derived neurotrophic factor (BDNF), Copine 6, the triggering receptors expressed on myeloid cells (TREMs), synapsin-1

### A Corrigendum on

### Quercetin Alleviates LPS-Induced Depression-Like Behavior in Rats via Regulating BDNF-Related Imbalance of Copine 6 and TREM1/2 in the Hippocampus and PFC

by Fang K, Li H-R, Chen X-X, Gao X-R, Huang L-L, Du A-Q, Jiang C, Li H and Ge J-F (2020). Front. Pharmacol. 10:1544. doi: 10.3389/fphar.2019.01544

An author name was incorrectly spelled as "Fang Ke." The correct spelling is "Ke Fang".

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2020 Fang, Li, Chen, Gao, Huang, Du, Jiang, Li and Ge. 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.