# MODULATING GLIAL CELLS PHENOTYPE: NEW FINDINGS AND THERAPIES

EDITED BY : Alberto Javier Ramos, Yolanda Diz-Chaves and Maria Jose Bellini PUBLISHED IN : Frontiers in Aging Neuroscience, Frontiers in Molecular Neuroscience and Frontiers in Neuroscience

#### 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 978-2-88966-183-1 DOI 10.3389/978-2-88966-183-1

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

# MODULATING GLIAL CELLS PHENOTYPE: NEW FINDINGS AND THERAPIES

Topic Editors: Alberto Javier Ramos, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina Yolanda Diz-Chaves, University of Vigo, Spain Maria Jose Bellini, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

Citation: Ramos, A. J., Diz-Chaves, Y., Bellini, M. J., eds. (2020). Modulating Glial Cells Phenotype: New Findings and Therapies. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88966-183-1

# Table of Contents


María Velasco-Estevez, Myrthe Mampay, Hervé Boutin, Aisling Chaney, Peter Warn, Andrew Sharp, Ellie Burgess, Emad Moeendarbary, Kumlesh K. Dev and Graham K. Sheridan


Meng-meng Jin, Fen Wang, Di Qi, Wen-wen Liu, Chao Gu, Cheng-Jie Mao, Ya-Ping Yang, Zhong Zhao, Li-Fang Hu and Chun-Feng Liu

*54 Parkinsonian Neurotoxins Impair the Pro-inflammatory Response of Glial Cells*

Neus Rabaneda-Lombarte, Efren Xicoy-Espaulella, Joan Serratosa, Josep Saura and Carme Solà


#### *146 Microglial Ultrastructure in the Hippocampus of a Lipopolysaccharide-Induced Sickness Mouse Model* Julie C. Savage, Marie-Kim St-Pierre, Chin Wai Hui and Marie-Eve Tremblay

*155 Boosting Antioxidant Self-defenses by Grafting Astrocytes Rejuvenates the Aged Microenvironment and Mitigates Nigrostriatal Toxicity in Parkinsonian Brain* via *an* Nrf*2-Driven* Wnt*/*β*-Catenin Prosurvival Axis* Maria Francesca Serapide, Francesca L'Episcopo, Cataldo Tirolo, Nunzio Testa, Salvatore Caniglia, Carmela Giachino and Bianca Marchetti

# Editorial: Modulating Glial Cells Phenotype: New Findings and Therapies

#### María José Bellini <sup>1</sup> \*, Yolanda Diz-Chaves <sup>2</sup> and Alberto Javier Ramos <sup>1</sup>

<sup>1</sup> Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina, <sup>2</sup> Laboratory of Endocrinology, The Biomedical Research Centre (CINBIO), University of Vigo, Vigo, Spain

Keywords: aging, inflammation, microglia, astroglia, therapies

**Editorial on the Research Topic**

#### **Modulating Glial Cells Phenotype: New Findings and Therapies**

Aging and neurodegenerative diseases are closely related, and the common point between both scenarios is neuroinflammation. Neuroinflammation is mainly mediated by glial cells, astrocytes, and microglia; while endothelial cells transfer the pro-inflammatory signals from periphery. The mechanism that underlies glial cells activation and neuroinflammation-related damage is not fully understood, being the study of glial cells during aging as well as neurodegenerative diseases, a key point to identify targets and to develop therapies that modulate adverse outcomes and mitigate neurodegeneration.

This Research Topic has produced a highly informative collection of original research and reviews, that cover multiple aspects for delving neuroinflammation and glial phenotypic changes in neurological diseases and aging. Researchers have presented their work and views to explain cellular and molecular mechanisms acquired by glial cells as well as possible interventions that can modify their functions and phenotypes associated with a broad spectrum of neurological diseases.

Serapide et al. started by focusing on the pivotal role of astrocytes activation in Parkinson's disease (PD), the most common age-dependent movement disorder. These authors tried a new approach to uncover the interaction between the astrocytes and neurons in PD mice brains. They demonstrate that grafting astrocytes alleviate oxidative stress in a parkinsonian brain by using MPTP-induced PD mice. Adding new information, Rabaneda-Lombarte et al. investigated the effect of MPP+ or rotenone to microglial/astroglial functions. Their manuscript provides information on the impact of PD-related toxins on the phenotype of glia, based on the evaluation of the mRNA levels of various gene products.

#### Deepen on astrocytes functions, the astrocyte-specific enzyme glutamine synthetase plays an essential role in supporting neurotransmission and in limiting ammonium toxicity. Moreover, deficits in Glutamine synthetase activity contribute to epilepsy and neurodegeneration. In this regard, Huyghe et al. extensively investigated the regulation of glutamine synthetase by PKA at T301 using biochemical and cell biological approaches. The authors successfully created two antiphospho-GS antibodies and demonstrated that phosphorylation of T301 is critical for its activity in astrocytes. This modification results in decreased enzyme activity in a mouse model of epilepsy.

To provide new insights into microglial implication in behavioral changes induced by infections and mediated by pro-inflammatory cytokines, Savage et al. describe for the first time the effects of acute LPS on the ultrastructure of microglia in the mouse hippocampus. The ultrastructural changes found by the authors, provide an excellent starting point to determine pro-inflammatory and phagocytic pathways engaged by in sickness behavior.

Another interesting aspect of glial cells participation on pathological conditions was reviewed by Zang et al.. Their work was focused on the role of epigenetic regulation in the pathogenesis of

#### Edited and reviewed by:

Thomas Wisniewski, New York University, United States

#### \*Correspondence:

María José Bellini mariajosebellini@yahoo.com; mariajosebellini@med.unlp.edu.ar

Received: 14 August 2020 Accepted: 18 August 2020 Published: 23 September 2020

#### Citation:

Bellini MJ, Diz-Chaves Y and Ramos AJ (2020) Editorial: Modulating Glial Cells Phenotype: New Findings and Therapies. Front. Aging Neurosci. 12:594870. doi: 10.3389/fnagi.2020.594870 glioma and discuss its value as therapeutic target for the treatment of this disease. Authors expand our knowledge on four epigenetic phenomena that occur in glioma: DNA methylation, abnormal microRNA, histone modifications, and chromatin remodeling. They review current epigenetic treatments for glioma, such as inhibitors of DNA methyltransferase and histone deacetylation inhibitors and argument the limitations of their applications due to the complicated pathogenesis of glioma.

Various articles in this collection focus on the role and mechanisms of astrocytes and microglia in the aging brain. Current issues and novel findings in the modulation of glial cells and the pathways able to promote healthy aging will provide a better understanding of cellular and molecular mechanisms toward a neuroprotective or a deleterious glial phenotype. Related with this topic, Matias et al. introduced the concept of astrocyte heterogeneity, which is a hot topic in the field. The review addressed diverse phenotypes of astrocytes with an emphasis on aging and neurodegenerative conditions. The authors discussed historical overview of astrocytes, astrocyte heterogeneity in healthy brain and pathological conditions, concept of "inflammaging," and selective vulnerability. Also, they present the possibility of using glia as target for therapeutic strategies.

Expanding the knowledge of the changes undergone by astrocytes during aging, Zárate et al. investigated the role of Humanin, an astrocyte-release mitochondrial peptide, and the influence of hormonal milieu. They analyzed the role of this protein on synaptic dysfunction in the context of hormonal decline as a model of aging. The study of Humanin regulation by ovarian hormones in the brain is novel and highly interesting, given the plethora of molecular targets that these hormones have and that participate in mediating their neuroprotective and neuro-restorative.

Continuing with the study of glial-cells changes during aging, Falomir-Lockhart et al. investigated the effects of IGF-1 gene therapy on microglia. Importantly, these authors found that IGF1 therapy leads to a region-specific modification of aged microglia population. Microglia cells become dystrophic with aging; which contributes to basal central nervous system neuroinflammation being a risk factor for age-related neurodegenerative diseases. These authors' findings suggest that such a therapy may be useful to alter the function of dystrophic microglia effects.

Another study by Trias et al. explores the emergence of senescent microglia in spinal cord during paralysis progression in amyotrophic lateral sclerosis (ALS) rat model (SOD1G93A). The authors showed nuclear p16INK4a expression in activated microglia during paralysis progression as well as in a subpopulation of spinal motor neurons and astrocytes. They suggest that cellular senescence is closely associated with inflammation and motor neuron loss occurring after paralysis onset in SOD1G93A rats. The emergence of senescent cells could mediate key pathogenic mechanisms in ALS.

With the increment of life expectancy, preventive measures, and treatments for age-related cognitive decline, and dementia are of utmost importance. The work by Ano et al. shows that both chronic and acute administration of Iso-alpha-acids (IAA) prevents the neuroinflammation and cognitive decline associated with aging. The use of natural compounds as IAA, which are hop-derived bitter compounds in beer, represent an interesting therapeutic tool to ameliorate age-related negative outcomes.

Later on, Velasco-Estevez et al. used an aging transgenic rat model of Alzheimer Disease (AD) to study the expression of mechanosensing Piezo1 ion channels in amyloid plaque-reactive astrocytes. Piezo1 is a non-selective mechanosensory mediating cation channel. The authors found that aging and peripheral infection augment amyloid plaque-induced upregulation of mechanoresponsive ion channels, such as Piezo1, in astrocytes. Additional work is required to investigate the role of astrocytic Piezo1 in the Alzheimer's brain and its possible use as a novel drug target for age-related dementia.

Microglia polarization to an anti-inflammatory phenotype has been described to be neuroprotective. In the study by Jin et al., they linked microglial pro-inflammatory polarization to neuroinflammation and autophagy dysfunction, that are closely related to the development of neurodegeneration in Parkinson's disease. They demonstrate that TNF-α inhibits autophagy in microglia through AKT/mTOR signaling pathway, and autophagy enhancement can promote microglia polarization toward anti-inflammatory phenotype and inflammation resolution. This study adds information to establish which specific factors are responsible for disturbing the pro-/anti-inflammatory balance in areas where inflammatory stimuli induce neurodegeneration. Besides the study contributes to understanding the relationship between inflammation and autophagy processes.

Together these articles have brought several interesting understandings of neuroinflammation and glial cells activation in a range of neurological diseases and aging. We expect that this topic would expand our knowledge of the mechanisms whereby reactive glial cells could be detrimental or neuroprotective. Moreover, this comprehension could contribute for the development of therapeutics to limit neuroinflammation or even slow down the process of aging.

# AUTHOR CONTRIBUTIONS

MB, YD-C, and AR contributed equally to this editorial, in its conception, writing, and revision. All authors approved the submitted version.

**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 Bellini, Diz-Chaves and Ramos. 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.

# Infection Augments Expression of Mechanosensing Piezo1 Channels in Amyloid Plaque-Reactive Astrocytes

María Velasco-Estevez1,2, Myrthe Mampay<sup>1</sup> , Hervé Boutin<sup>3</sup> , Aisling Chaney3,4 , Peter Warn<sup>5</sup> , Andrew Sharp<sup>5</sup> , Ellie Burgess<sup>5</sup> , Emad Moeendarbary6,7 \*, Kumlesh K. Dev<sup>2</sup> \* and Graham K. Sheridan<sup>1</sup> \*

<sup>1</sup> Neuroimmulology & Neurotherapeutics Laboratory, School of Pharmacy and Biomolecular Sciences, University of Brighton, Brighton, United Kingdom, <sup>2</sup> Drug Development, Department of Physiology, School of Medicine, Trinity College Dublin, Dublin, Ireland, <sup>3</sup> Wolfson Molecular Imaging Centre, Faculty of Biology, Medicine and Health and Manchester Academic Health Sciences Centre, The University of Manchester, Manchester, United Kingdom, <sup>4</sup> Department of Radiology, Stanford University, Stanford, CA, United States, <sup>5</sup> Evotec (UK) Ltd., Manchester Science Park, Manchester, United Kingdom, <sup>6</sup> Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, United States, <sup>7</sup> Department of Mechanical Engineering, University College London, London, United Kingdom

#### Edited by:

Alberto Javier Ramos, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Jingwen Niu, Temple University, United States Vladimir Parpura, The University of Alabama at Birmingham, United States

#### \*Correspondence:

Emad Moeendarbary e.moeendarbary@ucl.ac.uk Kumlesh K. Dev devk@tcd.ie Graham K. Sheridan G.Sheridan@brighton.ac.uk

Received: 19 June 2018 Accepted: 01 October 2018 Published: 22 October 2018

#### Citation:

Velasco-Estevez M, Mampay M, Boutin H, Chaney A, Warn P, Sharp A, Burgess E, Moeendarbary E, Dev KK and Sheridan GK (2018) Infection Augments Expression of Mechanosensing Piezo1 Channels in Amyloid Plaque-Reactive Astrocytes. Front. Aging Neurosci. 10:332. doi: 10.3389/fnagi.2018.00332 A defining pathophysiological hallmark of Alzheimer's disease (AD) is the amyloid plaque; an extracellular deposit of aggregated fibrillar Aβ1−<sup>42</sup> peptides. Amyloid plaques are hard, brittle structures scattered throughout the hippocampus and cerebral cortex and are thought to cause hyperphosphorylation of tau, neurofibrillary tangles, and progressive neurodegeneration. Reactive astrocytes and microglia envelop the exterior of amyloid plaques and infiltrate their inner core. Glia are highly mechanosensitive cells and can almost certainly sense the mismatch between the normally soft mechanical environment of the brain and very stiff amyloid plaques via mechanosensing ion channels. Piezo1, a non-selective cation channel, can translate extracellular mechanical forces to intracellular molecular signaling cascades through a process known as mechanotransduction. Here, we utilized an aging transgenic rat model of AD (TgF344- AD) to study expression of mechanosensing Piezo1 ion channels in amyloid plaquereactive astrocytes. We found that Piezo1 is upregulated with age in the hippocampus and cortex of 18-month old wild-type rats. However, more striking increases in Piezo1 were measured in the hippocampus of TgF344-AD rats compared to age-matched wild-type controls. Interestingly, repeated urinary tract infections with Escherichia coli bacteria, a common comorbidity in elderly people with dementia, caused further elevations in Piezo1 channel expression in the hippocampus and cortex of TgF344- AD rats. Taken together, we report that aging and peripheral infection augment amyloid plaque-induced upregulation of mechanoresponsive ion channels, such as Piezo1, in astrocytes. Further research is required to investigate the role of astrocytic Piezo1 in the Alzheimer's brain, whether modulating channel opening will protect or exacerbate the disease state, and most importantly, if Piezo1 could prove to be a novel drug target for age-related dementia.

Keywords: Alzheimer's disease, amyloid plaques, astrocytes, dentate gyrus, mechanosensitive ion channel, Piezo1, TgF344-AD rats, urinary tract infection

### INTRODUCTION

fnagi-10-00332 October 17, 2018 Time: 13:57 # 2

The etiology of early-onset Alzheimer's disease (AD) has a strong genetic component and mutations in genes that encode for amyloid precursor protein (APP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2) give rise to a rare but severe form of familial dementia characterized by cortical and hippocampal atrophy and progressive cognitive decline (Saunders, 2001; Nacmias et al., 2014; Lanoiselee et al., 2017). The amyloid cascade hypothesis of AD (Hardy and Allsop, 1991) states that aberrant cleavage of APP by γ-secretase results in the accumulation of cytotoxic oligomeric amyloid-β1−<sup>42</sup> (Aβ1−42) which aggregates and forms β-sheet fibrillar structures that, in turn, clump together and generate extracellular amyloid plaques (Lukiw, 2012; Morris et al., 2014). Amyloid plaques contribute to cortical and hippocampal neurodegeneration and neuroinflammation (Nobakht et al., 2011; Dansokho and Heneka, 2018), although extensive plaque pathology is not necessarily a good indicator of the clinical severity of dementia (Hyman et al., 1993). Structurally, amyloid plaques are very brittle and stiff (with a stiffness of ∼3 × 10<sup>9</sup> Pa) (Smith et al., 2006), in contrast to brain tissue which we have previously shown to be very soft (with a stiffness of ∼200– 500 Pa) (Moeendarbary et al., 2017). Therefore, neurons and glia surrounding the amyloid plaque experience a large perturbation to their mechanical microenvironment. Astrocytes, the most abundant glial cell type in the brain, respond to amyloid plaque accumulation by adopting a 'reactive' phenotype characterized by an increase in intermediate filament expression, e.g., GFAP (Wilhelmsson et al., 2006; Sofroniew, 2009), and synthesis of proinflammatory cytokines (Buffo et al., 2010; Sofroniew, 2014). We hypothesize that, in addition to biochemical signals, astrogliosis in the AD brain is triggered in part by a detectable increase in the stiffness of the mechanical microenvironment experienced by resting glial cells.

Neurons and glia are highly mechanosensitive (Previtera et al., 2010; Blumenthal et al., 2014) and detect changes in the stiffness of underlying cell culture substrates via stretchactivated membrane-spanning ion channels (Cox et al., 2016; Bavi et al., 2017). Microglial cells, for example, display durotaxis, i.e., they migrate toward stiffer areas when cultured on stiffness gradient hydrogels (Bollmann et al., 2015). Astrocytes grown on stiff hydrogels possess longer, thicker processes and appear much more reactive than those cultured on soft hydrogels (Moshayedi et al., 2010). Moreover, astrocytes in vivo respond to hard foreign body implants by releasing proinflammatory cytokines, such as Il-1β (Moshayedi et al., 2014). In AD, the build-up of extracellular amyloid plaques causes astrogliosis and, given the stiffness of plaques, we hypothesize that astrocyte reactivity is mediated partly by activation of mechanosensitive ion channels. Therefore, our aim here was to investigate cellular expression of mechanosensitive Piezo1 channels in response to amyloid plaque pathology in an aging rat model of AD.

To carry out this aim, we utilized the TgF344-AD rat model (Cohen et al., 2013) that overexpress human APP with the Swedish mutation (APPSWE) and mutated human PSEN1 (PSEN11E9). In the healthy human brain, Piezo1 mRNA is expressed by neurons but is absent from astrocytes. Post-mortem analysis of AD brains, however, revealed that Piezo1 mRNA is down-regulated in neurons and up-regulated in astrocytes surrounding amyloid plaques (Satoh et al., 2006). Therefore, Piezo1 was previously known as Mib (membrane protein induced by beta-amyloid treatment). Piezo1 is a large protein (>2,500 amino acids) predicted to contain 30–40 putative transmembrane segments (Kamajaya et al., 2014), assembles as a ∼900 kDa homotrimer (Ge et al., 2015) and forms a non-selective calcium (Ca2+)-permeable cation channel (Gnanasambandam et al., 2015). Mechanosensitive ion channels translate external physical stimuli, such as shear stress (Ranade et al., 2014), pressure (Gottlieb and Sachs, 2012; Bae et al., 2015) and extracellular matrix (ECM)-associated traction forces (Pathak et al., 2014) to intracellular biochemical signals via mechanotransduction events (Martinac, 2004; Moeendarbary and Harris, 2014). Piezo1 mRNA is highly expressed in tissues subject to regular deformations or shear forces such as lungs, bladder, skin and vasculature endothelium (Coste et al., 2010; Ranade et al., 2014; Li et al., 2015). To date, however, very little is known about the role of Piezo1 in the CNS. We have recently shown that Piezo1 is involved in axon guidance of optic tract retinal ganglion cells (RGCs) in the developing Xenopus laevis brain (Koser et al., 2016). Here, we will investigate if glial cells can physically sense and respond to the accumulation of amyloid plaques by quantifying expression of astrocytic Piezo1 channels in the hippocampus and cerebral cortex of the aging TgF344-AD rat model of AD.

Because of their advanced age, elderly people are prone to co-morbidities and peripheral infections that cause low-grade systemic inflammation (Salles and Mégraud, 2007; Doulberis et al., 2018). Interestingly, peripheral infections increase older people's risk of developing AD (Holmes et al., 2009; Tate et al., 2014) and reportedly exacerbate amyloid plaque deposition in the mouse brain (McManus et al., 2014), although the mechanisms underlying this crosstalk between the periphery and CNS is not well-understood. Although most transgenic animal models recapitulate many of the amyloid-related hallmarks of AD, it is important to use clinically-relevant animal models with at least one comorbidity, such as a peripheral infection, to investigate AD pathogenesis. Therefore, we investigated if a peripheral urinary tract infection (UTI) with E. coli bacteria, one of the most likely causes of infection in elderly people with dementia (Natalwala et al., 2008; Rowe and Juthani-Mehta, 2013), modifies astrocytic mechano-responsiveness to stiff extracellular amyloid deposits (Leyns and Holtzman, 2017). Our results suggest that restoring homeostatic mechanotransduction pathways and neuron/glial crosstalk around amyloid plaques may lead to the discovery of novel drug targets for neurodegeneration and AD.

**Abbreviations:** Aβ42, amyloid-β 1–42 peptide; BSA, bovine serum albumin; CMM, conditioned microglial media; CNS, central nervous system; DMEM, Dulbecco's modified Eagle's medium; FBS, fetal bovine serum; GFAP, glial fibrillary acidic protein; HBSS, Hank's buffered salt solution; HRP, horse radish peroxidase; iTG, transgenic with infection; iWT, wild-type with infection; NaCl, sodium chloride; NFH, neurofilament H; PBS, phosphate-buffered saline; PVDF, polyvinylidene difluoride membrane; SDS, sodium dodecylsulfate; TG, transgenic; TNF-α, tumor necrosis factor-alpha; WT, wild-type.

# MATERIALS AND METHODS

fnagi-10-00332 October 17, 2018 Time: 13:57 # 3

#### Ethics Statement

All experiments involving animals and Schedule 1 protocols used to obtain brain tissue were approved by the Animal Welfare and Ethical Review Bodies (AWERB committees) of the University of Brighton and The University of Manchester, as well as the Animal Research Ethics Committee (AREC) in Trinity College Dublin. This study was conducted in accordance with the principles of the Basel Declaration and adhered to the legislation detailed in the United Kingdom Animals (Scientific Procedures) Act 1986 Amendment Regulations (SI 2012/3039). All efforts were taken to maximize animal welfare conditions and to reduce the number of animals used in accordance with the European Communities Council Directive of September 20th, 2010 (2010/63/EU).

#### Mouse Neuron Cultures

Primary neuron cultures were prepared from the cerebral cortical tissue of male and female postnatal day 0 or 1 (P0 or P1) C57BL/6 mice bred in the BioResources Unit (BRU) of University of Brighton. Briefly, cortices were dissected following decapitation and collected in Ca2+- and Mg2+-free Hank's balanced salt solution (HBSS) (Thermo Fisher Scientific, 14185052). Cells were dissociated using Papain [20 U/mL] (Worthington, 3119) in Ca2+- and Mg-free HBSS supplemented with <sup>L</sup>-cysteine [1 mM] (Sigma, 168149), EDTA [0.5 mM] (Thermo Fisher Scientific, 15575020) and DNAse I solution [0.125 U/mL] (Thermo Fisher Scientific, 90083). Dissociation was inhibited using Type II-O ovomucoid trypsin inhibitor [0.1%] (Sigma, T9253) in HBSS with bovine serum albumin [0.1%] (Sigma, A9647) and D-APV [D-(-)-2-Amino-5-phosphonopentanoic acid] [50 µM] (Abcam, ab120003). A single-cell suspension was created by gentle manual trituration using fire-polished Pasteur pipettes. Cells were plated on Poly-D-Lysine (Millipore, A-003-E) coated coverslips in 24-well plates [6 × 10<sup>4</sup> cells/well] with serumfree Neurobasal-A medium (Life Technologies, 10888-022) supplemented with 2% B27 (Thermo Fisher Scientific, 17504044) and L-glutamic acid [25 µM] (Sigma, G1251). Primary neuron cultures were maintained at 37◦C and 5% CO<sup>2</sup> in a humidified incubator and after 1 day in vitro, half of the plating medium was replaced with maintenance medium (without L-glutamic acid) composed of Neurobasal-A medium supplemented with 2% B27, 2 mM GlutaMAXTM (Gibco, 35050061), and 1% penicillin/streptomycin (Thermo Fisher Scientific, 15140122). This was repeated every 3 days, i.e., 50% of the cell culture medium was replenished with fresh maintenance medium. After 14 days in culture, cells were fixed in ice-cold formalin solution (10%) for immunofluorescent labeling.

#### Mouse Astrocyte Cultures

Mixed glial cell cultures were prepared from the cerebral cortical tissue of male and female P1 wild-type C57BL/6 mice bred in the BioResources Units of Trinity College Dublin and University of Brighton, following previously described protocols (O'Sullivan et al., 2017, 2018). Briefly, following decapitation the cortices were dissected with sterile forceps and placed in HBSS (Gibco, 14025-050). Cortical tissue was cross-chopped with a sterile scalpel and then placed in prewarmed Dulbecco's modified Eagle's medium/F12 (DMEM/F12, Hyclone, SH30023) supplemented with 10% FBS (Labtech; FB-1090) and 1% penicillin/streptomycin (Sigma, P4333) and gently triturated until a clear solution was obtained. The resulting solution was passed through a cell strainer (40 µm) and plated into separate T75 flasks. Cells were cultured at 37◦C and 5% CO<sup>2</sup> in a humidified incubator and grown in the supplemented DMEM/F12 media previously described. After 12 days in culture, microglial cells were separated from mouse astrocytes by placing them on a rotating shaker for 2 h at 37◦C. The supernatant, with detached microglia, was removed and the enriched astrocyte cultures were trypsinised with 0.5% trypsin for 7 minutes (min) and plated accordingly. Mouse astrocyte purity was >97%, as previously described (Healy et al., 2013).

# Aβ1−<sup>42</sup> Conditioned Microglial Media

Microglial cultures were prepared from the cerebral cortical tissue of male and female P1 wild-type C57BL/6 mice bred in the BioResources Unit of Trinity College Dublin, as previously described (Dempsey et al., 2017). Briefly, after 48 h in culture, microglia were treated with 10 µM amyloid-β1−<sup>42</sup> peptide (Invitrogen, United Kingdom) for 24 h. Aβ1−<sup>42</sup> conditioned microglial media (CMM) was collected and used to treat the enriched astrocyte cultures described above.

#### Immunocytochemistry

After each biochemical treatment, mouse astrocytes were fixed in 10% formalin solution (Sigma, F1635) for 5 min on ice. Cells were blocked overnight at 4◦C in blocking solution, i.e., 1% BSA + 0.1% Triton X-100. Primary antibodies included chicken anti-neurofilament H (NFH; Millipore, AB5539, RRID: AB\_177520) 1:1000 dilution, rabbit anti-GFAP (Abcam, ab7260, RRID: AB\_305808) 1:1000 dilution, goat anti-PIEZO1 (N-15; Santa Cruz, sc-164319, RRID: AB\_10842990) 1:500 dilution, and rabbit anti-FAM38A (Abcam, ab128245, RRID: AB\_11143245) 1:500 dilution in blocking solution, were used and incubated overnight at 4◦C. Cells were washed with PBS + 0.1% Triton X-100 for 5 min, three times. Incubations with secondary antibodies were performed in the dark at 22 ± 2 ◦C for an hour. In the case of rabbit anti-FAM38A, it was also incubated with biotinylated goat anti-rabbit antibody (Thermo Fisher, A24535, RRID: AB\_2536003) and avidin Alexa 488 conjugate (Life Technologies, A21370) to amplify the fluorescent signal. After washes, cells were coverslipped and mounted on a microscope slide (Clarity, C361) with antifade reagent (Thermo Fisher S36936). Imaging was performed using a Leica SP8 confocal microscope.

#### Western Blotting

Mouse astrocyte samples were obtained by scraping the cells in radioimmunoprecipitation assay buffer (RIPA) containing 150 mM NaCl (Sigma, A3014), 1% Triton-X (Sigma, T8787), 0.1% SDS (Fisher, S/5200/53) and 50 mM Tris pH 8.0 (Fisher, T/3710/60). Samples were sonicated three times for 10 s at 20% amplitude using a vibracell VCX 130 (Sonics,

United States). Briefly, samples were mixed in 1:1 dilution with Laemmli sample buffer 2× (BioRad, 161-0737) and boiled at 95◦C for 5 min. They were run in a 6% polyacrylamide separating gel with 4% stacking gel (Applichem Panreac, A1672) in running buffer at constant voltage 120 V, followed by wet transfer to a PVDF (Millipore, IPVH00010) for 3 h at constant 70 V on ice. Blocking of the membrane was done with 5% BSA (Santa Cruz, sc-2323) in 0.05% PBS/Tween 20 for an hour at 22 ± 2 ◦C. Incubation with primary antibody rabbit anti-FAM38A (Abcam, ab128245, RRID: AB\_11143245) or goat anti-PIEZO1 (Santa Cruz, sc-164319, RRID: AB\_10842990) was performed at a 1:500 dilution for 48 h at 4◦C. The membrane was washed and incubated with secondary antibody HRP-conjugated goat anti-rabbit (GE Healthcare, NA934, RRID: AB\_772206) or HRP-conjugated rabbit anti-goat (Sigma, A4174, RRID: AB\_258138), respectively, at 1:5000 dilution for 2 h at 22◦C. Following further washes, membranes were developed using chemiluminescent HRP substrate (Millipore, WBKLS0500) in a Fujifilm LAS-3000 (Fujifilm, Japan).

#### Wild-Type Rats

Female Lister hooded rats (Harlan Laboratories, United Kingdom) were socially housed in groups of four and given ad libitum access to food and water. The holding room was kept on a 12 h light/dark cycle at 22 ± 2 ◦C. Postnatal day 37 (5-weeks) rats were sacrificed by pentobarbital overdose (intraperitoneal injection) and death was confirmed by decapitation. Immunohistochemical experiments were conducted using post-mortem brain tissue. All experiments on animal tissue were performed in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986 Amendment Regulations (SI 2012/3039).

#### TgF344-AD Rats

Two male and female WT Fisher and TgF344-AD (TG) rats with the APPSWE and PSEN11E9 mutations were purchased from Prof T. Town laboratory (University of Southern California, United States) and were set-up as breeding pairs in-house at the Biological Services Unit (BSU) in The University of Manchester, United Kingdom. TG rats and WT littermates were randomly split into four groups, i.e., WT and TG non-infected (WT/TG) and infected (iWT/iTG), bearing in mind that the UTI needed to be performed on the entire box of rats. The number of rats per group (n) were as follows; 12m WT (n = 5); 18m WT (n = 5); 12m TG (n = 6); 18m TG (n = 6); 12m iWT (n = 5); 18m iWT (n = 5); 12m iTG (n = 6); 18m iTG (n = 6). All animals were housed in groups of 2–4 in cages with individual ventilation, environmental enrichment and constant access to food and water. A 12 h light/dark cycle was used, with light from 7 am until 7 pm. All experiments were carried out following internal ethical review and approval by the AWERB committee of The University of Manchester.

#### Urinary Tract Infection

A UTI was administered to the iWT and iTG groups at 7–8, 10–11, 13–14, and 16–17 months of age using Escherichia coli (E. coli) UTI89. For infections at 7–8 and 10–11 months of age, a concentration of 1.3 × 10<sup>9</sup> colony forming units/mL (cfu/mL) was used, and for infections at 13–14 and 16–17 months of age, a concentration of 2.6 × 10<sup>9</sup> cfu/mL was used. Therefore, before sacrifice, 12-month rats received two rounds of infection and 18 month rats received four rounds. Prior to infection, stocks of E. coli were stored as 1.02 × 10<sup>11</sup> cfu/mL in 15% glycerol and were rapidly thawed and brought to the correct concentration using sterile PBS. Rats were anesthetized and infected with 0.1 mL of E. coli UTI89 by inserting a catheter (polyethylene tubing covering 30-G hypodermic needles) into the urethra and injecting the inoculum into the bladder. Prior to infection, buprenorphine (0.03 mg/kg) was administered and the animals' bladders were emptied by gently pressing on the abdomen to release any bladder content. Infected animals were placed into clean cages and were monitored for 5 days for signs of illness. Urine samples were taken 2 and 5 days post-infection to confirm infection levels. Urine was plated on CLED agar plates and incubated at 37◦C for 18–24 h after which colony forming unit counts were performed and scored according to the following classification system: 0 = Zero colonies; 0.5 = Scanty growth (<5 cfu); 2 = Scanty growth (<20 cfu); 3 = moderate growth (>20 cfu); 4 = confluent growth. Non-infected rats tested for infection all scored ≤ 0.5.

#### Tissue Sectioning

Coronal slices of Lister hooded rat brains (3–4 mm thick) were fixed in 10% formalin solution for 8 h at 4◦C and then transferred to 30% sucrose solution overnight (4◦C) for cryoprotection. Brain slices were then covered in OCT (Optimal Cutting Temperature compound) and frozen and stored at −80◦C. The slices were cryosectioned (14 µm thick) coronally at approximately −3.0 and −10.8 mm with respect to Bregma (Paxinos and Watson, 2014) in order to reveal the dorsal hippocampus and cerebellum regions, respectively. Cryosections were adhered to glass slides coated with poly-L-lysine solution [0.1% (w/v) in H2O, Sigma]. Sections were permeabilized using a solution of 0.1% TritonTM X-100 (Sigma) in PBS for 25 min.

TgF344-AD and WT rats were sacrificed using an isoflurane overdose, confirmed by cervical dislocation. The brains were collected, snap frozen using isopentane on dry ice and stored at −80◦C. Sagittal brain sections (20 µm thick) between 1 and 3.36 mm lateral to Bregma were taken using a cryostat (Leica CM3050s, Leica Biosystems Nussloch GmbH, Germany) and stored at −80◦C. Prior to immunofluorescence, frozen sections were allowed to air dry at 22 ± 2 ◦C for 20 min before fixation in 70% ethanol for 30 min. They were permeabilized in 0.2% Triton-X/PBS for 30 min, followed by several PBS washes, and then blocked with 5% BSA/PBS for 1.5 h at 22 ± 2 ◦C.

#### Immunohistochemistry

Sections from WT Fisher and TgF344-AD rats were incubated overnight at 22 ± 2 ◦C with the following primary antibodies, (1) 1:500 dilution of goat anti-PIEZO1 (N-15; Santa Cruz, sc-164319, RRID: AB\_10842990), an affinity-purified goat polyclonal antibody raised against a peptide mapping to the N-terminus of PIEZO1 of human origin; (2) 1:1000 dilution of rabbit anti-amyloid β1−<sup>42</sup> (mOC98, Abcam, ab201061), a

conformation-specific antibody that recognizes a discontinuous epitope of Aβ that maps to segments AEFRHD and EDVGSNK; (3) 1:1000 dilution of chicken anti-GFAP (Abcam, ab4674, RRID: AB\_304558). Slides were then washed in PBS and incubated for 4 h at 22◦C with the following secondary antibodies, (1) donkey anti-goat Alexa 488 (Abcam, ab150133), (2) donkey anti-rabbit 555 (Sigma, SAB4600061), (3) donkey anti-chicken IgY (H + L) CFTM 633 (Sigma, SAB4600127). Finally, slides were washed and coverslipped using ProLong <sup>R</sup> Gold antifade mounting medium with DAPI (Life Technologies, United Kingdom, P36935).

#### Microscopy and Image Analysis

All images used to quantify Piezo1 expression in sagittal cryosections were captured using an Axio Scan.Z1 slide scanner (Zeiss, Germany) with a 20× magnification objective (Plan-Apochromat). Individual images were montaged together automatically in ZEN software (Zeiss, Germany) to reconstruct a single image of the whole brain section. For quantitative fluorescence intensity analysis, all images were captured in a single uninterrupted run (∼50 h imaging time) and uniform microscope settings were maintained throughout the session. The images were exported as 8-bit tif files for Piezo1 fluorescence intensity quantification. Image analysis was conducted using the software package FIJI<sup>1</sup> . Briefly, Piezo1 expression was quantified by manually selecting 10 regions of interest (ROI) for each brain structure and calculating the average fluorescence intensity within each ROI. Two or three brain sections from 5 or 6 different animals per age group/genotype were analyzed (n = 5–6). More specifically, the number of rats per group were as follows; 12m WT (n = 5); 18m WT (n = 5); 12m TG (n = 6); 18m TG (n = 6); 12m iWT (n = 5); 18m iWT (n = 5); 12m iTG (n = 6); 18m iTG (n = 6). Therefore, 100–180 fluorescence intensity (ROI) measurements were captured depending on the age and genotype of the rat. Piezo1 fluorescence intensities were normalized to the 12m WT value for each region and the percentage (%) change relative to the 12m WT group is represented in each bar graph. To quantify Piezo1 expression in amyloid plaquereactive astrocytes versus astrocytes located at least 200 µm away from any noticeable amyloid deposits, 25 plaques from the frontal cortex of 18m TG rat brains were randomly selected for analysis. For each image (25 in total) Piezo1 fluorescence intensity was quantified in six astrocytes within the plaque core and in six astrocytes ∼200–500 µm away from the plaque. The average Piezo1 fluorescence intensity values for astrocytes located "outside" the plaque versus "inside" the plaque were calculated for each image, i.e., 150 astrocytes per group were analyzed.

#### Statistical Analysis

All statistical analysis was performed using GraphPad <sup>R</sup> Prism 7 (RRID: SCR\_015807). Assessment of the normality of data was carried out using column statistics with D'Agostino analysis before any other statistical test was performed. For Western blot assays, repeated measures analysis of variance (ANOVA) tests were performed because data in every experiment was matched. Holm-Sidak multiple comparisons post hoc tests were

<sup>1</sup>http://fiji.sc/Fiji

run in conjunction with one-way ANOVAs and all groups were compared with one another. Data are presented as the mean ± standard error of the mean (SEM). To analyze changes in Piezo1 fluorescence intensities in vivo, where three independent variables (i.e., age, genotype, and infection) were present, a threeway ANOVA was performed on the raw fluorescence values, followed by a Holm-Sidak multiple comparisons post hoc test. Changes in Piezo1 expression in vivo were only considered significant when they passed two criteria, i.e., the relative percentage (%) change was > 20% and the p-value < 0.01. To analyze differences in average astrocytic Piezo1 expression outside versus inside amyloid plaques in the frontal cortex of 18m TG rat brains, a two-tailed paired t-test was performed. Finally, to correlate Piezo1 expression in the DG with that of GFAP and Aβ1−<sup>42</sup> fluorescence intensities, Pearson r-values were calculated using the raw fluorescence intensity values. A correlation was only deemed significant if r > 0.5 and p < 0.0001.

# RESULTS

#### Mechanosensitive Piezo1 Channel Expression in the Juvenile and Adult Rat Brain

To characterize Piezo1 channel expression in distinct regions of the young rat brain, coronal sections were cut at −3.0 and −10.8 mm with respect to Bregma revealing hippocampal (**Figure 1A**) and cerebellar (**Figure 1B**) structures. In agreement with our previous findings in the brain of embryonic Xenopus laevis (Koser et al., 2016), the juvenile rat optic tract expresses relatively high levels of Piezo1 channels (**Figure 1A**, red arrow). Moreover, axonal tracts of the corpus callosum, hippocampal fimbriae, and arbor vitae of the cerebellum (**Figure 1B**, white arrow) express high levels of Piezo1 channels. In contrast, the molecular and granule layers of the cerebellum, and non-neuronal areas of the hippocampus, express much fewer Piezo1 channels (**Figures 1A,B**). These regional differences were confirmed by Western blot (WB) using two distinct antibodies, i.e., one which binds the N-terminus of Piezo1 (Santa Cruz; **Figures 1C,D**) and one that binds the C-terminus (Abcam; **Figures 1E,F**). For WB, the brains of juvenile rats were dissected into four areas, i.e., cerebellum (Crb), hippocampus (Hip), cerebral cortex (Ctx), and brain stem/thalamus (BS/T) combined. There were no statistically significant differences in Piezo1 expression between brain structures (repeated measures oneway ANOVA, p > 0.05). However, the hippocampus consistently displayed lower levels of Piezo1, both with immunofluorescence and WB techniques.

Next, sagittal sections from adult (12 months) rat brains were imaged to characterize Piezo1 localisation in several other structures along the anterior–posterior axis (**Figures 2A–D**). Similar to juvenile rats, Piezo1 was highly expressed along axonal tracts of the corpus callosum and cerebellar arbor vitae as well as on Purkinje cell bodies (**Figure 2J**), medulla, pons, thalamus (**Figure 2M**) as well as on Purkinje cell bodies (**Figure 2J**), medulla, pons, thalamus (**Figure 2M**), striatum (**Figure 2N**) and

the rostral migratory stream which links the subventricular zone to the olfactory bulb (**Figure 2B**, blue arrow). Areas of low Piezo1 expression included the dorsal midbrain (**Figure 2K**) and nonneuronal regions of the hippocampus. Pyramidal and granule neurons of the CA1 (**Figure 2E**), CA3 (**Figure 2F**), and dentate gyrus (**Figure 2G**), however, expressed higher levels of Piezo1. Interestingly, layer V cortical cells (**Figure 2H**) expressed more Piezo1 channels than prefrontal cortical cells (**Figure 2I**). It was also noted that epithelial cells of the choroid plexus within the lateral ventricle (**Figure 2L**) express relatively high levels of Piezo1 protein.

# Reactive Astrocytes Surrounding Amyloid Plaques Upregulate Piezo1

The pattern of expression in the rat brain suggested that Piezo1 is predominantly localized to neurons, in agreement with human studies of mRNA (Satoh et al., 2006). To examine if Piezo1 protein expression in mouse brain in situ matched that of the rat brain at the same age, we sectioned and immunofluorescentlylabeled juvenile (5 weeks) mouse brains for Piezo1 and observed a strikingly similar expression pattern to the young rat brain; a result that matches the highly conserved amino acid sequence between mouse and rat Piezo1<sup>2</sup> . It was noted that Piezo1 expression in mouse brain also appeared predominantly neuronal (data not shown). To confirm cellular localisation, primary neuron cultures and enriched astrocyte cultures (∼97% pure) were prepared from newborn C57BL/6 mice and double-stained for either Piezo1 and GFAP (**Figures 3A–D**) or Piezo1 and neurofilament H (NFH; **Figures 3E–H**). Piezo1 co-localized with neurofilament H in mouse cortical neurons but was almost completely absent from primary astrocyte cultures under basal conditions.

To investigate if neuroinflammatory cell stressors are capable of upregulating Piezo1 channels in astrocytes, we exposed astrocytes to 5 µM hydrogen peroxide (H2O2) or to a cytokine cocktail consisting of TNF-α (10 ng/mL) and IL17A (250 ng/mL) for 24 h. Neither stressor induced Piezo1 expression in astrocytes (**Figures 3M,N**). Satoh et al. (2006), however, have reported previously that Piezo1 mRNA (then known as membrane protein induced by beta amyloid, Mib), is upregulated in astrocytes surrounding amyloid plaques in human post-mortem brain tissue. To examine if mouse astrocytes in culture can also be forced to express Piezo1 under similar pathophysiological conditions, we first treated mouse microglial cells with 10 µM amyloid beta-42 (Aβ1−42; Invitrogen, United Kingdom) for 24 h and collected the inflammatory cell culture media. Mouse astrocytes were then incubated with Aβ1−<sup>42</sup> CMM for 48 h and stained for Piezo1 channels (**Figures 3I–L**). Aβ1−<sup>42</sup> CMM induced Piezo1 expression in 30–40% of cortical astrocytes (**Figures 3I,M,N**).

Next, we investigated if upregulation of Piezo1 channels is also evident in vivo in an aging rat model of Alzheimer's disease pathology (TgF344-AD). To achieve this, the brains of 18-month old TgF344-AD (18m TG) rats were sectioned in the sagittal plane and triple-immunolabeled for Piezo1, conformationspecific Aβ1−<sup>42</sup> [mOC98], and GFAP. High magnification (100×) z-stack projections clearly demonstrate an upregulation of Piezo1 in reactive astrocytes surrounding prefrontal cortical

<sup>2</sup>http://www.uniprot.org/

FIGURE 2 | Piezo1 channel expression in the adult rat brain. (A) The brains of adult (12 month) Fisher rats were sectioned in the sagittal plane (lateral –1.9 mm) and stained for (B) Piezo1 (N-terminal antibody), (C) DAPI, and (D) GFAP (n = 5). Images were captured using a 20× magnification objective on an Axio Scan.Z1 slide scanner (Zeiss, Germany) and montaged in ZEN software. 20× magnification images of the hippocampal (E) CA1 (scale bar = 150 µm), (F) CA3 and (G) dentate gyrus (DG) show that Piezo1 expression is greater in pyramidal and granule neurons compared to GFAP-positive astrocytes. In the cerebral cortex, (H) layer V pyramidal neurons (LV Ctx) appear to express more Piezo1 channels than (I) prefrontal cortical (PFC) neurons. In the cerebellum (J), Purkinje neuron cell bodies and axons (arbor vitae) express higher levels of Piezo1 compared to neurons in the granule and molecular cell layers. (K) The dorsal midbrain (DM) expresses moderate levels of Piezo1, whereas (L) epithelial cells of the choroid plexus (CP) express greater amounts of Piezo1 protein. It was also noted that white matter tracts in (M) the thalamus (Th) and (N) striatum (Str) express more Piezo1 than gray matter areas of these brain regions. The areas depicted in images (E–N) are shown as yellow boxes in (C). Relative fluorescence intensity (arbitrary fluorescence units, a.f.u.) of Piezo1 staining in each brain region was quantified (O). Axonal tracts including the cerebellar arbor vitae, hippocampal fimbriae, cingulum and corpus callosum, as well as the medulla and pons of the brainstem, had the highest expression of Piezo1 protein.

FIGURE 3 | Aβ1−<sup>42</sup> CMM upregulates Piezo1 expression in cortical astrocyte cultures. To check cellular localisation of Piezo1 in the brain, enriched astrocyte cultures (A–D) and cortical neuronal cultures (E–H) from newborn mice were stained for (A,E) Piezo1 (N-terminal antibody) and (C) GFAP or (G) neurofilament H (NFH), respectively (n = 3). Piezo1 co-localized with NFH staining in neurons (H) but was almost completely absent from astrocytes under basal conditions (D). To investigate if neuroinflammatory stimuli can induce Piezo1 expression in astrocytes, mouse microglia were exposed to human Aβ1−<sup>42</sup> (Invitrogen, United Kingdom) for 24 h and the inflammatory media was collected. Enriched mouse astrocyte cultures were then exposed to this Aβ1−<sup>42</sup> CMM for 48 h. Astrocytes were immunofluorescently stained for (I) Piezo1 (C-terminus antibody), (J) DAPI, (K) GFAP, and (L) is the merge of each channel. Scale bar = 20 µm. Exposure to Aβ1−<sup>42</sup> CMM upregulated Piezo1 protein expression in approximately 30–40% of astrocytes (M,N). Astrocytes were also treated for 24 h with either 5 µM hydrogen peroxide (H2O2) or a cytokine cocktail consisting of TNF-α (10 ng/mL) and IL17A (250 ng/mL). Neither H2O<sup>2</sup> nor TNF-α/IL17A increased Piezo1 expression indicating that Piezo1 upregulation is not a common response to all cell stressors but specific to certain inflammatory stimuli including Aβ1−<sup>42</sup> and factors released by reactive amyloid-stimulated microglia. Data are represented as mean ± SEM (n = 8). A repeated measures one-way ANOVA with Holm-Sidak post hoc test was performed. <sup>∗</sup>Represents a statistically significant difference (p < 0.05) from control (Ctrl), TNF-α/IL17A, and H2O<sup>2</sup> treatment groups.

amyloid plaques (**Figures 4A–E**). On average, there was a threefold upregulation in Piezo1 expression in amyloid plaquereactive astrocytes compared to astrocytes located at least 200 µm away from any plaque in the prefrontal cortex (**Figures 4F,G**). One must remember, however, that GFAP staining only covers ∼15% of the total volume of astrocytes (Sun and Jakobs, 2012) and therefore, the reported percentage of cells expressing Piezo1 might be somewhat of an underestimation. Interestingly, in vivo Piezo1 expression closely matches our in vitro results following incubation of enriched astrocyte cultures with Aβ1−<sup>42</sup> CMM, i.e., only a proportion of astrocytes (∼30%) express Piezo1 channels and the expression appears perinuclear in reactive astrocytes in vivo.

# Aging and Amyloid Plaque Pathology Increase Piezo1 Expression in Anterior Cortical Regions

We next characterized spatiotemporal changes in Piezo1 expression in different areas of the cerebral cortex in 12- and 18-month old wild-type (WT) and TgF344-AD (TG) rats. Low magnification (20× montage) images of the prefrontal cortex reveal an upregulation of Piezo1 and GFAP around amyloid plaques (**Figures 5A–J**). Piezo1 expression in the prefrontal cortex increased 32 ± 3% from 12 to 18 months of age in WT rats (**Figure 5K**; p < 0.0001). However, frontoparietal (**Figure 5L**) and audiovisual (**Figure 5M**)

immunofluorescently stained for (A) conformation-specific Aβ1−42, (B) GFAP, (C) Piezo1 (N-terminal antibody), and (D) DAPI. Scale bar = 40 µm. (E) High magnification (100× objective) z-stack projections of prefrontal cortex amyloid plaques reveal an upregulation of GFAP in reactive astrocytes surrounding the amyloid plaque. Piezo1 protein is also upregulated in a proportion of astrocytes in and around the core of the plaque. Piezo1 appears localized predominantly to the perinuclear compartment of Aβ1−<sup>42</sup> reactive astrocytes. (F) Astrocytic Piezo1 fluorescence intensity was calculated for astrocytes inside the amyloid plaque versus astrocytes located 200–500 µm away from any noticeable plaque deposits. Piezo1 fluorescence intensity was averaged for six astrocytes "inside" and six astrocytes "outside" the plaque and 25 images from the frontal cortex of 18 m TG rats (n = 6) were analyzed. There was no correlation (Pearson r = 0.468) between Piezo1 expression inside versus outside amyloid plaques. (G) However, on average there was a 190% increase in Piezo1 expression in astrocytes within the plaque core (Inside) versus astrocytes at least 200 µm away from the edge of any amyloid plaque (Outside). A two-tailed paired t-test was performed. <sup>∗</sup>Represents a statistically significant difference (p < 0.001) inside versus outside of the plaque.

cortical regions did not display any significant changes in Piezo1 expression in WT rats from 12 to 18 months. Piezo1 channel expression also increased in the prefrontal (34 ± 2%) and frontoparietal (31 ± 2%) cortices of TG rats from 12- to 18-months of age but not in the audiovisual area (17 ± 2%). Taken together, both aging and amyloid plaque pathology trigger increases in Piezo1 channel expression, particularly in the prefrontal cortex, of TgF344-AD rats (**Figures 5K–N**).

## Peripheral Bacterial Infection Enhances Piezo1 Expression in the Cerebral Cortex of TgF344-AD Rats

Twelve month old TgF344-AD rats had previously been exposed to a UTI (E. coli) at 8 and 11 months of age (iTG). This bacterial infection caused a significant upregulation of Piezo1 channels in the prefrontal (21 ± 3%; **Figure 5K**) and frontoparietal (23 ± 3%; **Figure 5L**) cortices of 12-month old iTG rats, but not the audiovisual cortex (16 ± 2%; **Figure 5M**). In contrast, in older 18-month iTG rats that had been infected with E. coli at 8, 11, 14, and 17 months of age, no changes in Piezo1 channel expression were measured in prefrontal (12 ± 2%; **Figure 5K**) or frontoparietal (12 ± 3%; **Figure 5L**) cortices compared to noninfected TgF344-AD counterparts. However, Piezo1 expression in the audiovisual cortex did display significant upregulations (21 ± 3%; **Figure 5M**) in 18-month iTG rats in response to repeated peripheral infections. Interestingly, peripheral E. coli infections did not increase Piezo1 expression in the prefrontal cortex of WT rats at 12 months (15 ± 3%) nor at 18 months (12 ± 3%) of age compared to WT controls. Taken together, our results indicate that repeated UTI enhances Piezo1 expression in

FIGURE 5 | Mechanosensing Piezo1 channel expression increases in the cerebral cortex of TgF344-AD rats with a urinary tract infection (UTI). Sagittal sections (lateral –1.9 mm) of wild-type (WT) Fisher and TgF344-AD rat brains were immunofluorescently stained for Piezo1 (N-terminal antibody), conformation-specific Aβ1−42, GFAP, and DAPI. Shown are representative images of the prefrontal cortex of an 18-month WT rat (A–E) and an 18-month TgF344-AD (18 m TG) (F–J). Scale bar = 150 µm. Elevations in Piezo1 channel expression with age, Alzheimer's disease (AD) pathology and peripheral infection were measured in three distinct areas of the cerebral cortex, i.e., the prefrontal cortex (K), the frontoparietal cortex (L), and the audiovisual cortex (M). In addition, Piezo1 expression was averaged over all three regions to illustrate changes in channel expression in the total cerebral cortex (N). Data were normalized to 12 month wild-type (12m WT) values for each cortical region and are represented as the mean ± SEM of the percentage change from the 12m WT groups. Three-way ANOVAs with Holm-Sidak post hoc tests were performed to test for interactions between age, genotype, and peripheral infection. <sup>∗</sup>Represents a change > 20% and a p-value < 0.01. There were 5–6 rats per group and 2–3 brain sections per animal. Ten regions of interest (ROI) were analyzed per cortical region per section, i.e., 100–180 ROIs analyzed per group.

the cerebral cortex of TgF344-AD rats over and above that seen in response to amyloid plaques alone (**Figure 5N**).

### Piezo1 Upregulation in the Hippocampus of TgF344-AD Rats Precedes Changes in the Cerebral Cortex

Just as normal aging increased expression of Piezo1 channels in the prefrontal cortex, increases in Piezo1 were also evident in the CA1 (34 ± 4%) and DG (24 ± 3%) of the hippocampus, but not in the CA3 region, from 12 to 18 months in WT rats (**Figures 6A–W**). Notably, there were also large relative increases in Piezo1 expression in the DG (99 ± 7%; **Figure 6U**), CA3 (58 ± 6%; **Figure 6V**) and CA1 (78 ± 6%; **Figure 6W**) of 12 month TG rats compared to their 12-month WT conspecifics. Interestingly, however, there were no further significant increases in hippocampal Piezo1 expression in TG rats from 12 to 18 months of age, suggesting that amyloid deposition and plaque load begin in the hippocampus before spreading further to cortical regions.

# Peripheral Infection Upregulates Hippocampal Piezo1 Expression in Wild-Type and TgF344-AD Rats

Unlike the cerebral cortex, upregulations in Piezo1 expression were measured in the hippocampal CA1 region (28 ± 5%) in response to peripheral E. coli infection at 12 months of age in iWT rats (**Figure 6W**). However, no further increases in Piezo1 expression were evident in the CA1 of infected 18-month iWT rats. Peripheral infection at 8 and 11 months did, however, cause an upregulation of Piezo1 channels in the CA3 region of iTG rats at 12 months of age (20 ± 7%; **Figure 6V**). Moreover, repeated peripheral infection at 8, 11, 14, and 17 months of age caused a large increase in Piezo1 channels in the dentate gyrus of 18 month iTG rats (25 ± 8%) compared to their non-infected and age-matched TG counterparts (**Figure 6U**).

## Piezo1 Expression Strongly Correlates With Amyloid Plaque Deposition in the Hippocampal Dentate Gyrus

It is clear from our immunofluorescence data that both GFAP reactivity and Piezo1 expression are greater around amyloid plaques, especially in the hippocampus. Apart from the molecular layer of the cerebellum, the dentate gyrus displayed the lowest levels of Piezo1 expression in 12-month old WT rats (**Figure 2O**). Because basal levels of Piezo1 in the DG are low, the relative (%) increases in Piezo1 protein expression in 12-month old, 18-month old, and 18-month infected Tg-344AD rats were large (**Figure 6U**). To assess the relationship between Piezo1 expression and amyloid plaque pathology in adult and aged rats with and without repeated peripheral infections, we correlated Piezo1 fluorescence intensity with both GFAP and Aβ1−<sup>42</sup> expression in the dentate gyrus of all groups (**Figure 7**). In agreement with our in vitro studies, Piezo1 expression did not correlate with GFAP expression in 12-month WT, 18-month WT, or 12-month iWT (**Figures 7A–C**). However, DG Piezo1 expression showed a moderate correlation with GFAP intensity in 18-month iWT rats exposed to repeated peripheral infections (**Figure 7D**; Pearson r = 0.523, p < 0.0001). Similarly, there was no correlation between GFAP fluorescence intensity and Piezo1 expression in the DG of 12-month TG, 18-month TG, or 12-month iTG (**Figures 7E–G**). In 18-month iTG, however, increases in GFAP reactivity in the DG did correlate with increased Piezo1 expression (**Figure 7H**; Pearson r = 0.519, p < 0.0001). Peripheral infection of older rats led to an upregulation of GFAP in the hippocampal DG and these reactive astrocytes were more likely to express Piezo1 channels. GFAP expression in the DG of older iTG rats that had repeated peripheral infections also correlated with high amyloid plaque pathology (**Figure 7L**; Pearson r = 0.586, p < 0.0001), when compared to 18-month TG rats without infection (**Figure 7K**). Finally, Piezo1 expression in the DG displayed very strong correlations with Aβ1−<sup>42</sup> fluorescence intensity in 12- and 18-month TgF344-AD rats either with or without peripheral infection (**Figures 7M–P**). This suggests that whilst peripheral E. coli infection alone can upregulate astrocytic Piezo1; Aβ1−<sup>42</sup> CMM and in vivo amyloid plaque pathology are stronger triggers of mechanosensing Piezo1 channel expression in reactive astrocytes.

# DISCUSSION

TgF344-AD transgenic rats overexpress mutated human PSEN1 (PSEN11E9) and human APP with the Swedish mutation (APPSWE) and were developed with the intention of creating an improved animal model of AD that recapitulates a broader spectrum of neuropathological features seen in the human AD brain (Cohen et al., 2013). As such, TgF344-AD rats display neuronal tau deposition in contrast to what is generally observed in transgenic mice with these particular gene mutations (Duyckaerts et al., 2008; Stancu et al., 2014). Unless tauassociated genes are also mutated, transgenic mice tend not to spontaneously form neurofibrillary tau tangles (Holcomb et al., 1998; Jankowsky et al., 2001). By combining this improved transgenic rat model with a UTI, a common comorbidity of elderly people who develop AD, the model we present here more closely mimics the systemic inflammation and the spatiotemporal deposition of amyloid plaques observed in the hippocampus and cortex of aging human AD subjects.

Utilizing the TgF344-AD rat model (Cohen et al., 2013), we show here that Piezo1 protein expression increases in several key brain structures in response to aging, peripheral infection and amyloid plaque pathology. Piezo1 channels increase in the prefrontal cortex and the hippocampal CA1 and dentate gyrus with normal aging. These brain regions are particularly important for processing salient sensory information (Barbas and Zikopoulos, 2007; Lester et al., 2017), for learning and memory (Kane and Engle, 2002; Guo et al., 2018) and for higher executive functions (Arnsten and Li, 2005; Suwabe et al., 2017), all of which deteriorate in the elderly (Zimprich and Kurtz, 2013; Thomé et al., 2016; Wiegand et al., 2016). Since an increase in stretchactivated Ca2<sup>+</sup> channel expression could, in theory, predispose

FIGURE 6 | Aging, Aβ1−<sup>42</sup> pathology and peripheral infection upregulate Piezo1 expression in the hippocampus of TgF344-AD rats. Sagittal sections (lateral −1.9 mm) of WT Fisher and TgF344-AD rat brains were immunofluorescently stained for Piezo1 (N-terminal antibody), conformation-specific Aβ1−42, GFAP, and DAPI. Shown are representative images of the hippocampus of an 18-month wild-type (18m WT) rat (A–E), an 18-month TgF344-AD 18m TG (F–J), an 18-month WT with infection (18m iWT) (K–O), and an 18-month TgF344-AD with infection (18m iTG) (P–T). TgF344-AD rats with and without an infection show clear upregulations of GFAP (I,S) in hippocampal areas with dense amyloid deposition (H,R), particularly the hilus, the outer molecular layer of the dentate gyrus, and the stratum oriens of the CA1. Moreover, amyloid plaques surrounded by a dense meshwork of GFAP-positive astrocytes generally showed concomitant upregulations of Piezo1 expression, particularly in 18m TG rats with or without an infection (G,Q). Next, changes in Piezo1 expression with age, infection and genotype were quantified in the neuronal cell body layers of the dentate gyrus (U), CA3 (V), and CA1 (W). Piezo1 expression increased from 12- to 18-months in WT rats in the DG and CA1, but not CA3 neurons. A much more potent trigger of Piezo1 expression in the hippocampus, however, was genotype with TgF344-AD rats expressing much larger levels of Piezo1 in the DG (U), CA3 (V), and CA1 (W) at 12-months of age. There were no further increases in Piezo1 expression in 18-month old TgF344-AD rats. However, peripheral infection caused a significant increase in Piezo1 expression in the dentate gyrus (U) of TgF344-AD rats (18m iTG). Data were normalized to 12m WT values for each hippocampal region and are represented as the mean ± SEM of the percentage change from the 12m WT groups. Three-way ANOVAs with Holm-Sidak post hoc tests were performed to test for interactions between age, genotype, and peripheral infection. <sup>∗</sup>Represents a change of > 20% and a p-value < 0.01. There were 5–6 rats per group and 2–3 brain sections per animal. Ten ROI were analyzed per hippocampal region per section, i.e., 100–180 ROIs analyzed per group.

FIGURE 7 | Piezo1 expression in the hippocampal dentate gyrus correlates with GFAP in aged rats with peripheral infection and strongly correlates with amyloid plaques in TgF344-AD rats. Sagittal brain sections were triple-stained for Piezo1, GFAP and conformation-specific Aβ1−<sup>42</sup> and Pearson correlations (r values) were performed to measure changes in Piezo1 channel expression in astrocytes (GFAP vs. Piezo1) with age and peripheral infection in WT (A–D) and TgF344-AD (E–H) rats. There was a moderate Pearson correlation between GFAP and Piezo1 in the infected 18-month WT (r = 0.523) and infected 18-month TgF344-AD rats (r = 0.519). However, the linear regression (R 2 ) values were relatively weak for both groups suggesting that the GFAP fluorescence intensity is not a good predictor of Piezo1 channel expression. Next, Pearson correlations were performed to measure changes in GFAP expression around amyloid plaques (GFAP vs. Aβ1−42) in the dentate gyrus with age and peripheral infection in TgF344-AD rats (I–L). There was a moderate Pearson correlation between GFAP and Aβ1−<sup>42</sup> in the infected 18-month old TgF344-AD rats (r = 0.586) but a relatively weak linear regression (R 2 ) value suggesting that Aβ1−<sup>42</sup> fluorescence intensity is not a good predictor of GFAP expression. Finally, Pearson correlations were performed to measure changes in Piezo1 channel expression in and around amyloid plaques (Aβ1−<sup>42</sup> vs. Piezo1) in the dentate gyrus with age and peripheral infection in TgF344-AD rats (M–P). There were strong Pearson correlations between Aβ1−<sup>42</sup> and Piezo1 in 12-month old TgF344-AD rats (r = 0.917), 12-month old TgF344-AD rats with peripheral infection (r = 0.736), 18-month old TgF344-AD rats (r = 0.770), and 18-month old TgF344-AD rats with peripheral infection (r = 0.856). In addition, the high linear regression (R 2 ) values for each group (M–P) suggests that Aβ1−<sup>42</sup> fluorescence intensity is a good predictor of Piezo1 channel expression in the dentate gyrus of TgF344-AD rats.

aging neurons to Ca2<sup>+</sup> dysregulations (Nagarajan et al., 2014), it would be interesting to further investigate if Piezo1 channel expression modulates neuronal Ca2<sup>+</sup> homeostasis. Intrinsic neuronal properties and synaptic plasticity are intimately linked with tight control of cellular Ca2<sup>+</sup> concentrations (Catterall et al., 2013; Maggio and Vlachos, 2014) and, therefore, mechanosensitive channels that regulate calcium levels may play an important role in cognition in the aging hippocampus or prefrontal cortex. In this regard, various members of the transient receptor potential (TRP) channel family have been implicated in both mechanosensation (Kwan et al., 2006) and Ca2<sup>+</sup> dysregulations in neurons and astrocytes in AD (Yamamoto et al., 2007). In the present study, the age-dependent increases in Piezo1 channel expression measured in the prefrontal cortex, CA1 and DG of wild-type rats appeared predominantly neuronal whereas in the TgF344-AD rat model, peripheral infection and amyloid plaque pathology caused an upregulation of Piezo1 predominantly in reactive astrocytes, although concomitant increases in neurons and reactive microglia are also likely to occur. This is interesting because the role of Piezo1 in CNS neurons could be different to its function in astrocytes since neurons constitutively express mechanosensing Piezo1 channels on axonal growth cones and the outer cell membrane, whereas astrocytes exposed to neuroinflammatory stimuli, such as Aβ1−<sup>42</sup> CMM, upregulate and accumulate Piezo1 channels within the perinuclear compartment. Indeed, in epithelial cell types Piezo1 reportedly co-localizes with the endoplasmic reticulum where it regulates intracellular Ca2<sup>+</sup> release from stores (McHugh et al., 2010). Moreover, separating the physiological role of Piezo1-mediated extracellular calcium influx from Piezo1 regulated Ca2<sup>+</sup> release from intracellular stores will reveal subsets of novel mechanotransduction cascades that originate from either outer membrane-spanning Piezo1 channels or from intracellular Ca2<sup>+</sup> stores (Clapham, 2007). Our results suggest that Piezo1 channels regulate glial mechanosensation and may trigger astrogliosis in the AD brain. Moreover, amyloid plaque-associated microglia and the biochemical factors they release when activated likely play a key role in controlling Piezo1 expression in reactive astrocytes. As such, TgF344-AD rats infected with E. coli show enhanced Piezo1 expression in the cortex and hippocampus compared to non-infected transgenic animals. Moreover, microglia have been shown to regulate the pathogenic activities of astrocytes in a mouse model of multiple sclerosis (Rothhammer et al., 2018). Therefore, determining whether Piezo1 regulates neurodegenerative or neuroprotective mechanotransduction signaling in reactive astrocytes surrounding amyloid plaques is an important topic for future research.

#### Astrocytic Piezo1 in Alzheimer's Disease

Satoh et al. (2006) discovered that astrocytes surrounding amyloid plaques in the cerebral cortex of AD patients express hMib mRNA. Coste et al. (2010) renamed Mib (also known as FAM38A), Piezo1, and described how it forms a stretchactivated calcium-permeable channel in the neuroblastoma cell line, Neuro2A. Elegant new research has revealed the tertiary structure of Piezo1 at near atomic level (Ge et al., 2015) and describes how non-selective cationic channels are formed from homotrimers which exhibit large mechanosensitive 'propeller-like' structures on the outer surface of membranes (Coste et al., 2012; Li et al., 2014). Astrocytes are very sensitive to changes in their local mechanical environment (Leipzig and Shoichet, 2009; Blumenthal et al., 2014) and our data suggests that perturbations caused by stiff amyloid plaques trigger the transcriptional upregulation of mechanosensing Piezo1 channels in astrocytes which are normally devoid of Piezo1 protein (**Figure 3E**). But why might a sub-population of reactive astrocytes upregulate mechanosensing Piezo1 channels? Interestingly, Liddelow et al. (2017) showed that ∼60% of GFAP-positive astrocytes in the prefrontal cortex of human AD patients were of the A1 phenotype, i.e., a neurotoxic subtype of reactive astrocyte that drives neuroinflammation and causes neurotoxicity. It would be interesting to determine if the 30– 40% of Piezo1-expressing astrocytes around amyloid plaques in the TgF344-AD rat model stain positive for A1 astrocyte markers (e.g., Serping1), or for astrocytes of the A2 phenotype (e.g., S100A10) which reportedly release neurotrophic factors and are mainly neuroprotective (Liddelow et al., 2017). More research is needed to determine what makes Piezo1-expressing astrocytes around amyloid plaques different to their Piezo1 negative neighbors and to characterize both subpopulations based on their 'neuroinflammatory phenotype.'

#### Piezo1 as a Regulator of Neuron/Glial Communication

Blumenthal et al. (2014) suggest that neuronal Piezo1 is important for sensing the topography of cells at the nanometre scale. Hippocampal neurons cultured in vitro form many neuron/glial interactions when grown on relatively smooth glass substrates but blocking mechanosensitive channels with GsMTx4, a peptide component of the Grammostola rosea tarantula venom, promotes decoupling of neurons from astrocytes and leads to fewer cell contacts. This suggests that Piezo1 can regulate the physical communication between neurons and astrocytes. Moreover, they also showed that the surface topography of brain tissue becomes rougher in response to amyloid plaque pathology which can increase the nanoroughness of astrocytes, thus contributing to neurodegeneration (Blumenthal et al., 2014). If Piezo1 does indeed regulate neuron/glial communication in vivo, this supports our previous work which revealed that Piezo1 channels are involved in neuronal mechanosensation and axon guidance during brain development (Koser et al., 2016), enabling migrating growth cones to detect changes in the stiffness of their local environment. Here, we confirm that the rat optic tract also expresses relatively high levels of Piezo1 channels (**Figure 1A**), as well as the corpus callosum and cerebellar arbor vitae, suggesting that Piezo1 may play a more general role in axonal pathfinding during brain development. Indeed, migrating axons and growth cones are highly mechano-responsive and likely receive constant feedback about the stiffness of the local ECM or the apparent density of glial cell networks as they grow through different brain regions toward their final targets. Moreover, glial cells are softer than neurons (Lu et al., 2006)

and recent studies have shown that SH-SY5Y cells (a human neuroblastoma cell line) that overexpress mutated APPSWE are stiffer than normal SH-SY5Y cells (Lu et al., 2017). Therefore, astrocytes capable of engulfing and phagocytosing neurotoxic Aβ1−<sup>42</sup> peptides may change their mechanical properties or surface topography in the AD brain. Neurons in contact with reactive astrocytes that engulf neurotoxic amyloid peptides may be able to sense a change in their cellular stiffness or membrane nanoroughness and detach from them. Therefore, widespread loss of cell contacts in the hippocampus and cortex, or a disruption to neuron/glial communication caused by changes in the mechano-responsiveness of cells, may accelerate neurodegeneration (Blumenthal et al., 2014) and cognitive decline in the AD brain.

Recent evidence also suggests that dysregulated mechanosensitive signaling can contribute to astrogliopathy in Alexander disease, leading to an increase in brain stiffness (Wang et al., 2018). Alexander disease is a rare neonatal neurodegenerative disorder caused by accumulation of pathogenic Rosenthal fibers due to mutations in astrocytic GFAP. In contrast, we have recently shown that glial scars rich in GFAP- and Vimentin-expressing glial cells, which form following cortical stab injury or contusive spinal cord injury, are much softer than surrounding healthy CNS tissue (Moeendarbary et al., 2017). Therefore, astrogliosis can apparently lead to either increases or decreases in brain tissue stiffness depending on the underlying causes of tissue injury or disease. As such, reactive astrogliosis can be classified as isomorphic (conserved morphology) or anisomorphic (astrocyte hypertrophy and proliferation) (Verkhratsky et al., 2013). These reactive states differ, not only in morphology, but also in their (patho)physiological functions. For instance, isomorphic gliosis is thought to aid regeneration of neuronal networks by providing a permissive biochemical microenvironment for growth cones to migrate through on their journey to relocate their correct synaptic targets. On the other hand, anisomorphic astrogliosis contributes to glial scar formation which is initially beneficial because it prevents extensive neurodegeneration, but as the primary trauma heals, the glial scar can also inhibit axonal re-growth across the lesion site due to secretion of factors such as chondroitin sulfate proteoglycans (Bradbury et al., 2002; Fawcett, 2006). Interestingly, we have shown that addition of chondroitin sulfate to the Xenopus brain in vivo causes widespread softening of brain tissue (Koser et al., 2016). Therefore, differences in the secretome of phenotypically and functionally distinct types of reactive astrocyte could, in theory, explain the dichotomy in brain tissue stiffness observed in Alexander disease (Wang et al., 2018) versus cortical glial scars caused by a traumatic stab injury (Moeendarbary et al., 2017). In both cases, however, astrogliosis will likely disrupt neuronal mechanobiological signaling and may trigger either neurodegenerative or neuroprotective/proregenerative mechanisms. An important area for future research will involve characterizing the role of Piezo1 in neurons and glial cells in the injured or degenerating brain. With this aim in mind, Piezo1 channel opening reportedly regulates the Hippo signaling pathway (Pathak et al., 2014), which controls cell proliferation and apoptosis. Piezo1 knockdown also decreases levels of the mechanoresponsive transcriptional coactivator, Yesassociated protein (YAP), in neural stem cells (Pathak et al., 2014) suggesting that Piezo1 activation triggers YAP transcriptional activity. Notably, YAP is highly expressed by astrocytes and deletion of the YAP gene induces astrogliosis (Huang et al., 2016). In Alexander disease, astrocytes overexpress nuclear YAP, potentially as a result of increased brain tissue stiffness (Wang et al., 2018). Whether upregulation of YAP is detrimental to astrocytes or induces a protective phenotype is not yet known. YAP has some protective functions and can activate suppressor of cytokine signaling (SOCS) gene expression and, therefore, can prevent astrocyte reactivity by downregulating the JAK–STAT inflammatory pathway (Huang et al., 2016). Taken together, the upregulation of astrocytic Piezo1 that we see around amyloid plaques in the hippocampus and cortex of TgF344-AD rats could, on the one hand, contribute to AD-associated astrogliopathy and neurodegeneration but, on the contrary, astrocytic Piezo1 may instead drive anti-inflammatory signaling cascades through YAP-mediated suppression of cytokine signaling that work in concert to reduce reactive astrogliosis and neuroinflammation in the aging AD brain.

#### CONCLUSION

The data presented here suggests that mechanoresponsive astrocytes surrounding stiff amyloid plaques in the AD brain upregulate Piezo1 channels for, as yet, unknown reasons. Moreover, peripheral infection, a common comorbidity in elderly people with AD, augments Piezo1 channel expression in astrocytes. This suggests a potential role for pro-inflammatory microglia in regulating astrocytic Piezo1 levels. Future studies will focus on the signaling cascades triggered by Piezo1 mediated currents in reactive astrocytes and Piezo1 as a potential therapeutic target for amyloid plaque-induced neurodegeneration in AD.

#### AUTHOR CONTRIBUTIONS

GS and KD conceived the project and designed the research. MV-E, MM, HB, AC, EM, and GS performed the experiments. MV-E, KD, and GS analyzed the data. PW, AS, and EB performed the infections in WT and TG rats. MV-E and GS wrote the manuscript with contributions from all authors. All authors discussed the data and results.

#### FUNDING

This work was supported by an Irish Research Council Ph.D. scholarship for MV-E, a University of Brighton Ph.D. scholarship for MM, and a Ph.D. studentship from the BioImaging Institute, The University of Manchester for AC. Purchase of TgF344-AD rats was jointly supported by the European Union's Seventh Framework Programme

(FP7/2007-2013) under Grant Agreement No. HEALTH-F2-2011-278850 (INMiND) and Alzheimer Research United Kingdom network funds. Breeding and maintenance costs for the TgF344-AD rat colony was supported by the European Union's Seventh Framework Programme (FP7/2007- 2013) under Grant Agreement No. HEALTH-F2-2011-278850 (INMiND). EM was grateful for funding from Cancer Research United Kingdom [C57744/A22057]. GS was grateful for a Rising Stars award and for support from the Chemistry-Biology Collaborative Research Fund, University of Brighton.

#### REFERENCES


#### ACKNOWLEDGMENTS

We thank Karen Bryson (Trinity College Dublin) for providing the Aβ1−<sup>42</sup> CMM for cell culture experiments and Sinead O'Sullivan (University of Chicago) for experimental support during the early parts of this study. We would like to thank Christopher Thrasivoulou and Tim Robson (University College London) for technical support with fluorescence imaging. We also thank the Centre for Stress and Age-Related Disease (University of Brighton) for supporting this research project.

non-phlogistic clearance of amyloid-beta and cognitive function in APP/PS1 mice. Brain Behav. Immun. 61, 306–316. doi: 10.1016/j.bbi.2016.12.014


individual-differences perspective. Psychon. Bull. Rev. 9, 637–671. doi: 10.3758/ BF03196323



Zimprich, D., and Kurtz, T. (2013). Individual differences and predictors of forgetting in old age: the role of processing speed and working memory. Neuropsychol. Dev. Cogn. B Aging Neuropsychol. Cogn. 20, 195–219. doi: 10. 1080/13825585.2012.690364

**Conflict of Interest Statement:** Evotec (UK) Ltd. was involved in the design of the urinary tract infection study performed on wild-type and TgF344-AD rats and analyzed the urine samples for bacterial load. Evotec (UK) Ltd. was not involved in the design or analysis of any other experiments.

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 Velasco-Estevez, Mampay, Boutin, Chaney, Warn, Sharp, Burgess, Moeendarbary, Dev and Sheridan. 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.

# Potential Epigenetic-Based Therapeutic Targets for Glioma

Lanlan Zang1,2 , Shukkoor Muhammed Kondengaden<sup>3</sup> , Fengyuan Che1,4 , Lijuan Wang<sup>1</sup> \* and Xueyuan Heng<sup>4</sup> \*

<sup>1</sup>Central Laboratory and Key Laboratory of Neurophysiology, Linyi People's Hospital, Shandong University, Linyi, China, <sup>2</sup>Department of Medicinal Chemistry, Key Laboratory of Chemical Biology (Ministry of Education), School of Pharmaceutical Sciences, Shandong University, Jinan, China, <sup>3</sup>Chemistry Department and Center for Diagnostics and Therapeutics, Georgia State University, Atlanta, GA, United States, <sup>4</sup>Department of Neurology, Linyi People's Hospital, Shandong University, Linyi, China

Glioma is characterized by a high recurrence rate, short survival times, high rates of mortality and treatment difficulties. Surgery, chemotherapy and radiation (RT) are the standard treatments, but outcomes rarely improve even after treatment. With the advancement of molecular pathology, recent studies have found that the development of glioma is closely related to various epigenetic phenomena, including DNA methylation, abnormal microRNA (miRNA), chromatin remodeling and histone modifications. Owing to the reversibility of epigenetic modifications, the proteins and genes that regulate these changes have become new targets in the treatment of glioma. In this review, we present a summary of the potential therapeutic targets of glioma and related effective treating drugs from the four aspects mentioned above. We further illustrate how epigenetic mechanisms dynamically regulate the pathogenesis and discuss the challenges of glioma treatment. Currently, among the epigenetic treatments, DNA methyltransferase (DNMT) inhibitors and histone deacetylase inhibitors (HDACIs) can be used for the treatment of tumors, either individually or in combination. In the treatment of glioma, only HDACIs remain a good option and they provide new directions for the treatment. Due to the complicated pathogenesis of glioma, epigenetic applications to glioma clinical treatment are still limited.

#### Edited by:

Alberto Javier Ramos, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Marianela Candolfi, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina Timothy J. Jarome, Virginia Tech, United States

#### \*Correspondence:

Lijuan Wang wanglj730@163.com Xueyuan Heng hengxueyuan12@163.com

Received: 12 July 2018 Accepted: 16 October 2018 Published: 15 November 2018

#### Citation:

Zang L, Kondengaden SM, Che F, Wang L and Heng X (2018) Potential Epigenetic-Based Therapeutic Targets for Glioma. Front. Mol. Neurosci. 11:408. doi: 10.3389/fnmol.2018.00408 Keywords: glioma, epigenetics, DNA methylation, miRNA, chromatin remodeling, histone modifications

#### INTRODUCTION

Glioma is the most common form of primary malignant brain tumors, accounting for nearly 30% of all brain tumors, and also one of the most lethal (Chien et al., 2016). Some symptoms of glioma may be subtle and gradually worsen, while others may be present as an acute illness. The exact mechanism of occurrence of glioma remains unclear.

Recently, the World Health Organization (WHO) Classification of Tumors of the Central Nervous System (CNS; Wen and Huse, 2017) categorized glioma into four grades. Grades I and II are considered low-grade glioma, whereas grades III and IV are high-grades, with the degree of malignancy increasing in the higher grades. The types of glioma include low-grade glioma (WHO II: mainly refers to diffuse astrocytoma, oligodendroglioma, and oligodendrocyte astrocytoma), anaplastic glioma (WHO III), glioblastoma (GBM, WHO IV), brain gliomatosis (pathologically astrocytoma-based, can be divided into WHO II, III, IV) and ependymomas (WHO II, III). Each grade has a relatively specific prognosis to guide the clinical treatment, but most of the glioma are WHO III grade anaplastic glioma and WHO grade IV GBM, which is devoid of cure with the current therapeutic options.

Unfortunately, most high-grade gliomas do not have a definitive cure. The current treatment for these high-grade tumors mainly focuses on surgical resection, followed by chemotherapy and radiotherapy. Temozolomide (TMZ) can traverse the blood-brain barrier and it is often used to treat these tumors. However, various side effects and adverse prognosis exist, such as the hematological toxicity (Stupp et al., 2005), and glioma have proven to be particularly resistant to radiotherapy and chemotherapy. The main mechanism of drug resistance is correlated with the O6-methyl guanine-DNA methyltransferase (MGMT; Hegi et al., 2005), which can bind to an alkyl compound on the O6 of DNA guanine and reduce it so that protein complexes cannot recognize the mismatched base pair, resulting in cell tolerance to O6 guanine methylation. As a result, the efficacy of these treatments is not promising and the prognosis of patients remains poor.

Therapies aimed to overcome these limitations have been presented in recent preclinical and clinical studies, including targeted molecular therapy, immunotherapy, gene therapy and stem cell therapy. The success of targeted molecular therapy for some tumor types, such as non-small cell lung cancer (Antonelli et al., 2016), malignant melanoma (Sullivan and Atkins, 2009) and chronic myeloid leukemia (Smirnikhina et al., 2016), has important guiding significance for targeted malignant glioma therapy.

With the advancement of the molecular pathology of malignant glioma, it is now evident that epigenetic abnormalities, including aberrant DNA methylation, abnormal microRNA (miRNA), chromatin remodeling and histone modifications, are closely related to the occurrence of glioma (Kondo et al., 2014; Hashizume, 2017; **Figure 1**). Multiple enzymes and genes that regulate the epigenetic modifications have become new targets not only for glioma treatments but for the treatment of other cancers as well.

This review summarizes the potential epigenetic-based therapeutic targets and related drugs for glioma. We illustrate how the epigenetic mechanisms dynamically regulate the pathogenesis of the disease. To provide information that will assist clinicians, we discuss preclinical and clinical trials of epigenetic-based treatments of glioma and include the results from these studies. This review also highlights the limitations of the current treatment and suggests future potential areas for research and advancements in the prognosis of the disease.

#### DNA METHYLATION

DNA methylation is one of the earliest discovered epigenetic modifications pathways. There are four possible sites for DNA methylation (Jones, 2012; Lovkvist et al., 2016)—the N-6 position of adenine, the N-4 position of cytosine, the N-7 position of guanine and the C-5 position of cytosine. In mammalian cells, DNA methylation occurs predominantly in cytosine of 5'-CpG-3' to produce 5-methylcytosine (5mC). The methylation reaction is site-specific and it is performed by the enzyme DNMT with the help of methyl donor s-adenosylmethionine (SAM) as a co-factor (Bird, 2002). Currently, according to the differences in their structure and function, these DNMTs are divided into three categories, DNMT1, DNMT2, DNMT3 (DNMT3a and DNMT3b), with DNMT1 and DNMT3 as the representatives. DNMT1 is involved in the maintenance and extension of methylation, a necessity for non-CpG site methylation. DNMT3a and DNMT3b are de novo methyltransferases that methylate CpG. De novo methyltransferases may be involved in the regulation of cell growth and differentiation, in which DNMT3b plays an important role in tumor gene methylation (Chédin, 2011). DNMT2 binds to a specific site on DNA and its main target is tRNA (Kaiser et al., 2017). A number of studies (Hashimshony et al., 2003; Krausz et al., 2012; Shimooka et al., 2013) show that DNA methylation can contribute to changes in chromatin structure, DNA conformation, DNA stability, interactions between DNA and proteins, and can also control gene expression. DNA methylation has additionally become an important study of epigenetics and epigenetic genomics due to the close relationship between DNA methylation, human development and tumor diseases, particularly the transcriptional inactivation of tumor suppressor genes caused by methylation of CpG islands content.

#### Aberrant DNA Methylation in Glioma

Many studies (de Souza et al., 2018; Liao et al., 2018) have shown that the patterns of DNA methylation in glioma cells are different from those in normal cells. Most notably, the coexistence of extensive hypomethylation and CpG island hypermethylation are characteristic of tumor cells. The DNA methylation status of some relevant genes in glioma is therefore a good biomarker for clinical glioma diagnosis (**Figure 1**; Qu et al., 2013; Choudhury et al., 2015). Hypermethylation of the gene promoter region is the clearest epigenetic change that occurs in the tumor. The methylation status of the promoter region in the human genome regulates gene expression. Almost all housekeeping genes and roughly half of the tissue-specific genes are under the control of the promoter region. Under normal physiological conditions, most CpG islands are in a hypomethylated state; however, some housekeeping genes, such as DNA repair genes and tumor suppressor genes, are often hypermethylated in tumor tissues (Jin et al., 2017). This aberrant methylation can lead to gene transcription suppression and the loss of their biological function. DNA-5-hydroxymethylcytosine (5hmC), an epigenetic mark resulting from 5mC oxidation, correlates with the progression of glioma (Fernandez et al., 2018). Recent evidence (Kraus et al., 2012) has shown that DNA 5hmC negatively correlates with tumor grade. The CpG island methylator phenotype (G-CIMP) is also used as an indicator of glioma prognosis in infancy, pediatric and adults (Mack et al., 2014; Malta et al., 2018). Importantly, another study (Jha et al., 2014) found that pediatric GBM have a distinct methylome compared with that of adults, suggesting that the G-CIMP indicator of glioma prognosis in adult GBM cannot simply be extrapolated to pediatric GBM and there is a strong need for identification of separate prognostic markers.

Abnormal DNA methylation is an important indicator of the inactivation of tumor suppressor genes. Many tumor suppressor genes in glioma have also been identified, including p16INK4a (Lee et al., 2000), p14ARF (Watanabe et al., 2007), MLH1 (Gömöri et al., 2007) and NDRG2 (Kolodziej et al., 2016; see **Table 1**). The p16INK4a gene maintains the dephosphorylated activating state of the retinoblastoma tumor suppressor protein (pRb) in the normal cyclinD-Rb pathway to control cell cycle progression. A high frequency (more than 50%) of homozygous p16INK4a gene deletion has been demonstrated in GBM tissues, and p16INK4a is altered in 80% of glioma cell lines. Therefore, restoration of p16INK4a would suppress cell proliferation and induce cell growth arrest (Lee et al., 2000). MGMT is a crucial DNA damage repair gene that can repair alkyl damage caused by BCNU. Esteller et al. (2000) found that MGMT promoter hypermethylation existed in approximately 40% of glioma tissues. The methylation level is related to the occurrence and prognosis of the tumor, which is more important than the prognosis of age and tumor grade (Mur et al., 2015). Besides, the methylation level of MGMT promoter is the most important indicator to assess TMZ sensitivity in glioma treatment, and down-regulated MGMT can substantially restore TMZ chemosensitivity in vitro and in vivo (Xipell et al., 2016; Yu et al., 2018). Except for the genes mentioned above, the CpG island methylation in gene promoter regions of p73 (Watanabe et al., 2002), LATS1, LATS2 (Jiang et al., 2006), and the genes that are listed in **Table 1** are also closely related with the occurrence and development of glioma.

Isocitrate dehydrogenases 1 (IDH1) is not only a major source of NADPH in the human brain (Bleeker et al., 2010) but also in other body tissues (Kim et al., 2009). The mutations of methylation regulatory proteins IDH1/2 can be detected in most low-grade diffuse astrocytoma (75% mutation rate) and anaplastic astrocytoma (66% mutation rate) as well as oligodendroglioma, mixed promyelocytoma and secondary sex polymorphic neuroblastoma (GBM, 76% mutation rate). Noushmehr et al. (2010) found that the IDH1 mutation is highly correlated with the G-CIMP, and its mutation was associated with the prognosis of secondary GBM and positively correlated with the survival rate of patients (Rossetto et al., 2011). Studies have shown that the hypermethylated phenotype of IDH mutations causes insulator proteins to separate from IDH mutant binding sites (Flavahan et al., 2016). These free insulator proteins are then linked to the normally resting platelet-derived growth factor receptor (PDGFRA) oncogene promoter to form a complex that can further stimulate tumor

#### TABLE 1 | DNA Methylation in glioma.


cell growth. By using demethylated drugs to restore normal function in these insulator proteins, PDGFRA can inhibit the growth of IDH mutant astrocytoma cells (Flavahan et al., 2016). In another study, they associated known subtypes with specific alterations in NF1 and PDGFRA/IDH1 in order to provide a framework for the investigation of targeted therapies (Verhaak et al., 2010). As a critical gene of glioma, IDH has vast potential for the diagnosis, treatment and prognosis of glioma.

# MicroRNA

MiRNAs are short non-coding endogenous RNAs that involve in post-transcriptional gene expression regulation in animals and plants (Bartel, 2004). miRNAs can target complementary regions of the 3'-UTR of the mRNA, inhibiting post-transcriptional processes or degrading mRNA and ultimately reducing protein levels (Pileggi et al., 2013). There are many miRNAs in the human genome, targeting tens of thousands of mRNAs.

#### MicroRNAs and Glioma

In recent years, some studies have shown that miRNAs play key roles in the transcriptional regulation and growth and proliferation of various tumor genes (Yan et al., 2017). Therefore, miRNA-based individual therapy and gene editing methods may play important roles in the diagnosis and treatment of glioma. It is currently estimated that about half of the miRNA genes are located in glioma cancer genes or their fragile sites (**Table 2**), and these miRNA genes can regulate 3% of the entire glioma tumor genes and 30% of the coding genes. Also, a single miRNA can simultaneously affect a hundred mRNAs of GBM (Berindan-Neagoe et al., 2014), whereas a single mRNA of glioma can be modulated by one or more miRNAs (Lakomy et al., 2011).

miRNAs play many critical roles in the progression of glioma diseases. In particular, miRNAs regulate the expression of cancer-related genes, participate in the regulation of tumorigenesis and regulatory pathways, regulate the differentiation of glioma stem cells, and are encoded by oncolytic


viruses and involved in tumor processes. Zhang C. et al. (2012), for instance, found miR-221/222 positively correlated with the degree of glioma infiltration and cell invasion, whereas knockdown of miR-221/222 decreased cell invasion via modulating the levels of the TIMP3 target. Knockdown of miR-221/222 additionally increased TIMP3 expression and considerably inhibited tumor growth in a xenograft model. Another study indicated that the over-expression of miR-221/222 reduced p27kipl levels (Zhang C. et al., 2009). P27kipl prevented cell cycle from G1 to S phase by binding to CDK2 and cyclin E complexes. Therefore, down-regulated miR-221/222 can up-regulate p27kipl to inhibit tumor proliferation.

#### CHROMATIN REMODELING

Chromatin remodeling complexes (CRCs) have ATPase activity and they rely on the hydrolysis of ATP to provide energy to complete the chromatin structure changes (Stanton et al., 2017). Depending on the different subunits that can hydrolyze ATP, the complexes can be divided into SWI/SNF, ISW I and other types of complexes (**Figure 1**). The SWI/SNF complex and the ISW I complex family were the first to be found in yeast and Drosophila (Biegel et al., 2014). The human SWI/SNF complex is a polymer with many molecules, including BRG1, hBMR and tumor suppressor protein Hsnf5, which mainly activates gene transcription and is also involved in the recombination of immunoglobulin and TCR genes (Pulice and Kadoch, 2016). The ISW I complex family include three complexes—RSF, HuCHRAC and CAF1 (Loyola et al., 2003). RSF is a heterodimer that mainly consists of Hsnfh, which is involved in transcription initiation (Sheu et al., 2010); HucHRAC contains Hsnf-2 h and chromatin assembly factor Hacf1, which is related to the maintenance of the heterochromatin replication status (Hanai et al., 2008); CAF1 is involved in chromatin assembly, altering the state of chromatin to correlate with DNA function (Endo et al., 2017). These complexes and related proteins are associated with activation and inhibition of transcription, DNA methylation, DNA repair and cell cycle.

#### Chromatin Remodeling and Glioma

Human diseases caused by abnormal chromatin remodeling are often due to mutations in the key proteins of the remodeling complex. This can lead to the failure of chromatin remodeling in which nucleosomes cannot be correctly positioned, preventing basic transcriptional machinery and the complexes that can repair DNA damage from accessing DNA, which can lead to aberrant gene expression. If these mutations lead to abnormalities of tumor suppressor genes or proteins that regulate cell cycle, they can finally lead to the occurrence of cancer (Marfella et al., 2008; Choi et al., 2015).

Liau et al. (2017) recently indicated that chromatin remodeling regulated GBM drug resistance. GBM stem cells (GSC) can reversibly change to a slow-cycling, long-lasting state when targeted by kinase inhibitors. Under this state, the notch signaling pathway is activated and histone demethylase KDM6A/B is significantly up-regulated. This leads to the removing of trimethylation of H3K27 in genome cis-regulatory region and further leads to the increased levels of H3K27Ac. Chromatin remodeling played a key role in this cellular shift, and this research provided a novel target for the development of effective treatments in the future. By targeting epigenetic and developmental pathways, it is possible to eradicate drug-resistant tumor cells and prevent disease recurrence. Another study (Xiao et al., 2017) revealed evidence demonstrating up-regulated chromatin remodeling factor lymphoid-specific helicase (LSH) promoted the development of glioma. Research (Xiao et al., 2017) indicates that the up-regulated transcription factor E2F1 and glycogen synthase kinase-3β (GSK-3β, an intact complex of E2F1) in astrocytomas and GBM were associated with the progression of glioma and correlated with LSH expression. The depletion of E2F1 decreased LSH expression and cell growth, while inhibition of GSK3β increased the enrichment of E2F1 to the LSH promoter, and increased LSH expression. Lipoprotein receptor-related protein 6 (LRP6), an upstream regulator of GSK3β signaling pathway, was also over expressed in glioma tissue. Knockdown of LRP6 reduced LSH expression level through decreased recruitment of E2F1 to the LSH promoter, finally leading to inhibition of cell growth. Taken together, a mechanistic link between LSH expression and activation of the LPR6/GSK3β/E2F1 axis in glioma illustrates a novel role of LSH in malignant astrocytomas and GBM. Understanding the roles of LSH in glioma progression will not only enrich our knowledge of glioma but also frame LSH as a potential therapeutic target for the treatment of these deadly brain cancers.

#### HISTONE MODIFICATIONS

In the mammalian epigenome, histone modifications can occur in many ways. The basic unit of histone, called nucleosome, is an octamer that consists of two H2A, two H2B, two H3, two H4 and 147 base pairs wound outside of the composition (de Ruijter et al., 2003; **Figure 1**). The core histone has C-terminal and N-terminal binding regions. Of these, the N-terminal is particularly relevant, since lysine residue of the N-terminal extends out of the nucleosome and is accessible for modifications, including acetylation, methylation, phosphorylation, ubiquitination and ADP ribosylation. These modifications subsequently alter the expression of the gene without altering the base pair (Wang et al., 2014; Mathias et al., 2015). The total process, known as epigenetic regulation, involves a variety of enzymes. Scientists have classified these enzymes according to their functions (Liu et al., 2011): ''writers'' (enzymes that add groups such as methyl, acetyl and glycans), ''erasers'' (enzymes that remove post-translational modifications) and ''readers'' (enzymes that recognize these epigenetic markers and regulate epigenetic effects). The protein complexes that promote the movement of nucleosomes on chromatin are called ''movers.'' From this enormous pool, many therapeutic targets are derived as a single target or as combinations, apart from the most prominent DNMTs and histone deacetylases (HDACs).

# Dysregulation of Histone Modifications in Glioma

Aberrant histone modifications can lead to transcriptional abnormalities in gene expression that eventually lead to the development and progression of glioma. Among the various histone modification proteins, two have attracted more attention than the others—HDACs, which cause histone deacetylation, and histone methyltransferases, which cause methylation at various sites of histone (Zang et al., 2014). Among the different classes of HDAC enzymes, HDAC1 (Wang et al., 2017), HDAC2, HDAC3 (Leng et al., 2016), HDAC5 and HDAC9 (Milde et al., 2010) have undergone significant changes in glioma cells. H3 acetylation levels are elevated in high-grade astrocytoma compared to low-grade medulloblastoma and normal brain tissues, as are the expression of HDAC5 and HDAC9 (Milde et al., 2010) in high-grade medulloblastoma. One survey demonstrated that mRNA levels of class II and IV HDACs were down-regulated in GBM compared to low-grade astrocytomas and normal brain tissues (Lucio-Eterovic et al., 2008). The use of HDAC inhibitors (HDACIs) for the treatment of cancer is an area of active investigation. In glioma treatment, HDACIs have been used for the treatment of GBM in combination with RT therapy and chemotherapy (Shi et al., 2016; Ghiaseddin et al., 2018). The anti-tumor mechanism of HDACIs includes blocking cell cycle and promoting cell differentiation, and inducing apoptosis and inhibiting angiogenesis, which can inhibit proliferation and apoptosis of various tumor cells (Marks and Breslow, 2007).

In glioma cells, histone methyltransferases G9a, EZH2, MLL1 and MLL2 (Chang et al., 2009; Cheung et al., 2012; Liu F. et al., 2013; Kondengaden et al., 2016; Wiese et al., 2016; Banasavadi-Siddegowda et al., 2018) regulated the methylation level of lysine located in histone (**Table 3**); these modifications were closely related to gene transcription regulation and genome integration (Heddleston et al., 2012; Zhou et al., 2016). Protein arginine methyltransferase 5 (PRMT5) is another candidate gene for the diagnosis and treatment of glioma, its nuclear expression correlates with poor survival in glioma patients. Banasavadi-Siddegowda et al. (2018) revealed that GBM cells treated with PRMT5 inhibitor mirrored the effects of PRMT5 knockdown, wherein it led to apoptosis of differentiated GBM cells and drove undifferentiated primary patient-derived GBM cells into a non-replicative senescent state.

Inhibiting the activities of histone methyltransferases or HDACs can suppress glioma cell proliferation and induce apoptosis (Sharma et al., 2010; Vargas et al., 2014), suggesting that the inhibitors of these proteins could be candidate drugs for the treatment of glioma. Recently, Ghildiyal and Sen (2017) reported that histone methyltransferase G9a that regulated H3K9 dimethylation has also correlated with the development and progression of glioma, and its inhibitors have also been reported as potential agents for the treatment of glioma (Guo et al., 2016).

It is worth noting that no post-translational modifications processes exist in isolation but rather act with mutual influence and coordination, usually referred to as histone crosstalk. In our earlier research (Zang et al., 2017), we found that simultaneously inhibiting the activity of HDAC and G9a would yield better effects than inhibiting single targets. In our anti-proliferation experiment, multiple cancer phenotypes including leukemia, prostatic carcinoma, hepatocellular carcinoma and pulmonary carcinoma and breast carcinoma were used in this study. Compared to the single target suppression effect, simultaneous inhibition of the activities for two protein targets showed a better anti-proliferation effect in parts of the tumor cell lines, such as breast carcinoma. As illustrated in **Figure 1**, under the cooperation of G9a inhibitor and HDACIs, more sites in histone tails are free to be acetylated. This acetylation status can activate cancer suppression gene transcription and alleviate the disease. Thus, developing high activity HDAC and G9a hybrid inhibitors is another effective route in targeted epigenetic therapy.

The Polycomb group (PcG) protein family is a group of gene regulatory factors that play a role in embryonic development (Zhao et al., 2017). They are divided into two protein complexes based on function, namely PRC1 and PRC2 (Collinson et al., 2016). PRC2 is a multi-protein complex responsible for the methylation of H3 at lysine 27 (H3K27Me). Zeste Gene Enhancer Homolog 2 (EZH2) is a catalytic subunit that constitutes the PRC2 protein complex (Collinson et al., 2016). Current studies (Orzan et al., 2011; Chen et al., 2017) indicated that EZH2 is over expressed in many tumor tissues, including glioma, and is closely related to the malignant progression, invasion and metastasis of the tumors. In the cellular level of research, EZH2 gene silencing technology or using EZH2 inhibitors prevented glioma cell proliferation (Kurmasheva et al., 2017). Therefore, focusing on EZH2 as a new target may pave a new way for the treatment of clinical glioma (Zhang Y. et al., 2017). In pediatric GBM, H3F3A involves two critical singlepoint mutations in the histone tail at lysine (K) 27 (K27M) and glycine (G) 34 (G34R/V) that are both involved with key regulatory post-transcriptional modifications (Schwartzentruber et al., 2012). The discovery of these molecular markers provides a basis for the diagnosis of this type of glioma and


will lay the foundation for further diagnosis and treatment research.

## PRESENT CLINICAL WORKFLOW

DNA methylation by DNMT leads to the gene silencing of tumor suppressor genes (Dammann et al., 2017). The inhibition of DNMTs can achieve reactivation of transcription of these critical genes (Castillo-Aguilera et al., 2017). Thus, the study of DNMT inhibitors has become a new bright spot in the treatment of glioma. Among the DNMT inhibitors that have now entered the clinical trials, 5-aza-2<sup>0</sup> -deoxycytidine (Chu et al., 2013) is the most representative. In tumor cells, 5-aza-2<sup>0</sup> deoxycytidine is blended with DNA in the form of phosphate, and then inhibit DNMT activity, eventually leading to the desired low methylation status to exert antitumor effects (Sun et al., 2011). Currently the clinical research of 5-aza-2 0 -deoxycytidine is limited in leukemia (Roboz et al., 2018) and part of solid tumors (Garrido-Laguna et al., 2013; Fan et al., 2014; lung cancer, etc., not including glioma). For glioma treatment, most of the studies with 5-aza-2<sup>0</sup> -deoxycytidine are still in pre-clinical research stage (Oi et al., 2009; Zhang et al., 2014).

HDACIs can inhibit glioma oncogene transcription and have a variety of effects on cell life cycle. HDACIs can arrest cell division in G1 and G2 phases, induce cell differentiation and apoptosis (Hazane-Puch et al., 2016), destroy the combination of heat shock protein and substrate protein, promote degradation of oncoprotein, and also inhibit the growth and proliferation of glioma by inhibiting tumor angiogenesis (Pei et al., 2016). DNMT inhibitors and HDACIs can be used for the treatment of various tumors, either individually or as a synergistic combination (Xu et al., 2014; Pathania et al., 2016). In glioma, as a new therapeutic drug, HDACIs provide new directions for the treatment of glioma. Already many HDACIs have entered the phase I/II clinical trials (**Table 4**) alone or in combination with other chemotherapeutic agents such as TMZ and radiotherapy for the treatment of various types of glioma, including diffuse intrinsic pontine glioma (DIPG), progressive, or recurrent GBM (Lee et al., 2012; Krauze et al., 2015; Kim et al., 2017).

One phase I study conducted by the Children's Oncology Group (COG) investigated vorinostat with TMZ in relapsed or refractory primary CNS tumors (Hummel et al., 2013). Five-day cycles of vorinostat in combination with TMZ were well tolerated in children with recurrent CNS malignancies, with myelosuppression as the dose-limiting toxicities (DLT). Accumulation of acetylated H3 in peripheral blood mononuclear cells (PBMC) was observed after administration of vorinostat. One phase II trial of vorinostat on recurrent GBM was reported by the North Central Cancer Treatment Group (Galanis et al., 2009). In this study, vorinostat monotherapy was well tolerated in patients with recurrent GBM, and there were obvious increases in acetylation levels of H2B, H3 and H4 after treatment. Microarray RNA analysis showed changes in genes regulated by vorinostat, such as up-regulation of E-cadherin.


Zang et al. Potential Glioma Treatments

With regards to panobinostat, one phase II study of panobinostat in combination with bevacizumab (BEV) was attempted in individuals with recurrent GBM and anaplastic glioma, but part of this study did not meet the criteria for continued accrual and was closed (Lee et al., 2015). Prior to closure, the treatment was reasonably well tolerated in both cohorts, but the addition of panobinostat to BEV did not significantly improve 6-month progression-free survival (PFS6) compared to historical controls of BEV monotherapy in either cohort. More preclinical evidences have shown that panobinostat may act as a radiosensitizer. A phase I trial combining panobinostat with stereotactic re-irradiation in patients with recurrent HGG has been reported (Shi et al., 2016). The results were more promising than panobinostat with BEV, with a PFS6 of 83% in the panobinostat and stereotactic re-irradiation therapy group, compared to 30.4% in the panobinostat with BEV group.

Another phase I study of valproic acid (VPA) in pediatric patients with refractory solid or CNS tumors was conducted by COG (Su et al., 2011). Histone hyperacetylation was observed in half of the patients at steady state. Krauze et al. (2015) recently conducted a phase II study of concurrent radiation therapy, TMZ, and VPA for patients with GBM. The results of this study demonstrated that the addition of VPA to concurrent RT/TMZ in patients with newly diagnosed GBM was well tolerated. VPA may result in improved outcomes compared to historical data and merits further study.

Overall, HDACIs as monotherapy or a combination therapy seem promising in improving prognosis in this difficulty to treat malignancy glioma (**Table 4**). For the possible toxicity that epigenetic drugs may present during treatment, improving the dosage regimen (Issa et al., 2004) or developing new epigenetic therapies or gaining knowledge of how to synchronize them with other treatment modalities are good choices to alleviate this toxic effect. So far, HDAC inhibitors, vorinostat and valproic acid can both combine with TMZ and/or RT to exert good effects in clinical trials to treat children with refractory or recurrent CNS malignancies or adult GBM (Lee et al., 2012; Hummel et al., 2013; Krauze et al., 2015). This is a good trend for our future clinical research. Vorinostat combined with erlotinib or panobinostat with BEV did not show obvious effects (Peereboom et al., 2010; Lee et al., 2015). Though the results are not optimistic, they provide valuable information for future research. In recent years, immunotherapy combined with other drugs has been a hot topic in cancer treatment, but few clinical trials were reported on combining the HDACIs with Gene-Mediated Cytotoxic Immunotherapy (G-MCI). Most G-MCI were preferred to combine with TMZ and standard of care (SOC) radiation after surgery, and survival outcomes were most notably improved in patients with minimal residual disease after gross total resection (Chiocca et al., 2011; Wheeler et al., 2016).

Although there are few clinical trials on the combination of glioma immunotherapy and epigenetics, many basic studies on immunotherapy and epigenetics of glioma were reported recently (Gallagher et al., 2017; Bhat et al., 2018). Studies have shown that tumor cells can use epigenetic mechanisms to alter their autoimmune origin and disrupt the process of recognition between tumor cells and the immune system. By DNA methylation or histone modifications, tumor cells can directly or indirectly down-regulate the expression levels of key molecules in the tumor immunoreaction process, thereby destroying the immune recognition and killing tumor cells (Maio et al., 2015). At present, the immunotherapeutic drugs that can combine with epigenetic drugs mainly include cytokines immunosuppressive agents, polypeptide vaccines, immunological adjuvants and tumor cell vaccine agents. Epigenetic drugs combined with tumor immunotherapy drugs will become an important research direction for the future treatment of tumors, and also provide new ideas for the treatment of glioma. With the development of improved medical standards, we believe more high-quality phase trials in newly diagnosed and recurrent GBM are imperative.

# CONCLUSIONS AND PERSPECTIVES

Glioma is a common primary malignant brain tumor with high recurrence rates, short survival times, high rates of mortality and treatment difficulties. For all grades of glioma patients, the largest range of safe resection remains the central step in current comprehensive treatment strategies (Kreth et al., 2013). Prior treatments of these tumors had taught us that conventional surgeries and chemo-radiotherapy protocols can only minimally improve the quality of life and slightly prolong the survival of some patients. The postoperative treatments, including radiation, chemotherapy, the dose and the cycle, should be implemented after comprehensive evaluation based on factors such as patient's age, operation circumstances, histopathological classification and molecular characteristics (Jiang et al., 2016). Further investigations and reviews of the treatment strategies for malignant glioma are needed.

The genetic instability and heterogeneity of glioma are prominent. The mutual regulation mechanisms of related signal transduction pathways, which are not yet fully understood, are essential for the determination of therapeutic targets and drug development process. In the study of epigenetic therapy, multitarget combined inhibition is a critical concept in targeted drug development. In addition to single drug with multi-target inhibition, the combined use of multiple targeted drugs is also important, including those mentioned in this study (e.g., HDAC and G9a). However, even if the targeted drugs are tested in combination, the number of trials is quite large and pre-clinical pre-screening of drugs is therefore necessary. Whether the targeted drug can act on the expected target site and whether it can effectively inhibit the downstream signaling pathways, the potential toxic and side effects, are related to the safety and efficacy of targeted therapy. These issues need to be studied in depth.

Overall, epigenetic modifications are closely related to glioma proliferation, metastasis, invasion and prognosis. Various epigenetic modifications closely interact to participate in the occurrence and progression of glioma. Breakthroughs in the treatment of glioma require advances in scientific research, improvements in therapeutic technologies and protocols, and the development of diagnosis and treatment around individualized protocols. The epigenetic phenomena of glioma, including DNA methylation, abnormal microRNA, chromatin remodeling and histone modifications, have excellent potential significance and application to the diagnosis, treatment, and prognosis of glioma. Among the four fields, the methylation level of the gene promoter region can be taken as a guide for glioma diagnosis, and also related to the prognosis of the glioma (Mur et al., 2015), miRNAs play many critical roles in the progression of glioma (Zhang C. et al., 2012), the protein levels of enzymes that regulate histone modifications are candidates for the diagnosis of the glioma, and its inhibitors are good candidate drugs for the treatment of glioma (Sharma et al., 2010; Vargas et al., 2014). These phenomena might also help monitor high-risk groups, and assist in tumor risk assessment, judgment of tumor recurrence, prediction of tumor treatment efficacy and prognosis and development of specific new target drugs. It is believed that with the improvement of detection methods and experimental methods, promising results will be achieved in the fields of glioma prevention, diagnosis and treatment. Biotherapy, including gene therapy, immunotherapy and targeted molecular therapy, provides a new hope for the treatment of glioma. Some of these therapies have been shown to be effective in preclinical studies and safe in phase I clinical trials (Maio et al., 2015). Yet, clinical trials in phase II and III have been conducted. Based on our review, except HDACIs, a new therapeutic and epigenetic drug that can be taken alone or combined with other drugs or other treatments, the combination of epigenetic drugs with biotherapy is also

#### REFERENCES


a particularly interesting and novel direction for the future treatment of tumors. Due to the complicated pathogenesis of glioma, epigenetic applications to glioma clinical treatment are still limited. Discovering more effective therapeutic targets, developing novel targeted drugs, improving the efficacy of existing drugs in clinical research, and reducing the side effects of existing drugs are the problems that we need to face and solve in clinical treatment. With continued research, epigenetics understanding is certain to improve and, so is the epigeneticbased treatment of glioma.

#### AUTHOR CONTRIBUTIONS

XH and LZ conceived and designed the project. Each author has contributed significantly to the submitted work. LZ drafted the manuscript. SK, FC and LW revised the manuscript. All authors read and approved the final manuscript.

## FUNDING

This work was supported by the China Postdoctoral Science Foundation (number 2018M632679); Medical and Health Science and Technology Development Project of Shandong Province (number 2017WS639); Linyi Science and Technology Development Project (number 201717024); Linyi People's Hospital Doctoral Research Foundation (number 2016LYBS02); Key Research Project program of Shandong Province (number 2016GSF201056); and the Natural Science Foundation of Shandong Province (number ZR2014HM077).


with fusions of dendritic and glioma cells in patients with glioblastoma. Neurosurgery 81:N11. doi: 10.1093/neuros/nyx263


immunotherapy. Clin. Cancer Res. 21, 4040–4047. doi: 10.1158/1078-0432.ccr-14-2914


through the mitochondrial pathway of apoptosis. Int. J. Oncol. 40, 119–129. doi: 10.3892/ijo.2011.1179


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

The reviewer MC and handling editor declared their shared affiliation at the time of the review.

Copyright © 2018 Zang, Kondengaden, Che, Wang and Heng. 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 Critical Role of Autophagy in Regulating Microglia Polarization in Neurodegeneration

Meng-meng Jin1,2,3† , Fen Wang<sup>4</sup>† , Di Qi<sup>4</sup> , Wen-wen Liu<sup>4</sup> , Chao Gu2,4, Cheng-Jie Mao<sup>2</sup> , Ya-Ping Yang<sup>2</sup> , Zhong Zhao<sup>1</sup> , Li-Fang Hu4,5 \* and Chun-Feng Liu1,2,4 \*

<sup>1</sup> Department of Neurology, Suzhou Clinical Research Center of Neurological Disease, Suzhou Municipal Hospital of Nanjing Medical University, Suzhou, China, <sup>2</sup> Department of Neurology, The Second Affiliated Hospital of Soochow University, Suzhou, China, <sup>3</sup> Department of Neurology, University Hospital Carl Gustav Carus, Technical University of Dresden, Dresden, Germany, <sup>4</sup> Jiangsu Key Laboratory of Neuropsychiatric Diseases, Institute of Neuroscience, Soochow University, Suzhou, China, <sup>5</sup> Department of Pharmacology, College of Pharmaceutical Sciences, Soochow University, Suzhou, China

#### Edited by:

Alberto Javier Ramos, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Carsten Culmsee, Philipps-Universität Marburg, Germany Cristina Pintado, Universidad de Castilla-La Mancha, Spain

#### \*Correspondence:

Li-Fang Hu hulifang@suda.edu.cn Chun-Feng Liu liuchunfeng@suda.edu.cn †These authors have contributed

equally to this work

Received: 06 June 2018 Accepted: 31 October 2018 Published: 20 November 2018

#### Citation:

Jin M-m, Wang F, Qi D, Liu W-w, Gu C, Mao C-J, Yang Y-P, Zhao Z, Hu L-F and Liu C-F (2018) A Critical Role of Autophagy in Regulating Microglia Polarization in Neurodegeneration. Front. Aging Neurosci. 10:378. doi: 10.3389/fnagi.2018.00378 Neuroinflammation and autophagy dysfunction are closely related to the development of neurodegeneration such as Parkinson's disease (PD). However, the role of autophagy in microglia polarization and neuroinflammation is poorly understood. TNF-α, which is highly toxic to dopaminergic neurons, is implicated as a major mediator of neuroinflammation in PD. In this study, we found that TNF-α resulted in an impairment of autophagic flux in microglia. Concomitantly, an increase of M1 marker (iNOS/NO, IL-1β, and IL-6) expression and reduction of M2 marker (Arginase1, Ym1/2, and IL-10) were observed in TNF-α challenged microglia. Upregulation of autophagy via serum deprivation or pharmacologic activators (rapamycin and resveratrol) promoted microglia polarization toward M2 phenotype, as evidenced by suppressed M1 and elevated M2 gene expression, while inhibition of autophagy with 3-MA or Atg5 siRNA consistently aggravated the M1 polarization induced by TNF-α. Moreover, Atg5 knockdown alone was sufficient to trigger microglia activation toward M1 status. More important, TNFα stimulated microglia conditioned medium caused neurotoxicity when added to neuronal cells. The neurotoxicity was further aggravated when Atg5 knockdown in BV2 cells but alleviated when microglia pretreatment with rapamycin. Activation of AKT/mTOR signaling may contribute to the changes of autophagy and inflammation as the AKT specific inhibitor perifosine prevented the increase of LC3II (an autophagic marker) in TNF-α stimulated microglia. Taking together, our results demonstrate that TNF-α inhibits autophagy in microglia through AKT/mTOR signaling pathway, and autophagy enhancement can promote microglia polarization toward M2 phenotype and inflammation resolution.

Keywords: microglia polarization, TNF-α, autophagy, inflammation, neurodegeneration

# INTRODUCTION

Parkinson's disease (PD) is a leading neurodegenerative movement disorder, caused by the demise of dopaminergic neurons in the substantia nigra (SN). Although the molecular mechanisms that underlie dopaminergic neuron losses are still unclear, it is increasingly recognized that microgliamediated neuroinflammation serves as a pathogenic factor in PD (Glass et al., 2010). Accumulation

of proinflammatory cytokines, such as tumor necrosis factorα (TNF-α), interleukin-1β (IL-1β), and interleukin-6 (IL-6) was reported in the brain, cerebrospinal fluid and serum of patients with PD, and animal models as well (Hunot and Hirsch, 2003). The polymorphism of human TNF-α encoding gene increases the risk of developing PD. The plasma TNF-α level correlates with some of the non-motor symptoms in PD including cognition decline, sleep disruption, and depression (Menza et al., 2010). At the molecular level, our previous study identified the autophagy deficits in TNF-α treated dopaminergic cells (Wang et al., 2015). All these suggest that TNF-α is a critical mediator in PD pathogenesis (Mccoy et al., 2011).

The microglia density is remarkably higher in the SN compared to other regions (Kim et al., 2000). Microglia exerts neurotoxic and neuroprotective effects, depending on their diverse functional phenotypes in response to specific stimuli (Orihuela et al., 2016). "Classic M1 polarization" of microglia is characterized by the production of pro-inflammatory mediators. In contrast, alternatively polarized M2 microglia limit inflammation and are typically characterized by the yield of anti-inflammatory factors (Kigerl et al., 2009; Hu et al., 2012). The enhanced microglial M1 activation contributes to neurodegeneration, such as PD and Alzheimer's disease (Wyss-Coray and Rogers, 2012; Vivekanantham et al., 2015). However, the molecular pathway that drives the microglia phenotypic changes during neurodegeneration remains unclear.

Autophagy is a catabolic mechanism which removes unnecessary or dysfunctional proteins and damaged organelles via the lysosome machinery (Mizushima, 2007). Autophagy dysregulation, along with neuroinflammation, is implicated in PD pathogenesis (Plaza-Zabala et al., 2017). The interaction between autophagy and inflammation is very complicated and controversial. For instance, in lipopolysaccharide (LPS) challenged macrophages, Atg16L1 (an autophagy-related protein) deficiency enhanced IL-1β production (Saitoh et al., 2008). Nevertheless, there was also study reported that rapamycin, an autophagy activator, can inhibit M2 macrophage polarization (Wang et al., 2017). However, it remains unclear how autophagy regulates the shifting between M1 and M2 phenotypes in microglia, the macrophage counterpart in the brain. Therefore, our present study was to explore if autophagy has any impact on the microglial polarization upon TNF-α exposure, and the molecular mechanisms involved.

#### MATERIALS AND METHODS

#### Reagents and Antibodies

The chemicals such as 3-methyladenine (3-MA), bafilomycin A1 (BafA1), resveratrol, rapamycin were purchased from Sigma-Aldrich (St. Louis, MO, United States). TNF-α was purchased from PEPROTECH (Rocky Hill, NJ, United States). The Perifosine was purchased from Selleck Chemicals (Houston, TX, United States). The sources for primary antibodies were listed as follows: anti-β-actin, anti-p62, anti-lysosome-associated membrane protein 1 (LAMP1) and anti-β-tubulin from Sigma-Aldrich (St. Louis, MO, United States); anti-LC3, anti-histone 2B and anti-LAMP2 (Abcam, Cambridge, United Kingdom), anti-p-mammalian target of rapamycin (mTOR) [Cell Signaling Technology (CST), 5536s], anti-mTOR (CST, 2983s), anti-p-AKT (CST, 4060), anti-AKT (CST, 4691) and anti-cleaved caspase-3 (CST, 9664); anti-transcription factor EB (TFEB) (Proteintech, Chicago, IL, United States), and anti-GAPDH (Millipore, Billerica, MA, United States). All the reagents for cell culture were bought from Gibco (Grand Island, United States).

#### Cell Culture and Treatment

Murine BV2 microglial cells were cultured in Dulbecco modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin in an incubator with 95% air atmosphere/5% CO<sup>2</sup> at 37◦C. MES23.5 cells were cultured in DMEM/F12 with Sato's components containing 5% heat-inactivated FBS with 1% penicillin/streptomycin in the incubator. Cells were regularly subcultured three times a week and seeded into 35 mm dishes or 24-well plates prior to experimentation.

Primary microglia culture was prepared from 1 to 2-dayold neonatal C57BL/6J mice. Briefly, the cortex was dissected, minced and dissociated in 0.125% Trypsin for 4 min at 37◦C. Trypsin was then neutralized with complete media (DMEM/F12 supplemented with 15% heat-inactivated FBS) and strained through a 200 µm mesh filter. After brief centrifugation, cells were harvested and plated in T75 cell culture flasks. The culture medium was replaced every 3 days. Once the lower layer reached confluence (about 10 days after culture), microglia cells were harvested by mechanical agitation at 180 rpm for 60–90 min and subsequently plated in DMEM/F12 supplemented with 10% FBS at a desired density for further experimentation.

#### Quantitative PCR

Total RNA was extracted using Trizol reagent and reverse transcribed using cDNA synthesis kit (Fermentas, Vilnius, Lithuania). Quantitative PCR (qPCR) was performed as we previously elaborated (Zhou et al., 2014). The primers used were listed in **Table 1**. 18s RNA was used as an internal control. The results were normalized and expressed as ratios of the target gene over 18s mRNA level.

#### Nitrite Quantification

The nitric oxide (NO) production in the culture supernatant was detected using a kit from Beyotime Biotechnology (Shanghai, China). Briefly, 50 µl culture supernatant was collected after treatment and transferred into a 96-well plate. Then, 50 µl Griess reagent was added and incubated at room temperature for 10 min. The absorbance was measured by a Microplate Reader (Tecan M200, Grodig) with a wavelength at 540 nm. The NO level was calculated with a standard curve plotted by sodium nitrite.

#### Cytokine Assay

The content of the cytokines IL-6 and IL-10 in the culture supernatant was detected using the ELISA kits from BOSTER (Wuhan, China), according to the manufacturer's instructions.

#### TABLE 1 | The Primers Used for Quantitative PCR.

fnagi-10-00378 November 16, 2018 Time: 20:48 # 3


experiments. Student t-test. <sup>∗</sup>P < 0.05 vs. untreated controls.

#### Western Blot Analysis

Whole cell lysates were washed prepared by washing cells with phosphate-buffered saline (PBS) and lysed in ice-cold lysis buffer (150 mM NaCl, 25 mM Tris, 5 mM EDTA, 1% Nonidet P-40, pH 7.5) as previously described (Du et al., 2014). Lysates were separated by 10% sodium dodecyl sulfate-polyacrylamide gels and transferred onto nitrocellulose membranes. Next, membranes were blocked in 5% (w/v) non-fat dry milk powder in 0.1% Tris-buffered saline/Tween 20 (TBST) for 1 h and then incubated with the primary antibodies at optimized dilutions at 4 ◦C overnight. After that, membranes were washed with TBST and incubated with HRP-conjugated secondary antibodies for another 1 h. Blots were finally visualized with chemiluminescence (Thermo Company, West Chester, PA, United States) and band densities were analyzed with ImageJ software (National Institute of Health, United States).

#### RFP-GFP-Tandem Fluorescent-Tagged LC3 (tf-LC3) and Atg5 siRNA Transfection

For transfection, 0.8 µg RFP-GFP-LC3 cDNA (Addgene, United States) were transfected into BV2 cells using 2 µl Lipofectamine 2000 (Invitrogen, Eugene, OR, United States). To monitor autophagic flux, cells were exposed to TNFα or vehicle with or without the lysosome inhibitor BafA1 at 24 h post-transfection with tf-LC3 plasmids. The yellow and red LC3 puncta were manually counted, and at least 30 cells per group were randomly selected for counting. The small-interfering RNAs (siRNA) targeting mouse Atg5 (50GACGUUGGUAACUGACAAATT3<sup>0</sup> and 5 <sup>0</sup>UUUGUCAGUUACCAACGUCTT3<sup>0</sup> ) and scrambled siRNA duplexes were synthesized by GenePharma (Shanghai, China), and were transfected using Lipofectamine RNAiMAX (Thermo Company). The knockdown efficiency was determined by western blotting at 24 h post-transfection.

#### Subcellular Fractionation

Cells were lysed in the fractionation buffer including 3 mM CaCl2, 2 mM MgAc, 320 mM sucrose, 0.1 mM EDTA, 1 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), and 0.5% NP-40, kept on ice for 20 min and then centrifuged at 600 g 4 ◦C for 15 min. The supernatants were collected as the cytosolic fraction. The pellets were washed with fractionation buffer without NP-40 twice, lysed with the nuclear lysis buffer comprising 280 mM KCl, 0.2 mM EDTA, 1 mM DTT, 0.5 mM PMSF, 20 mM Hepes (pH 7.9), 25% glycerol, 1.5 mM MgCl2, and 0.3% NP-40 and then centrifuged. The resulting supernatant was then used for the nuclear fraction.

#### Conditioned Medium-Induced Neuronal Cell Death Assay

BV2 cells were transfected with Atg5 siRNA or control siRNA for 48 h as mentioned above. For rapamycin group, cells were pretreated with 0.2 µg/ml rapamycin for 0.5 h. Cells were then treated with TNF-α for 24 h, and replaced by culture in fresh medium for another 24 h to produce the microglia conditioned medium (CM). After that, the microglia CM was transferred into MES23.5 cells and cultured for 24 h. Thereafter, cells were incubated with Hochest 33342 (Sigma) and propidium iodide (PI, Beyotime, Shanghai, China) for 5 min, and then washed with PBS. Images were taken using an inverted IX71 microscope system (Olympus, Tokyo, Japan). The hochest and PI stained cells were counted manually, and at least 500 cells per group were counted.

#### Statistical Analysis

All results are presented as mean ± SEM, obtained from a minimum of 3 independent experiments unless otherwise stated. Statistical significance was analyzed using Student t-test for twogroup comparison or one-way analysis of variance (ANOVA) followed by Tukey post hoc analysis using the GraphPad Prism. The significance level was set at P < 0.05.

FIGURE 2 | Autophagy stimulation enhances M2 but reduces M1 markers in primary microglia. (A–F) qPCR analysis of M1 and M2 genes mRNA levels in TNF-α-treated primary microglia with or without autophagy stimulation by different methods. (A,B) 0.2 µg/ml rapamycin pretreatment. (C,D) serum deprivation. (E,F) autophagy stimulation by 50 µM resveratrol pretreatment. Graphs show relative mRNA levels after normalization with those of corresponding housekeeping gene 18 s, n = 4. (G–I) Cytokines including NO (G), IL-6 (H) and IL-10 levels (I), measured by Griess reagent and ELISA, in cell-free culture supernatant from TNF-α-challenged primary microglia with or without rapamycin pretreatment. N = 4. All of the results are expressed as mean ± SEM, one-way ANOVA followed by Tukey analysis. <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 vs. controls. NS, not significant.

#### RESULTS

#### TNF-α Enhances M1 but Reduces M2 Markers in Microglia

Microglia polarization is commonly characterized by the expression of signature genes that are associated with M1 or M2 phenotype (Mantovani et al., 2013). Neuroinflammation is featured by an alteration of microglia polarization toward M1 status. In this study, we observed that 5 ng/ml TNF-α stimulation not only induced an elevation in M1 genes (iNOS and IL-1β) expression (**Figure 1A**), but also caused a reduction in the M2 signature genes level (Arginase1 and Ym1/2, **Figure 1B**) in mouse primary microglia, indicating a shift toward M1 phenotype following TNF-α stimulation.

# Autophagy Activation Enhances M2 but Reduces M1 Markers in Microglia

In order to assess the role of autophagy in microglia polarization, we induced autophagy activation in primary microglia using serum deprivation or two structure-unrelated autophagy inducers (rapamycin and resveratrol). Although none of the three tested autophagy inducers produced any obvious effect on the basal level of M1 genes, they increased the Arginase1 mRNA levels under normal conditions. A mild but not significant elevation of Ym1/2 mRNA was also observed. Importantly, both serum deprivation and pharmacologic autophagy inducers (rapamycin and resveratrol) potently suppressed the increase of M1 (iNOS and IL-1β) mRNA level and alleviated the decline of M2 gene expression (Arginase1 and Ym1/2) in TNF-α-stimulated

one-way ANOVA followed by Tukey analysis.

primary microglia (**Figures 2A–F**). Measurement of the cytokines level in the culture supernatant displayed that the autophagy activator rapamycin reduced the NO (**Figure 2G**) and IL-6 (**Figure 2H**) production but enhanced IL-10 yield (**Figure 2I**) in TNF-α-stimulated primary microglia. These data imply that autophagy activation is able to promote microglia polarization toward M2 phenotype under both basal and inflamed status.

#### Autophagy Inhibition Enhances M1 but Reduces M2 Markers in Microglia

To evaluate the role of autophagy inhibition in microglia phenotype transition, we used both the autophagy inhibitor 3-MA and the siRNA against autophagy-related protein Atg5 and observed the changes of polarization markers in TNFα-challenged microglia cells. The autophagy inhibitor 3-MA enhanced the basal and TNF-α-induced iNOS and IL-1β transcription (**Figure 3A**). Meanwhile, it downregulated the basal level of Arginase1 and Ym1/2 and caused a further decline of these two M2 markers in TNF-α-exposed primary microglia (**Figure 3B**). In line with this, Atg5 knockdown in BV2 cells was also found to increase the basal level of M1 gene expression (iNOS, IL-1β) and reduce the M2 gene expression (Arginase1, Ym1/2) (**Figures 3C–D**), and reinforced the alterations of the polarization markers in TNF-α-stimulated cells. These observations were confirmed by cytokines assay. As shown in **Figures 3E–J**, autophagy inhibition with 3- MA or Atg5 siRNA consistently caused a further increase of NO and IL-6 generation and a moderate decrease of IL-10 in the supernatant of TNF-α-treated microglia. Western blot study revealed the successful knockdown of Atg5 by small RNA interference in TNF-α-treated BV2 cells (**Figure 3K**). These data indicate that autophagy not only negatively regulates microglia activation under basal conditions but also affects microglia polarization in response to inflammatory stimulation.

(Continued)

#### FIGURE 5 | Continued

fnagi-10-00378 November 16, 2018 Time: 20:48 # 8

and then added with BafA1(50 nM) for 2 h before subjected to western blotting. Actin served as loading controls in panels A,B. N = 4. (C,D) Autophagic assay in microglia treated with TNF-α, BafA1, or in combination. BV2 cells were transfected with RFP-GFP-tandem fluorescent LC3 cDNA for 24 h before treatment. Confocal microscope pictures showing yellow (GFP and RFP overlap) and red LC3 puncta formation in different groups. Scale bar, 10 µm. LC3 dots were visualized and quantified from at least 30 cells per group. (E) Effect of TNF-α on lysosomal biogenesis. BV2 cells were treated with 5 ng/ml TNF-α for 3, 12, or 24 h. The cytosolic and nuclear fractions were subjected to western blotting analysis of TFEB, with GAPDH and Histone 2B as the cytosolic and nuclear loading controls, respectively. (F,G) Effect of TNF-α on lysosomal protein LAMP1 (F) and LAMP2 (G) levels in BV2 cells. N = 3. <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 vs. controls. NS, not significant. One way ANOVA followed by Tukey analysis.

### TNF-α Disrupts Microglia Autophagic Flux

To examine the effect of TNF-α on microglia autophagy, we study several autophagy-related markers including LC3 and p62 (also named sequestosome1, SQSTM1) protein expressions. Western blotting analysis showed that TNF-α treatment resulted in an increase of microglia LC3II and p62 in a time- and dose-dependent manner (**Figures 4A–B**). As several studies previously reported the p62 transcription by inflammatory stimuli (Fujita et al., 2011), here we also examined p62 mRNA level. qPCR analysis showed a substantial increase of p62 transcription in response to TNF-α stimulation (**Figure 4C**).

To determine whether the LC3II increase resulted from autophagy induction or lysosomal degradation impairment, we studied the effect of TNF-α on LC3II in the presence of BafA1, a vacuolar H+-ATPase inhibitor that blocks autolysosome degradation. As shown in **Figure 5A**, showed that 50 nM BafA1 was sufficient to fully block of lysosomal degradation as 100 nM BafA1 did not result in a further increase of LC3II level in BV2 cells. Therefore, 50 nM BafA1 was used in the following study for lysosomal inhibition. TNFα failed to further enhance the LC3II protein level in the presence of BafA1 (**Figure 5B**), indicating that the TNFα-induced LC3II increase may result from lysosome degradation impairment. This is consistent with the tf-LC3 assay which allows us to monitor the autophagic flux (Mizushima et al., 2010). As shown in **Figure 5C**, tf-LC3 transfected BV2 microglia had some basal level of autophagy, as exhibited by a few green/red dots staining in control. Similar to the impact of BafA1, TNF-α caused an obvious elevation in the number of yellow dots, with almost no detectable red dots in cells. TNF-α plus BafA1 co-treatment did not show any further impact on the yellow/red dots when compared to BafA1 treatment alone (**Figures 5C,D**). These data suggest the disruption of autophagic flux in TNF-α challenged microglia.

Anti-transcription factor EB served as a master regulator of the autophagy-lysosome system which coordinates autophagy and lysosomal biogenesis (Sardiello et al., 2009). To clarify the role of TFEB in TNF-α treated microglia, we studied and found that a time-dependent TFEB translocation from cytosol to nuclear fraction in response to TNF-α compared with untreated BV2 cells (**Figure 5E**). Moreover, the lysosome proteins LAMP1 and LAMP2 expression also increased after TNF-α treatment (**Figures 5F,G**).

#### AKT Signaling Is Involved in TNF-α Induced Autophagy Flux Disruption in Microglia

The signaling mechanism that underlies TNF-α induced autophagy flux disruption was also examined. We observed a rapid activation of AKT in BV2 cells following 5 ng/ml TNFα treatment, as AKT phosphorylation (Ser473) increased and reached its peak at 6 h after treatment (**Figure 6A**). This was accompanied by a consistent elevation in the phosphorylation of mTOR, the master regulator of autophagy and downstream target of AKT (**Figure 6B**). As expected, treatment with the AKT-specific inhibitor perifosine (100 nM) caused a LC3II increase, correlated with a reduction in Akt phosphorylation in BV2 cells. Of note, TNF-α did not further enhance the LC3II protein level in the presence of perifosine compared with perifosine treatment group (**Figures 6C,D**). This implies that the AKT/mTOR signaling pathway was involved in the autophagy regulation in TNF-α-stimulated microglia.

### TNF-α-Challenged Microglia Conditioned Medium Produced Toxicity to Co-cultured Neuronal Cells

In order to evaluate the effect of microglia polarization on neurons around, we established a system in which the cell-free culture supernatant from TNF-α or vehicle challenged BV2 cells (defined as a microglia conditioned medium) was harvested and then transferred into MES23.5 cells, a dopaminergic cell line. To assess the CM-associated impact and avoid the carryover from microglia treatment, BV2 cells were challenged with TNF-α for 24 h, washed, and cultured with fresh medium for another 24 h. After that, the microglia culture supernatant was collected as CM and applied in subsequent experimentation. To determine the influence of microglia autophagy regulation on neurons survival, BV2 cells were transfected with Atg5 siRNA or pretreated with rapamycin before exposure to TNF-α. The hochest and PI staining showed that treatment with TNF-α challenged CM for 24 h resulted in a 13.1 ± 1.2% cell death in MES23.5 cells compared to control CM group (2.2 ± 0.8%), implying TNF-α challenged CM produced toxicity to neuronal cells. This neurotoxicity was enhanced by Atg5 knockdown but attenuated when BV2 cells pretreatment with rapamycin (**Figures 7A,B**). The results were confirmed with the caspase-3 assay. Western blotting showed that treatment with TNF-α challenged CM caused an increase of cleaved caspase-3 level in MES23.5 cells. Transfection with Atg5 siRNA in BV2 microglia led to a further increase in cleaved caspase-3 compared to control siRNA. In

∗∗∗P < 0.001 vs. controls. One way ANOVA followed by Tukey analysis. NS, not significant.

contrast, microglia pretreatment with rapamycin attenuated the cleaved caspase-3 increase (**Figure 7C**). These results indicate that modulation of microglia autophagy could have an impact on the neuronal survival probably by affecting microglia polarization and associated neuroinflammation.

#### DISCUSSION

In this study we provide the evidence that pro-inflammatory cytokine TNF-α inhibits autophagy flux and drives microglia shift toward M1 phenotype via activating AKT/mTOR signaling. Our findings demonstrate that autophagy enhancement could guide microglia activation toward M2 status and attenuate the neurotoxicity of TNF-α challenged microglia CM, while microglia autophagy inhibition produces the opposite effects (as summarized in **Figure 8**).

Neuroinflammation and autophagy dysregulation closely linked with neurodegeneration. Recent studies identify a complex interaction between these two processes. Here we clearly show that TNF-α inhibits microglia autophagy flux, and reveal a role of autophagy flux inhibition in M1 microglia polarization and thereby the neurotoxicity. This is consistent with recent reports that autophagy may regulate LPS-stimulated inflammation in

microglia and its associated neurotoxicity (Bussi et al., 2017; He et al., 2018). These lines of evidence highlight a possibility that autophagy serves as the common machinery in regulating microglia-mediated inflammation in response to a multitude of inflammatory stimuli. Yet, another study demonstrated that LPS injection induced autophagy activation in rat hippocampus (Pintado et al., 2017). Autophagy is a catabolic process that can be activated during starvation, stress and nutrient deprivation by different signaling pathways. It is conceivable that microglia autophagy could be altered by diverse conditions. Also, the isolation and culture conditions could yield an impact on microglia in vitro. Thus, the in vivo study, in combination with microglia-specific genetic manipulation, may further our understanding of the relationship between inflammation and autophagy.

We previously reported that TNF-α can disrupt the autophagic flux and result in α-synuclein accumulation in PC12 cells and midbrain neurons (Wang et al., 2015). This and the present study consistently show TNF-α causes autophagy dysfunction in both neuron and microglia, which may produce a positive loop between microglia and neurons and thus lead to neurodegeneration. However, we did not observe any change in TFEB or LAMP1 expression in TNF-α-treated neurons (Wang et al., 2015). By contrast, in this study we detected TFEB nuclear translocation induced by TNF-α in microglia, correlating with the increases of LAMP1 and LAMP2 expression. This may act as a compensatory mechanism to autophagy flux disruption during TNF-α stimulation. This discrepancy may be attributed to the higher activity of lysosomal biogenesis in microglia. The TNF-α induced autophagy flux inhibition seemed to be dependent, at

least in part, on AKT activation. An early activation of AKT, accompanied by mTOR activation, was observed in microglia after TNF-α stimulation. In the presence of the AKT inhibitor, TNF-α failed to alter the autophagic marker LC3II. Therefore, activation of AKT/mTOR signaling may contribute to TNFα-induced autophagy inhibition in microglia. A recent study also identifies the role of AKT activation in regulating autophagy during the late stage (Matsuda-Lennikov et al., 2014). Phafin2 (EAPF or PLEKHF2), a lysosomal protein interacts with and recruits cytosolic Akt to the lysosome, and thus affects the autophagy-lysosome pathway.

The identification of different phenotypes is a breakthrough in microglia biology. M1 or M2 polarization reflects different microglia functions. Generally, M1 is pro-inflammatory and neurotoxic while M2 is anti-inflammatory and neuroprotective (Moehle and West, 2015). This opens up a novel avenue for the treatment of neuroinflammation related disorders. Notably, M1 or M2 markers are not exclusively expressed in microglia. Their phenotypic changes during neurodegeneration may be driven by extracellular and intracellular factors. The extracellular triggers were extensively reported, such as betaamyloid and LPS (Qin et al., 2007; Hickman et al., 2008). Only a few molecules including AMPK, hydrogen sulfide, and A1 adenosine receptor, have been reported to modulate microglia polarization (Tsutsui et al., 2004; Zhou et al., 2014; Xu et al., 2015). Hence, the intracellular machinery warrants intensive study. Our study reveals autophagy as an endogenous "brake" on microglia phenotype shift toward M1 polarization. Atg5 (an autophagy-related protein) knockdown was sufficient to trigger M1 microglia activation in the absence of inflammatory stimulation. In TNF-α-challenged microglia, autophagic flux was impaired, associated with the increase of M1 marker (iNOS/NO, IL-1β, and IL-6) and the decrease in M2 marker (Arginase1, Ym1/2, and IL-10) expression. Upregulation of autophagy with serum deprivation or with pharmacologic activators (rapamycin and resveratrol) promoted microglia polarization toward M2 phenotype while inhibition of autophagy with 3-MA or Atg5 siRNA consistently aggravated the M1 polarization induced by TNF-α. These findings suggest that autophagy acts as an endogenous "brake" on microglia shift toward M1 in normal situation. In disease conditions, microglia autophagy deficiency may sensitize the cells to simulation and boost neuroinflammation.

Persistent microglia activation toward M1 phenotype could have physiologic relevance in brain regions with high microglia density such as SN (Pintado et al., 2011; Leal et al., 2013). This may contribute to dopaminergic neurodegeneration in PD. In support of this, here we found TNF-α-challenged microglia CM produced toxicity to the cocultured MES23.5 cells. The neurotoxicity was boosted if inhibiting microglia autophagy, but alleviated when microglia autophagy was activated. This CM-derived toxicity may derive from an accumulation of neurotoxic and pro-inflammatory molecules in the challenged CM. Consistently, the cytokines assay (**Figures 2**, **3**) demonstrated a role of autophagy in modulating cytokines generation in microglia. Sure, other untested factors may also contribute to the toxicity.

In sum, our study demonstrates that TNF-α inhibits autophagy flux in microglia through activating AKT/mTOR signaling pathway, and identifies that autophagy serves as a negative regulator of microglia polarization in both normal and inflamed conditions. This study also highlights the potential of autophagy inducers in the treatment of neuroinflammationassociated degeneration.

#### AUTHOR CONTRIBUTIONS

fnagi-10-00378 November 16, 2018 Time: 20:48 # 12

C-FL and L-FH designed this study. M-mJ, FW, and W-wL carried out all the experiments. CG, DQ, ZZ, C-JM, and Y-PY

#### REFERENCES


revised the manuscript. L-FH and M-mJ contributed to writing the manuscript. All authors read and approved the final version of the manuscript.

#### FUNDING

This work was supported by grants from the National Natural Science Foundation of China (81571233 and 81671250), Jiangsu Provincial Medical Key Discipline Project (ZDXKB2016022), Jiangsu Key Laboratory of Neuropsychiatric Diseases (BM2013003), Suzhou Clinical Research Center of Neurological Disease (Szzx201503), Jiangsu Province's Young Medical Talents Program (QNRC2016872), and also partly supported by Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).

step in the induction of autophagy. PLoS One 9:e79795. doi: 10.1371/journal. pone.0079795


and tissue repair. Int. J. Neurosci. 125, 717–725. doi: 10.3109/00207454.2014. 982795


**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 Jin, Wang, Qi, Liu, Gu, Mao, Yang, Zhao, Hu and Liu. 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.

# Parkinsonian Neurotoxins Impair the Pro-inflammatory Response of Glial Cells

Neus Rabaneda-Lombarte1,2 , Efren Xicoy-Espaulella<sup>1</sup> , Joan Serratosa<sup>1</sup> , Josep Saura<sup>2</sup> and Carme Solà<sup>1</sup> \*

<sup>1</sup>Department of Brain Ischemia and Neurodegeneration, Institut d'Investigacions Biomèdiques de Barcelona (IIBB)—Consejo Superior de Investigaciones Científicas (CSIC), Institut d'Investigacions Biomèdiques August-Pi i Sunyer (IDIBAPS), Barcelona, Spain, <sup>2</sup>Biochemistry and Molecular Biology Unit, School of Medicine, Institut d'Investigacions Biomèdiques August-Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain

In the case of Parkinson's disease (PD), epidemiological studies have reported that pesticide exposure is a risk factor for its pathology. It has been suggested that some chemical agents, such as rotenone and paraquat, that inhibit the mitochondrial respiratory chain (in the same way as the PD mimetic toxin 1-methyl-4-phenylpyridinium, MPP+) are involved in the development of PD. However, although the neurotoxic effect of such compounds has been widely reported using in vivo and in vitro experimental approaches, their direct effect on the glial cells remains poorly characterized. In addition, the extent to which these toxins interfere with the immune response of the glial cells, is also underexplored. We used mouse primary mixed glial and microglial cultures to study the effect of MPP+ and rotenone on glial activation, in the absence and the presence of a pro-inflammatory stimulus (lipopolysaccharide plus interferon-γ, LPS+IFN-γ). We determined the mRNA expression of the effector molecules that participate in the inflammatory response (pro-inflammatory cytokines and enzymes), as well as the nitric oxide (NO) and cytokine production. We also studied the phagocytic activity of the microglial cells. In addition, we evaluated the metabolic changes associated with the observed effects, through the measurement of adenosine triphosphate (ATP) production and the expression of genes involved in the control of metabolic pathways. We observed that exposure of the glial cultures to the neurotoxins, especially rotenone, impaired the pro-inflammatory response induced by LPS/IFN-γ. MPP+ and rotenone also impaired the phagocytic activity of the microglial cells, and this effect was potentiated in the presence of LPS/IFN-γ. The deficit in ATP production that was detected, mainly in MPP+ and rotenone-treated mixed glial cultures, may be responsible for the effects observed. These results show that the response of glial cells to a pro-inflammatory challenge is altered in the presence of toxins inhibiting mitochondrial respiratory chain

**Abbreviations:** ATP, adenosine triphosphate; BSA, bovine serum albumin; (COX-2), cyclooxygenase-2; Carkl, carbohydrate kinase-like protein; CNS, central nervous system; Glut, glucose transporter; gp91phox, NADPH oxidase, catalytic subunit; Hif, hypoxia-inducible factor; Hk, hexokinase; IL, interleukin; IFN-γ, interferon-γ; iNOS, inducible nitric oxide synthase; LPS, lipopolysaccharide; MPP+, 1-methyl-4-phenylpyridinium; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; NO, nitric oxide; PBS, Phosphate-buffered saline; PCR, polymerase chain reaction; PFK-P, phosphofructokinase; PI, propidium iodide; PD, Parkinson's disease; qRT-PCR, quantitative real time PCR; Rot, rotenone; TNF, tumor necrosis factor.

#### Edited by:

Yolanda Diz-Chaves, University of Vigo, Spain

Reviewed by: Tarja Maarit Malm, University of Eastern Finland, Finland Hong Qing, Beijing Institute of Technology, China

> \*Correspondence: Carme Solà carme.sola@iibb.csic.es

Received: 02 October 2018 Accepted: 07 December 2018 Published: 10 January 2019

#### Citation:

Rabaneda-Lombarte N, Xicoy-Espaulella E, Serratosa J, Saura J and Solà C (2019) Parkinsonian Neurotoxins Impair the Pro-inflammatory Response of Glial Cells. Front. Mol. Neurosci. 11:479. doi: 10.3389/fnmol.2018.00479 activity, suggesting that the glial immune response is impaired by such agents. This may have relevant consequences for brain function and the central nervous system's (CNS's) response to insults.

Keywords: glial activation, mixed glia, microglia, immune response, MPP+, rotenone, glial metabolism, Parkinson's disease

## INTRODUCTION

Microglia are the main endogenous immune cells of the central nervous system (CNS). Under physiological conditions, they constantly patrol the CNS parenchyma, ready to detect alterations that could interfere with normal brain function. In response to noxious stimuli, microglial cells develop a wide range of reactive phenotypes aimed at re-establishing cerebral homeostasis and minimizing neuronal damage. In this way, they can respond to alterations in the CNS homeostasis due to the presence of exogenous pathogens or anomalous protein aggregates resulting from pathological processes. In addition, they are also able to respond to neuronal damage resulting from brain lesions, brain ischemia, neurodegenerative diseases or exposure to neurotoxic agents (Salter and Stevens, 2017).

In fact, exposure to neurotoxic agents, such as pesticides, may contribute to the development of some neurodegenerative diseases (Mostafalou and Abdollahi, 2013). In 1982, the accidental exposure to 1-methyl-4 phenyl-1,2,3,6-tetrahydropyridine (MPTP) in drug abusers caused parkinsonism (Langston et al., 1983). Epidemiological studies show that exposure to the pesticides rotenone and paraquat, which are functional and structural analogs of MPTP respectively, as well as to other pesticides, is a risk factor for Parkinson's disease (PD; Tanner et al., 2011; Goldman, 2013; Kamel, 2013). MPTP and its analogs are inhibitors of the mitochondrial respiratory chain and it has been suggested that mitochondrial dysfunction is involved in the induction of oxidative damage in dopaminergic neurons in parkinsonism (Dauer and Przedborski, 2003; Goldman, 2013). Due to the particular sensitivity of dopaminergic neurons to the effect of these neurotoxins, experimental models of PD have been developed by exposing neuronal cell cultures or laboratory animals to these agents (Bové and Perier, 2012). These experimental models are useful for studying mechanisms of dopaminergic neuronal cell degeneration and testing potential therapeutic approaches. However, although the toxic effect of MPTP [or its active metabolite 1-methyl-4-phenylpyridinium (MPP+)] and rotenone on dopaminergic neurons has been widely described using both in vivo and in vitro approaches, reports of their direct effect on glial cells are scarce. In addition, there is some controversy regarding the results already obtained. Either no direct effect of MPP+ on microglial cell function (Gao et al., 2003; Jin et al., 2012), or an increase in the expression of pro-inflammatory markers in microglial cells after MPP+ exposure (Bournival et al., 2012; Chen et al., 2015) has been reported. Similarly, either an increase in the expression of pro-inflammatory factors (Gao et al., 2013; Yuan et al., 2013; Du et al., 2014) or no direct effect on the production of inflammatory factors (Klintworth et al., 2009) has been observed in rotenone-treated microglial cell cultures.

Since reciprocal communication exists in the CNS between neuronal and glial cells, alterations in neuronal function may affect glial function and vice versa. In fact, a possible role of glial activation in the development of neuronal damage in neurodegenerative diseases has been repeatedly proposed (Perry et al., 2010; Colonna and Butovsky, 2017). In addition, communication also exists between glial cells, and alterations in a given cell type may affect the function of other glial cell types. Consequently, alterations in glial function due to exposure to neurotoxic compounds merit study, especially in the context of neurodegenerative diseases in which such exposure is considered a risk factor. The aim of this study was therefore to characterize the effects of MPP+ and rotenone on glial activation using primary mixed glial cultures, (mainly composed of astrocytes and microglia) and microglial cultures. We determined the direct effect of these neurotoxins on glial cell function, and also whether they could interfere with glial activation induced by a classical pro-inflammatory stimulus such as lipopolysaccharide (LPS)/interferon-γ (IFN-γ). We observed that MPP+ and rotenone did not induce the expression of pro-inflammatory markers by glial cells per se. However, the LPS/IFN-γ-induced pro-inflammatory response was modified in glial cultures in the presence of MPP+ and rotenone. These neurotoxins induced modifications in the mRNA expression of pro-inflammatory genes and phagocytic activity. Alterations in adenosine triphosphate (ATP) production could account for the effects observed. These results show that insults affecting the metabolic activity of glial cells, result in an altered immune response, which may have relevant consequences for normal brain function and the CNS response to insults.

#### MATERIALS AND METHODS

Experiments were carried out in accordance with European Union directives (86/609/EU) and Spanish regulations (BOE 67/8509-12, 1988) on the use of laboratory animals, and were approved by the Ethics and Scientific Committees of the University of Barcelona and CSIC.

#### Cell Cultures

Primary mixed glial cultures were prepared from the cerebral cortex of 1–3-day old C57Bl/6 mice as previously described (Gresa-Arribas et al., 2010). The culture medium used was the Dulbecco's modified Eagle medium-F12 nutrient mixture (GIBCO) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Invitrogen, Molecular Probes, Eugene, OR, USA), 20 U/mL penicillin-20 µg/mL streptomycin (Invitrogen), and 0.5 µg/mL amphotericin B (Fungizoner, Invitrogen). The cells were seeded at a density of 3.5 × 10<sup>5</sup> cells/mL (100 µL, 300 µL and 2.5 mL per well into 96-, 48- and 6-well culture plates) and cultured at 37◦C in a humidified 5% CO<sup>2</sup> atmosphere. The medium was replaced once a week. The cultures were used at 21 DIV.

Primary microglia enriched cultures were obtained from 21 DIV mixed glial cultures using the mild trypsinization method as previously described (Saura et al., 2003). Microglia enriched cultures were used 24 h after isolation by this procedure.

# Cell Culture Treatments

LPS and IFN-γ treatment: Cells were treated with 100 ng/ml LPS (E. coli 026:B6, Sigma-Aldrich, St. Louis, MO, USA) and 0.1 ng/ml IFN-γ (Sigma-Aldrich) for 6 h or 24 h. Stock solutions of 1 mg/mL LPS in a serum-free culture medium and 10 µg/mL IFN-γ in a serum-containing culture medium, were prepared and stored at −20◦C.

MPP+ and rotenone treatment: Cells were treated with 10, 25, 50 and 100 µM MPP+ or 20, 40, 100 and 150 nM rotenone (both from Sigma-Aldrich) for 6 h or 24 h, in the absence or presence of LPS/IFN-γ. Stock solutions of 50 mM MPP+ in milliQ H2O and 10 mM rotenone in DMSO were freshly prepared on the day of treatment. DMSO in the cell cultures was always below 1/1,000.

Treatments were added directly to the culture medium.

# Nitric Oxide Production

Nitric oxide (NO) production was estimated from the nitrite accumulation in the culture supernatant using the colorimetric Griess reaction. Briefly, the culture supernatant from the glial cells seeded into 96-well culture plates, was collected 24 and 48 h after treatments and stored at −20◦C until used. Fifty microliter aliquots of the culture supernatant were incubated with equal volumes of the Griess reagent for 10 min at 20–25◦C. Optical density at 540 nm was measured using a microplate reader (Multiskan Spectrum, Thermo Fisher Scientific, Vantaa, Finland). Nitrite concentration was determined from a sodium nitrate standard curve.

#### Cell Viability Measurements

Glial cells seeded into 96-well culture plates were used to estimate cell viability from the metabolic activity by a 3-(4,5 dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) colorimetric assay, 24 h after treatments. Briefly, MTT (Sigma-Aldrich) was added to the cell cultures to reach a final concentration of 1 mg/mL. After incubation for 30 min (mixed glial cultures) or 90 min (microglial cultures) at 37◦C, the medium was removed and 200 µL of DMSO was added to each well. The optical density of the resulting blue formazan was measured at 570 nm using a microplate reader (Multiskan Spectrum, Thermo Fisher Scientific). Readings were taken at 650 nm to obtain background levels. Results were expressed as percentages of the control.

Propidium iodide (PI) and Hoechst labeling were performed to corroborate data obtained in the MTT assay. Briefly, cells were incubated with PI (7.5 µg/ml, Molecular Probes, Eugene, OR, USA) and Hoechst 33342 (3 µg/ml, Molecular Probes) for 10 min. Microscopy images were obtained using an Olympus IX70 microscope (Olympus, Okoya, Japan) and a digital camera (CC-12, Olympus Soft Imaging Solutions GmbH, Hamburg, Germany). The extent of cell death was calculated from the ratio between PI positive nuclei, corresponding to dead cells, vs. Hoechst positive total nuclei.

## RNA Extraction and Quantitative Real Time PCR

Glial cells seeded into six-well culture plates were used (one or two wells per experimental condition for mixed glia and microglia, respectively) to assess the mRNA expression of pro-inflammatory markers by quantitative real time polymerase chain reaction (PCR) 6 h after treatments. A High Pure RNA Isolation Kit (Roche Diagnostics Scheiwz AG, Rotkreuz, Switzerland) was used to isolate the total RNA from the mixed glial cultures. A PureLink RNA micro kit (Invitrogen) was used to isolate the total RNA from the primary microglial cultures. The RNA (0.5–1 µg) was reverse transcribed with random primers using the Transcriptor Reverse Transcriptase Kit (Roche Diagnostics). Three nanograms of cDNA were used to perform quantitative real time PCR (qRT-PCR) with the IQ SYBRGREEN SuperMix (Bio-Rad Laboratories, Hercules, CA, USA) using an iCycler MyIQ apparatus (Bio-Rad Laboratories) as previously described (Dentesano et al., 2014). The primers used (Integrated DNA Technology, IDT, Skokie, IL, USA) are shown in **Table 1**. Samples were run for 40 cycles (95◦C for 15 s, 60◦C for 30 s, and 72◦C for 15 s). The amplification specificity was confirmed by the analysis of melting curves. Relative gene expression values were calculated using the ∆∆Ct method (Livak and Schmittgen, 2001). β-Actin and 18S ribosomal RNA (Rn18s) were used as the reference genes.

# ELISAs

The interleukin (IL) 1β, IL6 and tumor necrosis factor α (TNFα) release in the culture supernatant was determined using ELISA kits specific for each cytokine (mouse IL1-β ELISA Ready-SET-GO!, mouse IL6 ELISA Ready-SET-GO! and mouse TNFα ELISA Ready-SET-GO!, eBioscience-Affimetrix, Inc., San Diego, CA, USA), following the manufacturer's instructions. The culture supernatant from 48-well culture plates was collected 24 h after treatments and stored at −80◦C until use. IL1β, IL6 and TNFα concentrations were determined from the standard curves.

#### Phagocytosis Assay

The phagocytic activity of the microglial cells was assessed 24 h after treatments. Briefly, the microglial cell cultures in 48-well plates were incubated for 1 h at 37◦C with fluorescent latex beads (FluoSpheres, carboxylate-modified microspheres, 2.0 µm, red fluorescent (580/605), 2% solids; Thermofisher Scientific; 1/1,000) 23 h after treatments. Then, the cells were washed three times with a phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde for 15 min.

Immunocytochemistry was performed using a rabbit polyclonal anti-Iba1 primary antibody (1/500, WAKO; Japan), a specific marker for microglial cells. Cells were first incubated with 0.3% Triton-X-100 in PBS containing 1% bovine serum



β-Actin, Actin, beta; Carkl, sedoheptulokinase; COX2, cyclooxygenase-2; Glut1, Glucose transporter 1; gp91phox, catalytic subunit of NAPH oxidase; Hifα, hypoxia inducible factor 1, alpha subunit; Hk1, hexokinase 1, Hk2, hexokinase 2; IL1β, interleukin 1, beta; IL6, interleukin 6; iNOS, inducible nitric oxide synthase; Pfkp, phosphofructokinase, platelet; TNFα, tumor necrosis factor alpha; Rn18s, 18S ribosomal RNA.

albumin (BSA) and 10% normal donkey serum for 20 min at room temperature, and then overnight at 4◦C with the primary antibody. Once they had been rinsed in PBS, cells were incubated for 1 h at room temperature with an ALEXA 488 donkey anti-rabbit secondary antibody (1/1,000; Invitrogen). Antibodies were diluted in 0.3% Triton X-100 in PBS containing 1% BSA and 10% normal donkey serum.

Images of three microscopic fields using a 20× objective were obtained with an Olympus IX70 fluorescence microscope and a digital camera (CC-12, Olympus Soft Imaging Solutions GmbH). Two to three wells per experimental condition were processed and each experimental condition was repeated at least four times. Visual counting of the FluoSpheres was performed. The percentage of phagocytic cells and the average number of fluorescent microspheres per microglial cell were calculated. To further characterize the phagocytic activity, we also calculated the percentage of cells showing lower phagocytic activity (microspheres/cell) than the controls and the percentage of cells showing higher phagocytic activity than the controls.

#### ATP Production

The intracellular production of ATP was determined using a luminescence assay kit (ATPlite Luminescence ATP Detection Assay System, PerkinElmer, Waltham, MA, USA) following the manufacturer's instructions. Briefly, cells in 96-well plates (mixed glia) or 6-well-plates (microglia) were lysed 24 h after treatments and the ATP concentration was measured based on the production of light, caused by the reaction of the ATP with added luciferase and D-luciferin. The emitted light was quantified using a luminometer (Orion Microplate Luminometer, Berthold Detection System, Germany). The ATP concentration in the samples was calculated from an ATP standard curve.

#### Data Presentation and Statistical Analysis

The results are presented as the mean + SEM. At least three independent experiments were performed for analysis. Data were statistically analyzed with the GraphPad Prism software. Statistical analyses were performed using the one-way analysis of variance (ANOVA) followed by the Newman-Keuls posttest, and a two-way ANOVA followed by the Bonferroni post-test. Values of p < 0.05 were considered statistically significant.

# RESULTS

#### Effects of MPP+ and Rotenone on Glial Cell Viability

In a preliminary study, we performed dose-response experiments in order to select working concentrations of MPP+ and rotenone that did not result in significant alterations in cell viability after 24 h exposure. We evaluated glial cell viability after treating the mixed glial or the microglial cultures with increasing concentrations of MPP+ (10, 25, 50 and 100 µM) or rotenone (20, 40, 100 and 150 nM), both in the absence and in the presence of LPS/IFN-γ, considering the MTT assay and PI staining. In mixed glial cell cultures, MPP+ induced a concentrationdependent decrease in MTT reduction that was accentuated in the presence of LPS/IFN-γ (**Figure 1A**). On the contrary, no alterations in MTT reduction were observed in microglial cell cultures treated with MPP+, both in the absence and presence of LPS/IFN-γ (**Figure 1B**). Rotenone-treated mixed glial cell cultures showed a significant decrease in MTT reduction from 100 nM rotenone. In the presence of LPS/IFN-γ, there was a significant decrease in MTT reduction even at 20 nM rotenone (**Figure 1C**). As in the case of MPP+ treatments, no alterations in MTT reduction were observed in microglial cell cultures treated with rotenone or rotenone and LPS/IFN-γ (**Figure 1D**). To determine whether the decrease in MTT reduction in MPP+ and rotenone-treated mixed glial cell cultures was due to a decrease in metabolic activity or due to cell death, PI staining was performed. Mixed glial cultures treated with 50 and 100 µM MPP+ showed a significant increase in the percentage of PI positive nuclei. This effect was accentuated in the presence

of LPS/IFN-γ, and a significant increase was also detected at 25 µM MPP+ (**Figure 1E**). No alterations in the percentage of PI-positive nuclei were observed in microglial cultures treated with MPP+ (**Figure 1F**). In addition, no significant increases in the percentage of PI-positive nuclei were observed in rotenone-treated mixed glial cell cultures or microglial cultures (**Figures 1G,H**), with the exception of cells treated with 100 µM rotenone and LPS/IFN-γ.

The concentrations of 10 and 25 µM MPP+ and 40 and 100 nM rotenone were used in subsequent studies. Representative images of the cultures in these experimental conditions are shown in **Figure 2**, which corroborate the lack of a toxic effect of the concentrations of MPP+ (**Figure 2A**) and rotenone (**Figure 2B**) selected for further experiments.

#### MPP+ and Rotenone Induce Alterations in the Expression of Pro-inflammatory Genes in LPS/IFN-γ-Treated Primary Glial Cultures

We next determined whether MPP+ and rotenone induced a pro-inflammatory phenotype in primary glial cell cultures, as well as whether they had some effect on the development of the pro-inflammatory response induced by LPS/IFN-γ. We determined the mRNA expression of the cytokines IL1β, IL6 and TNFα and the enzymes inducible NO synthase (iNOS), cyclooxygenase 2 (COX2) and gp91phox (the catalytic subunit of NADPH oxidase), as markers of a pro-inflammatory response. In general, MPP+ (**Figure 3**) and rotenone (**Figure 4**) treatment did not significantly induce the mRNA expression of these pro-inflammatory markers in the primary glial cell cultures, although a trend towards increased expression was observed for some mRNAs, especially in rotenone-treated mixed glial cultures. On the contrary, 6 h after LPS/IFN-γ treatment, the mRNA expression of all the pro-inflammatory markers tested was clearly induced (**Figures 3**, **4**). However, MPP+ and especially rotenone induced alterations in the pattern of expression of these markers in LPS/IFN-γ treated cultures. When glial cell cultures were treated with LPS/IFN-γ in the presence of MPP+, the induction of IL1β mRNA expression was significantly inhibited in the mixed glial and microglial cultures (**Figures 3A,B**), while COX2 mRNA expression was further increased in the mixed glial cultures (**Figure 3A**) and gp91phox mRNA was induced in the microglial cultures (**Figure 3B**). More importantly, rotenone exposure significantly abrogated LPS/IFN-γ induction of the mRNA expression of all pro-inflammatory markers in the mixed glial cultures (**Figure 4A**), as well as IL1β, IL6 and COX2 mRNA expression in microglial cultures (**Figure 4B**).

## MPP+ and Rotenone Inhibit LPS/IFN-γ-Induced NO and Pro-inflammatory Cytokine Production in Primary Glial Cultures

We also analyzed the effect of MPP+ and rotenone on NO, IL1β, IL6 and TNFα release into the culture medium. MPP+ alone induced a decrease in NO production and an increase

in IL6 release into the culture medium in mixed glial cell cultures 24 h after treatment (**Figure 5A**). The latter effect was also observed in the MPP+-treated microglial cultures (**Figure 5B**), as well as in the rotenone-treated mixed glial (**Figure 5C**) and microglial cultures (**Figure 5D**). LPS/IFN-γtreatment clearly increased NO production in the mixed glial cultures, and MPP+ (25 µM) and rotenone (40 and 100 nM) significantly inhibited this effect (**Figures 5A,C**). Significant NO production was not detected in the microglial cultures treated with LPS/IFN-γ for 24 h (**Figures 5B,D**). However, an increase in NO production was observed when the microglial cultures were treated with LPS/IFN-γ for 48 h, but MPP+ and rotenone did not modify this effect (data not shown). With regards to cytokine release, LPS/IFN-γ-treatment resulted in drastic increases in IL1β, IL6 and TNFα levels in the mixed glial (**Figures 5A,C**) and microglial (**Figures 5B,D**) cultures. MPP+ exposure (25 µM) significantly inhibited LPS/IFN-γinduced release of IL1β and IL6, but not TNFα, in the mixed glia (**Figure 5A**), while it had no significant effect on the production of these cytokines in the microglial cell cultures (**Figure 5B**). Interestingly, rotenone (40 and 100 nM) significantly inhibited IL1β, IL6 and TNFα release induced by LPS/IFN-γ in both the mixed glial (**Figure 5C**) and microglial cultures (**Figure 5D**).

#### MPP+ and Rotenone Treatment Inhibit the Phagocytic Activity of Microglial Cells

We then evaluated whether MPP+ and rotenone modified the phagocytic activity of the microglial cells, another important parameter used to characterize the microglial activation phenotype. Both MPP+ and rotenone treatment showed a tendency to decrease the percentage of phagocytic microglial cells that were statistically significant when the cells were also treated with LPS/IFN-γ (**Figure 6A**). In addition, MPP+ and rotenone treatment resulted in a significant increase in the percentage of microglial cells showing low phagocytic activity (number of microspheres per cell lower than control) and a subsequent significant decrease in the percentage of cells showing high phagocytic activity (number of microspheres per cell higher than control; **Figure 6B**). These effects were accentuated in the presence of LPS/IFN-γ.

#### ATP Production Is Compromised in MPP+ and Rotenone-Treated Glial Cell Cultures

In an attempt to better characterize the metabolic status of the cells, we determined the intracellular ATP production in response to MPP+, rotenone and LPS/IFN-γ treatments. In general, ATP production was modified in the mixed glial cultures in our MPP+ and rotenone experimental models (p < 0.001, one-way ANOVA; **Figure 7**). In particular, ATP production was significantly decreased after 25 µM MPP+ treatment (**Figure 7A**). On the contrary, a significant increase in ATP production was detected in the LPS/IFN-γ-treated mixed glial cultures, which was abrogated in the presence of MPP+ and rotenone (**Figures 7A,B**). ATP production was also modified in the microglial cultures in the MPP+ and rotenone experimental models (p < 0.05, one-way ANOVA), but to a lesser extent than in the mixed glial cultures (**Figures 7C,D**).

# Metabolic Changes in LPS/IFN-γ-Treated Glial Cultures: Effect of MPP+ and Rotenone

In immune cells, the development of specific immune responses is associated with specific metabolic changes. Increased glycolysis and potentiation of the pentose phosphate pathway, together with the inhibition of oxidative phosphorylation has been reported for immune cells showing a pro-inflammatory phenotype. We checked whether this was the case in our glial cultures treated with LPS/IFN-γ and whether MPP+ and rotenone were able to modify it. We determined the mRNA expression of genes encoding critical proteins for the switch to glycolysis: glucose transporter 1 (Glut1; glucose entrance into the cell), key glycolytic enzymes such as hexokinase 1 (Hk1) (glycolysis initial rate limiting step) and phosphofructokinase 1 (PFK1) (master regulator of glycolysis), the glycolysis activator hypoxia-inducible factor 1α (Hif1α) and carbohydrate kinase-like protein (Carkl), involved in the control of the pentose phosphate pathway. In the microglial cell cultures, MPP+ and rotenone alone did not modify the expression of these genes per se (**Figure 8**). On the contrary, as expected, LPS/IFN-γ treatment induced an increase in their

statistical significance of the effects of MPP+ alone.

only considering the L/I-free groups. This latter analysis was performed to detect whether the high values observed in the L/I group may hinder the detection of

expression (**Figure 8**), with the exception of Carkl mRNA (**Figures 8I,J**), which showed a decrease. MPP+ treatment further increased LPS-IFN-γ-induced Glut1 mRNA expression (**Figure 8A**), while rotenone inhibited LPS-IFN-γ-induced Pfk1 (**Figure 8F**) and Hif1α (**Figure 8H**) mRNA expression. We also evaluated the expression of these mRNAs in the mixed glial cultures. MPP+ and rotenone alone increased the Glut1 mRNA (**Figures 9A,B**), MPP+ and Hif1α mRNA expression (**Figure 9G**). LPS/IFN-γ treatment inhibited the expression of the glycolytic genes Glut1 (**Figures 9A,B**), Hk1 (**Figures 9C,D**) and Pfk1 (**Figures 9E,F**), as well as Carkl mRNA expression (**Figures 9I,J**), and increased the expression of Hif1α (**Figures 9G,H**). Rotenone inhibited LPS/IFN-γ-induced Hif1α mRNA expression (**Figure 9H**).

#### DISCUSSION

In this study, we show that the response of glial cells to a pro-inflammatory stimulus is modified by the neurotoxic agents MPP+ and rotenone. MPP+ and rotenone treatment did not induce a significant pro-inflammatory phenotype in the primary mixed glial and microglial cultures per se. However, these neurotoxic agents, mainly rotenone, did impair the development of a pro-inflammatory phenotype in the LPS/IFN-γ-treated glial cultures. This effect was observed in the absence of significant cell death but in the presence of impaired metabolic activity.

The toxic effects of MPP+ and rotenone on neurons have repeatedly been demonstrated using primary neuronal cultures, with dopaminergic neurons showing the highest sensitivity to

100 nM Rot for 24 h, in the absence or in the presence of LPS/IFN-γ (L/I). Bars are means ± SEM of four independent experiments. <sup>∗</sup>p < 0.5, ∗∗p < 0.01 and ∗∗∗p < 0.001 vs. control (C); #p < 0.05, ##p < 0.01 and ###p < 0.001 vs. L/I; one-way ANOVA and Newman-Keuls post-test. \$p < 0.05, \$\$p < 0.01 and \$\$\$p < 0.001 MPP+ and Rot alone vs. C, one-way ANOVA and Newman-Keuls post-test only considering the L/I-free groups. This latter analysis was performed to detect whether the high values observed in the L/I group may hinder the detection of a statistical significance of the effects of MPP+ and Rot alone.

the toxic effects of these compounds. In mouse primary cultures, dopaminergic neuron death is observed at concentrations from 0.1 µM MPP+ (1-week exposure; Kinugawa et al., 2013) or 3 µM MPP+ (48 h exposure; Henze et al., 2005), and 10 nM (1-week exposure; Gao et al., 2003) or 5 nM rotenone (48 h exposure; Radad et al., 2006). Exposure to higher concentrations of these neurotoxins is necessary to induce the death of non-dopaminergic neurons (Gao et al., 2003; Henze et al., 2005). The presence of microglial cells in neuronal cultures

has been associated to the increased neurotoxicity of MPP+ and rotenone (Gao et al., 2002, 2003; Emmrich et al., 2013; Kinugawa et al., 2013). However, as MPP+- and rotenonedamaged neurons induce reactive microgliosis, which has a neurotoxic effect, it is difficult to establish the contribution of a direct effect of the toxins on glial cells in the neurotoxicity observed. In fact, although the neurotoxic effect of MPP+ and rotenone has been widely described using in vivo and in vitro experimental approaches, their direct effects on glial cells remain

(number of microspheres/cells > C). Bars are means ± SEM of four independent experiments. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 and ∗∗∗p < 0.001 vs. corresponding C;

poorly characterized (Gao et al., 2003; Klintworth et al., 2009; Bournival et al., 2012; Du et al., 2014; Chen et al., 2015; Zhou et al., 2016). Most of the studies performed until now using glial cell cultures have tested whether these neurotoxins induce a pro-inflammatory phenotype in the microglial cells, and the results obtained are controversial. The range of concentrations used in these studies are higher than that used in neuronal cultures (0.1–500 µM MPP+ and 10 nM-1 µM rotenone; Gao et al., 2003; Henze et al., 2005; Klintworth et al., 2009; Bournival et al., 2012; Jin et al., 2012; Du et al., 2014). Some authors have reported no alterations (Klintworth et al., 2009; Ferger et al., 2010; Jin et al., 2012), but others have shown the induction of pro-inflammatory markers in the MPP+- and rotenone-treated microglial cultures (Du et al., 2014; Zhang et al., 2014; Liang et al., 2015; Zhou et al., 2016). Differences in the pattern of neurotoxin treatment (concentration and duration of the treatment) and the cell types used (primary cultures and cell lines from different species) may partially account for the differences observed. Most studies have considered microglial cell lines, while studies using

#p < 0.05 and ##p < 0.01 vs. low-phagocytic cells; two-way ANOVA and Bonferroni post-test.

primary microglial cultures are scarce. In this study, we show that concentrations of MPP+ and rotenone that did not affect cell viability in primary glial cultures at 24 h did not result in the induction of a significant pro-inflammatory phenotype (with the exception of IL6 production), but they interfered with the development of the pro-inflammatory phenotype induced by LPS/IFN-γ. Thus, MPP+ and rotenone inhibited pro-inflammatory cytokine production induced by LPS/IFN-γ in glial cells (IL1β in the case of MPP+, and also IL6 and TNFα in the case of rotenone). They also modified the expression of pro-inflammatory enzymes (iNOS, COX2 and/or gp91phox). In general, the alterations observed were more pronounced in the mixed glia than in the microglial cell cultures. In addition, the effect of rotenone was stronger than that of MPP+, although the concentrations of rotenone used were three orders of magnitude below those of MPP+. MPP+ and rotenone treatment also interfered in the phagocytic activity of the microglial cells, which was clearly inhibited after neurotoxin treatment, especially in the presence of LPS/IFN-γ. Altogether, these results suggest

that MPP+ and rotenone directly impair the ability of the glial cells to respond to a pro-inflammatory insult. In this sense, exposure to stimuli that affect the mitochondrial activity, such as hypoxia or respiratory chain inhibitors, has been suggested to alter the immune response of macrophages (Wiese et al., 2012).

At the cellular level, the main target of both MPP+ and rotenone is the mitochondrial electron transport chain, where they selectively inhibit complex I (Dauer and Przedborski, 2003). As a consequence, ATP production is compromised, O<sup>−</sup> 2 levels increase, and subsequent oxidative stress occurs. This is critical in neuronal cells, where energy production depends mainly on ATP synthesis through oxidative phosphorylation (reviewed in Bélanger et al., 2011). In contrast, astrocytes are mainly glycolytic (reviewed in Bélanger et al., 2011). In addition, astrocytes can generate lactate from glycogen via glycolysis under metabolic activation (Hertz et al., 2007). Macrophages/microglia have the capacity to generate ATP by both glycolytic and oxidative pathways (although in the case of microglial cells the field is still underexplored; Van den Bossche et al., 2017; Ghosh et al., 2018). Indeed, they are able to shift from oxidative phosphorylation to aerobic glycolysis (production of lactate in the presence of oxygen) to obtain ATP from different pathways according to the metabolic demands of their activation status (Haschemi et al., 2012; Galván-Peña and O'Neill, 2014; Orihuela et al., 2016). The classical activation or M1/pro-inflammatory phenotype is associated with inhibition of the respiratory chain and the potentiation of aerobic glycolysis, which results in more rapid ATP production to satisfy the metabolic demands associated with the quick pro-inflammatory response of the M1 phenotype. In this situation, the glycolytic and pentose phosphate pathways are potentiated, and oxidative phosphorylation is inhibited (Haschemi et al., 2012). A metabolic-epigenetic crosstalk is suggested to control macrophage activation (Baardman et al., 2015). In contrast, pro-inflammatory stimuli increase tricarboxylic acid activity in astrocytes (Gavillet et al., 2008). In mixed glial cultures, LPS/IFN-γ treatment increased ATP production, an effect that was clearly inhibited by MPP+ and rotenone. These results suggest that MPP+- and rotenonetreated cultures suffer metabolic stress that is aggravated when the cells increase their energetic demands after LPS/IFN-γ treatment. Consequently, activated glial cells may not fulfill their

genes involved in the control of glycolysis in the mixed glial cell cultures. mRNA expression of the glucose transporter Glut1 (A,B) and the glycolytic enzymes Hk1 (C,D) and Pfk1 (E,F), as well as Hif1α (G,H) and Carkl (I,J). The primary mixed glial cultures were treated for 6 h with 10 and 25 µM MPP+ or 40 and 100 nM rotenone (Rot), both in the absence and the presence of LPS/IFN-γ (L/I). Rn18s and β-actin were used as housekeeping genes. Bars are means ± SEM of four independent experiments. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 and ∗∗∗p < 0.001 vs. control (C); #p < 0.05 vs. L/I; one-way ANOVA and Newman-Keuls post-test.

one-way ANOVA and Newman-Keuls post-test.

∗∗p < 0.01 and ∗∗∗p < 0.001 vs. control (C); #p < 0.05 and ##p < 0.01 vs. L/I;

metabolic demands in the presence of MPP+ and rotenone, which would explain why mixed glial cultures exposed to these toxicants were not able to produce an appropriate pro-inflammatory response to LPS/IFN-γ. The effects of MPP+ and rotenone on ATP production were less drastic in the LPS-IFN-γ-treated microglial than the mixed glial cultures. ATP production was not significantly compromised in the LPS/IFN-γ-treated microglial cultures exposed to MPP+ and rotenone, with the exception of 25 µM MPP+ treatment. In addition, the response to LPS/IFN-γ was also impaired, to a lesser extent, in the microglial cultures than in the mixed glial cultures. Altogether, these results suggest that the microglial cells can better cope with the metabolic alterations induced by MPP+ and rotenone than astrocytes can, which accounts for 75% of the cells in the mixed glial cultures. A possible explanation is that while microglial cells developing a pro-inflammatory phenotype switch to glycolysis (Haschemi et al., 2012; Galván-Peña and O'Neill, 2014; Orihuela et al., 2016), astrocytes exposed to pro-inflammatory stimuli increase the activity of the tricarboxylic acid cycle (Gavillet et al., 2008), which in the presence of MPP+ and rotenone will encounter truncated oxidative phosphorylation. However, the involvement of a differential response of the activated microglial cells to the toxins in the presence of astrocytes (or impaired astrocytes) cannot be ruled out.

To assess whether the glycolytic switch mentioned above was behind the ATP production in activated glial cultures, we evaluated the expression of genes encoding critical proteins for the glycolytic pathway. In the microglial cultures, a switch to the glycolytic pathway in the LPS/IFN-γ-activated microglial cultures was suggested by the observed increase in the expression of Glut1, HK1, Pfk1 and Hif1α mRNA and the decreased expression of the Carkl mRNA. Increased expression of the glucose transporter Glut1 may result in more glucose uptake, while the increased expression of Hk1 and Pfk1, which regulate critical steps in glycolysis, may increase the glycolytic rate. It has been suggested that Hif activation contributes to macrophage polarization, and that Hifα-dependent glycolysis favors polarization to a M1 phenotype (Palazon et al., 2016; Taylor et al., 2016). In addition, metabolic intermediates such as succinate play a role in Hif1α stabilization and subsequent IL1β expression in LPS-treated macrophages (Tannahill et al., 2013). Inhibition of the Carkl expression potentiates the flux through the pentose phosphate pathway (Haschemi et al., 2012). Whereas MPP+ exposure resulted in a further increase in Glut1 expression in the LPS/IFN-γ-treated microglial cultures, rotenone exposure partially inhibited the LPS/IFN-γ-induced Pfk and Hifα mRNA expression. Consequently, in the case of microglial cultures, the attenuated pro-inflammatory response to LPS-IFN-γ mostly observed in the presence of rotenone may result from some alterations to the glycolytic switch. In addition, it cannot be ruled out that ATP production through oxidative phosphorylation may also partially contribute to the energy demand required to develop a pro-inflammatory response, even in situations where the switch to glycolysis occurs. In this sense, Wang et al. (2018) showed that 2-deoxyglucose, which blocks glycolysis and partially inhibits glycolytic-dependent oxidative phosphorylation, has a stronger inhibitory effect on the IFN-γinduced inflammatory response in macrophages than inhibiting glycolysis when replacing glucose in the cell culture medium with galactose, which reduces glycolytic flux without interfering with oxidative phosphorylation. Interestingly, control of the pro-inflammatory macrophage response, through metabolic reprogramming, has been suggested as a potential therapeutic strategy to promote remission in chronic inflammatory diseases (Mills and O'Neill, 2016).

We observed contrasting effects in the LPS/IFN-γ-treated mixed glia and microglial cultures in terms of the mRNA expression of the glycolytic enzymes Glut1, Hk1, and Pfk1. Their expression was inhibited in the mixed glia, suggesting the contribution of astrocytes to the effects observed. Although astrocytes are mainly glycolytic (Bélanger et al., 2011), pro-inflammatory stimuli increase tricarboxylic acid activity in astrocytes (Gavillet et al., 2008). The decreased expression of the glycolytic enzymes we observed may reflect this switch. Consequently, the impaired response of the mixed glial cultures to LPS/IFN-γ in the presence of MPP+ and rotenone may be explained by the fact that oxidative phosphorylation, which would be responsible for the main ATP production from products of the tricarboxylic acid cycle, is inhibited by MPP+ and rotenone. However, the involvement of the microglial cells in the response of the mixed glial cultures to MPP+ and rotenone plus LPS/IFN-γ cannot be ruled out.

Finally, some studies show that MPP+ causes DNA damage (Zhang et al., 1995) and oxidative DNA damage in neuronal cells (Chen et al., 2005). In addition, rotenone-induced DNA damage (Goswami et al., 2016) and DNA methylation (Scola et al., 2014) in neurons have also been described. Although there are no reports on MPP+- and rotenone-induced DNA alterations on glial cells, we cannot discard that these alterations may be behind the decreased expression of inflammatory markers we detected in LPS/IFN-γ–treated glial cultures exposed to the neurotoxins.

# CONCLUSION

In summary, the results of the present study show that the pro-inflammatory response induced by LPS/IFN-γ in mouse primary glial cell cultures, is impaired under MPP+ and rotenone exposure, mainly when both the astrocytes and microglia are present. This suggests that the immune response of the glial cells is compromised in the presence of neurotoxins that inhibit the mitochondrial electron transport chain. We are currently studying the possible effects of MPP+ and rotenone on the development of an anti-inflammatory phenotype by the glial cells. Although the involvement of the glial cells in the development of neurodegenerative diseases is widely accepted, the precise role they play in every neurodegenerative disorder remains to be established. As many genetic and environmental factors are probably involved in the etiopathogenesis of neurodegenerative diseases, many factors may also determine when and how the glial cells take part in the pathological process. In the case of pathologies where the exposure to certain neurotoxicants is a risk factor, such as PD, the direct effect of the toxic agents on glial cell function may be an additional factor to take into account, as alterations in glial function will have an effect on neuronal function and CNS homeostasis. In this context, our results suggest that glial metabolic alterations induced by neurotoxin exposure compromises the brain's immune response. This impaired immune response may imply a more vulnerable brain, which can be a further aspect contributing to the development of PD.

#### DATA AVAILABILITY

All data generated or analyzed during this study are included in this published article.

#### AUTHOR CONTRIBUTIONS

NR-L and EX-E performed most of the experiments and analyzed the data. JSe participated in the processing of the samples and quantified the phagocytosis assay. JSa provided critical guidance and contributed to the final version of the manuscript. CS

#### REFERENCES


conceived and coordinated the experiments, provided guidance in the production of data and drafted the manuscript. All authors provided input and ideas throughout the process, and critically revised and approved the final manuscript.

#### FUNDING

NR-L was recipient of an FPU grant (FPU13/05491) from the Spanish Ministerio de Educación, Cultura y Deporte. This study was supported by grants PI14/00302 and PI15/00033 from the Instituto de Salud Carlos III (Spain) with joint financing by FEDER funds from the European Union.

#### ACKNOWLEDGMENTS

We thank Lucas Blasco and Andrés Jurado for their technical assistance. We acknowledge support for the publication fee, by the CSIC Open Access Publication Support Initiative through its Unit of Information Resources for Research (URICI).

properties of microglia. J Neuroinflammation 7:45. doi: 10.1186/1742- 2094-7-45


**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 Rabaneda-Lombarte, Xicoy-Espaulella, Serratosa, Saura and Solà. 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, Hop-Derived Bitter Components of Beer, Attenuate Age-Related Inflammation and Cognitive Decline

#### Yasuhisa Ano1,2\*, Rena Ohya1,2 , Keiji Kondo<sup>2</sup> and Hiroyuki Nakayama<sup>1</sup>

<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, Yokohama, Japan

With the aging population rapidly increasing worldwide, preventive measures and treatments for age-related cognitive decline and dementia are of utmost importance. We have previously demonstrated that the consumption of iso-α-acids (IAA), which are hop-derived bitter compounds in beer, prevents the formation of disease pathology in a transgenic mouse model of Alzheimer's disease (AD). However, the effect of IAA consumption on age-related cognitive decline is unknown. In the present study, we examined the effect of long-term and short-term dietary consumption of IAA, on age-related memory impairments and inflammation in the hippocampus of aged mice. When compared with young mice, aged mice showed impairment in spatial working memory during the Y-maze spontaneous alternation test, impairment in object recognition memory during the novel object recognition test (NORT), a pro-inflammatory hippocampal microglial phenotype with increased CD86 expression and inflammatory cytokine production, increased levels of glutamate and amyloid β1–42, and decreased levels of dopamine (DA). In aged mice fed IAA for 3 months, the age-related alterations in memory, microglial inflammation, and glutamate, amyloid β1–42, and DA levels were all significantly attenuated. Additionally, the oral administration of IAA for 7 days in aged mice with memory impairment, also improved spatial and object recognition memory. These results suggest that IAA consumption prevents inflammation in the hippocampus and ameliorates age-related cognitive decline.

#### Keywords: aging, cognitive decline, hippocampus, inflammation, iso-α-acids, memory

#### INTRODUCTION

The increasing burden of dementia and cognitive impairment in rapidly expanding aging populations is shouldered not only by patients and their families but also by national healthcare systems. The lack of an effective disease-modifying therapy for dementia has garnered increasing attention on preventive approaches, such as diet, exercise, and learning. Etiological studies on lifestyle suggest that low to moderate consumption of alcohol, such as wine and beer, may reduce the risk of cognitive decline and the development of dementia. Individuals who consumed low to moderate levels of alcoholic beverages on a daily basis showed a significantly lower

#### Edited by:

Maria Jose Bellini, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Fernando Correa, CONICET Centro de Estudios Farmacológicos y Botánicos (CEFYBO), Argentina Florencia Labombarda, Instituto de Biología y Medicina Experimental (IBYME), Argentina

#### \*Correspondence:

Yasuhisa Ano yasuhisa\_ano@kirin.co.jp

Received: 15 November 2018 Accepted: 17 January 2019 Published: 04 February 2019

#### Citation:

Ano Y, Ohya R, Kondo K and Nakayama H (2019) Iso-α-acids, Hop-Derived Bitter Components of Beer, Attenuate Age-Related Inflammation and Cognitive Decline. Front. Aging Neurosci. 11:16. doi: 10.3389/fnagi.2019.00016

**Abbreviations:** Aβ, amyloid β; DA, dopamine; DOPAC, 3,4-dihydroxyphenylacetic acid; ECD, electrochemical detection; DI, discrimination index; HVA, homovanillic acid; HPLC, high-performance liquid chromatography; IAA, iso-α-acids; NORT, novel object recognition test; PPAR-γ, peroxisome proliferator-activated receptor-γ.

risk of developing a neurodegenerative disease than those who either abstained from alcoholic beverages or drank heavily (Matsui et al., 2011; Neafsey and Collins, 2011; Horvat et al., 2015). These risk-reducing effects are thought to be because of the profile of compounds found in alcoholic beverages. Red wine is known to contain resveratrol, a polyphenolic compound with neuroprotective properties (Neafsey and Collins, 2011; Porquet et al., 2014; Witte et al., 2014). Conversely, beer has remained the most-consumed alcoholic beverage in the world for more than a thousand years; to date there are few reports on which constituents of beer could be beneficial for preventing cognitive decline.

Hops, the female inflorescences of the hop plant (Humulus lupulus L.), have been used in beer brewing since 822 AD and are used as both a preservative and a flavoring agent in the beer-brewing process. The bitter taste of beer originates from the α-acids found in hops. Owing to the fact that iso-α-acids (IAA) activate the peroxisome proliferator-activated receptorγ (PPAR-γ; Yajima et al., 2004), the IAA found in beer have antioxidant and anti-metabolic syndrome properties and are also reported to prevent diet-induced obesity in rodents and to improve hyperglycemia, a result which has been confirmed in humans (Obara et al., 2009). We previously demonstrated that the long-term intake of IAA prevented Alzheimer's pathology in transgenic model mice (Ano et al., 2017). IAA suppressed microglial inflammation induced by the deposition of amyloid β (Aβ) in the brain and prevented cognitive decline. Additionally, IAA activated PPAR-γ and regulated microglial phagocytosis and inflammation. Our group has also demonstrated that IAA suppressed microglial inflammation in tauopathy mice and improved obesity-induced cognitive impairment by suppressing inflammation induced by a high-fat diet (Ano et al., 2018b; Ayabe et al., 2018). However, the effects of IAA on age-related cognitive decline and neuronal dysfunction have not yet been demonstrated. Most cases of cognitive decline are associated with aging; consequently, in this study, we examined the effects of IAA consumption on cognitive impairment and brain inflammatory processes in aged mice.

# MATERIALS AND METHODS

### Preparation of Iso-α-acids (IAA)

The α-acids predominantly comprised of the following three congeners: cohumulone, humulone, and adhumulone. During the brewing process, they were each isomerized into two epimeric isomers: cis-IAA and trans-IAA. We used isomerized hop extract (IHE; Hopsteiner, Mainburg, Germany) containing 30.5% (w/v) IAA and 65% H2O as the IAA sample for our experiments. Based on the analysis in our previous report, IHE comprises six isomers: trans-isocohumulone (1.74% w/v), cis-isocohumulone (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; Ano et al., 2017).

#### Animals

Male C57BL/6J mice (Charles River Japan, Tokyo, Japan) were maintained at the Kirin Company Ltd. The Animal Experiment Committee of Kirin Company Limited approved all experiments, which were conducted between 2016 and 2017 in strict accordance with their guidelines. We made every possible effort to minimize suffering. Mice were fed a standard purified rodent diet (AIN-93M, Oriental Yeast, Tokyo, Japan).

Our previous study showed that dietary intake of 0.05% (w/w) IAA reduced inflammation in the brain of Alzheimer's model mice (Ano et al., 2017). To study the effect of long-term IAA intake on age-related cognitive decline, aged mice (68 weeks of age) were fed AIN-93M with or without 0.05% (w/w) IAA for 3 months. Young control mice (7 weeks of age) were fed AIN-93M diet without IAA for 3 months. After a 3-month dietary intervention, the aged mice fed IAA (n = 10), control aged mice not fed IAA (n = 9), and control young mice (n = 12) underwent behavioral evaluation, after which brain samples were obtained for subsequent biochemical evaluation. The body weight of the aged mice with and without IAA did not differ between groups (**Supplementary Figure S1A**). To study the effect of short-term IAA intake, mice aged at 22 months were intragastrically administered IAA at either 0 mg/kg (distilled water as vehicle, n = 13) or 1 mg/kg (n = 14) for 9 days. Young mice aged at 7 months, that were administered with vehicle (n = 15), served as additional controls. At 7 days, mice were subjected to the spontaneous alternation test; at 8 and 9 days, mice were subjected to the novel object recognition test (NORT) at 1 h after the IAA administration.

#### Spontaneous Alternation Test

We evaluated spatial memory by performing a spontaneous alternation test using a Y-maze in accordance with our previous study (Ano et al., 2018a). The Y-maze is a 3-arm maze with equal angles between each black polyvinyl plastic arm (25 cm long × 5 cm wide × 20 cm high). Each mouse was initially placed in the start arm of the maze, and the number and sequence of subsequent arm entries was recorded over 8 min. We defined the alternation score (%) for each mouse as the ratio of the actual number of alternations to the total possible number of alternations (defined as the total number of arm entries minus 2) multiplied by 100 as follows: alternation score (%) = [(number of alternations)/(total arm entries − 2)] × 100.

# NORT

We evaluated episodic memory by performing NORT in accordance with our previous study (Ano et al., 2018a). NORT was performed during the light period in a polyvinyl chloride box (25 cm × 40 cm × 20 cm) without a roof. For the acquisition trial, we used 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); for the retention trial, we used one triangle pole or pyramid as the familiar object and a golf ball (4.5 cm diameter) as the novel object In all trials, we placed the objects 7.5 cm from the corner of the box. In the acquisition trial, each mouse was allowed 10-min exploration time in the box with the two identical objects at 1 h after intragastric administration of the test sample. At 24 h after the acquisition trial and 1 h after intragastric administration, mice were allowed to explore the box with the novel and familiar objects for 5 min. The discrimination index (DI) was calculated by dividing the difference in time taken to explore the novel object and the familiar object, by the total time spent exploring both objects, that is, DI = (novel object exploration time − familiar object exploration time)/(total exploration time). Using this method, equal exploration of both objects was indicated by a DI of 0.

## Measurements of Activity, Food Intake, and Water Intake in Home Cage

We monitored the amount of food intake, water intake, and ambulatory activity in the home cages using a three-point meter for 72 h (O'HARA & Co., Ltd., Tokyo, Japan). To monitor home cage activity, the interruption of infrared beams positioned on the X and Y axes around the cage detected the position of the mouse, automatically measuring their position and movement in their home cages. In this experiment, we measured the moving distance of each mouse over 5 min for a total of 72 h.

# Aβ and Cytokine Measurement

To measure levels of cytokines, Aβ, and Tau, we homogenized the hippocampus of the left hemisphere in TBS buffer (Wako) with a multi-beads shocker (Yasui Kikai, Osaka, Japan). After centrifugation at 50,000× g for 20 min, we collected the supernatant. We measured the total protein concentration of each supernatant with a BCA protein assay kit (ThermoScientific, Yokohama, Japan). We assayed the supernatant to quantify soluble Aβ1–42 (Wako), Tau (ThermoScientific), and phosphorylated Tau (pTau, pS199, ThermoScientific) by ELISA and cytokines by a Bio-Plex assay system (Bio-Rad, Hercules, CA, USA).

#### Microglial Analysis

We isolated microglial cells from the mouse brain by magnetic cell sorting after conjugation with anti-CD11b antibodies, as described previously (Ano et al., 2015). Isolated CD11bpositive cells (>90% pure, as evaluated by flow cytometry) were cultured in DMEM/F-12 (Gibco, Carlsbad, CA, USA) supplemented with 10% fetal calf serum (Gibco, Carlsbad, CA, USA) and 100 U/ml penicillin/streptomycin (Sigma-Aldrich, St. Louis, MO, USA). We treated the microglia with a leukocyte activation cocktail containing GolgiPlus (BD Biosciences, San Jose, CA, USA) for 12 h, fixed and permeabilized them with a Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences, San Jose, CA, USA), and stained the microglia with FITC-conjugated anti-mouse TNF-α (MP6-XT22, eBioscience, San Diego, CA, USA), APC/Cy7-conjugated anti-mouse CD11b (M1/70, BD Pharmingen), and APC-conjugated anti-mouse CD86 (GL1, eBioscience, San Diego, CA, USA) antibodies. We analyzed populations of cytokine-producing cells and

the expression of cell markers with a flow cytometer (BD FACSCantoII, BD Biosciences, San Jose, CA, USA). We have expressed flow cytometry data as the mean median mode of the fluorescence intensity.

#### Monoamine Analysis

To evaluate the levels of the monoamine dopamine (DA) and its metabolites in the brain, we homogenized tissue in 0.2 M perchloric acid (Wako) containing 100 µM disodium EDTA (Sigma-Aldrich, St. Louis, MO, USA). After centrifugation, we analyzed the supernatant using high-performance liquid chromatography (HPLC) with an EICOMPAK SC-5ODS column and a PREPAK column (Eicom, Kyoto, Japan) with electrochemical detection (ECD). The mobile phase comprised of 83% 0.1 M acetic acid in citric acid buffer (pH 3.5), 17% methanol (Wako), 190 mg/ml sodium 1-octanesulfonate (Wako), and 5 mg/ml disodium EDTA. For ECD, we applied a voltage of 750 mV against an Ag/AgCl reference electrode.

#### Statistical Analyses

Data are presented as the mean with error bars representing the standard error of the mean (SE). We analyzed data by the one-way analysis of variance (ANOVA), followed by the Tukey-Kramer post hoc test. All statistical analyses were performed with Ekuseru-Toukei 2012 software (Social Survey Research Information, Tokyo, Japan). We considered a p-value of <0.05 to be statistically significant.

#### RESULTS

#### Preventive Effects of IAA on Age-Related Memory Impairment in Aged Mice

To evaluate the effects of IAA on memory impairment in aged mice, we fed mice a diet containing IAA for 3 months at 7 weeks old (young mice) and 68 weeks old (aged mice) and subjected them to behavioral memory evaluations using the spontaneous alternation test and NORT. In the Y-maze spontaneous alternation test, aged mice showed significantly lower alternation scores compared with young mice (**Figure 1A**), whereas the number of total arm entries was not different between the experimental groups (**Figure 1B**). In NORT, the time spent exploring the novel object (**Figure 1C**) and DI (**Figure 1D**) in aged mice were significantly decreased compared with those in young mice, whereas the total time taken to approach each object did not differ between groups. These results indicated that spatial working memory and object recognition memory declined with aging. However, IAA-fed aged mice showed significantly higher spontaneous alternations in the Y-maze (**Figure 1A**) and DI in NORT (**Figure 1D**) compared with the control aged mice, which indicated that dietary IAA intake for 3 months prevented age-related memory impairment. Food and water intake among the groups did not differ significantly (**Supplementary Figures S1B,C**, respectively).

#### Preventive Effects of IAA on Inflammation in Aged Mice

To evaluate the effects of IAA on age-related inflammation in the brain, we measured the level of cytokines in the hippocampus and analyzed the phenotype of brain microglia. Aged mice showed significant increases in TNF-α and IL-1β in the hippocampus compared with young mice (**Figures 2A,B**, respectively). The percentage of CD11b-positive microglia producing TNF-α (as gated in **Figures 2C,D**) and the microglial expression of CD86 were significantly increased in aged mice compared with young mice (**Figures 2E,F**, respectively). These results suggested that age-related microglial-mediated inflammation was present in the brain. Long-term dietary IAA in aged mice was able to attenuate the age-induced increase in hippocampal TNF-α and IL-1β levels and microglial inflammation, indicating that IAA downregulated aged-related inflammation in the brain. These results indicate that aging induces inflammation in the mouse hippocampus and that this inflammation can be reduced by IAA consumption.

# Effects of IAA on Molecules Inducing Inflammation in Aged Mice

We measured the level of Aβ1–42, the ratio of pTau to total Tau (pTau/Tau), and the level of glutamate in aged mice to evaluate the effect of IAA on the production of inflammationinducing molecules. Aβ is a major component of senile plaques (Hsiao et al., 1996), and the phosphorylation of Tau is known to cause neurofibrillary tangles (Wood et al., 1986). Glutamate can mediate neurotoxic effects and can induce inflammation in the brain (Novelli et al., 1988). We found that the hippocampus of aged mice had significantly higher levels of both TBS-soluble Aβ1–42 and glutamate compared with the hippocampus of young mice (**Figures 3A,C**, respectively), but the ratio of pTau/Tau and the level of the inhibitory neurotransmitter GABA in the hippocampus did not differ between groups (**Figures 3B,D**, respectively). The age-related increases in Aβ and glutamate levels were significantly attenuated in aged mice provided with dietary IAA. These results indicated that Aβ1–42 and glutamate, which have both been linked to inflammation and cognitive decline in the brain (Lambert et al., 1998; Brown and Bal-Price, 2003), increase in the hippocampus with aging and that dietary IAA was able to reduce the age-related increase in these molecules.

# Effects of IAA on Monoamine Production in Aged Mice

DA is thought to be crucial for hippocampus-dependent memory (Li et al., 2003; Chan et al., 2017). To evaluate the effects of aging and IAA on monoamine production, we used an HPLC-ECD system to measure the levels of DA and DA metabolites, 3,4 dihydroxyphenylacetic acid (DOPAC), and homovanillic acid (HVA) in the hippocampus. The levels of DA and DOPAC were significantly decreased in aged mice compared with young mice (**Figures 4A,B**, respectively). However, the level of HVA and ratio of (DOPAC + HVA)/DA were not significantly altered by aging (**Figures 4C,D**, respectively). The level of norepinephrine did not differ between the experimental groups. These results indicated that the dopaminergic system function declined with aging but that the age-related decreases in DA and DOPAC could be significantly attenuated in aged mice that are administered dietary IAA.

# Effects of Short-Term IAA Intake on Memory Impairments in Aged Mice

We demonstrated that the long-term intake of IAA suppressed cognitive decline and inflammation in the brain of aged mice. Next, to evaluate the effects of the short-term intake of IAA

dietary IAA. (A–C) The amount of DA (A), 3, 4-dihydroxyphenylacetic acid (DOPAC; B), and homovanillic acid (HVA; C) in the hippocampus. (D) The ratio of (DOPAC + HVA)/DA. Data are presented as mean ± SE. We calculated the p-values shown in the graph by one-way ANOVA, followed by the Tukey-Kramer test. <sup>∗</sup>p < 0.05.

on memory function, we administered aged mice (22 months of age) IAA at 1 mg/kg for 9 days. In this experiment, both the spontaneous alternation score in the Y-maze and DI in NORT were significantly lower in aged mice than in young mice (7 months of age; **Figures 5A,D**, respectively). The total number of arm entries in the Y-maze did not differ between groups (**Figure 5B**). However, aged mice that received short-term oral administration of IAA showed significant improvements in the Y-maze spontaneous alternation score, the time taken to approach a novel object, and DI compared with control aged mice without orally administered IAA (**Figures 5A,C,D**, respectively). These results indicated that the short-term intake of IAA improved spatial and object recognition memory impairment that occurred in aging.

# DISCUSSION

Given the worldwide increase in aging populations, it is imperative to discover novel preventive and treatment therapies for cognitive decline. Our study demonstrated that the consumption of IAA reduced inflammation in the brain and prevented the cognitive impairment associated with normal aging in mice. Aged mice displayed microglial inflammation in the hippocampus and impairments in spatial working memory and object recognition memory; but both the inflammation and the cognitive impairments were reduced by long-term dietary administration of IAA. In addition to these long-term preventive effects, short-term IAA administration also improved age-related cognitive impairment in aged mice.

In aged mice, the Aβ level in the hippocampus was increased and occurred in conjunction with a microglial inflammatory response and cognitive decline. Aβ, a well-known agent of Alzheimer's disease (AD; Kametani and Hasegawa, 2018), is gradually deposited in the brain over a long period, inducing chronic inflammation and accelerating disease pathology (Heppner et al., 2015). In aged mice, microglial inflammation is induced by Aβ and other antigens; activated microglia produce pro-inflammatory cytokines and reactive oxygen species, both of which lead to cognitive impairment (Sarlus and Heneka, 2017; Wendeln et al., 2018). Glutamate, a neurotransmitter that can confer neurotoxicity, is also produced by activated microglia (Barger et al., 2007; Takaki et al., 2012). Increased glutamate in the hippocampus of aged mice is thought to be involved in memory impairment (Tamminga et al., 2012). The short-term spatial memory tested in the Y-maze and long-term object recognition memory tested in NORT are hippocampusdependent memory types (Cohen et al., 2013; Albani et al., 2014; Pioli et al., 2014). Inflammation in the hippocampus is reported to have impaired these types of memory, and the suppression of inflammation subsequently improved these memory impairments (Miwa et al., 2011; Abareshi et al., 2016). Evidence from these reports together with the findings in our present study suggest that dietary IAA is capable of suppressing microglial inflammatory responses in aged mice, which might be associated with the prevention of the decline in hippocampal memory associated with aging. The behavioral evaluations in the present study cannot discriminate which specific step in memory (acquisition, retention, or recall) was improved by IAA consumption. Further evaluation of hippocampus-dependent memory acquisition and retention using the Morris water maze or radial arm maze will further elucidate the effects of IAA on memory impairments with aging.

We have previously demonstrated, as a mechanism underlying the suppression of inflammation by IAA in aged mice, that IAA activates PPAR-γ (Yajima et al., 2004), increases the expression of the ''Aβ receptor'' CD36, promotes phagocytosis of Aβ (Yu and Ye, 2015), and promotes microglial activation toward the M2 anti-inflammatory type (Ano et al., 2017). In AD model mice, Aβ deposition and inflammation in the hippocampus were reduced by IAA administration (Ano et al., 2017). In the present study, the increased expression of CD86 and production of TNF-α in aging are characteristic of the M1 inflammatory type of microglia (Zhou et al., 2017). The administration of IAA reduced these characteristics, suggesting that IAA suppressed the activity of M1 microglia. These results suggest that IAA consumption shifts microglial activation toward an anti-inflammatory phenotype in aged mice by activating PPAR-γ. Some studies report that the longand short-term administration of PPAR-γ agonists, such as pioglitazone and rosiglitazone, suppressed cognitive decline in aged transgenic mice (Heneka et al., 2005; Escribano et al., 2010), but to date, the effects of PPAR-γ agonists on normally aged mice have not been demonstrated. In the present study, we demonstrated for the first time that IAA, a PPAR-γ agonist, prevent age-related cognitive decline in naturally aged mice.

Our study also demonstrated that the levels of DA and its metabolite DOPAC in the hippocampus of aged mice were reduced compared with those of young mice, and this reduction was ameliorated in aged mice by IAA. Levels of DA and DOPAC in the nucleus accumbens of aged mice were reported to be reduced when compared with those of young mice, whereas those in the striatum did not differ between groups, but to date, there has been no report on DA levels in the hippocampus of aged mice

#### REFERENCES


(Winner et al., 2017). In the hippocampus, DA is crucial for both spatial memory and object recognition memory. Inflammation in the brain is also reported to reduce the production of DA (Coffeen et al., 2010); accordingly, lipopolysaccharide endotoxin treatment reduced DA and DOPAC levels in rats (Noworyta-Sokolowska et al., 2013). We suggest that inflammation in the aging brain induces the reduction of DA level in the hippocampus, resulting in memory impairment.

In conclusion, our study demonstrated that consumption of IAA, the hop-derived component that imparts a bitter taste to beer, suppresses microglial activation and attenuates aged-related memory impairment in aged mice. This finding supports those of existing epidemiological studies and our previous research. Various brain disorders, including dementia, depression, and chronic fatigue, are associated with inflammation in the brain. The consumption of IAA might support the treatment or even reversal of various inflammation-related conditions such as cognitive decline.

#### DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

#### AUTHOR CONTRIBUTIONS

YA conducted the experiment and wrote the manuscript. RO performed the experiment and the behavioral evaluations. KK and HN designed and conducted the research.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi. 2019.00016/full#supplementary-material

FIGURE S1 | Weight and food/water consumption in aged mice. We fed C57BL/6J mice aged 7 weeks (young; n = 12) and 68 weeks (aged) diets containing 0% or 0.05% (w/w) IAA for 3 months (aged with dietary IAA, n = 10; aged without dietary IAA, n = 9). (A) Body weight was monitored every month. (B,C) Food (B) and water (C) consumption over 72 h. Data are presented as mean ± SE. We calculated the p-values shown in the graph by one-way analysis of variance (ANOVA), followed by the Tukey-Kramer test.


**Conflict of Interest Statement**: YA, RO and KK are employed by Kirin Company Ltd.

The remaining 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 Ano, Ohya, Kondo and Nakayama. 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.

# Emergence of Microglia Bearing Senescence Markers During Paralysis Progression in a Rat Model of Inherited ALS

Emiliano Trias<sup>1</sup>† , Pamela R. Beilby2,3† , Mariángeles Kovacs<sup>1</sup> , Sofía Ibarburu<sup>1</sup> , Valentina Varela<sup>1</sup> , Romina Barreto-Núñez<sup>1</sup> , Samuel C. Bradford<sup>3</sup> , Joseph S. Beckman2,3 \* and Luis Barbeito<sup>1</sup> \*

1 Institut Pasteur de Montevideo, Montevideo, Uruguay, <sup>2</sup> Department of Biochemistry and Biophysics, Oregon State University, Corvallis, OR, United States, <sup>3</sup> Linus Pauling Institute, Oregon State University, Corvallis, OR, United States

#### Edited by:

Alberto Javier Ramos, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Neil Cashman, University of British Columbia, Canada Kaoru Tominaga, Jichi Medical University, Japan

#### \*Correspondence:

Joseph S. Beckman joe.beckman@oregonstate.edu Luis Barbeito barbeito@pasteur.edu.uy; barbeito2006@gmail.com

†These authors have contributed equally to this work

Received: 12 December 2018 Accepted: 13 February 2019 Published: 28 February 2019

#### Citation:

Trias E, Beilby PR, Kovacs M, Ibarburu S, Varela V, Barreto-Núñez R, Bradford SC, Beckman JS and Barbeito L (2019) Emergence of Microglia Bearing Senescence Markers During Paralysis Progression in a Rat Model of Inherited ALS. Front. Aging Neurosci. 11:42. doi: 10.3389/fnagi.2019.00042 Age is a recognized risk factor for amyotrophic lateral sclerosis (ALS), a paralytic disease characterized by progressive loss of motor neurons and neuroinflammation. A hallmark of aging is the accumulation of senescent cells. Yet, the pathogenic role of cellular senescence in ALS remains poorly understood. In rats bearing the ALSlinked SOD1G93A mutation, microgliosis contribute to motor neuron death, and its pharmacologic downregulation results in increased survival. Here, we have explored whether gliosis and motor neuron loss were associated with cellular senescence in the spinal cord during paralysis progression. In the lumbar spinal cord of symptomatic SOD1G93A rats, numerous cells displayed nuclear p16INK4a as well as loss of nuclear Lamin B1 expression, two recognized senescence-associated markers. The number of p16INK4a-positive nuclei increased by four-fold while Lamin B1-negative nuclei increased by 1,2-fold, respect to non-transgenic or asymptomatic transgenic rats. p16INK4a-positive nuclei and Lamin B1-negative nuclei were typically localized in a subset of hypertrophic Iba1-positive microglia, occasionally exhibiting nuclear giant multinucleated cell aggregates and abnormal nuclear morphology. Next, we analyzed senescence markers in cell cultures of microglia obtained from the spinal cord of symptomatic SOD1G93A rats. Although microglia actively proliferated in cultures, a subset of them developed senescence markers after few days in vitro and subsequent passages. Senescent SOD1G93A microglia in culture conditions were characterized by large and flat morphology, senescence-associated beta-Galactosidase (SA-β-Gal) activity as well as positive labeling for p16INK4a, p53, matrix metalloproteinase-1 (MMP-1) and nitrotyrosine, suggesting a senescent-associated secretory phenotype (SASP). Remarkably, in the degenerating lumbar spinal cord other cell types, including ChAT-positive motor neurons and GFAP-expressing astrocytes, also displayed nuclear p16INK4a staining. These results suggest that cellular senescence is closely associated with inflammation and motor neuron loss occurring after paralysis onset in SOD1G93A rats. The emergence of senescent cells could mediate key pathogenic mechanisms in ALS.

Keywords: microglia, ALS, senescence, astrocytes, motor neurons, aging, SASP

# INTRODUCTION

fnagi-11-00042 February 26, 2019 Time: 15:10 # 2

Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative disease characterized by progressive upper and lower motor neuron degeneration, leading to muscle weakness and paralysis (Tsai et al., 2017). Although the etiology of ALS remains unknown, age is considered the strongest independent risk factor, most patients being diagnosed between the ages of 50 and 85 (Kiernan et al., 2011). ALS is also characterized by the ineluctable progression of motor deficits, with a variable but short survival of about 20 months (Hardiman et al., 2017). Age of diagnosis is also considered a strong predictor of survival, with hazard ratios progressively increasing each decade for individuals older than 50 years (Crockford et al., 2018). Age-dependence of motor phenotypes has also been described in rodent and fly models (Iguchi et al., 2013; Sreedharan et al., 2015), further supporting ALS as an aging-related condition.

Various studies indicate that motor neuron degeneration in ALS is often associated with increased oxidative and nitrative damage, mitochondrial dysfunction, ER-stress, defective RNA processing, and protein homeostasis (Cassina et al., 2008; Morgan and Orrell, 2016). In parallel, glial cells also become activated, proliferate and display inflammatory features characteristic of gliosis (Philips and Rothstein, 2014; Trias et al., 2018a). These kinds of cellular stresses combined with DNA damage or strong mitogenic signaling in vulnerable cells have the potential to induce cellular senescence (Rodier et al., 2009), a basic and heterogenous mechanism by which damaged cells adapt to maintain survival and prevent potentially deleterious expansion or oncogenic transformation during aging (Munoz-Espin and Serrano, 2014). A fundamental feature of cellular senescence is the arrest of the cell cycle through p16INK4A-mediated pathway, which is usually associated with p53 nuclear expression (Prieur et al., 2011). p53 becomes activated in response to a variety of cellular stressors including DNA damage and oxidative stress leading to an increased half-life of the p53 protein, phosphorylation and nuclear translocation. In turn, nuclear p53 can function as a transcription factor to regulate the cell cycle, apoptosis, genomic stability or senescence response (Rufini et al., 2013). Nuclear expression of p16INK4A is considered a robust molecular marker of cellular aging, as its expression increases in a variety of aged tissues (Baker et al., 2011). Another remarkable senescence-associated marker is the loss of nuclear Lamin B1 (Freund et al., 2012), which together with other lamins, is essential to maintain nucleus stability, size and shape (Dechat et al., 2008). The loss of nuclear Lamin B1 in particular is recognized as a senescence marker, functionally associated with the induction of p16INK4A and p53 (Freund et al., 2012).

In addition, senescent cells develop profound phenotypic and functional changes, including an increase in senescenceassociated beta-galactosidase (SA-β-Gal) activity, reflecting an increased number of lysosomes (Dimri et al., 1995). In addition, senescent cells enlarge and flatten with a tendency to form multinucleated cell aggregates (Leikam et al., 2015), accumulate oxidative and nitrative damage (Lamoke et al., 2015) and typically display a senescent-associated secretory phenotype (SASP) (Tchkonia et al., 2013), releasing trophic factors, pro-inflammatory signaling molecules, extracellular matrix components and proteases (Rodier et al., 2009). Recent evidence indicate cells expressing senescence markers contribute to the chronic inflammatory environment and progressive degeneration in different tissues from aged animals (Childs et al., 2015), thus acquiring pathogenic significance.

Previous studies in neurodegenerative conditions show that the emergence of glial and neuronal senescent phenotypes displaying inflammatory features contribute to synaptic and neuronal loss (Arendt et al., 1996; Frost, 2016), with the senescence marker p16INK4a being frequently found in a subpopulation of astrocytes (Bhat et al., 2012). In accordance, a senescence phenotype in human astrocytes can be induced by toxic species of amyloid beta in cell cultures (Bhat et al., 2012). Also, brain astrocytes bearing senescence markers have been identified in normal aging and disease conditions (Salminen et al., 2011; Chinta et al., 2013). Both in ALS animal models and patients, aged astrocytes develop senescence markers such as p16INK4A, p53, p21, and SA-β-gal, becoming toxic for motor neurons (Martin, 2000; Das and Svendsen, 2015; Turnquist et al., 2016), suggesting a causal pathogenic role in mediating motor neuron loss. To what extent activated microglia follow senescence-associated phenotypes during the course of paralysis progression in ALS remains to be analyzed.

Microgliosis is a recognized pathological feature in ALS patients (Brettschneider et al., 2012). Extensive microglia activation has also been described in transgenic rodent models of inherited ALS carrying SOD1 mutations (Lewis et al., 2014). In SOD1G93A rats, the rapid spread of paralysis is associated with marked microglial cell activation in the surroundings of motor neurons, leading to the emergence of aberrant phenotypes including astrocyte-like hypertrophic cells and giant multicellular clusters (Fendrick et al., 2007; Diaz-Amarilla et al., 2011; Trias et al., 2013). Activated microglia expressing mutant SOD1 in ALS have the potential to induce motor neuron death (Liao et al., 2012; Frakes et al., 2014). Removal of mutant SOD1 transgene from microglia and neurons significantly increases survival in of SOD1G37R mice (Boillee et al., 2006). The unique nature of microglia with the potential for self-renewal and telomere shortening led to the hypothesis that these cells can exhibit senescence (Eitan et al., 2014; Caldeira et al., 2017). Age-dependent and senescence-driven impairments of microglia functions and responses have been suggested to play essential roles during the onset and progression of neurodegenerative diseases (Luo et al., 2010; Spittau, 2017). However, it remains unknown whether deleterious gliosis and phenotypically aberrant glia in ALS are causally associated with the emergence of senescent cells in the degenerating spinal cord.

In this study, we analyzed the expression of senescence markers in the spinal cord and primary cultures of microglia from adult SOD1G93A rats. In an attempt to determine the relationship between the emergence of senescent glia phenotypes and progressive motor neuron loss, we analyzed senescence markers at disease onset and then at advanced paralysis, a time period of only 2 weeks while rapid paralysis develops in SOD1G93A rats.

# MATERIALS AND METHODS

fnagi-11-00042 February 26, 2019 Time: 15:10 # 3

#### Animals and Study Approval

All procedures using laboratory animals were performed in accordance with the international guidelines for the use of live animals and were approved by either the Oregon State University Institutional Animal Care Use Committee or for experiments performed in Uruguay in strict accordance with the requirements of the Institut Pasteur de Montevideo Bioethics Committee under the ethical regulations of the Uruguayan Law N◦ 18.611 governing animal experimentation. Uruguayan law follows the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (United States). Male hemizygous NTac:SD-TgN(SOD1G93A)L26H rats (Taconic), originally developed by Howland et al. (2002), were bred locally by crossing with wildtype Sprague–Dawley female rats. Male SOD1G93A progenies were used for further breeding to maintain the line. Rats were housed in a centralized animal facility with a 12-h light-dark cycle with ad libitum access to food and water. Symptomatic disease onset was determined by a periodic clinical examination for abnormal gait, typically expressed as subtle limping or dragging of one hind limb. Rats were killed well before they reached the end stage of the disease.

#### Experimental Conditions

At least three male rats were analyzed for each experiment. Four different conditions were studied as follow: (1) non-transgenic (NonTg) rats of 160–180 days; (2) transgenic SOD1G93A rats of 125–135 days (asymptomatic); (3) transgenic SOD1G93A rats of 170–180 days (onset); and (4) transgenic SOD1G93A rats of 190–200 days (symptomatic 15d paralysis).

### Determination of Disease Onset and End-Stage

As described previously (Trias et al., 2017), all rats were weighed and evaluated for motor activity daily. Disease onset was determined for each animal when pronounced muscle atrophy was accompanied by abnormal gait, typically expressed as subtle limping or dragging of one hind limb. When necessary, endstage was defined by a lack of righting reflexes or the inability to reach food and water.

#### Immunohistochemical Staining of Rat Spinal Cords

Animals were deeply anesthetized and perfused transcardially with 0.9% saline and 4% paraformaldehyde in 0.1 M PBS (pH 7.2–7.4) at a constant flow of 1 mL/min. The fixed spinal cord was removed, post-fixed by immersion for 24 h, and then cut into transverse serial 25 µm sections with a cryostat. Serial sections were collected in PBS for immunohistochemistry. Free-floating sections were permeabilized for 30 min at room temperature with 0.3% Triton X-100 in PBS, passed through washing buffered solutions, blocked with 5% BSA:PBS for 1 h at room temperature, and incubated overnight at 4◦C in a solution of 0.3% Triton X-100 and PBS containing the primary antibodies overnight at 4 ◦C. After washing, sections were incubated in 1:1000-diluted secondary antibodies during 3 h at room temperature. Using a stereological approach, p16INK4a-positive nuclei, Iba1-/p16INK4a positive cells, Lamin B1/DAPI and ChAT-/p16INK4a-positive cells were counted in 25-µm spinal cord sections using confocal microphotograph with a magnification of 25×. At least 15 sections per spinal cord were analyzed (n = 3). ImageJ software was used for analysis. For p53 quantification in the spinal cord, p53 density was measured using ImageJ. At least five sections per animal were analyzed (n = 3) as previously described (Trias et al., 2018b).

## Antibodies Used

Primary antibodies: 1:200 mouse monoclonal anti-CDKN2A/p16INK4a (abcam, #ab54210), 1:300 rabbit polyclonal anti-p53 (abcam, #ab131442), 1:400 rabbit polyclonal anti-MMP-1 (Novus Biologicals, #NBP1-72209), 1:300 mouse monoclonal anti-Iba1 (Merck, #MABN92), 1:400 muse monoclonal anti-CD68 (abcam, #ab31630), 1:400 rabbit polyclonal anti-ChAT (Merck, #AB143), 1:500 rabbit polyclonal anti-GFAP (Sigma, #G9269), 1:300 mouse monoclonal anti-S100β (Sigma, #S2532), 1:250 rabbit polyclonal anti-Lamin B1 (abcam, #ab16048), 1:300 mouse monoclonal anti-misfolded SOD1 B8H19 (Medimabs, # MM-0070-P), and 1:250 rabbit polyclonal anti-Nitro tyrosine (abcam, #ab42789). Secondary antibodies: 1:500 goat anti-rabbit-AlexaFluor488 or AlexaFluor546 (Thermo Fisher Scientific, #A11035 or #A11034), 1:500 goat anti-mouse-AlexaFluor488, AlexaFluor546 or AlexaFluor633 (Thermo Fisher Scientific, #A11029, #A11030, or #A21052).

# Microglia Cell Culture From Adult Symptomatic SOD1G93A Rats

Microglia cells were isolated from adult symptomatic SOD1G93A rats as previously described with slight modifications (Trias et al., 2013). Rats were terminally anesthetized and the spinal cords were dissected with the meninges carefully removed. The cords were mechanically chopped then enzymatically dissociated in 0.25% trypsin for 10 min at 37◦C. Fetal Bovine Serum (FBS) 10% (vol/vol) in Dulbecco's Modified Eagle Medium (DMEM) was then added to halt trypsin digestion. Repetitive pipetting thoroughly disaggregated the tissue, which was then strained through an 80-µm mesh and spun down. The pellet was re-suspended in culture medium [DMEM + FBS 10% (vol/vol), HEPES buffer (3.6 g/mL), penicillin (100 IU/mL), and streptomycin (100 µg/mL)] and plated in glass-bottom p35 culture dishes for confocal microscopy or 25-cm<sup>2</sup> tissue culture flasks for flow cytometry analysis. Culture medium was replaced every 48 h.

#### Analysis of Aberrant Glial Cells After Phenotypic Transformation

As previously characterized (Trias et al., 2013), primary adult microglia isolated from symptomatic SOD1G93A rats transitioned into aberrant glial cells after 12–15 days in culture. These aberrant glial cells can be maintained in culture for several passages (Diaz-Amarilla et al., 2011). In the present study, passages 2–4 of aberrant glial cells maintained in vitro (DMEM-10% FBS) in

glass-bottom p35 culture dishes for several days were analyzed for different senescent markers.

## Senescence-Associatedβ-Galactosidase (SA-β-Gal) Activity in Cell Cultures

Protocol for β-galactosidase staining was followed as described by manufacturer cell staining kit (Cell Signaling, #9860). Briefly, growth media was removed from the cells and washed with PBS. The 1X fixative solution was added for 15 min at room temperature. After two PBS washes, 1 mL of β-galactosidase staining solution was added overnight at 37◦C in a dry incubator. After blue color was developed, β-galactosidase staining solution was removed and plates were mounted using 70% glycerol for long-term storage at 4◦C. Both microglia and aberrant glial cells were analyzed at different time points during 12 days. 10×, 20×, and 100× images were acquired using an Olympus CX41 microscope connected to a EvolutionTMLC Color camera and using ImagePro Express software for acquisition. At least 10 fields per plate were acquired for quantitative analysis using ImageJ software.

#### Immunocytochemical Staining of Cultured Cells

Cultured cells were fixed with 4% PFA for 20 min at 4◦C and then were washed three times with 10 mM PBS (pH 7.4). Cells were permeabilized using 0.3% Triton-X100 for 20 min. Nonspecific binding was blocked by incubating fixed cells with 5% BSA in PBS for 1 h at room temperature. Corresponding primary antibodies were diluted in blocking solution and incubated 3 h at room temperature. After washing, cells were incubated with secondary antibodies in blocking solution for 1 h at room temperature. For p16INK4a and p53 staining, cells were permeabilized using 2M HCl solution during 15 min at room temperature before incubation with blocking solution. DAPI was used for nuclei staining. At least 10 fields per plate were acquired in a confocal microscope for quantitative analysis using ImageJ software.

## Flow Cytometry of Senescence-Associated-β-Galactosidase (SA-β-gal) Activity

After 12 days in vitro, microglia were quantitatively analyzed for SA-β-Gal activity. Briefly, cells were treated with Bafilomycin A1 to inhibit lysosomal acidification, followed by incubation with C12FDG (Molecular Probes/Life Technologies), a fluorogenic substrate for β-galactosidase for 2 h at 37◦C with 5% CO2. Microglia were then rinsed with PBS, harvested by trypsinization, centrifuged, and re-suspended in ice-cold PBS. Cells were immediately run on a Beckman-Coulter FC500 flow cytometer. Data were analyzed using Winlist (Verity Software).

# Flow Cytometry of Cell Cycle Progression

Cells were trypsinized, washed, and centrifuged. The cell pellet was then resuspended in ice-cold 70% ethanol and incubated at −20◦C for 30 min for fixation. Subsequently, cells were washed, centrifuged and re-suspended in 0.1% Triton X-100 in Dulbecco's Phosphate-Buffered Saline (DPBS). RNase A (10 µg/mL) and propidium iodide (20 µg/mL) were added and cells were incubated for 60 min at room temperature. They were then filtered through a 37-µm mesh and run on Beckman-Coulter FC500 flow cytometer and analyzed using Multi-Cycle (Phoenix Software).

## Fluorescence Imaging

Fluorescence imaging was performed with a laser scanning Zeiss LSM 800 confocal microscope with either a 25× (1.2 numerical aperture) objective or 63× (1.3 numerical aperture) oilimmersion objective using Zeiss Zen Black software. Maximum intensity projections of optical sections were created with Zeiss Zen software.

# Statistical Analysis

Quantitative data were expressed as mean ± SEM. Two-tailed Mann–Whitney test or Kruskal–Wallis followed by Dunn's multiple comparison tests were used for statistical analysis, with p < 0.05 considered significant. GraphPad Prism 7.03 software was used for statistical analyses.

# RESULTS

#### Expression of Senescence Markers p16INK4a and Lamin B1 in the Spinal Cord of SOD1G93A Rats During Paralysis Progression

Based on a previous report showing an increase of p16INK4a RNA levels in symptomatic SOD1G93A rats (Das and Svendsen, 2015), we examined the number of p16INK4a-positive nuclei and Lamin B1 expression in the ventral horn of the lumbar cord during paralysis progression. Immunohistochemistry analysis revealed a continuous increase in p16INK4a nuclear expression in rats expressing mutant SOD1 as compared with nontransgenic rats (**Figure 1A**). The number of p16INK4a-positive nuclei was significantly increased by 2.3-fold and 3.5-fold at paralysis onset and 15d of paralysis progression, respectively (graph in **Figure 1A**).

On the other hand, nuclear levels of Lamin B1 significantly declined during paralysis progression in SOD1 rats, 1 out of 4 nuclei exhibiting loss of Lamin B1 at 15d post-paralysis (arrows in **Figure 1D**), which is significantly different from nontransgenic and asymptomatic SOD1G93A rats (**Figures 1B,D**). Moreover, the decline in Lamin B1 expression and nuclear Lamin B1 invaginations (asterisk in **Figure 1D**) were associated with aberrant nuclear shapes (arrowheads in **Figure 1D**).

Because cellular senescence is characterized by cell cycle arrest through p16INK4a- and p53-mediated pathways (Prieur et al., 2011), we also assessed p53 expression in the lumbar ventral horn. As shown in **Supplementary Figure S1**, p53 immunoreactivity significantly increased in mutant SOD1 rats at paralysis onset and advanced paralysis with frequent colocalization of p16INK4a

(**Figure 1C**). p53 expression levels increased by 1.5- and 2-fold at onset and advanced paralysis, respectively, with respect to age-matched non-transgenic littermates (Graph in **Supplementary Figure S1**).

# Nuclear p16INK4a and Lamin B1 Expression in Spinal Cord Microglia During Paralysis Progression

Next, we analyzed whether p16INK4a and Lamin B1 were expressed in Iba1-positive microglia, that typically proliferate and become hypertrophic near spinal motor neurons in symptomatic SOD1G93A rats (Trias et al., 2013). As shown in **Figure 2A**, Iba1-positive microglia express high levels of nuclear p16INK4a (white arrows) in rats developing paralysis. Compared with nontransgenic controls, p16INK4a expression at onset and 15d of paralysis progression significantly increased by 2.6- and 4.8-fold, respectively (graph in **Figure 2A**). Remarkably, a high density of p16INK4a nuclei was identified in multinucleated microglia clusters (**Figure 2B**) that are frequently found in the ventral horn of symptomatic SOD1G93A rats (Fendrick et al., 2007), further indicating the correlation of senescence with microglia bearing aberrant phenotypes.

In addition, nuclear expression of Lamin B1 progressively declined in Iba1-positive cells during advance paralysis. **Figure 2C** shows subpopulation of Iba1-positive microglia that devoid of nuclear Lamin B1 (arrows) coexisting with microglia displaying normal pattern of Lamin B1 staining (arrowheads). Furthermore, nuclear Lamin B1 decline was observed in senescent multinucleated microglia clusters in the lumbar spinal cord (**Figure 2D**).

Next, we analyzed whether misfolded SOD1 was associated with senescent microglia in SOD1 rats. Misfolded SOD1 is a recognized hallmark of neuronal pathology in ALS linked to

microphotograph shows a cluster of multinucleated microglia where Lamin B1 expression is absent in several nuclei (DAPI) at 15d post-paralysis. Scale bar: 20 µm.

SOD1 mutations (REF). As shown in **Supplementary Figure S2**, misfolded SOD1 was mainly detected in degenerating neuronal somas and dendrites in symptomatic SOD1G93A rats and was not observed in non-transgenic or asymptomatic transgenic rats. However, the presence of misfolded SOD1 in microglia appeared to correspond to neuronal debris being engulfed by phagocytic microglia (arrows in **Supplementary Figure S2**).

### Nuclear p16INK4a Staining in a Subset of Spinal Motor Neurons and Astrocytes During Advanced Paralysis

Previous reports have shown astrocytes bearing senescent markers in the spinal cord of symptomatic SOD1G93A rats (Das and Svendsen, 2015) as well as in post-mitotic neurons submitted to stress or aging (Jurk et al., 2012). Thus, we looked for p16INK4a-expressing astrocytes and motor neurons in the lumbar spinal cord of SOD1G93A rats during onset and 15d of paralysis progression. As shown in **Figure 3A**, a subset of ChAT-positive motor neurons expressed significant levels of nuclear p16NK4a during the period of rapid motor neuron loss in advanced paralysis. In comparison, motor neurons bearing healthy morphology in asymptomatic SOD1G93A rats were negative to p16NK4a, suggesting senescence develops only in damaged motor neurons. Nuclear p16INK4a was also observed in numerous GFAP-positive astrocytes that typically surround motor neurons in the ventral horn of symptomatic SOD1G93A rats (**Figure 3B**).

#### Senescence-Associated β-Galactosidase Activity (SA-β-gal) in Primary Cultures of Microglia From Symptomatic SOD1G93A Rats

We have previously shown that primary spinal cord cultures from symptomatic SOD1G93A rats yield >98% of microglia (Trias et al., 2013). **Figure 4A** summarizes the behavior of these microglia cultures and its ability to actively proliferate and transform into flat enlarged cells after serial passages. In this context, we explored whether cultured microglia from symptomatic SOD1G93A rats could develop senescence markers as observed in the degenerating spinal cord. Primary cultures of SOD1G93A microglia maintained for 12 daysin vitro progressively developed positive chromogenic SA-β-gal staining, with ∼8-fold increase between 1 DIV and 12 DIV (**Figure 4B**). Senescent microglia in cell cultures demonstrated an enlarged, flattened morphology (arrows in **Figure 4B**), morphological features previously described in other senescent cells (Carnero, 2013).

Flow cytometer analysis of microglia maintained in culture for 12 DIV showed 50% of the cells exhibiting SA-β-gal fluorescent staining (**Supplementary Figure S3**), with two distinct cell subpopulations, based on size as seen in the scatter diagram of the cells (**Supplementary Figure S3**). The subpopulation of smaller cells displays only 8% of SA-β-Gal activity and normal cell cycle behavior, corresponding to non-senescence cells (**Figures 4C,D**). In contrast, 92% of large size cells exhibited SA-β-gal activity and also significant S-phase arrest (**Figures 4E,F**), the latter being usually associated with inhibition of cell growth, proliferation and senescence in cell cultures (Blagosklonny, 2011).

#### Expression of Senescence Markers in Cell Cultures of SOD1G93A Microglia

Next, we analyzed the phenotypic features of Iba1- and CD68 positive SOD1 microglia at 2- and 12-DIV to identify senescence cellular markers. As shown in **Figures 5A,B**, approximately 50% of microglia expressed p16INK4a or p53 nuclear staining at 12DIV, as compared with approximately 15% at 2DIV (graphs in **Figures 5A,B**, and **Supplementary Figures S4A,B**). 12DIV microglial cells also displayed high levels of MMP-1 and NO2Tyr in comparison with 2DIV isolated cells (**Supplementary Figure 4C**). In addition, p16INK4a, p53, MMP-1, and NO2Tyr were also found in multinucleated cell aggregates that are frequently found in culture conditions (**Figure 5C**), reproducing the aberrant features found in the degenerating spinal cord in vivo.

The emergence of senescent cells was also observed in serially passaged SOD1G93A microglia cultures, which have undergone a phenotypic transformation (Trias et al., 2013). As shown in **Supplementary Figure 5A**, the number of SA-β-Gal-positive cells rapidly increased in the following 5 days after plating, ∼50% of these cells also displaying increased p16INK4a and p53 nuclear staining (**Supplementary Figures 5B,C**).

### DISCUSSION

Amyotrophic lateral sclerosis has been modeled as a multi-step process associating senescence-driven tissue dysfunction with underlying genetic defects and risk factors (Al-Chalabi et al., 2014). In this context, here we report that paralysis progression in a rat model of ALS is characterized by the emergence of numerous microglia, astrocytes and motor neurons displaying phenotypic markers of senescence. Senescent cells seem to be acutely induced after paralysis onset, suggesting a deleterious effect mediated by the ALS neurodegenerative cellular microenvironment and coincident to motor neuron loss. Senescence markers were also observed in cultures of microglia isolated from symptomatic SOD1G93A rats, further indicating the inherent ability of these cells to develop a senescence program with secretory features. In agreement with previous reports showing senescence microglia in aged rodents (Rawji et al., 2016; Theriault and Rivest, 2016), the present data show evidence of a yet unknown mechanism associating microglia activation and cell senescence, with the emergence of secretory phenotypes in a rat model of ALS.

Activation of the p16INK4a-pathway is essential for the induction of senescence in a variety of cell types (Prieur et al., 2011). The tumor suppressor p53 also contribute to the induction of cellular senescence in glial cells (Turnquist et al., 2016). We found that the basal levels in p16INK4a and p53 expression were significantly increased in SOD1G93A rats at asymptomatic and paralysis onset stages, respect to age-matched non-transgenic controls. Strikingly, p16INK4a and p53 levels sharply increased after paralysis onset, coincident with extensive spinal cord microgliosis and motor neuron loss occurring in SOD1G93A rats

(Howland et al., 2002). Increased levels of p16INK4a and p53 were shown to induce nuclear loss of Lamin B1 (REF). Such a decline in Lamin B1 level constitutes a recognized biomarker of cellular senescence (REF). This is the first report showing a significant increase in nuclear Lamin B1 loss in the degenerating spinal cord of SOD1G93A rats, which was associated with other pathological features of Lamin B1 and nuclear misshape. Senescent microglia showed Lamin B1 loss as well as abnormalities in nuclear Lamin B1 localization pattern. These findings agree with previous reports showing disruption of nuclear Lamin B1 in neural cells associated with Parkinson's disease and Tau pathologies (Frost et al., 2016; Chinta et al., 2018).

p16INK4a expression and nuclear Lamin B1 decline in microglia were typically observed in cells surrounding the damaged motor neurons. These cells also displayed large size, multinucleated formations as well as MMP-1 and nitrotyrosine staining in culture, suggesting phenotypic aberrations and secretory features. Thus, senescent microglia emerging in the degenerating spinal cord may explain the origin of aberrant glial phenotypes previously described during paralysis progression in SOD1G93A (Diaz-Amarilla et al., 2011; Trias et al., 2013). Taking together, these observations suggest that senescence microglia may result as a consequence of microglia activation, which involves the production of inflammatory mediators and oxidative

population.

R1 indicates the percentage of total population encompassed by this subset (45%). The gate for the smaller cell population indicates almost 8% of these cells are senescent. (D) The diagram shows the cell cycle analysis for the smaller cell population. (E) Scatter diagram for larger cell population (inside the white outline, R2). In the larger cell population, over 90% of the cells demonstrate SA-β-Gal activity. (F) The scatter diagram to the right shows the cell cycle analysis for the larger cell

considered statistically significant. Scale bar: 50 µm. (C) After 12 DIV, SOD1G93A isolated microglia form Iba1-/CD68-positive multinucleated giant cells, which express several senescence markers such as p16INK4a, p53, and MMP1. Also, these multinucleated cells express high levels of NO2Tyr. Scale bars: 20 µm.

stress with potential genotoxic activity (Spittau, 2017). Thus, p53 induction in activated microglia from paralytic SOD1G93A rats might not be only related to the senescence program but may also contribute to modulate the inflammatory phenotype as previously described (Aloi et al., 2015).

The finding that microglia isolated from symptomatic SOD1G93A rats develop senescence features in culture conditions further support the inherent ability of these cells to undergo a senescence program. As cultures aged during several days, an increasing number of cells displayed senescence markers such

as SA-β-Gal activity, p16INK4a, and MMP-1. SA-β-Gal activity, commonly used to distinguish senescent cells (Dimri et al., 1995), is perceptible due to the increased lysosomal content present in senescent cells (Kurz et al., 2000). Interestingly, senescent microglia in vitro expressed MMP-1, a marker of SASP (Strzyz, 2016), suggesting this phenotype could define a specific type of microglia polarization in ALS. Levels of matrix metalloproteinases increase with age in many tissues and organs and are associated with the SASP (Freund et al., 2010). MMP-1 levels in glial cells have been shown to be increased in Alzheimer's disease pathology (Bhat et al., 2012). Recent studies also suggest that metalloproteinases become increasingly dysregulated during disease progression in ALS (Soon et al., 2010), although this has not yet been considered in connection to cell senescence. In addition, senescent microglia isolated from symptomatic SOD1G93A rats showed a tendency to develop cell fusion and multinucleation. This cellular atypia has been previously described in the degenerating spinal cord of SOD1G93A rats (Fendrick et al., 2007). Our finding of microglia bearing SASP is in accordance with our previous reports in microglia in SOD1G93A rats displaying increased transcriptional expression of senescence-associated cytokines and inflammatory factors (Trias et al., 2016), as well as ultrastructural alterations in organelles occurring in cell senescence (Jimenez-Riani et al., 2017).

Cultures containing senescent microglia from symptomatic SOD1G93A rats were characterized by the fact that the emergence of senescent cells was coincident with a robust proliferation capacity of neighboring cells, which could be passaged many serial passages, as previously described (Diaz-Amarilla et al., 2011). Here, we have identified by flow cytometry that senescent microglia exhibited large size and cell cycle arrest, clearly differentiating from a subpopulation of smaller, SAβ-Gal-negative cells, with high proliferative capacity. Thus, SASP microglia in SOD1G93A rats could strongly promote the proliferation of neighboring non-senescent microglia by secretion of soluble factors. In accordance, we have shown that transplantation of SOD1G93A microglia into discrete sites of the lumbar spinal cord on non-transgenic rats, induced a massive microgliosis along the entire spinal cord (Ibarburu et al., 2017).

Because cultured SOD1G93A microglia from the rat paralytic spinal cord shows a high degree of activation, oxidative/nitrative stress and expression of inflammatory genes (Boillee and Cleveland, 2008; Thonhoff et al., 2012), we speculate that the triggering of the senescence program is a consequence of exacerbated cell damage or genotoxic stress, rather than aging per se. In accordance, we found that senescent microglia accumulate nitrotyrosine in proteins, indicating oxidative stress producing tyrosyl-radical formation and nitric oxide production (Zhao et al., 2004; Thonhoff et al., 2012). Increased levels of nitrotyrosine residues have been associated with endogenous production of peroxynitrite, a potent cellular oxidant and nitrating agent (Ischiropoulos et al., 1992; Pacher et al., 2007), which has not been previously associated with cellular senescence. In accordance, inflammatory stimulation of macrophages involving increase production of nitric oxide and superoxide also results in p16INK4a expression and SA-β-Gal activity (Hall et al., 2017).

Finally, we found evidence that motor neurons and astrocytes also express nuclear p16INK4a during the symptomatic stage, which might be related to the intriguing accumulation of misfolded SOD1 in motor neuron during advanced paralysis in SOD1G93A rats. This agrees with previous reports showing senescent neurons in aged mice and animal models of Alzheimer's disease (Jurk et al., 2012; Musi et al., 2018). Because neurons can develop a SASP, they can contribute to induce in inflammation in neighboring cells through the secretion of soluble factors (Appel et al., 2011; Komine and Yamanaka, 2015). Similarly, the finding of senescent astrocytes expressing nuclear p16INK4a in symptomatic SOD1G93A rat spinal cord suggest a role of defective astrocytes in ALS pathology. Astrocytes might exert their neurotoxic effect on motor neurons via the SASP, releasing several proinflammatory cytokines and trophic factors, such as

IL-6 (Haidet-Phillips et al., 2011; Das and Svendsen, 2015) and NGF species (Pehar et al., 2004). Astrocytes in ALS rodent models express different senescence markers which potentially turn them into a neurotoxic phenotype for motor neurons both in vitro and in vivo (Das and Svendsen, 2015; Turnquist et al., 2016). Thus, senescence-associated phenotypes in glial cells and neurons might be relevant pathogenic mechanisms in ALS. It remains unknown, however, whether prevention or eradication of senescence cells in ALS could result in delayed disease progression as has been reported in other neurological diseases (Bussian et al., 2018).

#### CONCLUSION

In conclusion, as summarized in **Figure 6**, here we show for the first time that senescent and secretory microglia emerge during paralysis progression in a rat model of inherited ALS. Risk factors such as aging together with mitochondrial dysfunction and nitro-oxidative damage linked to inflammation likely promote the emergence of senescent glial cells. Subsequently, senescent cells may promote profound changes in the cellular microenvironment through SASPs, exacerbating progressive neuroinflammation and motor neuron toxicity.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

#### AUTHOR CONTRIBUTIONS

ET, PB, LB, and JB designed the research. ET, PB, MK, SI, VV, RB-N, and SB performed the research. ET, PB, SB, LB, and JB analyzed the data. ET, PB, LB, and JB wrote the paper.

#### FUNDING

This work was supported by Institut Pasteur de Montevideo – FOCEM Mercosur (COF 03/11), the Amyotrophic Lateral Sclerosis Association (00482), Department of Defense (AL140108), Agencia Nacional de Investigación e Innovación (ANII), and Programa de Desarrollo de las Ciencias Básicas (PEDECIBA).

#### ACKNOWLEDGMENTS

We want to thank the staff from the Transgenic and Experimental Animal Unit from Institut Pasteur de Montevideo. Finally, we

#### REFERENCES

Al-Chalabi, A., Calvo, A., Chio, A., Colville, S., Ellis, C. M., Hardiman, O., et al. (2014). Analysis of amyotrophic lateral sclerosis as a multistep process: a population-based modelling wish to acknowledge the Confocal Microscopy Facility of the Center for Genome Research and Biocomputing at Oregon State University.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi. 2019.00042/full#supplementary-material

FIGURE S1 | Progressive increase in spinal cord p53 expression during paralysis progression Confocal microphotographs show the staining for p53 (red) among analyzed groups. The graph to the right shows the quantitative analysis of p53 intensity. Note the increase in p53 expression with disease progression. Data are expressed as mean ± SEM; data were analyzed by Kruskal–Wallis followed by Dunn's multiple comparison tests, p < 0.05 was considered statistically significant. Scale bars: 50 µm.

FIGURE S2 | Interaction of microglia with degenerating motor neurons expressing misfolded SOD1. Confocal microphotograph showing Iba1-positive microglia clusters (red) surrounding damaged motor neurons accumulating high levels of misfolded SOD1 (green). Arrowheads indicate the microglia/motor neuron clustering. Note that misfolded SOD1 is mainly expressed in neuronal structures during paralysis, while its expression in microglia appears to be associated with the phagocytosis of misfolded SOD1 contained in degenerating neuronal structures (arrows). Scale bar: 20 µm.

FIGURE S3 | Senescence-associated β-Galactosidase activity in primary cultures of microglia from symptomatic SOD1G93A rats. The scatter diagram, a population density heat map, indicates the gate for the sample and includes the entire population of cells. The diagram to the right shows that approximately 50% of the cells demonstrate SA-β-activity.

FIGURE S4 | Expression of senescence markers p16INK4a and MMP1 in cultured adult microglia from SOD1G93A symptomatic rats. Immunocytochemistry analysis of senescence markers on microglia isolated from SOD1G93A symptomatic rats. (A) Isolated Iba1-positive microglia after 2 days in culture express nuclear p16INK4a (A) and p53 (B) in a small subpopulation of cells. Arrows indicate the respective nuclear localization of both markers. Scale bar: 20 µm. (C) Progressive increase of MMP1 and NO2Tyr in adult cultured microglia. Note the increased expression of MMP1 and NO2Tyr between 2 DIV (upper panel) and 12 DIV (lower panel). Scale bar: 20 µm.

FIGURE S5 | Serially passaged SOD1G93A microglia cultures express senescence markers. Senescence marker analysis in phenotypic transitioned SOD1G93A microglia in culture. (A) Transitioning microglia population display increasing SA-β-Gal activity (red arrows) at different time points (Passage 4). The graph to the right shows the quantitative analysis of SA-β-Gal activity in transformed microglia. Data are expressed as mean ± SEM; data were analyzed by Kruskal–Wallis followed by Dunn's multiple comparison tests, p < 0.05 was considered statistically significant. (B) After several days in culture, transformed microglia express increasing levels of p16INK4a and p53. Also, note the high expression of NO2Tyr in those cells that express nuclear p16INK4a . Graphs to the right show the quantitative comparative analysis of p16INK4a and p53 at different time points. Data are expressed as mean ± SEM; data were analyzed by Kruskal–Wallis followed by Dunn's multiple comparison tests, p < 0.05 was considered statistically significant. Scale bars: 20 µm.

study. Lancet Neurol. 13, 1108–1113. doi: 10.1016/S1474-4422(14)70 219-4

Aloi, M. S., Su, W., and Garden, G. A. (2015). The p53 Transcriptional Network Influences Microglia Behavior and Neuroinflammation. Crit. Rev. Immunol. 35, 401–415.


**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 Trias, Beilby, Kovacs, Ibarburu, Varela, Barreto-Núñez, Bradford, Beckman and Barbeito. 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.

fnagi-11-00042 February 26, 2019 Time: 15:10 # 14

# IGF1 Gene Therapy Modifies Microglia in the Striatum of Senile Rats

Eugenia Falomir-Lockhart <sup>1</sup> , Franco Juan Cruz Dolcetti <sup>1</sup> , Luis Miguel García-Segura2,3 , Claudia Beatriz Hereñú<sup>4</sup> \* † and Maria Jose Bellini <sup>1</sup> \* †

<sup>1</sup>Laboratorio de Bioquimica del Envejecimiento, Instituto de Investigaciones Bioquímicas de La Plata (INIBIOLP), Facultad de Ciencias Médicas, UNLP-CONICET, La Plata, Argentina, <sup>2</sup> Instituto Cajal, CSIC, Madrid, Spain, <sup>3</sup>Centro de Investigación, Biomédica en Red de Fragilidad y Envejecimiento Saludable (CIBERFES), Instituto de Salud Carlos III, Madrid, Spain, 4 Instituto de Farmacología Experimental de Córdoba-CONICET, Departamento de Farmacología, Facultad de Ciencias Químicas, UNC-CONICET, Córdoba, Argentina

#### Edited by:

George E. Barreto, Pontificia Universidad Javeriana, Colombia

#### Reviewed by:

Andrzej Bartke, Southern Illinois University School of Medicine, United States Francisco G. Wandosell, Severo Ochoa Molecular Biology Center (CSIC-UAM), Spain

#### \*Correspondence:

Claudia Beatriz Hereñú cherenu@fcq.unc.edu.ar Maria Jose Bellini mariajosebellini@yahoo.com; mariajosebellini@med.unlp.edu.ar

†These authors have contributed equally to this work

Received: 27 November 2018 Accepted: 19 February 2019 Published: 05 March 2019

#### Citation:

Falomir-Lockhart E, Dolcetti FJC, García-Segura LM, Hereñú CB and Bellini MJ (2019) IGF1 Gene Therapy Modifies Microglia in the Striatum of Senile Rats. Front. Aging Neurosci. 11:48. doi: 10.3389/fnagi.2019.00048 Microglial cells become dystrophic with aging; this phenotypic alteration contributes to basal central nervous system (CNS) neuroinflammation being a risk factor for age related neurodegenerative diseases. In previous studies we have observed that insulin like growth factor 1 (IGF1) gene therapy is a feasible approach to target brain cells, and that is effective to modify inflammatory response in vitro and to ameliorate cognitive or motor deficits in vivo. Based on these findings, the main aim of the present study is to investigate the effect of IGF1 gene therapy on microglia distribution and morphology in the senile rat. We found that IGF1 therapy leads to a region-specific modification of aged microglia population.

Keywords: IGF1, microglia morphology, striatum, aging, gene therapy

#### INTRODUCTION

Microglia, the immune cells of the central nervous system (CNS), suffer a phenotypic alteration during aging that is characterized by decreased motility and inefficient surveillance (Streit et al., 2004; von Bernhardi et al., 2016; Koellhoffer et al., 2017). This dystrophic cellular phenotype of microglia is associated with the loss of their neuroprotective function, which contributes to increased basal CNS neuroinflammation with aging and may represent a risk factor for cognitive and motor impairment, depression or diverse age related neurodegenerative diseases (Streit and Xue, 2010; Cunningham, 2013; Patel et al., 2015; Pekny and Pekna, 2016; Ransohoff, 2016; Spittau, 2017).

A feasible approach to modulate microglia function in the aged brain is the use of neurotrophic factors that polarize these cells into a more neurotrophic/neuroprotective phenotype. Among these is insulin like growth factor 1 (IGF1; Arevalo et al., 2010; Suh et al., 2013; Acaz-Fonseca et al., 2014; Labandeira-Garcia et al., 2017), which exerts neuroprotective actions in the CNS (Piriz et al., 2011; Torres Aleman, 2012; Morel et al., 2016), including the aged brain (Piriz et al., 2011; Deak and Sonntag, 2012; Labandeira-Garcia et al., 2017). Previous studies have shown the viability of IGF1 gene therapy to target brain cells in vivo (Hereñú et al., 2009) and to

**Abbreviations:** IGF1, insulin like growth factor 1; RAd, recombinant adenovirus; DsRed, Discosoma Red fluorescent protein; CNS, central nervous system.

decrease behavioral functional impairments in aged rats (Nishida et al., 2011; Pardo et al., 2016). In the present study we report that IGF1 therapy lead to a region-specific modification in microglia number and morphology in the aged brain.

#### MATERIALS AND METHODS

#### Adenoviral Vectors

We employed recombinant adenoviral vectors (RAd) previously constructed in our laboratory (Hereñú et al., 2007) as carriers to deliver either the therapeutic cDNA of IGF1 gene (RAd-IGF1) or the red fluorescent protein from Discosoma sp DsRed (RAd-DsRed).

## Animals and Experimental Procedures

Female senile Sprague-Dawley rats (28 months old) were used. These rats have a maximum life expectancy of 36 months (Mansilla et al., 2016). Animals were housed in a temperaturecontrolled room (22 ± 2 ◦C) on a 12:12 h light/dark cycle and fed ad libitum, with a standard chow diet containing 12.08 kJ/g calories: 69.5% from carbohydrates, 5.6% from fat, and 24.9% from protein (Association de Cooperativas Argentinas-S.E.N.A.S.A. No. 04-288/A). All experiments with animals were performed according to the Animal Welfare Guidelines of NIH (INIBIOLP's Animal Welfare Assurance No A5647-01). The ethical acceptability of the animal protocols used here has been approved by our institutional IACUC (Protocol #T09-01-2013).

On day 0 (D0), rats were anesthetized with ketamine hydrochloride (40 mg/kg; i.p.) plus xylazine (8 mg/kg; i.m.) and placed in a stereotaxic apparatus. Rats were randomly divided into two groups (n = 10 per group): DsRed group, which received an injection of RAd-DsRed; and IGF1 group, which received an injection of RAd-IGF1. Bilateral injections in the lateral ventricles were performed placing the tip of a 26 G needle fitted to a 10 µL syringe at the following coordinates relative to the Bregma: −0.8 mm anteroposterior, −4.2 mm dorsoventral and ±1.5 mm mediolateral (Paxinos and Watson, 2007). Rats were injected bilaterally with 8 µL per side of a suspension containing 10<sup>10</sup> plaque forming units (pfu) of the appropriate vector. Body weight was determined every 2 or 3 days from day −5 before surgery to experimental day 18. Animals were sacrificed at experimental day 18.

#### Immunohistochemistry

Animals were placed under deep anesthesia and perfused with phosphate buffered paraformaldehyde 4%, (pH 7.4) fixative. The brains were removed and stored in paraformaldehyde 4%, (pH 7.4) overnight at 4◦C. Brains were kept in cryoprotective solution at −20◦C until use. For immunohistochemical assessment, brains were cut coronally in 40 µm-thick sections with a Vibratome (Leica).

All immunohistochemical techniques were performed on free-floating sections under moderate shaking. Washes and incubations were done in 0.1 M phosphate buffer pH 7.4, containing 0.3% triton X-100 (washing buffer). The endogenous peroxidase activity was quenched for 15 min at room temperature in a solution of 3% hydrogen peroxide in 50% methanol. After several washes in buffer, sections were incubated overnight at 4◦C with an Iba1 rabbit polyclonal antibody diluted 1:1,000 (WAKO CTG2683), marker of microglia/macrophages. Sections were then washed in buffer and incubated for 2 h at room temperature with an anti-rabbit biotinylated secondary antibody (1:1,000, BA-1000; Vector Labs). After several washes in buffer, sections were incubated for 90 min at room temperature with avidin-biotin-peroxidase complex (diluted 1:500, PK-6100; Vector ABC Elite Kit). The reaction product was revealed by incubating the sections with 3,3-diaminobenzidine (Sigma-Aldrich) and 0.01% hydrogen peroxide in 0.1 M phosphate buffer. Then, sections were dehydrated, mounted on gelatinized slides with mounting medium (Vectamount, Vector) and used for image analysis.

#### Morphometric Analysis

From each rat, 1 in every 12 brain serial sections was selected. For stereological analysis, we used an Olympus BX-51 microscope attached to an Olympus DP70 CCD video camera (Tokyo, Japan). All morphological parameters were assessed bilaterally. The density of microglia in the striatum (caudate-putamen), dorsal tier of substantia nigra pars compacta (SNCD), motor cortex, dorsolateral entorhinal cortex (DLEnt) and perirhinal cortex (PRh) was assessed. Areas of interest were defined in accordance with the rat brain atlas of Paxinos and Watson (2007). All Iba1-immunoreactive cells were manually quantified, according to the optical disector method, using a counting frame of 83 × 83 µm at 600× magnification. A total of 40–85 counting frames were analyzed. Cells in the uppermost focal plane and/or intersecting the exclusion boundaries of the counting frame were not counted. Cells counts are expressed as number/mm<sup>3</sup> .

Iba1-immunoreactive cells were classified as: (1) nonreactive/ramified microglia; cells with small cell body and few large to no branches; and (2) reactive/ameboid microglia; cells with large or ameboid cell bodies, numerous retracted processes and intense Iba1 immunostaining (Acaz-Fonseca et al., 2015). The proportion of reactive vs. total microglia was determined for each brain region analyzed.

## Statistical Analysis

Data shown in the figures are presented as the mean ± standard error of the mean (SEM). The size of the experimental groups is indicated in each figure legend. Gaussian distribution of data sets was assessed by Kolmogorov-Smirnov test. Statistical analysis was performed by using the software GraphPad Prism 6 (GraphPad Software). To determine significant differences in microglia densities and reactivity between groups, we used t-test analysis. P-values < 0.05 were considered to be significant.

# RESULTS

#### Body Weight Gain

Animal weight was recorded every 2 or 3 days. At day two after administration of RAd-DsRed or RAd-IGF1, both groups experienced a transient loss of body weight due to the surgical procedure. Subsequently, both groups of rats recovered the initial

Striatum/caudate-putamen) and (K; PRh, perirhinal cortex; DLEnt, dorsolateral entorhinal cortex; and SNCD, substantia nigra compact part dorsal) show a representation of the coronal brain sections where the analyzed regions; are highlighted. Panels (J,L) show microglia densities in the different groups in the analyzed regions. Data are given as means ± SEM (N = 5/group). <sup>∗</sup>Significant differences (p < 0.05).

weight and remained stable until the end of the experiment (**Supplementary Figure S1**).

#### Gene Therapy Affects the Number and Reactivity of Microglia

Iba-1 immunoreactive cells were analyzed in the motor cortex, striatum, substantia nigra, DLEnt and PRh. As shown in **Figure 1**, significant differences in the number of Iba-1 immunoreactive cells were detected in the striatum between the animals injected with RAd-DsRed and the animals injected with RAd-IGF1. Thus, there was a higher number of Iba-1 immunoreactive cells in the animals injected with RAd-IGF1 than in the animals injected with RAd-DsRed (**Figure 1**). No significant differences were detected in the number of Iba-1 immunoreactive cells in the motor cortex, substantia nigra and the PRh of the animals injected with RAd-IGF1 and the animals injected with RAd-DsRed (**Figure 1**).

In addition to the differences in total number, obvious qualitative differences were observed in the morphology of Iba-1 immunoreactive cells between the different experimental groups; with more cells with a reactive phenotype in the striatum of animals injected with Rad-IGF1 (**Figure 2**). Indeed,

(N = 5/group). <sup>∗</sup>Significant differences (p < 0.05).

the quantitative analysis of reactive/ameboid cells and nonreactive/ramified cells showed a significant increase in the proportion of reactive cells in the striatum of the animals injected with RAd-IGF1 in comparison with the striatum of RAd-DsRed injected rats. In contrast, no significant differences in the proportion of reactive and non-reactive phenotypes were detected in the other brain regions analyzed (**Figure 2**).

In summary, these findings indicate that the injection of RAd-IGF1 is able to increase the number of Iba-1 immunoreactive cells and the proportion of Iba-1 immunoreactive cells with a reactive phenotype in a specific brain region.

#### DISCUSSION

We previously demonstrated that ICV IGF1 gene therapy is effective to transduce brain ependymal cells with high efficiency, achieving effective release of transgenic IGF1 into the cerebrospinal fluid (CSF; Hereñú et al., 2009) and restoring motor performance in aged animals (Nishida et al., 2011). Thus, we decided to use the ependymal route to implement IGF1 gene therapy in aging rats with the goal to characterize microglia number and distribution in different brain regions related with cognition or motor performance.

ICV gene therapy caused a significant but transient decrease in weight gain in both DsRed and IGF1 groups. Body weigh may be affected by surgery and also by a temporary induction of endogenous neurotrophic factors, such as GDNF, which has been reported to cause body weight loss (Morel et al., 2010). This effect could be greater in the IGF1 group, since IGF1 may regulate the hypothalamic system controlling body weight and energy expenditure (Werner and LeRoith, 2014).

We expected that IGF1 gene therapy will decrease microglia reactivity, since this therapy has been shown to reduce the reactivity of astrocytes in response to proinflammatory stimuli in vitro (Bellini et al., 2011). And to exert neuroprotective and neuroreparative actions in experimental animal models of stroke (Zhu et al., 2008; Liu et al., 2017). However, other studies have reported detrimental actions of IGF1 receptor signaling in the brain of mouse models of Alzheimer's disease (Cohen et al., 2009; Freude et al., 2009; Gontier et al., 2015). In this study we found is that IGF1 gene therapy increased the number and the proportion of microglia with a reactive phenotype in a region dependent manner.

We have not a definitive explanation for the regional effect of RAd-IGF1 on striatal microglia. It is well established that microglia (Breese et al., 1996; Walter et al., 1999; Chesik et al., 2004; Suh et al., 2013; Rodriguez-Perez et al., 2016; Trueba-Saiz et al., 2017) and other cell types in the brain (Fernandez and Torres-Alemán, 2012) express IGF1 receptors. It is possible that regional differences in the expression of these receptors may cause a different sensitivity to IGF1 released and delivered to the ventricles by infected ependymal cells.

The functional consequences of these changes in microglia are unknown. However, it has been described that treatments that leads to the depletion of microglia in injury models increases neuronal death (Vinet et al., 2012). Reciprocally, repletion of microglia or introduction of exogenous microglia result in neuronal rescue, driven possibly by microglial production of trophic growth factors, or via the clearance of deleterious byproducts of metabolism and neurotransmission (Streit et al., 2004; Nissen, 2017). Therefore, it is possible that the increased number of microglia cells induced by IGF1 gene therapy may exert a protective function in the striatum of older rats. Indeed, previous studies have shown that the administration of RAd-IGF1 in the brain, following the protocol used in the present study, results in an improvement in motor function of senile rats (Nishida et al., 2011).

In summary, our findings indicate that ICV IGF1 gene therapy specifically modifies the number and phenotype of microglia in the striatum of senile rats, suggesting that such a therapy may be useful to alter the function of dystrophic microglia, at least in specific brain regions. Further studies should determine whether the effect of IGF1 gene therapy on microglia is age-dependent and is also detected in male animals.

#### DATA AVAILABILITY

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

#### AUTHOR CONTRIBUTIONS

EF-L and MB designed the experiments. EF-L and FD performed the experiments. EF-L, FD, LG-S, CH and MB analyzed the data. EF-L, LG-S, CH and MB wrote the manuscript.

#### FUNDING

This study was supported in part by grants #PICT13-1119 from the Argentine Agency for the Promotion of Science and Technology and grant PIP0618 from the Argentine Research

#### REFERENCES


Council (CONICET) to MB, grant M184 from the Universidad Nacional de La Plata to CH and grant from Agencia Estatal de Investigación, Spain (BFU2017-82754-R), CIBERFES and Fondos Feder to LG-S.

#### ACKNOWLEDGMENTS

We thank Natalia Scelsio for technical assistance as well as to Oscar Vercellini, Araceli Bigres for animal care and to Mario Ramos for assistance with graphics design.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi. 2019.00048/full#supplementary-material

FIGURE S1 | Time curve of body weight gains of RAd-DsRed and RAd-IGF1 rats. <sup>∗</sup>Significant difference (p < 0.05). N = 5/group.

amyloid-β clearance. J. Neurosci. 35, 11500–11513. doi: 10.1523/JNEUROSCI. 0343-15.2015


in hippocampal excitotoxicity. J. Neuroinflammation 9:27. doi: 10.1186/1742- 2094-9-27


**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 Falomir-Lockhart, Dolcetti, García-Segura, Hereñú and Bellini. 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.

# Astrocyte Heterogeneity: Impact to Brain Aging and Disease

Isadora Matias † , Juliana Morgado† and Flávia Carvalho Alcantara Gomes\*

Laboratory of Cellular Neurobiology, Institute of Biomedical Sciences, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

Astrocytes, one of the largest glial cell population in the central nervous system (CNS), play a key function in several events of brain development and function, such as synapse formation and function, control of neurotransmitters release and uptake, production of trophic factors and control of neuronal survival. Initially described as a homogenous population, several evidences have pointed that astrocytes are highly heterogeneous, both morphologically and functionally, within the same region, and across different brain regions. Recent findings suggest that the heterogeneity in the expression profile of proteins involved in astrocyte function may predict the selective vulnerability of brain regions to specific diseases, as well as to the age-related cognitive decline. However, the molecular mechanisms underlying these changes, either in aging as well as in brain disease are scarce. Neuroinflammation, a hallmark of several neurodegenerative diseases and aging, is reported to have a dubious impact on glial activation, as these cells release pro- and anti-inflammatory cytokines and chemokines, anti-oxidants, free radicals, and neurotrophic factors. Despite the emerging evidences supporting that reactive astrocytes have a duality in their phenotype, neurotoxic or neuroprotective properties, depending on the age and stimuli, the underlying mechanisms of their activation, cellular interplays and the impact of regional astrocyte heterogeneity are still a matter of discussion. In this review article, we will summarize recent findings on astrocyte heterogeneity and phenotypes, as well as their likely impact for the brain function during aging and neural diseases. We will focus on the molecules and mechanisms triggered by

#### Edited by:

Maria Jose Bellini, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Argentina

#### Reviewed by:

Kyoungho Suk, Kyungpook National University, South Korea Luca Steardo, Università degli Studi della Campania Luigi Vanvitelli Caserta, Italy

#### \*Correspondence:

Flávia Carvalho Alcantara Gomes fgomes@icb.ufrj.br

†These authors have contributed equally to this work

> Received: 08 January 2019 Accepted: 01 March 2019 Published: 19 March 2019

#### Citation:

Matias I, Morgado J and Gomes FCA (2019) Astrocyte Heterogeneity: Impact to Brain Aging and Disease. Front. Aging Neurosci. 11:59. doi: 10.3389/fnagi.2019.00059

**98**

**Abbreviations:** AD, Alzheimer's disease; Aldh1L1, aldehyde dehydrogenase 1 family, member L1; ALS, amyotrophic lateral sclerosis; ATP, adenosine triphosphate; AQP4, aquaporin-4; APs, amyloid plaques; Aβ, amyloid-β; AβO, Aβ oligomers; BBB, blood-brain barrier; BDNF, brain-derived neurotrophic factor; BLBP, brain lipid binding protein; CaM-kinase, Ca2+/calmodulin-dependent protein kinase; CNS, central nervous system; Cx30, connexin-30; Cx43, connexin-43; C1q, complement component 1q; C3, complement component 3; C4b, complement component 4b; EAAT1, excitatory amino acid transporter 1; EAAT2, excitatory amino acid transporter 2; ER, endoplasmic reticulum; GABA, gamma-aminobutyric acid; GFAP, glial fibrillary acidic protein; GLAST, astrocyte-specific glutamate-aspartate transporter; GLP1R, glucagon-like peptide-1 receptor; GLT-1, glutamate transporter-1; GDNF, glial-derived neurotrophic factor; GS, glutamine synthetase; HD, Huntington's disease; IFs, intermediate filaments; IFN-γ, interferon-γ; IL-6, interleukin-6; IL-8, interleukin-8; IL1R1, interleukin 1-receptor 1; IL-1α, interleukin-1α; IL-1ß, interleukin-1ß; JAK2, janus kinase 2; Kir4.1, inwardly rectifying potassium channel 4.1; LPS, lipopolysaccharide; LTD, long-term depression; LTP, long-term potentiation; NGF, nerve growth factor; PD, Parkinson's disease; RG, radial glia cells; SNpc, substantia nigra pars compacta; SPARC, secreted protein-acidic and rich in cysteine; STAT3, signal transducer and activator of transcription 3; TGF-β, transforming growth factor-β; TGF-β1, transforming growth factor-β1; TLR4, toll-like receptor 4; TNF-α, tumor necrosis factor-α; TNT, tunneling nanotubes; TSP, thrombospondin; TSP-1, thrombospondin-1; WHO, World Health Organization.

astrocyte to control synapse formation in different brain regions. Finally, we will discuss new evidences on how the modulation of astrocyte phenotype and function could impact the synaptic deficits and glial dysfunction present in aging and pathological states.

Keywords: astrocyte, glial reactivity, aging, heterogeneity, neurodegenerative diseases

#### ASTROCYTES: AN OVERVIEW

The term ''astrocytes'' comprises a heterogeneous group of non-neuronal cells, involved in fundamental functions of the central nervous system (CNS). Even though astrocytes tile the entire brain, these cells have not always been considered active partners of neurons in the transfer of neural information as they are now, in the beginning of the XXI century.

In the XIX century, Virchow coined the term nervenkitt (neuroglia), to refer to the passive, connective elements in the brain, in which the other elements, the excitable ones, were embedded (Somjen, 1988). Further discoveries about the nature of neural cells came in the end of the XIX century, with new techniques of tissue staining developed by Italian physician and cytologist, Camillo Golgi and the Spanish neurohistologist, Ramón y Cajal. Golgi and Cajal were the first to highlight that neuroglia and nerve cells represented different populations and to further identify a variety of glial shapes and forms, as well as the glial network formed by these cells and other non-neuronal cells, such as the glial endfeet in close proximity to blood vessels (De Carlos and Borrell, 2007). At the end of the XIX century, the Hungarian anatomist and histologist, Lenhossék, introduced the term astrocyte to refer to a star-shaped glial cell; he raised the concept that even though astrocytes were electrically silent, they had functions as important as nerve cells (Somjen, 1988; Verkhratsky and Butt, 2013; Verkhratsky and Nedergaard, 2018).

In the last decades, an increasing amount of data has provided new insights on the plethora of functions performed by astrocytes. In the healthy tissue, these cells occupy unique spaces in which their extensive branching of fine processes occupy contiguous non-overlapping domains (Bushong et al., 2002). Their processes can contact synapses, other glial cells, blood vessels, and depending on the brain area, they have more specific roles. One of these key roles is in synaptic regulation, as astrocytes can act not only in the formation and maturation of synapses (Diniz et al., 2014a), but also in the maintenance, pruning and remodeling of synapses in the development, aging and diseases (Chung et al., 2013, 2015, 2016; Liddelow et al., 2017). Beyond the well-established concept of the tripartite synapse, in which perisynaptic astroglial processes are fundamental participants in the synapse, along with the pre- and post-synaptic components (Araque et al., 1999), it is now argued that the astroglial synaptic coverage could be far more extensive. It is theorized that the astroglial perisynaptic processes form a ''synaptic cradle'' around the synapse, embracing it, allowing the astrocyte to provide proper maintenance of the synapse, maintaining neurotransmitter, ion and volume homeostasis, releasing neuromodulators and keeping the specificity of the signaling, and providing synaptic isolation (Verkhratsky and Nedergaard, 2014, 2018). As astrocytes express a diversity of neurotransmitters' receptors and transporters, they can control the levels and activity of several neurotransmitters such as glutamate, gammaaminobutyric acid (GABA), adenosine triphosphate (ATP) and D-serine (Rothstein et al., 1994, 1996). Some of these molecules, also known as gliotransmitters, can be secreted by astrocytes in the synaptic space and activate neuronal receptors, allowing astrocytes to act as modulators of neuronal activity (Volterra and Meldolesi, 2005).

Emerging evidence from the last decade have strongly challenged the concept that brain function is a result of solely neuronal networks' activity (Araque and Navarrete, 2010; Di Castro et al., 2011). Changes in synaptic function can cause the release of other gliotransmitters by astrocytes, such as thrombospondins (TSPs; Christopherson et al., 2005; Eroglu and Barres, 2010), hevin and secreted protein-acidic and rich in cysteine (SPARC; Kucukdereli et al., 2011), glypicans, and cytokines, including tumor necrosis factor-α (TNF-α; Stellwagen and Malenka, 2006) and transforming growth factor-β1 (TGFβ1; Diniz et al., 2012, 2014b). This could happen because neural activity cause excitability in the astrocyte's membrane, triggering the release of intracellular Ca2+, which in turn, causes the release of some gliotransmitters (Perea and Araque, 2007; Araque et al., 2014); however, the mechanisms that govern astrocytes' influence on synaptic activity are not completely clear yet. Further, neuronal activity has also been implicated in the generation of spontaneous syncytial signaling through calcium waves in astrocytes (Shigetomi et al., 2010; Di Castro et al., 2011), even though this is not entirely dependent on neuronal activity (Nett et al., 2002). Physiological stimulation creates a flux of Ca2<sup>+</sup> ions, changing the Ca2<sup>+</sup> concentration throughout cell compartments, thus generating the waves. The waves can spread out (Kanemaru et al., 2014) and propagate the signal from one cell to the other through gap junction channels (De Bock et al., 2014). This could represent a specific kind of cell-cell communication, especially for long-range information transfer in astrocytic syncytia (Scemes and Giaume, 2006). However, the proper physiological role of calcium waves and the mechanism through which it happens remains to be explored.

Besides their roles in the synaptic function, astrocytes are responsible for ion homeostasis in the CNS, regulating the extracellular levels of potassium, chlorine and calcium ions, water homeostasis and maintenance of the cellular pH. They are also responsible for providing energetic metabolites required by neurons in neural networks, such as glucose (Rouach et al., 2008) and lactate (Qu et al., 2000; Figley, 2011; Sotelo-Hitschfeld et al., 2015), and trophic factors essential for neuronal survival and differentiation (Gomes et al., 1999, 2005; Martinez and Gomes, 2002; Nones et al., 2012; Dezonne et al., 2013). Additionally, astrocytes maintain intimate contact along with endothelia and pericytes, thus contributing to the formation and function of the blood-brain barrier (BBB; Siqueira et al., 2018). More recently, a novel function has been attributed to astrocytes as part of the glymphatic system, in which astrocyte's endfeet contact the vasculature, surrounding it and forming a system of tunnels, through which compounds such as glucose and amino acids are distributed, and the excess of toxic waste products is removed, such as large proteins including β-amyloid and tau (Weller et al., 2009; Iliff et al., 2012; Jessen et al., 2015).

Given the large number of functions performed by astrocytes **(Figure 1)**, it is to be expected that deficits in these cells have a major impact on brain functioning. Such impact, however, due to the great diversity amongst these cells, vary among different brain areas and structures. This review article will focus on astrocyte diversity, highlighting the main findings on their morphological, functional and molecular differences amongst brain regions. We further discuss how such heterogeneity might contribute to shape the way astrocytes respond to different insults, such as those that rise from aging and in specific neurodegenerative diseases. Finally, we will argue that modulation of astrocyte phenotypes and functions, might shed light on new avenues to control aging and neural pathology.

## MOLECULAR AND FUNCTIONAL HETEROGENEITY OF ASTROCYTES IN THE HEALTHY BRAIN

The neurocentric view of the brain has persisted for many years, and several studies have identified numerous neuronal

FIGURE 1 | Schematic representation of the main astrocyte functions in the central nervous system (CNS). Astrocytes in the healthy murine CNS are highly ramified cells responsible for a number of functions in the CNS development, homeostasis and function. Glial fibrillary acidic protein (GFAP; in red) and DAPI (in blue) staining of murine hippocampus tissue. Scale bars, 20 µm.

populations that differ molecularly and functionally (Kepecs and Fishell, 2014; Jiang et al., 2015; Mancinelli and Lodato, 2018). Even though astrocytes are in close contact to neurons, it was not clear whether astrocytes would be as diverse as neurons. Astrocyte heterogeneity was an underexplored topic for many years, and just recently, the plurality of functions that they can exert in different neuronal circuits has shed a light on the fact that perhaps these cells have more specialized roles locally than previously thought.

In the mammal CNS, it has long been recognized a variety of astrocyte subtypes that differ regarding their developmental origin, morphology, physiology and metabolism amongst regions (Zhang and Barres, 2010; Oberheim et al., 2012). The first evidences of morphological heterogeneity of astrocytes were introduced by Golgi and Cajal (for review, Garcia-Lopez et al., 2010), and by William Lloyd Andriezen, World Health Organization (WHO), in 1893, described two different types of glial cells present in white and gray matter, the fibrous and protoplasmic glia, respectively (Andriezen, 1893), even though he thought these cells had different developmental origins. Since then, numerous evidence have corroborated the concept that protoplasmic and fibrous astrocytes are different subsets of astrocytes based on morphologic and molecular criteria (Vaughn and Pease, 1967; Raff et al., 1984; Raff, 1989). The protoplasmic astrocytes represent the biggest population of astrocytes in the gray matter, mainly found in the hippocampus and cerebral cortex; their cell bodies are extremely ramified, which probably allows them to touch numerous synapses typical of these regions, thus performing a neuromodulatory role (Bushong et al., 2002; Oberheim et al., 2012). The fibrous astrocytes are organized along white matter tracts; they are smaller and have lesser branching points than the protoplasmic astrocytes and can contact the nodes of Ranvier, thus contributing to keep the homeostasis in the region (Lundgaard et al., 2014).

Besides these two main subtypes of astrocytes, other astrocyte-like cells are described: radial glia cells (RG), initially described as merely playing a role in neuronal migration during cerebral cortex development, today RG are known as the main neuronal/glial progenitor present during brain development (Sild and Ruthazer, 2011); Bergmann glia, astrocyte type specific for cerebellum that ensheaths and controls cerebellar synapses, and is involved in granular cell migration (Rakic, 1971; Gregory et al., 1988; Grosche et al., 2002); Müller glia, astrocyte type specific for retina, involved in cell migration, neuronal generation and control of synapses of the retina (Reichenbach, 1989; Reichenbach and Bringmann, 2010). These specific subtypes of astrocytes and their heterogeneity will not be the scope of this review, but they provide an insight on how diverse the astrocyte population can be, especially as these subsets can vary amongst themselves.

The astrocytes that populate the human CNS are unique. Not only the cortical protoplasmic astrocytes are nearly three-fold larger in diameter, they also extend ten times more primary processes than their rodent counterparts and contact nearly 100-fold more synapses located in their territorial domains (Oberheim et al., 2009). The human brain also has exclusive subtypes of astrocytes, such as the interlaminar astrocytes, which cell body is located at the top cortical layer and sends long processes towards the deeper layers (Colombo et al., 1995) and the varicose projection astrocytes, which processes have many varicosities and extend 1–5 very long fibers towards all directions in the cerebral cortex (Oberheim et al., 2009, 2012). However, the whole extension of this diversity and its functional significance still remains to be investigated.

In rodents, by the end of gestation, RG-astrocyte differentiation is characterized, among several molecular mechanisms, by the replacement of RG markers, such as brain lipid binding protein (BLBP) and the intermediate filament (IF) protein, nestin, by astrocytic markers such as vimentin, the glial fibrillary acidic protein (GFAP), the glutamate transporter GLAST and the calcium binding protein S100β (Pixley and de Vellis, 1984) as discussed below. The correct timing of RG-astrocyte transformation is a crucial step to ensure correct number of neurons and cerebral cortex lamination. All things considered, the identification of astrocytic markers is crucial to the proper visualization of these cells, which may be the first step towards understanding how diverse they are. Although GFAP has become a classical marker for the identification of mature astrocytes, it has many limitations. One of the main ones is that it is not broadly expressed by all astrocytes across brain areas in the healthy CNS (Walz and Lang, 1998; Eng et al., 2000; Hol and Pekny, 2015).

GFAP has at least eight different isoforms generated by different splicing patterns, and they may be variably expressed in specific subsets of astrocytes (Middeldorp and Hol, 2011). Perhaps due to its importance in modulating astrocyte motility and shape by providing structural stability to astrocytic processes, GFAP expression is highly regulated including in injury and disease (Eng et al., 2000), and during aging in specific areas of the rodent brain, such as the corpus callosum, basal ganglia and hippocampus (Morgan et al., 1997, 1999) and in humans (Nichols et al., 1993). Another limitation is that GFAP has also been detected in peripheral glia, such as enteric glia (Kato et al., 1990), Schwann cells (Bianchini et al., 1992), and other non-neuronal cells, such as fibroblasts, and myoepithelial cells (Hainfellner et al., 2001). Other astrocytic markers include the glutamate transporters such as GLT-1 excitatory amino acid transporter 2 (EAAT2) and GLAST EAAT1, that stain a variety of astrocytes' subtypes, e.g., RG, Bergmann glia, Müller cells, etc. (Schmitt et al., 1997; Williams et al., 2005), glutamine synthetase (GS) which stains a wide array of astrocytes' subtypes in many regions where GFAP is not an effective marker (Anlauf and Derouiche, 2013); the glycoprotein S100β that is a more broadly expressed marker of astrocytes than GFAP in some regions (Ogata and Kosaka, 2002). However, these markers are less specific to astrocytes than GFAP, as S100β also stains oligodendrocytes and ependymal cells (Hachem et al., 2005; Steiner et al., 2007); GLT-1, GLAST and GS expression was also reported in neurons and oligodendrocytes (Cammer, 1990; D'Amelio et al., 1990; Schmitt et al., 2002). A few other markers have also been used to identify astrocytes such as aquaporin 4 (AQP4) and connexins 30 (Cx30) and Cx43, though these markers are mostly concentrated at astrocytes' endfeet, rather than throughout the cell soma (Nagy et al., 1999; Nagelhus and Ottersen, 2013).

The advent of genetics era in recent decades and identification of the transcriptional profile of astrocytes, represented a new tool to study the diversity of the astroglial lineage and provided useful insights in the physiological state of these cells. The genetic profiling of astrocytes has identified a wider array of new markers such as the aldehyde dehydrogenase 1 family, member L1 (Aldh1L1) gene. This metabolic enzyme is responsible for the folate metabolism, which is especially important during neurulation. Aldh1L1 is widely expressed throughout the whole cell, and it seems to be expressed by most astrocytes, but not by other cell types (Cahoy et al., 2008), which might make it a better tool to visualize astrocytes in the developing mouse brain.

All markers identified so far, have showed some kind of specificity depending on the area: while Aldh1L1 was reported mainly staining cortical astrocytes (Waller et al., 2016); GS labeled mainly astrocytes in the entorhinal cortex (EC) comparing to GFAP (Anlauf and Derouiche, 2013), whereas astrocytes in the hippocampus are widely stained for GFAP (Bushong et al., 2002; Chai et al., 2017).

Altogether, although these data suggest that there is some kind of selectivity of astrocytic protein expression among different areas, the lack of a universal astrocyte marker or specific astrocytic regional markers still represents a challenge for the identification of intrinsic differences between astrocytes from distinct brain regions.

Recently, Chai et al. (2017) demonstrated that astrocytes from the murine striatum and hippocampus differ morphologically, functionally and molecularly. They found that even though the cell density in both regions were equally high, astrocytes in the striatum have a larger territorial size and contact nearly twice as many neuronal somas than hippocampal astrocytes, but the latter display stronger and more numerous physical interactions with excitatory synapses. At the functional level, these cells displayed different Ca2+-signaling dynamics and gap-junctional coupling. The transcriptomic and proteomic analysis of the astrocytes from these distinct regions showed that hippocampal and striatal astrocytes are molecularly distinct cell populations. Genes responsible for proteins with functional importance in astrocytes such as the ones encoding GLT-1, SPARC, the potassium channel inwardly rectifying potassium channel 4.1 (Kir4.1) and a sodiumdependent GABA transporter were amongst the 40 most highly expressed genes common between the hippocampus and striatum, but the most highly expressed gene in the hippocampal astrocyte was the one encoding GFAP, and in the striatal astrocyte was that encoding the protein µcrystallin, a protein related to the regulation of the thyroidhormone T3 (Vié et al., 1997), whose proper function in the brain is still unknown. Although these data corroborated the already known limitations of GFAP as an astrocyte marker, it suggests that µ-crystallin might be a molecular marker specific to striatal astrocytes (Francelle et al., 2015; Chai et al., 2017). Interestingly, they also found a gradient of expression of µ-crystallin within the striatum, suggesting that these cells may differ even depending on their subregion (Chai et al., 2017). Such differences have already been observed among astrocytes from the CA1 and CA3 subareas of the hippocampus demonstrating different physiological properties (D'Ambrosio et al., 1998), even though expressing a similar molecular marker profile (Sharif et al., 2004).

The functional implication of astrocyte heterogeneity in different neural circuits is still a matter of discussion. However, considering the key role of astrocytes in synapse formation, maturation and maintenance in different brain regions, it is likely that interactions between astrocytes and synapses vary accordingly to regional demands. Our group has recently compared the synaptogenic properties of astrocytes from four different brain regions: cerebral cortex, hippocampus, midbrain and cerebellum. We found that distinct populations of astrocytes have distinct synaptogenic profiles due to the differential expression of synaptogenic molecules such as TSP-1, glypicans 4 and 6, hevin, SPARC, TNF-α, brainderived neurotrophic factor (BDNF) and TGF-β1 (Buosi et al., 2018). The synaptogenic properties of these astrocytes might reflect the differences in the requirement of astroglia's synaptic coverage amongst different brain regions: some regions, like the cerebellum, have nearly all their synapses covered by astrocytes endfeet, whilst less than 50% of cortical and hippocampal synapses are tripartite synapses (Lippman et al., 2008; Witcher et al., 2010). Further, the pool of synaptogenic and anti-synaptogenic molecules produced by astrocytes might contribute to shape distinct neural circuits. In the cerebral cortex, for example, astrocytes can control the balance between excitatory and inhibitory synapses by activating different downstream signaling pathways, as described by our group. Astrocytes secrete TGF-β1, which induces the formation of either excitatory synapses via activation of D-serine production (Diniz et al., 2012), or inhibitory synapses via the Ca2+/Calmodulin-dependent protein kinase (CAM-kinase) pathway (Diniz et al., 2014b).

Together, data discussed here supports the idea that astrocytes display distinct inter- and intra-regional features, however, the functional significance of this is still a matter of discussion. The fact that astrocytic processes fill the local environment in non-overlapping domains suggest that a potential advantage of region-specified astrocytes might be their capacity to locally regulate neural circuit function (Emsley and Macklis, 2006; Oberheim et al., 2012). This raises a few unanswered questions: Can astrocyte heterogeneity primarily determine susceptibility of brain regions to different insults? Does astrocyte diversity impact the way brain regions age? Do astrocytes' dysfunctions contribute equally to neurodegenerative diseases throughout the nervous system? In the next sections, we will discuss some recent advances and alternative explanations for the distinct vulnerability of these cells to different insults, diseases and aging amidst brain areas and the roles that astrocytes may play in these processes.

#### HETEROGENEITY OF ASTROCYTE'S RESPONSE TO DIVERSE INSULTS: REACTIVE ASTROCYTES AND THEIR DIFFERENT PHENOTYPES

Astrocytes acquire different phenotypes in response to numerous pathological stimuli, such as stroke, neurodegenerative disorders, tumors, trauma, infection, ischemia and aging. These can trigger a response known as astrocyte reactivity, characterized by changes in the profile of astrocytes' gene expression, leading to both morphological and functional changes that may vary from cellular hypertrophy, or atrophy, to proliferation and scar formation. The extent of the astrocytic reaction is severely impacted by the size of the affected area, the nature of the insult and its severity, the intensity of BBB disruption and the inflammatory response.

The first concept of astrocyte reactivity emerged with Virchow, who characterized a necrotic spinal lesion surrounded by a thick, highly fibrillary scarring, probably composed by densely packed astrocyte processes (Virchow, 1856). Even though astrocytes themselves were not formally described at that moment, accumulating pieces of evidence since then resulted in the classical hallmarks of reactive astrocytes: overexpression of GFAP, which frequently is related to the degree of reactivity, hypertrophy of cell body and increase in the number of cellular processes (Bignami and Dahl, 1976; Eng et al., 2000; Anderson et al., 2016). More recently, additional reactive astrocyte markers have been described, such as lipocalin-2 (Lcn-2), an acute phase protein (Lee et al., 2009), overexpressed by reactive astrocytes induced not only by ischemic stroke and lipopolysaccharide (LPS) exposure (Zamanian et al., 2012), but also by inflammation and excitotoxicity (Chia et al., 2011); and Serpin3n, a serine protease inhibitor (Zamanian et al., 2012), that was already reported to have its levels increased after injury in the rat facial and hypoglossal nuclei (Gesase and Kiyama, 2007). Identification of specific astrocytic reactivity markers is a crucial step towards the understanding of the unique mechanisms underlying this cellular process in different diseases.

Recently, the idea of astrocyte reactivity has been revisited. Rather than an all-or-none phenomenon, the process of becoming activated is gradated, and may vary from subtle changes to a deep irreversible morphological alteration, regulated in a context-dependent manner (Wilhelmsson et al., 2006; Sofroniew, 2009; Anderson et al., 2014). Further, although it is recognized that astrocytes can present gain or loss of function at the site of the insult, it is not clear whether this is beneficial or damaging to their surroundings (Sofroniew, 2009; Pekny et al., 2014). The reason for this controversy is due to the fact that apart from the protective roles that can be enhanced by astrocyte's activation, such as production of anti-inflammatory factors, these cells can also acquire a toxic reactive phenotype, thus producing cytokines that exacerbate injuries in the spinal cord (Brambilla et al., 2009), increasing β-amyloid production (Nagele et al., 2003), or becoming atrophied and losing their neuroprotective functions in Alzheimer's disease (AD), as will be discussed soon (Diniz et al., 2017).

Astrocytic reactivity is coordinated by complex molecular mechanisms, mainly initiated through an inflammatory response to the insults, in which gene expression and cellular changes are regulated by a series of inter- and intracellular signaling. Zamanian et al. (2012), was one of the first to describe reactive astrogliosis from a molecular standpoint. By using two injury models, a middle cerebral artery occlusion to induce ischemia and a systemic LPS injection, they found that reactive astrogliosis consisted of a rapid change in gene expression, but the reactive astrocyte phenotype varied according to the type of insult. Both models induced inflammation and caused concomitantly microglia activation and reactive astrogliosis, although by different mechanisms. They found that over 1,000 genes related to many different biological processes had their levels at least twice as high, when compared to quiescent astrocytes, and 260 of these were induced at least four-fold more in expression, hinting at the highly complex change that the astrocytic activation represents. They also found high levels of IF proteins such as GFAP and vimentin, and of many cytokines and their receptors, such as interleukin-6 (IL-6) and IL-1 receptor 1 (IL-1R1), confirming a few already described hallmarks of reactive gliosis. They found that the insults had overlapping but distinct sets of induced genes: from 263 reactive glial genes identified, 150 were preferentially expressed by ischemic stroke-reactive astrocytes, and 57 by LPS reactive astrocytes, and while ischemia induced a reactive astrocyte phenotype that was more protective, the insult with LPS produced a reactive astrocyte that was neurotoxic (Zamanian et al., 2012).

Recently, two different types of reactive astrocytes have been described in response to different insults (Liddelow et al., 2017), A1 and A2. In an LPS-induced neuroinflammation scenario, activated microglia induced a neurotoxic reactive astrocyte phenotype by secreting IL-1α, TNFα and complement component 1q (C1q). The so-called A1 astrocyte showed loss of normal functions and gain of new, harmful, functions at the same time. They upregulate many of the markers mentioned previously on this review article, such as LCN-2, Serpina3, and the classical GFAP. Liddelow et al. (2017) have demonstrated that the A1 astrocyte loses normal functions, such as the ability to induce synapse formation and this could be due to the increased expression of many classical complement cascade genes that are toxic to synapses, even though the specific molecule toxic to neurons remains unknown.

On the other hand, other insults, such as ischemia, can induce a protective reactive astrocyte. In this case, reactive astrocytes may secrete neuroprotective molecules, such as TSPs that upregulate many neurotrophic factors, and promote CNS recovery and repair (Papadopoulos et al., 2004; Zador et al., 2009). This second phenotype is called A2 astrocytes, and they are known to secrete neuroprotective cytokines.

We previously demonstrated that, in a sepsis model, activated astrocytes were observed both in the hippocampus and cerebral cortex, whilst activated microglia was mainly observed in the hippocampus; moreover, both of these areas showed impairment of synapse function (Moraes et al., 2015). To further investigate the molecular effects of sepsis on these cell populations, we have used stimulation with LPS to mimic the inflammatory effect observed during sepsis. As result, LPS stimulation of glial cells generated distinct secretory profiles in astrocytes and microglia, which had direct, though contrasting, impacts on synapses. The conditioned medium of LPS-stimulated microglia induced synaptic elimination, whilst the medium of astrocytes treated with LPS increased synapse number. Both cell types showed increased production of TNFα and IL-6, and while astrocytes had increased production of TGFβ1, an anti-inflammatory and synaptogenic cytokine, microglia showed elevated secretion of IL-1β, a pro-inflammatory cytokine, especially in the hippocampus (Moraes et al., 2015). These data highlight different responses of glial cells under an insult; whereas LPS-activated microglia present a neurotoxic phenotype; LPS-activated astrocytes, presented a neuroprotective role, similar to those called A2 astrocytes.

The knowledge of how these cells act in the healthy brain just adds another layer of complexity to the immense possibilities of phenotypes that they can acquire after an insult or aging. It is theorized that there is a gradient between the A1, neurotoxic phenotype and the A2, neuroprotective phenotype, and, possibly, an interplay between different profiles (Liddelow and Barres, 2017). It is imperative to consider the diversity of astrocyte's roles and their heterogeneity in health in order to understand more deeply their roles in disease and aging, as we will discuss in the next sections.

#### AGING AND NEURODEGENERATIVE DISEASES: IMPACT ON ASTROCYTE REACTIVE PHENOTYPES

The concept of astrocyte heterogeneity has profoundly changed our view on astrocyte functions, astrocyte-neuron interactions as well as their impact in aging and neurodegenerative diseases. The first evidence on astrocyte heterogeneity referred to their inter and intra-regional diversity in morphology in the CNS (Oberheim et al., 2009). Moreover, besides morphology, astrocytes are highly diverse in terms of molecular expression pattern and functions, as discussed in the previous sections.

Emerging lines of evidences have shown that global features of astrocytic reactivity, such as changes in astrocyte morphology, transcriptional profile and function, are distinctly observed in specific brain regions during early stages of many neurodegenerative diseases, such as AD, Parkinson's disease (PD), Huntington's disease (HD), and amyotrophic lateral sclerosis (ALS; Phatnani and Maniatis, 2015). These observations raise two main questions: ''Is astrocyte heterogeneity maintained throughout the aging and in neural diseases?'' and ''How might astrocyte heterogeneity affect the onset and progression of age-related cognitive decline and age-related diseases?'' The next sections will shed light on the possible involvement of changes in astrocyte phenotype to brain aging and pathology.

#### Aging

Astrocyte reactivity is also a hallmark of physiological aging in rodents, non-human and human primates (Nichols et al., 1993; Rodríguez et al., 2014; Robillard et al., 2016). It is interesting to note that similarly to some neurodegenerative diseases, during aging, astrocyte reactivity is mainly observed in specific brain regions primarily targets for synaptic loss and age-related cognitive decline, such as the hippocampus and frontal cortex (FC; Rodríguez et al., 2016). Nevertheless, the actual contribution of astrocytes to regional vulnerability of the nervous system to specific diseases remains to be investigated.

Interestingly, it has also been reported an intra-regional heterogeneity in astrocyte's response to brain aging. Astrocytes in the hippocampal dentate gyrus and CA1 presented an age-dependent hypertrophy, as observed by a reorganization of the cytoskeleton protein, GFAP, leading to an increased surface, volume and somata volume of astrocytes in aged mice; while the opposite effect was observed in the EC. On the other hand, S100β-immunostaining, which is a cytosolic protein and hence shows a more complete astrocyte arborization than GFAP, revealed an increased surface area and volume in aged astrocytes in the EC and dentate gyrus, but not in CA1. Therefore, these observations imply that morphological changes based only on GFAP-staining may not reflect the entire complexity of astrocytes, as observed in the EC, and hence it may cause misinterpretations. In conclusion, this study suggested that astrocytes undergo complex regionspecific morphological and molecular changes upon aging (Rodríguez et al., 2014).

In agreement with these results, an age-dependent increase in hippocampal GFAP expression has also been shown by other groups in rodents (Hayakawa et al., 2007; Lynch et al., 2010; Cerbai et al., 2012) and most prominently in the human hippocampal formation (David et al., 1997), but also in the FC and temporal cortex (TC; Nichols et al., 1993). Nevertheless, functional implication of GFAP upregulation during aging is still a matter of discussion, as well as how and why particularly hippocampal astrocytes undergo these changes.

Within this context, several studies have been trying to elucidate the transcription profile of astrocytes from different brain regions during aging, taking advantage of new techniques for the isolation of purified cell populations from the mouse brain or post-mortem human brain tissues. These new approaches, combined with transcriptome analysis, have shed light on the molecular signature of aged astrocytes as well as their regionspecific changes during the aging process.

The physiological aging is characterized by a chronic, low-grade and systemic inflammation, referred as ''inflammaging'', which is an important risk factor for morbidity and mortality in elderly people (Franceschi et al., 2000). It is also known that glial cells, specially microglia and astrocytes, play a preponderant role in controlling neuroinflammation. As previously discussed, glial cells undergo morphological, molecular and functional changes that will result in pro or anti-inflammatory phenotypes, depending on the pathological context and age (Colombo and Farina, 2016). However, only recently our knowledge on the role of astrocytes in inflammaging and their regional differences in response to the aging process have beginning to be better elucidated.

Orre et al. (2014) showed that astrocytes isolated from the aged mouse cerebral cortex present increased inflammatory phenotype, although decreased GFAP expression compared to young astrocytes. It is noteworthy that although an upregulation of GFAP is usually associated with astrocyte reactivity, this feature itself may not be entirely precise to characterize this process, either in aging or brain disease. This is due to several reasons, such as: the number of GFAP positive cells and the basal GFAP expression level notably vary between different brain regions (Kimelberg, 2004); alteration in GFAP expression is usually region-specific (Rodríguez et al., 2014) and it may depends on the type and stage of the disease, or even the time after injury (Kamphuis et al., 2014; Diniz et al., 2017; Cunha et al., 2018).

Supporting astrocyte heterogeneity in aging, evidence have shown that astrocytes from distinct brain regions present an unique molecular signature in both aged mouse and human brain tissue (Soreq et al., 2017; Boisvert et al., 2018; Clarke et al., 2018). Boisvert et al. (2018) performed RNA-seq on astrocytes isolated from four brain regions of adult and aged mice: motor and visual cortex, hypothalamus and cerebellum. Surprisingly, they found significantly more changes in these cells from the hypothalamus and cerebellum, than from the cortex. They found increased expression of genes for inflammatory response and astrocyte reactivity, including GFAP and Serpin3n, and synapse elimination pathways, mainly represented by proteins of the complement system, such as complement component 3 (C3) and complement component 4b (C4b), and decreased cholesterol synthesis enzymes (Boisvert et al., 2018).

Similarly, Clarke et al. (2018) compared the expression profile of astrocytes isolated from the hippocampus, cortex and striatum of young and aged mice. They observed that hippocampal and striatal astrocytes showed more pronounced changes compared to cortical astrocytes. Corroborating previous data, they observed the most prominent class of upregulated genes upon aging was related to astrocyte reactivity, immune response and synapse elimination (Clarke et al., 2018).

Therefore, these studies have suggested that murine astrocytes undergo age-dependent changes in gene expression that may contribute to age-related synapse loss and neuroinflammation. Further, they reinforce the region-specificity of astrocyte responses to the aging process. Nevertheless, there is still a lack of evidence concerning the functional implications of these changes in gene expression to astrocyte function and cellular interplays in aging. Moreover, whether human astrocytes behave similarly to their murine counterparts is still a matter of discussion.

Within this context, Soreq and collaborators reported the first evidence of glial region-specific gene expression in the human brain aging, by analyzing ten different brain regions from the post-mortem human tissue of individuals aged from 16 to 102 years. They demonstrated that while a general upregulation of microglial and endothelial genes was observed upon aging in all brain regions analyzed; astrocyte- and oligodendrocyte-specific genes showed a more complex shift in their expression pattern, prominently observed in the hippocampus and substantia nigra (SN). On the other hand, these changes were not evident for the neuron-specific genes (Soreq et al., 2017). Altogether, these data might suggest another view on the concept of ''selective vulnerability'' in aging and age-related diseases; a concept that primarily refers to an increased vulnerability of specific brain regions or groups of cells to a pathological state or injury. The vulnerability of neuronal populations has been mainly studied and characterized in neurodegenerative diseases, as reviewed by other works (Saxena and Caroni, 2011; Fu et al., 2018). However, recently new evidence has pointed that major changes in the transcriptional profile and function of glial cells might also contribute to predict the human brain aging and to understand the selective region-specific vulnerability in these age-related neurodegenerative diseases.

#### Neurodegenerative Diseases

During the last decades, a growing number of evidence has suggested that astrocytes and microglia play a key role in synaptic dysfunction, neuronal loss and cognitive impairments associated with several neurodegenerative diseases, such as AD; ALS, HD and PD. Interestingly, changes in astrocytes arise early in the course of these conditions and most prominently in specific brain regions known to be primarily vulnerable to a particular pathology. However, the cellular and molecular basis that define the selective vulnerability of specific regions in neurodegenerative diseases remain poorly understood. Understanding astrocyte heterogeneity in this context certainly represents an important step to unveil the mechanisms underlying the onset and progression of neurodegenerative diseases.

In this section, we will discuss recent studies on how astrocytes' changes and heterogeneity might impact in the onset of two neurodegenerative diseases, AD and PD.

#### Alzheimer's Disease

AD is the most common cause of dementia and one of the main public health concerns nowadays (Scheltens et al., 2016). According to the WHO a considerable increase in incidence of dementia and age-related diseases, especially AD, is expected for the next few years. Within this context, many efforts have been done to better elucidate the mechanisms involved in AD pathology and unveil possible cellular and molecular targets for diagnosis, prevention and/or therapy.

Clinically, AD is characterized by cognitive deficits, mainly represented by learning impairments and memory loss. Histopathologically, the two major hallmarks of AD are extracellular amyloid plaques (APs) depositions, composed of amyloid β (Aβ) peptides, and intracellular neurofibrillary tangles, constituted by hyperphosphorylated Tau protein (Ballard et al., 2011). Initially, it was thought that Aβ deposition would be the primary event in AD pathology, triggering neurofibrillary tangle formation and neuronal death, which led to the ''amyloid cascade hypothesis'' (Hardy and Higgins, 1992). However, it has been demonstrated that the most neurotoxic forms of Aβ are their soluble ligands, Aβ oligomers (AβOs), shedding light on the ''Aβ oligomer hypothesis'' (Lambert et al., 1998).

AβOs accumulate at early stages in AD pathology, before plaque formation, and clinical symptoms, in humans and animal models (Gong et al., 2003; Noguchi et al., 2009; Pham et al., 2010; Lesné et al., 2013). Currently, it is well known that AβOs induce memory impairment, disruption of long-term potentiation (LTP) and long-term depression (LTD; Townsend et al., 2006; Shankar et al., 2008), synapse dysfunction and loss (Brito-Moreira et al., 2017; Diniz et al., 2017), oxidative stress (Sponne et al., 2003; De Felice et al., 2007), endoplasmic reticulum (ER) stress (Umeda et al., 2011; Lourenco et al., 2013) and neuroinflammation (Forny-Germano et al., 2014).

AD pathology initiates in a region-specific manner, meaning that specific brain regions or group of cells are more vulnerable to AD than others. The selective vulnerability of neuronal populations has been better elucidated and reported in the EC, subiculum and the hippocampal CA1 region (Reilly et al., 2003; Stranahan and Mattson, 2010), but also in the basal forebrain (Whitehouse et al., 1982) and locus coeruleus (Bondareff et al., 1987). For a more complete view on neuronal selective vulnerability we recommend additional review (Fu et al., 2018). Besides neuronal contribution, several evidences have shown that glial changes are also present early, before neuronal death and clinical symptoms, during the pathogenesis of AD.

The likely involvement of astrocytes in AD pathology has been already reported by Alzheimer (1910), through the observation of association between glial cells and senile plaques in demented brains (Alzheimer, 1910; Verkhratsky et al., 2010). Subsequent studies have shown that astrocytes undergo complex morphological, transcriptional and functional alterations in AD. Morphologically, astrocytes may become either atrophic or hypertrophic depending on the stage of the disease and proximity to APs depositions (Rodríguez-Arellano et al., 2016; **Figure 2**).

Astrocyte atrophy has been observed during early stages of AD pathology as well as in non-plaque associated astrocytes along disease progression in the EC, medial prefrontal cortex, dentate gyrus and CA1 hippocampal region in 3xTg-AD mouse model (Olabarria et al., 2010; Kulijewicz-Nawrot et al., 2012; Yeh, 2013), PDAPP-J20 transgenic mice (Beauquis et al., 2013), and in CA1 of adult mice intracerebroventricularly injected with AβOs (Diniz et al., 2017). The functional meaning of this observation is still a matter of discussion. A possibility is that a reduction in astrocyte territorial domains might result in diminished synaptic coverage and impaired neuro-vascular unit; two important features that might be involved in synaptic dysfunction and loss of metabolic support at early stages of AD. Nevertheless, the underlying mechanisms of astrocyte atrophy as well as the functional impact of that to the onset of AD pathology have not been completely elucidated so far.

Within this context, recent data from our group have reinforced that astrocytes are targets for AβOs and that these neurotoxins induce molecular and functional changes in these cells, leading to astrocyte activation and impairments in their synaptogenic and synaptical-protective capacity (Diniz et al., 2017). We showed that AβOs rapidly interacted with astrocytes and triggered astrocyte activation in vitro. Further, the intracerebroventricular injection of AβOs in adult mice acutely induced astrocyte atrophy in the CA1 hippocampal region, as observed by reduced number of primary processes, area and intensity of GFAP per cell. Following these changes, we verified that AβOs reduced the levels of TGF-β1 in astrocytes, an event that was related to an impairment in the synaptogenic and neuroprotective potential of astrocytes (Diniz et al., 2017). These data together with the fact that TGF-β1 is an important cytokine that promotes inhibitory and excitatory synapse formation in the CNS (Diniz et al., 2012, 2014b; Araujo et al., 2016), shed light on a new possible mechanism involved in synapse dysfunction and loss at early stages of AD pathology, mediated by astrocytes.

In contrast, many evidences have shown that during later stages of AD, astrocytes nearby plaques become highly hypertrophic and produce increased levels of GFAP (Olabarria et al., 2010; Heneka et al., 2015). Such features have also been observed in post-mortem human brain tissue, which seems to accompany the Braak stage progression (Simpson et al., 2010; Kamphuis et al., 2014). In addition, a transcriptomic analysis of astrocytes isolated from aged APPswe/PS1dE9 transgenic AD mice revealed that these cells acquired a pro-inflammatory phenotype, as well as a less supportive capacity to neuronal communication, represented by reduced expression of genes involved in that process (Orre et al., 2014). Altogether, these evidences suggest that the reactive phenotype of astrocytes in AD might contribute to neuroinflammation, synaptic deficits and neuronal dysfunctions during disease progression. Interestingly, a recent study has corroborated that hypothesis through the genetic inhibition of astrocyte reactivity in an AD mouse model. Ceyzériat et al. (2018) have shown that the inhibition of the astrocytic janus kinase 2 (JAK2)/signal transducer and activator of transcription 3 (STAT3) pathway efficiently controlled the reactivity of these cells in the hippocampus of AD mice, which was followed by a reduction in amyloid deposition, synaptic deficits and spatial learning improvements.

Therefore, altogether, these data strongly point to the key involvement of reactive astrocytes in the progression of AD pathology, revealing their detrimental actions on neuronal communication and cognitive performance. Nevertheless, apart from being the two major hallmarks of astrocyte reactivity, it remains elusive the functional meaning of astrocyte hypertrophy and upregulation of GFAP levels at later stages of AD.

Furthermore, since these alterations are mostly reported in vulnerable brain regions to AD pathology, in which neuronal degeneration is mainly observed, it is possible that astrocyte heterogeneity plays a role in this selective vulnerability. It has been suggested that the pre-clinical stage of AD begins decades before clinical symptoms, which is called the cellular phase of AD pathology (De Strooper and Karran, 2016). During this phase, extensive changes occur in glial cells and vasculature, which may orchestrate subsequent neuronal deficits. In fact, Soreq et al. (2017) have found that age-related changes in gene expression profile of astrocytes take place particularly in the hippocampus and SN in human tissue, two regions primarily affected during the pathogenesis of AD and PD. Therefore, these observations strengthen the likely involvement of astrocyte heterogeneity to the regional selective vulnerability in AD onset and progression. However, a direct link between astrocytes morphological/molecular changes to functional alterations still remains to be determined.

#### Parkinson's Disease

PD is the second most common type of neurodegenerative disease, affecting 2%–3% of the elderly population. Similarly to AD, the majority of PD cases is sporadic, while less than 10% are associated to genetic causes (Poewe et al., 2017). Clinically, PD is primarily characterized by motor abnormalities, mainly represented by tremor at rest, rigidity, bradykinesia and postural

to amyloid plaques (APs; A). Astrocyte atrophy occurs at early stages of AD, before plaque formation, as well as in non-plaque associated astrocytes at later stages of disease (B). Atrophic astrocytes present reduced number and thinner cellular processes, which is accompanied by diminished volume and somata volume, compared to healthy or resting astrocytes (C). Astrocyte hypertrophy is mainly observed in astrocytes surrounding APs at later stages of AD. They show opposite features in comparison to atrophic astrocytes.

instability (Jankovic, 2008), symptoms that have been linked to a progressive and selective degeneration of dopaminergic neurons in SN pars compacta (SNpc) during early stages of the disease (Dauer and Przedborski, 2003). However, several other CNS and peripheral nervous system (PNS) structures are also affected early or during the disease progression, a fact that is responsible for the non-motor symptoms of PD, which include, for example, autonomic dysfunction, olfactory deficits and cognitive decline (Schapira et al., 2017).

Histopathologically, aggregates of the synaptic protein α-synuclein are usually found in the cell bodies and neurites, called respectively Lewy bodies and Lewy neurites, in specific classes of neurons in several brain regions in all PD patients (Poewe et al., 2017). Although it is still debated, Braak et al. (2003) have proposed that Lewy bodies' pathology spread from one region to another following a specific and predictive progression during the course of PD. Interestingly, more recently it has been suggested a prion-like propagation of α-synuclein aggregates. In this hypothesis, α-synuclein might be transmitted from neuron to neuron, along axons, secreted into the extracellular space and internalized by neighboring cells, including neurons and glial cells. Once internalized, it may induce the aggregation of the native proteins in the host, contributing and exacerbating the aggregate spreading (Brundin et al., 2010). Although this hypothesis has been proven in both in vitro and in vivo models (Desplats et al., 2009; Hansen et al., 2011), it remains to be determined whether this mechanism explains the spreading of Lewy's body pathology in humans.

Several lines of evidence have shown that impairments in α-synuclein proteostasis strongly contribute to PD pathogenesis, as well as may impact other cellular pathways and functions also implicated in PD, such as mitochondria dysfunction, oxidative stress and neuroinflammation (Poewe et al., 2017). Moreover, astrocytes and microglia seem to play an important role in all of these pathological mechanisms, contributing to PD onset and progression (Halliday and Stevens, 2011).

Astrocytes accumulate α-synuclein aggregates in the SN and extra nigral regions in patients with PD, which is correlated with the severity of neuronal loss (Wakabayashi et al., 2000). Interestingly, it has also been reported that particularly protoplasmic astrocytes presented inclusions of protein aggregates in PD, which was not observed in fibrous astrocytes (Song et al., 2009). Braak et al. (2007) have shown that α-synuclein immunoreactive astrocytes are present in several brain regions of PD patients, which appears to accompany the formation of Lewy bodies and neurites. Moreover, these immunoreactive astrocytes were also found in the striatum and dorsal thalamus, regions where Lewy's bodies do not develop, suggesting that astrocytes might internalize α-synuclein released by dysfunctional axon terminals in these regions. This is supported by observations that astrocytes can take up α-synuclein released by neurons, which, in turns, induce gene expression changes in astrocytes, mainly represented by an up-regulation of pro-inflammatory genes (Lee et al., 2010). Conversely, an in vitro study has proposed that human astrocytes are similarly able to transfer aggregated α-synuclein to neighboring astrocytes, via direct contact and tunneling nanotubes (TNTs; Rostami et al., 2017). Moreover, accumulation of α-synuclein in astrocytes may impair their lysosomalautophagosomal machinery as well as lead to ER swelling and mitochondria damage (Rostami et al., 2017). Nevertheless, it remains to be investigated whether these events also occur in vivo and their overall impact to the aggregate spreading in PD pathology.

The direct functional impact of non-cell autonomous mechanisms on neurodegeneration in PD has been reported in mouse models in which solely astrocytes expressed PD-related A53T α-synuclein mutation. In this case, α-synuclein aggregation in astrocytes resulted in severe astrogliosis, disruption in the BBB and down-regulation of glutamate transporters. Moreover, microglial activation was also pronounced especially in the brainstem and SN, in which neuronal loss occurred (Gu et al., 2010). Additional studies have unraveled the molecular mechanisms underlying astrocyte and microglia activation in PD. The treatment of astroglial and microglial cell cultures with recombinant α-synuclein induced cell activation, as observed by a strong inflammatory response and production of reactive oxygen species (ROS), effects dependent on Toll-like receptor 4 (TLR4) activation in both cell types. Further, the ablation of TLR4 in α-synuclein-treated microglia and astrocytes suppressed their pro-inflammatory response, although the uptake of α-synuclein by astrocytes was independent of TLR4 (Fellner et al., 2013; Rannikko et al., 2015). These studies have shed light on new signaling pathways involved in α-synuclein-induced neuroinflammation in PD, as well as the key role of glial cells in disease progression.

Although the underlying mechanisms involved in PD pathology have been extensively investigated, there is still a lack of evidence concerning the cellular and molecular determinants of regional selective vulnerability in PD. Curiously, as discussed above, astrocytes changes are largely present in regions primarily affected during the onset of PD in humans (Wakabayashi et al., 2000; Braak et al., 2007); further, the non-cell autonomous mechanisms of dopaminergic neuronal loss have also been proved in PD animal models (Gu et al., 2010; Booth et al., 2017). However, although these evidences point to the likely involvement of astrocytes in PD, they still do not explain why degeneration begins at specific brain regions. Within this context, new studies on the inter-regional molecular heterogeneity of astrocytes have shed light on the role of these cells in predicting the human brain aging, with implications to age-related diseases, such as PD.

Notably, it has been shown that the main age-related shifts in gene expression profile of astrocytes occur in the hippocampus and SN from post-mortem human tissue (Soreq et al., 2017). Since the aging process is a risk factor for PD, it is possible that the glial alterations in SN along aging may play a role in determining the selective vulnerability of this region to PD.

Recently, an outstanding study has strengthened the role of astrocyte-microglia interplay in PD pathology. It is well known that the state of microglia activation may have a direct impact on the astrocyte phenotype in different pathologies (Liddelow and Barres, 2017; Shinozaki et al., 2017). Recently, this hypothesis was strongly demonstrated for PD. By using the agonist of the glucagon-like peptide-1 receptor (GLP1R), NLY01, the authors control the pro-inflammatory profile of microglia treated with α-synuclein preformed fibrils (α-syn PFF) in vitro. NLY01 pre-treatment indirectly controls astrocyte reactivity by reducing microglial secretion of molecules involved in astrogliosis, such as IL-1β, TNF-α and C1q. Similarly, astrocyte reactivity in the ventral midbrain was inhibited in PD mice pre-treated with NLY01. These effects were accompanied by a prevention of dopaminergic neuronal loss, reduced dopamine levels and motor deficits in PD mice (Yun et al., 2018).

Several evidences have highlighted the key involvement of glial cells in neurodegenerative diseases. However, many unanswered questions remain concerning the contribution of glial cells to the selective vulnerability of the nervous system to these diseases. Altogether, elucidating the molecular and functional heterogeneity of astrocytes arise as a fundamental step to understand the underlying mechanisms of age-related cognitive decline and associated neurodegenerative diseases. Further, it might reveal new therapeutic targets to modulate astrocyte phenotype in order to prevent and/or delay these processes.

#### FUTURE DIRECTIONS: GLIAL CELLS AS NOVEL TARGETS FOR THERAPEUTIC STRATEGIES

Astrocyte reactivity is a common feature in aging and age-related diseases. As discussed in this review article, although its functional implications are not completely understood yet, new evidence have suggested that reactive astrocytes comprise a heterogeneous group constituted by subpopulations of astrocytes that may differ from each other in terms of molecular signature, function and response to several physiological and pathological stimuli. These cells can undergo atrophy and hypertrophy amongst other changes according to the insult (**Figure 3**), therefore altering their functions in the region.

Differently from previously thought, astrocyte reactivity is not just a hallmark of aging and brain diseases, but a key mechanism involved in the pathogenesis and progression of these conditions. It is worth noting that, as there is a crosstalk between microglia and astrocytes, these cells can greatly influence each other's phenotype (for review see Jha et al., 2018), therefore, by modulating microglia phenotypes, it would be possible to change the way astrocytes respond to injury and disease as well. As described before, microglia can respond to inflammatory insults by becoming activated and influence astrocyte's response by secreting cytokines and complement cascade molecules such as TNF-α and C1q (Liddelow et al., 2017), and during aging, extensive neuroinflammation is also observed (Franceschi et al., 2000). Therefore, as it can determine the fate of astrocytes, it is important to consider microglia when designing new drugs that could modulate astrocyte's phenotypes, especially as the modulation of microglia towards an anti-inflammatory phenotype may provide neuroprotective effects, contributing to the recovery of the diseased brain (Cherry et al., 2014; Song and Suk, 2017). Suk (2017), suggested that more importantly than blocking glial cells reactivity, the focus should be on new strategies that could modulate glial cells activated phenotype.

Thus, modulation of astrocyte reactivity state emerged as an important venue for the development of new preventive and therapeutic strategies (Yun et al., 2018). Within this context, molecules and compounds, including natural or synthetics, able to control astrocyte reactivity and function have been viewed as potential drugs to neurodegenerative diseases.

Within the natural compounds, the biggest class of polyphenols in nature and also in the human diet comprises the flavonoids. These molecules are ubiquitously found in vegetables, cereals, fruits, tea, chocolate and wine (Tsao, 2010). Epidemiological studies have shown that the long-term consumption of polyphenols is associated to a reduced risk in developing cardiovascular diseases, diabetes, cancer, and neurodegenerative diseases (Arts and Hollman, 2005; Graf et al., 2005).

The consumption of polyphenol-rich diets has been linked to an enhanced cognitive performance and delay in age-related cognitive decline in humans (Devore et al., 2012; Brickman et al., 2014). In agreement, the administration of flavonoids or polyphenol-rich extracts may promote learning and memory improvements in aged animals (Zeng et al., 2012; Bensalem et al., 2018), as well as in several mouse models for neurodegenerative diseases, including AD (Zhang et al., 2014), ALS (Koh et al., 2006) and PD (Rojas et al., 2012), for review see (Solanki et al., 2015). However, the cellular and molecular mechanisms underlying the flavonoids actions are still not fully understood. Besides their remarkable antioxidant effect, flavonoids can also directly interact with cellular receptors as well as modulate signaling pathways in neurons and glial cells (Williams and Spencer, 2012).

Due to the prominent involvement of astrocyte reactivity/inflammation in aging and neurodegenerative diseases, a growing number of studies has pointed that these cells may be important targets for the natural compounds in the CNS (Matias et al., 2016). Strikingly, flavonoids have been also reported to modulate astrocyte reactivity and hence attenuating neuroinflammation. It has been shown that the flavonoids luteolin and quercetin were able to inhibit IL-1β-induced astrocyte activation, which was characterized by a reduced secretion of pro-inflammatory cytokines and chemokines, such as IL-6, IL-8, IP-10, MCP-1, an up-regulation of antioxidant enzymes, SOD1 and thioredoxin (TRX1), as well as an down-regulation of the GFAP levels in astrocytes (Sharma et al., 2007). In agreement, Bahia et al. (2008) demonstrated that the flavonoid (−) epicatechin specifically induced the activation of the antioxidant response element (ARE) pathway in astrocytes, which was followed by an increase in the glutathione (GSH) levels, an important antioxidant enzyme.

Furthermore, the pre-treatment of glial cell cultures (astrocytes and microglia) with the flavonoids naringenin, hesperidin, (+)-catechin or (−)-epicatechin significantly attenuated the production of TNF-α by these cells when exposed to LPS/interferon-γ (IFN-γ). Most prominently, naringenin not only reduced glial cell activation, but also promoted neuronal survival in a co-culture system with neurons and glia (Vafeiadou et al., 2009). Altogether, these studies have pointed that the modulation of astrocyte reactivity may have an indirect beneficial effect on neuronal physiology.

Further effects of flavonoids in controlling astrocyte reactivity have also been suggested by other groups in animal models of aging and neurodegenerative diseases. Aged rats treated with the polyphenol resveratrol showed increased neurogenesis and microvasculature and reduced astrocyte hypertrophy, as well as more ramified microglia (non-activated) within different hippocampal regions. These effects were accompanied by improvement in memory performance (Kodali et al., 2015).

In APPswe/PS1dE9 double transgenic AD mouse model, the long-term oral administration of fisetin significantly reduced astrocyte reactivity, mainly observed by a reversal of astrocyte hypertrophy and a down-regulation of the GFAP levels in AD mice (Currais et al., 2014). Similar results have also been observed in p25Tg AD mouse model treated with the phenolic compound Curcumin (Sundaram et al., 2017).

Data from our group have recently suggested beneficial effects of flavonoids also in the healthy brain. We have shown that hesperidin and casticin increased the neuroprotective potential of astrocytes in vitro, by modulating the secretion of protective factors (de Sampaio e Spohr et al., 2010; Nones et al., 2012). Although the identity of those factors is not fully characterized, further studies have suggested a number of neurotrophic factors that might be target for the flavonoids' actions. A screening of thirty-three different flavonoids revealed that calycosin, isorhamnetin, luteolin, and genistein were more promising in inducing the synthesis and secretion of the nerve growth factor (NGF), glial-derived neurotrophic factor (GDNF), and BDNF by astrocytes in culture (Xu et al., 2013).

More recently, our group raised evidences of the beneficial effects of the flavonoids in the healthy brain in vivo. The treatment of adult mice with the flavanone hesperidin induced improvements in memory performance, which were accompanied by increased synaptic density in the hippocampal CA1 region, as well as activation of the TGF-β pathway in the same region. We also showed that hesperidin was able to enhance the synaptogenic potential of astrocytes, by modulating the TGF-β1 pathway and levels (Matias et al., 2017), a pathway previously involved in the regulation of synapse formation in the CNS (Diniz et al., 2014a). Deficits in this pathway is implicated in neurodegenerative diseases, including AD (Diniz et al., 2017, 2018). Whether the effect of hesperidin/TGF-β1 in astrocytes and thus in cognitive decline is related to their anti-inflammatory activities or not, remains to be elucidated. Nevertheless, modulation of the astrocytic TGF-β1 pathway might arise as a valuable strategy in promoting cognitive improvements in both physiological and pathological contexts. Based on that, targeting of specific signaling pathways that modulates astrocyte phenotypes emerged as a promise therapeutic strategy.

Recently, modulation of the JAK/STAT3 pathway has been shown as a useful strategy to control astrocyte reactivity in vivo. By specifically inhibiting the astrocytic JAK/STAT3 activation in an AD mouse model, Ceyzériat et al. (2018) reduced astrocytes reactivity and amyloid deposition, thus leading to improvements in their cognitive performance. Similarly, as previously discussed, block of A1 astrocyte conversion by microglia by the agonist (NLY01) of the GLP1R has proven to be neuroprotective in models of PD. The treatment of PD mouse models with NLY01 significantly prevented the dopaminergic neuronal loss in the ventral midbrain, restored dopamine levels as well as ameliorated the motor deficits observed in PD mice (Yun et al., 2018). Despite all the new advances, there are still many unanswered questions in this field, such as ''Is it possible to convert a neurotoxic reactive astrocyte in a neuroprotective one?''; ''Are the human astrocytes more diverse in terms of reactive phenotypes?''; ''Are the mechanisms involved in the induction of A1 and A2 astrocytes similar in different species, such as rodents and humans?''.

In addition, due to the large genetic variability in human population and the fact that human astrocytes are unique in terms of morphological and functional complexity (Colombo et al., 1995; Oberheim et al., 2009), new studies are needed to investigate whether our knowledge on murine astrocytes can be applied to their human counterparts. Interestingly, a new study has shown that changes in astrocyte gene signature may have functional implications to the age-related cognitive decline as well as may predict the human aging with greater precision than the neuron gene signature. Curiously, these changes were more evident in the hippocampus and substantia nigra, regions where AD and PD pathology seem to begin. Therefore, they

#### REFERENCES


have highlighted the likely involvement of human astrocyte heterogeneity in predicting the regional selective vulnerability in aging and age-related diseases, such as AD and PD.

Altogether, the fact that astrocyte reactivity and dysfunction play a key role in AD and PD pathology make these cells interesting targets for the actions of synthetic and natural compounds in ameliorating cognitive performance or even restoring brain function in disease contexts, based on astrocyte biology.

#### AUTHOR CONTRIBUTIONS

All authors wrote sections of the manuscript and contributed to manuscript revision. All authors approved the submitted version of the manuscript.

#### FUNDING

This work was supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Departamento de Ciência e Tecnologia do Ministério da Saúde (Decit), and Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ).

cortical astrocytes. J. Neurochem. 106, 2194–2204. doi: 10.1111/j.1471-4159. 2008.05542.x


cortex of a triple transgenic mouse model of Alzheimer's disease. J. Anat. 221, 252–262. doi: 10.1111/j.1469-7580.2012.01536.x


nonradioactive in situ hybridization and comparative immunocytochemistry. J. Neurosci. 17, 1–10. doi: 10.1523/jneurosci.17-01-00001.1997


**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 Matias, Morgado and Gomes. 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.

# Phosphorylation of Glutamine Synthetase on Threonine 301 Contributes to Its Inactivation During Epilepsy

Deborah Huyghe1† , Andrew R. Denninger 2† , Caroline M. Voss <sup>3</sup> , Pernille Frank <sup>3</sup> , Ning Gao<sup>2</sup> , Nicholas Brandon4,5 , Helle S. Waagepetersen<sup>3</sup> , Andrew D. Ferguson<sup>6</sup> , Menelas Pangalos <sup>7</sup> , Peter Doig<sup>2</sup> \* and Stephen J. Moss 1,8 \*

<sup>1</sup>Department of Neuroscience, Tufts University School of Medicine, Boston, MA, United States, <sup>2</sup>Mechanistic Biology & Profiling, Discovery Sciences, IMED Biotech Unit, AstraZeneca, Boston, MA, United States, <sup>3</sup>Department of Drug Design and Pharmacology, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark, <sup>4</sup>Neuroscience, IMED Biotech Unit, AstraZeneca, Boston, MA, United States, <sup>5</sup>AstraZeneca Tufts Laboratory for Basic and Translational Neuroscience, Boston, MA, United States, <sup>6</sup>Structure & Biophysics, Discovery Sciences, IMED Biotech Unit, AstraZeneca, Boston, MA, United States, <sup>7</sup> IMED Biotech Unit, AstraZeneca, Cambridge, United Kingdom, <sup>8</sup>Department of Neuroscience, Physiology and Pharmacology, University College, London, United Kingdom

#### Edited by:

Maria Jose Bellini, National Council for Scientific and Technical Research (CONICET), Argentina

#### Reviewed by:

Yijuang Chern, National Research Program for Biopharmaceuticals, Taiwan Arthur Joseph Cooper, New York Medical College, United States

#### \*Correspondence:

Peter Doig peter.doig@astrazeneca.com Stephen J. Moss stephen.moss@tufts.edu

†Joint-first authors

Received: 06 December 2018 Accepted: 25 April 2019 Published: 21 May 2019

#### Citation:

Huyghe D, Denninger AR, Voss CM, Frank P, Gao N, Brandon N, Waagepetersen HS, Ferguson AD, Pangalos M, Doig P and Moss SJ (2019) Phosphorylation of Glutamine Synthetase on Threonine 301 Contributes to Its Inactivation During Epilepsy. Front. Mol. Neurosci. 12:120. doi: 10.3389/fnmol.2019.00120 The astrocyte-specific enzyme glutamine synthetase (GS), which catalyzes the amidation of glutamate to glutamine, plays an essential role in supporting neurotransmission and in limiting NH<sup>4</sup> <sup>+</sup> toxicity. Accordingly, deficits in GS activity contribute to epilepsy and neurodegeneration. Despite its central role in brain physiology, the mechanisms that regulate GS activity are poorly defined. Here, we demonstrate that GS is directly phosphorylated on threonine residue 301 (T301) within the enzyme's active site by cAMP-dependent protein kinase (PKA). Phosphorylation of T301 leads to a dramatic decrease in glutamine synthesis. Enhanced T301 phosphorylation was evident in a mouse model of epilepsy, which may contribute to the decreased GS activity seen during this trauma. Thus, our results highlight a novel molecular mechanism that determines GS activity under both normal and pathological conditions.

Keywords: glutamine synthetase, phosphorylation, epilepsy, astrocyte, cAMP-dependent protein kinase

#### INTRODUCTION

The glutamate/GABA-glutamine cycle plays a fundamental role in the central nervous system (CNS) by detoxifying the brain of excess neurotransmitters and ammonia as well as supplying neurons with glutamine, which is a precursor of both GABA and glutamate. Glutamine synthetase (GS) catalyzes an essential step in this pathway, the ATP-dependent condensation of glutamate and ammonia into glutamine (Schousboe et al., 2013).

GS is expressed in several organs; however, its expression in the CNS is mainly restricted to astrocytes (Suárez et al., 2002). Previous studies have shown that GS is essential to survival, both in rodents and in humans (Häberle et al., 2005; He et al., 2010). Moreover, inhibition of GS activity in vivo induces seizures in animal models (Rowe and Meister, 1970; Eid et al., 2008; Boissonnet et al., 2012). Consistent with its central role in neurotransmitter recycling and neurotransmission, GS expression and/or activity are dramatically decreased in several neurological disorders, including mesial temporal lobe epilepsy

(MTLE; Eid et al., 2004; van der Hel et al., 2005). Moreover, a recent study has demonstrated that conditional ablation of GS expression selectively in the mouse cortex of mice induces spontaneous seizures (Zhou et al., 2018).

MTLE is a recurrent, complex pathology characterized by neuronal loss and astrogliosis in the hippocampus and is the most common type of drug-resistant epilepsy (Asadi-Pooya et al., 2017). Several studies have shown that the cycling of glutamate to glutamine is slower and glutamate levels are increased in MTLE patients (Petroff et al., 2002; Cavus et al., 2005, 2008). Consistent with these findings, GS activity is significantly reduced in the sclerotic hippocampus (Eid et al., 2004; van der Hel et al., 2005), leading to the hypothesis that the loss of the enzyme's function might contribute to the pathological state of this disorder. However, the molecular mechanisms regulating the activity of GS within astrocytes remain to be explored.

Here, we report that cAMP-dependent protein kinase (PKA) phosphorylates GS on both threonine 301 (T301) and serine 343 (S343). Additionally, we show that the main site of phosphorylation, T301, plays a fundamental role in regulating the enzyme's activity and is significantly increased in a mouse model of epilepsy. Taken together, these results suggest that PKA-dependent phosphorylation of T301 regulates the activity of GS. Moreover, enhanced phosphorylation of this residue may contribute to the deficits in GS activity arising during epilepsy.

# MATERIALS AND METHODS

#### Animals

Eight to 12-week-old C57BL/6 male mice were used for phosphospecific antibody characterization and kainate injections. All mice were housed in a 12-h light/dark cycle. All mice were bred in-house at the Tufts University School of Medicine and handled according to protocols approved by the Institutional Animal Care and Use Committee (IACUC).

#### Antibodies and Expression Constructs

The following antibodies were used for Western-blotting: GS mouse antibodies (Millipore, cat number MAB302 and Santa Cruz, cat number sc-398034), tubulin mouse antibody, pT301 and pS343 antibodies (created by PhosphoSolutions). GS rabbit antibody (Sigma, cat number G2781) was used for immunoprecipitation. GS was cloned into prK5 from mouse genome (ATTG) and a MYC tag was inserted between amino acids 372–373. A mutagenesis kit was used to substitute T301 and S343 into alanine (T>A: 5 <sup>0</sup>AACGCCCGGCGTCTGGCTGGATTCCACGAAACC, S>A: 5 <sup>0</sup> GAAGACCGTCGGCCTGCTGCCAATTGTGACCCC).

#### Cell Culture and Transfection

COS-7 cells were maintained in Dulbecco's modified Eagle's medium/F12 (1:1) nutrient mix with 10% fetal bovine serum and 1% of PenStrep (Thermo Fisher, cat number 11330–057). The cells were transfected by electroporation (3 µg of DNA/condition) and used 24 h after transfection. Astrocytes were prepared from forebrain of 1–4 day-old mouse pups of either sex as previously described (Schildge et al., 2013). Cells were grown in Advanced modified Eagle's medium supplemented with 10% FBS and 1% P/S (Thermo Fisher, cat number 11995073). Confluent cultures were shaken to obtain an enriched astrocytes culture. The cells were trypsinized once a week and used 3 weeks after shaking.

### Western-Blotting

Cells or brain tissues were sonicated in lysis buffer consisting of Tris, pH 8 20 mM, NaCl 150 mM, Triton 1%, EDTA 5 mM, NaF 10 mM, Na3VO<sup>4</sup> 2 mM, and Na pyrophosphate 10 mM. In order to detect GS, pT301 and pS343 signal, 15, 30 and 45 µg were loaded respectively from cell in culture lysates and 15, 45 and 60 µg were loaded respectively from brain lysates. Proteins separated by SDS-PAGE (10% gel) were then transferred to nitrocellulose membranes and blocked in 6% milk in PBST for 6 h. Membranes were further incubated with the appropriate primary antibody (6% milk in PBST overnight), and after extensive washes, they were probed with HRP-conjugated secondary antibodies for 1 h. Western blots were developed using an enhanced chemiluminescence system as per the manufacturer's instructions (Amresco). Membranes were imaged (ChemiDoc MP, Bio-Rad, Hercules, CA, USA) and analyzed using ImageJ (National Institutes of Health, Bethesda, MD, USA). For quantification, the ratio between GS, pT301 and pS343 and the reporter protein (tubulin) was first analyzed. We then normalized this value to the total GS to compare the variation of T301 and S343 phosphorylation change between each condition. In order to purify the signal of pT301 antibody, it was premixed with non-phospho peptide overnight before each use.

### Immunoprecipitation and in vitro PKA Assay

Cells were lysed in lysis buffer consisting of 20 mM Tris, pH 8, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 10 mM NaF, 2 mM Na3VO4, and 10 mM Na pyrophosphate for 1 h at 4◦C on a circular rotor. Lysates were precleared with rabbit IgG attached to protein G beads for 4 h at 4◦C. Lysates were then incubated overnight with 50 µL of protein G sepharose and 3 µg of GS antibody. Precipitated immunocomplexes were twice washed in lysis buffer. Half of the sample was then incubated in 50 µl of 40 mM Tris-HCl, pH 7.5, 5 mM magnesium acetate with 200 µm ATP, and 0.1 µg of the catalytic subunit of PKA for 30 min at 30◦C, and the other half was incubated in the same reaction buffer without ATP as negative control.

#### Cloning, Expression and Purification of Mouse GS

The gene sequence encoding mouse GS (2–373, Uniprot assession #P15105) was synthesized with an N-terminal 6×His-AVI-TEV tag (MHHHHHHGLNDIFEAQKIEWHEENLYFQG) and cloned into pET-24a(+) to yield pNG189 (Blue Sky Bioservices, Worcester, MA, USA). GS mutants T301A and S343A were synthesized and cloned similarly (numbering refers to the native mouse sequence). All sequences were codon optimized for expression in Escherichia coli. GS mutants T301E and T301V were generated with the QuikChange II Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA, USA) using pNG189 as a template according to the manufacturer's protocol.

For expression of GS, pNG189 was transformed into BL21 (DE3) cells (New England Biolabs, Ipswich, MA, USA). Cultures were grown in LB medium with 25 µg/mL kanamycin at 37◦C with shaking until they reached an OD of 0.6, at which point they were induced by addition of isopropyl β-D-1 thiogalactopyranoside to a final concentration of 500 µM. After induction, cultures were incubated at 16◦C overnight with shaking. Cells were harvested by centrifugation and stored at −20◦C until further use. For protein purification, all steps were performed at 4◦C. Two liters worth of cells were thawed and resuspended in lysis buffer containing 50 mM sodium phosphate pH 7.5, 500 mM NaCl, 10% glycerol, 500 µM TCEP, 10 mM imidazole, and 1× Halt Protease Inhibitor Cocktail, EDTA-Free (Thermo Scientific, Waltham, MA, USA). Cells were lysed by passage through a French press. Lysates were cleared by high-speed centrifugation and loaded onto a HisTrap HP column (GE Healthcare, Chicago, IL, USA) pre-equilibrated with 50 mM sodium phosphate pH 7.5, 500 mM NaCl, 10% glycerol, 500 µM TCEP, and 10 mM imidazole. The column was washed with the same buffer before eluting with a gradient of 1 M imidazole. Fractions containing GS were identified by SDS-PAGE, pooled, and concentrated to ∼2 mL. Crude GS was then loaded onto a HiLoad 16/600 Superdex 200 pg column (GE Healthcare, Chicago, IL, USA) pre-equilibrated with 50 mM sodium phosphate pH 7.5, 500 mM NaCl, 10% glycerol, and 500 µM TCEP. GS was eluted with the same buffer at a flow rate of 1 mL/min. Fractions containing GS were identified by SDS-PAGE, pooled, supplemented with 1 mM MgCl<sup>2</sup> and ATP, and concentrated. Protein concentration was determined by the Bradford method (Bio-Rad; #500–006) using bovine serum albumin as a standard (Thermo Scientific, Waltham, MA, USA; #23209). GS was stored at −80◦C. GS mutants were expressed and purified similarly.

# Thermal Shift Assays

Twenty-five-microliter reactions containing 5 µM wild-type (WT) or mutant GS, 5× SyproOrange (Life Technologies, Carlsbad, CA, USA; #S6651) and varying concentrations of MgCl<sup>2</sup> and ATP in 20 mM HEPES pH 7.5, 300 mM NaCl, 10% glycerol, and 500 µM TCEP were prepared in white 96-well PCR plates (ThermoFisher, Waltham, MA, USA; #AB0700/W). Plates were incubated in a Bio-Rad CFX96 thermocylcer at 25◦C for 2 min then heated to 95◦C at a rate of 0.2◦C/5 s. Plates were read at every 0.2◦C interval. Melting temperatures (Tm) were determined by the maximum point of the dF/dT curve of the HEX channel.

#### In vitro Phosphorylation Assays

Fifty-microliter reactions containing 5 µM WT or mutant GS and 2.5 µM Protein Kinase A (PKA; Sigma; #P2645) in 50 mM imidazole pH 7.5, 10 mM MgCl2, 100 µM EDTA, 2 mM DTT, and 1 mM ATP were prepared and incubated overnight at room temperature with or without 5 µM PKI 6-22 (Calbiochem; #539684). Kinase reactions lacking PKI were stopped by the addition of PKI. Intact mass spectrometry analysis was performed on a TripleTOF 5,600+ (AB Sciex) equipped with a Duo Spray Ion Source and a Shimadzu LC 20-AD HPLC system (Shimadzu Scientific Instruments). Samples were diluted to 1 µM in 0.1% formic acid and separated on a Poroshell 300SB-C8 75 × 2.1 mm, 5 µm column (Agilent) at 30◦C with a gradient of acetonitrile (5%–95%) in 0.1% formic acid. The mass spectrometer was operated in positive ion and intact protein mode. LC-MS data were acquired in TOF MS mode for m/z+ between 600 and 2,000. Peak areas for protein species were determined following spectrum deconvolution using PeakView (version 2.2) software.

# GS Activity Assays

GS activity was measured using the γ-glutamyl hydroxamate assay (Pamiljans et al., 1962). For recombinant GS, reactions were prepared in 384-well plates (Greiner; #781101) in a final volume of 10 µL by mixing 5 µL of a concentrated substrate mix containing 0–700 mM glutamate, 40 mM ATP, and 40 mM hydroxylamine in assay buffer (40 mM imidazole pH 7.4, 20 mM MgCl2, 1 mM DTT, and 0.01% Triton X-100) with 5 µL of concentrated enzyme mix containing either 150 nM WT GS, 700 nM GS T301A, 700 nM GS T301E, or 200 nM GS T301V in assay buffer. Negative controls lacking glutamate were prepared in parallel. Reactions were allowed to proceed at room temperature for 1 h. Reactions were terminated by the addition of 50 µL of a stop/developer solution containing 163 mM FeCl3, 98 mM trichloroacetic acid, and 202 mM HCl. Stopped reactions were mixed thoroughly, and their absorbances were read at 540 nm. Background absorbance from negative controls lacking glutamate was subtracted from all measurements. These data were compared to standard curves containing genuine γ-glutamyl hydroxamate (Sigma; #G2253), which were prepared in assay buffer and developed as above; kinetic parameters were extracted from non-linear fits of these data.

Assays in tissue lysates were performed as above with slight modifications. Samples were lysed by sonication in assay buffer supplemented with 1% Triton X-100, 150 mM NaCl, 25 mM β-glycerophosphate, 2.5 mM sodium orthovanadate, and 1× Protease Inhibitor Cocktail (Sigma; #P1860) and centrifuged to remove insoluble material. Several common protease and phosphatase inhibitors, including sodium fluoride, sodium pyrophosphate, phenylmethylsulfonyl fluoride, and 4-(2 aminoethyl) benzenesulfonyl fluoride were found to inhibit GS (not shown) and should be avoided. Reactions were prepared in 200 µL PCR tubes in a final volume of 20 µL by mixing 10 µL of cleared lysate and 10 µL of a concentrated substrate mix containing 700 mM glutamate, 40 mM ATP, and 40 mM hydroxylamine in assay buffer. Negative controls lacking glutamate were prepared in parallel. Reactions were allowed to proceed at room temperature for 30, 40, 50, or 60 min to ensure linearity. Reactions were terminated by adding 100 µL of stop/developer solution, after which they were centrifuged to remove insoluble material. One-hundred microliters of each supernatant was transferred to the well of a 384-well plate, and their absorbances were read at 540 nm. Background absorbance from negative controls lacking glutamate was subtracted from all measurements. Activity in lysates was compared to standard curves containing genuine γ-glutamyl hydroxamate as above and normalized by protein concentration using the Bradford method.

#### PKA Inhibition Assay

The effect of methionine sulfoximine (MSO; Sigma; #M5379) on PKA (EMD Millipore; #539481) activity was measured using the PKA Colorimetric Activity Kit (Invitrogen; #EIAPKA) according to the manufacturer's protocol.

#### Data Analysis

Data were analyzed using GraphPad PRISM, and statistical significance was determined at p < 0.05 using one-way ANOVA followed by Dunnett's multiple comparison post hoc test or Student's t-test for two groups.

#### RESULTS

#### GS Is Phosphorylated by PKA in vitro

In mammals, GS is a homomeric decamer comprising two stacked pentameric rings that catalyzes the ATP-dependent condensation of ammonia and glutamate to glutamine (Krajewski et al., 2008). Protein sequence comparison between rodents and human shows a high percentage of similarity (98%, **Figure 1A**). Interestingly, both species exhibit two consensus sites for PKA-dependent phosphorylation near the active site of the protein localized on threonine 301 and serine 343 (T301 and S343, **Figure 1A**). However, neither of these predicted sites of phosphorylation are found in prokaryotic forms of GS (**Figure 1B**), suggesting that phosphorylation may play a species-specific role in controlling enzyme activity.

In order to assess the possible role that phosphorylation plays in determining GS activity, we initially expressed WT and mutant versions of mouse GS modified with an N-terminal 6× His-Avi-TEV tag in E. coli. WT GS and mutants in which T301 and S343 were mutated to alanine (T301A and S343A) were affinity purified and exposed to the catalytic subunit of PKA in the presence or absence of a peptide inhibitor of PKA (PKI). Intact mass spectrometry analyses revealed that WT GS and GS S343A were substrates for PKA-dependent phosphorylation (**Figure 1C**), as indicated by an increase in mass consistent with the addition of a single phosphoryl group in the absence of PKI. Mutating T301 to alanine completely abolished GS phosphorylation, thus indicating that T301 is the primary phosphorylation site of PKA-mediated phosphorylation of GS in vitro.

#### PKA Phosphorylation Decreases GS Activity

T301 is located within the ''glutamate flap,'' a flexible active site loop that enhances the binding of both glutamate and ammonia and is thought to shield the active site from water, thereby preventing premature hydrolysis of reactive intermediates (**Figure 2A**, **Supplementary Figure S1**; Liaw and Eisenberg, 1994; Alibhai and Villafranca, 1994; Gill and Eisenberg, 2001; Gill et al., 2002; Krajewski et al., 2008). Previous mutational studies of GS from Bacillus subtilis have demonstrated that mutations in the flap can dramatically affect the enzymatic activity of GS (Fisher et al., 2002; Wray and Fisher, 2010). In order to understand how phosphorylation of T301 affects GS activity, we treated GS with PKA in the presence or absence of PKI and measured the initial velocities of γ-glutamyl hydroxamate formation at various concentrations of glutamate (Pamiljans et al., 1962). We found that unphosphorylated GS (+PKI) was highly active and displayed a low affinity for glutamate (Km app = 21 ± 1 mM; n = 3; **Figure 2B**), which is consistent with previous studies on GS from other rodents (Richterich-Van Baerle et al., 1957; Wu, 1963; Lund, 1970). Importantly, treatment of GS with PKA in the absence of PKI decreased the specific activity of GS by approximately 40% at the highest glutamate concentration tested but did not significantly alter glutamate affinity. The decrease in activity correlated with a 74 ± 7% (n = 3) degree of phosphorylation of GS as measured by intact mass spectrometry (**Figure 1B**).

#### PKA Mediates Its Effects on GS Activity via T301

To determine if the effects of PKA on GS activity are mediated via direct phosphorylation we used site-directed mutagenesis. First, we mutated T301 to glutamate (T301E), whose negative charge mimics that of a phosphoryl group. We then tested the activity of GS T301E and found, as expected, that it had very low activity (about 25% of that of WT at maximum glutamate concentration; **Figure 2B**) and was insensitive to PKA-treatment.

To further explore structure-function relationships at position 301, we tested the activity of GS T301A, which we expected to exhibit WT-like activity but lack the capacity for regulation by phosphorylation. GS T301A was insensitive to PKA-treatment; however, its activity was similar to GS T301E (approximately 20% of that of WT, **Figure 2B**). In light of the low activity observed for GS T301A, we questioned whether WT-like activity would be maintained by the introduction of a bulkier valine residue, which is found at this position in several bacterial species (Krajewski et al., 2008; Wray and Fisher, 2010). Therefore, we generated a T301V mutant and tested its activity. GS T301V exhibited a similar specific activity to WT GS but displayed a significantly lower affinity for glutamate (Km app = 145 ± 12 mM, **Figure 2B**). As expected, PKA-treatment did not affect T301V activity.

In addition to alterations in glutamate-binding and activity, mutations within the glutamate flap have been shown to reduce the sensitivity of GS to the inhibitor MSO (Wray and Fisher, 2010). MSO binds to the glutamate binding site of GS in an initially reversible manner but is rapidly phosphorylated by GS in the presence of ATP; in WT GS, the glutamate flap forms strong hydrogen bonds with MSO phosphate, resulting in non-covalent but nearly irreversible inhibition (Eisenberg et al., 2000; Gill and Eisenberg, 2001; Jeitner and Cooper, 2014). We found that the IC<sup>50</sup> of MSO for WT GS was 182 ± 14 µM (**Figure 2C**). The IC<sup>50</sup> of MSO for GS T301V was 40-fold higher (7.6 ± 0.5 mM) and >100 mM for GS T301E and GS T301A.

In order to rule out the possibility that the observed differences in GS activity resulted from changes in GS stability or Mg- or ATP-binding, we performed thermal shift assays on GS in the presence of varying concentrations of MgATP, which has been shown to stabilize GS (Krajewski et al., 2008). We found that the melting temperatures of all GS constructs used in this study were similar (approximately 55◦C) and that incubation with MgATP increased their melting temperatures to similar extents, indicating that the respective mutations do not impact on enzyme stability or ability to bind MgATP (**Figure 2D**). Collectively, these results indicate that residue T301 is essential for efficient catalysis and that alterations at this site, either by phosphorylation or mutation, can be a strong negative regulator of GS activity. Additionally, these findings are consistent with the idea that T301 phosphorylation exerts its effects on GS activity by perturbing the position or dynamics of the glutamate flap.

## Analyses of GS Phosphorylation in Cell Lines and the Brain Using Mass Spectrometry

In order to assess the significance of our measurements using purified GS, we expressed GS in COS-7 cells (Huyghe et al., 2014). GS was immunoprecipitated from transfected COS-7 cells

Activity of purified WT and mutant GS pretreated with MSO for 1 h at the indicated concentrations (with 20 mM ATP in assay buffer) as measured by the γ-glutamyl hydroxamate assay. Results are normalized to the maximum activity for each protein. Graph shows the mean ± SEM (n = 3–5). (D) Melting temperatures of purified WT and mutant GS in the presence of various concentrations of MgATP as measured by differential scanning fluorimetry. Graph shows the mean ± SEM (n = 3). and incubated with active PKA in the presence or absence of ATP. The samples were then subjected to SDS-PAGE and Thus, our experiments in COS-7 cells and brain reveal that in common with our experiments in vitro, T301 is phosphorylated by PKA. In addition, these findings suggest that GS is also phosphorylated on residue S343 in cell lines and in

#### Characterization of Phospho-specific Antibodies Against GS

The results obtained using in vitro measurements of GS phosphorylation and our LC-MS/MS studies prompted us to create phospho-specific antibodies against T301 and S343 (pT301 and pS343, respectively). Accordingly, rabbits were immunized with a highly purified (>95%) synthetic peptide centered and chemically phosphorylated on the residue corresponding to T301 (bold) in GS: KGGLDNARRL**T**GFHETSNIND. Antibodies were raised

stained with Coomassie blue. The major band at 45 kDa was then digested with trypsin and phosphorylation was examined using liquid chromatography coupled with mass spectroscopy (LC-MS/MS). Possible sites of phosphorylation were then ranked according to their ambiguity score (Ascore; Beausoleil et al., 2006). Both T301 and S343 exhibited Ascores >19 compared to control (-ATP), suggesting both residues are substrates for PKA-dependent phosphorylation (**Figure 3A**). Encouraged by our experiments using ''back-phosphorylation,'' we assessed phosphorylation of GS in the brain. To do so, GS was immunoprecipitated from mice hippocampal lysates. Following SDS-PAGE and Coomassie blue staining, the principal band at 45 kDa was subjected to LC-MS/MS as detailed above. Ascores >19 were seen for both T301 and S343 (**Figures 3B,C**).

the brain.

(C) Representative LC/MS/MS spectra associated for two phosphopeptides identified from GS immunoprecipitated from brain samples.

against S343 by injecting mice with synthetic peptide centered and chemically phosphorylated on the residue corresponding to S343 (bold): KKGYFEDRRP**S**ANCDPYAVTE. High titer antisera were then subject to tandem affinity purification on the phospho- and dephospho-antigens (Jovanovic et al., 2004; Saliba et al., 2012).

To test the specificity of these antibodies we expressed WT GS, T301A and S343A in COS-7 cells (**Figures 4A,D** respectively). Our laboratory previously determined that COS-7 cells do not express GS endogenously; therefore, those cells are a great system to study the specificity of GS phospho antibodies (Huyghe et al., 2014). Extracts of cells expressing the respective constructs were then immunoblotted with pT301, pS343 and GS antibodies. The pT301 antibody recognized a major band of 45 kDa in cells expressing WT GS, as well as a lighter band at the same molecular weight with GS T301A (**Figure 4A**, n = 4). We then performed a peptide competition assay by pre-mixing the antibody with the phospho or non-phospho peptide (**Supplementary Figure S2**). We observed that the pT301 signal disappears when pre-mixed with the phospho peptide but not with the non-phospho peptide, demonstrating that the pT301 antibody specifically recognizes the T301 phosphorylation site; however, the lighter band observed with the T301A point mutant also suggests that pT301 antibody recognizes a second epitope on the phospho peptide. Therefore, we tested whether we could increase the specificity of this antibody by adding an extra purification step consisting of pre-mixing pT301 with the non-phospho peptide. Detection of this band was prevented by incubation with the de-phospho-antigen (**Supplementary Figure S2**); therefore, we added this extra step of antibody purification for the rest of the study.

The same method was used to assess the specificity of the pS343 antibody. In contrast to T301, no basal phosphorylation of S343 was seen in COS-7 cells. In order to determine whether S343 phosphorylation could be detected by increasing PKA activity, the cells were treated with forskolin for 15 min (FSK; 20 µM), which resulted in the detection of a 45-kDa band in cells expressing WT GS but not GS S343A (**Figure 4D**). Significantly, the detection of this band was blocked by absorption with the phospho- but not the dephospho-antigen and by pretreatment with λ-phosphatase (**Supplementary Figure S2**).

forebrain were used for peptide competition assays to test pS343 specificity in brain. Total lysates were immunoblotted with tubulin and pS343 antibodies (n = 3). (F) Lysates from WT mice forebrains were used for lambda-phosphatase assays and immunoblotted with tubulin and pS343 antibodies (n = 3).

In agreement with our studies conducted in COS-7 cells (**Figures 4A,D**, **Supplementary Figure S2**), pT301 and pS343 antibodies recognized bands of 45 kDa in brain extracts, the detection of which was prevented by peptide competition (**Figures 4B,E**) and λ-phosphatase treatment (**Figures 4C,F**). Collectively, these findings show that the antibodies directed against pT301 and pS343 recognize specifically the phosphorylated sites of GS. Moreover, these results suggest that the enzyme is phosphorylated on residues T301 and S343 in cell lines and the brain.

# GS Is Preferentially Phosphorylated on T301 in COS-7 Cells and Astrocytes

To further assess GS phosphorylation, we expressed GS in COS-7 cells and compared the effects of FSK treatment on the phosphorylation of both residues T301 and S343 over a time course of 60 min. Consistent with our initial studies, only T301 exhibited basal phosphorylation, and FSK significantly increased phosphorylation of T301 and S343 over time. Significantly, PKA-dependent phosphorylation of T301 occurs faster (**Figure 5A**, 15 min = 1.77 ± 0.21, p = 0.011, n = 4) than on S343 (**Figure 5A**, 30 min = 0.13 ± 0.009, p = 0.046, n = 4). Next, we examined the levels of basal phosphorylation for each site in cultured astrocytes. While robust phosphorylation of T301 was seen in these primary cells, S343 phosphorylation was not detected (**Figure 5B**, n = 3). Collectively, these results in cultured cells suggest that GS is preferentially phosphorylated on T301.

# MSO Binding Decreases T301 Phosphorylation

MSO is a well-characterized GS inhibitor that binds irreversibly to the glutamate binding site within this enzyme (Eisenberg et al., 2000; Cloix et al., 2010). Given the strategic

was used as an additional control (n = 3).

localization of T301 within the active site, we further sought to assess whether MSO affects the ability of PKA to phosphorylate GS. To do so, COS-7 cells transfected with WT GS were exposed to 1 mM MSO for 60 min followed by 20 µM FSK for 30 min. Pre-treatment with MSO partially blocked the FSK-induced potentiation of T301 phosphorylation compared to control (**Figure 6A**, FSK = 3.41 ± 1.15, MSO+FSK = 2.19 ± 0.88, n = 4). Likewise, we observed that MSO tended to partially reduce FSK-induced potentiation of S343 phosphorylation; however, a significant increase was still observed compared to control condition (**Figure 6A**, FSK = 0.46 ± 0.11, MSO+FSK = 0.35 ± 0.12, n = 4, p = 0.03).

To assess the significance of the results obtained in COS-7 cells for events in the brain, mice were injected with either saline or a pro-convulsive dose of MSO (75 mg/kg, Bernard-Helary et al., 2000; Cloix et al., 2010). All MSO-injected mice started seizing between 4 and 6 h after injection. After 8 h, the mice were sacrificed and total GS levels and phosphorylation of T301 and S343 were analyzed from hippocampal lysates. MSO treatment significantly decreased T301 phosphorylation relative to vehicle-treated controls (**Figure 6C**, MSO = 0.70 ± 0.06, p = 0.0185, n = 5). We did not observe any significant change in S343 phosphorylation in MSO-injected mice compared to controls (**Figure 6C**, MSO = 1.11 ± 0.07, n = 5). These results suggest that MSO treatment specifically reduces phosphorylation of residue T301 in the hippocampus of WT mice. However, a previous study has shown that MSO binding impairs the antibody-antigen recognition that might lead to misinterpreting

corresponding control followed by forskolin treatment (FSK, 20 µM for 30 min). Lysates were analyzed by immunoblotting with tubulin, GS, pT301 and pS343 antibodies. Graphs show the mean ± SEM (∗p < 0.05, ∗∗p < 0.01, ns, not significant, ANOVA Dunnett's multiple comparison test, n = 4). (B) Relative activity of purified mouse PKA treated with either PKI (1 mM) or MSO as measured by an ELISA-based assay using an immobilized peptide substrate. Graph shows the mean ± SEM (n = 3). (C) WT male mice (8–12 weeks) were injected intraperitoneally by saline or MSO (75 mg/kg). Dissections of the hippocampi were performed 8 h after injection and the tissues lysates were analyzed by immunoblotting with tubulin, GS, pT301 and pS343 antibodies. Graphs show the mean ± SEM (∗p < 0.05, ns, not significant, unpaired t-test, n = 5). (D) The schema represents the antigen region recognized by different GS antibodies. Hippocampal lysates used in (C) were analyzed by Dot-blot for pT301 and two different GS antibodies (GS<sup>1</sup> and GS<sup>2</sup> ). Graphs show the mean ± SEM (∗∗p < 0 0.01, unpaired t-test, n = 5).

the reduction of the signal for a decrease in protein levels (Bidmon et al., 2008). To circumvent this technical issue, the hippocampal samples from the saline or MSO-injected mice were used to perform dot-blot analyses to compare their respective immunoreactivity after an additional denaturation step (10 min at boiling temperature). As an additional control, a GS antibody targeting the region surrounding the T301 residue was also used (GS<sup>2</sup> ). We did not detect any change in GS signal using the regular GS antibody (GS<sup>1</sup> ) or GS<sup>2</sup> , however, a reduction in the T301 signal was still evident (**Figure 6D**). Therefore, MSO binding leads to a decrease of T301 phosphorylation in vivo. Finally, to investigate any possible direct inhibitory effects of MSO on PKA activity, we directly tested how MSO affected the enzyme activity of PKA in vitro. Under these conditions, MSO did not directly modify PKA activity (**Figure 6B**). Accordingly, these results provide further evidence that T301 phosphorylation

is intimately linked to the catalytic mechanism of GS in the hippocampus of WT mice.

#### T301 Phosphorylation Is Increased and GS Activity Is Decreased in Chemico-induced Status Epilepticus

Epilepsy leads to deficits in GS activity (Eid et al., 2004; van der Hel et al., 2005); however, the underlying molecular mechanisms are ill-defined. Thus, we assessed if alterations in GS phosphorylation may be of significance in a mouse model of epilepsy. To do so, we injected mice with the chemico-convulsant kainic acid (KA; 20 mg/kg, Silayeva et al., 2015). Subsequent to KA injection, the development of Status Epilepticus (SE) was measured by the development of stage V seizures as defined on the Racine Scale (Racine et al., 1972). Sixty minutes after entrance into SE, mice were sacrificed and hippocampal lysates were subjected to immunoblotting with GS, pT301 and pS343 antibodies. This revealed that T301 phosphorylation was significantly increased in SE relative to control (**Figure 7A**, SE = 1.71 ± 0.12, n = 5, p = 0.0017). In contrast, pS343 signal (SE = 1.41 ± 0.14, n = 5, p = 0.27) and total GS (SE = 1.11 ± 0.04, n = 5, p = 0.85) were not affected in KA-treated mice. Finally, we assessed if in addition to modifying GS phosphorylation, SE impacts on its activity using the γ-glutamyl hydroxamate assay. The results revealed that GS activity was reduced by about 25% in KA-injected mice relative to saline-injected controls (**Figure 7B**, control = 5.5 ± 0.17, SE = 3.4 ± 0.49, n = 6, p = 0.02). Thus, increased phosphorylation of T301 may contribute to the reduced activity of GS activity observed in epilepsy.

#### DISCUSSION

In the brain, GS expression is restricted to astrocytes where it plays a critical role in regulating glutamate levels and ammonia by catalyzing their conversation to glutamine. Thus, GS is accepted to play a central role in regulating amino acid-mediated neurotransmission. Studies in peripheral tissues and the CNS suggest that GS activity is modulated by posttranslational mechanisms that include acetylation, tyrosine nitration, oxidation, and ubiquitination (Bidmon et al., 2008; Castegna et al., 2011; Nguyen et al., 2016).

GS shows remarkable phylogenic conservation, but it is notable that all mammalian isoforms contain consensus sites for phosphorylation by some protein kinases, including PKA, that are not found in prokaryotes. To assess the importance of phosphorylation—the most common protein covalent modification—in regulating GS activity, we expressed and purified the murine enzyme from E. coli. In vitro, GS is phosphorylated by PKA to a final stoichiometry of 70% solely on T301. T301 is within the glutamate flap region of GS, which plays a critical role in substrate binding and is essential for efficient catalysis (Alibhai and Villafranca, 1994; Liaw and Eisenberg, 1994; Gill and Eisenberg, 2001; Gill et al., 2002). Phosphorylation of GS significantly reduced its ability to synthesize glutamine, an effect that was critically dependent on T301, as demonstrated by mutation.

Mutation of T301 to either a glutamate residue, which mimics the negative charge of a phosphoryl group or a conservative alanine residue dramatically decreases enzyme activity and increases the apparent K<sup>m</sup> for glutamate. In prokaryotes, a hydrophobic valine residue is found at this position, and conversion of T301 to valine in murine GS results in a relatively mild reduction in glutamate affinity. In addition to these effects on the catalytic properties of GS, mutating T301 greatly reduced the affinity of GS for its inhibitor MSO. Importantly, mutations of T301 did not modify GS stability or its ability to bind ATP and magnesium, providing further evidence that these mutations selectively impact glutamate binding. Collectively these results suggest a vital role for the phosphorylation of T301 in modulating the activity of GS, consistent with its role in substrate binding (Krajewski et al., 2008; Wray and Fisher, 2010).

To better understand the significance of our in vitro measurements, we examined the phosphorylation of GS when expressed in COS-7 cells and the brain. Confirming our experiments with purified PKA, LC-MS/MS revealed that T301 was phosphorylated in both systems. In addition, a second site of phosphorylation, S343, was detected. In contrast to T301, S343 is buried near the monomer-monomer interface (**Supplementary Figure S1**), and its role remains unknown. In agreement with our LC-MS/MS measurements, several high-throughput mass spectrometry studies have detected both T301 and S343 phosphorylation in the murine brain (Goswami et al., 2012; Trinidad et al., 2012). To further study GS phosphorylation, we produced phospho-specific antibodies against T301 and S343. These tools confirmed that both residues were phosphorylated in COS-7 cells and brain lysates. Significantly, in COS-7 cells, phosphorylation of T301 was faster than S343 upon activation of PKA. In cultured astrocytes, phosphorylation of S343 was not detected under basal conditions, but robust phosphorylation of T301 was seen. Collectively, these results suggest that T301 is preferentially phosphorylated upon the activation of PKA. They further suggest that S343 phosphorylation may be regulated in a cell type- and/or context-specific manner.

To further assess the relationship between phosphorylation of T301 and GS structure, mice were injected with MSO, which irreversibly binds to the enzyme's active site. Eight hours after MSO injection, decreased T301 phosphorylation was evident in the brain, while S343 and total GS levels were unaffected. Thus, these results suggest that phosphorylation of T301 is intimately linked with the catalytic mechanism of GS in the brain. Finally, we assessed the effects of SE on GS phosphorylation and activity. Deficits in GS activity and increased T301 phosphorylation were evident in mice exhibiting SE, while total GS levels and S343 phosphorylation were unaltered. Therefore, enhanced T301 phosphorylation may contribute to the deficits in GS activity that have been reported in human patients and animal models of epilepsy (Eid et al., 2004; van der Hel et al., 2005; Eid et al., 2012).

Collectively, our studies have revealed that GS is subject to PKA-mediated phosphorylation, which leads to its inhibition after SE. Thus, preventing its phospho-dependent inactivation may be a potent mechanism to upregulate GS activity, which may be of therapeutic value in epilepsy. Consistent with this notion, PKA activity has been shown to be elevated in animal models of SE and in human epileptic foci (Rakhade et al., 2005; Lee et al., 2007; Bracey et al., 2009). It is important to note that T301 in addition to S343 may also be subject to phosphorylation by other protein kinases such as protein kinase C. Thus, the phosphorylation of GS may be subject to modulation by multiple cell signaling pathways under control conditions and during seizures. Finally, as decreased GS activity has been implicated in schizophrenia (Steffek et al., 2008) and Alzheimer's disease (Smith et al., 1991), alterations in its phosphorylation may also be relevant to these pathologies.

# ETHICS STATEMENT

All mice were bred in-house at the Tufts University School of Medicine and handled according to protocols approved by the Institutional Animal Care and Use Committee (IACUC).

# AUTHOR CONTRIBUTIONS

DH performed cell culture, dissections, mutagenesis, antibody characterization, in vivo injection and Western blotting, and data analysis. AD performed enzyme purifications, in vitro phosphorylation, enzyme assays, thermal shift assays and data analysis. CV and PF performed antibody characterization, cell culture, Western-blotting. NG performed intact mass spectrometry. Experiments were designed by DH, AD, PD and SM. The manuscript was written by DH, AD and SM with input from PD, CV, NB, PF, HW, NG, AF, MP. DH and SM conceived the project.

# FUNDING

This work was supported by NIH-NINDS grants NS101888, NS081735, NS087662, NIMH grants MH097446, MH106954 and NIH-DA grant DA037170.

# ACKNOWLEDGMENTS

AD is a fellow of the AstraZeneca postdoc programme. We thank Emily Schwenger for technical assistance.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnmol.2019. 00120/full#supplementary-material

#### REFERENCES


**Conflict of Interest Statement**: AD, NG, NB, PD and MP are current employees of AstraZeneca. SM acts as a consultant for AstraZeneca and SAGE Therapeutics, relationships that are governed by Tufts University.

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.

Copyright © 2019 Huyghe, Denninger, Voss, Frank, Gao, Brandon, Waagepetersen, Ferguson, Pangalos, Doig and Moss. 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.

# Humanin, a Mitochondrial-Derived Peptide Released by Astrocytes, Prevents Synapse Loss in Hippocampal Neurons

Sandra Cristina Zárate1,2\*, Marianela Evelyn Traetta3,4 , Martín Gabriel Codagnone3,4‡ , Adriana Seilicovich1,2† and Analía Gabriela Reinés 3,4†

1 Instituto de Investigaciones Biomédicas (INBIOMED, UBA-CONICET), Facultad de Medicina, Universidad de Buenos Aires, Buenos Aires, Argentina, <sup>2</sup>Departamento de Histología, Embriología, Biología Celular y Genética, Facultad de Medicina, Universidad de Buenos Aires, Buenos Aires, Argentina, <sup>3</sup> Instituto de Biología Celular y Neurociencias "Prof. E. De Robertis" (IBCN, UBA-CONICET), Facultad de Medicina, Universidad de Buenos Aires, Buenos Aires, Argentina, <sup>4</sup>Departamento de Farmacología, Facultad de Farmacia y Bioquímica, Universidad de Buenos Aires, Buenos Aires, Argentina

#### Edited by:

Maria Jose Bellini, National Council for Scientific and Technical Research (CONICET), Argentina

#### Reviewed by:

Alexei Verkhratsky, University of Manchester, United Kingdom Estefanía Acaz-Fonseca, Spanish National Research Council (CSIC), Spain

#### \*Correspondence:

Sandra Cristina Zárate szarate@fmed.uba.ar

†These authors have contributed equally to this work

#### ‡Present address:

Martín Gabriel Codagnone, APC Microbiome Ireland, University College Cork, Cork, Ireland; Department of Anatomy and Neuroscience, University College Cork, Cork, Ireland

> Received: 16 January 2019 Accepted: 09 May 2019 Published: 31 May 2019

#### Citation:

Zárate SC, Traetta ME, Codagnone MG, Seilicovich A and Reinés AG (2019) Humanin, a Mitochondrial-Derived Peptide Released by Astrocytes, Prevents Synapse Loss in Hippocampal Neurons. Front. Aging Neurosci. 11:123. doi: 10.3389/fnagi.2019.00123 Astroglial cells are crucial for central nervous system (CNS) homeostasis. They undergo complex morpho-functional changes during aging and in response to hormonal milieu. Ovarian hormones positively affect different astroglia parameters, including regulation of cell morphology and release of neurotrophic and neuroprotective factors. Thus, ovarian hormone loss during menopause has profound impact in astroglial pathophysilogy and has been widely associated to the process of brain aging. Humanin (HN) is a secreted mitochondrial-encoded peptide with neuroprotective effects. It is localized in several tissues with high metabolic rate and its expression decreases with age. In the brain, humanin has been found in glial cells in physiological conditions. We previously reported that surgical menopause induces hippocampal mitochondrial dysfunction that mimics an aging phenotype. However, the effect of ovarian hormone deprivation on humanin expression in this area has not been studied. Also, whether astrocytes express and release humanin and the regulation of such processes by ovarian hormones remain elusive. Although humanin has also proven to be beneficial in ameliorating cognitive impairment induced by different insults, its putative actions on structural synaptic plasticity have not been fully addressed. In a model of surgical menopause in rats, we studied hippocampal humanin expression and localization by real-time quantitative polymerase chain reaction (RT-qPCR) and double immunohistochemistry, respectively. Humanin production and release and ovarian hormone regulation of such processes were studied in cultured astrocytes by flow cytometry and ELISA, respectively. Humanin effects on glutamate-induced structural synaptic alterations were determined in primary cultures of hippocampal neurons by immunocytochemistry.

**Abbreviations:** CNS, Central nervous system; GFAP, glial fibrillary acidic protein; HN, humanin; HNr, rattin; AD, Alzheimer's disease; ORF, open reading frame; rRNA, ribosomal RNA; NMDA, N-methyl-D-aspartate; CNTFR-α, ciliary neurotrophic factor α; DMEM, Dulbecco's Modified Eagle Medium; FCS, fetal calf serum; OVX, ovariectomized; HPRT, hypoxanthine-guanine phosphoribosyltransferase; FITC, fluorescein isothiocyanate; DAPI, 4<sup>0</sup> ,6 diamidino-2 phenilindoledihydrocloride; E, 17β-estradiol; P, progesterone; MTT, 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide; DIV, day in vitro; GH, growth hormone; IGF-1, insulin-like growth factor-1; TFAM, mitochondrial transcription factor A; mtDNA, mitochondrial DNA; DG, dentate gyrus; GABA, γ-aminobutyric acid.

Humanin expression was lower in the hippocampus of ovariectomized rats and its immunoreactivity colocalized with astroglial markers. Chronic ovariectomy also promoted the presence of less complex astrocytes in this area. Ovarian hormones increased humanin intracellular content and release by cultured astrocytes. Humanin prevented glutamate-induced dendritic atrophy and reduction in puncta number and total puncta area for pre-synaptic marker synaptophysin in cultured hippocampal neurons. In conclusion, astroglial functional and morphological alterations induced by chronic ovariectomy resemble an aging phenotype and could affect astroglial support to neuronal function by altering synaptic connectivity and functionality. Reduced astroglial-derived humanin may represent an underlying mechanism for synaptic dysfunction and cognitive decline after menopause.

Keywords: astrocytes, ovarian hormones, mitochondria, humanin, synapse, hippocampus

# INTRODUCTION

Proper functioning of the central nervous system (CNS) requires the tight intercommunication between two main cell types: neurons and glial cells. It is now well accepted that glia functions go much beyond mere structural and metabolic support to neurons, indeed they are now recognized as key players in diverse physiological processes such as synaptic communication and plasticity, homeostasis and network-level activity in the adult brain (Allen and Lyons, 2018). Neuron-glia interplay is highly dynamic and prone to changes as a result of the natural process of aging as well as alterations associated to different pathologies (Verkhratsky et al., 2014). During aging, both structural and physiological changes that occur in the brain have been attributed to changes in glial cells, which reduce their number and modify gene expression (Palmer and Ousman, 2018). In fact, glial rather than neuronal-specific genes have been proposed to be better predictors of age (Soreq et al., 2017). Among glial cells, astrocytes are crucial for maintaining synaptic connectivity throughout life by means of creating a perisynaptic sheath that accumulates molecules responsible for synaptic support (Verkhratsky et al., 2015). Aged astrocytes undergo profound morpho-functional alterations, which involve retraction of their cytoplasmic processes and a decline in the production of metabolic and trophic factors. These alterations have a direct impact in CNS health, as aged astrocytes reduce their neuroprotective and homeostatic capacity, altering, in turn, neuronal synaptic connectivity and functionality (Verkhratsky et al., 2010, 2014; Palmer and Ousman, 2018).

Ovarian hormones estradiol and progesterone have trophic effects involved in the maintenance of both reproductive and non-reproductive functions in different tissues (Nilsen and Brinton, 2002; Morrison et al., 2006). In the brain, they exert potent antioxidant and neuroprotective actions that promote cognitive health. In fact, their loss during aging and natural or induced menopause has been linked to several pathological conditions, such as neuroinflammation, mitochondrial dysfunction, synaptic decline, cognitive impairment and increased risk of neurodegenerative disorders (Zárate et al., 2017b). It has been reported that the density of dendritic spines as well as synapse number in the CA1 area of the hippocampus decrease with natural or surgical loss of ovarian hormones (Gould et al., 1990; Woolley and McEwen, 1992, 1993; Adams et al., 2001). Many of the beneficial effects of these hormones are mediated through their direct actions on neurons. However, astrocytes are also cellular targets of ovarian hormones and thus are highly involved in the protective and reparative actions of estradiol and progesterone (Acaz-Fonseca et al., 2014). Through binding to sex hormone receptors in astrocytes, ovarian hormones regulate several cellular, molecular and functional parameters in these cells, including the growth of astroglial cytoplasmic processes, the expression of glial fibrillary acidic protein (GFAP), glutamate transport and the release of neurotrophic and neuroprotective factors (Acaz-Fonseca et al., 2014, 2016; Palmer and Ousman, 2018). Considering the active role of astrocytes in regulating synaptic maintenance and plasticity and the dependence on ovarian hormone signaling for several of their neuroprotective functions, it is expected that decline in sex hormones after menopause results in impaired astroglial synaptic function.

Humanin (HN) is a cytoprotective 24 amino acid peptide which was originally isolated from a cDNA library constructed from nervous tissue of a patient with familial Alzheimer's disease (AD). Since then, it has been identified in different species such as mice, nematodes and rats (Niikura et al., 2004). The rat humanin homolog rattin (HNr) is a 38 amino acid peptide encoded and translated from an open reading frame (ORF) within the mitochondrial 16S ribosomal RNA (rRNA) gene (Caricasole et al., 2002; Paharkova et al., 2015). Several studies have demonstrated that HN is a potent pro-survival factor for neurons exposed to multiple cell stressors, such as Aβ oligomers and over-expression of familial AD-related genes (Hashimoto et al., 2001; Caricasole et al., 2002), serum deprivation (Kariya et al., 2002), stroke (Xu et al., 2006; Gao et al., 2017) and N-methyl-D-aspartate (NMDA)-induced excitotoxicity (Cui et al., 2014). HN and its derivatives have also proven to be beneficial in ameliorating cognitive impairment induced by Aβ, muscarinic receptor antagonists and aging in rodents (Mamiya and Ukai, 2001; Krejcova et al., 2004; Tajima et al., 2005; Niikura et al., 2011; Zhang et al., 2012; Yen et al., 2018). Moreover, HN and HNr were reported to prevent Aβ-induced spatial learning and memory impairments in rats by a mechanism involving changes in long-term potentiation (Chai et al., 2014; Wang et al., 2014), synaptic protein expression, dendritic branch number and spine density in the hippocampus (Chai et al., 2014).

It has been reported that HN exerts its neuroprotective action from the extracellular space through binding to a trimeric IL-6-receptor-related receptor(s) on the cell surface involving the receptor for ciliary neurotrophic factor α (CNTFR-α), WSX-1 and glycoprotein 130 kDa (gp130) subunits (Matsuoka and Hashimoto, 2010) and further modulation of tyrosin kinase, ERK1/2, AkT, STAT3 and JNK signaling cascades (Hashimoto et al., 2005; Matsuoka and Hashimoto, 2010; Takeshita et al., 2013; Kim et al., 2016). HN has been ubiquitously detected in different adult tissues with high metabolic rate, including skeletal and cardiac muscle, cerebral cortex, hippocampus and liver both in humans and in rodents (Caricasole et al., 2002; Kariya et al., 2005; Muzumdar et al., 2009). The mechanisms regulating HN expression are not fully elucidated. It has been shown that HN levels decrease with age in human and mice plasma as well as in the rat hypothalamus (Muzumdar et al., 2009; Bachar et al., 2010). Also, there is a sexual dimorphism in the expression of HN in rat anterior pituitary cells, suggesting that sex hormones are involved in the regulation of HN biosynthesis in this gland (Gottardo et al., 2014). In the brain, HN has been widely localized both in neurons and glial cells, albeit the former only in pathological conditions (Tajima et al., 2002). Thus, glial cells have been suggested to be the main production sites of this peptide in the brain in physiological conditions (Niikura et al., 2004). However, astroglial HN expression and release and the regulation of such processes by ovarian hormones have not been studied so far.

All things considered, the aim of this work was to evaluate the expression of HNr in the hippocampus of a rat model of surgicalinduced menopause. We also aimed at determining whether astroglia is able to produce and release HNr in vitro and ovarian hormone dependence on these processes. HN putative actions on structural synaptic plasticity in a model of glutamate-induced dendritic atrophy were also studied.

# MATERIALS AND METHODS

#### Drugs

All drugs and reagents were obtained from Sigma Chemical Co., St. Louis, MO, USA except for Dulbecco's Modified Eagle Medium (DMEM) and supplements (Gibco, Invitrogen Carlsbad, CA, USA) fetal calf serum (FCS; Natocor, Córdoba, Argentina) and the materials indicated below.

# Animals

Adult female Wistar rats were housed in groups of four in controlled conditions of light (12 h light-dark cycles) and temperature (20–22◦C). Rats were fed standard lab chow and water ad libitum and kept in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Animal protocols were previously approved by the Ethics Committee of the School of Medicine, University of Buenos Aires (Res. No. 2249).

For in vivo experiments, rats were ovariectomized (OVX) or sham-operated (SHAM) at 3 months of age under ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.) anesthesia and ketoprofen (5 mg/kg) for analgesia. Beginning on week 10 postsurgery, rat hormonal status was monitored daily by vaginal smears. SHAM animals had 4–5 days estrous cycles while OVX animals presented continuous diestrus status. Eleven weeks after the surgery, rats were subjected to behavioral tests as described below. Twelve weeks after the surgery, rats were either deeply anesthetized (100 mg/kg ketamine and 6 mg/kg xylazine, i.p.), transcardially perfused with heparinized saline solution and fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (for free-floating immunostaining of brain sections) or euthanized in a CO<sup>2</sup> chamber followed by decapitation (for real-time quantitative polymerase chain reaction (RT-qPCR) assays).

# Behavioral Tests

#### Open Field Test

The open field test was performed to evaluate animal general locomotor activity and exploratory behavior (Gould et al., 2009). The arena consisted of a squared open field (60 × 60 cm) limited by a 40 cm-height wall with a grided floor divided into squares (15 × 15 cm) by lines. Animals were individually placed in the center of the open field arena and were allowed to freely explore for 10 min. The frequency with which the animal crossed grid lines with all four paws (crossings) was recorded as a measure of locomotor activity. After each animal was tested, the open field was cleaned with a 10% ethanol-damp cloth. Testing was performed between 10:00 and 14:00 h in a quiet room illuminated with a 75 W electric bulb, hung 75 cm above the open field apparatus.

#### Y-Maze Spontaneous Alternation Test

Spontaneous alternation behavior in a Y-maze was recorded to evaluate animal spatial working memory (Miedel et al., 2017). The apparatus consisted of three identical black arms (50 × 10 × 40 cm, length × width × height). Animals were habituated to the testing room for at least 30 min prior to the test. At the beginning of the session, animals were placed individually at the end of one same arm of the Y-maze and allowed to freely explore for 6 min. The whole session was recorded using a SONY CCD-TRV75 video camera recorder connected to a personal computer with AVerTV A833 video capture. The number of total arm entries and the number of triads (referred as entries to a different arm of the maze in each of three consecutive arm entries) were recorded and the percentage of alternation was calculated as [number of alternations/(total arm entries-2)] × 100 (Miedel et al., 2017).

#### Elevated Plus Maze

The elevated plus maze test was carried out to evaluate animal anxiety-like behavior (Walf and Frye, 2007). The apparatus consisted of a plus-shaped maze containing two open arms (50 × 10 cm, length × width) and two enclosed arms by 40 cm high walls arranged such that the two open arms were opposite to each other. The apparatus was placed on four legs so that it was elevated 50 cm off the floor. Testing was performed between 10:00 and 16:00 h in a quiet room illuminated with a dim light hung 75 cm above the center of the maze. At the beginning of the session, each animal was placed in the center of the maze facing a close arm and allowed to freely explore for 5 min. The whole session was recorded and behavior was assessed offline. The number of times the animal entered an arm with all four paws was recorded. The percentage of open arm entries was calculated as [number of open arm entries/total arm entries] × 100.

#### Forced Swimming Test

The forced swimming test was performed to evaluate animal depressive-like behaviors by quantifying their mobility and immobility and associated behaviors (Overstreet, 2012). Each rat was individually placed in a plastic cylinder (diameter, 40 cm; height, 35 cm) containing water (23−25◦C) up to 25 cm from the bottom for 5 min. At 5 s intervals throughout the test session, the predominant behavior was assigned to one of the followings categories: (1) immobility: lack of movement, except for the ones needed to keep the head above water; (2) swimming: swimming movement throughout the cylinder or (3) climbing: vigorous movements of the forepaws in and out of the water, usually directed against the walls. After each animal was tested, the water was changed and the cylinder rinsed with clean water. Following the session, each animal was dried and placed to its housing cage in a temperature-controlled room. All swimming sessions were carried out between 10:00 and 16:00 h.

#### RNA Isolation and Reverse-Transcription Real-Time Quantitative Polymerase Chain Reaction (RT-qPCR)

Immediately after decapitation, hippocampi from SHAM and OVX rats were dissected on ice, snap-frozen and kept at −80◦C until use. Total RNA was extracted from frozen tissue using QuickZol reagent (Kalium Technologies, Buenos Aires, Argentina) according to the manufacturer's protocol. One µg of total RNA was treated with 2 U DNAse (Promega Corp., Madison, WI, USA) and then reverse transcribed using SuperScript II Reverse Transcriptase (Invitrogen, Thermo Fisher Scientific) following the manufacturer's instructions. Amplification of the products from RT reactions was performed in duplicate using specific primers (HNr forward 5<sup>0</sup> -GAG GGT TCA ACT GTC TCT TAC TTT CA-3<sup>0</sup> , reverse 5<sup>0</sup> -GTG AAG AGG CTG GAA TCT CCC-3<sup>0</sup> ; HPRT forward 5<sup>0</sup> -CTC ATG GAC TGA TTA TGG ACA GGA C-3<sup>0</sup> , reverse 5<sup>0</sup> -GCA GGT CAG CAA AGA ACT TAT AGC C-3; Invitrogen, Thermo Fisher Scientific) and SYBR Green Select Master Mix (Invitrogen, Thermo Fisher Scientific) on a StepOneTM Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). PCR product specificity was verified by a melting curve analysis. Negative RT controls were performed by omitting the addition of the reverse transcriptase enzyme in the RT reaction, while negative template controls were performed by addition of nuclease-free water instead of cDNA. Both primer sets were pre-validated to check similar efficiency ∼2 and for the use of 2−∆∆Ct method as the quantification method (Livak and Schmittgen, 2001). Lack of statistically significant variation of endogenous reference gene HPRT expression between SHAM and OVX animals was determined (**Supplementary Figure S1**). Gene expression was normalized to HPRT using Step-One Software (Applied Biosystems, Thermo Fisher Scientific), and expressed as fold-changes relative to the control group.

#### Free-Floating Immunostaining of Tissue Sections

After animal intracardiac perfusion, brains were dissected, postfixed with 4% paraformaldehyde in 0.1 M phosphate buffer and equilibrated in 25% (w/v) sucrose in the same buffer. The hippocampus was serially sectioned in a freezing microtome and the free-floating coronal 30-µm-thick tissue sections were stored at −20◦C in 25% (w/v) sucrose in phosphate buffer until use. The sections were blocked with 10% (v/v) normal goat serum and incubated for 48 h with primary antibodies against HNr (1:1,500), anti-GFAP (1:500, Millipore), NF-200 (1:1,000), OLIG2 (1:50,000, Millipore) or S100B (1:1,000) followed by an hour incubation with Alexa 594- or FITC-labeled secondary antibodies (Jackson ImmunoResearch). 4 0 ,6 diamidino-2-phenilindoledihydrocloride (DAPI) was used for DNA staining. Negative controls were incubated in the absence of primary antibodies (**Supplementary Figure S2**).

#### Astroglial Cell Culture

Astroglial cell cultures were prepared from neonatal rat pups of 3–4 days old. Hippocampi and cortices from four to six pups were isolated, cut into small fragments using scissors and then cells were mechanically dispersed by extrusion through a Pasteur pipette in Hank's Balanced Salt Solution. After decanting tissue, supernatant was transferred to a new tube and pelleted. The procedure was repeated twice and finally the cells were plated in poly-D-lysine coated bottles with high glucose DMEM supplemented with 10% FCS, 100 µg/ml penicillin-streptomicin (DMEM-S) and 1 µg/ml fungizone. After the first 24 h and every 3–4 days, media was replaced with fresh DMEM-S. When the cells reached confluence (10–12 days), they were subjected to shaking at 180 rpm for two consecutive 24 h-periods at 37◦C to detach microglia and oligodendrocytes. The cells were then incubated with 0.625% 5-fluorouracil for further 24 h, washed and incubated in fresh DMEM supplemented with 10% FCS previously treated with 0.025% dextran-0.25% charcoal (FCS-DCC) to remove steroids for additional 48 h. Then the cells were tripsinized and re-seeded onto 12 well plates in DMEM FCS-DCC for 24–48 h. Cultures obtained with this procedure showed 93 ± 0.04% GFAP-positive astrocytes (n = 3 independent cultures), as reported in the literature (Villarreal et al., 2014). Then, the cells were cultured for 4 h in DMEM FCS-DCC containing 1 nM 17β-estradiol (E) and 1 µM progesterone (P) or vehicle (ethanol, 20 µl/l), washed and cultured for further 20 h in fresh DMEM FCS-DCC.

#### Expression of HNr in Astroglial Cells by Flow Cytometry

Cultured astroglial cells were harvested with 0.025% trypsin-EDTA, washed in cold PBS, fixed with 4% paraformaldehyde and permeabilized with 0.1% saponin in PBS (MP Biomedicals Inc., OH, USA) for 10 min. Then, the cells were incubated with rabbit anti-HNr antibody (1 µg/µl) in PBS-0.05% saponin for 1 h at 37◦C followed by 40-min incubation with a FITC-conjugated anti-rabbit secondary antibody (1:100) in the same buffer. To determine the cut-off for HNr fluorescence, cells were incubated with secondary antibody only. Cells were washed, resuspended in PBS and analyzed by flow cytometry using a FACScan (Becton Dickinson). Data were analyzed with WinMDI 98 software.

#### ELISA

HNr levels in supernatants from astroglial cultures were measured by ELISA using a commercial kit HN(N) (Rat)- EIA Kit (Phoenix Pharmaceuticals) following the manufacturer's instructions and normalized to 50 µg of total protein in cell lysates from corresponding wells. Attached astrocytes were harvested as described above. Cells were lysed in lysis buffer containing 150 mM NaCl, 1% Igepal, 0.02% sodium azide, 0.1% sodium dodecyl sulfate (SDS), in 50 mM Tris-HCl pH 7.4 and a protease inhibitor cocktail (1:50). Cell lysate was collected and clarified by centrifugation at 16,000× g for 30 min and total protein content was determined by the Bradford protein assay (BioRad Laboratories, CA, USA) using bovine serum albumin as standard.

#### Metabolic Activity of Viable Cells

The metabolic activity of viable cells was determined by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Promega, Madison, WI, USA). Cells were washed twice and incubated for 4 h in 100 µl Krebs buffer plus 50 µg MTT reagent in PBS at 37◦C. The developed crystals were dissolved in 100 µl 0.04 N HCl in isopropanol and the OD was read in a microplate spectrophotometer at a wavelength of 600 nm. The quantity of formazan product is directly proportional to the number of living cells in culture.

#### Primary Neuronal Cultures and Glutamate Treatment

Hippocampal neuronal cultures were prepared from dissected embryonic day 18 hippocampi as previously reported (Reinés et al., 2012; Podestá et al., 2014). Briefly, hippocampal tissue was trypsinized and mechanically dissociated. Then, dispersed cells were seeded on poly-D-lysine-coated glass coverslips at a density of 2 × 10<sup>4</sup> cells/cm<sup>2</sup> in Neurobasal medium supplemented with 2% (v/v) B27 and 0.5 mM glutamine. On day 13 in vitro (DIV), neurons were treated with 5 µM glutamate or vehicle for 3 min at 37◦C, media was immediately removed and cells were washed with Hank's balanced salt solution. Immediately after, neurons were incubated with Humanin peptide (HN, 0.01–1 µM; Genemed Synthesis, Inc., San Antonio, TX, USA) in supplemented culture media for 24 h and then fixed for immunostaining as described below.

#### Immunostaining of Fixed Cells

Neurons in culture (14 DIV) were fixed in 4% (w/v) paraformaldehyde/4% (w/v) sucrose in PBS solution, pH 7.2 for 20 min at room temperature (RT) and permeabilized with 0.2% (v/v) Triton X-100 for 10 min at RT followed by a blockade with 5% (v/v) normal goat serum for 1 h at RT. The cells were then incubated overnight at 4◦C with MAP-2 (1:500) or SYN (1:2,000; Chemicon, Millopre) primary antibodies in PBS. The next day, cells were washed and incubated for 1 h at RT with the appropriate secondary antibodies followed by DAPI for DNA staining. Negative controls were incubated in the absence of primary antibodies (**Supplementary Figure S3**). Finally, cells were mounted using Mowiol.

#### Image Acquisition and Quantification

A Zeiss Axiophot microscope (Carl Zeiss, Oberkochen, Germany) equipped with an Olympus Q-Color 5 camera or a Olympus IX81 microscope equipped with a CCD model DP71 digital camera were employed in immunohistochemistry or immunocytochemistry assays, respectively. Confocal images were acquired using Fluoview version 3.3 software using an Olympus FV300 confocal microscope. Immuno-positive structures were quantified as relative immunoreactive area (immunoreactive area/total area) as described elsewhere using the ImageJ (NIH) software (Reinés et al., 2008; Aviles-Reyes et al., 2010; Codagnone et al., 2015). In immunohistochemistry assays, the total area corresponded to the hilus, granular and molecular layers in the superior and inferior blades of the dentate gyrus (DG) and to pyramidal and stratum lucidum and radiatum in CA1. Briefly, microscopic images were captured with a digital camera, transformed to an 8-bit gray scale and an interactive threshold was determined. Then, the area fraction covered by immunostained structures obtained using this threshold, which remained fixed for the entire experiment, was quantified with the particle counting tool of the software. In immunocytochemistry assays, dendritic tree area per neuron was calculated by subtracting the area corresponding to the immunolabeled soma from the total MAP-2 immunostaining. The number of synaptic puncta, total puncta area and individual puncta area for SYN was calculated as previously described (Podestá et al., 2014). Figures were prepared using Adobe Photoshop 7.0 software (Adobe Systems Inc.). Brightness and contrast were kept constant between the experimental groups. Each immunohistochemistry assay consisted of 5–6 hippocampal serial sections of each animal per group. The average data obtained from the quantification of sections from the same animal was considered n = 1. Results of immunohistochemistry assays are expressed as mean values (±SEM) of n = 3–5 animals per group. Each experiment was repeated 2–4 times. Each immunocytochemistry assay consisted of 1–3 coverslips per experimental condition. Results of immunocytochemistry assays are expressed as mean values (±SEM) of 20–40 neurons per experimental condition from two to three independent cultures. The experiments were repeated at least twice.

#### Statistical Analysis

Results are expressed as mean ± SEM and evaluated by unpaired Student's t-test or nonparametric Mann-Whitney U test. HNr content in conditioned media determined by ELISA was evaluated by one-way ANOVA followed by Tukey's test. Dendritic tree area, total SYN puncta area, number and individual SYN puncta area were evaluated by two-way ANOVA followed by Tukey's test. Differences were considered significant if p < 0.05.

#### RESULTS

#### Behavioral Characterization of OVX-Induced Model of Menopause

It is now accepted that ovariectomized (OVX) adult rodents share common features with aged animals regarding brain function, including changes in mitochondrial function, synaptic plasticity, behavior and cognition (Zárate et al., 2017b). To characterize our animal model of surgical menopause in adult Wistar rats at the behavioral level, we assessed parameters of spatial working memory, anxiety and depression, which are known to be modulated by ovarian hormones (Diz-Chaves et al., 2012; Kiss et al., 2012; Cao et al., 2013; Rodríguez-Landa et al., 2017; Hampson, 2018; da Silva Moreira et al., 2016). Twelve-week ovarian hormone-deprived rats showed lower spontaneous alternation behavior in the Y-maze, indicating impaired spatial working memory (**Figure 1A**). They also spent less time in the open arms and more time immobile when subjected to the elevated plus maze and the forced swimming test, respectively. These results show that long-term ovarian hormone deprivation increases anxiety-like and depressive-like behaviors (**Figures 1B,C**). There were no differences in spontaneous locomotion and exploratory behavior between OVX and control groups, as assessed in the open field test (**Figure 1D**), indicating that the observed behaviors in OVX rats cannot be attributed to altered locomotion.

#### HNr Expression in the Hippocampus

Circulating levels of mitochondrial-encoded HN are known to decline with age in mice and humans and in the rat hypothalamus (Muzumdar et al., 2009; Bachar et al., 2010). Indeed, this peptide has been suggested to participate in the endocrine regulation of the aging process (Lee et al., 2014). We have previously reported that long-term ovarian hormone deprivation induces functional and structural alterations in hippocampal mitochondria, which resembles a mitochondrial aging phenotype (Zárate et al., 2017a). We thus evaluated HNr expression in the hippocampus of OVX rats by RT-qPCR. Hippocampal expression of HNr was lower in OVX rats compared to control rats (**Figure 2**). In order to identify cell types expressing this peptide, we performed immunohistochemistry for both HNr and different brain cell markers. HNr immunostaining

FIGURE 1 | Behavioral characterization of an animal model of surgical menopause in adult rats. Adult Wistar female rats were ovariectomized (OVX) or sham-operated (SHAM). Twelve weeks after surgery, the animals were subjected to behavioral tests as described in "Materials and Methods" section. (A) Y-maze spontaneous alternation test: for each animal, the number of total arm entries and the number of triads in 6 min was recorded. (B) Elevated plus maze test: each animal was allowed to freely explore for 5 min and the number of times the animal entered an arm with all four paws was recorded. (C) Forced swimming test: every 5 s, a time-sampling technique was used to score the presence of immobility, swimming or climbing behavior. (D) SHAM or OVX rats were placed individually in the center of a field marked with a grid of 16 equal squares for 10 min. The number of times the animal crossed each line was registered. Each column represents the mean ± SEM of (A) the percentage of alternation, (B) the percentage of open arm entries, (C) the number of counts or (D) the number of crossings per session (n = 4–12 animals/group); <sup>∗</sup>p < 0.05, ∗∗∗p < 0.001, Student's t-test.

colocalized with astroglial marker S100B and GFAP, but not with neuronal or oligodendroglial markers NF-200 and OLIG-2, respectively (**Figures 3**, **4A**). Noteworthy, some

RNA was evaluated in hippocampi from SHAM and OVX rats by real-time quantitative polymerase chain reaction (RT-qPCR). Each column represents the mean ± SEM of the concentration of HNr RNA relative to its internal control HPRT expressed as fold-changes relative to SHAM group (AU; n = 3 animals/group); <sup>∗</sup>p < 0.05, Student's t-test.

HNr staining could be observed in the extracellular space, suggesting the presence of secreted HNr. Further evaluation of HNr and GFAP expression levels in different hippocampal subregions showed that HNr relative immunoreactive area was lower in the CA1 region of the hippocampus of OVX rats while there was no difference between HNr relative immunoreactive area from OVX and control rats in the DG (**Figure 4B**). OVX rats also showed lower relative immunoreactive area for GFAP in all hippocampal subregions studied (**Figure 4B**). Remarkably, there was a positive correlation for HNr and GFAP levels in both OVX and control groups (**Figure 4C**).

#### HNr Production and Release by Astroglia

Glial cells have been proposed as the main site of HNr production in the brain in physiological conditions (Tajima et al., 2002). Indeed, our results show that HNr immunoreactivity colocalizes with astroglial markers and that OVX rats display lower expression of HNr in the hippocampus. However, the production and release of this peptide by astroglial cells as well as the regulation of these processes by ovarian hormones remain to be determined. To address this issue, cultured astrocytes were incubated with ovarian hormones 17β-estradiol (E) and progesterone (P) alone or in combination and released

image (yellow). PYR, pyramidal layer; CC, corpus callosum. Scale bar = 50 µm.

p < 0.05, Pearson correlation test). PYR, pyramidal layer; CC, corpus callosum; DG, dentate gyrus. Scale bar = 50 µm.

HNr content was determined in the conditioned media of cultured astrocytes by ELISA. There was a two-fold increase in HNr levels in the conditioned media of astrocytes incubated with E + P (**Figure 5A**). Then, intracellular astroglial HNr levels were determined by immunocytochemistry and flow cytometry. Ovarian hormones increased the intracellular content of HNr per cell without changing the number of astrocytes expressing this peptide (**Figures 5B,C**), indicating that ovarian hormones increase both the expression and the release of HNr by these cells.

# HN Effect on Structural Synaptic Plasticity in the Hippocampus

It has been extensively reported that ovariectomy decreases the number of synapses and induces dendritic alterations in hippocampal neurons (Gould et al., 1990; Woolley and McEwen, 1992, 1993; Adams et al., 2001). However, HN effects on these synaptic parameters have not been explored yet. To this aim, cultured hippocampal neurons were subjected to a brief exposure to a low glutamate concentration, a condition that has previously been reported to induce dendritic atrophy in the absence of neuronal death (Podestá et al., 2014). Immediately after glutamate exposure, neurons were incubated with HN in two different concentrations and cell viability and synaptic parameters were studied 24 h later. Neither glutamate nor HN affected cell viability, as assessed by MTT assay (**Figure 6**). As expected, immunostaining for the specific dendritic marker MAP-2 was reduced 24 h after glutamate insult, rendering into a decreased neuronal dendritic area. Remarkably, HN prevented glutamate-induced dendritic atrophy in both concentrations studied (**Figure 7**). Pre-synaptic marker SYN immunostaining was also altered in glutamate-treated neurons, as previously reported (Podestá et al., 2014). Remarkably, while glutamate decreased SYN puncta number and total puncta area, both concentrations of HN prevented glutamate actions over SYN synaptic profile. Neither

#### glutamate nor HN induced changes in SYN individual puncta area (**Figure 8**).

# DISCUSSION

Mitochondrial DNA has been classically described as a maternally-inherited, small DNA encoding only 13 mitochondrial proteins involved in oxidative phosphorylation and 24 structural RNAs required for their translation. However, recent research has challenged this view (Capt et al., 2016; Luo et al., 2018). In fact, several mitochondrial-derived peptides encoded as genes-within-genes in short ORFs dispersed throughout the mitochondrial genome have been recently described (Capt et al., 2016; Kim et al., 2017). Among them, humanin (HN) has received much attention due to its neuroprotective effect against different types of stress and disease models (Lee et al., 2013). Encoded within the 16S rRNA, HN has been detected in several tissues and in circulation in rodents and humans (Caricasole et al., 2002; Kariya et al., 2005; Muzumdar et al., 2009) and its expression is age-dependent (Muzumdar et al., 2009; Bachar et al., 2010). HN circulating levels both in mice and humans are regulated by the growth hormone and insulin-like growth factor-1 (GH/IGF-1) axis, a well conserved

endocrine system that controls the process of aging. In fact, HN has been suggested to be a mitochondrial signaling peptide that is secreted and acts as a hormone involved in the endocrine regulation of the aging process (Lee et al., 2014). We previously reported that 12-week ovarian hormone deprivation, a model of surgical menopause, induces mitochondrial dysfunction in the rat hippocampus (Zárate et al., 2017a). These alterations comprise slower active respiration and ATP production rates as well as decreased membrane potential together with changes in the lipid composition of mitochondrial membranes. Remarkably, these functional and structural features are found in mitochondria from aged animals (Pamplona, 2008; Gómez and Hagen, 2012), further supporting the idea that ovarian hormone loss promotes an accelerated aging phenotype, as previously suggested (Yao et al., 2009). Ovarian hormones, especially estrogens, are well-known regulators of mitochondrial function, which is crucial in organs and tissues with high energy demand like the CNS (Zárate et al., 2017b). Estrogen regulation of mitochondrial metabolism, biogenesis and morphology has been reported to occur in neuronal tissue (Garcia-Segura et al., 1998; Nilsen and Brinton, 2003; Arnold et al., 2008; Brinton, 2008; Hara et al., 2014; Kemper et al., 2014; Klinge, 2017). Since ovarian hormone loss has been associated to the process of aging and estrogens are well-known regulators of mitochondrial gene expression (Virbasius and Scarpulla, 1994; Kang et al., 2007), we hypothesized that the levels of a mitochondrial-encoded

gene like HN would be decreased in hormone-deprived animals. Indeed, herein we found that long-term ovariectomy reduced the expression of HNr in the hippocampus. It is to note that, although initially transcribed as long polycistronic precursor transcripts (Ojala et al., 1981), the levels of individual mitochondrial RNAs varies in different tissues and cell types due to post-transcriptional processing mechanisms (Mercer et al., 2011; Sanchez et al., 2015). Such mechanisms have been shown to respond to the tissue hormonal milieu (Sanchez et al., 2015). In fact, estrogen treatment has been shown to increase the amount of mitochondrial-encoded RNAs (Stirone et al., 2005; Klinge, 2008). Moreover, 16S rRNA levels have been reported to decrease with aging due to reduced rate of mitochondrial RNA transcription (Calleja et al., 1993). Remarkably, female rats express more 16S rRNA than males of the same chronological age (Borras et al., 2003), suggesting a putative role of sex hormones in the regulation of its expression. Considering that HNr is encoded within the mitochondrial 16S rRNA molecule, it can be speculated that similar post-transcriptional processing mechanisms induced by hormonal environment can affect HNr expression. Also, HN peptide seems to be associated with lipids in different tissues (Tajima et al., 2002), suggesting that alterations in tissue lipid composition induced by age or disease may affect HN effective levels. Interestingly, hippocampal mitochondria membranes from OVX rats display an altered lipid profile consisting of increased membrane peroxidability index together with decreased cardiolipin levels (Zárate et al., 2017a), which could also affect HN levels and/or distribution within this tissue.

Different cells comprise the highly specialized and complex nervous tissue. Although an early report showed the presence of HN immunoreactivity in both normal and AD human brains (Tajima et al., 2002), the specific cell types expressing this peptide remained elusive. Tajima et al. (2002) showed HN immunoreactivity to be located in round cells resembling glia widely distributed in the brain, mainly in the hippocampus. However, the identity of such glial cells was not investigated then. HN immunoreactivity was also detected in some neurons of the occipital lobe in the AD brain but not in an age-matched normal brain (Tajima et al., 2002). By means of double immunohistochemistry using neuronal, astroglial and olidodendroglial markers in the hippocampus, we detected HNr immunoreactivity to be located only in astrocytes, both in OVX and control animals. However, co-immunostaining with microglial markers was not studied and HNr localization in microglial cells cannot be ruled out. Considering that microglia are highly regulated by both estrogen and progesterone, further studies are warranted to elucidate whether microglial cells also express HNr in vivo.

The role of astroglial cells as supportive cells in the CNS is well established. They are not only the major contributors to cell homeostasis in the CNS but also play a central role in the control of synaptic transmission through different mechanisms. Perisynaptic astroglial membranous sheaths are known to cover a high number of all synaptic contacts in the hippocampus, thus finely regulating synaptic transmission by means of establishing physical contact with synapses and also by secreting a plethora of bioactive agents (Verkhratsky and Nedergaard, 2018). These astroglial processes have been reported to be devoid of organelles but may contain mitochondria, which, in light of our results, could be the source of mitochondria-derived peptides with a local role in the neighboring synapse. Considering the effect of HN in the prevention of glutamate-induced structural synaptic alterations shown herein, this peptide could be considered as another bioactive molecule locally secreted by astroglia to regulate synaptic plasticity in pathophysiological conditions. In this way, our results add new evidence for astroglial role as a major contributor to proper signal transmission in the CNS and therefore to higher cognitive function.

In vitro assays using cultured astrocytes further confirmed that this cell type is able to produce and release HNr and that ovarian hormones positively regulate these processes. It is well known that estradiol and progesterone exert their actions on astrocytes through classical and non-classical receptor signaling initiated at the nucleus, membrane or cytoplasm levels (Acaz-Fonseca et al., 2016). By inducing the transcription of nuclear-encoded mitochondrial transcription factor A (TFAM), estrogens are able to promote transcription of mitochondrial DNA (mtDNA) through a classical mechanism (Virbasius and Scarpulla, 1994; Kang et al., 2007). Also, estrogen and progesterone receptors have been localized within mitochondria, suggesting that they might regulate mitochondrial transcription through direct binding to hormone-response element-like sequences in mtDNA (Demonacos et al., 1996; Chen et al., 2004). Further studies are warranted to determine ovarian hormone mechanism involved in HNr expression and release by astroglial cells.

Estradiol and progesterone are well-known regulators of astroglial cell morphology and GFAP expression. Evidence supporting this includes changes in the surface density of GFAP-immunoreactive cells in the DG of the hippocampus along the estrous cycle (Luquin et al., 1993). Moreover, GFAP-immunoreactive cell density in this brain area decreases after ovariectomy and is increased by treatment with estradiol alone or in combination with progesterone in a dose-dependent manner (Luquin et al., 1993). It has been suggested that ovarian hormones promote the increase in the size and/or branching of cell processes without affecting astrocyte cell number (Tranque et al., 1987). We evidenced the presence of qualitatively smaller, less complex astrocytes with thinner cytoplasmic processes, which resulted in lower GFAP relative immunoreactive area in all subregions of the hippocampus from OVX animals. Similar alterations in astrocyte phenotype have been described in aged rodents, primates and humans (Castiglioni et al., 1991; Amenta et al., 1998; Kanaan et al., 2010; Cerbai et al., 2012; Jyothi et al., 2015; Robillard et al., 2016). For example, Cerbai et al reported the presence of fewer, smaller and less complex GFAP-positive astrocytes in the CA1 region of the hippocampus from 22-month old rats (Cerbai et al., 2012). These morphological changes are accompanied with loss of homeostatic function, which represents an underlying mechanism for impaired neuroprotection and disrupted neuronal connectivity (Verkhratsky et al., 2014). Indeed, several recent reports have shown that the atrophic astrocyte is the main astroglial phenotype not only in natural aging but also in the early stages of neurodegenerative diseases (Verkhratsky et al., 2014). This phenotype is characterized not only by smaller and less complex astrocytes but also by a decrease in glutamate uptake and glutamate synthase activity, which may contribute to the observed general imbalance in both excitatory and inhibitory neurotransmission as well as alterations at the synapse level (Verkhratsky et al., 2014). It is of note that astrocyte cytoplasmic processes form close structural contacts with synapses to regulate all aspects of synaptic function via the secretion of several factors (Pfrieger, 2010). Thus, it can be speculated that OVX may negatively impact astroglial support to neuronal function by means of reduced synapse maintenance.

Accumulating evidence indicates that the CA1 region of the hippocampus highly responds to natural or surgical ovarian hormone loss by decreasing synapse number and spine density in rats (Gould et al., 1990; Woolley and McEwen, 1992, 1993; Adams et al., 2001), which has been directly associated to cognitive impairment. Interestingly, quantification of HN immunoreactivity in the hippocampus of OVX rats evidenced lower HN protein levels in the CA1 region without changes in the DG between experimental groups, which was temporally correlated with impaired spatial working memory. A HN derivative has been reported to improve cognitive impairments in different genetic mouse models of AD (Niikura et al., 2011; Zhang et al., 2012) as well as blunt learning and memory decline induced by Aβ peptides (Tajima et al., 2005). Also, HN ameliorates spatial working memory deficits induced by cholinergic antagonism- and GABA agonist-induced amnesia in mice (Mamiya and Ukai, 2001; Krejcova et al., 2004; Tajima et al., 2005). It has been recently reported that HN treatment improves cognition in aged mice. Also, there is a positive correlation between decreased HN circulating levels and accelerated cognitive aging in humans (Yen et al., 2018). Moreover, intrahippocampal injection of HN into the CA1 region is able to prevent Aβ-induced memory deficits (Chai et al., 2014). Thus, it can be hypothesized that decreased HN expression in the CA1 region of the hippocampus could be the underlying mechanism for cognitive impairment induced by OVX.

It is well recognized that structural synaptic plasticity, which involves changes in synaptic architecture and number, is an important biological basis of learning and memory (Lamprecht and LeDoux, 2004). In vivo treatment with HN has been shown to prevent Aβ-induced dendritic atrophy in the CA1 region of the hippocampus by promoting dendritic branching and spine density. These HN-induced effects occur concomitantly with an increase in pre- and post-synaptic proteins in this hippocampal subregion (Chai et al., 2014). To our knowledge, ours is the first in vitro study showing that HN has a direct effect at the synaptic level preventing glutamate-induced structural synaptic alterations in hippocampal neurons.

HN has been shown to protect cultured rat cortical neurons from NMDA-induced neurotoxicity, an effect that seems to be time- and concentration-dependent (Cui et al., 2014, 2017; Yang et al., 2018). In this study, we aimed at studying HN effects on structural synaptic plasticity in glutamate-induced dendritic atrophy and synapse alterations (Podestá et al., 2014). This approach, which consists on subjecting cultured hippocampal neurons to a brief exposure to a low glutamate concentration, proved to be mediated by NMDA receptor (Podestá et al., 2014). Our results show that HN is able to prevent glutamateinduced dendritic atrophy even at the lowest concentration studied. Interestingly, we also detected a reduction in synapse number, evidenced by decreased SYN puncta number and total SYN puncta area, which occurs simultaneously with dendritic retraction. Remarkably, HN prevented these structural synaptic alterations at concentrations that are comparable to HNr levels detected in astrocyte conditioned media by ELISA, suggesting that HN could be exerting its actions in pathophysiological conditions.

In summary, our results indicate that long-term ovarian hormone deprivation promotes structural changes in hippocampal astrocytes, which is positively correlated with reduced HN expression in this brain area. Our results in vitro show that ovarian hormones positively regulate astroglial HN expression and release and that this peptide prevents glutamateinduced structural synaptic alterations of cultured hippocampal neurons. Thus, OVX-induced functional and morphological alterations in astrocytes discussed above could impair astroglial support to neuronal function and may represent an underlying mechanism for synaptic dysfunction after menopause. Our study could help find new therapeutic targets for interventions that may promote a healthier lifespan for post-menopausal women.

# ETHICS STATEMENT

Animals were kept in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Animal protocols were previously approved by the Ethics Committee of the School of Medicine, University of Buenos Aires (Res. N◦ 2249).

# AUTHOR CONTRIBUTIONS

SZ and AR contributed to the conception and design of the study. SZ and MT performed the experiments. MC contributed to the design and performance of behavioral tests. SZ wrote the first draft of the manuscript. SZ, AR and AS discussed results. AR and AS revised the first draft of the manuscript. All authors contributed to the revision of the final version of the manuscript, read and approved the submitted version.

# FUNDING

This study was supported by Agencia Nacional de Promoción Científica y Tecnológica (ANPCYT; PICT 2014-1769 and PICT 2014-0334), Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET; PIP 11220130100212CO) and Universidad de Buenos Aires (UBA; UBACyT 20020150200250BA, 20020130100020BA and 20020170100478BA). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi. 2019.00123/full#supplementary-material

FIGURE S1 | Validation of the 2−∆∆Ct method for quantitative polymerase chain reaction (qPCR) data quantification. The efficiency of amplification of (A) the target gene (HNr) and (B) housekeeping control (HPRT) was examined using real-time PCR and SYBR Green detection. Using reverse transcriptase, cDNA was synthesized from 1 µg total RNA isolated from sham-operated (SHAM) and

ovariectomized (OVX) rat hippocampi. Serial dilutions of cDNA were amplified by real-time PCR using gene-specific primers. The most concentrated sample contained 200 ng of cDNA. (C) The ∆CT (CtHNr-CtHPRT) was calculated for each cDNA dilution. The data were fit using least-squares linear regression analysis (n = 2–3). (D) Validation of HPRT as housekeeping gene. The expression of HPRT RNA was evaluated in hippocampi from SHAM and OVX rats by RT-qPCR. Each column represents the mean ± SEM of the concentration of HPRT RNA expressed as fold-changes relative to SHAM group (AU; n = 3 animals/group); p = 0.0802, Student's t-test.

FIGURE S2 | Negative controls for immunohistochemistry assays. Coronal sections from the hippocampus of SHAM or OVX animals were processed for double immunohistochemistry for HNr and GFAP. (A) Representative microphotographs from CA1 stratum oriens show the lack of signal in the 488 (green) and 594 (red) channels in sections incubated only with secondary antibodies (double negative control). (B) Representative microphotographs from

#### REFERENCES


CA1 stratum oriens show the expression of GFAP (green) or HNr (red) together with the lack of signal in the 594 (red) or 488 (green) channels respectively in sections incubated in the presence of each primary antibody as indicated in the figure and both secondary antibodies (single negative control). Nuclear staining is shown in blue (DAPI). Scale bar = 50 µm.

FIGURE S3 | Negative controls for immunocytochemistry assays. Hippocampal neurons in culture (DIV 13) were immunostained for double immunohistochemistry for MAP-2 and SYN. (A) Representative microphotographs show the lack of signal in the 488 (green) and 590 (red) channels in neurons incubated only with secondary antibodies (double negative control). (B) Representative microphotographs show the expression of SYN (red) or MAP-2 (green) together with the lack of signal in the 488 (green) or 590 (red) channels respectively in neurons incubated in the presence of each primary antibody as indicated in the figure and both secondary antibodies (single negative control). Nuclear staining is shown in blue (DAPI). Scale bar = 50 µm.

A DNA Mapp. Seq. Anal. 27, 3098–3101. doi: 10.3109/19401736.2014. 1003924


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

The handling Editor declared a shared affiliation, though no other collaboration, with the authors.

Copyright © 2019 Zárate, Traetta, Codagnone, Seilicovich and Reinés. 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 Ultrastructure in the Hippocampus of a Lipopolysaccharide-Induced Sickness Mouse Model

Julie C. Savage1,2, Marie-Kim St-Pierre<sup>2</sup> , Chin Wai Hui1,2,3 and Marie-Eve Tremblay1,2 \*

<sup>1</sup> Axe Neurosciences, Centre de Recherche du CHU de Québec-Université Laval, Quebec, QC, Canada, <sup>2</sup> Départment de Médecine Moléculaire, Faculté de Médecine, Université Laval, Quebec, QC, Canada, <sup>3</sup> Division of Life Science, The Hong Kong University of Science and Technology, Kowloon, Hong Kong

#### Edited by:

Yolanda Diz-Chaves, University of Vigo, Spain

#### Reviewed by:

Robert Dantzer, University of Texas MD Anderson Cancer Center, United States Harris A. Gelbard, University of Rochester Medical Center, United States

> \*Correspondence: Marie-Eve Tremblay tremblay0202@gmail.com

#### Specialty section:

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

Received: 27 June 2019 Accepted: 27 November 2019 Published: 20 December 2019

#### Citation:

Savage JC, St-Pierre M-K, Hui CW and Tremblay M-E (2019) Microglial Ultrastructure in the Hippocampus of a Lipopolysaccharide-Induced Sickness Mouse Model. Front. Neurosci. 13:1340. doi: 10.3389/fnins.2019.01340 Sickness behavior is a set of behavioral changes induced by infections and mediated by pro-inflammatory cytokines. It is characterized by fatigue, decreased appetite and weight loss, changes in sleep patterns, cognitive functions, and lost interest in social activity. It can expedite recovery by conserving energy to mount an immune response involving innate immunity. To provide insights into microglial implication in sickness behavior with special focus on cognitive and social impairment, we investigated changes in their ultrastructure and interactions with synapses using a toxemia mouse model. Adult mice were injected with 1 mg/kg lipopolysaccharide (LPS) or saline, and assayed for signs of sickness behavior. LPS treated mice displayed reduced activity in openfield tests 24 h post-injection, while social avoidance and weight gain/loss were not significantly different between treatment groups. Microglia were investigated using electron microscopy to describe changes in their structure and function at nanoscale resolution. Microglial cell bodies and processes were investigated in the hippocampus CA1, a region responsible for learning and memory that is often impacted after peripheral LPS administration. Microglia in LPS treated animals displayed larger cell bodies as well as less complex processes at the time point examined. Strikingly, microglial processes in LPS injected animals were also more likely to contact excitatory synapses and contained more phagocytic material compared with saline injected controls. We have identified at the ultrastructural level significant changes in microglia-synapse interactions shortly after LPS administration, which draws attention to studying the roles of microglia in synaptic rewiring after inflammatory stimuli.

Keywords: sickness behavior, lipopolysaccharide, neuroinflammation, microglia, phagocytosis, hippocampus, mouse, electron microscopy

# INTRODUCTION

Sickness behavior is a well-defined set of cognitive and behavioral adaptations recruited in response to bacterial or viral infection or peripheral increases in proinflammatory cytokines. It is characterized by fatigue, joint and muscle pain, coldness, and reduced appetite (Dantzer, 2009) accompanied by psychological, emotional, and behavioral disturbance (Dantzer et al., 2008). Sepsis

is a life-threatening condition caused by the body's immune response to infection. Massive increases in blood and serum levels of cytokines and chemokines can cause organ failure and breakdown of the blood-brain barrier (BBB). Septic-associated encephalopathy is a serious complication, with symptoms including cognitive impairment and seizures.

Peripheral macrophages are able to recognize viruses, bacteria, fungi, and other invading pathogens using cell-surface receptors which recognize pathogen and danger associated molecular patterns (PAMPs and DAMPs). Upon recognition of dangerous materials, these cell surface receptors initiate signaling cascades, and culminate into a robust inflammatory response to kill and phagocytose the microorganism. PAMPs often cause the initial cascade of proinflammatory signaling in the early stages of illness, and this inflammation can damage nearby cells, thus releasing DAMPs into circulation and initiating a snowball proinflammatory signaling cascade (Hanisch and Kettenmann, 2007). Macrophages recognize gram-negative bacterial endotoxin lipopolysaccharide (LPS) using a receptor complex including tolllike receptor 4 (TLR4), CD14, and complement proteins, and secrete reactive oxygen and reactive nitrogen species to destroy the pathogen and proinflammatory cytokines in order to call nearby microglia to aid in the response (Lund et al., 2006). While this response is beneficial in most cases of infection, high levels of PAMPs and DAMPs also cause immune cells to release dangerously high levels of cytokines. As such, PAMPs, DAMPs, and cytokines have been targeted in recent therapeutics designed to treat sepsis (Gruda et al., 2018).

Microglia, the brain's resident macrophages, survey the brain and respond to any disturbances in their environment, resulting in morphological and functional changes during peripheral inflammation and infection (Butovsky and Weiner, 2018). As the main immune effector cell within the brain, they play intimate roles in the mechanisms behind sickness behaviors. During peripheral inflammation driven by increased circulating levels of proinflammatory cytokines, microglia express high levels of inducible nitric oxide producing enzymes, which remain elevated for several days. High levels of nitric oxide can result in destruction of bacteria, but also cause apoptosis of nearby neurons (Heneka et al., 1998). Elevated nitric oxide levels have been linked to synaptic loss in a protein kinase-G dependent manner (Sunico et al., 2010). Additionally, complementmediated synaptic pruning has been implicated in normal brain development, cognitive aging, and neurodegenerative disease (Presumey et al., 2017). Systemic LPS injection has also been shown to have both disruptive and non-disruptive effects on the BBB during and after inflammation (Varatharaj and Galea, 2017).

Various mouse models of endotoxemia have revealed reduction of synapses in the CA1 region of the hippocampus (Moraes et al., 2015; Zhang et al., 2017). Reduced levels of NMDA protein were reported in the CA1 region of two different mouse sickness models (Zhang et al., 2017), while other studies have found reductions in excitatory synapse number in the hippocampus using cecal ligation and puncture (CLP) models (Moraes et al., 2015). Behavioral deficits, including spatial memory tasks, have been found to be microglia-dependent as pre-emptive minocycline treatment reduced neuroinflammation, oxidative stress, and neuronal dysfunction following CLP (Michels et al., 2015, 2017). Additionally, hippocampal-dependent context discrimination memory is impaired in rats following a single peripheral injection of LPS (Czerniawski and Guzowski, 2014). Recent human studies have similarly found long-term cognitive dysfunction in sepsis survivors. Sickness behavior in humans also includes memory impairment (Capuron et al., 1999; Reichenberg et al., 2001) and even very low doses of endotoxins can cause anxiety and depressive symptoms in humans (Reichenberg et al., 2001; Krabbe et al., 2005). Cognitive dysfunction and sickness behavior was exacerbated in aged mice in response to intracerebroventricular LPS administration (Huang et al., 2008). Synaptic loss is considered the best known correlate of cognitive dysfunction, but the direct role of microglia in synaptic loss in sepsis or sickness behavior has not yet been investigated.

To provide insights into this possible involvement, the present study aimed to investigate changes in microglial ultrastructure and interactions with synaptic elements in the strata radiatum and lacunosum-moleculare of the hippocampal CA1 region, 24 h after peripheral LPS administration in mice. We chose a 24 h timepoint as previous studies have found changes in ultrastructural interactions between microglia and cortical neurons, and disruptions in inhibitory synapses 24 h after peripheral LPS injection (Chen et al., 2014). We focused on the CA1 as it is the main region implicated in spatial memory tasks, where deficits were seen in prior studies in both human cases and mouse models of illness (Michels et al., 2015; Calsavara et al., 2018; Barichello et al., 2019). Our quantitative analysis determined that microglial cell body and process ultrastructure, as well as interactions with the neuropil, including synaptic clefts, were significantly modified following peripheral LPS administration.

#### MATERIALS AND METHODS

#### Animal Model

All experimental procedures were performed in agreement with the guidelines of the Institutional Animal Ethics committees, in conformity with the Canadian Council on Animal Care and the Animal Care Committee of Université Laval. Animals were group housed three to five animals per cage under a 12-h light-dark cycle at 22–25 ◦C with free access to food and water. Four month old CX3CR1-GFP heterozygous mice on C57BL/6J background (The Jackson Laboratory) were injected intraperitoneally (i.p.) with saline or 1 mg/kg of LPS derived from Escherichia coli serotype O55:B5 (Sigma Aldrich). CX3CR1-GFP mice were used considering that a subset of mice from the same protocol were imaged using two-photon in vivo microscopy (Abiega et al., 2016; Paris et al., 2018). The dose and timing of LPS was defined by the minimal dose required to induce sickness behavior while preventing mortality in our and other studies, and coinciding with changes in microglia-neuron interactions in previous studies in mouse cortex (Chen et al., 2014; Hoogland et al., 2015). A small cohort (four saline and five LPS injected animals) were treated to verify this dose and afterward utilized for two-photon in vivo

microscopy (Abiega et al., 2016; Paris et al., 2018). Following LPS injection, murine sickness score (MSS) was assessed every 2 h as previously described (Shrum et al., 2014) by an observer blinded to the experimental conditions. Coat appearance, level of consciousness, activity, response to stimulus, eye appearance, and respiration rate and quality were assessed every 2 h until 10 h post-injection. Male and female mice were split evenly between groups. Seven saline and seven LPS injected animals were used for open-field behavior studies, keeping only those with an optimal ultrastructural preservation (five saline and six LPS injected animals) for electron microscopy studies.

# Open-Field Test

Twenty four hours after injection, mice were subjected to openfield testing (Hui et al., 2018). Carefully, one mouse was placed at the center of the apparatus (i.e., 50 × 50 cm white laminated cardboard box) and allowed to move freely for ten minutes. The movement was recorded with the ANY-maze system (version 4.8, Stoelting, Wood Dale, IL, United States). The total distance traveled, lines crossed, distance traveled at the center, entrances into the center, body rotations, freezes and immobile episodes were determined. The apparatus was cleaned between each mouse with 70% ethanol.

### Animal Sacrifice and Tissue Processing

Immediately after open-field testing, mice were anesthetized with a cocktail of 80 mg/kg ketamine and 10 mg/kg xylazine. The animals were then transcardially perfused with ice-cold phosphate-buffered saline (PBS; 50 mM at pH 7.4) followed by 3.5 % acrolein and 4% paraformaldehyde (PFA) both diluted in phosphate buffer (PB; 100 mM at pH 7.4). Brains were harvested and post-fixed 2 h in ice-cold 4% PFA. Following postfixation, brains were washed with PBS to remove excess PFA. Fifty-micrometer thick coronal brain sections were generated in PBS using a vibratome (Leica VT1000s). Brain sections were stored in a solution of cryoprotectant and stored at −20◦C (Bisht et al., 2016a).

# Tissue Preparation Staining for Electron Microscopy

Immunohistochemistry was performed against ionized calcium-binding adaptor protein 1 (IBA1) which provides an excellent visualization of microglial fine processes by immunocytochemical electron microscopy, as described previously (Savage et al., 2018). Briefly, brain sections between −2.0 mm and −2.3 mm Bregma levels were selected and washed in PBS to remove cryoprotectant, then incubated in 0.3% hydrogen peroxide followed by 0.1 % sodium borohydride. Sections were incubated 1 h in blocking buffer (10% fetal bovine serum, 3% bovine serum albumin, 0.03% Triton X-100) then overnight at 4◦C in primary rabbit anti-IBA1 antibody in blocking buffer (Wako). Sections were next incubated for 90 min in goat-anti-rabbit IgGs conjugated to biotin (1/300, diluted in TBS, Jackson ImmunoResearch) followed by ABC reagent (Vector Laboratories) and developed with a solution containing 0.05% 3,3<sup>0</sup> -diaminobenzidine and 0.015% hydrogen peroxide. A 30-min incubation of sections with 1% osmium tetroxide to fix lipids was performed followed by an ethanol dehydration of increasing concentration, washing in propylene oxide and overnight infiltration in Durcupan resin. The next day, the sections were embedded with Durcupan resin between ACLAR embedding films (Electron Microscopy Sciences) for 72 h at 55◦C.

The CA1 of the dorsal hippocampus was excised, affixed to a resin block and cut into 70–75 nanometer-thick sections using an ultramicrotome (Leica Ultracut UC7). The ultrathin sections were collected on copper mesh grids. In each animal, 10 microglial cell bodies and 150–250 microglial processes were randomly selected and imaged at a magnification of 6800× using a transmission electron microscope (FEI Tecnai Spirit G2) equipped with an ORCA-HR digital camera (Hamamatsu; 10 MP). Microglial cell bodies were identified based on their positive staining for IBA1 and their unique ultrastructure. Microglia generally have smaller cell bodies and nuclei than neighboring astrocytes or neurons, characteristic heterochromatin patterns in their nuclei, as well as long and narrow stretches of endoplasmic reticulum (ER) (Savage et al., 2018). Microglial processes were identified based on their positive staining for IBA1 and their lack of nucleus, ER or Golgi apparatus. Intracellular organelles were identified as previously described (El Hajj et al., 2019).

# Ultrastructural Analysis of Microglia

Images of the microglial cell bodies and processes were blinded to the experimental conditions prior to analysis to prevent bias. The area, perimeter, circularity, solidity, number of phagosomes, percentage of cells with phagosomes and presence of ER dilation was determined for each cell body using FIJI. Additionally, the maximum distance from nuclear membrane to cellular membrane was measured, as well as the number of cells displaying proximal processes. A proximal process was defined as a region which narrows to below 0.3 microns and does not contain ER or Golgi apparatus. Excitatory synapses (asymmetric synapses) were defined by the presence of a presynaptic axon terminal containing 40 nanometer vesicles in close apposition to a postsynaptic dendritic spine displaying an asymmetric postsynaptic density thickening (Colonnier, 1968). The perimeter, area, percentage touching synapses, percentage with phagosomes and percentage associated with extended extracellular space pockets was determined for each process (El Hajj et al., 2019). Phagosomes were identified by their ovoid shape with a clear cytoplasm. ER was characterized as dilated if the distance between the two membranes enclosing the lumen was greater than 60 nanometers. Extracellular space pockets were identified by clear space surrounding the microglia, without delineating membranes and lacking acute angles seen in astrocytic processes (Tremblay et al., 2010).

#### Statistics

Data was analyzed using GraphPad Prism 7. LPS versus saline injected mice were compared using a non-parametric Mann–Whitney test. The sample size (n) refers to individual

microglial cell bodies or processes, as previously reported in our ultrastructural analyses (El Hajj et al., 2019). All data is reported as mean ± standard error of the mean (SEM).

#### RESULTS

To determine the ultrastructural changes of microglia and their interactions with synaptic structures during sickness behavior, we injected 4 month old CX3CR1-GFP heterozygous mice i.p. with 1 mg/kg LPS or saline (see **Figure 1A** for experimental paradigm). An observer blinded to animal treatment monitored the injected mice for sickness behavior during 10 h following injection (**Figure 1B**), and was able to correctly identify LPS from control mice, as control mice displayed a score of 0 on the MSS at every timepoint investigated. Mice injected with LPS displayed time-dependent increases in MSS (**Figure 1B**). Twenty-four hours after injection, mice were subjected to openfield behavior testing to verify sickness behavior. Mice displayed sickness behavior including decreased distance traveled both in total and in the center of the open-field, and decreases in the number of line crossings (**Figures 1C–G**), consistent with other rodent sickness behavior models (Shrum et al., 2014; Furube et al., 2018; Mansour et al., 2018).

Following verification of LPS injection causing sickness behavior, we analyzed microglia at nanoscale resolution to determine changes in ultrastructure and interactions with synapses. We focused on the strata radiatum and lacunosummoleculare of the dorsal CA1, which has been implicated in microglial-mediated behavioral deficits in mouse models of sickness behavior (Michels et al., 2015; Zhang et al., 2017). Microglial cell bodies and processes were identified by their immunoreactivity against IBA1, with cell bodies displaying characteristic bean-shaped nuclei, long stretches of ER and numerous mitochondria (**Figures 2A–D**).

Microglial cell bodies were more likely to contain phagosomes in LPS injected versus saline injected animals (84.62 percent versus 43.75 percent, **Table 1**). LPS injected animals also displayed increased numbers of phagosomes per microglial cell body (**Figure 2E**). Most of these phagosomes were lucent and contained fully digested contents. However, many phagosomes contained partially degraded membranes and in one case contained what may appear to be two partially digested postsynaptic densities. Cell bodies were also more likely to display attached proximal processes, defined by a narrowing to less than 300 nanometers in width at some point also devoid of ER and Golgi (**Figure 2E**). Attached proximal processes are very rarely seen in ultrathin sections when investigating healthy brain tissue, but were significantly increased after LPS injection (14.58 percent in saline injected controls versus 44.23 percent in LPS injected animals, **Table 1**). The cytosol of microglial cell bodies from LPS injected animals was also thicker and more expansive compared with saline injected controls, as determined by the longest distance measured between cellular and nuclear membranes (**Figure 2E**).

While microglial cell bodies in LPS injected animals contained a larger area of cytosolic space (10.75 µm<sup>2</sup> versus 6.72 µm<sup>2</sup> , TABLE 1 | Quantification of ultrastructural changes induced by LPS in 4 month old mice.


**Table 1**), they were not significantly different in total area or perimeter. Roughly 60 percent of cell bodies contacted pockets of extracellular space in both conditions (56.25 percent in saline injected versus 59.62 percent in LPS injected mice). Previous studies from our group have defined dark microglia, recognized by their electron dense cytoplasm and nucleoplasm, as cells displaying signs of metabolic stress including dilated ER (Bisht et al., 2016b). These cells are rare in healthy young adult mice, but they increase in number among the hippocampus CA1 strata radiatum and lacunosum-moleculare of maternal immune activation, chronic stress, aging or Alzheimer model mice (Bisht et al., 2016b; Hui et al., 2018). We found no dark microglia in either experimental group, indicating that acute sickness induced by a single LPS injection is an insufficient stressor to induce microglial cytoplasmic/nucleoplasmic condensation, the hallmark identifying ultrastructural feature of dark microglia, at least in this brain region and at the time point examined.

While no dark microglia were identified, occasional stressed or degenerating dendritic spines were observed based on their darkened cytoplasm and altered organelles, often in contact with microglial processes (**Figure 2B**).

In addition to investigating microglial cell bodies, we utilized anti-IBA1 staining to identify at the ultrastructural level microglial processes discontinuous to their cell body in ultrathin section. Because IBA1 is distributed throughout the cytosol, we were able to investigate processes and their interactions with the surrounding neuropil. Although microglial processes did not change size following LPS injection, they were rounder and increased in solidity compared with the saline injected controls (**Figures 2B,D**). This is in line with light-level analyses of microglia conducted in other sickness behavior models (Hoogland et al., 2015). Microglial processes in LPS injected animals were also significantly more likely to interact with excitatory synapses. In particular, processes from LPS injected animals were much more likely to directly touch synaptic clefts (15.53 percent versus 9.6 percent, **Table 1**) and interact with both presynaptic axon terminals and postsynaptic dendritic spines (**Figure 2F**) than those of saline injected controls.

# DISCUSSION

Microglia have been implicated in the initiation of neuronal death and damage across a myriad of neurodegenerative conditions including models of sickness behavior and sepsis survivors (Block et al., 2007; Sankowski et al., 2015; Zhao et al., 2019). We are reporting the first quantitative ultrastructural characterization of microglia within the hippocampus CA1 (strata radiatum and lacunosum-moleculare) of an LPS induced mouse sickness model. Our results uncovered increases in phagosomes 24 h after peripheral exposure. As we performed all of our TEM analysis on individual ultrathin sections and not serial sections, we cannot rule out the possibility that these phagosomes could

∗∗∗p < 0.001. bv – blood vessel, n – nucleus, lg – lipid granule, dn – dark neuron, A - astrocyte.

be incidences of extracellular digestion or "exophagy" (Haka et al., 2016) viewed in 70–75 nanometer-thick profile. However, the incidence of microglia-associated extracellular space pockets was not significantly different between LPS and saline injected controls. In addition to increases in putative phagosomes, we also detected increased microglial interactions with synaptic structures when observed 24 h after peripheral LPS injection.

Dark microglia have been described in the hippocampus CA1 region of various mouse models in conjunction with disruptions in their inflammatory signaling, including animals

subjected to chronic unpredictable stress, repeated social defeat, maternal immune activation, aging, Alzheimer pathology and CX3CR1 deficiency (Bisht et al., 2016b; Hui et al., 2018; El Hajj et al., 2019). While a single dose of LPS was sufficient to induce proinflammatory signaling and significant changes in microglial cell body and process ultrastructure, we found no evidence of dark microglia in the CA1 (strata radiatum and lacunosum-moleculare) of LPS injected animals. It appears that the acute reaction to a single dose of LPS is insufficient to induce the dark microglia phenotype in 4 month old animals. This could imply that such a shift in microglial phenotype requires chronic increases in inflammatory signaling. Further studies are required to investigate the long-term consequences of LPS induced sickness and recovery on microglial ultrastructure.

While we did not observe dark microglia in the CA1 of LPS injected animals, we identified dark neuronal spines and cell bodies, for the first time in a sickness behavior model. These dark neurons, which could represent a subset of susceptible neurons, were previously described in aged mice and in mice experiencing stressful challenges, including sensory loss (Peters et al., 1991; Tremblay et al., 2012), as well as neurodegenerative disease pathology (Turmaine et al., 2000). Our data is in line with previous studies showing that mice subjected to i.p. LPS display reductions in neuronal projections (microtubule associated protein 2 staining) accompanied by neuronal cell body loss in the hippocampus (Zhao et al., 2019). Additionally, mouse models of sepsis have lower levels of glutamatergic NMDA receptors, as well as reduced numbers of doublecortin-positive newborn neurons and parvalbumin interneurons in the hippocampus (Valero et al., 2014; Ji et al., 2015; Zhang et al., 2017). These alterations were shown to be triggered by inflammatory pathways, most likely through increased microglial secretion of proinflammatory cytokines and reactive oxygen species.

Microglia in LPS injected animals contained increased amounts of phagocytic material. It is well established in the literature that peripheral administration of LPS can cause a shift in microglial morphology toward a more amoeboid shape (Buttini et al., 1996; Hoogland et al., 2015). It is possible that exposing microglia to proinflammatory cytokines shifts their activity to a more pro-phagocytic state. In our study, microglial cell bodies in animals injected with LPS contained nearly three times the phagocytic cargo observed in those injected with saline. These data imply that proinflammatory cytokines, at least in the short term of 24 h, significantly increase microglial phagocytic activity. However, further studies focused on the expression of proinflammatory cytokines, chemokines, and phagocytic cell surface receptors by these cells is required to directly link LPS administration with increases in microglial phagocytosis.

While several models of microglial morphology have implied that more amoeboid cells are associated with cell body migration and phagocytosis at the expense of surveillance (Walker et al., 2014), microglia in LPS injected animals were significantly more likely to interact with excitatory synapses (both elements and clefts). This is not immediately intuitive as microglia were previously defined with phenotypes somewhere on a spectrum between surveillant (interacting with synapses) and reactive (increasing cytokine/chemokine response to invading pathogens followed by phagocytic clearance) (Kreutzberg, 1996; Butovsky and Weiner, 2018). However, microglia involved in the phagocytosis of newborn neurons generated through adult neurogenesis or synapses during normal brain development were previously shown to display a ramified morphology with 'ball and chain' structures of phagocytic pouches on ramified processes (Sierra et al., 2010; Tremblay et al., 2010; Schafer et al., 2012). In the context of chronic stress, microglia which are involved in neuronal circuit rewiring (Milior et al., 2016; Wohleb et al., 2018) were either shown to display reduced or hyper-ramified processes depending on the model and time course (Hinwood et al., 2013; Hellwig et al., 2016; Milior et al., 2016). The synaptic contacts we observed did not appear to be sites of active proteolytic degradation, as there was no change in the amount of microglia-associated extracellular debris, marked by partially degraded membranes located in the large pockets of extracellular space between microglial membranes and surrounding neuropil, between either group of microglia. However, several microglial cell bodies displayed putative phagosomes containing seeming intact neuropil, while partially digested membranes and possible postsynaptic densities were also found in LPS injected mice.

Microglia have previously been implicated in synaptic pruning in a cellular mechanism known as "trogocytosis" (Weinhard et al., 2018). Weinhard and colleagues demonstrated that microglial processes in postnatal day 15 animals removed small parts of axon terminals averaging between 0.1 and 0.5 µm<sup>3</sup> , much smaller than traditionally phagocytosed elements, without extracellular space between the neuronal and microglial membrane. It is possible that these excess synaptic contacts in LPS injected animals were undergoing trogocytosis, although further studies using 3-dimensional electron microscopy would be necessary to verify complete engulfment. It is also possible that these processes are engaged in synaptic stripping, by which a microglia remove synapses by interjecting a process into the synaptic cleft, physically separating the presynaptic terminal from the dendritic spine or shaft (Blinzinger and Kreutzberg, 1968). Synaptic stripping has been described in cortical samples from mice sacrificed 24 h after peripheral LPS injection (Chen et al., 2014).

Our study focused on the microglial response 24 h after a single injection of LPS, which has allowed us an ultrastructural snapshot of what occurs inside the brain during recovery from acute sickness. Numerous studies have studied acute (2–6 h post LPS injection) sickness behavior in rodents but have found increases in depressive-like behavior 24 h after injection (Dantzer et al., 2008). Human studies have also uncovered both acute and long-term implications of peripheral inflammation on the emergence of depression and other psychiatric disorders (Savitz and Harrison, 2018). Other studies in rodents have noted various long-term changes to microglial morphology after single or multiple exposures to LPS, being present in some cases as much as a year after injection (Hoogland et al., 2015). As this is the first ultrastructural characterization of microglia in the hippocampus in response to systemic LPS administration we focused on a single timepoint during initial sickness behavior. While we investigated both male and female mice there were no overt differences between the microglial ultrastructure between the two sexes in response to LPS, although the response to

LPS administration was rather robust and our sample size was insufficient to detect possibly small changes due to sex differences. In addition to sex differences, there are numerous questions to be addressed in follow-up studies, including exploration of acute versus chronic ultrastructural changes of microglia in response to a single dose or multiple doses of LPS, as well as investigation into other brain regions and along the aging trajectory. In complement, it would also be important to use additional techniques than electron microscopy, beyond the scope of this short communication study, to provide cellular and molecular insights into the microglial activities, such as neuronal circuit rewiring, mediated during sickness behavior.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

#### ETHICS STATEMENT

The animal study was reviewed and approved by the Université Laval.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

M-ET obtained funding for the study. JS designed and performed the experiments with CH. JS and M-KS-P prepared and analyzed the electron microscopy pictures and behavioral data. JS wrote the first draft of the manuscript. M-ET edited the subsequent versions to which all authors contributed. All authors read and approved the final version of the manuscript.

#### FUNDING

This study was supported by the ERA-NET TracInflam grant. JS and CH were supported by the Fonds de Recherche du Québec– Santé (FRQS) fellowships, M-KS-P holds a Canadian Institute of Health Research scholarship, while M-ET holds a Tier II Canada Research Chair of Neuroimmune plasticity in health and therapy.

#### ACKNOWLEDGMENTS

We would like to thank Nathalie Vernoux for technical assistance and experimental guidance, as well as Julie-Christine Lévesque at the Bio-Imaging platform of the Infectious Disease Research Centre, funded by an equipment and infrastructure grant from the Canadian Foundation Innovation (CFI).

in the adult brain. Nat. Commun. 5:4486. doi: 10.1038/ncomms 5486


mice. Brain. Behav. Immun. 55, 126–137. doi: 10.1016/j.bbi.2015. 11.008


drivers of neurodegeneration. Front. Cell. Neurosci. 9:28. doi: 10.3389/fncel. 2015.00028


**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 Savage, St-Pierre, Hui and Tremblay. 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.

# Boosting Antioxidant Self-defenses by Grafting Astrocytes Rejuvenates the Aged Microenvironment and Mitigates Nigrostriatal Toxicity in Parkinsonian Brain via an Nrf2-Driven Wnt/β-Catenin Prosurvival Axis

Maria Francesca Serapide<sup>1</sup> , Francesca L'Episcopo<sup>2</sup> , Cataldo Tirolo<sup>2</sup> , Nunzio Testa<sup>2</sup> , Salvatore Caniglia<sup>2</sup> , Carmela Giachino<sup>2</sup> and Bianca Marchetti 1,2 \*

<sup>1</sup>Pharmacology Section, Department of Biomedical and Biotechnological Sciences, Medical School, University of Catania, Catania, Italy, <sup>2</sup>Section of Neuropharmacology, OASI Research Institute-IRCCS, Troina, Italy

#### Edited by:

Alberto Javier Ramos, National Council for Scientific and Technical Research (CONICET), Argentina

#### Reviewed by:

Zhigang Liu, Northwest A&F University, China Juan Ferrario, National Council for Scientific and Technical Research (CONICET), Argentina

> \*Correspondence: Bianca Marchetti biancamarchetti@libero.it

Received: 04 October 2019 Accepted: 22 January 2020 Published: 12 March 2020

#### Citation:

Serapide MF, L'Episcopo F, Tirolo C, Testa N, Caniglia S, Giachino C and Marchetti B (2020) Boosting Antioxidant Self-defenses by Grafting Astrocytes Rejuvenates the Aged Microenvironment and Mitigates Nigrostriatal Toxicity in Parkinsonian Brain via an Nrf2-Driven Wnt/β-Catenin Prosurvival Axis. Front. Aging Neurosci. 12:24. doi: 10.3389/fnagi.2020.00024 Astrocyte (As) bidirectional dialog with neurons plays a fundamental role in major homeostatic brain functions, particularly providing metabolic support and antioxidant self-defense against reactive oxygen (ROS) and nitrogen species (RNS) via the activation of NF-E2-related factor 2 (Nrf2), a master regulator of oxidative stress. Disruption of As–neuron crosstalk is chiefly involved in neuronal degeneration observed in Parkinson's disease (PD), the most common movement disorder characterized by the selective degeneration of dopaminergic (DAergic) cell bodies of the substantia nigra (SN) pars compacta (SNpc). Ventral midbrain (VM)-As are recognized to exert an important role in DAergic neuroprotection via the expression of a variety of factors, including wingless-related MMTV integration site 1 (Wnt1), a principal player in DAergic neurogenesis. However, whether As, by themselves, might fulfill the role of chief players in DAergic neurorestoration of aged PD mice is presently unresolved. Here, we used primary postnatal mouse VM-As as a graft source for unilateral transplantation above the SN of aged 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) mice after the onset of motor symptoms. Spatio-temporal analyses documented that the engrafted cells promoted: (i) a time-dependent nigrostriatal rescue along with increased high-affinity synaptosomal DA uptake and counteraction of motor deficit, as compared to mock-grafted counterparts; and (ii) a restoration of the impaired microenvironment via upregulation of As antioxidant self-defense through the activation of Nrf2/Wnt/β-catenin signaling, suggesting that grafting As has the potential to switch the SN neurorescueunfriendly environment to a beneficial antioxidant/anti-inflammatory prosurvival milieu. These findings highlight As-derived factors/mechanisms as the crucial key for successful therapeutic outcomes in PD.

Keywords: Parkinson's disease, aging, astrocyte–neuron crosstalk, neuroinflammation, dopaminergic neurons, neuroprotection

# INTRODUCTION

Astrocyte (As) bidirectional dialog with neurons plays a fundamental role in major homeostatic brain functions. Besides their physical and metabolic support to neurons, As regulate central nervous system (CNS) synaptogenesis, promote neuronal development and plasticity, guide axon pathfinding, modulate the blood–brain barrier, and contribute to neuroprotection via the production of different growth and neurotrophic factors, antioxidant and anti-inflammatory molecules, through a concerted crosstalk with neurons (Marchetti and Abbracchio, 2005; Bélanger and Magistretti, 2009; Sofroniew and Vinters, 2010; Molofsky et al., 2012; Sun and Jakobs, 2012). Notably, As display region-specific properties, as the nature of As-derived factors can vary as a function of the CNS region, the age and sex of the host, and the type of brain lesion/injury (Gallo et al., 1995; Marchetti, 1997; Barkho et al., 2006; Jiao and Chen, 2008; Oberheim et al., 2012). Specifically, As of the ventral midbrain (VM-As) represent a primary source of survival, neurotrophic and neuroprotective molecules for dopaminergic (DAergic) neurons (Engele and Bohn, 1991; Takeshima et al., 1994; Morale et al., 2006; Sandhu et al., 2009; L'Episcopo et al., 2010a), the neuronal cell population that progressively degenerates in Parkinson's disease (PD).

Here, the selective death of DAergic neurons of the substantia nigra pars compacta (SNpc) and their terminals in the striatum (Str) are responsible for the gradual impairment of motor function leading to the classical motor features of PD (i.e., bradykinesia, rest tremor, rigidity, and postural instability; Schapira et al., 2014; Jankovic, 2019). While the causes and mechanisms are not completely understood, current evidence indicates that a complex interplay between several genes and many environmental factors affecting the regulation of crucial pathways involved in inflammatory glial activation, mitochondrial function, protein misfolding/aggregation, and autophagy contribute to DAergic neuron demise in PD (Marchetti and Abbracchio, 2005; Frank-Cannon et al., 2008; Gao et al., 2011; Lastres-Becker et al., 2012; Cannon and Greenamyre, 2013; Hirsch et al., 2013; Dzamko et al., 2015, 2017; Langston, 2017; Blauwendraat et al., 2019).

Especially, in this context, dopamine (DA) oxidative metabolism represents a vulnerability factor linking both mitochondrial and lysosomal dysfunctions to PD pathogenesis (Hirsch and Hunot, 2009; Johri and Beal, 2012), whereby As play a critical antioxidant self-protective role. Hence, oxidative stress upregulates the expression of Nuclear factor erythroid 2 like 2 (NFE2L2/Nrf2), which translocates to the nucleus and binds to antioxidant responsive elements (AREs; Tebay et al., 2015; Zhang et al., 2017). Besides other activated genes, the antioxidant, anti-inflammatory, and cytoprotective Heme oxygenase 1 (HO1) and superoxide dismutase 1 (SOD1; Chen et al., 2009; Sandhu et al., 2009; Surh et al., 2009; Zhang et al., 2017) likely play an important role in DAergic neuroprotection against oxidative damage (Burbulla et al., 2017; Giguère et al., 2018; Surmeier, 2018; Nguyen et al., 2019).

Notably, aging represents a chief risk factor for PD development, as with advancing age, nigrostriatal DAergic neurons progressively deteriorate (Rodriguez et al., 2015; Wyss-Coray, 2016; Poewe et al., 2017). With age, the ''adaptive'' or ''compensatory'' capacity of midbrain DAergic neurons gradually fails, which may contribute to the slow nigrostriatal degeneration of PD (Hornykiewicz, 1993; Bezard and Gross, 1998; Blesa et al., 2017). In fact, aging is associated to a gradual decline in the ability of DAergic neurons to recover upon an insult (Bezard and Gross, 1998; Ho and Blum, 1998; Boger et al., 2010). Aging exacerbates inflammation and oxidative stress, which are crucial hallmarks of PD and 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)–induced PD (Di Monte and Langston, 1995; Langston, 2017).

In fact, microglial cells show age-dependent and regionspecific changes in morphology such as structural deterioration or dystrophy, decreased expression of growth/neurotrophic factors, and an impaired phagocytic activity in the face of increased marker expression and upregulation of pro-inflammatory molecules (see Niraula et al., 2017, and references herein), all of which are associated to a gradual loss of As and microglia neuroprotective capacity. Reportedly, microglia switch to a so-called ''primed'' status, endowed with a strong neurotoxic, pro-inflammatory M1 phenotype with harmful consequences for As–neuron interactions (Miller and Streit, 2007; Liddelow et al., 2017; Rosciszewski et al., 2018, 2019) and DAergic neuron survival upon injury (L'Episcopo et al., 2011a,b,c, 2018a,b).

Notably, genetics, environmental toxicity, and particularly ''inflammaging'' differentially affect the nature, quality, and outcome of As–neuron crosstalk, directing to either neurodegeneration/repair (L'Episcopo et al., 2018a,b). Hence, the neuroprotective functions of VM-As are impaired, including their ability to mount a self-protective neurorepair strategy thanks to the expression of several pro-neurogenic, neurotrophic, and antioxidant molecules (Marchetti et al., 2013).

Besides others, we uncovered the Wnt/β-catenin signaling pathway, a chief player in neurodevelopmental processes (Salinas, 2012; Arenas, 2014; Joksimovic and Awatramani, 2014; Wurst and Prakash, 2014; Brodski et al., 2019; Marchetti et al., 2020), as a crucial signaling system involved in the physiopathology of nigrostriatal DAergic neurons (see Marchetti, 2018, for extensive review). The hallmark of the Wnt/β-catenin pathway after binding the Wnt's receptors, Frizleds (Fzds), is the cytoplasmic accumulation of β-catenin and its nuclear translocation, finally activating the transcription of Wnt target genes involved in DAergic neurogenesis and neuroprotection (Marchetti, 2018). Notably, VM-As express region-specific transcription factors including Wnt glycoproteins (Marchetti et al., 2013). Especially, wingless-related MMTV integration site 1 (Wnt1) is critically involved in DAergic neuroprotection against several neurotoxic and inflammatory insults (Marchetti and Pluchino, 2013).

Of specific mention, with age, Wnt signaling becomes dysfunctional, with potential consequences for neuron–glia crosstalk, DAergic neuron plasticity, and repair (Marchetti, 2018). Notably, deregulation of Wnt signaling has been reported in major neurodegenerative diseases including PD (Berwick and Harvey, 2012; Galli et al., 2014; Harvey and Marchetti, 2014; Marchetti, 2018; Tapia-Rojas and Inestrosa, 2018; Palomer et al., 2019). Additionally, an increasing number of reports corroborate a Wnt connection for neuron survival and regeneration (Singh et al., 2016; Wang et al., 2017; Marchetti et al., 2020).

To date, there are no effective treatments that can stop or reverse the neurodegeneration process in PD, and current treatments rely on DAergic drugs, including levodopa (L-DOPA) and DAergic agonists, which only temporarily alleviate motor symptoms (Olanow and Schapira, 2013; Obeso et al., 2017; Olanow, 2019). Thus, different lines of research are being pursued to develop novel therapeutic regimens for PD, including cell therapies, aimed at protecting or enhancing the intrinsic regenerative potential of DAergic neurons.

Hence, an increasing number of studies focus on the ability of neural stem progenitor cells (NSCs) harvested from the adult brain to engraft in the injured brain of experimental neurodegenerative diseases and exert positive effects (Yasuhara et al., 2006; Redmond et al., 2007) promoting local trophic support and immune modulation, thus synergizing with the restorative responses of the endogenous NSC population (Madhavan et al., 2009; Zuo et al., 2015; Bacigaluppi et al., 2016). Recently, we performed a carefully constructed time-course analysis of the degenerative changes occurring at the nigrostriatal level of aged male mice upon exposure to MPTP and studied the effects of the unilateral transplantation of syngeneic somatic NSCs within the SNpc. In this aged mouse model, the compensatory DAergic mechanisms are lost, and MPTP induces a long-lasting nigrostriatal degeneration with no repair (Collier et al., 2007).

Interestingly, we found that grafted adult NSCs mostly differentiated into reactive As activating intrinsic cues instructing endogenous As to incite DAergic neuroprotection/neurorestoration, thus efficiently counteracting neurotoxin-induced long-lasting DA degeneration (L'Episcopo et al., 2018b).

Based on this background, VM-As appear uniquely positioned to drive neuroprotective and regenerative programs in PD, and recent preclinical studies used co-transplantation of As to promote the regenerative effects of co-transplanted stem cells in rodent PD models (Yang et al., 2014; Song et al., 2018). However, how the aged brain responds to pharmacological or cellular therapeutical interventions is not well documented. In particular, whether transplantation of VM-As by themselves might fulfill the role of chief players in DAergic neurorestoration of aged PD mice is presently unresolved.

Here, we combined different recognized environmental risk factors for human PD, i.e., aging, male gender, inflammation, and exposure to MPTP (Langston, 2017), to explore the capacity of As grafting to mitigate the harmful SN microenvironment and DAergic toxicity. To this end, we used primary mouse postnatal [postnatal days 2–3 (P2–3)] VM-As as a graft source for unilateral transplantation above the SN of middle-aged mice treated with MPTP, after the onset of motor symptoms. We herein report that the engrafted VM-As survived, expressed As markers, and promoted a remarkable reduction of DAergic neuron loss within the MPTP-lesioned SN associated to striatal DAergic reinnervation and functionality. Gene expression analyses coupled to immunofluorescence and functional data in vivo, ex vivo, and in vitro, suggest the ability of AS grafts to increase Nrf2-antioxidant self-defense, to mitigate MPTP-induced oxidative stress and inflammation, thus switching the harmful As–neuron crosstalk, via activation of an Nrf2/Wnt/β-catenin prosurvival axis.

# MATERIALS AND METHODS

#### Mice and Treatments

Middle-aged (9- to 11-month-old) male C57BL/J (Charles River, Calco, Italy) mice were maintained under standard laboratory conditions. All surgeries were performed under anesthesia. The mice received n = 4 intraperitoneal (i.p.) injections of vehicle (saline) or MPTP-HCl (Sigma-Adrich, St. Louis, MO, USA) dissolved in saline, 3 h apart during 1 day, at a dose of 12 mg/kg-1 free base, according to titration studies that produced long-lasting depletion of DA end points with no recovery in both Str and SNpc of aged mice without causing toxicity (L'Episcopo et al., 2013). MPTP was handled in accordance with the reported guidelines (Jackson-Lewis and Przedborski, 2007). Based on our time-course analysis of the degenerative changes occurring in aging male mice upon exposure to MPTP (L'Episcopo et al., 2013, 2018b), a window of 7 days post-MPTP was selected for transplantation of VM-As (**Supplementary Figure S1**).

# Experimental Design

Middle-aged male mice exhibiting a long-lasting nigrostriatal DAergic toxicity with no recovery upon MPTP treatment (L'Episcopo et al., 2018b) were used to study the neuroprotective/neurorescue effects of unilateral transplantation of As derived from the ventral midbrain (tVM-As), above the SN (**Supplementary Figure S1**). We designed in vivo, ex vivo, and in vitro experiments. The studies included the SN and the Str and were performed both in basal condition and at different time points (tps) after MPTP ± VM-As/mock transplants, covering 1–5 weeks post-MPTP. The quantification of the different parameters studied included immunohistochemical, neurochemical, gene expression, and behavioral analyses (see **Supplementary Figure S1**). Based on our gene profiling analysis of the MPTP response in young and aged mice (L'Episcopo et al., 2011b, 2013, 2018b), this study focused on inflammation and oxidative stress, and Wnt signaling genes that are specifically altered in aging mice (see Marchetti, 2018). The microglial response was studied using quantitative immunohistochemistry, and at both gene and protein levels. For the As response, both ex vivo and in vitro experiments were carried out to address molecular and functional changes according to: (i) the different experimental groups; (ii) the different pharmacological challenges; and (iii) the different As–neuron coculture paradigms established with primary mesencephalic neurons, according to our previous studies (L'Episcopo et al., 2011b, 2014a, 2018a,b).

# Ventral Midbrain As Cultures

Primary astroglial cell cultures were obtained from mouse VM at P2–3 according to Booher and Sensenbrenner (1972), with slight modifications, as described in full detail (Gallo et al., 2000a; Gennuso et al., 2004). The cultures were allowed to grow and differentiate until they reached confluency, at which time (15–20 days in vitro, DIV) the loosely adherent microglial cells were separated by shaking for 2 h at 37◦C and 190 rpm. The attached cells were then washed with sterile phosphate buffered saline (PBS) and incubated for 1–2 h at 37◦C, at 5% CO<sup>2</sup> , before overnight shaking at 37◦C and 210 rpm. The supernatant media containing oligodendrocyte precursors and other cell types were discarded. The glial [more than 95% of the cells were glial fibrillary acidic protein (GFAP)–immunoreactive (IR) As] monolayers were then rinsed with sterile PBS and pulsed with the nucleotide analog bromodeoxyuridine (BrdU, 5 µM) 24 h before transplantation. Part of the cultures were replated at a final density of 0.4–0.6 × 10<sup>5</sup> cells/cm<sup>2</sup> in poly-D-lysine (10 µg/ml)–coated 6-, 12- or 24-well plates, or in insert membranes (0.4 µm polyethylene terephthalate) for direct or indirect coculture (BD Biosciences) with primary mesencephalic neurons, and processed as described for RT-PCR or immunocytochemistry (**Figures 1**, **2**).

#### Transplantation of VM-As

Upon verification of the purity, proliferation, and neuroprotective potential, VM-As, tagged in vitro with BrdU, were then transplanted at different concentrations in pilot experiments conducted for optimization of cell number (50–200 × 10<sup>3</sup> VM-As; n = 6 mice/experimental group) and timing of transplantation after MPTP (1, 7, or 21 days post-MPTP, n = 6 mice/tp). The dose of 150 × 10<sup>3</sup> VM-As and 7-day post-MPTP interval were selected, as a higher number of VM-As and TH<sup>+</sup> neurons were recovered by 1 week post-implant (wpt). On the day of transplantation (i.e., day 7 post-MPTP), MPTP-injected mice exhibiting a significant motor deficit were randomly (**Supplementary Figure S1**) assigned to MPTP + PBS, MPTP + VM-As grafts, or MPTP + mock [VM dead cells (VMCs), Madhavan et al., 2009]. Mice were anesthetized with chloral hydrate (600 mg/kg) and positioned in a stereotaxic apparatus. The following stereotaxic coordinates were used: 3.2 posterior to the Bregma, 1.5 mm lateral to the midline, and 3.6 mm ventral to the surface of the dura mater. VM-As or mock (150 × 10<sup>3</sup> ) was injected unilaterally above the left SN (over a period of 2 min). The needle was kept in place for 5 min after each infusion before retraction. Saline-injected controls received the same volume of PBS.

#### Motor Behavior Analysis With the Rotarod

An accelerating rotarod (five-lane accelerating rotarod; Ugo Basile, Comerio, Italy) was used to measure motor coordination in mice. Mice had to keep their balance on a horizontal rotating rod (diameter, 3 cm) and rotation speed was increased every 30 s by 4 rpm. Five mice were tested at the same time, separated by large disks. A trial started when the mouse was placed on the rotating rod, and it stopped when the mouse fell down or when 5 min were completed. Falling down activated a switch that automatically stopped a timer. On the testing day, each mouse was submitted to five trials with an intertrial interval of 30 min. Mice housed five per cage were acclimated to a 12 h shift in light/dark cycle so that the exercise occurred during the animals' normal wake period. Saline- and MPTP-treated mice (10/experimental group) were assessed for their rotarod performance on days −7, 1, 7, 14, 21, and 28 after MPTP injection.

#### Immunohistochemistry

On the day of sacrifice, mice were anesthetized and transcardially perfused with 0.9% saline, followed by 4% paraformaldehyde in phosphate buffer (pH 7.2 at 4◦C); the brains were carefully removed and processed as described in full detail (Morale et al., 2004). Tissues were frozen and stored at −80◦ C until further analyses. Serial coronal sections (14 µm thick), encompassing the Str (Bregma 1.54 to Bregma −0.46) and the SNpc (Bregma −2.92 to Bregma −3.8 mm) according to Franklin and Paxinos (2007), were collected, mounted on poly-L-lysine–coated slides, and processed as previously described in full detail (L'Episcopo et al., 2011b). The following pre-absorbed primary antibodies were used: rabbit anti-tyrosine hydroxylase (TH, Chemicon International, Temecula, CA, USA), the rate limiting enzyme in DA synthesis; rabbit anti-TH (Peel Freez Biochemicals, Rogers, AR, USA); mouse anti-TH (Boehringer Mannheim Bioc., Philadelphia, PA, USA), rat anti-dopamine transporter (DAT, Chemicon, Int. USA); mouse anti-neuron specific nuclear protein (NeuN, US Biologicals, Swampscott, MA, USA); rabbit anti-GFAP (GFAP, Dako, Cytomation, Denmark), mouse anti-GFAP (Sigma-Adrich, St. Louis, MO, USA) as an As-specific cell marker; goat anti-ionized calcium-binding adapter molecule 1 (IBA1, Novus Biologicals, Littleton, CO, USA) a microgliaspecific marker; goat anti-heme oxygenase 1 (anti-Hmox, 1:150, Santa Cruz Biotechnology, Santa Cruz, CA, USA); rabbit polyclonal, anti-inducible nitric oxide synthase (iNOS;1:200; Santa Cruz Biotechnology, Santa Cruz, CA, USA); and rabbit anti-3-nitrotyrosine (3-NT; 1:200; Millipore, Kankakee, IL, USA; see also complete list of Abs in **Supplementary Table S1**). Nuclei were counterstained with 4<sup>0</sup> ,6-diamidino-2-phenylindole (DAPI) in mounting medium (Vector Laboratories, Burlingame, CA, USA). Visualization of incorporated BrdU requires DNA denaturation performed by incubating the sections in HCl for 30 min at 65◦C. After overnight incubation, sections were washed extensively and incubated with fluorochrome (FITC, CY3, CY5)–conjugated species-specific secondary antibodies for immunofluorescent detection. TH immunoreactivity was also detected using biotinylated secondary antibodies (Vector Laboratories, Burlingame, CA, USA) and diaminobenzidine (DAB, Vector Laboratories, Burlingame, CA, USA) as the developing agent as described (L'Episcopo et al., 2011b). Cresyl violet (CV) was used to visualize the Nissl substance.

In all of these protocols, blanks were processed as for experimental samples except that the primary antibodies were replaced with PBS.

#### DAergic End Points

Quantitative analysis of DAergic neurons in the SNpc was carried out by serial section analysis of the total number of TH-positive (TH<sup>+</sup> ) and NeuN-positive (NeuN<sup>+</sup> ) neurons throughout the entire rostro-caudal axis of the SNpc (Franklin

vs. DAn alone) and DA uptake levels measured by [3H]DA incorporation (E; n = 3 independent experiments). Comparable results were obtained in the indirect coculture paradigms (DAn exposure to ACM or to VM-As inserts, As-), and values were pooled together. \*p ≤ 0.01 vs. -MPP<sup>+</sup> within the same experimental group; ◦p ≤ 0.01 vs. VM-As groups (direct and indirect coculture) by ANOVA followed by Newman–Keuls test.

and Paxinos, 2007) as previously described (L'Episcopo et al., 2011b). Total numbers of TH- and CV-stained neurons in adjacent tissue sections were estimated in parallel to validate TH<sup>+</sup> neuron survival, using Abercrombie correction (Abercrombie, 1946).

Striatal TH- and DAT-immunoreactive (IR) fiber staining was assessed in n = 3 coronal sections at three levels (Bregma coordinates: +0.5, +0.86, and 1.1 mm, respectively) of the caudate putamen (CPu), in n = 6 mice/group/time (Burke et al., 1990). Fluorescence intensity (FI) of TH- and DAT-staining above a fixed threshold used the corpus callosum for background subtraction.

For synaptosomal, high-affinity DA uptake, at the indicated time intervals, mice were sacrificed by cervical dislocation, and the brains rapidly removed and immediately placed on ice-cold saline. The right and left striata were then dissected on an ice-cold plastic dish and processed as described in full detail (L'Episcopo et al., 2011a,b). Synaptosomal, high-affinity DA uptake was assessed in the presence of 10 µM mazindol, according to Morale et al. (2004).

#### Glial Cell Counts

Cell counts were obtained for GFAP<sup>+</sup> As and ameboid IBA1<sup>+</sup> microglial cells (Kreutzberg, 1996). For SN cell counts, at least three sections were obtained from each animal representing each of the five representative planes from −2.92 mm to −3.8 mm relative to Bregma according to the stereotaxic coordinates of Franklin and Paxinos (2007). GFAP<sup>+</sup> As and IBA1<sup>+</sup> microglial cell number per unit of surface area was determined in 8–10 randomly selected fields per section on both sides, the counts averaged for each animal, and the mean number of cells per mm<sup>2</sup> was estimated. Classification of microglia activation was carried out according to Kreutzberg (1996); as described in **Supplementary Table S3**, the glial counts were confirmed by two different observers.

#### Confocal Laser Scanning Microscopy, Image Analysis, and Quantification of Immunostaining

All the quantifications were performed by investigators blind to treatment conditions. Immunostaining was examined using a Leica LCS-SPE confocal microscope. For FI assessments and colocalizations, midbrain sections were labeled by immunofluorescence, and images were acquired by sequential scanning of 12–16 serial optical sections (Gennuso et al., 2004; L'Episcopo et al., 2011a,b,c, 2013). Three dimensional reconstructions from z-series were used to verify colocalization in the x–y, y–z, and x–z planes. Serial fluorescent images were captured in randomly selected areas, the number of labeled cells per field (n = 6–8 fields/section) was manually counted in 4–6 midbrain sections per brain (n = 6/treatment group) using Olympus cellSense Dimension software, and cell counts obtained were averaged (mean ± SEM); the percentages of HO1<sup>+</sup> /GFAP<sup>+</sup> cells out of the total GFAP<sup>+</sup> cells were estimated in each condition, in ≥100 cells, read from at least four VM sections per brain, in six mice per experimental group, and results expressed as mean ±SEM (Gennuso et al., 2004).

#### RNA Extraction, Reverse Transcription, and Real-Time PCR

RNA was extracted from tissues/cell samples, as previously detailed (L'Episcopo et al., 2011a,b). Briefly, after purification using a QIAquick PCR Purification kit (Qiagen), 250 ng of cDNA was used for real-time PCR using pre-developed TaqMan assay reagents (Applied Biosystems). Real-time quantitative PCR was performed using the Step One Detection System (Applied Biosystems) according to the manufacturer's protocol, using the TaqMan Universal PCR master mix (#4304437). The assay IDs are reported in **Supplementary Table S2**. For each sample, we designed a duplicate assay. β-actin (Applied Biosystems #4352341E) was selected as the housekeeping gene, according to our previous (L'Episcopo et al., 2011b, 2018b) and present pilot studies indicating that it does not modify its expression between conditions. Quantification of the abundance of target gene expression was determined relative to β-actin with regard to the control group by using the delta C<sup>t</sup> (2−∆∆Ct) comparative method, with the results expressed as arbitrary units (AU). Relative fold changes over saline/PBS or MPTP/PBS are indicated.

#### Enriched Neuronal Cultures and Primary Midbrain Astroglial–Neuron Cultures

For in vitro establishment of primary mesencephalic neuronal cultures, timed pregnant Sprague–Dawley rats (Charles River Breeding Laboratories, Milan, Italy) were killed in accordance with the Society for Neuroscience guidelines and Italian law. Primary mesencephalic neurons were prepared from the brain on embryonic day 13–14, as detailed (L'Episcopo et al., 2011b). Briefly, mesencephalic tissues were isolated and dissociated with gentle mechanical trituration. Cells were diluted to 1.5 × 10<sup>6</sup> /ml in maintenance medium (MEM supplemented with 10% heat-inactivated FBS, 10% heat-inactivated horse serum, 1 g/L glucose, 2 mM glutamine, 1 mM sodium pyruvate, 100 µM nonessential amino acids, 50 U/ml penicillin, and 50 µg/ml streptomycin) and seeded into 24-well culture plates precoated with poly-D-lysine (20 µg/ml). Plates were maintained at 37◦C in a humidified atmosphere of 5% CO<sup>2</sup> and 95% air. To obtain neuron-enriched cultures, cytosine β-D-arabinofuranoside (Ara-c) was added to a final concentration of 6 µM 36 h after seeding the cells, to suppress glia proliferation (L'Episcopo et al., 2011b). Cultures were changed back to maintenance medium 2 days later and were used for treatment 7 DIV after initial seeding. Neuronal enrichment was verified by immunocytochemistry using GFAP-, TH-, and NeuN-Abs as described. Ara-c treatment reduced glial expression by 95%.

Both purified neuronal cultures and astroglial–neuron cultures at 7 DIV received MPP<sup>+</sup> (10 µM). The specificity of the As neuroprotective effect was further verified in purified neuronal cultures exposed to astroglial conditioned media (ACM) or to As inserts (As-i, indirect As–neuron coculture). In this experimental paradigm, the inserts containing the As monolayer were added on the top of the purified neurons. These inserts allowed diffusion of factors from the glia monolayer to the mesencephalic neurons and vice versa, without direct contact between cells (Gallo et al., 1995, 2000a,b). DAergic neuron survival was estimated after 24 h, by counting the number of TH<sup>+</sup> neurons over the DAPI-positive nuclei and TH<sup>+</sup> neurons expressed as percent (%) of control (-MPTP), and by determination of [3H]DA incorporation, which reflects DAergic cell count and functionality. Uptake of [3H]DA was performed essentially as previously described, by incubating the cell cultures for 20 min at 37◦ with 1 µM [3H]DA in Krebs-Ringer buffer [16 mM sodium phosphate, 119 mM NaCl, 4.7 mM KCl, 1.8 mM CaCl2, 1.2 mM MgSO4, 1.3 mM EDTA, and 5.6 mM glucose (pH 7.4)]. Non-specific DA uptake was blocked by mazindol (10 µM). Cells were then collected in 1 N NaOH after washing in ice-cold Krebs-Ringer buffer. Radioactivity was determined by liquid scintillation and specific [3H]DA uptake calculated by subtracting the mazindol counts from the wells without the uptake inhibitor (Morale et al., 2004; L'Episcopo et al., 2011a,b,c).

#### Ex vivo Isolation of As

Ex vivo isolation and culture of glial cells from the adult brain were previously detailed (Schwartz and Wilson, 1992; L'Episcopo et al., 2012, 2013). Isolated As from middle-aged mice of the studied groups (m-astro, >95%, GFAP<sup>+</sup> cells) were counted and plated at a final density of 0.4–0.6 × 10<sup>5</sup> cells/cm<sup>2</sup> in poly-D-lysine (10 µg/ml)–coated 6-, 12-, or 24-well plates, and their conditioned media were collected and stored at −70◦ . Some of the glial cells were exposed to different treatments and processed for qPCR or functional analyses or were used for direct coculture with primary mesencephalic neurons, as detailed in the previous section.

# Mitochondrial Activity With the 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium Bromide

#### Assay

The colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) assay was used to measure mitochondrial functionality in glial cells (Gennuso et al., 2004). Briefly, cells were incubated with 0.25 mg/ml MTT for 3 h at 37◦C, and mitochondrial enzyme activity was measured in culture supernatants in a spectrophotometer (Molecular Devices) at 570 nm, with a reference wavelength of 630 nm. Results are expressed as percentage changes of control.

# ROS and RNS

For reactive oxygen species (ROS) measurement, the redox membrane-permeant probe 2<sup>0</sup> ,70 -dichlorofluorescein diacetate (DCFH-DA, 50 µM) was added for 1 h at 37◦C, and cells were viewed under the confocal microscope (Gennuso et al., 2004). Measurement of iNOS-derived NO was carried out in cell free supernatant using Griess reagent (Marchetti et al., 2002; L'Episcopo et al., 2011c). To study the effect of inhibition of oxidative and nitrosative stress mediators, the freshly prepared GFAP<sup>+</sup> cells were cultured in the absence or the presence of the ROS antagonist, apocynin (Apo, 0.5 mM), and the specific iNOS inhibitor, L-Nil [L-N6-(1-iminoethyl)-lysine, 50 µM, Sigma] (Marchetti et al., 2002; Morale et al., 2004), applied after plating and determinations carried out 24–48 h after treatment (**Figure 6**).

#### Enzyme-Linked Immunosorbent Assay

Levels of cytokines were determined in tissue homogenates/cells using enzyme-linked immunosorbent assay (ELISA) kits (DuoSet ELISA Development System; R&D Systems, McKinley Place, MN, USA) following the manufacturer's protocol (Marchetti et al., 2002; L'Episcopo et al., 2011c).

# Data Analysis

Statistical significance between means ± SEM was analyzed by a two-way analysis of variance (ANOVA). Experimental series performed on different days were compared by the Student–Newman–Keuls t-test. A value of p < 0.05 was considered to be statistically significant.

# RESULTS

### In vitro Characterization of VM-As Displaying Neuroprotective Properties

As a first step of our transplantation protocol, primary VM-As cultures established from P2–3 and cultured as described for 14–20 DIV were processed for qPCR and immunocytochemical analyses to identify astrocytic (GFAP, S100B, ALDHL1) markers and verify the purity of the preparation by testing several non-astrocytic (i.e., microglial IBA1, oligodendrocyte, Olig2, and neuronal MAP2, NeuN) identity markers (**Figures 1A,B**). As observed, the astrocytic identity was confirmed by As co-expression with S100b, and ALDH1L1, two As-associated genes, whereas IBA1, Olig2, and MAP2 or NeuN were only poorly expressed, at both gene and protein levels (**Figure 1B**), thus confirming the full differentiation and purity of the VM-astro cultures. Additionally, when pulsed with the nucleotide analog, BrdU, a large proportion of GFAP<sup>+</sup> cells were co-stained after 24 h.

Our previous studies on neuron-As interactions first reported the region and growth factor specificity of As-derived molecules for neuronal development and growth, for acquisition of the mature neuronal phenotype, as well as for neuroprotection (Gallo et al., 1995; Morale et al., 2006; L'Episcopo et al., 2011a,b). Especially, P2–3 VM-As were shown to promote the differentiation of adult midbrain- but not subventricular zone (SVZ) NSCs into functionally active DAergic neurons, in vitro (L'Episcopo et al., 2011b, 2014a). Here the neuroprotective ability of VM-As was further verified in the coculture paradigm, where purified mesencephalic neurons (DAn) were layered on the top of VM-As in both the presence and the absence of the toxic MPTP metabolite, Mpp<sup>+</sup> (10 µM; **Figures 1C–E**). In accord with our previous findings (L'Episcopo et al., 2011a,b), in enriched primary mesencephalic neuronal cultures at 7 DIV, MPP<sup>+</sup> promoted the well-known neuron toxicity as demonstrated by the loss of TH<sup>+</sup> neurons, DAn atrophy, and inhibition of [3DA] incorporation (**Figures 1C, 2,D,E**). On the other hand, in VM-As–neuron cocultures, MPP<sup>+</sup> failed to induce DAergic cell death, as reflected by the greater TH<sup>+</sup> neuronal number and DA uptake levels measured (**Figures 1D,E**), with the TH<sup>+</sup> neurons cocultured with VM-As showing longer ramified neurites compared with those without As coculture (**Figure 1C2** vs. **Figure 1C4**). Accordingly, exposure of enriched DAn to either VM-As–conditioned medium (ACM) or VM-As insert paradigms similarly exerted a significant degree of TH<sup>+</sup> neuron protection against MPP<sup>+</sup> (**Figures 1C3,D,E**).

Together, these findings supported the differentiation, proliferative, and neuroprotective properties of our VM-As cultures, in vitro (**Figures 1A–C**).

interval respectively, by ANOVA followed by post hoc Newman–Keuls test.

#### Grafted VM-As Survive, Express Identity Markers, and Integrate Into the Aged MPTP-Lesioned Host SNpc

As a second step of our transplantation protocol, we verified the engraftment, proliferation, and distribution of VM-As tagged by incorporation of BrdU, which were unilaterally transplanted 7 days post-MPTP in the SNpc of middle-aged 9- to 11-month-old mice (**Figures 2A,B**). Spatio-temporal immunohistochemical, neurochemical, molecular, and motor behavioral analyses were carried out 1–3 wpt (**Supplementary Figure S1**).

By 1 week post-transplant (wpt), tVM-astros showed robust ipsilateral engraftment at all the SN rostro-caudal levels analyzed. Quantitative colocalization analyses in midbrain sections at 1 wpt (1 wpt = 100%) indicated that more than 70% of the engrafted BrdU+GFAP<sup>+</sup> cells expressed S100β, whereas only 3–5% of tVM-astros were IR for the oligodendroglial cell marker Olig2<sup>+</sup> and the microglial marker IBA1, and we did not observe colocalization of tVM-As with neuronal (NeuN) or DAergic (TH) markers (**Figure 2C**). Additionally, time-course analyses indicated that by 3 wpt, an almost 25–30% decline of GFAP<sup>+</sup> /BrdU<sup>+</sup> cells was observed within the lesioned SNpc (**Figure 2D**).

Together, tVM-As displaying in vitro well-recognized neuroprotective features, when transplanted unilaterally in the SNpc of middle-aged mice engrafted within the lesioned SNpc, expressed typical astrocytic, but not neuronal, nor microglial or oligodendroglial, markers and survived up to 4 wpt.

#### tVM-As Promote the Rescue of Endogenous TH<sup>+</sup> Neurons in the SNpc

To investigate the effect of the transplantation of tVM-astros on MPTP-induced neuronal loss in middle-aged mice, the number of DAergic cell bodies in the SNpc was assessed by TH immunoreactivity (**Figures 3A1–A3** and **Figure 3B**). Stereological quantification of TH<sup>+</sup> and Nissl<sup>+</sup> neurons in SNpc was performed to validate TH<sup>+</sup> neuron survival. tVM-astros grafted into MPTP-lesioned mice (MPTP/tVM-As) showed a significant (p ≤ 0.05) increase in endogenous TH<sup>+</sup> immunoreactivity and TH<sup>+</sup> neuron survival at both early (1 wpt) and later (4 wpt) tps (**Figures 3A1–A3** and **Figure 3B**) vs. control MPTP mice receiving an intranigral injection of PBS (MPTP/PBS; **Figures 3A1–A3** and **Figure 3B**). The observed increase in the number of TH<sup>+</sup> neurons appears specific for VM-As, since transplantation of a mock cell preparation (VMCs) failed to increase TH<sup>+</sup> neuron survival (not shown). By 4 wpt, the number of TH<sup>+</sup> neurons in MPTP/tVM-astro mice became almost comparable with unlesioned (saline/PBS) mice (**Figure 3B**).

Next, we studied the expression levels of several DAergic mRNA species by applying quantitative real-time polymerase chain reaction (qRT-PCR) in MPTP-injured SNpc tissues. Here, we found that in MPTP/tVM-astro mouse Th, the high-affinity DA transporter Slc6a3 (DAT), the vesicular monoamine transporter Slc18a2 (VMAT), and the DA-specific transcription factor Nr4a2 (Nurr1) required for the mature DA phenotype and survival (Kadkhodaei et al., 2009) showed an upregulation (by 2 to 3-fold), vs. MPTP/PBS control mice (**Figures 3C–F**).

Together, the tVM-As graft had a significant time-dependent rescue effect on endogenous TH neurons, which occurs at tissue, gene, and protein levels. Moreover, this effect was tVMastro–specific and was not attributable to a direct differentiation of transplanted tVM-astros into TH<sup>+</sup> neurons but, rather, to a rescue effect on endogenous cells.

#### tVM-As Grafts Counteract MPTP-Induced Loss of DAergic Innervation and Synaptosomal DA Uptake in the Str and Revert PD Motor Deficits

We next investigated the ability of tVM-astro grafts to increase the functionality of new TH<sup>+</sup> neurons, using quantitative confocal laser microscopy on striatal sections, the high-affinity synaptosomal DA uptake levels, and the analysis of motor behavior.

Hence, tVM-As grafts efficiently counteracted the MPTP-induced loss of striatal TH and DAT innervation (**Figures 4A1–A3** and **Figures 4B–C**). By contrast, corresponding levels in control MPTP/PBS mice were found to be significantly lower, compared with both MPTP/tVM-As (p ≤ 0.05) and saline/PBS controls (p ≤ 0.01) at each tp tested. Measuring the striatal uptake of radiolabeled DA([3H]-DA) in presynaptic terminals of middle-aged MPTP mice, we found that MPTP/tVM-As mice showed a significant (p ≤ 0.05) recovery of high-affinity striatal synaptosomal DA uptake (vs. MPTP/PBS) by 2 wpt (**Figure 4D**). This effect in MPTP/tVM-As mice increased over time, reaching values similar to those of saline/PBS controls by 4 wpt. On the contrary, MPTP/PBS mice showed a constant significant (p ≤ 0.01) reduction of synaptosomal DA uptake over time, vs. saline/PBS controls.

Behavior analyses confirmed that these structural and functional striatal changes were coupled with a full recovery of motor coordination deficits in MPTP/tVM-As mice, vs. MPTP/PBS controls, which started to be significant (p ≤ 0.05) at 2 wpt (**Figure 4E**).

Together, besides the effects of tVM-As on SNpc-DA neuronal cell bodies, tVM-As would promote a progressive recovery of the host striatal DA terminal region function, which further supports their role in enhancing endogenous recovery mechanisms in the aged brain.

#### tVM-As Rejuvenate the SN Microenvironment: Downmodulation of Microglial Pro-inflammatory Status

With age, both As and microglial cells become dysfunctional. As lose their neuroprotective, antioxidant, and pro-neurogenic potential, and microglia acquire a ''primed'' status (Streit, 2010; Njie et al., 2012; Niraula et al., 2017), capable of producing upregulated levels of pro-inflammatory cytokines when challenged with inflammatory/neurotoxic stimuli (L'Episcopo et al., 2010a,b, 2011c, 2018a,b). We thus questioned whether tVM-As might affect this ''harmful'' setting, thus ameliorating the injured microenvironment of middle-aged MPTP mice. To this end we first examined the effect of tVM-As on microglial response to MPTP. Notably, glial inflammatory mechanisms have long been recognized to contribute to both nigrostriatal degeneration and self-repair (see Marchetti and Abbracchio, 2005; Marchetti et al., 2005a,b,c; McGeer and McGeer, 2008; Marchetti et al., 2011; Przedborski, 2010; Gao et al., 2011; L'Episcopo et al., 2018a,b). We thus examined the microglial cell number and phenotype in vivo (**Figures 5A–D**) and found that tVM-As induced a significant (p ≤ 0.01) counteraction of MPTP-induced increased IBA1<sup>+</sup> microglial cells (**Figure 5B**) displaying the morphology of activated macrophage-like microglia (**Figures 5A1,A2**). Hence, in MPTP/PBS middle-aged mice, we observed a greater number of stage 3 IBA1<sup>+</sup> , ameboid microglia, showing a round-shaped body with short, thick, and stout processes, or stage 4 phagocytic glial cells (**Supplementary Table S3**), with round-shaped cells and no processes, indicative of an M1-activated phenotype (Kreutzberg, 1996). In stark contrast, MPTP/tVM-As mice displayed a ramified, more quiescent phenotype, with elongatedshaped cell bodies and long and thicker processes, comparable to stages 1–2 (**Figures 5A3,A4**, **Supplementary Table S3**), which suggested a switch towards the M2 less-reactive phenotype. Accordingly, using qPCR, we found that the expression levels of the pro-inflammatory M1-cytokines IL1-β, IL-6, and TNF-α were significantly (p ≤ 0.01) upregulated by about 5- to 6-fold in MPTP/PBS, as compared to saline-injected controls, in the face of no changes of the low levels of expression of the anti-inflammatory M2-cytokine, IL-10 (**Figure 5C**). By contrast, tVM-As significantly (p ≤ 0.01) reduced IL1-β, IL-6, and TNF-α to values almost comparable to saline/PBS mice, while IL-10 gene expression increased significantly (p ≤ 0.05) at both 2 and 4 wpt (**Figure 5D**). Supporting these results, when the cytokine protein levels were determined in VM tissues of the different groups at 4 wpt, by ELISA, we found significantly (p ≤ 0.01) greater levels of IL1-β, IL-6, and TNF-α in MPTP/PBS as compared to MPTP/tVM-As treated mice, exhibiting cytokine values comparable to those determined in saline/PBS-injected mice (**Figure 5E**). Additionally, MPTP/tVM-As mice, but not MPTP/PBS mice, had increased IL-10 protein levels (**Figure 5E**).

Thus, tVM-As efficiently override the microglial pro-inflammatory status and reduce the number of activated IBA1<sup>+</sup> cells within the middle-aged MPTP-lesioned SNpc during the studied experimental period. This effect was associated with a significant downregulation of the pro-inflammatory cytokines, TNF-α, IL-6, and IL-1β, associated to an increase of the anti-inflammatory cytokine, IL-10, at both gene and protein levels.

### tVM-As Upregulate As Antioxidant Self-defense in vivo and ex vivo: Contribution of Wnt/β-Catenin Signaling

Given the central role of As in antioxidant self-defense brain functions, we then looked at critical As-oxidative/nitrosative stress markers in SN tissues isolated from MPTP/PBS and MPTP/tVM-As, in vivo (**Figure 6A**). To differentiate the As-specific mRNAs vs. the SN-expressed mRNAs, we used ex vivo As cultures acutely isolated from MPTP/tVM-As and MPTP/PBS-As at 1 wpt (**Figure 6B**). Nrf2 is a conserved basic leucine zipper transcription factor which affords cytoprotection against xenobiotics and ROS through induction of antioxidant (ARE) and electrophile (EpRE) response elements (see Tebay et al., 2015; Zhang et al., 2017). HO1 is a principal mediator of cellular adaptive (i.e., antioxidant and anti-inflammatory) responses (Chen et al., 2009; Surh et al., 2009). SOD1 is critical for enhancing antioxidant self-defense during aging and neurodegenerative conditions, whereas its deficiency results in an accelerating aging phenotype (see Zhang et al., 2017).

Hence, we found that in response to MPTP, Nrf2 and the antioxidant gene, HO1, but not HO2, were sharply upregulated (p ≤ 0.01), together with SOD1, in aging MPTP/tVM-As–grafted vs. MPTP/PBS mice, which instead failed to activate an antioxidant self-defense response to the MPTP challenge (**Figures 6A,B**). Notably, tVM-astros greatly (p ≤ 0.01) increased Nrf2-antioxidant genes as compared to saline/PBS mice exhibiting very low transcript levels (**Figures 6A,B**). Within the nicotinamide adenine dinucleotide phosphate (NADPH) oxidases, Nox2 is the predominant oxidase family member expressed in As, at both the mRNA and protein level (Belarbi et al., 2017). Here, we detected decreased activation/expression of Nox2 in SN tissues from MPTP/tVM-As when compared to MPTP/PBS counterparts (**Figures 6A,B**) exhibiting an almost 3-fold increase over saline/PBS-injected mice. Additionally, tVM-As significantly counteracted the exacerbated expression of the harmful pro-inflammatory mediator, iNOS, vs. MPTP/PBS, showing instead a 3-fold upregulation (p ≤ 0.01, vs. saline-injected mice; **Figures 6A,B**). Remarkably, dual immunofluorescent stainings evidenced the neuroprotective effect of tVM-As and close interactions with both TH neurons and microglial cells, as suggested by the long GFAP<sup>+</sup> processes close to the rescued TH<sup>+</sup> neuronal cell body (**Figures 6D1,D2**), and in close contact with the long, ramified IBA1<sup>+</sup> processes (**Figures 6D4,D5**). On the

homogenate tissue samples (mean % ± SEM of n = 5 samples/experimental group), documents the ability of tVM-As to suppress the pro-inflammatory cytokines in the face of a significant increase in the anti-inflammatory cytokine, IL-10, when levels are compared to MPTP/PBS mice. <sup>∗</sup>p ≤ 0.05, ∗∗p ≤ 0.01 vs. saline/PBS; ◦◦p ≤ 0.01 vs. MPTP/PBS; ∗◦p ≤ 0.01 vs. saline/PBS and MPTP/PBS, at each time interval respectively, by ANOVA with post hoc Newman–Keuls.

other hand, a severe neuronal loss, poor GFAP+–TH<sup>+</sup> neuron interactions (**Figures 6C1,C2**) were observed in MPTP/PBS, where only round-shaped IBA1<sup>+</sup> cells were seen close to the reactive GFAP<sup>+</sup> As (**Figure 6C6**), thus supporting the failure of middle-aged MPTP-injured As to mount a neuroprotective response against MPTP.

Consistently, dual staining with GFAP and HO1 followed by confocal laser microscopic analyses uncovered an abundant co-expression of HO1 in tVM-As (p ≤ 0.01, vs. salineand MPTP-injected mice) as determined by the sharp increase in the percentage of GFAP<sup>+</sup> /HO1<sup>+</sup> out of the total GFAP<sup>+</sup> cells (77 ± 11%) in MPTP/tVM-As, when compared to MPTP/PBS (20 ± 4%) and saline/PBS mice (12 ± 4%; **Figure 6C3**). These results coupled to the abundant expression of iNOS in GFAP<sup>+</sup> As of MPTP/PBS (**Figure 6C5**) as opposed to MPTP/tVM-As mice suggested failure of aged As to activate the antioxidant and anti-inflammatory response upon MPTP challenge. Given that when iNOS and NADPH oxidase are present together, a potent toxin, peroxynitrite (ONOO-), may be generated which promotes the nitration of proteins (Gao et al., 2011), we then looked at the colocalization of 3-NT and found an

1 (SOD1), whereas Nox2 and inducible nitric oxide synthase (iNOS) are downregulated vs. MPTP/PBS both in SN tissues (A) and As-derived cultures (B). ∗∗p ≤ 0.01 vs. saline/PBS; ◦ ◦p ≤ 0.01 vs. MPTP/PBS, within each treatment group, respectively, by ANOVA followed by post hoc Newman–Keuls test. (C,D) Dual staining with TH (green) and GFAP (red) in midbrain sections from MPTP/PBS (C1,C2) and MPTP/tVM-As (D1–D3) showing the ability of VM-As grafts to revert MPTP-induced TH neuronal loss. In tVM-As, GFAP<sup>+</sup> cells extend the long process in close contact, almost embracing TH<sup>+</sup> neurons (D2, D3, arrows), and co-express HO1 (D4). Scale bars: (C1), 100 µm; (C2–C5), 25 µm. By contrast, in MPTP/PBS, GFAP<sup>+</sup> As (red) do not colocalize with HO (green, C3), but express iNOS (green) at high levels (C5). Likewise, the reactive oxygen (ROS) and nitrogen species (RNS) footprint, 3-nitrotyrosine (red), is abundantly expressed in GFAP<sup>+</sup> (green) cells (C3). Note that in tVM-As–grafted mice, dual staining of GFAP and ionized calcium-binding adapter molecule 1 (IBA1), showing a GFAP<sup>+</sup> cell in close contact with a ramified IBA1+cell phenotype (D5), as opposed to MPTP/PBS, IBA1<sup>+</sup> cells with a rounded morphology, is seen close to reactive GFAP<sup>+</sup> As (C6). Scale bars: (D1), 100 µm; (D2–D4), 25 µm. (E–G) Accordingly, in astrocyte cultures acutely isolated from MPTP/PBS, the production of ROS measured with the redox membrane-permeant probe 2<sup>0</sup> ,70 -dichlorofluorescein diacetate (DCFH-DA, F) and the generation of iNOS-derived RNS (G), in cell free supernatant, are sharply increased, resulting in a significant inhibition of mitochondrial reductase activity (MTT, E). These effects were efficiently counteracted in tVM-As cultures, with the application of the specific iNOS inhibitor, L-N6-(1-iminoethyl)-lysine (L-Nil) (E–G). Activation of Wnt/β-catenin signaling counteracts ROS and RNS production and reverts mitochondrial dysfunction of MPTP/PBS-As (E–G). By contrast, lipopolysaccharide (LPS) application further exacerbated oxidative/nitrosative stress in MPTP/PBS- vs. MPTP/tVM-As (E–G). <sup>∗</sup>p ≤ 0.05, ∗∗p ≤ 0.01 vs. saline/PBS; ∗◦p ≤ 0.01 vs. saline/PBS and MPTP/PBS; ◦ ◦p ≤ 0.01 vs. MPTP/PBS.

abundant expression of 3-NT in GFAP<sup>+</sup> As of MPTP/PBS (**Figure 6C4**) but not of MPTP/tVM-As (not shown). These findings suggested the potential of tVM-As to ameliorate the exacerbated endogenous oxidative/nitrosative glial status of the aged injured SN milieu via an upregulation of antioxidant functions.

As a proof of concept, we next studied some functional properties of tVM-As derived from MPTP/PBS and MPTP/tVM-As mice, looking at the mitochondrial activity with the MTT assay, the production of ROS with the redox membrane-permeant probe DCFH-DA, and the generation of iNOS-derived NO, in cell free supernatant using the Griess reagent (**Figures 6E–G**). To study the effect of inhibition of oxidative and nitrosative stress mediators, the freshly prepared GFAP<sup>+</sup> cells were cultured in the absence or the presence the specific iNOS inhibitor, L-Nil (Marchetti et al., 2002; Morale et al., 2004), whereas to study the effect of an exogenous inflammatory trigger, lipopolysaccharide (LPS) was applied at a dose of 100 ng/ml. Finally, the effect of Wnt/β-catenin activation was studied using AR, a drug that antagonizes GSK-3β (i.e., the kinase that phosphorylates β-catenin, leading to its degradation).

First we found that mitochondrial activity was increased in tVM-As when compared to MPTP/PBS-As metabolic activity, showing significantly (p ≤ 0.01) reduced activity vs. saline/PBS-As (**Figure 6G**). This effect is in line with the reduced amounts of both ROS and reactive nitrogen species (RNS) produced by MPTP/tVM-As vs. MPTP/PBS (**Figures 6E,F**), exhibiting significantly (p ≤ 0.01) greater levels vs. saline/PBS-As. These results thus support increased antioxidant properties of tVM-As. Accordingly, the specific iNOS-NO inhibitor, L-Nil, efficiently counteracted the sharp increase in oxidative and nitrosative stress of MPTP/PBS-As vs. MPTP/tVM-As counterparts (**Figures 6F,G**), with a beneficial effect on mitochondrial reductase activity (**Figure 6E**). Conversely, application of LPS counteracted the low ROS and RNS levels of MPTP/tVM-As, reducing mitochondrial activity, and further increased the already elevated ROS and RNS of MPTP/PBS-As (**Figures 6E–G**).

Given the decline of As-derived Wnts with age, we asked whether Wnt/β-catenin signaling activation might affect the exacerbated redox status of MPTP/PBS-As. Remarkably, we found a significant counteraction of both ROS and RNS upregulation of MPTP/PBS-As in AR-treated vs. untreated As cultures, to levels almost comparable to those measured in tVM-As (**Figures 6F,G**), which resulted in increased metabolic activity (**Figure 6E**).

Together, tVM-As grafts downregulate the exacerbated oxidative/nitrosative status of the aged MPTP-injured SNpc and As cultures via an increase of Nrf2-antioxidant genes, thereby shifting the pro-inflammatory and oxidative SN microenvironment. Additionally, the ability of Wnt/β-catenin activation to revert ROS- and RNS-exacerbated production of MPTP/PBS-As further suggested the contribution of Wnt/βcatenin signaling in redox properties of VM-As.

#### tVM-As Promote Upregulation of Wnt/β-Catenin Genes and Protect DAergic Neurons Against MPTP/MPP<sup>+</sup> : Crosstalk With Oxidative Stress Pathways

Previous findings clearly indicated that the aging process is associated with a sharp decline of As Wnts in the face of an increase of endogenous Wnt antagonists (L'Episcopo et al., 2011a,b, 2013, 2014b; Okamoto et al., 2011; Seib et al., 2013; Orellana et al., 2015; reviewed in Marchetti, 2018). We thus hypothesized that an As-driven disbalance of the Nrf2-ARE axis in middle-aged mice might contribute to aging-induced loss of Wnt signaling, and looked at the ability of tVM-As to override this hostile Wnt setting. Hence, both SNpc tissues from tVM-As (**Figure 7A**) and tVM-As–derived cultures (**Figure 7B**) enriched the expression of Wnt signature genes (p ≤ 0.01), including Wnt1 and β-catenin, together with Fzd-1 receptor, vs. MPTP/PBS SNpc tissues and MPTP/PBS-As cultures (**Figures 7A,B**). By contrast, a number of endogenous Wnt signaling inhibitors, including the Dkk1, sFrp1, and GSK-3β, were markedly (p ≤ 0.01) downregulated by 1 wpt in tVM-As, compared to MPTP/PBS exhibiting significantly (p ≤ 0.01) increased levels compared to saline-injected controls (**Figures 7A,B**), suggesting that tVM-As promoted a switch of the redox- and inflammationsensitive Wnt/β-catenin signaling response. Hence, by reverting the age-dependent decline of endogenous Wnt tone, this, in turn, might enhance the ensuing DAergic neuronal survival. To test this hypothesis, primary (E18) mesencephalic TH<sup>+</sup> neurons were grown on the top of MPTP/PBS-As (**Figure 7E1**) or tVM-As (**Figure 7E2**), for 5–7 DIV, and the survival and growth of TH<sup>+</sup> neurons as well as their functionality were studied using immunocytochemistry and DA uptake assay (**Figures 7C,D**). Hence, we found increased TH<sup>+</sup> neuron numbers and increased neurite extension and DA uptake levels when cocultured with tVM-As (**Figures 7C,D,E3,E4**). By contrast, TH<sup>+</sup> neurons grown on the top of MPTP/PBS-As exhibited a marked inhibition of neuronal survival, growth and functionality (**Figures 7C,D,E3,E4**). Accordingly, activating Wnt/β-catenin signaling with AR by lowering ROS and RNS and ameliorating MPTP/PBS-As metabolic activity (see **Figures 6E–G**) in turn increased both TH<sup>+</sup> neuron survival and DA uptake in MPTP/PBS-As cocultures (**Figures 7C,D,E5**), and further increased both parameters in the tVM-As–neuronal cocultures (**Figures 7C,D,E6**). By contrast, antagonizing Wnt/βcatenin signaling by the exogenous application of Dkk1 to tVM-As coculture reduced both neuronal numbers and function, thereby supporting increased As-Wnt tone as a crucial event involved in DAn growth. On the other hand, in MPTP/PBS-As overexpressing Dkk1, exogenous Dkk1 only slightly affected the already inhibited neuronal numbers found in MPTP/PBS-As–neuron cultures (**Figures 7C,D**). Finally, lowering ROS and RNS by the application of the antioxidant Apo, or L-Nil, in MPTP/PBS-As increased the number of TH<sup>+</sup> neurons and DA uptake levels vs. the untreated cocultures (p ≤ 0.05), but these effects were significantly (p ≤0.05) lower when compared to their tVM/As counterparts, where L-Nil and Apo induced a further increase in neuronal growth and DA uptake vs. the untreated tVM-As–neuronal cocultures (**Figures 7C,D**).

Together, these data suggest the potential of grafted VM-As to reverse the As decline of canonical Wnt's signaling components, downregulate major Wnt antagonists, and trigger the activation of an Nrf2/Hmox/Wnt/β-catenin axis favoring the ''rejuvenation'' of the aged, injured microenvironment, likely contributing to the DAergic neuroprotection and neurorescue herein observed.

FIGURE 7 | tVM-As grafts promote activation of Wnt/β-catenin signaling in SN tissue and As and reverse the impaired DAn growth in vitro. (A,B) Wnt/β-catenin signaling transcripts were analyzed by quantitative real-time PCR (qPCR) in SNpc tissues derived from MPTP/PBS- and MPTP/tVM-As mice, in vivo (A), and acutely isolated As, ex vivo, from both groups (B). Values (AU, mean % ± SEM of n = 5 samples/experimental group) are expressed as fold changes over control. VM-As grafts upregulated Wnt1, β-catenin, and Fzd1 receptor, whereas the endogenous Wnt antagonists, Dkk1, sFrp, and GSK-3β are downregulated vs. MPTP/PBS in both SN tissues (A) and As-derived cultures (B) at 1 wpi. \*\*p ≤ 0.01 vs. saline/PBS; ◦ ◦p ≤ 0.01 vs. MPTP/PBS; ∗◦ ◦p ≤ 0.01 vs. saline/PBS and MPTP/PBS, within each treatment group, respectively, by ANOVA followed by post hoc Newman–Keuls test. (C,D) Quantification of TH<sup>+</sup> neuron number at 7 DIV (C) and DA uptake levels measured by [3H]DA incorporation (D) in a direct coculture paradigm between MPTP/PBS- and MPTP/tVM-As acutely isolated, ex vivo, and primary mesencephalic DAn grown on the top of As, in the absence or the presence of the GSK-3β inhibitor AR, the Wnt antagonist Dkk1, the antioxidant apocynin (Apo), or the specific iNOS inhibitor L-Nil. Wnt signaling activation with AR increased TH neuron number and DA uptake levels in MPTP/PBS-As–DAn coculture (C–D,E5) and significantly magnified TH neuron number and functionality of MPTP/tVM-As–DAn coculture (C–D,E6). Lowering oxidative and nitrosative stress with Apo and L-Nil also increased TH neuron number and DA uptake in MPTP/PBS-As–DAn coculture, albeit to a lower extent vs. their tVM-As–DAn treated counterparts (C,D). Mean ± SEM values (n = 3 different culture preparations) are reported. \*p ≤ 0.05, \*\*p ≤ 0.01 vs. saline/PBS control; ◦ ◦p ≤ 0.01 vs. PBS; ◦ ◦ ◦p ≤ 0.01 vs. all treatments, within each group, respectively, by ANOVA followed by post hoc Newman–Keuls test. (E1–E6) Confocal images of GFAP<sup>+</sup> (red) in As cultures from MPTP/PBS (E1) and MPTP/tVM-As (E2), and TH neurons (green) and GFAP (red) in the coculture from MPTP/PBS- (E3) and MPTP/tVM-As (E4), showing TH+ neuron growth at 5–7 DIV neurons in the presence of tVM-As (E4), compared to the impaired TH neuron development in MPTP/PBS-As coculture (E3). Activation of Wnt/β-catenin signaling improved TH neuron growth in MPTP/PBS-As (E5), an effect that was magnified in MPTP/tVM-As–DAn coculture (E6). Bars: (E1,E2), 23 µm; (E3–E6), 20 µm.

#### DISCUSSION

Given that aging is the leading risk factor for PD development, and that by middle age, several parameters of astroglial and DAergic neuronal functionality start to be impaired, we focused on the middle-aged (9–11 months) male midbrain microenvironment to address the ability of primary fully differentiated tVM-As to activate intrinsic mechanisms that might rejuvenate As–neuronal dialog and promote DAergic neurorepair in MPTP-induced DAergic neurotoxicity. We herein report for the first time that grafting VM-As derived from postnatal (P2–3) midbrain within the middle-aged MPTP-injured SN can trigger a significant time-dependent endogenous nigrostriatal DAergic neurorescue. Clearly, various interactions between exogenous tVM-As and the pathological host milieu may underlie the improvement herein observed, and further time-course studies and in-depth molecular analyses both at a tissue and at a single-cell level are clearly needed to unravel how tVM-As grafts may drive a DAergic neurorescue program. However, from the presented results and based on our previous (see L'Episcopo et al., 2012, 2013, 2014b, 2018a,b; and as reviewed by Marchetti, 2018) and the recent literature findings (Tebay et al., 2015; Zhang et al., 2015, 2017; Zheng et al., 2017; Rizor et al., 2019), it seems tempting to speculate that the observed changes might result from a beneficial tVM-As–to–host SN crosstalk, promoting the rejuvenation of the host microenvironment, through the activation of an astrocytic Nrf2/ARE/Wnt/β-catenin prosurvival axis.

Notably, complex mechanisms are responsible for SNpc-DAergic cell death in PD, where the demise of this mesencephalic neuronal population is a process that is very long and is still not yet clarified (Del Tredici and Braak, 2012), as opposed to MPTP-induced DAergic degeneration (Langston, 2017), and extrapolations must be very careful. In the present study, our approach was aimed at combining: (i) the effect of age with; (ii) neurotoxin exposure; (iii) the consequent exacerbated inflammation; and (iv) male gender (representing four strong risk factors for PD), to address the ability of As grafting to mitigate the SN microenvironment and DAergic cell death.

Here, we found that, within the aged microenvironment, the proliferating tVM-As expressed S100β, but not microglia or neuronal markers, and survived for at least 4 wpt. At the SN level, tVM-As counteracted aging and MPTP-induced DAergic cell body loss, and at the striatal level, tVM-As promoted the recovery of DAT- and TH-IR, which was corroborated by the increased striatal synaptosomal DA uptake capacity and the reversal of motor impairment of aged MPTP mice, indicating the ability of tVM-As to restore nigrostriatal DAergic neurons and reinstate DAergic functionality at least for the studied time window of 4 wpt; further studies are needed to confirm, at a protein level, the capacity of tVM-As to revert the downregulation of the studied DA-signaling genes, as observed by qPCR analyses.

By analyzing microglial cells, we found that tVM-As efficiently counteracted aging and MPTP-induced increase of microglial cell number within the injured midbrain and shifted microglial morphological appearance from the M1 macrophage–like to the M2 more quiescent glial phenotype. Especially, in MPTP/PBS mice, glial cells showed round-shaped bodies with either short and thick processes or no processes, as observed in activated microglial cells (Kreutzberg, 1996), whereas in tVM-As mice, microglial cells exhibited more elongated cell bodies and longer, ramified processes, similar to the M2 glial phenotype. Accordingly, tVM-astros reversed the upregulated levels of the pro-inflammatory M1 cytokines, TNF-α, IL-6, and IL1-β, at both gene and protein expression levels in SN tissues, as compared to MPTP/PBS mice, but significantly increased the anti-inflammatory cytokine, IL-10, therefore supporting tVM-As ability to downregulate microglial exacerbation of aged MPTP/PBS mice. While further studies are clearly needed to address tVM-As–microglial interactions, their close contacts coupled to the observed change in glial phenotype may favor the possibility of a ''beneficial'' As–microglia crosstalk in tVM-As–grafted vs. MPTP/PBS mice, likely contributing to the downregulation of inflammation and DAergic neurorescue. In stark contrast, when the microglia is chronically activated, as observed in MPTP/PBS aging mice, As can lose both immunomodulatory and neuroprotective properties with harmful consequences for the dysfunctional DAergic neurons, as herein observed.

We then focused on the astrocytic compartment and found that tVM-As grafts sharply increased the expression of Nrf2, a chief astrocytic regulator of oxidative stress and inflammation, and upregulated the antioxidant and anti-inflammatory HO1 and SOD1 transcripts, in the face of a marked suppression of iNOS. Clearly, from our gene expression analysis, it is not possible to differentiate the exogenous tVM-As vs. the endogenous As-expressed mRNAs, albeit using ex vivo As cultures from MPTP/tVM-As and MPTP/PBS-As at 1 wpt, we delineated a clear upregulation of antioxidant and Wnt signaling genes in MPTP/tVM-As vs. their MPTP/PBS-As counterparts. Especially, we found VM-As of middle-aged MPTP-lesioned mice as a critical source of Wnt antagonists likely responsible for their failure to exert neuroprotection. Hence, at a functional level, tVM-As generated significantly lower amounts of ROS and RNS compared to the exacerbated reactive mediators of MPTP/PBS-As, resulting in As mitochondrial impairment, whereas Wnt signaling activation in MPTP/PBS-As efficiently reversed the upregulated ROS and RNS to levels measured in L-Nil–treated As levels, suggesting crosstalk between Wnt signaling and oxidative stress pathways. Remarkably, tVM-astros promoted Wnt/βcatenin activation within the middle-aged, MPTP-injured SN in vivo and in ex vivo direct As–neuron coculture paradigms, where MPTP/tVM-As promoted the growth and functionality of developing primary mesencephalic DAn, as compared to their counterparts derived from MPTP/PBS mice, which failed to exhibit the critical supportive properties and instead inhibited TH<sup>+</sup> neuron survival and functionality.

Thus, switching the SN neurorescue-unfriendly environment, grafted As promoted a beneficial antioxidant/anti-inflammatory ''Wnt-on'' prosurvival milieu, highlighting As-derived factors/mechanisms as the crucial keys for successful therapeutic outcomes in PD.

#### As Grafting Switched the Harmful SN Milieu of Aging MPTP Mice Driving the Master Regulator of the Oxidative Stress and Inflammatory Response

With advancing age, the decline of the nigrostriatal DAergic system coupled with the progressive loss of DAergic neuron adaptive potential is believed to contribute to the slow nigrostriatal degeneration of PD (Hornykiewicz, 1993; Bezard et al., 2000; Collier et al., 2007; de la Fuente-Fernández et al., 2011; Blesa et al., 2017). In fact, the activation of endogenous compensatory mechanisms is thought to mask the appearance of PD before the appearance of the first clinical symptoms (Zigmond et al., 2009; Blesa et al., 2017), which raises the possibility that some individuals with PD suffer from a reduction of these neuroprotective mechanisms and that treatments that boost these mechanisms may provide therapeutic benefit (see Zigmond et al., 2009). Interestingly, while young adult rodents experience a time-dependent recovery/repair from neurotoxic or immunological challenges, aging mice fail to recover for their entire life span (L'Episcopo et al., 2011a,b,c). With aging, the glial adaptive mechanisms are reduced, resulting in increased DAergic neuron vulnerability to various risk factors, including genetic, inflammatory, and environmental toxic exposures, such as MPTP (L'Episcopo et al., 2018a).

In fact, declined Nrf2-antioxidant signaling during aging leads to accumulation of ROS/RNS and oxidative stress, which is either causally linked or associated with numerous health problems including neurodegenerative conditions; thus, targeting Nrf2 has been suggested as a promising therapeutical avenue in neurodegeneration (Abdalkader et al., 2018). Remarkably, previous studies of Chen et al. (2009) indicated that Nrf2 expression restricted to As is sufficient to protect against MPTP toxicity, suggesting that As modulation of the Nrf2-ARE pathway is a promising target for therapeutics aimed at reducing or preventing neuronal death in PD (Chen et al., 2009). Significantly, Lastres-Becker et al. (2012) reported that α-synuclein expression and Nrf2 deficiency cooperate to aggravate protein aggregation, neuronal death, and inflammation in early-stage PD, and very recent findings further define As oxidative/nitrosative stress as a key etiopathogenetic factor in PD (Rizor et al., 2019). Additionally, we pinpointed that by middle age, a significant decrease of Nrf2/HO1 response to MPTP in striatal and SVZ-As played a prominent role and synergized with the heightened inflammatory SVZ milieu to downregulate SVZ neurogenesis (L'Episcopo et al., 2013). Here, we further uncovered that VM-As grafts promoted a switch of inflammatory M1 microglia phenotype and turned the aged As into a supportive and neuroprotective A2 phenotype.

In fact, amongst glial cytotoxic molecules, iNOS-derived NO, a superoxide from the plasma membrane NADPH oxidase, associated with a number of potent inflammatory cytokines, including TNF-α, IL-1β, IL-6, and IFN-γ, is known to exert detrimental effects in DAergic neurons (see Gao and Hong, 2008; Hirsch and Hunot, 2009). Indeed, when iNOS and NADPH oxidase are present together, a potent toxin, ONOO-, may be generated, which promotes the nitration of proteins (Gao and Hong, 2008), with harmful consequences for DAergic neurons. Conversely, As-secreted antioxidant factors represent a critical self-protective system against MPTP and 6-OHDA cytotoxicity (Chen et al., 2009). In fact, the prolonged dysfunction of As and activation of microglia accelerate degeneration of DAergic neurons in the rat SN and block compensation of early motor dysfunction induced by 6-OHDA (Kuter et al., 2018).

Here, we uncovered the ability of VM-As grafts to override the strong oxidant and pro-inflammatory status of the SN milieu of middle-aged MPTP-injured As overexpressing iNOS and the harmful RNS footprint, 3NT, by decreasing the number activated macrophage-like microglia and the expression of major inflammatory transcripts and proteins, in vivo, resulting in suppression of the exacerbated ROS and RNS production of MPTP-injured As, ex vivo.

Coupled to the critical role of As mitigating mitochondrial dysfunctions in human DAergic neurons derived from iPSC (Du et al., 2018), these present findings support the ability of VM-As grafts to boost antioxidant self-defenses, as a potential mechanistic link in the herein observed DAergic neurorescue effects.

#### As Grafting Triggering Wnt/Nrf2/HO1 Crosstalk Mitigates Inflammation Promoting DAn Neurorescue

The Wnt/β-catenin signaling pathway is of utmost importance owing to its ability to promote tissue repair and regeneration of stem cell activity in diverse organs, and in light of its crucial role in age-related pathogenesis and therapy of disease (Nusse and Clevers, 2017; García-Velázquez and Arias, 2017; Marchetti et al., 2020). In the last decade, we characterized As-derived Wnt1 as crucial actor in As–DAn crosstalk involved in neuroprotection against several neurotoxic and inflammatory insults (Marchetti and Pluchino, 2013; Marchetti et al., 2013; Marchetti, 2018) and the regulation of neurogenesis and SVZ plasticity via crosstalk with inflammatory and oxidative stress signaling pathways during aging and MPTP-induced parkinsonism (Marchetti et al., 2020).

Indeed, a major finding of aging is the decline of Wnt signaling and that this ''Wnt-off '' state very likely drives the decline of neurogenesis and the exacerbation of inflammation, thus increasing DAn vulnerability to a number of cytotoxic and inflammatory insults (Marchetti and Pluchino, 2013; L'Episcopo et al., 2018a,b; Marchetti, 2018). Here, we further unveiled that VM-As grafts promoted an enriched expression of canonical Wnt signature genes in the middle-aged MPTP-injured VM, which included Wnt1, βcatenin, and Fzd1 receptor, thus triggering a Wnt-on state, likely contributing to DAergic neurorescue. Accordingly, a number of endogenous Wnt signaling antagonists, such as GSK-3β, Dkk1, and sFrp1, were downregulated by 1 wpt in tVM-As SN tissues and As cultures, when compared to their MPTP/PBS-As counterparts exhibiting a significant upregulation of endogenous Wnt antagonists, supporting the Wnt-off state of the aged MPTP-injured midbrain, as observed in our previous studies (Marchetti, 2018). Here, As overexpression of Wnt antagonists and upregulated ROS and RNS levels of MPTP/PBS mice impaired As–neuron interactions in vivo and ex vivo, thus resulting in a marked inhibition of DAn survival and growth, whereas activating Wnt/βcatenin signaling, as observed in VM-As–grafted mice and VM-As–derived cultures, or MPTP/PBS-As–derived cultures treated with the GSK-3β antagonist, powerfully reverted oxidative/nitrosative stress markers promoting DAn survival and growth.

Thus, in light of the role of Wnt signaling in the inflammatory and oxidative stress response and the crosstalk between inflammatory and Wnt signaling components (Chong et al., 2007, 2010; Halleskog et al., 2011; Kilander et al., 2011; L'Episcopo et al., 2011b; Schaale et al., 2011; Halleskog and Schulte, 2013; Marchetti and Pluchino, 2013; Ma and Hottiger, 2016; Zheng et al., 2017; L'Episcopo et al., 2018a,b; Marchetti, 2018), it seems tempting to suggest that the Wnt-off state of MPTP/PBS mice likely exacerbates the pro-inflammatory microglial phenotype of MPTP/PBS mice, in turn responsible for the hostile SN milieu that synergizes with the downregulation of MPTP/PBS-As antioxidant self-defenses.

In stark contrast, Wnt/β-catenin activation of tVM-As–grafted mice might well promote a beneficial effect, switching the microglial M1 phenotype to a likely more quiescent anti-inflammatory state (see L'Episcopo et al., 2018a,b). Especially, a lack of As-derived Wnt-microglial dialog might well contribute to the loss of major Nrf2-antioxidant genes responsible for As failure to protect and rescue/repair the injured DAn of middle-aged MPTP mice.

Together, these findings argue in favor of reciprocal As/microglial/neuron interactions and suggest the Nrf2/HO1/Wnt/β-catenin axis as a critical mediator in promoting neuroprotection.

#### CONCLUSIONS

We have shown that grafting tVM-astros can override the aged hostile SN milieu and drive DAergic neurorescue in MPTP-induced nigrostriatal toxicity triggering the activation of a ''beneficial'' astrocytic Nrf2/Hmox1/Wnt/β-catenin axis that rejuvenates the SN microenvironment and favors DAergic neurorescue/neurorepair.

In light of the emerging implications of dysfunctional As in major human NDs (Endo et al., 2015; Anderson et al., 2016; L'Episcopo et al., 2016, 2018a,b; Booth et al., 2017; Patel et al., 2019), together with the critical role of As and As-derived factors in both pharmacological and cell therapeutical interventions (Kondo et al., 2014; Yang et al., 2014; Nicaise et al., 2015; Hall et al., 2017; Rivetti di Val Cervo et al., 2017; Barker et al., 2018; Du et al., 2018; Kuter et al., 2018; Song et al., 2018; Bali et al., 2019; Klapper et al., 2019; Rizor et al., 2019), our findings highlight

#### REFERENCES


As-focused therapies as the crucial key for beneficial outcomes in PD.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### ETHICS STATEMENT

This study has been approved by the local Ethics Committee ''Comitato Etico IRCCS Sicilia-Oasi Maria SS'' of the Associazione Oasi Maria SS., based in Troina (Italy), Via Conte Ruggero, 73.

#### AUTHOR CONTRIBUTIONS

Authors contributing to the presented experimental findings and manuscript editing. BM: conception and design, data analysis and interpretation, manuscript writing. MS: astrocyte transplantation, histopathology, data analyses and interpretation, manuscript writing and final approval of manuscript. CT, NT, FL'E, SC, and CG: performed experiments, data analyses and interpretation, final approval of manuscript.

#### FUNDING

This research program has received support from different funding agencies, in particular, the Italian Ministry of Health (Ricerca Corrente-2016-2019), the Italian Ministry of Education, University and Research (MIUR), CHANCE Grant 2016–2019, the OASI Research Institute–IRCCS, Troina, Italy, and BIOMETEC at the University of Catania, Medical School, Catania, Italy.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi. 2020.00024/full#supplementary-material.

by upregulating glutamate transporter GLT-1 in astrocytes. J. Neurosci. 36, 10529–10544. doi: 10.1523/JNEUROSCI.1643-16.2016


systematic review. Mol. Neurodegener. 12:84. doi: 10.1186/s13024-017- 0225-5


to aggravate protein aggregation, neuronal death and inflammation in early-stage Parkinson's disease. Hum. Mol. Genet. 21, 3173–3192. doi: 10.1093/hmg/dds143


endogenous precursor cell response in a rat model of Parkinson's disease. J. Comp. Neurol. 515, 102–115. doi: 10.1002/cne.22033


status and the pathways through which it attenuates degenerative disease. Free Radic. Biol. Med. 88, 108–146. doi: 10.1016/j.freeradbiomed.2015. 06.021


**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 Serapide, L'Episcopo, Tirolo, Testa, Caniglia, Giachino and Marchetti. 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.

digital media

of impactful research

article's readership