# PREVENTING DEVELOPMENTAL BRAIN INJURY – FROM ANIMAL MODELS TO CLINICAL TRIALS

EDITED BY : Masahiro Tsuji, Stéphane V. Sizonenko and Olivier Baud PUBLISHED IN : Frontiers in Neurology and Frontiers in Pediatrics

#### Frontiers Copyright Statement

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

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

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

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

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

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

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-075-2 DOI 10.3389/978-2-88963-075-2

#### 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

# PREVENTING DEVELOPMENTAL BRAIN INJURY – FROM ANIMAL MODELS TO CLINICAL TRIALS

Topic Editors:

Masahiro Tsuji, Kyoto Women's University, Japan Stéphane V. Sizonenko, University of Geneva, Switzerland Olivier Baud, University of Geneva, Switzerland

Citation: Tsuji, M., Sizonenko, S. V., Baud, O., eds. (2019). Preventing Developmental Brain Injury – From Animal Models to Clinical Trials. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-075-2

# Table of Contents

#### 1. EDITORIAL

*07 Editorial: Preventing Developmental Brain Injury—From Animal Models to Clinical Trials*

Masahiro Tsuji, Stéphane V. Sizonenko and Olivier Baud

### 2. REVIEWS


Clément Chollat, Loïc Sentilhes and Stéphane Marret

*36 Nitric Oxide Synthase Inhibition as a Neuroprotective Strategy Following Hypoxic–Ischemic Encephalopathy: Evidence From Animal Studies*

Laurent M. A. Favié, Arlette R. Cox, Agnes van den Hoogen, Cora H. A. Nijboer, Cacha M. P. C. D. Peeters-Scholte, Frank van Bel, Toine C. G. Egberts, Carin M. A. Rademaker and Floris Groenendaal

*50 Neuroprotective Drugs in Infants With Severe Congenital Heart Disease: A Systematic Review*

Raymond Stegeman, Kaya D. Lamur, Agnes van den Hoogen, Johannes M. P. J. Breur, Floris Groenendaal, Nicolaas J. G. Jansen and Manon J. N. L. Benders

### 3. ANIMAL STUDIES

#### 3.1. NEONATAL ENCEPHALOPATHY


Anton Kichev, Ana A. Baburamani, Regina Vontell, Pierre Gressens, Linda Burkly, Claire Thornton and Henrik Hagberg

*85 Early Detection of Hypothermic Neuroprotection Using T2-Weighted Magnetic Resonance Imaging in a Mouse Model of Hypoxic Ischemic Encephalopathy*

Sydney E. Doman, Akanksha Girish, Christina L. Nemeth, Gabrielle T. Drummond, Patrice Carr, Maxine S. Garcia, Michael V. Johnston, Sujatha Kannan, Ali Fatemi, Jiangyang Zhang and Mary Ann Wilson

*96 Bone Morphogenetic Protein (BMP)-3b Gene Depletion Causes High Mortality in a Mouse Model of Neonatal Hypoxic-Ischemic Encephalopathy*

Yuko Ogawa, Masahiro Tsuji, Emi Tanaka, Mikiya Miyazato and Jun Hino


Takeo Mukai, Arinobu Tojo and Tokiko Nagamura-Inoue

*125 Brain Metabolism Alterations Induced by Pregnancy Swimming Decreases Neurological Impairments Following Neonatal Hypoxia-Ischemia in Very Immature Rats*

Eduardo F. Sanches, Yohan Van de Looij, Audrey Toulotte, Analina R. da Silva, Jacqueline Romero and Stephane V. Sizonenko

*141 Modulation of Microglial Activation by Adenosine A2a Receptor in Animal Models of Perinatal Brain Injury*

Marina Colella, Manuela Zinni, Julien Pansiot, Michela Cassanello, Jérôme Mairesse, Luca Ramenghi and Olivier Baud

*155 Intravenous Administration of Bone Marrow-Derived Mesenchymal Stem Cell, but not Adipose Tissue-Derived Stem Cell, Ameliorated the Neonatal Hypoxic-Ischemic Brain Injury by Changing Cerebral Inflammatory State in Rat*

Yuichiro Sugiyama, Yoshiaki Sato, Yuma Kitase, Toshihiko Suzuki, Taiki Kondo, Alkisti Mikrogeorgiou, Asuka Horinouchi, Shoichi Maruyama, Yoshie Shimoyama, Masahiro Tsuji, Satoshi Suzuki, Tokunori Yamamoto and Masahiro Hayakawa


#### 3.2. BRAIN INJURIES ASSOCIATED WITH PREMATURITY

*189 Repetitive Neonatal Erythropoietin and Melatonin Combinatorial Treatment Provides Sustained Repair of Functional Deficits in a Rat Model of Cerebral Palsy*

Lauren L. Jantzie, Akosua Y. Oppong, Fatu S. Conteh, Tracylyn R. Yellowhair, Joshua Kim, Gabrielle Fink, Adam R. Wolin, Frances J. Northington and Shenandoah Robinson

#### *200 Mild Intrauterine Hypoperfusion Leads to Lumbar and Cortical Hyperexcitability, Spasticity, and Muscle Dysfunctions in Rats: Implications for Prematurity*

Jacques-Olivier Coq, Maxime Delcour, Yuko Ogawa, Julie Peyronnet, Francis Castets, Nathalie Turle-Lorenzo, Valérie Montel, Laurence Bodineau, Phillipe Cardot, Cécile Brocard, Sylvie Liabeuf, Bruno Bastide, Marie-Hélène Canu, Masahiro Tsuji and Florence Cayetanot

*213 Effects of Intrauterine Inflammation on Cortical Gray Matter of Near-Term Lambs*

Vanesa Stojanovska, Anzari Atik, Ilias Nitsos, Béatrice Skiöld, Samantha K. Barton, Valerie A. Zahra, Karyn Rodgers, Stuart B. Hooper, Graeme R. Polglase and Robert Galinsky

#### 3.3. OTHER BRAIN INJURIES

*223 Dietary Iron Repletion Following Early-Life Dietary Iron Deficiency Does not Correct Regional Volumetric or Diffusion Tensor Changes in the Developing Pig Brain*

Austin T. Mudd, Joanne E. Fil, Laura C. Knight and Ryan N. Dilger


Hemmen Sabir, John Dingley, Emma Scull-Brown, Ela Chakkarapani and Marianne Thoresen


Kazuhiro Osato, Yoshiaki Sato, Akari Osato, Machiko Sato, Changlian Zhu, Marcel Leist, Hans G. Kuhn and Klas Blomgren

#### 4. CLINICAL STUDIES

*282 Lymphocytes Contribute to the Pathophysiology of Neonatal Brain Injury* Arshed Nazmi, Anna-Maj Albertsson, Eridan Rocha-Ferreira, Xiaoli Zhang, Regina Vontell, Aura Zelco, Mary Rutherford, Changlian Zhu, Gisela Nilsson, Carina Mallard, Henrik Hagberg, Jacqueline C. Y. Lai, Jianmei W. Leavenworth and Xiaoyang Wang

#### *293 Investigation of EEG Activity Compared With Mean Arterial Blood Pressure in Extremely Preterm Infants*

Sujith S. Pereira, Stephen T. Kempley, David F. Wertheim, Ajay K. Sinha, Joan K. Morris and Divyen K. Shah

### *300 Combined Analysis of* Interleukin-10 *Gene Polymorphisms and Protein Expression in Children With Cerebral Palsy*

Lei Xia, Mingjie Chen, Dan Bi, Juan Song, Xiaoli Zhang, Yangong Wang, Dengna Zhu, Qing Shang, Falin Xu, Xiaoyang Wang, Qinghe Xing and Changlian Zhu

#### *307 Cerebral Lactate Concentration in Neonatal Hypoxic-Ischemic Encephalopathy: In Relation to Time, Characteristic of Injury, and Serum Lactate Concentration*

Tai-Wei Wu, Benita Tamrazi, Kai-Hsiang Hsu, Eugenia Ho, Aaron J. Reitman, Matthew Borzage, Stefan Blüml and Jessica L. Wisnowski

# Editorial: Preventing Developmental Brain Injury—From Animal Models to Clinical Trials

#### Masahiro Tsuji <sup>1</sup> , Stéphane V. Sizonenko<sup>2</sup> \* and Olivier Baud3,4

<sup>1</sup> Department of Food and Nutrition, Kyoto Women's University, Kyoto, Japan, <sup>2</sup> Division of Development and Growth, Department of Pediatrics, Gynecology and Obstetrics, School of Medicine, University of Geneva, Geneva, Switzerland, <sup>3</sup> Division of Neonatology, Department of Pediatrics Gynecology and Obstetrics, School of Medicine, University of Geneva, Geneva, Switzerland, <sup>4</sup> Robert Debré Hospital, INSERM U1141, Paris-Diderot University, Paris, France

Keywords: neonatal encephalopathy, brain injury, prematurity and low birth weight, animal model, clinical study

#### **Editorial on the Research Topic**

#### **Preventing Developmental Brain Injury—From Animal Models to Clinical Trials**

A variety of brain injuries originating from fetal, perinatal, and neonatal periods may cause mortality, and many of the survivors of those injuries suffer from life-long morbidity. It is our task as researchers to prevent and reduce such brain injuries in infants. This is why we Topic Editors chose the theme "Preventing developmental brain injury—from animal models to clinical trials."

#### Edited by:

Jo Madeleine Wilmshurst, University of Cape Town, South Africa

#### Reviewed by:

Mary Ann Rutherford, King's College London, United Kingdom Richard Joseph Burman, University of Oxford, United Kingdom

> \*Correspondence: Stéphane V. Sizonenko Stephane.Sizonenko@unige.ch

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 20 April 2019 Accepted: 03 July 2019 Published: 19 July 2019

#### Citation:

Tsuji M, Sizonenko SV and Baud O (2019) Editorial: Preventing Developmental Brain Injury—From Animal Models to Clinical Trials. Front. Neurol. 10:775. doi: 10.3389/fneur.2019.00775

Among those developing brain injuries, neonatal encephalopathy in full-term infants has been the most crucial issue for decades (1). Neonatal encephalopathy refers to newborns with symptoms of acute cerebral dysfunction such as depressed consciousness, abnormal muscle tone, respiratory distress, feeding difficulties, and seizures (2, 3). Hypoxic-ischemic injury is the most common (up to 85% of cases), and ischemic/hemorrhagic stroke is the second most common, etiology and pathophysiology of neonatal encephalopathy (3). Approximately 3% of babies born in the world die by 4 weeks of age, and asphyxia, i.e., neonatal encephalopathy accounts for 23% of neonatal mortality (4); hence, the mortality rate due to neonatal encephalopathy is estimated to be 70/1,000 livebirths. Almost all (99%) neonatal mortality occurs in resource-limited countries (4). The incidence of neonatal encephalopathy in term newborns has been declining in resource-rich countries, and is currently 1/1,000 livebirths or less (5, 6). Nevertheless, nearly half of infants with neonatal encephalopathy in resource-rich countries die or develop permanent neurological disabilities, such as cerebral palsy, intellectual disability, and epilepsy (7). Therefore, the burden of neonatal encephalopathy remains high for patients, caregivers, and society.

Currently, the most common brain injuries caused during fetal, perinatal, and neonatal periods are those associated with prematurity, i.e., fetal growth restriction, preterm birth, and low birth weight (8). Approximately 15 million newborns were estimated to be born preterm (<37 weeks of gestation) in 2010, which accounted for 11% of all livebirths worldwide, ranging from 5% in some European countries to 18% in some African countries (9). The preterm birth rate has been increasing over the past two decades (9). Similarly, the rate of low birth weight (birth weight is <2,500 g) has been increasing: 7–9% of all live births in resource-rich countries (10, 11) and 11–12% in resource-limited countries (12). Intrauterine hypoperfusion and inflammation are the two leading causes of brain injuries associated with prematurity (13–16). Although most infants do not present obvious neurological symptoms during the neonatal period, they may develop cerebral palsy and neurodevelopmental disorders such as attention-deficit/hyperactive disorders and autistic spectrum disorders during childhood (17, 18). Survivors of very low birth weight (<1,500 g) may present cerebral palsy in 5–10% of the individuals and cognitive/behavioral/attentional deficits

**7**

in ∼50% of these individuals (19, 20). The main focus of outcome studies in neonatal care has been the extremely low-birth-weight (<1,000 g) infants and extremely preterm infants [born before 28 weeks of gestation (9)]. Recent studies, however, have shown that even infants born with mild prematurity (birth weight is 1,500– <2,500 g; gestational age is 32–<37 weeks) have significantly higher risks of developing numerous neurological, psychological, and general health problems later in life (21). This phenomenon is now widely known as DOHaD (developmental origin of health and diseases) (22). While many preclinical studies have been conducted in neonatal encephalopathy, a limited number of preclinical studies have been conducted regarding brain injuries associated with prematurity.

Neonatal encephalopathy and brain injury associated with prematurity seem to be contrasting problems, as the former is a medical emergency during the neonatal period, while the latter is a gradually emerging problem in childhood. While it is important to shed light on rare brain injuries, neonatal encephalopathy in term infants and brain injuries associated with prematurity are the most prominent issues that researchers and funding sources need to address considering their impact on patients and society.

This Research Topic collects 29 articles: 20 animal studies (including 1 in vitro study), 3 clinical studies, 1 combined study with clinical and animal studies, and 5 reviews. Of the 21 animal studies, the majority used rodents, and 3 studies used large animals, i.e., piglets or lambs. Hypoxic-ischemic injury was the most frequently used model (11 articles) in this Research Topic collection. Other models included intrauterine ischemia/inflammation (3 articles) and irradiation (2 articles). As the theme of this Research Topic was "preventing developing brain injury," many of the animal studies explored the effects of novel therapies. Five studies examined the effects of cellbased therapies, 2 studies examined erythropoietin, and others examined melatonin, inhaled carbon dioxide therapy, and protective ventilation strategies. Some studies examined the effects of certain genomic modifications as possible targets of novel therapies. Clinical studies investigated EEG and blood pressure, IL-10 gene polymorphism, and cerebral lactate in infants with brain damage. Despite the Topic Editors' expectations, there was no manuscript submission of an

#### REFERENCES


interventional clinical trial, which may suggest a scarcity of clinical trials and difficulty in conducting clinical trials of novel therapies in sick infants. The review articles summarized the effects of therapies, such as oxytocin, magnesium, and nitric oxide synthase inhibition.

With respect to the demographics of this Research Topic collection, the corresponding author's laboratories are located in a variety of countries: Japan (6 articles), the United States (5 articles), France (4 articles), the United Kingdom (3 articles), Sweden (3 articles), the Netherlands (2 articles), Switzerland (1 article), Germany (1 article), China (1 article), South Korea (1 article), Canada (1 article), and Australia (1 article). Thirteen articles represent international collaborative studies.

The Topic Editors are pleased to have a wide variety of 29 articles from many countries around the world. None of the articles in this collection, however, are from the resource-limited world. It is important to encourage research in this area in resource-limited countries where developmental brain injuries are expected to be most prevalent. We hope that research on neonatal brain injuries will become even more active and will contribute to better preventions and treatments for these injuries in the near future.

#### AUTHOR CONTRIBUTIONS

MT wrote the draft. SS and OB critically reviewed the manuscript.

### FUNDING

MT was supported by JSPS KAKENHI Grant Number 17K10200, JSPS Bilateral Open Partnership Joint Research Projects, and the Joint Research Project of the Institute of Medical Science, the University of Tokyo (2017-1017). SS was supported by Swiss Excellence Scholarship for Foreign Scholars, by the Swiss National Fund No. 33CM30-124101/140334 and the Fondation pour Recherches Médicales, Geneva.

#### ACKNOWLEDGMENTS

We thank all the reviewers who provided useful suggestions to improve the manuscript.


**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 Tsuji, Sizonenko and Baud. 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.

# Role of Perinatal inflammation in Neonatal Arterial ischemic Stroke

*Antoine Giraud1,2†, Clémence Guiraut2†, Mathilde Chevin2†, Stéphane Chabrier3 and Guillaume Sébire2 \**

*1EA 4607 SNA EPIS, Jean Monnet University, Saint-Etienne, France, 2Child Neurology Division, Department of Pediatrics, McGill University, Montréal, QC, Canada, 3 French Center for Pediatric Stroke and Pediatric Rehabilitation Unit, Department of Pediatrics, Saint-Etienne University Hospital, Saint-Etienne, France*

Based on the review of the literature, perinatal inflammation often induced by infection is the only consistent independent risk factor of neonatal arterial ischemic stroke (NAIS). Preclinical studies show that acute inflammatory processes take place in placenta, cerebral arterial wall of NAIS-susceptible arteries and neonatal brain. A top research priority in NAIS is to further characterize the nature and spatiotemporal features of the inflammatory processes involved in multiple levels of the pathophysiology of NAIS, to adequately design randomized control trials using targeted anti-inflammatory vasculoand neuroprotective agents.

#### *Edited by:*

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Bruno J. Gonzalez, University of Rouen, France Maxime Gauberti, INSERM, France Justin Dean, University of Auckland, New Zealand*

#### *\*Correspondence:*

*Guillaume Sébire guillaume.sebire@mcgill.ca*

*† These authors have contributed equally as first authors.*

#### *Specialty section:*

*This article was submitted to Stroke, a section of the journal Frontiers in Neurology*

> *Received: 21 August 2017 Accepted: 02 November 2017 Published: 16 November 2017*

#### *Citation:*

*Giraud A, Guiraut C, Chevin M, Chabrier S and Sébire G (2017) Role of Perinatal Inflammation in Neonatal Arterial Ischemic Stroke. Front. Neurol. 8:612. doi: 10.3389/fneur.2017.00612*

Keywords: NAIS, risk factors, physiopathology, chorioamnionitis, vasculitis, immunothrombosis, treatment, neuroprotection

#### INTRODUCTION

Neonatal arterial ischemic stroke (NAIS) is defined by a symptomatic arterial ischemic stroke occurring in the neonatal period, i.e., the first 28 days of life (1). NAIS has to be differentiated from the other subtypes of perinatal stroke, namely cerebral sinovenous thrombosis, neonatal hemorrhagic stroke, and arterial presumed perinatal ischemic stroke (2). Each of these entities differs from NAIS on several aspects, either: (i) affecting different vessel types each of them, such as vein, artery or capillary, featured by specific biological aspects as well as responses to stresses (3), (ii) occurring within distinct—even if somehow overlapping—developmental time frames across gestational and postnatal ages (2), (iii) having different clinical and imaging presentations and outcomes (2), or (iv) being consequently associated with distinct sets of risk factors and causes.

Neonatal arterial ischemic stroke is one of the commonest forms of pediatric stroke, causing a heavy burden of life-long motor, cognitive, and/or behavioral disabilities (2). The pathophysiology of NAIS remains largely unknown (2, 4); hence, there is no evidence-based preventive or curative vasculo- or neuroprotective strategy available for patients affected by NAIS (5).

**Abbreviations:** AI, autoimmune; Apaf-1, apoptotic protease activating factor-1; AS, Apgar score; BBB, blood–brain barrier; CCL, C-C chemokine ligand; CD, cluster of differentiation; CI, confidence interval; CRP, C-reactive protein; CXCL, C-X-C chemokine ligand; Cyt-c, cytochrome-c; DAMP, damage-associated molecular patterns; Drp1, dynamin-related protein 1; FADD, Fas-associated death domain; Fas-L, Fas ligand; FHR, fetal heart rate; FIRS, fetal inflammatory response syndrome; GA, gestational age; HI, hypoxia–ischemia; HT, hypothermia therapy; ICAM, intracellular adhesion molecule; IL, interleukin; IL-1Ra, interleukin-1 receptor antagonist; IL-1RacP, interleukin-1 receptor accessory protein; iNOS, inducible nitric oxide synthase; IUGR, intrauterine growth restriction; LPS, lipopolysaccharide; MAPK, mitogen-activated protein kinases; MCA, middle cerebral artery; MLKL, mixed lineage kinase domain-like; vWF, von Willebrand factor.

After having reviewed the literature to assess epidemiological, clinical and fundamental data, we discuss the role of perinatal inflammation in the causal pathways leading to NAIS.

#### EPIDEMIOLOGY OF NAIS

According to the five population-based studies currently available, the range of prevalence of NAIS varies from 6 to 17/100,000 (6–11). This represents one-fourth of all perinatal strokes syndromes (6).

Based on the review of the literature, the risk factors of NAIS as defined from the seven case–control studies dedicated to NAIS – or other studies in which specific data about NAIS could be extracted—have been summarized in the **Table 1** (8, 9, 12–16).

Inflammatory markers are found as risk factors of NAIS in all case–control studies in which they were studied (8,12–14,16) (**Table 1**). Several direct markers of active perinatal infection/ inflammation are independent risk factors of NAIS, namely maternal fever [odds ratio (OR): 4.0–10.2, in the three studies including multivariate analysis (12, 14, 16)], and neonatal infection [OR: 5.8–9.5, in two of these studies (12, 14)]. The only other independent and consistent across studies risk factor of NAIS is *peripartum* asphyxia (12, 16). *Peripartum* asphyxia can either be secondary to infection and subsequent inflammation triggered by systemic exposure to pathogen-associated molecular patterns (PAMPs), or be a powerful inducer of sterile inflammation *via* the systemic or intracerebral release of damage-associated molecular patterns (DAMP) (cf. see Primary Phase of Neonatal Arterial Ischemic Brain Injury) (17).

Materno-fetal and postnatal inflammation is mostly caused by infection. Neonatal bacterial meningitis is classically complicated by arterial ischemic stroke due to focal arteritis (14, 18–22). Histological chorioamnionitis has not been studied as a possible risk factor (8, 9, 12–16). However, it is well possible that most mothers displaying fever suffer of clinical chorioamnionitis. Further investigations of the potential association between chorioamnionitis and NAIS need to be performed.

Few non-infectious/inflammatory features are associated with NAIS. Male sex was found as an independent risk factor in only one of the seven case–control studies (16) (**Table 1**). Genetic prothrombotic risk factors are not associated with NAIS occurrence. The only study which identified thrombophilia from genetic origin as an independent risk factor of neonatal ischemic stroke was based on a heterogeneous cohort of term, late preterm, and early preterm newborns (12). Several other studies assessed the association between constitutive prothrombotic risk factors and NAIS, with contradictory findings (23–25). The one with the most reliable methodology found a similar rate of thrombophilia at 12 months between the NAIS and the control groups (23). All these studies only investigated constitutive/genetic coagulation markers. To our knowledge, no controlled study was performed in close temporal relationship with the NAIS to assess the expected acute activation of prothrombotic factors.

In sum, perinatal infection/inflammation is the only independent risk factor of NAIS consistently reported up to now. Genetic prothrombotic risk factors do not appear to be associated with NAIS occurrence.

### PHYSIOPATHOLOGY OF THE ARTERIAL OCCLUSION LEADING TO NAIS

#### Role of Inflammation in the Disruption of the Cerebral Arterial Blood Flow in NAIS

Given the tight reciprocal activation between inflammatory and coagulation cascades, it is quite possible that inflammation promotes thrombus formation within placental, umbilical cord or other vessels feeding the cerebral blood flow. According to a classic pathophysiological hypothesis of NAIS, such thrombus would then migrate and occlude cerebral arteries leading to embolic stroke (26). This embolic hypothesis is also supported by the preponderant distribution of NAIS in the middle cerebral arterial territories, and in few instances by the detection of thrombotic/embolic events proximal or distal to the NAIS (27).

However, this embolic hypothesis is challenged by: (i) the imbalanced distribution of NAIS between the anterior *versus* posterior intracranial arterial territories even when the asymmetry of anterior *versus* posterior blood flows is taken into account (13, 28, 29); (ii) the infrequent occurrence of extracerebral infarcts concomitant to NAIS (13, 28, 29); and (iii) angiographic findings from newborns with NAIS showing that 22–65% of them present focal disruptions of the anterior circulation which, in certain cases might correspond to arterial wall defects, or to thrombi generated *in situ* from an inflamed arterial wall (28, 29). Based on these elements, we hypothesized that maternofetal inflammation induces a focal arteritis specifically affecting NAIS susceptible cerebral arteries, namely the middle cerebral artery (MCA), anterior carotid artery and intracranial internal carotid artery (13, 28, 29).

Using a preclinical rat model of chorioamnionitis induced by pathogen components [lipopolysaccharide (LPS) from *Escherichia coli*], we showed that end-gestational inflammation combined with a classic prothrombotic stress (transcutaneous laser exposure of the artery of interest), but not sole prothrombotic stress, targeting the MCA triggers NAIS (30). On one hand, the walls from neonatal cerebral arteries susceptible to stroke displayed a constitutively higher expression of proinflammatory cytokines in NAIS-susceptible *versus* non-susceptible arteries. On the other hand, pups from LPS-exposed dams presented a cerebral arteritis characterized by an increased number of inflammatory cells and expression of proinflammatory cytokines [interleukin (IL)-1/IL-1 receptor antagonist (IL-1Ra) ratio] within NAIS-susceptible, but not non-susceptible, arteries (30).

These preclinical results support, beside the embolic hypothesis, the contribution of a focal arteritis and thrombosis in the pathophysiology of NAIS.

#### Links between Perinatal Inflammation and Thrombosis within NAIS Susceptible Arteries

#### Established Mechanistic Links between Systemic Inflammation and Thrombosis

Inflammation has been well characterized as a potent procoagulant phenomenon.

#### TABLE 1 | Identified risk factors of NAIS.


*Summary of the seven case–control studies published between 1997 and 2016.*

*a Hematological diseases: child sickle cell disease (OR 11.4; 95% CI 3.3–39.4), child sickle cell trait (OR 6.9; 95% CI 1.7–28.1), and child constitutive thrombophilia (OR 413.2; 95% CI 111.1–* > *1,000).*

*AI, autoimmune; AS, Apgar score; CI, confidence interval; ECS, emergency C-section; FHR, fetal heart rate; GA, gestational age; IUGR, intrauterine growth restriction; ND, not done; OR, odds ratio; P, prospective; R, retrospective; Ref, reference; UC, umbilical cord.*

*Inflammatory risk factors are in bold type.*

The C-reactive protein (CRP) and proinflammatory cytokines, including IL-1β, IL-6, and tumor necrosis factor (TNF)-α, are well-established mediators which are implicated in the materno-fetal inflammation, combining maternal immune activation and/or fetal inflammatory response syndrome (FIRS) (31–33) (**Figure 1**). CRP is known to increase the tissue factor (TF) activity *in vivo* (34). Exposure of endothelial cells to proinflammatory cytokines, such as TNF-α, induces the endothelium activation, its production of TF and release of von Willebrand factor (vWF)-propeptide, which are interacting key

FIGURE 1 | Crosstalk between inflammation and thrombosis. C-reactive protein (CRP) and proinflammatory cytokines [interleukin (IL-β), IL-6, tumor necrosis factor (TNF)-α] are mediators implicated in the materno-fetal inflammation (31–33). CRP increases tissue factor (TF) activity *in vivo* (34). TNF-α induces the endothelium activation: the production of TF and release of von Willebrand factor (vWF)-propeptide (35, 36). Microparticles (MPs), released from platelets, macrophages, and endothelial cells upon activation (37), trigger coagulation cascades *via* TF and vWF binding sites (37). Activated mononuclear cells attract and activate platelets *via* TF inducible expression (36–38). Monocyte activation leads to TF-dependent thrombin generation and activation of coagulation (38, 39). Neutrophil extracellular traps (NET) regulate both inflammation and coagulation (40). Proinflammatory cytokines and activated monocyte/macrophages are present within the wall of NAIS-susceptible arteries (30). Glycosaminoglycan synthesis and anticoagulant activity is decreased under inflammation through TF pathway inhibitor and antithrombin interactions with their serine proteinases is impaired (37, 39, 41). Antithrombin activity is downregulated due to its consumption to counteract the mononuclear cells activation and thrombin generation (37). Thrombomodulin is downregulated upon TNF-α exposure (37, 39). IL-1β and TNF-α contribute to vasoconstriction *via* the upregulation of endothelin-1 (42–44). Endothelin-1 increases superoxide anion production, cytokine release (45), and induces prothrombotic effect close to the endothelin-1-induced vasospasm. The *ductus arteriosus* closure is triggered by a vasospasm caused by the decreased plasma concentrations in PGE2 and the increased O2 tension (46, 47). This mechanism could also happen in NAIS-susceptible fetal cerebral arteries. The thrombotic process is associated with the recruitment at the coagulation site of innate immune cells, this is the immunothrombosis. Thrombin induces the expression of proinflammatory cytokines and chemokines by the endothelial cells (37). Platelets are also involved in the trapping and clearing of bacterial agents (40, 48, 49). Activated platelets support neutrophils and monocytes recruitment (e.g., through CXCL1, CXCL4, CXCL5, CCL3, CCL5, CCL7 expression), adhesion (e.g., platelet P-selectin, CD40 ligand expression), and activation [e.g., receptor expressed on myeloid cell (TREM)-1 induced proinflammatory activity] (36, 37, 40).

determinants of platelet activation and aggregation (35, 36) (**Figure 1**). Microparticles (MPs) are known to: (i) be produced by a large variety of activated cells, and (ii) be released from platelets, macrophages and endothelial cells during bacterial infection (37). MPs carry on their surface a range of molecules implicated in triggering coagulation cascades *via* TF and vWF binding sites (37). Activated mononuclear cells (upon bacterial infection through TNF-α or IL-6 exposure) are able to attract and activate platelets *via* TF inducible expression (36–38) (**Figure 1**). Cytokine-mediated monocyte activation leads to TF-dependent thrombin generation and activation of coagulation (38, 39). Some structures, such as neutrophil extracellular traps (NET) are involved in the regulation of both inflammation and coagulation (40). Such cellular and molecular processes bridge the activation of coagulation with data from a preclinical model of NAIS showing that proinflammatory cytokines and activated monocyte/macrophages are present within the wall of NAIS-susceptible arteries from pups exposed to perinatal inflammation (30). Under inflammation, procoagulant proteins are not the only ones to be affected: natural anticoagulant mechanisms are also downregulated. For instance, the glycosaminoglycan synthesis is downregulated on the endothelial surface under inflammation and its anticoagulant activity through TF pathway inhibitor and antithrombin interactions with their serine proteinases is impaired (37, 39, 41). Antithrombin activity is downregulated due to its consumption to counteract the mononuclear cells activation and thrombin generation (37). Another anticoagulant protein, thrombomodulin, which acts by inhibiting the thrombin procoagulant activity, has been shown to be downregulated upon TNF-α exposure (37, 39).

Altogether, clinical data and preclinical modeling show that inflammatory processes happening within placenta, systemic circulation, and wall of NAIS-susceptible arteries are part of the causal pathway of NAIS.

#### Contribution of Systemic Inflammation to Vasoconstriction

Proinflammatory cytokines such as IL-1β and TNF-α have been shown, using endothelial cell culture and MCA occlusion models, to contribute to vasoconstriction *via* the upregulation of endothelin-1 (42–44). Endothelin-1 plasma levels are increased during endotoxemia (45). Endothelin-1 is also known to be able to increase superoxide anion production, cytokine release (45), and to induce subsequent prothrombotic effects adjacent to the endothelin-1-induced vasospasm. According to this mechanism, vasoconstriction might be a key factor in NAIS pathophysiology.

Only few hours after birth, the *ductus arteriosus* remodeling occurs. This remodeling is provoked by the decreased newborn's plasma concentrations in prostaglandin E2 (PGE2; placental production) and the increased dioxygen (O2) tension. This combination triggers a vasoconstriction (46, 47) and platelets are recruited during this process (50). These two mechanisms could act similarly on the fetal cerebral arteries and trigger a focal vasospasm (**Figure 1**).

Premature babies are more often affected by chorioamnionitis than term babies; however, they are less affected by NAIS (51, 52). This observation might be seen as challenging the inflammatory hypothesis of NAIS, which happens mostly in term babies. However, the inflammatory response of preterm newborns is immature and weaker compared to the one from term newborns (53, 54). In addition, premature babies have immature baroreflex which might prevent the occurrence of vasospasm and subsequent ischemia (52, 55–57). These elements might contribute to prevent the occurrence of NAIS in preemies.

#### Contribution of Thrombosis to Inflammation

Inflammation and coagulation are tightly interrelated processes in bidirectional ways. For instance, thrombin, which is a potent activator of the coagulation cascade, is also able to induce the expression of proinflammatory cytokines and chemokines by the endothelial cells (37). Moreover, platelets are not only involved in coagulation in order to protect the host from arterial wall damage but also involved in the trapping and clearing of bacterial agents [e.g., *E. coli* binding through Toll-like receptor (TLR) 4 and cluster of differentiation (CD) 62 (48)] (40, 49). Consequently, the thrombotic process is associated with the recruitment at the coagulation site of innate immune cells, the so-called immunothrombotic process. Activated platelets have an important role in immunothrombosis; they have been shown to support neutrophil and monocyte recruitment [e.g., through chemokines C-X-C chemokine ligand (CXCL) 1, CXCL4, CXCL5, C-C chemokine ligand (CCL) 3, CCL5, CCL7 expression], adhesion (e.g., platelet P-selectin, CD40 ligand expression), and activation (e.g., triggering receptor expressed on myeloid cells (TREM)-1 induced proinflammatory activity) (36, 37, 40). These immune cells could induce a deleterious inflammation within the wall of a thrombosed artery and trigger further occlusion and thrombosis as observed in disseminated intravascular coagulation subsequent to sepsis (36, 37, 39, 40).

Hence, immunothrombotic processes occurring within the lumen and the resulting inflammation of the adjacent wall of NAIS-susceptible arteries might be part of a vicious circle driving the pathophysiology of NAIS, even in the frame of a primary embolic etiopathogenic process triggering arterial wall inflammation (40).

#### INFLAMMATORY PATHWAYS INVOLVED IN BRAIN ISCHEMIC INJURIES OF THE TERM NEWBORN

Based on the above clinical and preclinical findings, NAIS results from a multiple hit mechanism combining perinatal inflammation and hypoxia–ischemia (HI). In sharp contrast to pre-natal HI or postnatal inflammation happening at least 24 h before HI, which have been shown in preclinical models to be neuroprotective (the so-called preconditioning effect) (58–60), prenatal inflammation sensitizes the brain to immediately postnatal HI injuries (61–63). In newborns, such pre-natal inflammatory exposure (e.g., due to chorioamnionitis) might be involved in aggravating the HI injury due to NAIS. The inflammatory pathways involved in HI- and infection/inflammation plus HI-induced NAIS are summarized below.

### Primary Phase of Neonatal Arterial Ischemic Brain Injury

The neural cell stress associated with NAIS is mainly due to the combination of energy failure, excess of intracellular Ca2<sup>+</sup>, and glutamate release, as well as ionic imbalance and oxidative stress (64–66). All these pathways are key mechanisms driving neural cell death (64, 66–68). We learned from a preclinical model of NAIS, that excitotoxic cell death, necrosis and programmed necrosis (necroptosis) occur between 0 and 6 h after exposures to sole HI or pathogen components plus HI (**Figure 2A**) (66, 69–72).

Pattern recognition receptors (PRR) such as TLR recognize pathogens as well as DAMP (73, 74, 90–94). Inflammation sensitizes the neonatal brain to subsequent HI injury (54, 73, 74, 95). For instance, Stridh et al. showed in a newborn mouse model [at postnatal day (P) 9, i.e., a level of brain development equivalent to the term human newborn] of NAIS that TLR-2 deficiency protected the brain from infarcts (94). Energy failure due to hypoxia and inflammation combine their effects to increase the oxidative stress (96, 97). Besides, LPS and other pathogen components interact with various TLR to increase the synthesis of a wide set of proinflammatory cytokines and chemokines (**Figure 2B**) (54, 73, 74, 95). For instance, in a preclinical model of NAIS, exposure to LPS plus HI leads to an autocrine/paracrine loop of neuronal self-injury, mediated by inflammatory molecules: IL-1β, TNF-α, reactive oxygen species (ROS) production, and mitogen-activated protein kinases (MAPK)-induced apoptosis (**Figure 2B**) (73, 74). LPS combined with HI leads to glial activation and neurotoxic

FIGURE 2 | Phases of injury occurring in neonatal arterial ischemic stroke (NAIS) and mechanistic pathways. (A) The first phase of injury in NAIS occurs between 0 and 6 h after the exposition to hypoxia–ischemia (HI) alone or infection/inflammation plus HI. This phase is characterized by different cell death types, including excitatory cell death, necrosis, and necroptosis. These primary cell deaths will induce several inflammatory cascades. Exposure to lipopolysaccharide (LPS) + HI releases DAMPs within neurons leading to an overexpression of IL-1β through inflammasome activation, which also leads to nuclear factor-κB (NFκB)-induced tumor necrosis factor (TNF)-α synthesis (73, 74). This will further result in the activation of the glial cells and the increase of the inflammation through the release of reactive oxygen species (ROS) and several inflammatory molecules by these cells. The secondary phase occurs between 24 and 72 h after NAIS and includes apoptosis, anoikis, and autophagy cell deaths. Overall, this will induce the activation of the endothelium of the brain vessels and can lead to the rupture of the blood–brain barrier (BBB) and the infiltration of leukocytes within the brain. (B) Cell death and inflammatory pathways at play within a neuron in the injured brain. Extrinsic apoptosis is induced by inflammatory molecules, such as Fas ligand (Fas-L) and TNF-α and further activation of their respective receptors FAS and TNFR-1. This leads to the recruitment of caspase-8. The activation of caspase-8 induces the recruitment of executioner caspases and subsequent cell death by apoptosis (75, 76). Activated caspase-8 negatively regulates necroptosis signaling by cleaving receptor interacting protein kinase (RIP)-1 (69, 77). In intrinsic apoptosis, caspase-8 can recruit and activate proapoptotic proteins, including Bax and Bak through the activation of t-Bid. The excess of proapoptotic protein as compared to antiapoptotic protein (Bcl2, Bcl-xL) results in an opening of the mitochondrial permeability transition pore, and the release of cytochrome-c (Cyt-c) into the cytoplasm. This leads to the apoptosome formation with the recruitment of apoptotic protease activating factor-1 (Apaf-1) and caspase-9, and the induction of the cell death by apoptosis. Necroptosis is induced by different signaling pathways, including Fas-L–Fas, TNF-α–TNFR-1, and LPS–TLR-4. This will induce the dimerization of RIP-1 and RIP-3, thus inducing the phosphorylation of RIP-3 (78). In the TLR-4-induced necroptosis, RIP-3 and MLKL are activated, but without RIP-1. Instead of RIP-1, TIR-domain-containing adapter-inducing interferon-β (TRIF) will associate with RIP-3 and subsequently induce necroptosis (79, 80). Upon RIP-3 activation, MLKL is recruited, phosphorylated, and translocated to the plasma membrane to initiate cell death through the disruption of the membrane integrity (80). Phosphorylated MLKL can also interact with the mitochondrial phosphatase PGAM5 and further induced ROS expression, and may activate dynamin-related protein 1 (Drp1) that could ultimately lead to cell death through mitochondrial fission (75, 80, 81). The release of mitochondrial ROS within the cytoplasm of the neuron, as well as the DAMPs can induce the activation of the inflammasome (82, 83). The activation of caspase-1 will induce the cleavage of the pro-IL-1β into IL-1β. The inflammation will pursue with the autocrine-paracrine loop of IL-1β activation and the activation of transcription factors, such as NFκB and P38/MAPK (83). Potential blocking agents (⊥) are as following: IL-1 blockers (e.g., IL-1Ra) (74), necrostatin-1 for RIP-1 (67), GSK'872 for RIP-3 (84), necrosulfonamide for MLKL (85), caspase inhibitors for apoptosis (86–88), and apocynin for nicotinamide adenine dinucleotide phosphate (NADPH) oxidase targeting (89). Color codes: black: cell death; red: inflammation; purple: vascularization; blue: blockades.

molecule release such as IL-1β-induced matrix metalloproteinase (MMP)-9, nitric oxide (NO), and inducible NO synthase (iNOS) (66, 73). In these LPS plus HI-exposed brains, NO and MMP-9 combine their effects to alter the blood brain barrier (BBB) by degrading the lamina of intracerebral blood vessels (67, 74). Such disruption enables proinflammatory and/or neurotoxic mediators to leak through the BBB, and thus increasing the extent of NAIS in preclinical models (74, 98) (**Figure 2A**).

Necroptosis is an early cell death pathway which is triggered by inflammatory mediators led by TNF-α and TNF family death receptor (TNFR): TNFR-1, FAS, and TLR, namely TLR-3 and TLR-4 (75, 79, 84, 99). The necrosome is a complex that requires the presence of activated receptor interacting protein kinase (RIP)-1, RIP-3 and the pseudo-kinase mixed lineage kinase domain-like (MLKL) to execute necroptosis (75, 99) (**Figure 2B**). All the mechanisms by which MLKL induces cell death are not totally elucidated (75, 81, 99). It was shown in a preclinical model that cerebral expression of TNF-α was triggered by LPS plus HI exposure (73). Accordingly, in this NAIS model RIP-3 expression was increased after LPS plus HI (74). IL-1β-induced MMP-9 can also activate necroptosis *via* Fas-ligand (Fas-L) interaction through Fas-associated death domain (FADD) (70, 96, 99). In line with these findings, it was shown that necrostatin-1—a RIP-1 inhibitor—administered after HI injury in P7 mice was neuroprotective: this inhibitor is able to prevent forebrain injury, as well as attenuates oxidative stress and mitochondrial dysfunction (69, 71). However, since RIP-1 is an important crosstalk molecule between apoptosis and necroptosis pathway, it has been observed in neonatal HI model that RIP-1 blockade increased apoptotic cell death (69).

## Secondary Phase of Neonatal Arterial Ischemic Brain Injury

The secondary phase of ischemic injury occurs between 24 and 72 h after NAIS and implicates apoptosis, anoikis and autophagy cell deaths. Intrinsic and extrinsic forms of apoptosis are known to be involved in NAIS due to pure HI or HI combined to pathogen exposure (67, 68, 72) (**Figure 2B**). Many preclinical studies characterized the involvement of apoptotic cell death in the genesis of NAIS and provided evidence in favor of neuroprotective strategies targeting apoptotic pathways (86–88, 100–102).

It is well known that apoptotic and autophagic cell death pathways crosstalk, and that autophagy can block apoptosis by sequestration of mitochondria (68) (**Figure 2B**). The induction of autophagy just after neonatal HI may be a neuroprotective mechanism by limiting apoptosis (68, 103). On the other hand, autophagy seems to be implicated in HI-induced cell death (104). Besides, another form of apoptosis triggered by cell detachment from the extracellular matrix—namely anoikis—could follow HI and/or inflammation exposures. Anoikis is induced by increased MMPs, including MMP-9 and activation of Fas receptor, which initiates the apoptosis cascade (105). This cell death pathway was assessed by our group: we showed an overexpression of MMP9 after HI plus inflammation injury (73, 74). Besides, the use of an MMP-9 competitive inhibitor shrank the size of LPS plus HI-induced brain infarcts (73, 74) (**Figure 2A**). To our knowledge, this is the first demonstration of anoikis-induced cell death in a model of combined inflammation and/or HI.

### NEW HYPOTHESES BRINGING NEW TREATMENTS

Given that the diagnosis of NAIS is often delayed due, in most cases, to the prenatal onset and/or to the absence, or paucity and diagnostic delays, of neonatal symptoms, future therapies should focus on the control of preinsult determinants most often acting prenatally—e.g., through anti-inflammatory intervention, such as IL-1Ra, or on postinsult neonatal mechanisms—e.g., through hypothermia therapy (HT)—rather than on less feasible perinsult acute interventions.

### IL-1 Blockade

Our team and others uncovered that the upregulation of IL-1 plays a key role in chorioamnionitis, and in associated neonatal ischemic brain injuries (54, 79, 82, 84, 94–97). Our preclinical studies, and others, showed that prenatal IL-1 blockade using IL-1Ra is protective against chorioamnionitis, associated FIRS, and subsequent brain injuries (73, 74, 106–108). Postnatal administration of IL-1Ra is also effective in alleviating mortality (from 40 to 18%) as well as morbidities arising from postnatal inflammatory-sensitized NAIS (82, 92, 94): 66% decrease of the core (cavitary lesion), and 54% decrease of the penumbra (rim of mild to moderately ischemic tissue lying between the core and the unaffected tissue), and preventing the loss of motor skills (73, 74). IL-1Ra is an already approved drug to treat chronic inflammatory conditions, including those affecting pregnant mothers and newborns (96). IL-1Ra is, among the various molecules interfering with the IL-1 signaling, the one which dominates its pharmacological field due to its: (i) blocking effect on both IL-1α and IL-1β; (ii) short 4–6 h half-life (blood levels falling within a few hours of treatment stoppage); (iii) multiple routes of administration; (iv) approval for several pediatric inflammatory conditions (1–10 mg/kg/24 h), knowing that the repurposing of well-studied drugs used in the pediatric population is a cost-effective and efficient strategy to identify new therapies for pediatric diseases; and (v) excellent safety record (absence of opportunistic infection; reversible increase of liver enzyme, decrease of polymorphonuclear cells, slight increase of infection, that are mostly observed in patients on chronic treatment) after more than 10 years of use in more than 150,000 patients (80). Altogether, this provides encouraging preclinical evidence in favor of the efficacy and feasibility of end-gestational or neonatal interventions using IL-1Ra.

### Hypothermia Therapy

Hypothermia therapy is now a mandatory standard of care for term newborns suffering for diffuse HI encephalopathy (109–111). However, cooling treatment is modestly effective and leaves 50% of the treated patients with major sequelae (109, 112). Besides, it is uncertain why HT is effective for some, but not all, human newborns. Clinical studies reported that HT might have less beneficial effects on newborns exposed to infection-inflammation plus HI, than those exposed to HI alone (113–115). Furthermore, evidence in favor of an antiinflammatory role of HT within the newborn brain is limited and conflicted. Only a few clinical or preclinical models address this question. However, a well-established anti-inflammatory effect of HT is the down regulation of oxidative stress within the brain (91, 97, 116–118). It has also been reported that HT is neuroprotective by limiting apoptotic cascades in human term newborns (110, 119–121). The potential effect of HT on neuroinflammatory cytokines expression has been poorly investigated up to now in preclinical models as well as in term newborns. It has been recently shown that HT did not modulate inflammatory molecules, including IL-1β, TNF-α, IL-1Ra and MMP-9, on LPS plus HI-exposed pups (118). Other downregulating effect of HT within preclinical NAIS brains remained unclear. In the clinical settings, HT has not been tested yet in NAIS patients, even though it is feasible and possibly effective.

### Erythropoietin

Erythropoietin presents anti-inflammatory and neuroprotective properties mainly through dampening free radical release and neural cells apoptosis that have been well-established on preclinical models of neonatal brain infarcts (122, 123). Erythropoietin seems to be well-tolerated and neuroprotective against perinatal brain lesions of premature newborns: Benders et al. performed a study in 21 consecutive NAIS patients diagnosed by magnetic resonance imaging (MRI) using erythropoietin (1,000 IU/kg intravenously administered just after the diagnostic confirmation by MRI, and repeated at 24 and 48 h) (124).

There was no adverse effect on blood cells counts, or coagulation. The residual *versus* initial MRI injuries were quantitatively compared at 3 months of age between the erythropoietin-treated patients *versus* 10 untreated matched historical controls. The percentage of tissue loss within the ischemic area was not different between the treated *versus* untreated group. Hence, the effectiveness of erythropoietin in terms of neuroprotection in the NAIS context remains to be established.

#### WHICH CLINICAL RESEARCH PRIORITIES FOR NAIS?

### Pre-NAIS Neuroinflammatory-Oriented Research Avenues

Before considering human therapeutic trials aiming to prevent the occurrence of NAIS, it is mandatory to improve our ability to rapidly and efficiently detect prenatal, and immediately postnatal inflammation, and to correlate inflammatory profiles with the occurrence of NAIS. Hence, large scale case–control studies are a necessary prerequisite to further identify subpopulation(s) of newborns at risk of NAIS (28), by comparing pertinent profiles of expression of inflammatory and prothrombotic markers—reflecting acute activation instead of genetic predisposition—in the umbilical cord tissues and cord blood samples, between NAIS *versus* non NAIS patients. In addition, there is an urgent need to design efficient diagnostic tools adapted to the rapid and non-invasive diagnosis of chorioamnionitis and other maternofetal infectious/inflammatory diseases. In this line, MRI of the placenta, which has been shown to rapidly detect

### REFERENCES


abnormal placental signals in a preclinical model of chorioamnionitis might be a promising diagnostic option (125). Based on these potential predictive biomarkers of NAIS, preventive anti-inflammatory strategies might then be tested.

### Post-NAIS Neuroinflammatory-Oriented Research Avenues

A low threshold for early head MRI should be applied in newborns presenting acute neurological symptoms compatible with neonatal stroke. Adapted and thorough protocol of parenchymal and angiographic imaging should be continuously updated and more systematically applied (126), as recently recommended (127). The acute activation of inflammatory and prothrombotic factors at the time of the NAIS is an important research avenue to improved our physiopathological knowledge, and provide much needed biomarkers of NAIS. This might be performed first in retrospect using immunoassay on dried neonatal blood, as previously described (128). IL-1 blockers, erythropoietin, and/or hypothermia seem to be the most promising avenues to be tested in multicentric randomized control studies in the aim to limit the extent of NAIS and/or to promote recovery. Some teams already carried out retrospective and prospective phase I–II studies addressing such therapeutic approaches (124, 129, 130).

## AUTHOR CONTRIBUTIONS

SC and GS contributed to the design of the manuscript. AG, CG, MC, SC, and GS drafted the manuscript. All the authors contributed to the editing of the manuscript.

## ACKNOWLEDGMENTS

We thank Noha Gerges for her help in the final revision of the manuscript.

## FUNDING

This work was funded by the Heart and Stroke Foundation Canada (grant number: G-14-0005756), the *Agence Régionale de Santé Auvergne-Rhône-Alpes*, the *Faculté de Médecine des Sciences de la Santé de l*'*Université de Sherbrooke*, the Integrated Program in Neuroscience, the Hoppenheim/Montreal Children Hospital Foundation, the Research Institute of McGill University Health Center, and the Foundation of Stars.


International Pediatric Stroke Study. *Pediatrics* (2011) 128:e1402–10. doi:10.1542/peds.2011-1148


in patients with hemolytic uremic syndrome. *Blood* (2006) 108:167–76. doi:10.1182/blood-2005-08-3219


following neonatal hypoxia-ischemia. *Neuroscience* (2012) 219:192–203. doi:10.1016/j.neuroscience.2012.05.002


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

*Copyright © 2017 Giraud, Guiraut, Chevin, Chabrier and Sébire. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Modulating the Oxytocin System During the Perinatal Period: A new Strategy for neuroprotection of the immature Brain?

*Manuela Zinni1 , Marina Colella1 , Aline Rideau Batista Novais1,2, Olivier Baud1,3,4† and Jérôme Mairesse1,3\*†*

*<sup>1</sup> INSERM U1141 Protect, Paris-Diderot University, Paris, France, 2Neonatal Intensive Care Unit, Robert Debré Children's Hospital, Paris, France, 3University of Geneva, Geneva, Switzerland, 4Division of Neonatology, Geneva Children's Hospital, Geneva, Switzerland*

Oxytocin is a neurohypophysal hormone known for its activity during labor and its role in lactation. However, the function of oxytocin (OTX) goes far beyond the peripheral regulation of reproduction, and the central effects of OTX have been extensively investigated, since it has been recognized to influence the learning and memory processes. OTX has also prominent effects on social behavior, anxiety, and autism. Interaction between glucocorticoids, OTX, and maternal behavior may have long-term effects on the developmental program of the developing brain subjected to adverse events during pre and perinatal periods. OTX treatment in humans improves many aspects of social cognition and behavior. Its effects on the hypothalamic–pituitary–adrenal axis and inflammation appear to be of interest in neonates because these properties may confer benefits when the perinatal brain has been subjected to injury. Indeed, early life inflammation and abnormal adrenal response to stress have been associated with an abnormal white matter development. Recent investigations demonstrated that OTX is involved in the modulation of microglial reactivity in the developing brain. This review recapitulates stateof-the art data supporting the hypothesis that the OTX system could be considered as an innovative candidate for neuroprotection, especially in the immature brain.

Keywords: intra-uterine growth restriction, neuro-inflammation, white matter brain injury, oxytocin, microglia, glucocorticosteroid, GABA, maternal behavior

### WHITE MATTER INJURY (WMI) FOLLOWING FETAL GROWTH RESTRICTION

Intrauterine growth restriction (IUGR) is a complication observed in 10% of the pregnancies (1) and represents the major causes of neonatal mortality and morbidity (2). Placental insufficiency resulting in fetal hypoxia and maternal malnutrition are two identifiable and major causes of IUGR (3). Due to its constant increase in both industrialized and developing countries, where 2.8 million children out of 135 million born in 2010 were born preterm and growth restricted (4), IUGR represents an important public health problem. Indeed, growth-restricted infants showed a higher risk of perinatal morbidity and of neurodevelopmental alteration with long-term cognitive and neurobehavioral handicaps (5, 6). Interestingly, studies based on magnetic resonance imaging have clearly evidenced that the cognitive and psychiatric deficits observed (7–9) are correlated to alterations of brain white and gray matter (7, 10, 11), including altered neural circuitry (12, 13). The importance of IUGR in the context of public health economy is further highlighted by the presence of a positive correlation between the

#### *Edited by:*

*Carl E. Stafstrom, Johns Hopkins Medicine, United States*

#### *Reviewed by:*

*Yehezkel Ben-Ari, Neurochlore, France Akira Monji, Saga University, Japan*

*\*Correspondence: Jérôme Mairesse j\_mairesse@hotmail.com*

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

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 24 January 2018 Accepted: 23 March 2018 Published: 13 April 2018*

#### *Citation:*

*Zinni M, Colella M, Batista Novais AR, Baud O and Mairesse J (2018) Modulating the Oxytocin System During the Perinatal Period: A New Strategy for Neuroprotection of the Immature Brain? Front. Neurol. 9:229. doi: 10.3389/fneur.2018.00229*

**21**

severity of IUGR and the risk to develop cerebral palsy (14), risk that is 10- to 30-fold higher in IUGR babies (15–18).

The mechanisms responsible for the induction of brain injury in preterm infants remain largely elusive and several putative inductor factors have been identified (oxidative stress, excitotoxicity, neuroinflammation) (19). In the hereinafter of this section, we will focus on the role of neuroinflammation and the possible cellular mechanisms responsible for inflammatory-induced brain damage.

Clinical studies showed that abnormal inflammatory responses in the fetus and/or in the neonate can contribute to white matter damage (20, 21). These clinical observations are well supported by studies conducted in rodents in which IUGR is not only associated with an abnormal neuroinflammatory response and myelinization defects (22–24), but is also a risk factor for the development of inflammatory-induced brain damage (25).

The brain inflammatory response is orchestrated by crosstalk between microglia and astrocyte (26). In particular, microglia (brain resident macrophage) colonizes the brain during development in two phases: the fetal development (first two trimesters in humans and between embryonic days (EDs) 10 and 19 in rodents) and the early postnatal days (PND) (27). An accurate regulation of their activation is critical for the development of a proper immune response and for maintaining brain homeostasis. Indeed, as recently demonstrated, abnormal microglia activity can influence cortical neurogenesis (28), neuronal migration, axonal growth (29, 30), and synaptic pruning (31). These events that occur during the fetal and the early postnatal period are critical for the development of a functional brain architecture and their alterations can generate "pre-symptomatic signatures" correlated to the manifestation of neurological disease later in life as suggested by the neuro-archeological hypothesis (32). Abnormal microglia activation can also negatively affect myelinization (27). In order to better understand the relation between microglia and myelinization, it is important to consider the developmental stages of myelinization. The process is defined by initial migration and proliferation of oligodendrocyte precursors followed by their differentiation first into pre-oligodendrocytes (pre-OL) and then into mature oligodendrocytes (33). In particular, pre-OL showed higher intrinsic vulnerability to environmental insults and exposure of the brain to free radical or to excitotoxic molecules dramatically affect their maturation and differentiation (34). Abnormal microglial activation is the third factor affecting pre-OL maturation, a pivotal player in the context of WMI (27, 34) (**Figure 1**). Thus considering this background, the early modulation of the microglia activity could represent a valid therapeutic option for the treatment of brain injury in prematurity and to prevent the printing of the "pre-symptomatic signature" of neurological disease.

### EARLY OVEREXPOSURE TO GLUCOCORTICOSTEROIDS (GCs): IMPACT ON NEUROINFLAMMATION AND OXYTOCIN PRODUCTION

The release of GCs is regulated by the hypothalamic–pituitary– adrenal axis (HPA) (35). HPA axis activation results in the release of corticotropin releasing factor (CRF) from the hypothalamic paraventricular nucleus (PVN) in the portal vessel system, inducing the secretion of adrenocorticotropic hormone (ACTH) from the pituitary that in turn stimulates the release of GCs from the adrenal gland.

Glucocorticosteroids, classically described as anti-inflammatory and immunosuppressive agents, have also displayed proinflammatory actions. Indeed, studies conducted in humans and in rodents showed that chronic exposure to stress or to high levels of GCs potentiate the inflammatory response both at central and peripheral levels (36–38).

The pro-inflammatory effects of GCs are long lasting and early life stress is able to shift the immune response toward a pro-inflammatory phenotype later in life (39–41) with a direct effect on microglia immunoreactivity and maturation (42–45). Concerning the latter point, the study (42) showed that exposure to prenatal stress between ED 10 and 20 affects microglia maturation by inducing a reduction of immature microglia in the corpus callosum and an increase in ramified microglia in other brain regions at PND 1. More recently, two different studies

demonstrated that exposure to maternal separation (MS) (43) or to prenatal stress (44) increase the activation of microglia cells in the hippocampus at PND15 and the number of activated microglia in the hippocampus and in the cortex of adult animals, respectively. In the same study (44) the authors showed, through an *in vitro* approach, that microglia isolated from prenatal stressed animals is more amoeboid and releases higher levels of pro-inflammatory cytokines. In addition, as reported in Ref. (45), exposure to prenatal stress is able to shift the hippocampal microglia morphology toward an activated phenotype not only in basal condition, but also in response to LPS stimulation in adults.

The mechanisms responsible for these effects are not yet well understood, however, an important role could be exerted by nuclear GCs receptors (GRs). GRs regulate the HPA activity by means of negative feedback (46) and the anti-inflammatory effects of GCs are promoted by the formation of a GC/GR complex (47). As previously reviewed, hippocampal GRs undergo epigenetic regulation of their expression that is influenced by early parental care (48). Interestingly, two human studies reported an increase in GR methylation in leukocytes and mononuclear cord blood cells in adults (49) and infants (50) exposed to childhood adversity, respectively. Moreover, a recent study evidenced a relation between GR methylation and inflammation at the central level (51). Rats exposed to MS showed, as adults, a higher methylation of hippocampal GR receptor that is linked to an increase in hippocampal astrocytes inflammatory response following sevoflurane administration. Interestingly, these effects can be reversed by treatment with an epigenetic regulator (51).

Because microglia are the resident immune cells of the brain, we hypothesize that stress or high levels of GCs can induce epigenetic modification of GRs on microglia too. The change in GR expression could, therefore, shift the microglia response toward the pro-inflammatory phenotype observed in premature infants and in animal models of IUGR. In this context, defining strategies to prevent exposure of the developing brain to high pro-inflammatory levels of GCs acquire greater importance.

Oxytocin is a neuropeptide released by the PVN and by the supraoptic nucleus of the hypothalamus. Studies conducted in rodents and in humans showed the existence of a bidirectional relation between the HPA axis and OTX: exposure to stress induced an increase in OTX plasma levels (52–54), while OTX administration counterbalanced axis activation reducing GCs release (55–59). The details of this inhibitory action were clarified *via* pharmacological approaches in several studies. In particular, intra-cerebroventricular administration of OTX induces a reduction in CRF mRNA levels in the PVN in response to stress (55, 57) and a reduction in ACTH and corticosterone plasma levels both in the basal condition (58) and in response to stress (55, 56, 58) (**Figure 2**). This proven ability to modulate the GCs release supports the hypothesis of

a functional interaction between the OTX system, HPA axis, and immune system. In this mutual communication between the three endogenous systems, OTX could exert an indirect anti-inflammatory action through the control of HPA axis activation. Therefore, during the early phase of life, OTX could have an important role to prevent the exposure of the brain to high and pro-inflammatory doses of GCs.

### RELATION BETWEEN OTX AND THE IMMUNE SYSTEM: EVIDENCES OF AN ANTI-INFLAMMATORY EFFECT

Oxytocin known for its role in labor and lactation is generally used in clinical practice for the induction and augmentation of labor (60). However, recent investigations in animals evidenced a pivotal role of OTX in the regulation of a central inflammatory response (61–63). The anti-inflammatory action in the brain was described for the first time in an animal model of brain stroke (MCAO) in combination with social housing and social isolation protocols (61). Social environment is associated with a reduction in incidence, mortality, and morbidity of stroke (61, 64). Housing in a social environment increases the synthesis of OTX mRNA in the hypothalamus and interestingly, this increase mediates the neuroprotective effects of the social environment (61). Indeed, intracerebral administration of an OTX receptor (OTXR) antagonist neutralized the neuroprotective effects of the social environment, whereas the administration of OTX to socially isolated animals before induction of cerebral arterial occlusion improved stroke outcome reducing infarct size, oxidative stress, stroke-induced gliosis, and neuroinflammation (61). In addition, a recent *in vivo* study demonstrated that intracerebral administration of OTX reduced pro-inflammatory gene expression in the hippocampus of adult animals exposed to MS (62).

Studies aimed at clarifying the cellular target of this antiinflammatory action pointed out a role for the OTX system in the regulation of microglia reactivity both *in vivo* and *in vitro* (61, 63). Regarding the *in vivo* evidence, the study (63) demonstrated that intranasal administration of OTX to adult mice reduced microglia activation and pro-inflammatory cytokine expression induced by an LPS injection. In addition, in the same study the authors demonstrated that OXT is able to reduce the LPS-induced activation both in primary microglia and in microglia cell lines (63). Similar results have been reported in microglia cells purified from socially isolated animals and stimulated *in vitro* with LPS (61).

The biological action of OTX is linked to the activation of OTXR, a selective seven transmembrane Gq/Gi-coupled receptor (65) expressed both in astrocytes and in microglia (61, 63). Exposure of microglia cells to inflammatory stimulus induced a time-dependent increase in OTXR expression (63) suggesting that the OTX system is an inducible system that undergoes a dynamic regulation to respond to the requests of an immune challenge. The molecular bases of neuroprotective action of OTX are not well known and modulation of the downstream ERK/MAPK pathway in microglia was reported only in one study (63). In addition, other molecular effectors of OTXR (e.g., NFkB, eukaryotic elongation factor 2) could mediate the observed effect. Finally, because microglia also express receptors for glutamate and other important neurotransmitter (e.g., GABA, Acetylcholine) (66), the existence of a functional crosstalk between OTXR and other neurotransmitter receptors cannot be ruled out.

#### EFFECTS OF OTX ON THE NEONATAL BRAIN AND "GABA SWITCH"

The early phase of life represents a period of maximum plasticity for the brain. Indeed, during this time it undergoes morphological changes that are fundamental for the development of correct excitatory and inhibitory neuronal circuits. An abnormal balance between excitatory and inhibitory transmission have been proposed as a causal factor for the occurrence of neurodevelopmental disorders (e.g., autism) (67) and in this context an interesting role is mediated by GABA (68, 69). GABA, the main inhibitory neurotransmitter in adults, exerts an excitatory effect in the immature brain switching transiently to an inhibitory action during delivery, and permanently during the first postnatal week (70, 71). The peculiar Cl<sup>−</sup> homeostasis that characterizes the immature brain is at the base of GABA excitatory action (67). Indeed, immature neurons express on their membrane high levels of Cl<sup>−</sup> importer NKCC1, and low levels of Cl<sup>−</sup> exporter KCC2 with a consequent increase in Cl<sup>−</sup> intracellular concentration (67). In presence of this ionic gradient, activation of the GABA receptor (GABAAR) induces an efflux of Cl<sup>−</sup> and the consequent generation of an excitatory membrane depolarization (67).

Alteration of GABAergic signaling is reported in several neurodevelopmental diseases, such as autism (69) and Fragile X (68, 72). Therefore, the understanding of the mechanisms underlying the GABA function in the immature brain acquires greater importance.

Oxytocin is a key player for the biphasic transition of GABA and during delivery a main role is exerted by maternal OTX. Parturition is indeed associated with a massive release of OTX (73) that easily crosses the placenta and reaches the fetus (74). Combining the electrophysiological approach with *in vivo* administration of an OTXR antagonist to pregnant rats, Tyzio et al. elegantly demonstrated that maternal OTX is necessary and sufficient to promote GABA switch (70) and that the inhibition of this OTX-mediated transition induces in the offspring an autisticlike phenotype (69). The modulation of NKCC1 activity is at the base of OTX-mediated GABA switch during delivery (70) and, as observed for OTX, the administration of an NKCC1 antagonist to pregnant rats reverts the abnormal electrophysiological phenotype in two animal models of autism (69). Concerning the role of OTX in the postnatal GABA switch, a recent research highlighted the involvement of the KCC2 transporter (67). Indeed, mutant OTXR<sup>−</sup>/<sup>−</sup> mice showed delayed GABA switch associated with reduced KCC2 hippocampal expression. On the contrary, wild type animals showed in the early postnatal period correct GABA transition and an increase in KCC2 expression that are promoted by activation of OTXR and of its downstream pathway Gq/protein kinase C. Interestingly, this OTXR-mediated modulation of KCC2 expression is time dependent and restricts to an early time point (67). This observation further highlighted the pivotal role of OTX in the first phase of life and supports the hypothesis of OTX as a novel neuroprotective agent in the immature brain. Indeed, a precocious treatment of preterm infants with OTX could represent a valid therapeutic strategy to ensure correct brain development and perhaps reduce the risk of developing neurodevelopmental disorders later in life.

### EFFECTS OF OTX ON THE NEONATAL BRAIN AND THE ROLE OF MODULATION OF MATERNAL BEHAVIOR

Oxytocin is an important hormone for the regulation of maternal behavior (75–80) and this interaction was first reported by Pedereson and Fahrbach (75, 76). Indeed, the authors demonstrated that intracerebral administration of OTX to virgin female rats reduced the latency to develop maternal care in response to exposure to forest pups (75, 76). In agreement with these results, intracerebral administration of an OTXR antagonist reduced maternal behavior and canceled the differences between high maternal care (High LG-ABN) and low maternal care (Low LG-ABN) mothers (78). High LG-ABN is associated with a higher level of OTX (77) and of OTXR in the medial preoptic area, a hypothalamic area important for the regulation of maternal care (78, 79). In humans, the increase in plasma OTX between the first and the second trimester of pregnancy is predictive of motherinfant bonding (81) and higher plasmatic and salivary levels of OTX are observed in mothers with high affectionate contact (80).

Maternal attachment is the first form of social interaction and its quality and quantity influence the behavioral and neuroendocrine outcomes of the organisms. Indeed, human studies reported that a low quantity and quality of maternal care are associated with a higher risk to develop adult psychopathy (82) and to worse cognitive performances later in life (83). In agreement with these human results, animal studies clearly demonstrated that rats reared by low LG-ABN showed, as adults, impaired cognitive performances (84, 85) increase in anxiogenic behavior (86) and fearfulness (87). Moreover, low LG-ABN showed hyperactivity of the HPA axis in response to stress (88).

Considering the ability of OTX to modulate the insurgence of maternal care and the positive effects of high levels of maternal behavior, it is possible to suggest a functional interaction between OTX and maternal care. Therefore, OTX could exert an indirect neuroprotective effect through modulation of maternal care.

### CONCLUSION

Intrauterine growth restriction is recognized to be an important public health problem and growth-restricted infants present an increased risk to develop cognitive and behavioral alterations later in life. As evidenced by clinical studies these increased risks are significantly correlated to the development of gray and white matter injury including altered neural circuitry. Preclinical and clinical studies have demonstrated that neuroinflammation, associated with abnormal microglia reactivity, is a causal factor for the development of WMI. Therefore, the modulation of inflammation could represent a valid therapeutic strategy for the treatment of brain injury in preterm infants. In this context, the neuropeptide OTX can exert a pivotal role due to its ability to modulate the immune system and shift its activity toward an anti-inflammatory phenotype. The studies discussed in this review demonstrate that this anti-inflammatory effect is exerted through the regulation of microglia activation. However, the beneficial effects of OTX are not only related to the modulation of neuroinflammation, but also to the development of correct neural circuitry. Indeed, its action is necessary to regulate the "GABA switch" and the proper balance between excitatory and inhibitory transmission whose alterations have been linked to the occurrence of neurodevelopmental disorders. Finally, the positive effects of OTX can not only be confined to a direct action on the immature brain. Indeed, OTX is an important regulator of maternal behavior and alterations of maternal care are correlated to the insurgence of behavioral and neuroendocrine alterations later in life. Thus, it is possible to suppose that the modulation of maternal care is one of the mechanisms at the base of OTXmediated neuroprotection. In conclusion, the data summarized here supports the hypothesis of OTX as a potential neuroprotective agent in the developing brain.

## METHODS

The present review summarizes clinical and preclinical data about causal relations between inflammation and neonatal brain injury, and recapitulates experimental evidences hypothesizing OTX as a novel anti-inflammatory and neuroprotective agent in the immature brain. A literature search was performed in December 2017–January 2018 using the PubMed library in English. No restriction of year and authors were applied and review papers were used as references only for the general concepts. The literature search relating to the pre-clinical studies was restricted to research conducted in rats and mice. Only papers that satisfied the following criteria were included: pertinence to the subject, presence of control groups, and clear descriptions of experimental procedures.

## AUTHOR CONTRIBUTIONS

MZ and JM did the literature review. All the authors collectively analyzed articles selected in this review paper. MZ, JM, and OB wrote the manuscript. All the authors revised and approved final version of the manuscript.

### ACKNOWLEDGMENTS

We thank Audrey Toulotte-Aebi for editing the manuscript draft.

## FUNDING

This study was supported by INSERM, "Agence Nationale de la Recherche" (ANR), and by "La Fondation Paralysie Cérébrale," France.

### REFERENCES


immune system in adult rats. *Psychoneuroendocrinology* (2007) 32(2):114–24. doi:10.1016/j.psyneuen.2006.11.005


axis in male and female rats: partial action within the paraventricular nucleus. *J Neuroendocrinol* (2001) 12(3):235–43. doi:10.1046/j.1365-2826.2000.00442.x


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

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

# Fetal neuroprotection by Magnesium Sulfate: From Translational Research to Clinical Application

*Clément Chollat1,2\*, Loïc Sentilhes3 and Stéphane Marret1,4*

*<sup>1</sup> INSERM U1245, Team 4 Neovasc, School of Medicine of Rouen, Institute of Innovation and Biomedical Research, Normandie University, Rouen, France, 2Department of Neonatal Intensive Care, Port-Royal University Hospital, APHP, Paris, France, 3Department of Obstetrics and Gynecology, Bordeaux University Hospital, Bordeaux, France, 4Department of Neonatal Pediatrics and Intensive Care – Neuropediatrics, Rouen University Hospital, Rouen, France*

Despite improvements in perinatal care, preterm birth still occurs regularly and the associated brain injury and adverse neurological outcomes remain a persistent challenge. Antenatal magnesium sulfate administration is an intervention with demonstrated neuroprotective effects for preterm births before 32 weeks of gestation (WG). Owing to its biological properties, including its action as an *N*-methyl-d-aspartate receptor blocker and its anti-inflammatory effects, magnesium is a good candidate for neuroprotection. In hypoxia models, including hypoxia-ischemia, inflammation, and excitotoxicity in various species (mice, rats, pigs), magnesium sulfate preconditioning decreased the induced lesions' sizes and inflammatory cytokine levels, prevented cell death, and improved long-term behavior. In humans, some observational studies have demonstrated reduced risks of cerebral palsy after antenatal magnesium sulfate therapy. Meta-analyses of five randomized controlled trials using magnesium sulfate as a neuroprotectant showed amelioration of cerebral palsy at 2 years. A meta-analysis of individual participant data from these trials showed an equally strong decrease in cerebral palsy and the combined risk of fetal/infant death and cerebral palsy at 2 years. The benefit remained similar regardless of gestational age, cause of prematurity, and total dose received. These data support the use of a minimal dose (e.g., 4 g loading dose ± 1 g/h maintenance dose over 12 h) to avoid potential deleterious effects. Antenatal magnesium sulfate is now recommended by the World Health Organization and many pediatric and obstetrical societies, and it is requisite to maximize its administration among women at risk of preterm delivery before 32 WG.

Keywords: magnesium sulfate, neuroprotection, preterm birth, cerebral palsy, animal studies, randomized controlled trials

## INTRODUCTION

Preterm brain injury remains a crucial and unresolved issue among neonatologists. The ensuing cerebral lesions (i.e., brain injury related to encephalopathy of prematurity, including white matter injury, periventricular leukomalacia, and intraventricular/intraparenchymal hemorrhage) are strongly associated with later cerebral palsy and neurobehavioral developmental disorders. The mechanisms leading to these forms of brain injury are numerous and may include inflammation or ischemic insult. Numerous risk factors may be present before, during, and after birth (e.g., intra- and extra-uterine growth restriction, systemic inflammation, or perinatal hypoxia-ischemia). Although

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Tomoki Arichi, King's College London, United Kingdom Sheffali Gulati, All India Institute of Medical Sciences, India*

> *\*Correspondence: Clément Chollat*

*clement.chollat@gmail.com*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 30 January 2018 Accepted: 28 March 2018 Published: 16 April 2018*

#### *Citation:*

*Chollat C, Sentilhes L and Marret S (2018) Fetal Neuroprotection by Magnesium Sulfate: From Translational Research to Clinical Application. Front. Neurol. 9:247. doi: 10.3389/fneur.2018.00247*

**29**

no single neuroprotective intervention is known to prevent preterm brain injury, neuroprotective strategies should be adopted to reduce the risk of neurodevelopmental anomalies in premature newborns. One such intervention is antenatal administration of magnesium sulfate (MgSO4) in women at risk of preterm birth. This mini review discusses the benefits of antenatal MgSO4 administration for fetal neuroprotection.

#### WHY IS MgSO4 A GOOD CANDIDATE FOR NEUROPROTECTION?

#### Biological Properties

Magnesium is the fourth most prevalent ion in the body and contributes to several physiological processes including storage, metabolism, and energy utilization. In the brain, magnesium is predominantly bound to chelators such as adenosine triphosphate (ATP) and is a cofactor in more than 300 enzymatic reactions (1, 2). Magnesium ions are essential for DNA, RNA, and protein synthesis. It contributes to glycolysis and ATP production and functions as a cell membrane stabilizer. In the central nervous system, magnesium is a non-competitive blocker of the *N*-methyld-aspartate (NMDA) glutamate receptor and modulates calcium influx. Its physiological role as a calcium channel blocker (3) and modulator of sodium and potassium flux through its action on ion pumps (e.g., Na+/K+ ATPase) and other membrane receptors (e.g., nicotinic acetylcholine receptor) (4) underlies its central role in heart function, muscle contraction, vascular tone, and nerve impulse conduction.

Sixty percent of magnesium is stored in bone, 20% in muscle, and 20% in soft tissue. Magnesium exists primarily in an ionized state (60%) but may also be complexed to proteins (33%) or anions (7%). Normal adult plasma concentration of magnesium is 0.75 mmol/L (95% confidence interval [CI]: 0.45–1.05) (5). In newborns, magnesium levels increase during the first week after birth (0.91 mmol/L [95% CI: 0.55–1.26]) (6).

#### Potential Mechanisms of Action Underlying the Neuroprotective Effect of Magnesium

Multiple mechanisms may underlie the neuroprotective impact of magnesium. Magnesium affects several pathways potentially involved in preterm brain injury. As a non-competitive NMDA receptor antagonist, magnesium prevents excitotoxic calciuminduced injury (7). Magnesium decreases extracellular glutamate under ischemic conditions, possibly reducing excitotoxicity (8). Magnesium limits calcium influx through voltage-gated channels, which may reduce the activation of apoptosis (9).

Magnesium also has anti-inflammatory properties as it reduces oxidative stress and reduces the production of pro-inflammatory cytokines interleukin-6 and tumor necrosis factor-α (10–14). Magnesium deficiency increases endothelial nitric oxide production, which can promote endothelial dysfunction (15, 16). This could involve decreased calcium influx and activation of phagocytic cells, inhibition of neurotransmitter release, or inhibition of nuclear factor kappa B.

### Neuroprotective Effects of Magnesium in Preclinical Studies

Since the 1980s, animal studies have investigated the neuroprotective role of magnesium. Early experiments involved adult animal models of hypoxia, stroke, or traumatic brain injury. In 1984, Vacanti and Ames demonstrated neuroprotective effects of MgSO4 in an adult rabbit spinal cord ischemia model (17). In 1987, MgSO4 administration to rat hippocampal slices reduced the effect of hypoxia (18). McIntosh et al. demonstrated in 1989 that post-traumatic MgSO4 injection decreased neurological disorders in a dose-dependent manner (19). In 1996, Marinov et al. showed that MgSO4 administration before a focal ischemic episode in rats could be neuroprotective by blocking NMDA receptors (20). The neurological impact of MgSO4 on the developing brain was evaluated in several lesion models. In 1990, McDonald et al. showed that cerebral lesions induced by intraspinal injection of NMDA in postnatal (P) day 7 rats were decreased after intraperitoneal administration of MgSO4 (21). Several studies have reported the importance of the timing of MgSO4 administration. Intraperitoneal administration of MgSO4 reduced the excitotoxic brain lesions in mice induced by intracerebral injection of ibotenate (a glutamate receptor agonist) on P5. However, there was no effect on brain lesions developing on the day of birth or on P10 (brain lesions induced by intracerebral injection of ibotenate in mice are comparable to those identified in preterm human infants by age, specifically P0–22 weeks of gestation (WG), P2–26 WG, P5–32 WG, P10–41 WG) (22). In this P5 model, MgSO4 prevented sensorimotor alterations in P6 and P7 and prevented motor impairment, fine motor skill alteration, and memory deficits in adolescent mice (P34–40) (23). In the seminal model of focal hypoxia-ischemia established by the Rice-Vannucci procedure (surgical ligature of the right carotid artery followed by a 1–2-h exposure to 8% oxygen) in rats, MgSO4 injection before the hypoxic episode on P7 led to reduced lesion sizes, decreased hippocampal apoptosis, and improved adult sensorimotor performances (9, 24). In that model, MgSO4 treatment preserved mitochondrial respiration and reduced inflammation, thus reducing the production of reactive oxygen species after hypoxia-ischemia (16).

Under hypoxic conditions (fraction of inspired oxygen 5–7%) in P2 piglet brains, MgSO4 prevented the changes induced by hypoxia in the function of neuronal nuclear membrane, which decreased the transcription of apoptotic proteins and kinase activity. These actions ultimately prevented programmed cell death (25, 26). The neuroprotective effect of MgSO4 was also assessed under inflammatory conditions. In pregnant rats, lipopolysaccharide (LPS)-induced inflammation affected progeny learning and memory capabilities at 3 months, which is improved by antenatal MgSO4 treatment (27). MRI abnormalities (increased T2 and diffusion coefficient levels in white and gray matter) were highlighted for pups of LPStreated dams, consistent with diffuse cerebral injury, which may be prevented by antenatal MgSO4 treatment (28). MgSO4 protected oligodendrocyte lineage cells *in vitro* in a model of hypoxic-ischemic injury (29).

### EFFECTS OF MgSO4 TREATMENT IN PREGNANCY

#### Use of MgSO4 in Obstetrics for Maternal Indication

MgSO4 has been used in obstetrics for decades as a tocolytic agent and for prevention or treatment of seizures in women with preeclampsia or eclampsia (30, 31). Despite strong evidence indicating effectiveness in preventing eclampsia, MgSO4 is ineffective in delaying preterm birth (32). Despite weak evidence, MgSO4 is still recommended by the American College of Obstetricians and Gynecologists for short-term pregnancy prolongation (up to 48 h) to allow the administration of corticosteroids (33). In a European population-based cohort study, 35% of women with severe pre-eclampsia, eclampsia, or HELLP syndrome received MgSO4 before delivery. Only 1 of 119 hospital units reported using MgSO4 as a first-line tocolytic (34).

#### MgSO4 Transplacental Passage

Fetuses are passively exposed to MgSO4 administered to pregnant women. In animals, fetal blood magnesium concentrations increase after maternal administration (35–37) and correlate with maternal blood levels (38, 39). The ratio of the mean fetal magnesium level to the mean maternal serum level at delivery was estimated at 0.94 ± 0.15 (40).

### OBSERVATIONAL STUDIES

Considering its use in obstetrics for maternal indications, its transplacental passage, and its neuroprotective action in animal studies, several observational studies have focused on the impact of MgSO4 on neurological outcomes in preterm neonates. Nelson and Grether showed that exposure to MgSO4 exposure was higher in the control group than in the group of children with cerebral palsy (odds ratio [OR], 0.14; 95% CI, 0.05–0.51) (41). In another cohort study, prenatal MgSO4 exposure was associated with a reduced risk of cerebral palsy at 3–5 years (OR, 0.11; 95% CI, 0.02–0.81) (42). Other observational studies have not shown effects of MgSO4 on infant neurological outcomes (43–52). A meta-analysis of these observational studies highlighted that antenatal MgSO4 treatment was associated with a significantly reduced risk of mortality (risk ratio [RR], 0.73; 95% CI 0.61–0.89) and cerebral palsy (OR, 0.64; 95% CI 0.47–0.89) (53). Antenatal MgSO4 treatment was also associated with a decreased incidence of apparent echodensities and echolucencies on neonatal cranial ultrasonography and cerebellar hemorrhage on MRI (54, 55).

#### RANDOMIZED CONTROLLED TRIALS (RCTs) OF MAGNESIUM AS A NEUROPROTECTANT

A total of five RCTs were performed in the 1990s and 2000s. Notably, two RCTs are ongoing: MASP (for administration of antenatal magnesium sulfate for the prevention of cerebral palsy

### Magnesium and Neurological Endpoints Trial (MagNET)

A total of 1,049 women in preterm labor at 25–33 WG (165 fetuses) treated at a single US center between October 1995 and January 1997 were included in the MagNET. Cases of triplet pregnancy or chorioamnionitis were excluded. In the tocolytic arm, women in active labor with cervical dilatation of at least 4 cm were randomly allocated to receive MgSO4 (4 g bolus then 2–3 g/h maintenance dose) or another tocolytic agent. In the neuroprotection arm, women with cervical dilatation of more than 4 cm were randomly allocated to receive MgSO4 (4 g bolus only) or 0.9% saline placebo.

The study was stopped prematurely in January 1997 due to significant mortality in the MgSO4 group (58, 59) and was widely discussed (60–63). The excessive number of mortalities occurred primarily in the tocolytic arm. The mortality rate in the MgSO4 group (11%) was consistent with that in previous reports of premature infants, whereas that in the placebo group was unusually low (1.4%). Moreover, causes of death were similar to those typical among premature children and were therefore difficult to attribute solely to MgSO4 treatment. Additionally, the confounding impact of multiple births was not accounted for, as more twin neonates were assigned to the treatment group than the placebo group. Finally, this increased mortality rate conflicted with the results of observational studies.

### Australasian Collaborative Trial of Magnesium Sulfate (ACTOMgSO4)

A total of 1,062 women in preterm labor before 30 WG from 16 centers were included in the Australasian Collaborative Trial of Magnesium Sulfate (ACTOMgSO4) between February 1996 and September 2000 (64). MgSO4 (4 g bolus followed by 1 g/h maintenance for 24 h or until birth) was randomly allocated to 535 women (629 live fetuses), and 527 women (626 live fetuses) received placebo. Although the primary study outcome, the rate of cerebral palsy at 2 years, was similar between the groups (5.7% in the MgSO4 group versus 6.7% in the control group; RR, 0.85; 95% CI, 0.55–1.31), the rate of motor dysfunction was significantly lower in the MgSO4 group (2.9 versus 5.4% in the control group; RR, 0.53; 95% CI, 0.30–0.92). Neonatal and pediatric mortality rates were also similar.

#### PREMAG Trial

The PREMAG trial included 573 women treated at 18 French centers between July 1997 and July 2003 (65), with 286 women (354 fetuses) randomly assigned to receive a 4-g bolus of MgSO4 and 278 women (341 fetuses), placebo. The trial was stopped after 6 years of enrollment. The primary outcomes (the rates of white matter injury and mortality) were similar between the groups (white matter injury, 10% versus 11.7%; OR, 0.78; 95% CI, 0.47–1.31; mortality, 9.4 versus 10.4%; OR, 0.79; 95% CI, 0.44–1.44). Combined death or gross motor dysfunction at 2 years was lower in the MgSO4 group (25.6 versus 30.8%; OR, 0.62; 95% CI, 0.41–0.93), but there was no difference in cerebral palsy (66).

### Beneficial Effects of Antenatal Magnesium Sulfate (BEAM)

The BEAM trial included 2241 women in preterm labor before 32 WG at 20 centers between December 1997 and May 2004 (67). Women were randomized to receive a 6-g bolus of MgSO4 followed by a 2-g/h maintenance dose for 12 h (1,096 women, 1,188 fetuses) or placebo (1,145 women, 1,256 fetuses). Antenatal MgSO4 administration had no impact on pediatric mortality. Although the primary outcome (composite of stillbirth or death by 1 year or cerebral palsy at 2 years) was similar in the two groups, moderate or severe cerebral palsy was significantly reduced in the MgSO4 group (1.9 versus 3.5%; RR, 0.55; 95% CI, 032–0.95).

### MAGnesium Sulfate for Prevention of Eclampsia (MAGPIE)

The MAGPIE trial, a large international trial to evaluate the impact of antenatal MgSO4 administration in the prevention of eclampsia, included 10,141 women with preeclampsia between July 1998 and November 2001: 1,544 women (1,593 fetuses) before 37 WG (68). The women were randomly allocated to receive either MgSO4 (4 g bolus followed by 1 g/h maintenance dose for 24 h) or placebo. A pediatric follow-up study including 4,483 children (2,254 and 2,229 in the MgSO4 and placebo groups, respectively) showed no difference in neurological outcomes (Ages and Stages questionnaire) or mortality at 18 months. Notably, only 19% of the children followed were born before 33 WG.

#### Meta-Analyses

These five RCTs have been the subject of four meta-analyses to date, with consistent findings and conclusions (69–73). In all meta-analyses, antenatal MgSO4 given to women at risk of preterm delivery was associated with a significantly reduced risk of cerebral palsy in children exposed *in utero*, with an RR ranging from 0.61 to 0.70 and no impact on mortality. The number of women needed to treat (NNT) to prevent one case of cerebral palsy ranged from 56 to 74 in infants born before 34 WG, and it was 29 in those born before 28 WG (**Table 1**). Minor maternal side effects (e.g., flushing, nausea or vomiting, sweating, injection site discomfort) were more frequent in the MgSO4 groups, but with no significant effect on serious maternal complications.

An individual participant data meta-analysis was also undertaken by the AMICABLE group (Antenatal Magnesium sulfate Individual participant data international Collaboration: Assessing the benefits for babies using the Best Level of Evidence) to explore the interaction between treatment and participant characteristics (74), which included the 5 above-mentioned RCTs (5,493 women and 6,131 babies). The overall RR of cerebral palsy among survivors after antenatal MgSO4 was 0.68 (95% CI, 0.54–0.87), and the NNT was 46. Interestingly, MgSO4 also reduced the combined risk of fetal/ infant death or cerebral palsy in the analysis of the 4 trials with fetal neuroprotective intent (RR 0.86, 95% CI, 0.75–0.99).

In all RCTs and meta-analyses to date, MgSO4 treatment had no impact on pediatric mortality or neonatal morbidity (respiratory distress syndrome, chronic lung disease, any intraventricular hemorrhage, cystic periventricular leukomalacia, necrotizing enterocolitis, patent ductus arteriosus, and retinopathy of prematurity). Similarly, MgSO4 treatment was not associated with serious maternal side effects. The benefit remained constant regardless of gestational age, cause of prematurity, total dose received, or maintenance dose administration after the loading dose. These data indicating persistent benefits of MgSO4 regardless of dose and support the use of low doses (e.g., 4 g loading dose ± 1 g/h maintenance dose for 12 h, 16 g maximum total dose) compared to high doses (e.g., 6 g loading dose + 2 g/h maintenance dose during 24 h, maximum total dose received: 54 g). Indeed, high MgSO4 dosage was implicated in the vasculopathy and high mortality observed in the MagNET trial (75). In a mouse preclinical model, MgSO4 demonstrated a dose-dependent, potentially deleterious effect on brain angiogenesis, vessel damage, and endothelial cell survival. The highest neuroprotective dose of MgSO4 induced cerebral hypoperfusion, whereas the lowest dose did not (76). These results support the use of MgSO4 in low doses.


*a Relative risk (95% CI).*

*bNumber needed to treat (95% CI).*

*CP, cerebral palsy; CI, confidence interval; IPD, individual participant data.*

#### Long-Term Follow-Up

Cohorts from the PREMAG and ACTOMgSO4 trials were followed up over their school-age years. From the PREMAG trial, 431 children were assessed at a mean age of 11 years (26.9% lost to follow-up) using a questionnaire completed by the parents (77). Although the ORs for motor, cognitive, behavioral outcomes, and school performance were favorable after magnesium treatment, the impact on neurodevelopment was not statistically significant. From the ACTOMgSO4, 669 children (21.3% lost to follow-up) were assessed at a mean age of 8 years using pediatric and psychological assessments and questionnaires completed by parents and teachers (78). Antenatal MgSO4 treatment had no impact on neurological, cognitive, behavioral, or school-related outcomes. Neither the PREMAG study nor the ACTOMgSO4 showed any effect on cerebral palsy at 2 years, likely because of limited sample sizes. Only the larger BEAM trial and meta-analyses reported reductions in cerebral palsy at 2 years. These long-term follow-up studies detected no harmful effects after antenatal MgSO4 treatment, although they were not designed for this purpose.

#### IMPLEMENTATION OF MAGNESIUM IN NEUROPROTECTIVE PROTOCOLS WORLDWIDE

In France, in 2015, only 60% of tertiary maternity hospitals used MgSO4 for fetal neuroprotection, with protocols that differed by maximum gestational age, possibility of retreatment, and monitoring (79). In Europe, in 2012, only 9 of 119 tertiary maternity hospitals (7.6%) used MgSO4 for fetal neuroprotection (34). Lack of experience and an absence of a written protocol or national guidelines, decision-making processes, environmental contexts, or beliefs about possible consequences seemed to represent barriers to widespread applications of MgSO4 in women at risk of preterm delivery (79, 80). Studies assessing MgSO4 protocol implementation found that nearly 70% of eligible women received MgSO4 before preterm delivery, and approximately 90% delivered within 24 h. The main reasons for not giving treatment

#### REFERENCES


were omission by the medical team and urgent delivery (81, 82). In an Australasian audit, the proportion of eligible women not receiving MgSO4 decreased significantly after publication of national guidelines, from 69.7% in 2010 to 26.9% in 2011, which was maintained in 2012 and 2013 (22.5%) (83). In Canada, a knowledge translation strategy (including national practice guidelines, online e-learning modules, educational rounds, and evaluation of barriers and feasibility) was associated with an 84% increase in optimal MgSO4 use (84). To improve the rates of MgSO4 administration to eligible women, implementing educational programs could be effective.

### CONCLUSION

Preterm birth is a major cause of death and a significant cause of long-term disability worldwide (85). MgSO4 is a safe and effective molecule that plays a key role in protecting the immature brain. It is a cost-effective, feasible, efficient, and safe intervention that contributes to the improvement of neurological outcomes. While MgSO4 has not been found to significantly improve cognition and behavior outcomes at school age, it prevents cerebral palsy at 2 years. Its use is now recommended by several pediatric and obstetrical societies, as well as the World Health Organization (strong recommendation based on moderate-quality evidence) for women at risk of imminent preterm birth before 32 WG. More work is needed to clarify the impact of MgSO4 on the cognitive outcome and efforts to improve the MgSO4 coverage of eligible women should be reinforced.

### AUTHOR CONTRIBUTIONS

CC, LS, and SM contributed equally to the writing of this review.

#### FUNDING

This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.


morbidity in < or = 1000-gram infants. *Am J Perinatol* (1998) 15:635–41. doi:10.1055/s-2007-994082


Network. Effects of antenatal exposure to magnesium sulfate on neuroprotection and mortality in preterm infants: a meta-analysis. *Obstet Gynecol* (2009) 114:354–64. doi:10.1097/AOG.0b013e3181ae98c2


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

# Nitric Oxide Synthase Inhibition as a Neuroprotective Strategy Following Hypoxic–Ischemic Encephalopathy: Evidence From Animal Studies

*Laurent M. A. Favié1,2\*, Arlette R. Cox <sup>3</sup> , Agnes van den Hoogen2 , Cora H. A. Nijboer <sup>4</sup> , Cacha M. P. C. D. Peeters-Scholte5 , Frank van Bel2,6, Toine C. G. Egberts1,7, Carin M. A. Rademaker1 and Floris Groenendaal 2,6*

*1Department of Clinical Pharmacy, University Medical Center Utrecht, Utrecht, Netherlands, 2Department of Neonatology, Wilhelmina Children's Hospital, University Medical Center Utrecht, Utrecht, Netherlands, 3Department of Pharmacy, Academic Medical Center, Amsterdam, Netherlands, 4 Laboratory of NeuroImmunology and Developmental Origins of Disease (NIDOD), University Medical Center Utrecht, Utrecht, Netherlands, 5Department of Neurology, Leiden University Medical Center, Leiden, Netherlands, 6Brain Center Rudolf Magnus, University Medical Center Utrecht, Utrecht, Netherlands, 7Department of Pharmacoepidemiology and Clinical Pharmacology, Faculty of Science, Utrecht University, Utrecht, Netherlands*

#### *Edited by:*

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Akira Yoshii, University of Illinois at Chicago, United States Sheffali Gulati, All India Institute of Medical Sciences, India*

> *\*Correspondence: Laurent M. A. Favié l.m.a.favie@umcutrecht.nl*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 27 December 2017 Accepted: 03 April 2018 Published: 19 April 2018*

#### *Citation:*

*Favié LMA, Cox AR, van den Hoogen A, Nijboer CHA, Peeters-Scholte CMPCD, van Bel F, Egberts TCG, Rademaker CMA and Groenendaal F (2018) Nitric Oxide Synthase Inhibition as a Neuroprotective Strategy Following Hypoxic–Ischemic Encephalopathy: Evidence From Animal Studies. Front. Neurol. 9:258. doi: 10.3389/fneur.2018.00258*

Background: Hypoxic–ischemic encephalopathy following perinatal asphyxia is a leading cause of neonatal death and disability worldwide. Treatment with therapeutic hypothermia reduced adverse outcomes from 60 to 45%. Additional strategies are urgently needed to further improve the outcome for these neonates. Inhibition of nitric oxide synthase (NOS) is a potential neuroprotective target. This article reviews the evidence of neuroprotection by nitric oxide (NO) synthesis inhibition in animal models.

Methods: Literature search using the EMBASE, Medline, Cochrane, and PubMed databases. Studies comparing NOS inhibition to placebo, with neuroprotective outcome measures, in relevant animal models were included. Methodologic quality of the included studies was assessed.

Results: 26 studies were included using non-selective or selective NOS inhibition in rat, piglet, sheep, or rabbit animal models. A large variety in outcome measures was reported. Outcome measures were grouped as histological, biological, or neurobehavioral. Both non-selective and selective inhibitors show neuroprotective properties in one or more outcome measures. Methodologic quality was either low or moderate for all studies.

Conclusion: Inhibition of NO synthesis is a promising strategy for additional neuroprotection. In humans, intervention can only take place after the onset of the hypoxic–ischemic event. Therefore, combined inhibition of neuronal and inducible NOS seems the most likely candidate for human clinical trials. Future studies should determine its safety and effectiveness in neonates, as well as a potential sex-specific neuroprotective effect. Researchers should strive to improve methodologic quality of animal intervention studies by using a systematic approach in conducting and reporting of these studies.

Keywords: nitric oxide synthase inhibition, neuroprotection, animal models, hypoxic–ischemic encephalopathy, 2-iminobiotin, review

## INTRODUCTION

Hypoxic–ischemic encephalopathy (HIE) following perinatal asphyxia (i.e., severe oxygen deprivation at birth) is one of the leading causes of neonatal death and adverse neuromotor outcome in term and near-term infants worldwide. In high-income countries, the incidence of HIE has been estimated between 0.5 and 1.0 for every thousand live births, although some sources have reported an incidence as high as 8 per 1,000 live births (1, 2). In low- and middle-income countries, the incidence of HIE is higher, affecting more than 1.1 million babies annually (3–5).

The overall burden of HIE is high, in terms of quality-adjusted life years, years of life lost, and years lived with disability, not to mention a great financial cost for both society and the families involved (6, 7). With an estimated annual one million deaths worldwide, HIE is accountable for roughly 25% of all deaths in the neonatal period (3, 8).

Hypoxic–ischemic brain injury is not a single event, evoked by the actual asphyxia, but rather an ongoing process that leads to significant neuronal cell death over hours to days after the initial insult (9, 10). Several distinct phases have been identified in this process. The primary energy failure takes place during the hypoxic–ischemic event, resulting in failure of oxidative metabolism, cytotoxic edema, and accumulation of excitotoxins (11). After resuscitation and restoration of cerebral circulation, a latent phase, lasting approximately 6 h, commences (12, 13). Subsequently, starting between 6 and 15 h after asphyxia, the brain experiences a secondary energy failure that can last for days. This phase is marked by seizures, renewed cytotoxic edema, release of excitotoxins, impaired cerebral oxidative energy metabolism, and finally, neuronal cell death (14).

Currently, the only treatment that has proven to effectively reduce hypoxic–ischemic brain injury following perinatal asphyxia is the application of therapeutic hypothermia (TH). During TH the brain temperature is lowered to 33–34°C which is maintained for 72 h (1). Since the introduction of TH, the combined adverse outcome of death and disability, such as hearing loss, cerebral palsy, and other neuromotor disorders, has been reduced from approximately 60–45% (15–17). TH has widely been implemented as the standard of care treatment for moderate to severe HIE in high-income countries. However, TH needs to be started within 6 h after birth, leaving clinicians with a narrow window for establishing the diagnosis and severity of HIE as well as transportation to a medical facility equipped for TH (18). Additional neuroprotective strategies for HIE are urgently needed to augment TH, but when hypothermia is not yet feasible, act as a first line treatment option (3, 4, 19).

A potential target for (additional) neuroprotection in patients with HIE is the inhibition of nitric oxide synthase (NOS, enzyme commission number 1.14.13.39). NOS is an enzyme catalyzing production of nitric oxide (NO) from l-arginine. After perinatal asphyxia, NO can react with the superoxide free radical to form toxic peroxynitrite, setting a pre-apoptotic pathway in motion, resulting in neuronal loss (10, 20). Nitrotyrosine, an end product of this process, has been demonstrated post mortem in neonatal brain and spinal cord tissue after severe HIE (21, 22).

Three isoforms of NOS have been identified: endothelial (eNOS), neuronal (nNOS), and inducible NOS (iNOS) (23). All isoforms are upregulated after asphyxia; both nNOS and eNOS immediately after reperfusion and iNOS from several hours onward (24). While eNOS is regarded to be critical in maintaining pulmonary blood flow, preventing pulmonary hypertension and thereby maintaining adequate oxygenation of tissues throughout the body, excessive activation of nNOS and iNOS is associated with deleterious effects on the brain (24, 25). To illustrate, in mice genetically deficient of eNOS, infarct size after middle cerebral artery occlusion is larger compared with wild-type animals, due to a reduction in regional cerebral blood flow (26). By contrast, nNOS knockout mice are protected against hypoxic–ischemic brain injury, while mice lacking iNOS showed a delayed reduction in brain injury (27–32).

The aim of this study is to review the available evidence on NOS inhibition as a potential neuroprotective strategy in animal models translational for neonatal HIE and to identify one or more NOS inhibiting compounds that could evolve from preclinical to clinical studies in the near future.

### METHODS

#### Search Strategy

Studies assessing the neuroprotective effects of NOS inhibitors in HIE models were identified. A literature search using the EMBASE, Medline, Cochrane, and PubMed databases was performed. The primary keywords were *Animals (newborn)*, *Hypoxia*, and *Nitric Oxide Synthesis*; the searches were limited to the English language. The complete search string is included in Supplementary Material. After the exclusion of duplicates, the titles and abstracts were independently screened by two researchers (Laurent M. A. Favié and Arlette R. Cox). A final selection was made after full text evaluation. Any discrepancies were resolved by a third researcher (Floris Groenendaal). In addition, the reference lists of the retrieved articles were searched for additional studies.

#### Selection Criteria

Studies were included based on the following inclusion criteria: animal models of a postnatal age in which brain development corresponds to near term or term brain development in humans, transient hypoxia or hypoxia–ischemia (HI), neuroprotection as outcome defined by histological, biochemical, and/or neurobehavioral parameters and inclusion of both a treatment group administering at least one NOS inhibitor and a control group that received sham treatment or consisted of untreated animals.

#### Data Synthesis

The year of publication, name of first author, the class and type of NOS inhibitor, the animal model, the method used to achieve HI, the dose and number of animals in each treatment group, the type of control group and number of control animals, the timing of administration with regards to the HI insult, and the results on the reported outcome parameters were recorded for each study. Each outcome parameter was categorized as histological, biochemical, or neurobehavioral.

#### Quality Assessment

The methodological quality of the included articles was assessed using the SYRCLE's risk of bias (RoB) tool for animal intervention studies (33). This tool is based on the Cochrane RoB tool and consists of 10 items on which an article can be scored. Each item was scored 0, 1, or 2 points by two researchers (Laurent M. A. Favié and Arlette R. Cox) independently. If no evidence for adherence or evidence for non-adherence was found, a score of 0 was awarded. When evidence for adherence was present but inconclusive, one point was scored. If the item was fully adhered to, two points were scored. Any discrepancies were resolved after consultation with a third researcher (Agnes van den Hoogen). Because of the nature of the included studies and the timing of the interventions, "allocation concealment" was deemed unfeasible and was not rated for any of the articles. Articles scoring 1–6 points were considered low quality, 7–12 points moderate quality, and 13–18 points high quality. An example of the tool is included in Supplementary Material.

#### RESULTS

#### Eligible Studies

The search yielded a total of 348 studies; 280 studies after removal of duplicates. After screening of title and abstract, 238 articles were excluded. Screening of the reference lists identified one additional article. 43 articles were thus assessed in full detail. Of these, 26 were deemed eligible for inclusion (**Figure 1**); the data were extracted from these studies, and these studies were assessed for methodological quality. Performing a meta-analysis was considered impossible because of the heterogeneity of the studies in outcome, administered NOS inhibitor, and animal models.

#### Study Characteristics

The included studies and their descriptive characteristics are summarized in **Table 1**. Eight studies (31%) tested a non-specific NOS inhibitor (34–41), another eight (31%) applied an nNOS-specific inhibitor (42–49); three studies (12%) used an iNOS-specific inhibitor (50–52); and six (23%) used an inhibitor of both nNOS and iNOS (53–58). One study (3%) used separate groups for nNOS and iNOS inhibition (59). Four different species of animals were used: rat (*n* = 11, 42%), piglet (*n* = 10, 38%), sheep (*n* = 3, 12%), and rabbit (*n* = 2, 8%).

Different models for HI were used, mostly dependent on the animal species. All rat studies applied the Vannucci–Rice model in P7–P14 pups. All newborn (P1–P5) piglet studies induced brain injury by hypoxia for 30–60 min, in 30% of studies combined with transient bilateral artery occlusion. In sheep

#### Table 1 | Study characteristics including RoB score.


(*Continued*)

Neuroprotection by NOS Inhibition

#### TABLE 1 | Continued


#### TABLE 1 | Continued


(*Continued*)

Neuroprotection by NOS Inhibition


First author

TABLE 1 | Continued

NOS inhibitor

Animal, age

 HI method  Dose, no.,

RoA

class, type

(year)

#### NP yes/no B, yes Significantly higher hippocampus and cortex neuropathology score vs vehicle Significantly higher ipsilateral/ contralateral hemisphere area ratio vs vehicle B, yes Significantly lower ipsilateral HSP70 level vs vehicle H, no No difference in nitrotyrosine levels vs vehicle Nijboer (2007) (54) nNOS and iNOS, 2-IB Rat (Wistar), 7 days Right carotid artery ligation and hypoxia (FiO2 0.08) for 120 min 10 mg/kg, *n* = NS, sc Vehicle, *n* = NS, sc Directly after insult, repeated at 12 and 24 h H, yes Significantly higher ipsilateral/ contralateral hippocampus area ratio vs vehicle in females only Significant reduction in cortical and hippocampal lesions vs vehicle in females only M (11)B, yes Significant reduction in cytochrome c release vs vehicle in females only Decrease in caspase-3 activity vs vehicle in females only No effect on nuclear translocation of apoptosis-inducing factor vs vehicle in both genders N, yes Less deaths in female pups compared with male pups Peeters-Scholte (2002) (55) nNOS and iNOS, 2-IB Piglet (Dutch Store) 1–3 days Bilateral carotid artery occlusion and hypoxia for 60 min 0.2 mg/kg, *n* = 11, iv Vehicle, *n* = 12, iv Directly after insult, repeated every 60 min, six doses in total H, yes 90% reduction of vascular edema vs vehicle 60–80% increase in normal neuronal cells vs vehicle M (7)B, yes 90% improvement of cerebral energy state vs vehicle Reduction of caspase-3 activity by 93% in cortex and 71% in striatum vs vehicle Peeters-Scholte (2002) (56) nNOS and iNOS, 2-IB Piglet (Dutch Store) 1–3 days Bilateral carotid artery occlusion and hypoxia for 60 min 0.2 mg/kg, *n* = 11, iv Vehicle, *n* = 12, iv Directly after insult, repeated every 60 min, six doses in total B, yes Preservation of endogenous IGF-1 production vs vehicle Reduction of caspase-3 activity vs vehicle L (4)B, no No significant decrease in cytokine production vs vehicle

Control, no.,

Timing Outcome

Result

Parameter H/B/N,

RoA

RoB L/M/H

(score)

(*Continued*)

TABLE 1 | Continued


Frontiers in Neurology


www.frontiersin.org

(*Continued*)

Neuroprotection by NOS Inhibition

Frontiers in Neurology


www.frontiersin.org


*Studies are grouped by class and type of NOS inhibitor and subsequently by type of animal tested and year of publication. Low quality studies are indicated by a gray background.*

*non-spec, non-specific; NNLA, N-nitro-l-arginine; l-NAME, N-nitro-l-arginine methyl ester; nNOS, neuronal nitric oxide synthase; 7-NI, 7-nitro indazole; iNOS, inducible nitric oxide synthase; S-MI, S-methyl-isothiourea; AG, aminoguanidine; IMC, indomethacin; 2-IB, 2-iminobiotin; GA, gestational age; FiO2, fraction of inspired oxygen; MABP, mean arterial blood pressure; RoA, route of administration; ip, intraperitoneal; iv, intravenous; sc, subcutaneous; NP, neuroprotection; H, histological; B, biochemical; N, neurobehavioral; HI, hypoxia–ischemia; HSP70, heat shock protein 70; EEG, electro-encephalogram; BBB, blood–brain barrier; RoB, risk of bias; L, low; M, moderate; NA, not available; NOS, nitric oxide synthase; NS, not specified.*

aged 2–11 days (one study), hypoxia for 30 min was combined with hypotension for 5 min. Also, two studies using sheep at 103–104 days gestation (term = 147 days) were included, in which brain injury was induced by hypoxia due to occlusion of the umbilical cord for 25 min. In rabbits, fetuses (embryonic day 22, 70% gestation) were subjected to an HI event by uterine ischemia for 40 min.

The dosing regimen of the included studies is summarized in **Table 2**. Seventeen studies (65%, all non-specific or nNOSspecific inhibitors) describe only a single administration, and nine studies (35%, all iNOS of combined nNOS and iNOS inhibitors) described repeated dosing. With regards to timing of the intervention, 12 studies (46%) administered the (first) dose before the onset of the HI event; 9 (35%) after the event; and the remaining 5 (19%) incorporated groups with administration both before and after the event.

#### Outcome

The results of the reported outcome parameters for each study are presented in **Table 1**. A wide variety of histological, biochemical, and neurobehavioral outcome parameters were reported. Histological parameters included ipsilateral/contralateral weight ratio disparity and analysis of cortical and striatal lesions. Biochemical parameters included free radical formation and other biomarkers for neurological damage, but also cerebral energy status and electrocortical brain activity. Neurobehavioral parameters included overall survival, survival with normal EEG and results of neurobehavioral tests.

In the group of non-specific NOS inhibitors, administration before onset of the insult proved neuroprotective in 7/8 settings (88%), while administration directly after the insult was partially beneficial in 2/3 settings (67%).

For nNOS inhibitors, administration before the insult showed neuroprotective properties in 9/10 settings (90%) and when administered directly after the insult (1/1). When administration was delayed by 15 min or more, neuroprotective properties were lost (4/4).

When treatment with an iNOS inhibitor was started before the insult, neuroprotection was achieved (4/4). Administering the first dose after induction of HI showed neuroprotection in 33% of the settings (1/3). Hsu et al. administered the iNOS inhibitor aminoguanidine (AG) 30 min before and 3 h after the insult as a single dose. Both were neuroprotective compared with the control group, although less parameters were tested in the post insult treatment group.

All studies testing combined inhibition of nNOS and iNOS reported (partially) neuroprotective outcome. van den Tweel et al. (53) showed that 2-iminobiotin (2-IB) is neuroprotective in rats in a dose-dependent matter.

A direct comparison between two different inhibitors was made in two studies. Yu et al. reported superior neuroprotection of the novel nNOS inhibitor JI-8 compared with 7-nitro indazole (7-NI) when administered before the insult in equimolar doses. Hsu et al. observed that both 7-NI and AG are neuroprotective when administered 30 min before HI and that 7-NI is superior to AG in this setting. When the compounds were administered 3 h after HI, 7-NI lost its neuroprotective effect while AG remained neuroprotective compared with both vehicle and 7-NI.

#### Methodological Quality

Eleven studies (42%) were ranked low quality, 15 (58%) were considered moderate quality; none of the studies were ranked in the high quality group. On average, RoB score was 7 (3–12). Overall, animal baseline characteristics, randomization for treatment allocation, blinding of investigators and/or outcome assessors, and random selection for outcome assessment were often not mentioned and therefore scored 0.

### DISCUSSION

This systematic review shows that both selective and non-selective NOS inhibitors have neuroprotective qualities in various animal models of HI brain damage using histological, biochemical, and neurobehavioral outcome parameters. In animal studies, induction of the insult and administration of the potentially neuroprotective agent (before and/or after the insult) can be timed precisely. By contrast, this is not the case in clinical practice. The onset of perinatal asphyxia is often sudden and unpredictable. Therefore, administration of any drug before the onset of the insult is impossible, and administration directly after the insult (i.e., directly after birth) is highly improbable. All non-selective NOS inhibitors reviewed in this study were administered before insult or directly after; there are no data on delayed administration. Furthermore, non-selective inhibitors will also target eNOS, which could counteract the potential neuroprotective effects of


*a One study tested both an nNOS and iNOS inhibitior in separate groups.*

*non-spec, non-specific; nNOS, neuronal nitric oxide synthase; iNOS, inducible nitric oxide synthase; HI, hypoxia–ischemia.*

nNOS and iNOS inhibition. For selective nNOS inhibitors, neuroprotection was lost when administration was withheld by as little as 15 min. For selective iNOS inhibitors, administration before the insult shows greater neuroprotective potential than postinsult treatment. The combination of nNOS and iNOS inhibition shows neuroprotective properties on histological, biochemical, and neurobehavioral outcome parameters when administered after the insult in a repeated dosing regimen. Thus, combined nNOS/iNOS inhibition with a repeated dosing regimen seems the most promising strategy to advance into human clinical trials. In fact, several phase II studies with 2-IB are currently underway, in addition to TH (NTR5221) as well as without TH in low-income countries (NCT01626924, EudraCT2015-003063-12).

Because of the wide variety in reported outcome measures, a clear-cut comparison between inhibitors based on outcome was difficult to make. Twelve studies report no neuroprotection on one or more outcome parameters after NOS inhibition. Potentially, this can be attributed to timing of the intervention or suboptimal dosing. When a NOS inhibitor is administered before the insult, the compound will be present in the tissues and circulation at the time of the actual insult, increasing the compound's potential to exhibit neuroprotective effects. Most studies have tested one NOS inhibitor in a single dose. In studies testing different dosages, a higher dose often shows a better neuroprotective outcome, although some studies indicate a U-shaped effect. For 2-IB, the optimal dose in rats appears to be 30 mg/kg intraperitoneal (53). In piglets, increasing the dose by five times to 1.0 mg/kg intravenous does not provide greater neuroprotective properties compared with 0.2 mg/kg (57). Although most studies measured histological and biochemical outcome parameters associated with neuroprotection, the clinically most relevant parameter of improved neurobehavioral outcome was reported in four studies only. Yu et al. (45) and Ji et al. (49) showed that nNOS inhibition administered before the insult resulted in less deaths, and less neurobehavioral abnormalities in rabbits. Nijboer et al. (54) and Bjorkman et al. (57) report a (partial) neuroprotective effect for 2-IB on neurobehavioral outcome parameters in rats and piglets, respectively. Assessing neurobehavioral outcome requires a longer follow up period, which often involves intensive hands-on trained personnel especially in larger animal models, as well as validated tools to score the desired outcome parameter, making it very expensive. Using histological and biochemical markers provides researchers with a more time- and cost-effective alternative. Although data are limited, results on neurobehavioral outcome parameters, combined with results from histological and biochemical parameters, identify NOS inhibition as a potential neuroprotective strategy in humans.

Important differences exist between the adult and the neonatal brain with regard to susceptibility to injury, plasticity and cell death pathways. Therefore, adult animal models are not suitable to examine neuroprotective interventions for HIE. Across species, key brain maturation events regarding susceptibility and regenerative capacities have been identified at different moments before and after birth and are related to the developmental stage of the human neonatal brain (60–62). It is generally accepted that rats, at postnatal days 7–14 (P7–14), are comparable to near term/ term human neonates with regards to cerebral cortex development (63, 64). The Vannucci–Rice model of unilateral common carotid artery ligation followed by a period of systemic hypoxia results in apoptotic-necrotic cell degeneration in P7–14 rats, similar to HIE (64–67). In term piglets aged 1–5 days, hypoxia leads to basal ganglia and somatosensory cortical injury, largely comparable to damage seen in human neonates after perinatal asphyxia (64, 68, 69). Introducing HI *in utero* to fetal rabbits provides animals with a motor phenotype similar to human cerebral palsy (64, 70). In term and preterm sheep models, hypoxia and asphyxia cause abnormalities in cerebral oxygen metabolism and hemodynamics as well as electrocortical brain activity comparable to human neonates after HI and basal ganglia injury representative for cerebral palsy (71–73).

Of interest is the potential role of sex-specific cell death pathways involved in HIE and possible sex-specific neuroprotective therapies. In general, females seem to be less susceptible to brain injury. This effect is seen across species, age groups, and origin of injury (74). In adult animal models, reduction in ischemic injury in females has been attributed to estradiol levels (74). Although estradiol will not be as predominant in prepubertal animal models, there is evidence of sexual dimorphism regarding sex steroids in central nervous system development in mice and rats (75, 76). Other studies show sex-specific cell death pathways leading to brain injury after HI both *in vitro* and *in vivo*. For instance, there is evidence that brain injury after HI in males is evoked by caspaseindependent pathways whereas in females, caspase-dependent pathways are responsible (77–82). Therefore, neuroprotective agents such as NOS inhibitors that interact, either upstream or downstream, with the caspase-dependent pathway may be effective in females only.

The role of sex was only sparsely investigated in the studies included. For the majority of the studies (65%), the sex of the animals used was not reported. Six studies (23%) used rats of both sexes but have not reported sex-specific outcome. Yu et al. reported no outcome differences between sex for 7-NI and JI-9 but this statement was not supported by statistical analysis, possibly due to the small sample size in each of the groups (45). Nijboer et al. showed a statistically significant difference in histological and biochemical outcome parameters between sexes in rats, concluding that 2-IB was neuroprotective in female rats only (54). Other studies with different neuroprotective agents in both animals and humans also indicate a (potential) neuroprotective effect in females only (81–84).

Methodological quality assessment using the SYRCLE's RoB tool resulted in only low and moderate scores for the publications used in this study. In all of the studies, at least on one or more items no information was available, forcing a score of 0 in that area. It is unknown whether these items were not adhered to during the experiment, or simply not included in the final manuscript due to regulations imposed by the editorial guidelines of the publishing journal. Unfortunately, it is not yet common practice to be as complete and precise in reporting data for animal studies as it is for human studies (33, 85). However, since this problem was addressed in a commentary published in the Lancet in 2002, awareness has been steadily increasing (86, 87). Fourteen of the studies included in this review were published in or before 2002; seven (50%) scoring low and an equal number scoring moderate. For the 12 included studies published in 2003 or later, 8 (67%) were awarded a moderate score. The SYRCLE's RoB tool proved to be an adequate tool to consistently score the methodological quality of the included studies. However, this tool was developed recently and experience is still sparse. We would like to encourage future researchers to adhere to the items listed in this tool when conducting and reporting animal intervention studies to improve the methodological quality of studies as well as to use this tool when attempting a systematic review of animal literature. To illustrate the need for improvement in methodological quality and because of the possibility that low scores reflect lack of reporting and not lack of quality in the design of the study, we decided not to omit low quality studies nor did we emphasize the RoB scores when comparing the NOS inhibitors discussed in this study.

An important limitation of this study is that no independent statistics could be applied due to the large heterogeneity in study designs. Ideally, all NOS inhibitors should be tested in identical animal models with identical outcome measures. In reality, researchers over the past decades have used various animal models, dose and timing of NOS inhibitors, and reported outcome parameters. For the purpose of this review, we choose to report all of these and base our conclusions on the best available evidence. Based on this heterogeneity, these conclusions should be interpreted with caution.

Despite the low to moderate methodological quality according to the RoB tool, presented in Supplementary Material, and the lack of independent statistics, the evidence presented in this systematic review still indicates NOS inhibition as a promising strategy for (additional) neuroprotection in human neonates after perinatal asphyxia. Combined inhibition of nNOS and

#### REFERENCES


iNOS started as soon as possible after birth and in a repeated dosing regimen seems to have the best potential based on the combined outcome parameters, translation to clinical practice and methodological quality. Human studies (phase 2, openlabel) with 2-IB, an inhibitor of both nNOS and iNOS, are currently taking place. Future clinical studies should make clear whether the sex-specific neuroprotective effect of drugs such as 2-IB observed in rats is present in humans as well. Furthermore, well designed placebo-controlled studies are needed to determine the safety of 2-IB in neonates and its effectiveness both with and without TH.

## AUTHOR CONTRIBUTIONS

LF, AC, and FG were involved in study selection; LF, AC, and AH conducted the methodological quality assessment. All the authors discussed the results and read and approved the final version of the manuscript. LF drafted the manuscript; AC, AH, CN, CP-S, FB, TE, CR, and FG provided critical feedback to each draft.

### ACKNOWLEDGMENTS

The authors would like to thank Justin van der Swaluw for his preliminary work in defining the search strategy.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fneur.2018.00258/ full#supplementary-material.


**Conflict of Interest Statement:** FB, FG, and CP-S are inventors of 2-iminobiotin as neuroprotective agent for neonates with HIE. CP-S is consultant for and shareholder of Neurophyxia BV's-Hertogenbosch, The Netherlands. The other authors report no potential conflict of interest.

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

# Neuroprotective Drugs in Infants With Severe Congenital Heart Disease: A Systematic Review

Raymond Stegeman1,2,3 \*, Kaya D. Lamur 1,2,3, Agnes van den Hoogen<sup>1</sup> , Johannes M. P. J. Breur <sup>2</sup> , Floris Groenendaal <sup>1</sup> , Nicolaas J. G. Jansen<sup>3</sup> and Manon J. N. L. Benders <sup>1</sup>

*<sup>1</sup> Department of Neonatology, University Medical Center Utrecht, Utrecht University, Wilhelmina Children's Hospital, Utrecht, Netherlands, <sup>2</sup> Department of Pediatric Cardiology, University Medical Center Utrecht, Utrecht University, Wilhelmina Children's Hospital, Utrecht, Netherlands, <sup>3</sup> Department of Pediatric Intensive Care, University Medical Center Utrecht, Utrecht University, Wilhelmina Children's Hospital, Utrecht, Netherlands*

Background: Perinatal and perioperative brain injury is a fundamental problem in infants with severe congenital heart disease undergoing neonatal cardiac surgery with cardiopulmonary bypass. An impaired neuromotor and neurocognitive development is encountered and associated with a reduction in quality of life. New neuroprotective drugs during surgery are described to reduce brain injury and improve neurodevelopmental outcome. Therefore, our aim was to provide a systematic review and best-evidence synthesis on the effects of neuroprotective drugs on brain injury and neurodevelopmental outcome in congenital heart disease infants requiring cardiac surgery with cardiopulmonary bypass.

#### Edited by:

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### Reviewed by:

*Kumar Sannagowdara, Medical College of Wisconsin, United States Marie Pascale Pittet Metrailler, Hospital for Sick Children, Canada*

> \*Correspondence: *Raymond Stegeman r.stegeman@umcutrecht.nl*

#### Specialty section:

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

Received: *05 January 2018* Accepted: *13 June 2018* Published: *03 July 2018*

#### Citation:

*Stegeman R, Lamur KD, van den Hoogen A, Breur JMPJ, Groenendaal F, Jansen NJG and Benders MJNL (2018) Neuroprotective Drugs in Infants With Severe Congenital Heart Disease: A Systematic Review. Front. Neurol. 9:521. doi: 10.3389/fneur.2018.00521* Methods: A systematic search was performed in PubMed, Embase and the Cochrane Library (PRISMA statement). Search terms were "infants," "congenital heart disease," "cardiac surgery," "cardiopulmonary bypass," and "neuroprotective drug." Data describing the effects on brain injury and neurodevelopmental outcome were extracted. Study quality was assessed with the Cochrane Risk of Bias Tool. Two reviewers independently screened sources, extracted data and scored bias. Disagreements were resolved by involving a third researcher.

Results: The search identified 293 studies of which 6 were included. In total 527 patients with various congenital heart diseases participated with an average of 88 infants (13–318) per study. Allopurinol, sodium nitroprusside, erythropoietin, ketamine, dextromethorphan and phentolamine were administered around cardiac surgery with cardiopulmonary bypass. Allopurinol showed less seizures, coma, death and cardiac events in hypoplastic left heart syndrome (HLHS) infants (OR: 0.44; 95%-CI:0.21–0.91). Sodium nitroprusside resulted in lower post cardiopulmonary bypass levels of S100ß in infants with transposition of the great arteries after 24 (*p* < 0.01) and 48 (*p* = 0.04) h of treatment. Erytropoietin, ketamine and dextromethorphan showed no neuroprotective effects. Phentolamine led to higher S100ß-levels and cerebrovascular resistance after rewarming and at the end of surgery (both *p* < 0.01). Risk of bias varied between studies, including low (sodium nitroprusside, phentolamine), moderate (ketamine, dextromethorphan), and high (erytropoietin, allopurinol) quality.

**50**

Conclusions: Allopurinol seems promising for future trials in congenital heart disease infants to reduce brain injury given the early neuroprotective effects in hypoplastic left heart syndrome infants. Larger well-designed trials are needed to assess the neuroprotective effects of sodium nitroprusside, erytropoietin, ketamine and dextromethorphan. Future neuroprotective studies in congenital heart disease infants should not only focus on the perioperative period, however also on the perinatal period, since significant brain injury already exists before surgery.

Keywords: infant, congenital heart disease, cardiac surgery, cardiopulmonary bypass, neuroprotective drugs, brain injury, neurodevelopmental outcome

#### INTRODUCTION

#### Rationale

Congenital heart disease (CHD) is the most common congenital malformation with an incidence varying from 4 to 50 per 1,000 live births (1). The incidence of severe forms of CHD severely ill patients presenting in the newborn period or early infancy—is about 6 per 1,000 live births, including infants with transposition of the great arteries (TGA), univentricular heart physiology (UVH), aortic arch anomalies, tetralogy of Fallot (ToF), and large ventricular septal defects (VSD) undergoing cardiac surgery with cardiopulmonary bypass (CPB) (1). The survival of infants with severe CHD until adulthood has increased substantially to almost 90% during the last decades as a result of improved surgical procedures and intensive care (2). However, delayed brain development, brain injury and related long-term neurodevelopmental impairments are relevant problems in infants with severe CHD, indicating the urgent need for neuroprotective drugs. Altered cerebral circulation and reduced cerebral oxygenation already starts before birth and is associated with impaired brain growth in fetuses and neonates with severe CHD (3–5). The delayed brain development "in utero" increases the vulnerability for hypoxic-ischemic brain injury in postnatal life. The periods around birth and cardiac surgery with CPB are the most critical periods for the occurrence of brain injury (6–8). Early postnatally and preoperatively magnetic resonance imaging (MRI) of the brain shows injury in up to 63% of the infants with severe CHD (6, 7). After cardiac surgery with CPB up to 78% shows additional brain injury on MRI (8). Most common forms of brain injury seen in severe CHD infants are white matter injury (WMI) and focal infarctions of the gray matter, which are known to be caused by hypoxicischemic events (5, 8). Hypoxia causes excessive production of excitotoxins with overactivation of the N-methyl D-aspartate (NMDA-) receptor and calcium influx into neurons leading to cell damage and the release of pro-radicals and increased levels of xanthine. Upon reperfusion and reoxygenation reactive oxygen ("oxidative stress") and nitrogen species are formed. This chain of events activates the inflammatory pathway with increased formation of pro- and anti-inflammatory cytokines resulting in inappropriate apoptosis and further brain injury. In addition neurotrophic factors are downregulated leading to diminished recovery of brain injury (9). Important consequences of brain injury in severe CHD infants are longterm neuromotor (standard deviation (SD) −1.5) and cognitive (SD −0.65) impairments (10), with even lower scores in infants with syndromic disorders (11). At school age, language disorders (20–30%), behavioral problems (20–40%), learning difficulties (30–50%) (12, 13) and impairments in executive functions are common (14). Therefore, it is of great importance to find ways to reduce hypoxic-ischemic brain injury and improve neurodevelopmental outcome (NDO) in this vulnerable population. A number of diagnostic and therapeutic neuroprotective strategies have been investigated. Most of these strategies focus especially on the period around cardiac surgery with CPB, including neuromonitoring with (amplitude integrated) electroencephalography, transcranial Doppler ultrasound and near-infrared spectroscopy and perfusion techniques as deep hypothermic circulatory arrest (DHCA), low flow CPB, and regional cerebral perfusion (15). Only avoidance of extreme hemodilution during hypothermic CPB (hematocrit level above 24%) is recommended. A low hematocrit strategy (mean 21.5%, SD 2.9) showed worse perioperative and neurodevelopmental outcomes in comparison to a higher hematocrit strategy (mean 27.8%, SD 3.2), as was indicated by higher lactate levels post-CPB (p = 0.03) and lower scores on psychomotor developmental index (82 vs. 90, p < 0.01) (16, 17). Some procedures or treatments are reasonable to consider, including deep hypothermia during CPB (18), avoiding hypoglycemia perioperatively (19), and postoperative normothermia (20). However, currently there is limited evidence for the effectiveness of the majority of the investigated neuroprotective strategies (15). The cascade leading to brain injury provides several pharmaceutical targets to intervene. Drugs that antagonize the NMDA-receptor, prevent oxidative stress, suppress the inflammatory response or upregulate neurotrophic factors could play a significant neuroprotective role in infants with severe CHD, both early postnatally, as well as perioperatively (9, 21). Currently, no standard neuroprotective drugs are used in infants with severe CHD. In light of the appearance of potential new neuroprotective drugs this systematic review appears to be useful.

#### Objectives

The aim was to provide a systematic review and best-evidence synthesis on the effects of neuroprotective drugs on brain injury and neurodevelopmental outcome in congenital heart disease infants requiring cardiac surgery with cardiopulmonary bypass.

#### Research Question

Which neuroprotective drugs have been studied in infants with severe congenital heart disease and what is the evidence of the effects of these agents on brain injury and neurodevelopmental outcome?

### METHODS

#### Study Design

A systematic review was performed following the steps of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (22).

### Inclusion and Exclusion Criteria

Studies reporting on the effects of neuroprotective drugs on brain injury and/or NDO in infants with severe CHD requiring cardiac surgery with CPB were included. Gestational age at birth was restricted to at least 35 weeks (near-term and term) and patients with syndromic or genetic disorders were excluded. Reviews, studies that investigated other neuroprotective strategies than neuroprotective drugs and studies not written in English language were excluded. No restriction was set on the years of publication of the articles identified.

#### Search Strategy

The main search terms were infants, congenital heart disease, cardiac surgery, cardiopulmonary bypass and neuroprotective drug. Besides title and abstract, MESH terms were used for all search terms (**Table 1**).

### Data Sources, Studies Selections and Data Extraction

PubMed, Embase and The Cochrane Library were searched in the period from inception to May 30th 2017 to identify suitable articles. Scopus and Web of Science were searched for additional


\**Truncation symbol was used to find terms with other endings or an alternative spelling.*

articles through reference screening. Citations of included articles were manually screened for relevant articles. After removal of duplicates, articles were screened on title and abstract and records not matching the inclusion criteria were excluded. The remaining articles were assessed full-text for eligibility. After exclusion of full-text articles not answering the research question, a decision was made of studies to be included in the final systematic review. The study characteristics and relevant findings of the included studies were recorded on a data extraction file. The following study characteristics were extracted: study design, number of infants, type of CHD, age at cardiac surgery with CPB, drug, moment of administration, dose and mode of administration, outcome and outcome assessment. The selection of studies and data extraction was performed independently by two researchers (RS, KDL) and any disagreements were resolved by involving a third researcher (NJGJ, AvdH).

### Data Analysis

The Cochrane Risk of Bias Tool was used to assess the methodological quality of the included studies (23, 24). The included studies were assessed on random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants/personal (performance bias), blinding of outcome assessors (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other forms of bias. The risk of bias and overall quality of the studies was assessed independently by two researchers (RS, KDL) and disagreements were resolved by involvement of a third researcher (NJGJ, AvdH). A total of six forms of bias were scored as low or high risk. Studies were considered of high quality when at least 5 forms of bias scored low risk, of moderate quality when 4 forms of bias scored low risk and of low quality in case 3 or less forms of bias scored low risk.

#### Best Evidence Synthesis

A best-evidence synthesis was performed since the outcome measures of the included studies were too heterogeneous for a meta-analysis. Both the outcome (**Table 2**) and the quality (**Figure 2**) of the included studies were taken into account.

#### RESULTS

#### Flowchart

The search identified 293 records, including 290 through database searching (Pubmed n = 83, Embase, n = 194, Cochrane n = 13) and 3 by reference screening. After removing duplicates, 216 studies were screened on title and abstract. Of these, 207 were excluded since they met the exclusion criteria and/or did not met the inclusion criteria. Nine full-text articles were assessed for eligibility of which 3 were excluded because only the abstract was available (n = 1) or no answer was given to the research question (n = 2). Finally, 6 studies were included in this systematic review (**Figure 1**: adapted from Moher et al. (22)).

#### Study Characteristics

A total of 5 randomized controlled trials (RCTs) and 1 prospective cohort study were eligible for the review comprising a total of 527 patients (ranging from 13 to 318 per study).


**53**

The age of the included infants with CHD in these studies, ranged from <30 days to 36 months. Various cardiac defects as TGA, UVH, aortic arch anomalies, VSD, and TOF were included. Six neuroprotective drugs were investigated: sodium nitroprusside (SNP), erythropoietin (EPO), ketamine, allopurinol, phentolamine, and dextromethorphan. All were administered intravenously around cardiac surgery with CPB, with the exception of dextromethorphan which was given orally by nasogastric tube. The drugs were compared with placebo, except SNP which was compared with standard care. Various outcome measures of brain injury (S100ß, neuron specific enolase, MRI, clinical seizures, coma) and NDO (Bayley II/III, Griffiths) were taken into account (**Table 2**). S100ß and neuron specific enolase are released in blood in the setting of brain injury from glial cells and neurons respectively, and are related to brain injury by MRI and early neurodevelopmental outcomes (25).

### Methodological Analysis

The overall methodological quality of the included studies was analyzed with the Cochrane Risk of Bias Tool (23, 24) There were 2 studies of high (26, 27) (EPO, allopurinol), 2 studies of moderate (28, 29) (ketamine, dextromethorpan) and 2 studies of low (30, 31) (SNP, phentolamine) methodological quality (**Table 2**, **Figure 2**).

### Sodium Nitroprusside (SNP)

In 2000, Abdul-Khaliq et al. published a prospective cohort study in which they evaluated the effect of continuous treatment with the nitric-oxide (NO-) liberator SNP on the brain injury marker S100ß in 53 neonates after cardiac surgery with hypothermic CPB for TGA. SNP was infused (1–5 microgram per kilogram bodyweight per minute depending on the hemodynamic status) after the induction of anesthesia, and during and after the termination of CPB for 2 days. SNP treated neonates (n = 25, 0.37 months) had significantly lower levels of S100ß 24 h (2.0 vs. 2.9 µg/L, p = 0.009) and 48 h (1.0 vs. 1.8 µg/L, p = 0.04) after surgery in comparison to non-treated infants (n = 28, 0.32 months). S100ß levels 24 h after surgery normalized to preoperative values in the SNP treated neonates, however remained significantly high in the non-treated infants (p = 0.01) (30). This study showed that continous low-dose treatment with the NO liberator SNP was safe and decreased the release of S100ß into the blood stream after corrective cardiac surgery with CPB for TGA infants. However, the overall quality of the study was low with both a high risk on selection bias as well on performance


Knowledge of the allocated interventions by participants and personnel during the study.

§Knowledge of the allocated interventions by outcome assessors.


¶Selective outcome reporting.

bias. Patients were not randomized and the cross-clamping time (minutes) at baseline was significantly higher in the SNPtreated group (median 98, range 50–174) compared to the standard-treated group (78, 67–114) (p = 0.004). Furthermore, parents of participating children and (treating) physicians were not blinded and were aware of the treatment-group (30).

#### Erythropoietin (EPO)

In 2013 Andropoulos et al. determined the anti-apoptotic, anti-excitatory and anti-inflammatory effects of EPO on brain injury and NDO at 12 months in a phase I/II safety and efficacy randomized, blinded, placebo-controlled trial. Fiftynine neonates (age <30 days) with TGA, hypoplastic left heart syndrome (HLHS) or aortic arch anomalies received 3 intravenous doses of EPO (500 or 1,000 U/kg) or placebo before and after hypothermic CPB. There were no differences between treated infants and controls in clinical events (as cardiac arrest, the need of extracorporeal membrane oxygenation and seizures), physical neurological examination preoperatively and before discharge, and mortality. Brain MRI was performed immediate preoperatively and postoperatively (at 7–10 days after surgery). MRI-scans were evaluated by blinded pediatric neuroradiologists and assessed on mild/moderate/severe white matter injury, intraparenchymal infarction, intraparenchymal or intraventricular hemorrhage, and sinovenous thrombosis. No differences in rate and severity of preoperative and postoperative brain injuries were observed. Neurodevelopmental testing with the Bayley Scales of Infant Development (BSITD-III, mean value 100, SD 15) at 12 months were not significantly different between the treated infants and controls, including cognitive (101 vs. 106; p = 0.19), language (89 vs. 92; p = 0.33), and motor composite scores (90 vs. 92, p = 0.51). This study showed no significant differences in safety profile (including brain injury) and NDO after perioperative EPO or placebo administration. The overall methodological quality of the study was high with only a high risk on attrition bias by incomplete outcome data (loss to follow-up 21%) (26).

#### Ketamine

In 2012, Bhutta et al. studied the effect of the anesthetizing, anti-inflammatory and anti-excitotoxic non-competitive NMDA receptor antagonist ketamine on S100ß, neuron specific enolase (NSE), brain injury (MRI and proton MR-spectroscopy) and NDO (BSITD-II) in a pilot RCT. Twenty-four infants with a mean age of 5.4 months received 2 mg/kg intravenous ketamine (n = 13) or placebo (n = 11) before CPB for VSD repair. Postoperative MR-spectroscopy showed a significant decrease in choline a marker of demylination and glutamate plus glutamine/creatine a marker of excitotoxic neuronal and glial cell death in frontal white matter of the brain. There were no structural abnormalities on pre- and post-operative MRI and no differences in S100ß and NSE at the end to 48 h after surgery. Preoperative and postoperative (2–3 weeks after surgery) BSITD-II scores showed no significant differences in mental (MDI) and psychomotor developmental index (PDI). This study did not find neuroprotective effects of ketamine on brain injury or NDO. The overall methodological quality of this study was moderate, since there was a high risk on selection bias. Significant differences in several clinical parameters (such as intraoperative cooling) were present at baseline and no procedure for allocation concealment was described (28).

#### Allopurinol

In 2001, Clancy et al. investigated in a single center, randomized, placebo-controlled, blinded trial the effects of the free radical scavenger allopurinol on clinical seizures, coma, death and cardiac events in infants undergoing cardiac surgery with DHCA. Cardiac events were defined as periods of acute, severe cardiorespiratory deterioration necessitating immediate resuscitation such as chest massage, defibrillation, and acute boluses of inotropics. A total of 318 HLHS (n = 131) and non-HLHS (n = 187, other forms of CHD than HLHS) infants (mean age 5.6 days) received intravenous allopurinol 5–20 mg/kg or placebo before, during and after surgery. There was no significant difference in the primary endpoint (death, clinical seizures, coma) between allopurinol treated infants and controls. However subgroup analysis showed that allopurinol in comparison to placebo resulted in a lower event rate of clinical seizures, coma, death and cardiac events in HLHS-infants (38 vs. 60%; OR 0.44; 95%-CI 0.21–0.91), but not in non-HLHS infants (30 vs. 27%; OR 1.17; 95%-CI 0.61–2.25). There were significantly less clinical seizures (4 vs. 18%, p = 0.05) and cardiac events (4 vs. 20%, p = 0.03) after allopurinol vs. placebo treatment in the HLHS-group, but there was no difference in mortality. In HLHS survivors allopurinol showed less endpoint events (clinical seizures, coma or cardiac event) compared to placebo (event free 85 vs. 55%, p = 0.002). Safety profile was similar between both groups. This study showed significant neurocardiac protection in HLHS infants. The overall methodological quality was high since there were no risks of bias identified (27).

#### Phentolamine

In 2003, Gazzolo et al. studied in a RCT the effect of the vasodilating non-selective catecholamine receptor blocker phentolamine on S100ß-levels and middle cerebral artery pulsatility index (MCA-PI) before, during and after surgery. Sixty patients (age 124–128 days) undergoing CHD surgery for TOF, (multiple) VSD(s), TGA or aortic stenosis received 0.2 mg/kg phentolamine (n = 30) or placebo (n = 30) before the cooling and rewarming phases of CPB. Cooling and rewarming times were shorter in the phentolamine-treated group (p < 0.01). Phentolamine treated infants had significantly higher levels of S100ß after rewarming (3.53 vs. 1.58 µg/L, p < 0.001) and at the end of surgery (2.95 vs. 0.79 µg/L, p < 0.001) than placebo-treated infants. The cerebrovascular resistance (MCA-PI values) was also significantly higher at the end of surgery in phentolamine treated infants (1.83 vs. 1.22, p < 0.01). This study showed that phentolamine administration to shorten the cooling and rewarming phases of CPB was correlated with increased brain damage and cerebrovascular resistance. The overall methodological quality was low since there was a high risk on selection, performance and detection bias. No procedure for allocation concealment was described and parents, study personnel and outcome assessors were aware and not blinded for treatment group respectively (31).

#### Dextromethorphan

In 1997, Schmitt et al. determined the effect of the noncompetitive NMDA antagonist dextromethorphan in a pilot RCT on brain injury (MRI, NSE), cerebral activity (electroencephalography or EEG) and neurodevelopmental outcome (Griffiths). Thirteen infants and children (3–36 months) with VSD or TOF received dextromethorphan 36–38 mg/kg/day (n = 6) or placebo (n = 7) by nasogastric tube before and after surgery with CPB. Pre- and post-operative MRI showed less ventricular enlargement (non-significant) in the dextromethorphan group. Periventricular white matter lesions were only seen in 2 placebo-treated children. Levels of NSE were not increased in both groups. Postoperative EEG showed significant less sharp waves in the dextromethorphan vs. placebo-group (2 vs. 7, p = 0.02). Griffiths developmental quotients (normal value 100) before the operation (mean 98 vs. 95), at hospital discharge (92 vs. 85) and after 3 months (92 vs. 93) were similar in both groups. Adverse effects were not observed. This study showed no significant neuroprotective effects of dextromethorphan on brain injury and early NDO (29). The study quality was moderate with a high risk on both selection and attrition bias. ToF was more often diagnosed in the placebo group and some outcome data were incomplete.

### Best Evidence Synthese

The current evidence of the neuroprotective effects of SNP (30), EPO (26), ketamine (28), and dextromethorphan (29) is too insufficient, due to the quality of these studies, to make any recommendation for clinical usage at this moment. Allopurinol is the only drug that may be considered as is shown in a high quality study. However the effect on structural brain abnormalities and longterm NDO is not well-established and requires further investigation (27). Phentolamine should not be recommended given the potential neurotoxic effects as shown in a study of low quality (31).

# DISCUSSION

This systematic review assessed the available evidence of potential neuroprotective drugs on brain injury and/or NDO in infants with severe CHD requiring cardiac surgery with CPB. The current evidence was limited as only 6 different drugs were studied in this population. SNP and allopurinol showed potential neuroprotective effects. EPO, ketamine and dextromethorphan showed no neuroprotective effects whereas phentolamine showed neurotoxic effects. However, the evidence of these studies was not sufficient enough to make any recommendation for usage in clinical practice. First, larger welldesigned trials are needed, for which allopurinol is a promising candidate.

Five studies were placebo-controlled (pilot) RCTs, while the study of Abdul Kahliq et al. (30) was the only one which prospectively compared SNP with standard treatment. The sample sizes were limited (n = 13–60) and only the study of Clancy et al. (27) studied allopurinol in a large number of 318 infants. A heterogeneous group of cardiac defects was studied. The age of the study participants varied and was not only limited to neonates. Ketamine (28) and phentolamine (31) were studied in infants whereas dextromethorphan (29) was studied in young children up to 36 months. All studies focused on the period around cardiac surgery with CPB and not (also) on the vulnerable perinatal and early postnatal period. Brain injury (clinical seizures, coma, S100ß, NSE, Doppler, EEG, MRI) and/or NDO (BSITD, Griffiths) were measured in different ways, making it not possible to compare the outcomes of the studies included.

The study of Abdul-Kahliq et al. (30) indicates that continues low-dose treatment with the NO-liberator SNP during and after surgery for TGA may give delayed neuroprotection by reducing astroglial cell activation and disintegration of the blood-brain barrier ("oxidative stress"). However, the overall methodological quality of the study was low. Patients were not randomized and the cross-clamping time was significantly higher in the SNPtreated group compared to the standard-treated group. This may have led to an underestimation of the neuroprotective effects of SNP in this study. Therefore, a well-designed study seems needed to evaluate the true effects of SNP on structural brain abnormalities and longterm NDO. Previous in vitro studies showed protective effects of nitric oxide on blood-brain barrier after hypoxia reoxygenation mediated injury, by effectively scavenging reactive oxygen species (32).

Andropoulos et al. (26) found no significant neuroprotective effects of perioperative EPO administration on NDO at 12 months. Despite the high quality of the study the power was not sufficient to demonstrate a NDO difference. In addition, the change in EPO dosage (from 1,000 to 500 units/kg) by the FDA during the study may have led to levels that may not be neuroprotective. Because of these limitations and the promising results of EPO on neurodevelopment-related outcomes in neonates with hypoxic-ischemic encephalopathy (HIE) (33, 34) and very low birthweight infants (35), a larger RCT would be required to definitively address the neuroprotective effects of EPO in this CHD population. In animal and in vitro models EPO protects the brain against cerebral insults and cell death by antiexcitatory, anti-inflammatory and anti-apoptotic mechanisms (36).

Bhutta et al. (28) found no significant neuroprotective effects of pre-CPB administration of Ketamine on structural brain injury and early postoperative NDO. However, the anti-excitotoxic and anti-inflammatory effects (NMDA-antagonism) of ketamine led to a significant decrease in myelin breakdown and cell death mediated excitotoxicity in the frontal white matter of the brain, as was indicated by functional MR spectroscopy. This study was of moderate quality, had a small sample size (n = 24) and showed significant baseline differences despite randomization. Apart from the results of this study doubts have recently emerged over the safety of anesthetics, including ketamine, in recent neonatal animal studies and children under the age of 3 (37, 38). Large doses of Ketamine given repeatedly or as continuous infusion for prolonged periods can induce apoptotic cell death (37). Currently, trials are underway to investigate these doserelated and exposure-time effects of anesthesia on longterm NDO in young children. Further research with ketamine in CHD neonates should be postponed, until the results of these trials are known (38).

Clancy et al. (27) showed that perioperative allopurinol administration was safe and provided a neurocardiac protective effect (death, coma, clinical seizures, cardiac events) in higherrisk HLHS infants. The methodological quality of this study was high given the low risks of bias. The neuroprotective effects in HLHS-infants were suggested by significantly fewer clinical seizures in the allopurinol vs. placebo-group. The occurrence of perioperative seizures is an early sign of new brain injury and associated with worse neurodevelopmental outcome (39). However, the effects of the xanthine-oxidase inhibitor allopurinol on amplitude integrated EEG and structural brain injury with pre- and post-operative MRI and longterm NDO were not assessed. The definitive neuroprotective effect of allopurinol in the CHD population should be demonstrated by including these study procedures in a future high quality study. The possible neuroprotective effects of allopurinol are based on several preclinical studies in rats, piglets and sheep and clinical studies in neonates with HIE (40). In neonates with HIE beneficial effects were found in three small studies (41–43) in which allopurinol was administered postnatally and a pilot (44) and multicenter (45) study in which allopurinol was administered antenatally. Longterm NDO was only beneficial after postnatal allopurinol treatment in infants with moderate HIE (46). The ALBINOtrial (NCT03162653) will investigate the neuroprotective effect of early postnatal allopurinol as add-on therapy to hypothermia on NDO in HIE-neonates.

The study of Gazollo et al. (31) indicates that phentolamine administration to shorten the cooling and rewarming phases of CPB is neurotoxic and increases brain damage. However, the study was of low quality since there was no blinding and no allocation concealment procedure was described. Notwithstanding the beneficial effects of the non-selective alpha-1, 2 receptor blocker phentolamine on the duration of CBP and surgery, it seems correlated with increased brain stress during CPB and should not be recommended. Other vasodilator agents used in this population should also be investigated on account of their possible undesired effects.

The study of Schmitt et al. (29) showed less abnormalities on EEG and MRI after perioperative administration of highdose oral dextromethorphan. However, this moderate quality study was too small and there were significant dissimilarities between the treatment groups, making conclusions about possible neuroprotective effects of the NMDA-antagonistic properties of dextromethorphan at this time not possible. Further larger well-designed studies are encouraged because of the good resorption and tolerance of orally administered dextromethorphan.

Some explanations can be given for the limited number of RCTs that have been performed concerning the effects of neuroprotective drugs in the population of severe CHD infants. First, the need for neuroprotective drugs has only recently become clear, since we now know more about the immature brain development, perinatal/perioperative brain injury, and consequently longterm neurodevelopmental impairments within this specific population. In addition, infants with severe CHD are a relatively rare (low number of patients per center) and heterogeneous population (different cardiac defects, prenatal and postnatal diagnosis). Therefore, a large number of patients is needed to perform a RCT and many centers have to be involved using the same treatment (study drug vs. placebo) and strategies at their intensive care unit and during the perioperative setting. Lastly, often funding is first needed for these RCTs which is challenging to accomplish because of the hight costs.

In recent reviews of Robertson et al. and Hagberg et al. overviews were given of neuroprotective agents in animal models and a term newborns with perinatal brain injury by hypoxic-ischemic encephalopathy, intracranial hemorrhage and stroke. Besides allopurinol and EPO, also other agents as tetrahydrobiopterin (BH4), melatonin, topiramate, nitric oxide inhibitors, xenon, N-acetylcysteine (NAC), vitamins C and E, and stem cells were discussed (9, 47). Although the pathophysiology of brain injury is different in neonates with HIE, these drugs can also apply for infants with severe CHD in the future.

Finally, a number of recommendations are worth mentioning for future studies investigating the effects of neuroprotective drugs on brain injury and NDO in infants with severe CHD. (1) Well-designed RCTs with adequate sample sizes are needed taking into account the heterogeneity of the CHD population. (2) Drugs should be administered in the neonatal period, since the greatest neuroprotective effect can be expected in this phase. This is the most vulnerable period in which the brain develops the fastest. (3) Neuroprotective drugs should be administered in both the perinatal/early postnatal phase and perioperative period. Recent research showed that brain injury occurs during both of these vulnerable phases (6, 8). (4) Brain imaging with preoperative and postoperative MRI is an important study procedure to assess brain injury quantitatively. (5) Longterm neurodevelopmental follow-up (including measurements of executive functioning) is necessary as neurodevelopmental impairments become more pronounced as these children grow up ("grow in their deficits").

#### AUTHOR CONTRIBUTIONS

All authors have made a significant contribution to the manuscript. RS and KL performed the literature search, screened and selected the studies, extracted data, assessed risk of bias, drafted the initial manuscript, and approved the final manuscript as submitted. AvdH coordinated the literature search, screening and selection of studies, extraction of data and assessment on risk of bias, reviewed and revised the manuscript, and approved the final manuscript as submitted. JB conceptualized the study, reviewed and revised the manuscript, and approved the final manuscript as submitted. FG conceptualized the study, coordinated the selection of studies, reviewed and revised the manuscript and approved the final manuscript as submitted. NJ conceptualized the study, coordinated the literature search, screening and selection of studies, extraction of data and assessment on risk of bias, reviewed and revised the manuscript, and approved the final manuscript as submitted. MB conceptualized the study, edited the final manuscript, and approved the final manuscript as submitted.

#### REFERENCES


congenital heart disease. Dev Med Child Neurol. (2017) 59:894–902. doi: 10.1111/dmcn.13461


agent in children undergoing surgery on cardiopulmonary bypass: a pilot randomized, double–blind, placebo–controlled trial. Pediatr Crit Care Med. (2012) 13:328–37. doi: 10.1097/PCC.0b013e31822f18f9


controlled trial. Arch Dis Child Fetal Neonatal Ed. (2015) 100:F216–23. doi: 10.1136/archdischild-2014-306769


**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 Stegeman, Lamur, van den Hoogen, Breur, Groenendaal, Jansen and Benders. 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.

# Dose-Dependent effect of intravenous administration of human Umbilical cord-Derived Mesenchymal stem cells in neonatal stroke Mice

*Emi Tanaka1,2, Yuko Ogawa1 , Takeo Mukai3 , Yoshiaki Sato4 , Takashi Hamazaki2 , Tokiko Nagamura-Inoue3 , Mariko Harada-Shiba1 , Haruo Shintaku2 and Masahiro Tsuji1 \**

*1Department of Regenerative Medicine and Tissue Engineering, National Cerebral and Cardiovascular Center, Suita, Japan, 2Department of Pediatrics, Osaka City University Graduate School of Medicine, Osaka, Japan, 3Department of Cell Processing and Transfusion, Institute of Medical Science, The University of Tokyo, Tokyo, Japan, 4Division of Neonatology, Center for Maternal-Neonatal Care, Nagoya University Hospital, Nagoya, Japan*

#### *Edited by:*

*Carl E. Stafstrom, Johns Hopkins Medicine, United States*

#### *Reviewed by:*

*Alexander Drobyshevsky, NorthShore University HealthSystem, United States Clotilde Des Robert, Assistance Publique Hôpitaux de Marseille, France*

*\*Correspondence:*

*Masahiro Tsuji mtsuji@ncvc.go.jp, mtsujimd@ybb.ne.jp*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 28 December 2017 Accepted: 22 February 2018 Published: 08 March 2018*

#### *Citation:*

*Tanaka E, Ogawa Y, Mukai T, Sato Y, Hamazaki T, Nagamura-Inoue T, Harada-Shiba M, Shintaku H and Tsuji M (2018) Dose-Dependent Effect of Intravenous Administration of Human Umbilical Cord-Derived Mesenchymal Stem Cells in Neonatal Stroke Mice. Front. Neurol. 9:133. doi: 10.3389/fneur.2018.00133*

Neonatal brain injury induced by stroke causes significant disability, including cerebral palsy, and there is no effective therapy for stroke. Recently, mesenchymal stem cells (MSCs) have emerged as a promising tool for stem cell-based therapies. In this study, we examined the safety and efficacy of intravenously administered human umbilical cord-derived MSCs (UC-MSCs) in neonatal stroke mice. Pups underwent permanent middle cerebral artery occlusion at postnatal day 12 (P12), and low-dose (1 × 104 ) or high-dose (1 × 105 ) UC-MSCs were administered intravenously 48 h after the insult (P14). To evaluate the effect of the UC-MSC treatment, neurological behavior and cerebral blood flow were measured, and neuroanatomical analysis was performed at P28. To investigate the mechanisms of intravenously injected UC-MSCs, systemic blood flowmetry, *in vivo* imaging and human brain-derived neurotrophic factor (BDNF) measurements were performed. Functional disability was significantly improved in the high-dose UC-MSC group when compared with the vehicle group, but cerebral blood flow and cerebral hemispheric volume were not restored by UC-MSC therapy. The level of exogenous human BDNF was elevated only in the cerebrospinal fluid of one pup 24 h after UC-MSC injection, and *in vivo* imaging revealed that most UC-MSCs were trapped in the lungs and disappeared in a week without migration toward the brain or other organs. We found that systemic blood flow was stable over the 10 min after cell administration and that there were no differences in mortality among the groups. Immunohistopathological assessment showed that the percent area of Iba1-positive staining in the peri-infarct cortex was significantly reduced with the high-dose UC-MSC treatment compared with the vehicle treatment. These results suggest that intravenous administration of UC-MSCs is safe for a mouse model of neonatal stroke and improves dysfunction after middle cerebral artery occlusion by modulating the microglial reaction in the peri-infarct cortex.

Keywords: mesenchymal stem cell, umbilical cord-derived mesenchymal stem cell, neonatal stroke, neonatal brain injury, intravenous administration

# INTRODUCTION

Neonatal stroke is a common cause of acute neonatal encephalopathy and frequently results in neurological impairments such as cerebral palsy, cognitive disorders, and seizures (1, 2). The incidence of neonatal arterial ischemic stroke is 1 per 3,500 to 7,700 neonates, and the frequency of neonatal arterial ischemic stroke diagnosis has increased due to the advancement of imaging techniques (3, 4). Although neonates with stroke may present features of hypoxic-ischemic encephalopathy (HIE) (5, 6), therapeutic hypothermia has not currently been proven effective for neonatal stroke despite its efficacy in HIE. In addition, most neonatal stroke patients are recognized only retrospectively with emerging seizures or hemiparesis days after birth, hence the patients are likely to miss the opportunity of receiving acute neuroprotective treatments. There is currently no effective treatment for neonatal stroke. A novel therapeutic strategy to improve the outcome of neonatal stroke is needed.

Recently, many researchers have focused on cell therapies as novel treatments for neonatal brain injury. Our group has focused on the use of umbilical cord blood (UCB) for its advantage in feasibility to clinical applications (7), and previously we have shown the benefit of systemic administration of human UCB cells in animal models of neonatal encephalopathy (8, 9). Subsequently, we are conducting a clinical trial of autologous UCB transplantation for neonates with HIE (National Institutes of Health, ClinicalTrials.gov: NCT02256618). However, more than half of neonatal stroke patients do not present signs or symptoms at birth, hence physicians and parents may not recognize the need for preserving their own UCB. Even if the need is recognized, it is not always possible to collect their own UCB, especially when babies are unexpectedly born with severe asphyxia. An alternative cell source for neonatal brain injury is needed, and mesenchymal stem cells (MSCs) have emerged as a promising candidate. MSCs are isolated from umbilical cord, UCB, and placenta, as well as from bone marrow and adipose tissue, and they can be easily expanded in culture (10). Regardless of the cell source, each MSC population has a self-renewal capacity, a multi-lineage differentiation ability, and the potential for migration toward an inflamed or injured site (11, 12).

Umbilical cord-derived MSCs (UC-MSCs) have been proposed as a preferable cell source for regenerative medicine and immunotherapy, as they possess faster self-renewal, higher multipotency, and less immunogenicity than the features of born marrow-derived MSCs (13–16) and adipose-derived MSCs (17, 18). Considering these advantages and the demand for a cell source for neonates who do not have their own UCB, we are preparing to produce clinical grade off-the-shelf UC-MSCs for the next clinical application and for broad spectrum use in neonates with brain injuries. We have shown that intravenous administration of UC-MSCs ameliorates injuries via attenuating reactive gliosis and hypomyelination in a neonatal mouse model of intraventricular hemorrhage (IVH) (19). However, there is no study on UC-MSC in a neonatal stroke model. Regarding the injection route, most studies in models of neonatal brain injury have used local administration of MSCs, such as intraventricular and intranasal injection (20–23). There is a limited number of studies on intravenous MSC treatment for neonatal brain injury, one in sheep model (24) and four in rat models (25–28), and there is no study in mouse model except for our IVH report (19). With respect to studies on UC-MSCs for treatment of neonatal brain injury, most of them examined the effects of a focal injection (29–31). In clinical settings, intravenous administration is more desirable for unstable, sick neonates because it does not require an additional invasive procedure. Our previous study showed that cell distributions vary with injection routes, and more cells were trapped in the lungs when injected intravenously (32). Therefore, it is necessary to demonstrate the safety and efficacy of intravenous administration of UC-MSCs in neonatal animal models. In addition, studies of dosing regimens were limited mainly to UCB cells (33–36), evidence on intravenous MSC dosing is lacking. The main objective of this study was to investigate whether intravenous administration of UC-MSCs is safe and whether this treatment can attenuate brain damage and to determine the optimal dose for ameliorating the neurodevelopmental deficits after neonatal stroke.

### MATERIALS AND METHODS

### Cell Preparation

Human umbilical cord tissues were obtained from women who underwent cesarean sections after informed consent was obtained. UC-MSCs were isolated as described by Mori et al. (37). Briefly, UC-MSCs were isolated by an improved explant method, and fragments were cultured with RM medium (kindly provided by ROHTO Pharmaceutical Co., Ltd., Osaka, Japan), which is a serum-free culture medium, at 37°C with 5% CO2. After confluence, adherent cells were trypsinized and replated (passage 1). The cells of fourth passage were cryopreserved in a cryoprotectant, STEM-CELLBANKER (ZENOAQ Resource Co., Ltd., Fukushima, Japan) which contains 10% dimethyl sulfoxide (DMSO). The expanded UC-MSCs were validated for their differentiation potential and cell surface molecules as previously described (38). Cells were rapidly thawed in a 37°C water bath just before use without washing. Of note, this cryoprotectant with DMSO has been proven safe for clinical use. When mostly thawed into liquid, the cells in a tube were kept on ice and used within 1–2 h. The viability of UC-MSCs was 91.75 ± 8.30%. The cryoprotectant was used as vehicle in the control group.

#### Neonatal Stroke Model

All animal research studies were approved by the Experimental Animal Care and Use Committee of the National Cerebral and Cardiovascular Center and was done according to the NIH Guide for the Care and Use of Laboratory Animals.

CB17 male and female mouse pups were used in the experiments (*n* = 90). Postnatal day 12 (P12) pups, which are thought to be equivalent to full-term human newborns at P0 (39, 40), were divided into a no-surgery control group (*n* = 6), a sham-surgery group (*n* = 12), and middle cerebral artery occlusion (MCAO) groups (*n* = 72). The pups were subjected to permanent MCAO as described previously (41). Under isoflurane anesthesia (4.0% for induction and 1.5–2.0% for maintenance), a hole was made in the left temporal bone. The left middle cerebral artery (MCA) was electrocauterized and disconnected just distal to its crossing of the olfactory tract. Pups in the sham group underwent open-skull surgery without MCA electrocoagulation. After the insult, mice were observed for any bleeding and was awakened in the 32°C infant-warmer and then returned to their dams. All mice tolerated the MCAO procedure, and there was no surgical mortality. The numbers of mouse pups used in each cohort are summarized in **Table 1**. The blood flow measurement, a battery of behavioral tests, and morphological evaluations were done in cohort A. A flow cytometry beads assay and *in vivo* imaging were done in cohorts B and C, respectively. All the other experiments were performed in cohort A (**Figure 1**).



*Mice in cohort A were used for the blood flow measurement, behavioral, and histological evaluations; cohort B for blood and cerebrospinal fluid (CSF) sampling; and cohort C for in vivo imaging of injected cells.*

#### Cell Administration

At 48 h after the MCAO procedure (at P14), mice were randomly divided into three groups: vehicle (*n* = 23), low-dose UC-MSCs (1 × 104 cells, *n* = 13), and high-dose UC-MSCs (1 × 105 cells, *n* = 36). From our previous study with a cell therapy (8) and from a clinical stand point, we thought that 48 h after the MCAO would be the optimal timing of cell administration for neonatal stroke; neonatal stroke is rarely diagnosed on the first day of life, rather diagnosed a few days after the birth in many cases. Pups in the UC-MSC groups were administered frozen-thawed UC-MSCs in 60 µl of STEM-CELLBANKER cryoprotectant. Pups in the vehicle group were administered cryoprotectant alone. After incising the skin in the left inguinal region, UC-MSCs were infused or vehicle was infused into the femoral vein using a 35-G needle under isoflurane anesthesia, slowly administered over 1 min as described previously (32). Under the microscope, we were able to directly observe the injection flow through the transparent vein when the needle insertion was optimal.

#### Systemic and Cerebral Blood Flowmetry

Blood flow of the body surface and cerebral surface were measured by a laser speckle flowmetry imaging system (OmegazoneOZ-1, Omegaweve Inc., Tokyo, Japan). Flowmetry of the body surface as a representative of systemic blood flow was performed before

(P14), they were administered vehicle or umbilical cord-derived mesenchymal stem cells (UC-MSCs) intravenously (i.v.). (Cohort A) Systemic and cerebral blood flows were measured at P14 and P15, respectively. Neurological behavioral measurements were performed at P15–P23. After sacrifice, morphological and immunohistopathological assessments were conducted in this cohort. (Cohort B) Mouse serum and cerebrospinal fluid (CSF) were collected at P15 for evaluation of a trophic factor. (Cohort C) *In vivo* imaging was performed at different time points after UC-MSC administration: 3 h, 24 h, 48 h, 4 days, and 7 days after injection.

and immediately after injection. First, under isoflurane anesthesia, pre-flow images were acquired every 5 s, and vehicle was administered or UC-MSCs were administered over 1 min followed by 30 s hemostasis with a swab. Second, post-flow images were sequentially taken every 5 s for 10 min. We set the region of interest in the chest, and the average blood flow per minute was compared with that from the pre-flow measurements (vehicle *n* = 7, low-dose *n* = 8, and high-dose *n* = 8).

Cerebral blood flow (CBF) was measured 24 h after injection (at P15) as described previously (42). We evaluated the three ROIs such as ischemic core, penumbra and MCA region (**Figure 2C**). Data were presented as CBF ratio, ipsilateral/contralateral hemisphere (sham *n* = 10, vehicle *n* = 10, low-dose *n* = 10, high-dose *n* = 12).

#### Behavioral Tests

To evaluate the effects on motor and sensory functions, the following four behavioral tests were performed: a cylinder test, a dynamic plantar test, a rotarod test, and an open-field test.

The cylinder test was performed on P15. Asymmetry of forelimb use was assessed during rearing in a transparent acrylic cylinder (43). We videotaped 20 rearing movements, counted wall touches by each forelimb separately, and analyzed forepaw use preference as follows: (nonimpaired side [left] − impaired side [right])/(nonimpaired + impaired sides) × 100 (sham *n* = 12, vehicle *n* = 18, low-dose *n* = 13, and high-dose *n* = 14).

The dynamic plantar test was performed on P16. We measured responses to von Frey filaments to assess sensory function as described previously (44, 45). Mice were acclimated in an elevated mesh floored cage and tested using the Dynamic Plantar Aesthesiometer (Ugo Basile, Varese, Italy). A filament was pushed up mechanically from beneath each paw, which increased in force gradually from 0 to 40 g over the course of 50 s until the mouse withdrew its paw. Each mouse received three trials per paw, the withdrawal time was noted, and the average was calculated as follows: (nonimpaired side − impaired side)/

(nonimpaired + impaired sides) × 100 (sham *n* = 12, vehicle *n* = 13, low-dose *n* = 13, and high-dose *n* = 15).

For the rotarod test, sensorimotor skills were evaluated on P16 and P23. The rotarod accelerated from 4 to 40 rpm over 5 min (Ugo Basile, Varese, Italy) (8). The length of time that the mice remained on the rotarod was measured. Five measurements with longer than 3-min intervals were performed, and the average time was recorded.

For the open-field test, locomotor and exploratory behaviors were evaluated at 8–10 days after the insult (P20–P22) as described previously (8). Animals were allowed to act freely in a box (30 cm × 30 cm) for 30 min in the light and for 30 min in the dark (Taiyo Electric Co., Ltd., Osaka, Japan). Infrared beams were mounted at specific intervals on the X-, Y-, and Z-banks of the open-field area. The number of beam crossings by an animal was counted and scored as locomotion for the horizontal movement and as rearing for the vertical movement.

#### Tissue Preparation and Morphology

A morphological evaluation of the brain was performed as previously described (46, 47). On P28, mice were sacrificed with pentobarbital i.p. and transcardially perfused with phosphate-buffered saline and then with 4% paraformaldehyde. Subsequently, the whole brain was removed and immersed overnight in the same fixative. Fixed brains were cut coronally into 1-mm slices and the hemispheric areas were measured using ImageJ software (National Institutes of Health, Bethesda, MD, USA). The hemispheric volume was estimated by integration of the hemispheric areas. The cerebral hemispheric volume ratio was calculated as follows: viable ipsilateral hemispheric volume/contralateral hemispheric volume.

#### Immunohistochemistry

The processed brains were embedded in paraffin, and 5-µm coronal sections were prepared for every 1 mm. For immunohistochemical detection, anti-mouse GFAP (glial fibrillary acidic protein, a marker of astrocytes) antibody (1:100; Merck Millipore, Burlington, VT, USA), and anti-rabbit Iba1 (ionized calcium binding adaptor molecule, a marker of microglia 1) antibody (1:200; WAKO, Osaka, Japan,) were used. The secondary antibody was a goat-anti-mouse IgG polyclonal antibody (Nichirei Bioscience Inc., Tokyo, Japan). Slices were stained with 3,3′-diaminobenzidine (DAB) with hydrogen peroxide. The GFAP-positive astrocytes and Iba1-positive microglia were segmented by applying an appropriate threshold in gray value in order to distinguish from non-specific background staining. The percent areas (segmented area/total 400 μm × 400 μm area) were calculated using ImageJ software. The regions of interest (ROIs), i.e., square frame, were laid not to include tissue absent area or necrotic area by an examiner who was blinded to the experimental group of the mice. Four regions were determined from a section containing both the hippocampus and striatum; the peri-infarct cortex, hippocampus, subcortical white matter, and non-infarct cortex in an area of 400 μm × 400 μm. We calculated the percent area for two ROIs in one slice per region per animal (*n* = 5 per group).

#### Flow Cytometry Beads Assay

In cohort B (**Figure 1**), the no-surgery control and MCAO mice were prepared. The MCAO mice were subjected to intravenous administration of vehicle or UC-MSCs (1 × 105 cells; control *n* = 6, vehicle *n* = 6, and UC-MSCs *n* = 13). Mouse serum and cerebrospinal fluid (CSF) were collected 24 h after injection, and the concentration of human brain-derived neurotrophic factor (BDNF) secreted from human UC-MSCs was measured using a bead immunoassay as previously described (19). CSF was collected by puncture of cisterna magna with a 25-G needle and a capillary glass under microscope. Subsequently, blood was collected by cardiac puncture using a 1 ml syringe and a 26-G needle, and centrifuged to collect the serum. We quantified human BDNF with the HQ-Plex kit (Bay Bioscience Co., Ltd., Kobe, Japan). All samples were analyzed in triplicate according to the manufacturer's instructions. Bead fluorescence readings were done by a BDTM FACS Canto II flow cytometer (BD Biosciences, San Jose, CA, USA).

#### *In Vivo* Imaging of UC-MSCs

In cohort C (**Figure 1**), eight MCAO mice received intravenous administration of vehicle (*n* = 1) or 1 × 105 UC-MSCs (*n* = 7) 48 h after insult. UC-MSCs were traced using an IVIS Lumina II imaging system (Xenogen Corporation, Almeda, CA, USA) at 2 h, 24 h, 48 h, 4 days, and 7 days after the intravenous injection. Mice were given an intraperitoneal injection of d-luciferin (WAKO, Osaka, Japan) at a dose of 150 mg/kg 10 min before each imaging time point. Images were captured with an exposure time of 1 min under isoflurane anesthesia in a chamber, and images were analyzed using Living Image 3.0 software (Caliper Life Sciences, Waltham, MA, USA).

#### Statistical Analysis

The mortality rate of the animals was analyzed using the Fisher's exact test. Calculated values such as CBF ratios, the cylinder and plantar test results, cerebral hemispheric volumes, and percent areas of immunohistological staining were assessed using a Kruskal–Wallis test followed by Dunn's test. Absolute values or temporal sequences that include systemic blood flow and the rotarod and open-field test results were assessed using two-way repeated measures ANOVA. The significance level was established at *p* < 0.05. The results are presented as the mean ± SEM in bar graphs or box plots (maximum, quartile point, medium, and minimum). All data were analyzed using JMP 12.2.0 software (SAS Institute, Cary, NC, USA).

#### RESULTS

#### Mortality

In this study, 62 mice survived for two weeks after the insult, and seven died (**Table 2**). One mouse in the high-dose UC-MSC group died approximately 10 min after induction of anesthesia and a few minutes after cell administration. The rest died during a remote period approximately 7 days after the insult, indicating that the deaths may not be due directly to cell administration but rather due to weakness from the MCAO injury.

#### Table 2 | Mortality rate.


*Mortality rates at 24 h (postnatal day 15, P15) and 2 weeks (P28) after the vehicle or umbilical cord-derived mesenchymal stem cell (UC-MSC) injection were not different between the groups.*

# Blood Flowmetry

#### Systemic Blood Flow

Blood flow of the thoracic body surface was measured to assess systemic blood flow after cell administration. The blood flow in all groups was slightly increased and returned gradually to the pre-injection level after 10 min (**Figures 2A,B**). These changes in blood flow were almost concordant and were not different among the groups. This indicates that the systemic blood flow was not exacerbated by UC-MSC administration.

#### Cerebral Blood Flow

Twenty-four hours after cell injection (P15), the CBF was measured. CBF was decreased in the ipsilateral hemisphere of MCAO mice, and the decrease was significant in all assessed ROIs, including the ischemic core, penumbra, and MCA region (**Figures 2C,D**) compared with the CBF in the contralateral hemisphere. The UC-MSC injection did not increase the CBF. The two pups that showed the lowest and the second lowest CBF ratio, i.e., <0.6, died the week after the insult; one was in the vehicle group and the other was in the high-dose group.

#### Behavioral Tests

We observed forelimb asymmetry to evaluate motor function by the cylinder test at P15. In this test, sham mice did not show paw preference (**Figure 3A**). MCAO caused significant asymmetry to use the left unimpaired forepaw in the vehicle group. After administration of high-dose UC-MSCs, performance in the cylinder test was significantly improved in comparison to the performance of the vehicle mice.

At P16, the dynamic plantar test was performed. Delayed withdrawal time from the filament was measured in forepaws and hind paws to assess the extent of sensory deficit. The changes were more obvious in the forepaw performance (**Figure 3B**). The vehicle mice showed significant asymmetrical dullness in the forepaw compared with that in the sham mice. Although the mice treated with UC-MSCs did not show significant improvement in performance compared with that in the vehicle mice, the high-dose UC-MSC mice showed a trend toward improvement (*p* = 0.092) in forepaw performance. There was no significant difference in hind paw sensory preference between the sham and vehicle groups (data not shown).

We performed the rotarod test at two time points: P16 and P23. The average number of times fallen from the rotarod cylinder shows sensorimotor capacity. There were no significant differences among the groups (**Figure 3C**).

Additionally, we evaluated spontaneous activity in an openfield test during P20–22. There were no differences among the groups (**Figures 3D,E**).

#### Morphology

Cerebral hemispheric volume was estimated in fixed brain tissue. The volumes were consistent across all animals in the sham group (**Figures 4A,B**). MCAO caused a significant volume loss in the vehicle group. There was no improvement in the volume by UC-MSC administration compared with the volume in the vehicle group.

#### Immunohistopathology

We observed inflammation in sections of the injured brains at P28. MCAO significantly increased the percent areas of GFAP- and Iba1-positive cells in the peri-infarct cortex and subcortical white matter in the vehicle group compared with those in the sham group. There was no significant increase in the areas of the hippocampus or non-infarct cortex of the vehicle group. In the peri-infarct cortex, high-dose UC-MSCs significantly reduced the Iba1-positive percent area, and low-dose UC-MSCs also showed the same trend toward reduction (*p* = 0.083). Therefore, this result shows that high-dose UC-MSCs decreased the microglial accumulation in the periinfarct cortex (**Figure 5**).

#### Detection of Human BDNF *In Vivo*

We assessed the presence of the neurotrophic factor BDNF from human UC-MSCs. Both serum and CSF were assessed 24 h after injection (P15). There was no detection of BDNF in the control group (**Figure 6**). Only one of the 13 mice treated with UC-MSCs showed an elevated level of human BDNF in CSF. A few mice had minimal equivocal levels of BDNF, which were inconclusive as to whether the observed increases were meaningful.

#### *In Vivo* Imaging

An *in vivo* imaging system was used to quantify the photon flux in the vehicle- and UC-MSC-treated groups at 3 h, 24 h, 48 h, 4 days, and 7 days after injection. The images revealed that intravenously injected UC-MSCs were rapidly trapped in the lungs. The intensity of photon flux was highest in the first image acquisition of the UC-MSC mice and then decreased over time, disappearing by 7 days. We did not detect the signals in other organs such as the brain or spleen (**Figure 7**).

#### DISCUSSION

In this study, we demonstrated that intravenous administration of human UC-MSCs following MCAO in neonatal mice was safe and attenuated damages in neurodevelopmental behaviors and glial cell reaction after neonatal stroke. We found that systemic blood flow was stable during 10 min after cell administration despite most UC-MSCs being trapped in the lungs and that there were no differences in mortality among the groups. Although the CBF and cerebral hemispheric volume were not restored after UC-MSC administration, neurological performance was significantly improved.

asymmetrical forepaw use was observed in the vehicle group. After administration of high-dose umbilical cord-derived MSCs, forepaw preference significantly improved. Wall touches by the left of right forepaw were measured separately, and paw preference with wall touches was calculated as (nonimpaired side [left] − impaired side [right])/(nonimpaired + impaired sides) × 100. \**p* < 0.05 (sham *n* = 12, vehicle *n* = 18, low-dose *n* = 13, and high-dose *n* = 14). (B) The dynamic plantar test was performed at P15 to assess the extent of sensory deficit. The forepaw of vehicle mice showed significant asymmetrical dullness compared with that of the sham mice. The high-dose UC-MSC mice showed a trend toward improvement. Average withdrawal time was calculated as (nonimpaired − impaired)/ (nonimpaired + impaired) × 100. \**p* < 0.05 and # *p* = 0.092 (sham *n* = 12, vehicle *n* = 13, low-dose *n* = 15, and high-dose *n* = 15). (C) In the rotarod test at P16 and P23, there were no significant differences among the groups (P16: sham *n* = 12, vehicle *n* = 13, low-dose *n* = 13, and high-dose *n* = 15; and P23: sham *n* = 10, vehicle *n* = 12, low-dose *n* = 13, and high-dose *n* = 11). (D,E) The open-field test was performed from P20 to P22. There were no significant intergroup differences with respect to either locomotion, i.e., spontaneous horizontal movement, or rearing, i.e., spontaneous vertical movement (*n* = 11–13 in each group).

This is the first report from the following two standpoints: (1) UC-MSCs were used in a neonatal stroke model, and (2) two cell doses were compared in a scheme of intravenous MSC injection for neonatal brain injury. To our knowledge, there is no study on the use of UC-MSCs for neonatal stroke. There are a few studies in the literature that examined the effects of UC-MSCs for neonatal brain injury and showed functional recovery (19, 27–30). As there is no proven effective therapy for neonatal stroke, we investigated whether UC-MSCs improve neurological outcomes of neonatal stroke.

No report has demonstrated an optimal dose of intravenous MSCs injection in a neonatal brain injury model. Our current study aimed to determine an optimal MSCs dose for neonatal brain injuries that can be translated into clinical settings. To this end, two different doses of UC-MSC were compared for their effectiveness and safety in the behavioral tests and histological assessments. The higher dose was chosen based on the results of our previous study where intravenous administration of 1 × 105 human UCB CD34<sup>+</sup> cells was found beneficial in the same mouse model of neonatal stroke (8). The lower dose of 1 × 104 cells was chosen to see whether these beneficial effects could be obtained even after lowering dosage. In light of clinical translation, lower dose is considered preferable because of limited availability of these cells and cost. Although there were no differences in the CBF or morphology, mice treated with the high-dose of UC-MSCs (1 × 105 cells) showed a significant improvement in neurological function and glial responses after MCAO, and mice treated with the low-dose of UC-MSCs (1 × 104 cells) showed a trend toward improvement. Studies of MSC dosing are limited only to focal injection (48–50); Donega et al. showed that

0.5 × 106 bone marrow-derived MSCs was the minimal effective dose in HIE model mice when administered intranasally. With regard to mononuclear cells (33–36), there are some reports of intravenously injected cells which also show dose effects. Taken together, a relatively high dose of UC-MSCs may be needed to exert therapeutic effects in neonatal stroke mice.

Another strength of this study is the route of intravenous administration to neonatal mice. This study showed that intravenously injected UC-MSCs did not increase mortality or decrease systemic blood flow in these small mice; although, *in vivo* imaging demonstrated that the administered cells were mainly trapped in the lungs. Intravenous administration of UC-MSCs is studied in many clinical trials in adults without major adverse events (51). However, considering the size and adhesive characteristics of MSCs, there is a concern about whether the intravenous injection is safe for clinically unstable newborns. Six reports were published that studied intravenous injection of MSCs in models of neonatal brain injury (19, 24–28). Our previous report is the only study performed in mice (19). Similar to the immunohistological evaluation in our previous study, *in vivo* imaging in the present study demonstrated that intravenously injected US-MSCs were mainly trapped in the lungs (32). We were concerned about the possibility of a pulmonary embolism caused by the administered cells, so we measured systemic blood flow before and after the intravenous injection to rule out that possibility. Although a study of large animals demonstrated that the oxygen saturation after intravenous transplantation of MSCs was not significantly changed (52), there is no such study in neonatal mouse model. The stability in systemic blood flow suggests that the intravenous injection of UC-MSCs does not cause significant blood vessel embolism and is safe for newborn humans as well.

We studied how intravenously injected UC-MSCs distribute and act in the injured brain. Contrary to expectations, only one of the 13 mice treated with human UC-MSCs exhibited an increase in human BDNF in serum and CSF, and *in vivo* imaging showed that no animal exhibited migration of UC-MSCs toward the brain. We previously showed that intravenous administration of human UC-MSCs increased the levels of human BDNF and nerve growth factor in serum and CSF in about half of the mouse pups that were subjected to IVH at P5 (19). The previous results demonstrated the time-dependent migration of UC-MSCs toward the brain at 3 days after the injection, whereas this study did not detect the cell migration toward the brain. This contradiction may be due to differences in the pathophysiology (hemorrhage-induced vs. ischemia-induced injury) and the ages of animals used. Extracellular hemoglobin and their metabolites from IVH cause inflammation and contribute to mRNA up-regulation of pro-inflammatory cytokines after 72 h (53). These cytokines increase blood-brain barrier (BBB) permeability, leading to a consequent increase in endothelial adhesion molecules and modifications of the molecular composition/functional state of tight junctions. In contrast, a report showed that MCAO-induced BBB dysfunction was rather moderate and stable than that by LPS- or cold-induced injury (54); although, the report did not directly compare the IVH and MCAO methods. We measured human BDNF excreted from exogenous UC-MSCs but not endogenous murine BDNF. Ahn et al. showed a protective effect of human UCB-derived MSCs and an increase in endogenous BDNF expression 2 and 5 days after an intraventricular injection to IVH model rats (55). Taken together, the variety of inflammatory responses among different models critically influences the fate and actions of MSCs, and therapeutic effect may be induced mainly by the endogenous response in the recipient and not by direct action of the donor cells.

As there was no evidence of migration toward the injured brain tissue or restoration in CBF, we do not consider that cell replacement and improved cerebral perfusion contributed to the neurological improvement in this study. Similar to our current results, studies in the literature often show behavioral improvement without neuroanatomical improvement. One of the mechanisms by which UC-MSCs exert neuroprotective/neurorestorative effects was via a significant reduction in microgliosis in the peri-infarct cortex by high-dose UC-MSC injection, which modulates neuro-inflammation. Although cell replacement is considered a pivotal component in stem cell therapies including

MSC therapies, accumulating studies have focused more on immunomodulation of the inflammatory microenvironment (56). Zhu et al. injected UC-MSCs intraperitoneally to a neonatal model of periventricular white matter damage and showed a decrease in reactive astrocytes and activated microglia in white matter (29). In post-ischemic tissue, M1 microglia are dominant and contribute to the inflammatory cascade and propagate cell death beyond the initial ischemic region (57, 58). With regard to macrophages, MSCs release factors that switch them from proinflammatory type 1 to an anti-inflammatory type 2 phenotype (59, 60) and produce TGF-β, which promotes the induction of Treg cells, ultimately leading to immune tolerance (61). Donega et al. demonstrated that transplanted MSCs modulate microglia to the immunosuppressive M2 type in neonatal ischemic brain injury (49). Our results also indicate that a reduction in microgliosis by intravenously injected UC-MSCs modulates the inflammatory cascade, which may contribute to the therapeutic efficacy rather than cell replacement. Further studies are needed to clarify mechanisms of stem cells actions, which may include electrophysiological and microstructural studies.

In conclusion, intravenous administration of UC-MSCs was safely performed in the small animal model of neonatal stroke mice, and the high dose (1 × 105 ) of UC-MSCs improved functional outcomes. We consider that one of the underlying mechanisms of the therapeutic effect was microglial immunomodulation.

#### ETHICS STATEMENT

All animal research was approved by the Experimental Animal Care and Use Committee of the National Cerebral and Cardiovascular Center and was done according to the NIH Guide for the Care and Use of Laboratory Animals.

### AUTHOR CONTRIBUTIONS

ET performed the experiments (behavioral tests, etc.), analyzed data, prepared the figures, and wrote the manuscript. YO and TM performed the experiments (behavioral tests, etc.) and contributed to the analysis and interpretation of data. MT designed the study, performed the experiments (animal model, etc.), and critically reviewed the manuscript. YS, TH, TN, MS, and HS supervised the project and revised the manuscript critically

#### REFERENCES


for important intellectual content. All authors approved the manuscript.

### ACKNOWLEDGMENTS

We would like to thank Satoshi Saito and Yumi Yamamoto for useful discussions and providing their expertise on experimental techniques. We thank Ritsuko Maki, Natsuki Hanada and Manami Sone for their excellent technical assistance.

### FUNDING

The research was supported by the Joint Research Project of the Institute of Medical Science, the University of Tokyo (2017-1017), and the Japan Agency for Medical Research and Development (AMED) under Grant Number JP17gk0110022.

encephalopathy. *Lancet* (2003) 361(9359):736–42. doi:10.1016/s0140-6736(03) 12658-x


severe intraventricular hemorrhage in newborn rats. *Cell Transplant* (2017) 26(1):145–56. doi:10.3727/096368916X692861


**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 potential conflicts of interest.

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

# TWeaK receptor Deficiency has Opposite effects on Female and Male Mice subjected to neonatal hypoxia–ischemia

*Anton Kichev1 \*, Ana A. Baburamani1 , Regina Vontell1 , Pierre Gressens1,2, Linda Burkly3 , Claire Thornton1 and Henrik Hagberg1,4,5*

*1Perinatal Brain Injury Group, Centre for the Developing Brain, School of Biomedical Engineering and Imaging Sciences, Kings College London, King's Health Partners, St. Thomas' Hospital, London, United Kingdom, 2PROTECT, INSERM, Université Paris Diderot, Sorbonne Paris Cité, Paris, France, 3Department of Neuroinflammation, Biogen, Cambridge, MA, United States, 4Perinatal Center, Institute of Clinical Sciences, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, 5Perinatal Center, Institute of Neuroscience and Physiology, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden*

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Susan Cohen, Medical College of Wisconsin, United States Claudia Espinosa-Garcia, Emory University, United States*

> *\*Correspondence: Anton Kichev anton.kichev@kcl.ac.uk*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 27 October 2017 Accepted: 23 March 2018 Published: 12 April 2018*

#### *Citation:*

*Kichev A, Baburamani AA, Vontell R, Gressens P, Burkly L, Thornton C and Hagberg H (2018) TWEAK Receptor Deficiency Has Opposite Effects on Female and Male Mice Subjected to Neonatal Hypoxia–Ischemia. Front. Neurol. 9:230. doi: 10.3389/fneur.2018.00230*

Tumor necrosis factor (TNF)-like weak inducer of apoptosis (TWEAK) is a multifunctional cytokine member of the TNF family. TWEAK binds to its only known receptor, Fn14, enabling it to activate downstream signaling processes in response to tissue injury. The aim of this study was to investigate the role of TWEAK signaling in neonatal hypoxia– ischemia (HI). We found that after neonatal HI, both TWEAK and Fn14 expression were increased to a greater extent in male compared with female mice. To assess the role of TWEAK signaling after HI, the size of the injury was measured in neonatal mice genetically deficient in Fn14 and compared with their wild-type and heterozygote littermates. A significant sex difference in the Fn14 knockout (KO) animals was observed. Fn14 gene KO was beneficial in females; conversely, reducing Fn14 expression exacerbated the brain injury in male mice. Our findings indicate that the TWEAK/Fn14 pathway is critical for development of hypoxic–ischemic brain injury in immature animals. However, as the responses are different in males and females, clinical implementation depends on development of sex-specific therapies.

#### Keywords: TWEAK, Fn14, perinatal brain injury, hypoxia–ischemia, brain, TweakR, sex differences

#### INTRODUCTION

Hypoxia–ischemia (HI) remains an important etiology of brain injury in term infants suffering from intrapartum asphyxia, and in preterms suffering from hypoxic and hypotensive exposures (1, 2). Children from both sexes are at risk of developing brain injury; however, sex differences in the response to HI especially among preterm children are well described (3–5). Most of the available data to date are suggesting that male sex is significant risk factor and male infants are more vulnerable to HI insult.

The pathophysiology is complex and multifactorial. Mechanisms such as mitochondrial dysfunction, oxidative/nitrosative stress, apoptosis, necroptosis, and inflammatory processes are involved (6). Experimental data support the concept that components in the immunoinflammatory reaction contribute to cell death after HI (7). However, our understanding of which inflammatory mediators induce damage of neurons and oligodendroglia precursor cells (OPCs) in neonatal HI is incomplete. It is important to mention that inflammatory responses to injury may be different in males and females as sex differences in microglial response in various neurodevelopment and neurodegenerative disorders have been reported (8–10). This difference is likely to contribute to the differences in outcome after HI insult observed in clinical studies between males and females.

During the inflammatory cascade, infiltrating cells and glial cells resident in the brain (part of both the adaptive and innate immune system) produce reactive oxygen species, release excitotoxic amino acids, pro-inflammatory cytokines, and chemokines (11, 12). Among the prominent cytokine pathways involved in cell death is a subset of the tumor necrosis factor (TNF) family, including TNF-α, TNF-β, FasL, TRAIL, and TWEAK (11, 13–17). We hypothesized that TNF superfamily members participate in the development of injury after HI since several TNF ligand– receptor pairs have been identified in the injurious cascade (18). TNF-α (19) and Fas signaling (13, 20, 21) are already implicated in HI brain injury, but there is limited knowledge about the role of TWEAK in the evolution of neonatal brain injury.

TWEAK is a ubiquitously expressed cytokine and the only confirmed signaling receptor for TWEAK is FGF-inducible molecule 14, Fn14 (also called TWEAKR, TNFRSF12A, and CD266), although another receptor, CD163, has been reported to bind TWEAK (22, 23). Fn14 signaling is mediated by TNF receptor-associated factors (24) and has been inferred to regulate various physiological processes such as cell proliferation, migration, survival, differentiation, and death, depending on the cellular context (25).

TWEAK and Fn14 are expressed at low levels in the brain under normal conditions (15). However, the TWEAK pathway has been implicated in the pathogenesis of several neurological conditions such as ischemic stroke and multiple sclerosis (15, 26–28). Experiments in rodent adult models of stroke showed that inhibition of this pathway by TWEAK neutralizing antibody, soluble decoy Fn14-Fc, or genetic deficiency (26, 27, 29) significantly reduced brain damage after middle cerebral artery occlusion. The beneficial effects of inhibition of TWEAK signaling in this model have been attributed to the increased blood–brain barrier (BBB) stability, which leads to decreased cerebral edema, decreased infiltration of inflammatory cells in the ischemic tissue (30), and decreased neuronal death.

The aim of this study was to test the effect of inhibiting TWEAK signaling in the context of a neonatal model of HI. Here we evaluate changes of the expression of both TWEAK and Fn14 after HI in mice, and the effect of Fn14 genetic deficiency on the development of the brain injury after HI.

#### MATERIALS AND METHODS

#### Neonatal HI

This study was approved by the KCL animal ethical committee (AWERB), and all animal experimentation was performed in compliance with the UK Home Office regulations (PPL 70/8376). C57BL/6 wild-type (WT) mice were obtained from Charles River (UK); Fn14 knockout (KO) animals were obtained from Biogen (Cambridge, MA, USA). For the experiments, the homozygote KO females were mated with WT C57BL/6 males to produce heterozygote offspring, those heterozygote animals were mated to produce mixed litters of WT, homo- and heterozygote KO animals which we have used in our experiments.

Neonatal HI was induced at postnatal day 9 according to methods described by Rice et al. (31), but modified for mice (32, 33). Mice of both sexes were anesthetized with isoflurane (3% for induction and 1.5% for maintenance) in nitrous oxide/ oxygen (1:1). The left common carotid artery was ligated with Prolene suture (6.0). After the surgical procedure, the wounds were closed and infiltrated with a local anesthetic. After 1 h of recovery with the dam, the pups were placed in a chamber perfused with a humidified gas mixture (10% oxygen in nitrogen) for 65 min at 36°C. The animals were kept in humidified air at 36°C for an extra 10 min before and after the hypoxic exposure. After the hypoxia, the pups were returned to their dams. This procedure results in brain injury in the ipsilateral hemisphere, consisting of cerebral infarction and selective neuronal death in the cortex, striatum, hippocampus, and the thalamus, leaving the contralateral hemisphere undamaged. Sham-operated animals were subjected to isoflurane anesthesia and incision only, without carotid ligation and hypoxia. After the HI procedure pups were killed by decapitation at various time points. Brains were removed and rapidly frozen on dry ice for RNA isolation or fixed in 4% paraformaldehyde for immunostaining. In total, 144 animals were used in this study, 36 for RT-qPCR analysis (**Figures 1A–D**), 5 for western blot analysis (**Figures 1E–H**), 3 for fluorescent immunohistochemistry (**Figure 2**), and 100 for histological evaluation of the brain injury (**Figures 3**–**6**).

#### Immunohistochemistry

The fixed brains were embedded in blocks of 25 brains, freezesectioned (30 µm) and sections collected at 10 different levels using MultiBrain® Technology (NSA Labs, USA). The multibrain sections were stained with MBP, Ischemia Contrast, MAP2 and Iba1 (NSA Labs, USA). For the fluorescent immunohistochemistry, 5 µm paraffin embedded brain sections were prepared, deparaffinized with xylene and ethanol, and pretreated with heating for 20 min in 10 mM citric acid, pH 6.0 with 0.1% Tween 20. Sections were blocked for 20 min with 5% horse serum, 1% BSA in PBS before incubation with primary antibodies overnight at 4°C. We used rabbit anti-TWEAK (Sigma, UK), rabbit anti-Fn14 (Abcam, UK), mouse anti-GFAP (Sigma, UK), mouse anti-Olig2 (Millipore, UK), and mouse anti-TUJ1 (Cambridge Bioscience, UK) antibodies. Immunohistochemistry controls performed with secondary Ab alone (data not shown) or on tissue from Fn14 KO animals with anti-Fn14 antibody (Figure S1 in Supplementary Material) showed no visible staining. For visualization, we used AlexaFluor488 conjugated anti-mouse antibody and AlexaFluor546-conjugated anti-rabbit antibody (Life Technologies). Sections were analyzed using a Leica DM6000 B fluorescent microscope and a Leica TCS SP5 confocal microscope. Images were processed using Leica LAS AF Lite and ImageJ.

Figure 1 | Effects of hypoxia–ischemia (HI) on the expression levels of TWEAK (A,C) and Fn14 (B,D) in neonatal mice. mRNA levels were measured in contralateral (Contra; white bars) and ipsilateral (Ipsi; gray bars) hemispheres of the brain 4, 24, and 72 h after the insult. The results are expressed as fold change from contralateral side. Western blot detection of TWEAK (E) and Fn14 (F) protein and densitometry (G,H) in sham-operated (Sham, white bars), ipsilateral (I, gray bars), and contralateral hemisphere (C, black bars) brain lysates 72 h after HI. Densitometry of the TWEAK and Fn14 bands was corrected to the corresponding β-actin bands. Bars represent mean + SEM; \**p* < 0.05; \*\**p* < 0.01; versus corresponding contralateral side [*n* = 6–9 (A–D) and *n* = 3 (E,F)].

#### Evaluation of Brain Injury

For our analysis, we have used seven coronal levels across each brain. Total amount of tissue loss was calculated using ImageJ. Briefly, the coronal brain images were segmented by the background and threshold for the Ischemia Contrast staining applied, the hemispheres were manually segmented and intensity of the immunostaining and area calculated for both contralateral and ipsilateral hemisphere. The ipsilateral hemisphere was represented as percentage of uninjured contralateral hemisphere.

Brain injury in different regions was also evaluated by an observer blinded to the study groups, using a semi quantitative neuropathological scoring system (32). Injury in the cerebral cortex was graded both with respect to hypotrophy (0–3) and injury/ infarction (0–4; with 0 being no observable injury and 4 being confluent infarction encompassing most of the hemisphere). The damage in the hippocampus, striatum, and thalamus was assessed regarding both hypotrophy (shrinkage) (0–3) and observable cell injury/infarction (0–3), resulting in a neuropathological score for each brain region (0–6) except for cerebral cortex 0–7. The total score (0–25) was the sum score for all four regions.

oligodendrocytes (E,F). Scale bar = 10 µm (A–D) and 30 µm (E,F).

#### Evaluation of Microglia Injury

For our analysis, we have used seven coronal levels across each brain. Total amount of Iba1 staining was calculated using ImageJ. Briefly, the coronal brain images were segmented by the background and threshold for the immunostaining applied, the hemispheres were manually segmented and the immunostaining area calculated for both contralateral and ipsilateral hemisphere. The area of positive Iba1 staining was calculated as percent of hemisphere area. The ipsilateral hemisphere was represented as percentage difference compared with uninjured contralateral hemisphere.

## RNA Isolation

Total RNA from mouse brains was isolated using the RNeasy mini kit (Qiagen) according to the manufacturer's instructions.

## Quantitative RT-PCR (qRT-PCR)

The reverse transcription reaction was performed using High Capacity cDNA reverse transcription kit (Applied Biosystems) according to the manufacturer's instructions. qRT-PCR experiments were performed using the StepOnePlus™ Real-Time PCR Systems, TaqMan probes and TaqMan Gene Expression Master Mix (Applied Biosystems). All reactions were conducted in duplicate and corrected to GAPDH expression. Data were analyzed using the delta threshold cycle (CT) method (34).

#### Western Blot

30 µg total protein lysate per sample was separated by electrophoresis using 10% Bis–Tris NuPAGE® Novex gels (Invitrogen) and XCell SureLock® Mini-Cell (invitrogen). Proteins were transferred to polyvinylidene fluoride membranes using iBlot® Gel Transfer Device (Invitrogen). Membranes were blocked with 5% non-fat dry milk in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) and immunoblotted overnight at 4°C with anti-TWEAK antibody (sigma) or anti-Fn14 antibody (abcam) diluted 1:2,500 in TBS-T. After washing with TBS-T, membranes were incubated for 1 h with HRP conjugated anti-goat antibody at room temperature. Membranes were washed with TBS-T and developed with SuperSignal West Pico Chemiluminescent Substrate (Thermo Life Science, City, State). The images were taken with ImageQuant™ LAS4000 digital imaging system (GE). Density of the bands was analyzed using FIJI (ImageJ), and the TWEAK and Fn14 bands were corrected to the corresponding β-actin bands.

#### Statistical Analysis

The size of samples for our study was calculated using the following formula *n* = (*Z*α/2 + *Z*β)2\*2σ 2/δ2, where *Z*α/2 is significance level; *Z*β is power; σ is SD of variable, and δ is minimum important difference. For this study, we have chosen significance level of 5% and power 80%.

The statistical analysis was performed using Prism GraphPad 5.0 software (GraphPad Software). Data are expressed as mean ± SEM. Comparisons between the experimental groups were made using one-way analysis of variance (ANOVA) followed Tukey posttest for comparing between more than two groups with one independent factor (**Figures 1G,H**) or using two-way ANOVA followed by Bonferroni posttest for comparing between more than two groups with two independent factors (**Figures 1A–D**, **3**, **4**, **5** and **6**).

## RESULTS

#### Neonatal HI Increases Expression of TWEAK and Fn14

The expression of TWEAK and Fn14 was evaluated in a neonatal mouse model of HI. We found a significant increase in TWEAK

genotype stained with Ischemia Contrast staining. These images are selected from an animal with tissue loss closest to the average tissue loss for the group.

The percent of the tissue loss is indicated below each brain section. WT (white bars), Het (gray bars), and KO (black bars).

mRNA levels in the ipsilateral compared with the contralateral hemisphere of male pups at the time points of 24 and 72 h after HI (**Figure 1A**). In the brains of female animals, we found a similar pattern in TWEAK expression in the ipsilateral side at 72 h after HI that did not reach significance (*p* = 0.052; **Figure 1C**). Fn14 mRNA expression was significantly induced at 4 h after HI in male animals and 24 h after HI in both male and female animals (**Figures 1B,D**). Unlike TWEAK where expression increased over

time, with highest values at 72 h, Fn14 expression peaked at 24 h and normalized 72 h after HI. At all the time points studied, no differences in TWEAK and Fn14 expression were observed between samples from contralateral compared with sham-operated animals as well as no difference in expression of TWEAK and Fn14 in males versus females in the contralateral hemisphere. Protein

expression was analyzed by western blot of brain lysates prepared at 72 h post insult. As observed with the mRNA results, TWEAK and Fn14 protein expression were significantly increased in the ipsilateral hemisphere after HI (**Figures 1E–H**).

To investigate which cells were responsible for the enhanced expression of TWEAK and Fn14, we analyzed brain sections by immunofluorescence from animals sacrificed 72 h after HI. TWEAK immunofluorescence was strongest in GFAP+ astrocytes in the ipsilateral hemisphere (**Figures 2A,B**); no co-localization of TWEAK was observed in neurons, microglia, or oligodendrocytes (data not shown). By contrast, Fn14 was expressed in neurons (**Figures 2C,D**) and in oligodendroglial precursor cells (**Figures 2E,F**) 72 h after HI. Expression was not observed in any other cell type (data not shown).

### Fn14 Gene Deletion Is Protective in Females but Augments Brain Injury in Males After HI

To examine the impact of TWEAK signaling on the evolution of brain injury, we subjected mice genetically deficient in Fn14 (Fn14 KO) to our neonatal HI protocol and determined the effect of gene KO on brain injury. We found no significant difference in the tissue loss between the female and male WT or heterozygote (Het) animals. However, in female Fn14 KO mice, mean tissue volume loss was significantly reduced compared with WT and Het females [21.3 ± 2.8% in KO females compared with 38.1 ± 2.0 and 41.0 ± 1.4% in WT and Het females, respectively (**Figures 3A,B**)]. Surprisingly, brain injury was consistently more pronounced in male Fn14 KO compared with WT and Het males [47.5 ± 1.7% in the KO males compared with 33.5 ± 2.0 and 39.6 ± 1.4% in WT and Het males, respectively (**Figures 3A,C**)]. Thus, there was a differential effect of Fn14 deficiency in female versus male KO mice. To investigate this differential effect further, we analyzed potential changes in white matter [myelin basic protein (MBP)] as well as microglial infiltration (Iba1) after HI injury (**Figures 4** and **5**). MBP staining was conserved in female Fn14 KO mice (**Figures 4A,B**), indicating increased white matter integrity, whereas there was a trend toward loss of MBP staining in male Fn14 KO, although this did not reach significance (**Figures 4A,C**). There were also no significant differences between the genotypes as well as between males and females observed in the sections stained for Iba1 (**Figures 5A–C**).

Finally, we carried out a detailed analysis of neuropathology scores associated with injury to specific brain regions. In line with our previous observations, we observed differential responses between male and female sections in cerebral cortex (**Figure 6A**), hippocampus (**Figure 6B**), striatum (**Figure 6C**), and thalamus (**Figure 6D**). In all regions examined, deletion of *Fn14* was associated with attenuation of brain injury in females and increased injury in the male brain (**Figure 6**).

To discard the possibility of any differences in the brain volume, myelin content, or microglia numbers between males and females and between different genotype groups, we have analyzed and compared contralateral sides of all brains stained with Ischemia Contrast, MBP and Iba1 (Figure S2 in Supplementary Material). We found no significant differences indicating that all differences found in the ipsilateral side were a result of the differences in brain injury.

#### DISCUSSION

In this study, we report a sex difference in neonatal HI where Fn14 deletion decreased injury in females and increased injury in males. We have shown that the expression of both TWEAK and its receptor Fn14 is increased in the injured side of the brain after HI *in vivo* and that the increased expression of TWEAK appears to be due to its expression in activated astrocytes. The increase of Fn14 expression was detected mainly in neurons and OPC proximal to the injury site. As neurons and oligodendrocytes are the main cell types lost after HI (35), this may suggest involvement of TWEAK signaling in the development of neonatal brain injury after HI. We also found a marked sex difference in the response to partial or complete *Fn14* gene deficiency. In females, brain tissue loss after HI was reduced in *Fn14* KO mice whereas gene deletion worsened injury in the males (**Figures 3A**, **4**, **5** and **6**). Existence of sex differences in the response to HI especially among preterm children is well described (3–5). Sex differences have been shown in HI experiments with neonatal WT mice, where males exhibit more severe brain injury than females (36), although studies using this model do not always find a difference in the extend of brain injury (37, 38). A more common finding is sex-specific responses to various treatments in neonatal animal models, where male and female pups respond differently to treatments administered after injury implicating that different injury mechanisms are at play (37, 39–44). In adult, it is well known that gonadal hormones such as estrogen and progesterone exert protective effect (45, 46). However, in neonatal animals, hormonal influences are less likely to contribute but sex-dependent basic cellular and genetic mechanisms seem to be critical (5).

The interesting finding in our study is that *Fn14* deficiency has an effect on both sexes but is in opposite directions for males and females. The differences that we observed were consistent in all brain areas examined (cerebral cortex, hippocampus, striatum, and thalamus) suggesting that no specific area accounts for the difference but rather a consistent effect across the whole brain.

We currently cannot provide an explanation regarding why TWEAK–Fn14 signaling increases injury in females whereas it is part of a protective response in males. From our experiments, in WT animals, it is obvious that male animals responded differently to HI with an earlier increase in the expression of TWEAK and Fn14 than the females (**Figures 1A–D**). We and others have previously found differences in the intrinsic apoptotic pathway between males and females (37, 42, 43). It appears that the poly(ADP-ribose) polymerase 1 and apoptosis-inducing factor are more critical components in the caspase-independent apoptotic cascade in neurons of males whereas those of females depend more on the cytochrome C–caspase pathway (37). We can speculate that there are also sex differences in the death receptor pathway that may relate either to the extrinsic apoptotic pathway or RIP1 kinase-dependent necroptosis. The latter has previously been shown to be different in male and female mice after neonatal HI (44). TWEAK treatment induces cell death in primary cortical neurons through NF-κB pathway activation (26, 47). However, NF-κB activation downstream of death receptor activation has been suggested to play a role both in cell survival and cell death depending on context (16, 48), which certainly could be different between the male and female CNS.

Another possible explanation for the sex difference is that TWEAK exerts its neurotoxicity on the neurons and oligodendrocytes not directly through activating their Fn14 receptors but indirectly by changing their environment, for example, through effects on BBB permeability, as observed in adult models of HI injury (30). TWEAK signaling is able to activate matrix metalloproteinase-9 activity in the brain and in primary astrocytes, which in turn increases the permeability of the BBB (49). If the BBB is compromised, this will allow the influx of immune cells from the blood, as well as of inflammatory cytokines and infectious agents, into the brain. In such a case, the state of peripheral inflammatory cells may have a profound effect on the development of brain injury.

We examined the amount of Iba1 positive cells in the brains after HI injury to study whether the sex differences in the injury can be attributed to differences in the microglial immune response. Sex differences in microglial response have been reported in various neurodevelopment and neurodegenerative disorders (8–10). In our study, we have limited our investigation on microglia only to Iba1 staining, which is an estimate for the number of microglia cells rather than their inflammatory status (e.g., secretion of cytokines, phagocytosis, etc.). We did not detect any significant difference in our experiment (**Figure 5**), meaning that the increase in the number of microglia cell after HI is equal in male and female as well as in different genotypes.

Sex differences in the inflammatory response of astrocytes to inflammation-inducing agents like lipopolysaccharide has been reported (50). This makes astrocytes and their interaction with the neurovascular unit a reasonable candidate for future studies on the sex-dependent role of TWEAK/Fn14 in perinatal brain injury.

More studies are needed to determine whether the sex differences between male and female KO animals can be attributed to differences in the inflammatory response and/or BBB permeability. The other receptor for TWEAK, CD163, is unlikely to contribute for any of the physiological effects observed in our study. CD163 is a scavenger receptor for hemoglobin and haptoglobin without intracellular signaling (51). Moreover, some studies were unable to find any interaction of TWEAK with CD163 (52). A certain degree of promiscuity can be anticipated as it is well known that the members of TNF family have significant structural similarity to each other and it is common that one ligand is signaling through multiple receptors (TNF-α can bind to both TNFR1 and TNFR2; TRAIL can bind to five different receptors in human, etc.). It is not unlikely, however that TWEAK is interacting with and signaling through receptors undescribed to date as a binding partner of TWEAK. Therefore, TWEAK that is present in high concentrations after HI (**Figure 1A**) could easily bind to other receptors with much lower affinity and trigger some cellular responses. However, to date no alternative binding partners of TWEAK are currently described.

In this study, we have found increase in the expression levels in both TWEAK and Fn14 in the ipsilateral side of the brain following HI and an opposite effect of the Fn14 deletion on the development of the injury for the males and females. These findings are opening different avenues for further investigation of the effects of TWEAK/Fn14 signaling in the context of perinatal brain injury.

#### ETHICS STATEMENT

This study was approved by the KCL animal ethical committee (AWERB), and all animal experimentation was performed in compliance with the UK Home Office regulations (PPL 70/8376).

#### REFERENCES


### AUTHOR CONTRIBUTIONS

AK was responsible for design and organization of the study. He performed experiments, data acquisition, statistical analysis, and interpretation and wrote the manuscript. AB performed experiments and data acquisition and reviewed the manuscript. RV performed data acquisition and reviewed the manuscript. LB critically discussed findings and reviewed the manuscript from a technical, statistical, and scientific perspective. PG reviewed the manuscript. CT and HH supported in designing the study, interpretation of the result, critical revision, and preparing the manuscript.

#### ACKNOWLEDGMENTS

We would like to thank to Dr. Richard Ransohoff for his advice and suggestions during preparation of this manuscript. We gratefully acknowledge the support of Action Medical Research and The Henry Smith Charity, the Department of Perinatal Imaging and Health and financial support from Wellcome Trust (WT094823), the Medical Research Council (MRC strategic grant contract P19381), Swedish Medical Research Council (VR 2015-02493), Brain Foundation (HH), Ahlen Foundation (HH), ALF-GBG (426401), ERA-net (EU; VR 529-2014-7551), and the Leducq Foundation (DSRRP34404) to enable this study to be completed. In addition, the authors acknowledge financial support from the Department of Health *via* the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre Award to Guy's and St. Thomas' NHS Foundation Trust in partnership with King's College London and King's College Hospital NHS Foundation Trust.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fneur.2018.00230/ full#supplementary-material.

Figure S1 | Representative images of confocal (A,B) and conventional (C,D) fluorescence microscopy of paraffin sections from wild type [WT (A,C)] and Fn14 knockout [Fn14 KO (B,D)] of the mouse brain. There was Fn14 immunoreactivity (red) in TUJ-positive (green) neurons (A) and Olig2 positive (green) oligodendrocytes (C) in the WT animals. Lack of any Fn14 staining in Fn14 KO animals (B,D). Scale bar = 10 µm (A,B) and 30 µm (C,D).

Figure S2 | No difference in total brain tissue (A), myelin basic protein (MBP) staining (B), and Iba1 staining (C) between males and females and between wild-type (WT), Fn14 knockout (KO), or heterozygote (Het) animals. Brain sections (seven different levels) were generated 7 days after hypoxia–ischemia and subjected to Ischemia Contrast staining (A), tissue area in the contralateral side was measured and presented as square millimeters. MBP immunostaining (B) and Iba1 immunostaining (C) of the contralateral side are presented as percentage of total brain area of the contralateral side. Bars represent average + SEM.


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

# Early Detection of Hypothermic Neuroprotection Using T2-Weighted Magnetic Resonance Imaging in a Mouse Model of Hypoxic Ischemic Encephalopathy

*Sydney E. Doman1†, Akanksha Girish1†, Christina L. Nemeth1 , Gabrielle T. Drummond1 , Patrice Carr1 , Maxine S. Garcia1 , Michael V. Johnston1,2, Sujatha Kannan1,3, Ali Fatemi1,2, Jiangyang Zhang4 and Mary Ann Wilson1,2,5\**

*1Hugo W. Moser Research Institute at Kennedy Krieger, Baltimore, MD, United States, 2Department of Neurology, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, 3Anesthesiology and Critical Care Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, 4Department of Radiology, New York University School of Medicine, New York, NY, United States, 5Department of Neuroscience, The Johns Hopkins University School of Medicine, Baltimore, MD, United States*

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Christiane Charriaut-Marlangue, Institut National de la Santé et de la Recherche Médicale (INSERM), France Shigeyoshi Saito, Osaka University, Japan*

#### *\*Correspondence:*

*Mary Ann Wilson wilsonm@kennedykrieger.org*

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

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 09 January 2018 Accepted: 18 April 2018 Published: 08 May 2018*

#### *Citation:*

*Doman SE, Girish A, Nemeth CL, Drummond GT, Carr P, Garcia MS, Johnston MV, Kannan S, Fatemi A, Zhang J and Wilson MA (2018) Early Detection of Hypothermic Neuroprotection Using T2-Weighted Magnetic Resonance Imaging in a Mouse Model of Hypoxic Ischemic Encephalopathy. Front. Neurol. 9:304. doi: 10.3389/fneur.2018.00304*

Perinatal hypoxic-ischemic encephalopathy (HIE) can lead to neurodevelopmental disorders, including cerebral palsy. Standard care for neonatal HIE includes therapeutic hypothermia, which provides partial neuroprotection; magnetic resonance imaging (MRI) is often used to assess injury and predict outcome after HIE. Immature rodent models of HIE are used to evaluate mechanisms of injury and to examine the efficacy and mechanisms of neuroprotective interventions such as hypothermia. In this study, we first confirmed that, in the CD1 mouse model of perinatal HIE used for our research, MRI obtained 3 h after hypoxic ischemia (HI) could reliably assess initial brain injury and predict histopathological outcome. Mice were subjected to HI (unilateral carotid ligation followed by exposure to hypoxia) on postnatal day 7 and were imaged with T2-weighted MRI and diffusion-weighted MRI (DWI), 3 h after HI. Clearly defined regions of increased signal were comparable in T2 MRI and DWI, and we found a strong correlation between T2 MRI injury scores 3 h after HI and histopathological brain injury 7 days after HI, validating this method for evaluating initial injury in this model of HIE. The more efficient, higher resolution T2 MRI was used to score initial brain injury in subsequent studies. In mice treated with hypothermia, we found a significant reduction in T2 MRI injury scores 3 h after HI, compared to normothermic littermates. Early hypothermic neuroprotection was maintained 7 days after HI, in both T2 MRI injury scores and histopathology. In the normothermic group, T2 MRI injury scores 3 h after HI were comparable to those obtained 7 days after HI. However, in the hypothermic group, brain injury was significantly less 7 days after HI than at 3 h. Thus, early neuroprotective effects of hypothermia were enhanced by 7 days, which may reflect the additional 3 h of hypothermia after imaging or effects on later mechanisms of injury, such as delayed cell death and inflammation. Our results demonstrate that hypothermia has early neuroprotective effects in this model. These findings suggest that hypothermia has an impact on early mechanisms of excitotoxic injury and support initiation of hypothermic intervention as soon as possible after diagnosis of HIE.

Keywords: hypothermia, neuroprotection, hypoxic-ischemic, neonatal encephalopathy, magnetic resonance imaging

## INTRODUCTION

Perinatal hypoxic-ischemic encephalopathy (HIE) is a childbirth complication most commonly induced by intrauterine asphyxia (1). Neonatal HIE can lead to a wide spectrum of neurodevelopmental disorders, most notably cerebral palsy (CP) (2). While CP is a non-progressive disorder, HIE injury is a dynamic process. Injury begins when cerebral blood flow and oxygen delivery to the brain are impaired, resulting in primary energy failure (3–5). Secondary energy failure occurs 6–48 h later, from a combination of oxidative stress, excitotoxicity, and inflammation (6, 7). The window between primary and secondary energy failure provides a critical period for intervention to mitigate or altogether prevent the effects of secondary energy failure (8). Early classification of HIE severity may allow more appropriate intervention, such as identification of patients who would benefit from more aggressive or pharmaceutical treatment (9).

Therapeutic hypothermia is the standard treatment for neonatal HIE, due to its proven efficacy in reducing infant mortality and brain injury by 25%, and in reducing the risk of CP by 34% (10). The timing and intensity of cooling during therapeutic hypothermia are critical in preserving neural function. In clinical practice, neonates begin hypothermia within 6 h of birth and are typically maintained at 33.5 ± 0.5°C for 72 h (11). Cooling below 33.5°C provided no further neuroprotection in a preclinical study of rat pups (12), and this has now been confirmed in a clinical trial (11). Moreover, studies across a variety of animal models have shown a negative correlation between overcooling and neuroprotection. Hypothermia performed at 8.5°C below normothermia in piglets increased brain cell death (13), and overcooling in the rat model to 18°C provided no neuroprotection and had a negative impact on feeding ability (12). We have validated a model of hypothermic neuroprotection in mice subjected to hypoxic ischemic insult on postnatal day 7 (P7) that replicates the neuroprotective effects of mild hypothermia in term human infants (14).

Magnetic resonance imaging (MRI) has numerous benefits as a non-invasive tool to assess brain injury in neonates. There is debate over the optimal timing for MRI in order to reliably assess and predict outcomes of brain injury. A report from the American College of Obstetricians and Gynecologists has affirmed that an initial diagnostic MRI obtained between 24 and 96 h shows distinguishable patterns of injury in myelinated regions of the brain, while scans obtained after 10 days of life offer the most conclusive evidence for the degree of cerebral injury (15). In infants with HIE, early MRI (≤6 days of age, typically at the end of therapeutic hypothermia) has been shown to be sensitive and specific for predicting neurodevelopmental outcome, with similar performance to MRI performed after 7 days of age (16). Furthermore, a study of newborns with HIE and MRI studies performed a median of 6 days after birth found that the prognostic value of MRIs was unaffected by therapeutic hypothermia (17). MRI is typically obtained at the end of hypothermia, but a recent study evaluated imaging carried out during hypothermia. Although MRI performed on day 1 of life tended to underestimate brain injury, brain injury examined using DWI at day 2–3 of life fully corresponded with assessments carried out using T1/T2-weighted imaging at day 10 of life (18). Establishing reliable MRI at earlier time points may facilitate the differentiation of mild, moderate, and severe cases of HI insult, which could have a decisive impact on treatment.

Immature rodent models of HIE, in which carotid artery ligation is followed by a period of hypoxia, produce a distribution of brain injury that resembles the injury observed in asphyxiated term human infants (19, 20). The timing of brain injury in these models differs from that reported in human infants, with a more rapid progression of injury detected in mouse models of stroke and HIE. A number of studies have shown that injury can be reliably assessed with diffusion-weighted or T2 MRI, 3–24 h after the initial insult in various rat and mouse models of neonatal brain injury (21–27). In this study, we first sought an efficient method for assessing initial injury in our mouse model of perinatal HIE, for use in studies of hypothermic neuroprotection and for later use in developing complementary drug therapies. We compared T2-weighted and diffusion-weighted MRI obtained 3 h after HI for predicting brain histopathological outcome at 7 days in this model and found that they detected a very similar distribution and extent of injury that corresponded well with histopathological outcome at 7 days. We then applied the more efficient T2-weighted MRI at 3 h after HI to characterize initial injury in mice treated with hypothermia and found differences between the treatment groups at this early time point. The final phase of this study examined whether hypothermic neuroprotection that was detected with MRI 3 h after HI was maintained at a longer survival, using both T2 MRI and histopathology 7 days after HI. Our results confirm the prognostic value of early T2-weighted MRI as a predictor of HI brain injury in this mouse model, in both untreated mice and in mice treated with therapeutic hypothermia. These data also reveal that neuroprotective effects of hypothermic treatment can be detected as early as 3 h after initiating hypothermia in this model of HIE. These findings suggest that hypothermia can ameliorate very early mechanisms of brain injury in HIE, which may have important implications for the development of complementary therapies for use with hypothermia.

### MATERIALS AND METHODS

#### Experimental Design

This study consisted of three phases, illustrated in **Figure 1**. All phases used a P7 mouse model of HIE (28), and Phases 2 and 3 evaluated hypothermic neuroprotection, carried out as described previously (14). Phase 1 compared early T2-weighted and diffusion-weighted MRI for prediction of histopathologic brain injury examined at P7 (**Figure 1A**). Phase 2 applied the validated T2-weighted MRI scale in mice treated with hypothermia and

**Abbreviations:** ADC, apparent diffusion constant; CP, cerebral palsy; DWI, diffusion-weighted MRI; EPI, echo planar imaging; FOV, field of view; HI, hypoxic ischemia; HIE, hypoxic ischemic encephalopathy; MRI, magnetic resonance imaging; P, postnatal day; RARE, rapid acquisition with relaxation enhanced; TE, echo time; TR, repetition time.

Brain injury was scored in T2 MRI using a qualitative scale. Mice were euthanized and histopathology was examined 7 days after HI, on P14. (B) Phase 2: mice were subjected to HI on P7, followed by a 6 h period of hypothermia or normothermia (brains were harvested for other studies not shown here). The validated T2 MRI score was determined at 3 h. (C) Phase 3: mice were subjected to HI on P7, followed by a 6 h period of normothermia or hypothermia. T2-weighted MRI was obtained and scored 3 h and 7 days after HI, and brain injury was examined in histological sections 7 days after HI.

in normothermic littermates, to characterize initial injury for a separate study (**Figure 1B**). Because Phase 2 showed striking differences in MRI injury scores between normothermic and hypothermic groups as early as 3 h after HI, Phase 3 was conducted to determine whether hypothermic neuroprotection demonstrated using T2-weighted MRI scores 3 h after HI persisted at a longer survival time, using both T2-weighted MRI and histopathologic brain injury examined 7 days after HI (**Figure 1C**).

### Animal Model of HIE

All animal procedures were carried out in accordance with the National Research Council Guide for the Care and Use of Experimental Animals (29) and were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University. Lactating CD1 female mice with litters consisting of five male and five female pups were obtained on P6 (Charles River Laboratories) and housed in barrier facilities with a 14 h light/10 h dark schedule and standard mouse chow and water *ad libitum*. The following day, pups were subjected to HI as described previously (14). Briefly, under isoflurane anesthesia, the right common carotid artery was permanently ligated. Pups recovered in an incubator at 36.5°C for approximately 90 min and were then placed in hypoxic chambers (10% O2, balance N2) for 20 min at 36.5°C. Pups were randomly assigned to hypothermic treatment (6 h at 33.5°C) or a control period of normothermia (6 h at 36.5°C), as described previously (14). The hypothermic temperature was selected to correspond to that used clinically for HIE, and the length of hypothermic therapy was estimated as the longest period that pups could be maintained in incubator chambers, separated from their dam. We have shown previously that these conditions result in moderate neuroprotection [(14), supplemental Figure S2]. The number of subjects in each treatment group for each phase is shown in **Table 1**. Hypothermic and Table 1 | Distribution of surviving pups by treatment group in each phase and mortality.


*a Pups were not kept until the 7-day time-point; mortality rate reflects that observed after 6 h treatment.*

normothermic body temperatures (average of readings at 30 min and 6 h) were 33.6 ± 0.4°C and 36.9 ± 0.3°C, respectively. All pups were then kept at 36.5°C for 15 min before returning to the dam.

### *In Vivo* MRI

Magnetic resonance imaging was carried out 3 h after HI in all pups across all phases and again 7 days after HI in Phase 3 pups. All MRI studies were performed in the F.M. Kirby Research Center High-Field Preclinical MR Facility at the Kennedy Krieger Institute on a Bruker BioSpec horizontal 11.7 T MRI system. A custom holder was designed to permit scanning of mice in pairs, using a receive-only rat head phased array coil in combination with a quadrature volume coil for excitation. Anesthesia was induced with 2% isoflurane and maintained with 1.0 to 1.5% isoflurane in a mixture of O2 and air. After an initial localizer scan, T2-weighted images were acquired using the rapid acquisition with relaxation enhanced sequence and the following parameters: echo time (TE)/repetition time (TR) = 50/5,500 ms, 4 signal averages, echo train length = 8, field of view (FOV) = 23 mm × 14 mm, matrix size = 256 × 128, an in-plane resolution of 0.09 mm ×

0.11 mm, 41 slices with a slice thickness of 0.5 mm that cover the entire brain, and a total imaging time of 5 min. Diffusionweighted images were acquired using the diffusion weighted echo planar imaging sequence and the following parameters: TE/ TR = 27/5,000 ms, 1 signal averages, 4 segments, FOV = 23 mm × 14 mm, matrix size = 174 × 96, an in-plane resolution of 0.13 mm × 0.15 mm, gradient duration/separation = 4/12 ms, 30 diffusionencoding directions with a diffusion-weighting (b) of 800 s/mm2 , 5 non-diffusion-weighted images, 20 slices with a slice thickness of 1 mm that cover the entire brain, and a total imaging time of 11 min. For studies in Phases 2 and 3, the bed and air temperatures in the MRI were adjusted to maintain normothermic or hypothermic body temperatures.

#### Histopathology

To evaluate brain injury in histological sections 7 days after HI, mice were perfused with ice-cold phosphate-buffered saline. Brains were removed, immersion fixed in methacarn fixative (60% methanol, 30% chloroform, 10% glacial acetic acid) and embedded in paraffin. Brains were sectioned at 20 µm, and a 1 in 10 series of sections was stained with cresyl-violet.

#### Histopathological Image Analysis

In each subject, a series of regularly spaced sections extending throughout the brain was imaged using a 1.25× objective on a Zeiss Axio Scope microscope. MCID Core 7.1 (Interfocus Imaging, LTD) was used to carry out a volumetric analysis of injury in the injured right hemisphere, compared with the uninjured contralateral left hemisphere. Brain injury was calculated as a percentage of the total contralateral volume, [(contralateral − ipsilateral/contralateral) × 100]. The contralateral hemisphere typically served as a reliable uninjured control. However, in very severely injured cases, injury in the contralateral hemisphere, typically in the cerebral cortex (cortex MRI score of 4), invalidated its use as a comparison control. A standard volume for the control hemisphere was calculated as the mean hemispheric volume of hemispheres in uninjured pups (MRI score of 0). Contralateral and ipsilateral hemispheric volumes did not differ significantly in the uninjured pups (paired *t*-test, *p* = 0.351), and thus both were used for an accurate control. This standard hemispheric volume was substituted for cases in which contralateral injury was observed and the contralateral volume was more than two SDs below the uninjured mean.

#### MRI Analysis

Unlike in MRI of patients with ischemic stroke, which often show delayed increases in T2 signals by several hours after changes in DWI signals, in this mouse model, increased T2 signals can be readily detected in edema regions 3 h after HI. DWI data were processed using ParaVision 5.1 DTI tensor reconstruction to generate fractional anisotropy, tensor trace, intensity, and trace weighted images, which were compared with T2 MRI for evaluation of injury. Brain injury was most clearly delineated in the T2 MRI and trace DWI, shown in **Figure 2**. The injury was comparable using these methods, but T2 MRI provided a more detailed assessment of the extent and regional distribution of injury, with less time required for imaging, at higher resolution, than that required for DWI. (A lower resolution for DWI was necessitated by overall imaging/anesthesia time constraints and a voxel size permitting adequate diffusion signals.) The higher resolution T2 MRI were obtained using much shorter imaging times that limited the amount of additional anesthesia and minimized potential interference with thermal control. T2 MRI was thus selected for scoring and used exclusively in the later experiments. Each series of coronal T2 images extending throughout the forebrain was examined by two investigators who were unaware of treatment status. Cortex, hippocampus, striatum, and thalamus were scored using an ordinal scale similar to that used previously to score histopathology in this model (28) as follows: (0) no apparent injury; (1) mild injury, consisting of small hyperintense areas (<30% of the regional volume); (2) moderate injury, larger hyperintense areas (30–60% of total regional volume); (3) severe injury (>60% of the regional volume); or (4) severe with bilateral injury in that region. The regional scores were summed to obtain a total brain injury score, with a maximum possible score of 16. Representative samples are shown in **Figure 2**. No significant difference was found between the two investigator's MRI scores (Wilcoxon signed-rank test, *p* = 0.48), and the average of their scores was used for analysis. A similar regional scoring model was used to evaluate brain injury 7 days after HI in Phase 3, when injury consisted of cystic infarction and atrophy: (0) no injury apparent; (1) mild injury (<30% regional volume loss); (2) moderate injury (30–60% regional volume loss); (3) severe injury (>60% regional volume loss); or (4) severe with bilateral injury in that region.

### Data Analysis

Statistical analyses were performed using IBM SPSS Statistics 24. Independent samples *t*-tests were used to analyze differences in volumetric brain injury across treatment groups. Paired *t*-tests were used to compare right and left uninjured hemispheres. Nonparametric tests were used for analysis of MRI scores: Spearman's non-parametric correlation, related samples Wilcoxon signedrank tests for comparison of MRI scores at 3 h and 7 days, independent samples Mann–Whitney *U*-tests for comparison of MRI scores between treatment groups. Differences were considered significant at *p* < 0.05.

### RESULTS

### Phase 1: Validation of MRI Score at 3 h to Predict Brain Injury at 7 Days

Increased T2 signals were readily detected in edema regions 3 h after HI (**Figure 2**), and T2 MRI and trace DWI showed a comparable extent and distribution of edema (**Figure 2**). T2 MRI provided a more detailed assessment of the extent and regional distribution of injury, with less time required for imaging, at higher resolution, than that required for DWI. T2 MRI scores at 3 h in normothermic pups were highly correlated with histopathological brain injury at 7 days (**Figure 3**, Spearman's rho = 0.93, *p* < 0.001). Thus, in this murine model of neonatal

of injury in DWI, with corresponding brain injury observed at 7 days in cresyl-violet stained brain sections. Brain injury was scored in the T2-weighted images, with the following total injury scores in the examples shown: (A) no injury (MRI score of 0), (B) mild injury (MRI score 3.5), (C) moderate injury (MRI score 8), (D) severe unilateral injury (MRI score 12), (E) severe bilateral injury (MRI score 15).

HIE, very early T2-weighted MRI provides a reliable method for predicting the severity of brain injury exhibited 7 days after HI. Refer to **Table 1** for number of animals used.

### Phase 2: Early Hypothermic Neuroprotection Detected by T2-Weighted MRI, 3 h After HI

In mice subjected to HI followed by a 6 h period of hypothermia or normothermia, T2-weighted MRI was obtained 3 h after HI, to evaluate initial injury. As the MRI was obtained, a qualitative assessment was noted, distinguishing no injury, mild, moderate, and severe injury in these subjects, which revealed a difference in apparent injury between hypothermic and normothermic pups. The regional scoring method validated in our Phase 1 study was then applied, and we found a striking reduction in brain injury in the hypothermic group, after only 3 h of hypothermia (**Figure 4**). This study establishes a surprisingly early time point for the detection of hypothermic neuroprotection. Refer to **Table 1** for number of animals used.

Figure 3 | Phase 1: correlation between T2 magnetic resonance imaging scores, 3 h after hypoxic ischemia (HI), and percent brain injury measured in histological brain sections, 7 days after HI. (*n* = 43. Samples shown in Figure 2 are colored red.)

#### Phase 3: Validation of Prognostic Value of Early T2-Weighted MRI and Confirmation of Sustained Hypothermic Neuroprotection at 7 Days

Phase 3 was undertaken to determine whether the early neuroprotective effects of hypothermia observed at 3 h correspond with the MRIs and histopathology at 7 days (**Figure 5**). Within the normothermic treatment group, there was no significant difference in T2 MRI scores at 3 h and 7 days (*p* = 0.21*).* However, in the hypothermic group, T2 MRI brain injury scores were lower 7 days after HI than 3 h after HI (related-samples Wilcoxon signed-rank test, *p* < 0.05), suggesting that hypothermic neuroprotection continued to evolve after the initial imaging at 3 h. With the smaller number of subjects used in this study (7 or 8 per treatment group), hypothermic pups showed a trend for lower MRI scores compared to normothermic pups at 3 h (Mann– Whitney *U*-test, *p* = 0.072), and this difference became significant at 7 days (Mann–Whitney *U*-test, *p*< 0.05). The histopathological analysis also showed less brain injury in hypothermic subjects than in normothermic subjects, 7 days after HI (*t*-test, *p* < 0.05). Refer to **Table 1** for number of animals used.

T2 MRI scores at 3 h and 7 days after HI in Phase 3 were compared with histopathological assessment of percent brain injury 7 days after HI by correlation analysis (**Figure 6**). These results confirm the very strong correlation between T2 MRI, at 3 h or 7 days after HI, with histopathological assessment of brain injury 7 days after HI. There is a somewhat greater spread of scores, especially for the mildly injured cases, in T2 MRI at 3 h than at 7 days after injury. This may reflect the ease of detecting small patches of hyperintense signal in the early scans, compared to the difficulty in detecting subtle atrophy at 7 days.

#### DISCUSSION

The goal of our Phase 1 study was to determine if, in this mouse model of neonatal HIE, initial injury could be consistently detected in T2-weighted or diffusion-weighted MRI at a very early time point, just 3 h after HI. A secondary goal was to optimize imaging for efficient assessment of initial injury in 10–20 subjects within an imaging session, while limiting the amount of additional anesthesia and potential interference with thermal control. We found that both imaging methods detected brain injury with similar extent and distribution. However, T2-weighted imaging provided a more detailed assessment of the extent and pattern of injury, with less time required for imaging at higher resolution. Injury was scored in four regions that were readily delineated in the T2 images, using a scale similar to that used previously to score brain injury in histopathological sections (28). T2 MRI injury scores 3 h after HI were highly correlated with a volumetric assessment of histopathological brain injury examined 7 days after HI.

Previous studies in rats subjected to HI on P7 demonstrated that T2 MRI and DWI methods show transient changes in signal intensity during or immediately after HI that return toward normal values 1–3 h later, with a secondary increase 12–24 h after HI (21, 22). Nedelcu and colleagues compared T2 MRI and DWI with cellular histopathology, which showed that the early MRI changes correlate with neuronal cytotoxic edema, while the delayed MRI signal changes correlate with neuronal edema, glial activation and interstitial edema (22). DWI detects restrictions to water molecule diffusion, and reductions in the calculated ADC are thought to reflect cell swelling associated with cytotoxic edema, while increased T2 MRI signal is thought to reflect both cytotoxic and vasogenic or interstitial edema (30–33). In a neonatal rat stroke model, MRI signal changes were observed at earlier intervals after injury. Injured volumes in ADC maps obtained immediately after reperfusion correlated well (slope near 1.0) with histopathological outcome at 48 h, but overestimated histopathological

injury 2–5 h after reperfusion. T2 maps accurately depicted the injured volume at this early interval [slopes 0.97 at 2 h and 1.04 at 4 h (27)]. The neonatal stroke model, in which permanent middle cerebral artery occlusion is combined with transient common carotid artery occlusion, produces a more focal lesion than HI, in which permanent common carotid artery occlusion is combined with a period of global hypoxia. Such differences between models may account for differences observed in the timing of MRI signal changes. Ådén and colleagues (23) examined T2 MRI and DWI in mice subjected to HI on P7 and found increased T2 values 3–6 h after HI and decreased ADC values 3 h after HI. Our data are similar, showing consistent increases in T2 signal at 3 h, with injury scores that are strongly correlated with brain histopathology at 7 days. Therefore, we chose T2 MRI for Phases 2 and 3, as it is faster, with minimal interruption to hypothermia and shorter periods of anesthesia for imaging, is less sensitive to motion, and offers higher resolution and better delineation of brain regions than DWI.

hypothermic group was confirmed by histopathology (\**p* < 0.05, *t*-test).

In a study of the effects of hypothermia on brain injury after HI in rats on P10, Patel and colleagues reported no hypothermic protection 24 h after HI but found delayed neuroprotection using T2 MRI at 2 weeks and histopathology at 12 weeks (34). Therefore, we did not expect to find a difference between treatment groups in T2 MRI 3 h after HI. However, we found significant early neuroprotection in the hypothermic group. We then confirmed that this early effect of hypothermia was sustained and provided even greater benefit when examined with T2 MRI and histopathology, 7 days after HI. Burnsed and colleagues (35) used T2 MRI to evaluate hypothermic neuroprotection 8 and 20 days after HI in immature C57Bl/6 mice. In their model, using a lower temperature for hypothermia (31°C) for a shorter period (4 h) than that used in the present study, hypothermia provided neuroprotection 8 days after HI but protection persisted only in males at 20 days. Further studies will be required to determine whether hypothermic neuroprotection persists at longer survival times in our model.

Frontiers in Neurology | www.frontiersin.org May 2018 | Volume 9 | Article 304

uptake and on cellular and interstitial edema.

**92**

T2 MRI at 7 days after HI was significantly greater than the initial protection observed at 3 h. This suggests that the additional 3 h of hypothermia after imaging at 3 h may improve neuroprotection, and/or that some of the neuroprotective effects of hypothermia may involve later mechanisms of injury such as inflammation. In neonatal rat HIE models, hypothermia reduces caspase-3 activation, apoptosis, and necrosis examined 24 h after HI (45), modifies complement factor expression (46), and reduces IL1β levels (47). Later effects of hypothermia on cytokine levels and microglial activation have been reported in an adult murine stroke model (48), and a shift in microglial polarization toward an M2 phenotype at 24 h has been observed after hypothermia in an adult rat TBI model (49). In human infants with HIE treated with hypothermia, a reduction in serum levels of the pro-inflammatory cytokine IL-6 and an increase in the anti-inflammatory cytokine IL-10 have also been observed (50). Other studies have demonstrated that cooled infants with favorable outcomes had low or declining serum levels of pro-inflammatory cytokines over time, while those with persistently elevated cytokine levels had worse outcomes (51–53). Thus, a reduction in delayed mechanisms of injury such as inflammation may be responsible for the increased neuroprotection that we observed at 7 days.

The neuroprotection observed in Phase 3 of this study using

A study in neonates evaluating early MRI, obtained during hypothermia, found that imaging on day 1 of life often underestimated the extent of injury. However, MRI obtained on days 2–3 detected injury (especially visible in DWI/ADC) that was highly predictive of the injury later on (18). Among the 43 asphyxiated newborns that were scanned, 60% had an injury seen on the early MRI that was persistent on the late MRI (days 10 and 30), while 40% did not have an injury on both early and late MRIs. The authors concluded that the presence of injury detected by

Figure 6 | Phase 3: T2 magnetic resonance imaging (MRI) scores at (A) 3 h or (B) 7 days after hypoxic ischemia (HI). At both MRI time points, we found a very strong correlation between MRI score and histopathologic brain injury at 7 days. However, inspection of the correlation plots reveals less sensitivity for detection of subtle injury in MRI scores at 7 days. MRI scores are generally lower at 7 days than at 3 h after HI, and the subtle atrophy that is detected by quantitative histopathology is not readily apparent in T2 MRI at 7 days. (Dashed lines show 95% confidence interval for the mean.)

Therapeutic hypothermia may provide neuroprotection *via* multiple mechanisms. These include a reduction in brain energy utilization, normalization of protein synthesis, and regulation of microglial activity and cytokine production (36). In astrocytes, preservation of ATP may support the Na<sup>+</sup> gradients that drive glutamate uptake (37), and changes in glutamate transporter gene expression may limit glutamate release from astrocytes (38). In an adult rat cardiac arrest model, post-ischemic hypothermia reduced striatal glutamate and dopamine release during reperfusion (39), and in an adult rat stroke model, intraischemic mild hypothermia blocked the increase in extracellular glutamate and reduced the release of dopamine after four-vessel occlusion (40). Thus, a reduction in neuronal release of neurotransmitters after HI may play a role in early hypothermic neuroprotection. In adult rats subjected to cardiac arrest and resuscitation, hypothermia reduced levels of reactive oxygen species and activation of caspase-3 in the hippocampus, 12–24 h after arrest (41). Such changes in early excitotoxic mechanisms may underlie the very early neuroprotective effects of hypothermia observed here. Hypothermic neuroprotection in human infants requires that hypothermia be initiated during the latent phase, within 6 h of birth asphyxia (10, 42). Cooling within 3 h of birth provides better psychomotor outcomes than cooling initiated 3–6 h after birth (43), supporting the hypothesis that this therapy has important effects on early mechanisms of injury. An 1H-MR spectroscopy study in infants with HIE treated with hypothermia reported decreased glutamate and aspartate levels in the brain and improved energy homeostasis during hypothermia, compared to after rewarming, which supports an early effect on excitotoxic mechanisms (44). Further studies will be needed to evaluate the effect of hypothermia on other early HI injury mechanisms such as astrocyte glutamate

Doman et al. Early Hypothermic Neuroprotection in HIE

MRI during hypothermia may help identify infants who would benefit from adjunct therapy, especially as the majority of patients with injury also had clinical evidence of severe encephalopathy. Several studies and sub-study analyses have demonstrated that therapeutic hypothermia decreases MRI lesions in infants (17, 54). Our current study, in which we demonstrate an early response to hypothermia following HI, has translational relevance, because evaluating the response to hypothermia at an early stage may help identify patients who will benefit from hypothermia alone vs. those who may need additional therapies.

This study demonstrates both early and delayed neuroprotective effects of hypothermia, using T2 MRI in a mouse model of HIE. Because the progression of injury is likely to vary in different brain injury models, different species, and at different ages, the optimal imaging modality and timing for imaging and interventions such as hypothermia will need to be determined for each model. In the P7 CD1 mouse model used for our studies, T2 MRI provides an effective and efficient method for assessing injury as early as 3 h after HI. The very early neuroprotective effects of hypothermia observed in this study suggest that hypothermia has a notable impact on early mechanisms of injury and support the initiation of hypothermic intervention as soon as possible after diagnosis of HIE.

### ETHICS STATEMENT

All animal procedures were carried out in accordance with the recommendations of the National Research Council Guide for the

### REFERENCES


Care and Use of Experimental Animals (29) and were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University.

### AUTHOR CONTRIBUTIONS

MW, JZ, CN, AF, and MJ contributed to the conception and design of the study; CN and GD conducted experiments for Phase 1; MW, PC, SD, AG, and MG conducted experiments for Phases 2 and 3; SD and MW performed the statistical analysis; SD and AG scored MR images, collected histopathological data, and wrote the first draft of the manuscript; and MW, JZ, and SK wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

#### ACKNOWLEDGMENTS

The authors would like to thank Kazi Akhter and Jiadi Xu for assistance with MR imaging.

### FUNDING

Research reported in this publication was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award numbers U54 HD079123 and T32 HD007414. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

among neonates with hypoxic-ischemic encephalopathy: a randomized clinical trial. *JAMA* (2017) 318:57–67. doi:10.1001/jama.2017.7218


regulating the reverse transport of glutamate by astrocytes as mediated by neurons. *Neuroscience* (2013) 237:130–8. doi:10.1016/j.neuroscience.2013. 01.056


**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 Doman, Girish, Nemeth, Drummond, Carr, Garcia, Johnston, Kannan, Fatemi, Zhang and Wilson. This is an open-access article distributed*  *under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Bone Morphogenetic Protein (BMP)-3b Gene Depletion Causes High Mortality in a Mouse Model of Neonatal Hypoxic-Ischemic Encephalopathy

#### Yuko Ogawa<sup>1</sup> , Masahiro Tsuji <sup>1</sup> , Emi Tanaka<sup>1</sup> , Mikiya Miyazato<sup>2</sup> and Jun Hino<sup>2</sup> \*

*<sup>1</sup> Department of Regenerative Medicine and Tissue Engineering, National Cerebral and Cardiovascular Center, Suita, Japan, <sup>2</sup> Department of Biochemistry, National Cerebral and Cardiovascular Center, Suita, Japan*

#### Edited by:

*Carl E. Stafstrom, Johns Hopkins Medicine, United States*

#### Reviewed by:

*Robert W. Dettman, Northwestern University, United States Francisco Capani, University of Buenos Aires, Argentina*

> \*Correspondence: *Jun Hino jhino@ncvc.go.jp*

#### Specialty section:

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

Received: *01 December 2017* Accepted: *15 May 2018* Published: *05 June 2018*

#### Citation:

*Ogawa Y, Tsuji M, Tanaka E, Miyazato M and Hino J (2018) Bone Morphogenetic Protein (BMP)-3b Gene Depletion Causes High Mortality in a Mouse Model of Neonatal Hypoxic-Ischemic Encephalopathy. Front. Neurol. 9:397. doi: 10.3389/fneur.2018.00397* Bone morphogenetic proteins (BMPs) are a group of proteins that induce the formation of bone and the development of the nervous system. BMP-3b, also known as growth and differentiation factor 10, is a member of the BMPs that is highly expressed in the developing and adult brain. BMP-3b is therefore thought to play an important role in the brain even after physiological neurogenesis has completed. BMP-3b is induced in peri-infarct neurons in aged brains and is one of the most highly upregulated genes during the initiation of axonal sprouting. However, little is known about the role of BMP-3b in neonatal brain injury. In the present study, we aimed to describe the effects of BMP-3b gene depletion on neonatal hypoxic-ischemic encephalopathy, which frequently results in death or lifelong neurological disabilities, such as cerebral palsy and mental retardation. BMP-3b knockout and wild type mice were prepared at postnatal day 12. Mice of each genotype were divided into sham-surgery, mild hypoxia-ischemia (HI), and severe HI groups (*n* = 12–45). Mice in the HI groups were subjected to left common carotid artery ligation followed by 30 min (mild HI) or 50 min (severe HI) of systemic hypoxic insult. A battery of tests, including behavioral tests, was performed, and the brain was then removed and evaluated at 14 days after insult. Compared with wild type pups, BMP-3b knockout pups demonstrated the following characteristics. (1) The males exposed to severe HI had a strikingly higher mortality rate, and as many as 70% of the knockout pups but none of the wild type pups died; (2) significantly more hyperactive locomotion was observed in males exposed to severe HI; and (3) significantly more hyperactive rearing was observed in both males and females exposed to mild HI. However, BMP-3b gene depletion did not affect other parameters, such as cerebral blood flow, cylinder test and rotarod test performance, body weight gain, brain weight, spleen weight, and neuroanatomical injury. The results of this study suggest that BMP-3b may play a crucial role to survive in severe neonatal hypoxic-ischemic insult.

Keywords: bone morphogenetic protein-3b, neonatal brain injury, hypoxia-ischemia, hypoxic-ischemic encephalopathy, mice, growth and differentiation factor 10

### INTRODUCTION

Bone morphogenetic proteins (BMPs) form a subgroup in the transforming growth factor-β (TGF-β) superfamily and act as powerful morphogens that perform crucial roles during embryonic development (1, 2). Although they were first discovered as osteoinductive proteins, they were later recognized as critical regulators of nervous system development (2). BMP signaling plays dynamic roles in the development of the brain in the early stages of life, during which they sequentially induce neurogenesis and then astrogliogenesis (3). BMPs also regulate neurite outgrowth from immature neurons during forebrain development and are crucial for maintaining adult neural stem cell niches in the subventricular and subgranular zones (2). Previous studies have suggested that while certain BMPs are beneficial (4– 6), other BMPs are detrimental after ischemic brain injury (7, 8).

In 1996, we discovered a protein that was structurally similar to BMPs, especially BMP-3. We therefore named it BMP-3b (9, 10), and it is currently also known as growth and differentiation factor 10 (GDF10) (11). Unlike other BMPs, BMP-3b/GDF10 suppresses osteoblast differentiation by activating Smad2/3 signaling via the ALK4/ActRIIA receptors (12). We previously reported that BMP-3b is essential for head formation in Xenopus embryos and acts as a neural inducing factor (13). BMP-3b is strongly expressed in developing skeletal structures in embryos and in bones, the brain (especially the cerebellum), the aorta, and adipose tissues in adult rodents (14, 15). Despite the high level of BMP-3b expression observed in these organs, no obvious abnormality was noted in the development of these tissues and organs in BMP-3b knockout (KO) mice (14). It is conceivable that BMP-3b plays a role in brain injury events because it is highly expressed in the brain throughout life. Recently, Li and colleagues reported that BMP-3b is induced in peri-infarct neurons and that it enhanced axonal sprouting and functional recovery in a mouse model of stroke (16). However, the role of BMP-3b in injury to the developing brain has not yet been explored.

Perinatal/neonatal hypoxic-ischemic encephalopathy (HIE) is the most serious problem encountered in neonatal neurology. HIE occurs in 1–2/1000 births and can be caused by a variety of events, such as abruptio placentae and umbilical cord prolapse. Although the development of therapeutic hypothermia to treat neonatal HIE was a landmark achievement in neonatal care, the mortality rate in HIE patients treated for hypothermia in neonatal intensive care units remains approximately 10% (17). Additionally, many of the infants who survive suffer lifelong neurological sequelae, such as cerebral palsy and mental retardation. Therefore, novel therapies need to be developed.

The aim of the present study was to examine the physiological roles of endogenous BMP-3b in neonatal HIE using mouse models in which BMP-3b has or has not been depleted. Specifically, we focused on the effects of BMP-3b gene depletion on outcomes related to neurological damage.

### MATERIALS AND METHODS

#### Animal Model of Hypoxic-Ischemic Encephalopathy

All procedures were performed according to protocols approved by the Animal Care and Use Committee of the National Cerebral and Cardiovascular Center. The BMP-3b/GDF10 KO mice were kindly provided by Se-Jin Lee. BMP-3b KO mice were backcrossed with C57BL/6 mice more than 10 times. In all, 41 BMP-3b KO males, 46 BMP-3b KO females, 22 wild type males, and 27 wild type females were prepared. Mouse pups were obtained at postnatal day 12 (P12), which is considered the equivalent of human term newborns at P0. Mice of each genotype were divided into the following four groups: no-surgery control, sham-surgery control, mild hypoxia-ischemia (HI), or severe HI (n = 12–45) **(Table 1**). Littermates were randomly assigned to one of the four groups.

We used the Rice-Vannucci model (18), which is the most widely used animal model of HIE. Under isoflurane anesthesia (4.0% for induction and 1.5 to 2.0% for maintenance), the left carotid artery was separated from the jugular vein and connective tissue, doubly ligated, and then severed between the two ligatures. All the surgeries were performed by a single experimenter to reduce variability between experimenters. After surgery, the mouse pups were allowed to recover for 1–2 h while separated from their dam. In the mice exposed to hypoxia, littermates were placed in an enclosed vented chamber that was flushed with a humidified mixture of 8% oxygen (balance nitrogen) for either 30 min (mild HI) or 50 min (severe HI). Mice in the sham-surgery control group underwent the same procedure except for artery ligation and exposure to hypoxia. The ambient temperature inside the chamber was kept at 33.0◦C, which is the temperature the pups are normally exposed to while huddling with their dam. The mice were then allowed to recover for 1 h in an incubator set to an ambient temperature of 33.0◦C, and they were then returned to their dams. All analyses were performed by investigators who were blinded to the experimental group.

### Cerebral Blood Flow Measurements

We measured cortical surface cerebral blood flow (CBF) with a laser speckle flowmetry imaging system (Omegazone, Omegawave Inc., Tokyo, Japan). Measurements were obtained through the intact skull and an open scalp at 24 h after the surgery, as previously described (19). We measured CBF in both the ischemic core region and the penumbra region (the broader region between the ischemic and intact regions) (**Figures 1A,B**). We also measured CBF in the corresponding region of the contralateral hemisphere.

#### Behavioral Tests

The no-surgery controls were tested only in cylinder and rotarod tests. The data obtained in both the no-surgery and sham-surgery controls were combined because there was no difference in performance between the groups.



*BMP-3b KO, bone morphogenetic protein-3b knockout; HI, hypoxia-ischemia. <sup>a</sup>After HI-50 min insult, the mortality rate in BMP-3b KO males was significantly higher than the rate in wild type males. p* < *0.01. <sup>b</sup>After HI-50 min insult, the mortality rate in BMP-3b KO males was significantly higher than the rate in BMP-3b females. p* < *0.05.*

FIGURE 1 | (A) A scheme of the experiment schedule. P; postnatal day. (B) Using laser speckle flowmetry, we measured cerebral blood flow (CBF) in the "penumbra" and "ischemic core" regions, both of which were set as shown, at 24 h after the surgery. A line was drawn from bregma to lambda. This line was used to draw a square. The line perpendicular to the bregma-lambda line was divided into four equal segments. The second quarter-rectangle from the center was defined as the "penumbra" region. The posterior half of the most lateral segment was defined as the "ischemic core" region. We also measured CBF in the corresponding region of the contralateral hemisphere.

#### Cylinder Test

Forelimb functional impairment was evaluated in cylinder tests 7 days after the surgery (P19), as previously described (20). We placed each mouse in a transparent cylinder (14 cm inner diameter and 30 cm high) and video recorded its behavior until it reared and touched the cylinder wall with its forepaw(s) a minimum of 20 times. In most mice, this took 3–5 min. We counted the number of contacts with the wall by the left or right forepaw during a rear separately. Paw preference in wall touch (i.e., asymmetry of forelimb use) was calculated using the following formula: (left [non-impaired side]—right [impaired side])/(left + right) × 100%.

#### Rotarod Test

Sensorimotor skills were evaluated in rotarod tests 8 days after the surgery (P20). The rotarod was accelerated from 4 to 40 rpm over 5 min (Ugo Basile, Co., Ltd., Gemonio, Italy). We recorded the time until the pup fell off the rotating drum in five consecutive sessions. The average time spent on the drum was used for the analysis.

#### Open-Field Test

Spontaneous activities and exploratory behaviors were evaluated in open-field tests 10–13 days after the surgery (P22–25), as previously described (21). Each animal was allowed to search freely in a box (30 × 30 cm) for 30 min in a light environment and then for 30 min in a dark environment. Infrared beams were mounted at specific intervals on the X-, Y-, and Z-banks of the open-field box (Taiyo Electric Co., Ltd, Osaka, Japan). The total number of beam crossings made by the mouse was counted and scored as "locomotion" for horizontal movements and "rearing" for vertical movements.

#### Neuroanatomical Analysis

Fourteen days after surgery, the mice were anesthetized with an overdose of pentobarbital and then perfused via the left ventricle with phosphate-buffered saline followed by 4% paraformaldehyde. After perfusion, the brains of the mice were removed and weighed. After they were placed in the same fixative for 2–3 days, the brains were coronally sectioned into 1 mmthick slices using a mouse brain slicer (Neuroscience Idea Co., Ltd., Osaka, Japan). Areas in the ipsilateral and contralateral hemispheres of each brain section were measured using the ImageJ program (NIH, Bethesda, MD, USA). Hemispheric volume was estimated by integrating the hemispheric areas. The brains were further thin-sectioned, 6 µm-thick, and stained with hematoxylin-eosin. We evaluated neuropathological injury semi-quantitatively in four brain regions (the cerebral cortex, hippocampus, striatum, and thalamus), as previously described (22). The total score (0–22) was obtained by calculating the sum of the ratings in the four brain regions. We also evaluated white matter injury semi-quantitatively in corpus callosum in 6 µmthick sections with Klüver-Barrera staining. The severity of white matter injury was scored (0–3) as previously described (23).

#### Statistical Analysis

The mortality rate of the animals was analyzed with the Fisher's exact test with a Bonferroni correction for multiple comparisons. Data on body, brain, and spleen weights; CBF; cylinder and rotarod tests; and hemispheric volumes were assessed using twoway analysis of variance (ANOVA) followed by a Bonferroni test to analyze the effects of HI insult and genotype. Because the injury scores were not distributed normally, they were assessed with the Kruskal-Wallis test followed by Dunn's multiple comparisons test. Data from the open-field tests were assessed using two-way repeated-measures ANOVA followed by a Bonferroni test. Although data of sham, HI-30 min, and HI-50 min groups were analyzed all together, data of HI-30 min and those of HI-50 min in the open-field test are presented separately with the same data of sham group in order to avoid busy figures. Data of males and females were analyzed and are presented separately. Some of the data, however, are presented males and females combined as there were no significant difference between males and females. Differences were considered significant at p < 0.05. Significant differences between groups with the same genotype or that received the same magnitude of insult are presented in this manuscript and figures. Other comparisons, such as comparisons between BMP-3b KO mice treated with sham-surgery and wild type mice treated with severe hypoxia (HI-50 min) insult, are not presented. The results are presented as the mean ± standard deviation (SD) except for data from the open-field tests, which are presented as the mean ± standard error of the mean (SEM).

## RESULTS

#### Body Weights and Mortality Rates

The mean body weight at the time of surgery (P12) was 6.20 ± 1.10 g in BMP-3b KO males, 5.77 ± 0.85 g in BMP-3b KO females, 5.58 ± 0.89 g in wild type males, and 5.48 ± 0.73 g in wild type females (**Figure 2A**). Body weights did not differ among these groups.

None of the animals subjected to a sham-surgery or the HI-30 min insult died during the 14-day observation period (until P26) regardless of the genotype. In animals subjected to the HI-50 min insult, the mortality rate was 70% in male BMP-3b KO pups, and this was significantly higher than the rate observed in male wild type pups (0%). Almost all deaths occurred during the last 15 min of hypoxic exposure or immediately after it ended. The mortality rate was 32% in female BMP-3b KO pups, and this was not significantly different from the rate observed in female wild type pups (22%) (**Table 1**,**Figure 2B**).

FIGURE 2 | (A) Body weights of mouse pups at the time of surgery (postnatal day 12, P12). Bone morphogenetic protein (BMP)-3b knockout (KO) males, *n* = 41; BMP-3b KO females, *n* = 46; wild type males, *n* = 22; and wild type females, *n* = 27. Mean ± standard deviation (SD). (B) The numbers of animals dead and survived in each group after the hypoxic-ischemic (HI) insult. One female animal died 6 days after the 50 min HI insult. (C) Body weight gain in male pups during the observation period from P12 to P26. (D) Body weight gain in female pups during the observation period. Sham, sham-surgery controls; HI-30 min, hypoxia-ischemia for 30 min; HI-50 min, hypoxia-ischemia for 50 min. \**p* < 0.05, \*\**p* < 0.01.

In the animals treated with a sham-surgery, body weight gains during the 14-day observation period were similar between wild type pups and BMP-3b KO pups. However, weight gain was lower in the HI-50 min insult groups than in the sham-surgery groups. The magnitude of this reduction did not differ according to genotype. When stratified by sex, BMP-3b KO males but not wild type males had significantly lower body weight gains. In contrast, wild type females but not BMP-3b KO females had significantly lower body weight gains (**Figures 2C,D**).

#### Cerebral Blood Flow

Surface CBF in the right hemisphere (i.e., the contralateral side) was not different between wild type and BMP-3b KO pups at 24 h after sham-surgery [40.0 ± 7.3 and 38.0 ± 6.0 (arbitrary units), respectively]. There was also no significant different in CBF in the contralateral hemisphere between wild type and BMP-3b KO pups at 24 h after HI-30 min insult [43.0 ± 3.6 and 38.1 ± 7.6 (arbitrary units), respectively]. In the ipsilateral hemisphere, CBF was lower after HI insult (**Figure 1B**). The ratio of CBF in the ischemic core and penumbral regions of the ipsilateral hemisphere to the corresponding regions in the contralateral hemisphere was not different between wild type and BMP-3b KO pups (**Figures 3A,B**). There was no difference according to gender within each group.

#### Cylinder Test

The cylinder test was performed at 7 days after the surgery (P19), and the results showed that motor impairment of the affected forelimb (i.e., the right side) had occurred in pups subjected to HI insult (**Figure 4A**). While forepaw preferences were significantly different between types of insult, they did not differ by genotype (two-way ANOVA). Post-hoc tests showed that significant group differences were observed only between the sham and HI-50 min groups among the wild type mice. There was no difference according to sex within any group.

### Rotarod Test

Rotarod performance was measured at 8 days after sham-surgery (P20), and no difference was detected between the wild type and BMP-3b KO pups (**Figure 4B**). However, the Rotarod test results demonstrated that sensorimotor impairment occurred in pups subjected to HI insult. Similar to the results of cylinder tests, there was no difference in time spent on the rotarod according to genotype, whereas they were significantly different according to the type of insult (two-way ANOVA). Post-hoc tests, however, did not reveal any significant group differences, and there was no difference within each group according to sex.

#### Open-Field Test

Spontaneous activities were evaluated in open-field tests performed at 10–13 days after the surgery (P22–25). With regard for locomotion (i.e., horizontal movement), in the animals submitted to sham-surgery, in a light environment, BMP-3b KO males were significantly more hyperactive than the wild type males were (**Figure 5A**). After HI-50 min insult but not after HI-30 min insult, in a dark environment, BMP-3b KO males became significantly more hyperactive than did the wild type

males (**Figure 5B**). There was no difference between the BMP-3b KO females and wild type females after sham-surgery, HI-30 min insult, or HI-50 min insult (**Figures 5C,D**).

With regard for rearing (i.e., vertical movement), after the HI-30 min insult, throughout the 60 min-session, including the light and dark environment periods, BMP-3b KO males were significantly more hyperactive than wild type males were (**Figure 5E**). However, after the HI-50 min insult, BMP-3b KO males were not more hyperactive than wild type males were (**Figure 5F**). After sham-surgery, BMP-3b KO females were significantly more hypoactive than wild type females during the last 5 min of the 60 min-session (in a dark environment) (**Figure 5G**). In contrast, after the HI-30 min insult, BMP-3b KO females were significantly more hyperactive than wild type females during the first 5 min of 60 min-session (in a light environment) (**Figure 5G**). After the HI-50 min insult, BMP-3b KO females did not exhibit significantly different spontaneous activities compared with the wild type females (**Figure 5H**).

#### Brain Weight

Brain weights were similar between wild type pups and BMP-3b KO pups at 14 days after sham-surgery (P26). The HI-50 min insult but not the HI-30 min insult resulted in lower brain

FIGURE 4 | (A) Hemiplegia, i.e., disuse of the affected right forepaw, was evaluated in cylinder tests 7 days after surgery (i.e., postnatal day 19, P19). The number of forepaw contacts made with the wall during rearing were counted. Paw preference was calculated as follows: (left [non-impaired side]—right [impaired side])/(left + right) × 100%. (B) Sensorimotor skills were evaluated in rotarod tests 8 days after surgery (P20). The average time until the animal fell off the rotating drum in 5 consecutive sessions was analyzed. The box shown extends from the 25 to 75th percentiles. The whiskers shown are drawn down to the 5th and up to the 95th percentile. Points below and above the whiskers are drawn as individual dots. The no-surgery controls were combined with the sham group because there was no difference in performance between these groups. In the cylinder test, Wild Sham, *n* = 12; Wild HI-30 min, *n* = 16; Wild HI-50 min, *n* = 8; BMP-3b KO Sham, *n* = 22; BMP-3b KO HI-30 min, *n* = 12; and BMP-3b KO HI-50 min, *n* = 17. In the rotarod test, Wild Sham, *n* = 7; Wild HI-30 min, *n* = 10; Wild HI-50 min, *n* = 10; BMP-3b KO Sham, *n* = 25; BMP-3b KO HI-30 min, *n* = 12; and BMP-3b KO HI-50 min, *n* = 9. \**p* < 0.05.

weight at P26 (**Figure 6A**). There were no genotype-dependent differences between animals treated with the same degree of insult. There were no sex-dependent differences in any genotype or insult.

#### Hemispheric Volume Loss

In animals examined at 14 days after insult, the HI-30 min insult caused a mild loss in hemispheric volume, but the volumes in these animals were not significantly different from those observed in the sham-surgery-treated animals in either males (**Figure 6C**) or females (**Figure 6D**). The HI-50 min insult caused a significant loss in hemispheric volume in wild type and BMP-3b KO males and wild type females (**Figures 6B–D**). There were no significant genotype- or sex-dependent differences.

#### Neuropathological Injury

Similar to the previous reports, we did not find any macroscopic brain anomaly or sign of a specific disease in BMP-3b KO mice with sham-surgery. The HI-30 min and HI-50 min insults caused neuropathological injury in animals examined at 14 days after insult (**Figures 7A,D**). Neuropathological injury evaluated in hematoxylin-eosin stained sections was not significantly different between the wild type and BMP-3b KO groups whether they were either analyzed according to brain region (i.e., the cerebral cortex, striatum, hippocampus, and thalamus) (**Figures 7A,B**) or by total injury score (**Figure 7C**). White matter injury evaluated in sections with Klüver-Barrera staining was not significantly different between the wild type and BMP-3b KO groups (**Figure 7E**). There was no significant sex-dependent difference.

### Spleen Weight

Spleen weight is an indicator of inflammation and was similar between the wild type and BMP-3b KO pups at 14 days after sham-surgery (P26). Both HI insults reduced spleen weight (two-way ANOVA) (**Figure 8**). However, post-hoc tests did not reveal any significant differences according to genotype or the magnitude of the insult.

### DISCUSSION

This is the first report to describe the phenotypes of BMP-3b/GDF10 KO mouse pups with brain injury. In sham-surgerytreated animals, BMP-3b gene depletion caused no changes in any parameter measured in this study except for hypoactive rearing in females. BMP-3b gene depletion resulted in (1) a striking increase in mortality in males treated with severe HI insult, (2) significantly hyperactive locomotion in males treated with severe HI insult, and (3) significantly hyperactive rearing in males and females treated with mild HI insult. However, BMP-3b gene depletion did not affect the results of any of the other tests (i.e., CBF, cylinder test, rotarod test, body weight gain, brain weight, spleen weight, % cerebral hemispheric volume loss, and semi-quantitative neuropathological injury score in four brain regions and white matter injury score). The most prominent finding in the current study was that BMP-3b may play a crucial role in helping animals survive severe HI insult.

During the last 15 min of the 50 min systemic hypoxic exposure period used in this study, the majority of BMP-3b KO male pups had a low rate of respiration, and respiration was shallow, ultimately leading to respiratory arrest. The specific mechanisms that lead to death during and immediately after HI insult are difficult to analyze in immature mice because physiological parameters, such as blood pressure, cannot be measured in these animals, especially during exposure to hypoxia. The mouse pups used in this study were subjected to unilateral carotid artery ligation followed by temporal systemic exposure to hypoxia. Hence, the HI insult was far more severe in the brain than in other organs, including the heart. The brain injury induced in this model was limited to the ipsilateral cerebral hemisphere, particularly the region that receives its blood flow from the middle cerebral artery. We hypothesize that brainstem dysfunctions inflicted by this type of cerebral injury may have been the cause of the observed deaths.

FIGURE 5 | Spontaneous activities were evaluated in open-field tests 10–13 days after surgery (P22-25). Mice were allowed to search freely in a box for 30 min while in a light environment and for the next 30 min while in a dark environment. The total number of times that the infrared beams set at specific intervals along the X-, Y-, and Z-banks of the box were crossed by each animal was counted and scored as "locomotion" for horizontal movement (males: A,B, females: C,D) and as "rearing" for vertical movement (males: E,F, females: G,H). The experiments were designed and performed in six groups, i.e., sham, HI-30 min, and HI-50 min groups, each containing both wild type and BMP-3b KO subgroups. To avoid the use of busy graphs with six lines, the HI-30 min and HI-50 min groups are presented separately, each with the same sham group. Mean ± standard error of the mean (SEM). AU, arbitrary unit. Males: Wild Sham, *n* = 6; Wild HI-30 min, *n* = 9; Wild HI-50 min, *n* = 7; BMP-3b KO Sham, *n* = 4; BMP-3b KO HI-30 min, *n* = 7; BMP-3b KO HI-50 min, *n* = 4. Females: Wild Sham, *n* = 7; Wild HI-30 min, *n* = 11; Wild HI-50 min, *n* = 7; BMP-3b KO Sham, *n* = 5; BMP-3b KO HI-30 min, *n* = 9; and BMP-3b KO HI-50 min, *n* = 13. #*p* = 0.08, \*\**p* < 0.01. a: *p* < 0.05, KO HI vs. KO Sham. b: *p* < 0.05, KO HI vs. Wild HI. b′ : *p* < 0.05, KO Sham vs. Wild Sham.

HI insult caused significantly more severe brain damage in the survivors with BMP-3b gene depletion than in those without the gene depletion. The BMP-3b KO pups were more hyperactive than the wild type pups were. These results suggest that BMP-3b plays a role in protecting against HI-induced brain damage and/or promoting recovery from this type of damage in survivors. We had expected that loss of BMP3b was more markedly detrimental to the brain both morphologically and behaviorally after the HI injury as suggested by the substantially high mortality during hypoxia in BMP3b KO mice. However, the high mortality in the KO group (as high as 70% in males) may have skewed the results, as only the resilient top 30% of pups subjected to the HI insult survived and were assessed. We considered that was the reason why the difference between wild type and BMP3b KO mice were not so remarkable and limited to only some of evaluations we performed.

\*\**p* < 0.01.

Semi-quantitative white matter injury scores were evaluated in Klüver-Barrera-stained sections. The maximum (i.e., worst) score was 3. Each group, *n* = 5. \**p* < 0.05,

No previous study has examined the role of BMP-3b after a brain injury except for one study performed by Li et al. (16) that evaluated adult stroke. They showed that BMP-3b expression is induced in peri-infarct neurons in mice and humans. They also demonstrated that BMP-3b induces axonal sprouting and enhances functional recovery after stroke in rodents and that knocking down BMP-3b blocks axonal sprouting and reduces recovery. Our results are basically in line with theirs. We found that suppressing the expression of the BMP-3b gene is deleterious to ischemic brain injury. However, the deleterious effects of BMP-3b gene suppression seemed to be more consistent and profound in their study. It may not be possible to directly compare the data from two such different studies. Several potential reasons for the observed differences are conceivable. For instance, (1) the ages of the models were different in that they used adult animals whereas we used immature pups; (2) the pathophysiology of the models was different in that theirs was stroke with photothrombosis whereas ours was HI; and (3) because the mortality rate observed in our study was high, it is possible that only inherently resilient sur'rs remained for evaluation (the mortality rate in the BMP-3b knock-down used in the report by Li and colleagues was not provided). Despite the fact that cerebral ischemia was the main cause of injury in both models, neonatal HIE and adult stroke involve very different conditions from a number of standpoints, including their mechanisms of disease progression, clinical symptoms, and outcomes. As is the case in adult stroke patients, a photothrombosis model generally involves the induction of clearly demarcated and small ischemic lesions in the cerebral cortex with no or little reperfusion. In contrast, as has been observed in human newborns with HIE, the effects of HI in neonatal models affect most parts of the cerebral hemisphere and reperfusion proceeds immediately after the end of hypoxic exposure. The results of the present study support the notion that preclinical studies should be performed in an animal model appropriate to the disease.

To the best of our knowledge, this is the first report that revealed the loss of BMP-3b affected the survival rate after injuries. Hence, there is no study that BMP-3b was used to revert the mortality after brain injuries or other types of insult. The functional roles of BMP-3b in events following injuries/diseases remain largely unknown, although recent studies have gradually begun to unveil them. For example, BMP-3b selectively activates TGF-β receptor (TGF-βI/II)-dependent Smad3 phosphorylation and attenuates tumor formation (24). BMP-3b also increases energy expenditure and protects high-fat diet-induced obesity by suppressing peroxisome proliferator-activated receptor γ (PPARγ) (25).

The roles of other BMPs in the events involved in cerebral ischemia have been explored in several studies. The results of these studies suggest that BMPs may play detrimental roles in ischemic brain injury. For example, Dizon and colleagues reported that the levels of BMP-4 protein increase following HI insult in neonatal mouse pups and that inducing the transgenic expression of noggin, a BMP antagonist that binds BMP-2/4 with high affinity, confers a survival advantage (e.g., 90% of noggin-overexpressing pups and 59% of wild type pups survived) and ameliorates HI-induced white matter injury (8). BMP-3b and BMP-2/4 are mutually antagonistic (12), and the results of this previous study showed that antagonizing BMP-2/4 enhanced survival in neonatal HI, in line with our results. These data suggest that endogenous BMP-3b may enhance survival in neonatal HI. Using a different model of adult mice with permanent middle cerebral artery occlusion (MCAO), the same laboratory reported that noggin overexpression increased the number of PDGFRα <sup>+</sup> cells (oligodendrocyte progenitor cells) in the ischemic boundary zone and ameliorated brain injury (7). Conversely, other studies have suggested that BMPs may exert beneficial roles in ischemic brain injury. For example, Wang and colleagues reported that the intracerebrally administering BMP-6 before inducing transient MCAO ameliorated brain injury in adult rats (5). The same group reported that BMP-7 signaling was increased in the brain after transient MCAO and that intravenously administering BMP-7 after inducing transient MCAO improved brain injury in adult rats (6). Treatment with BMP-7 did not change the mortality rate. Another group reported that the intracisternal administration of

#### REFERENCES


BMP-7 enhanced functional recovery in a rat model of stroke (4). Several possible factors may explain the seemingly contradictory effects of BMPs. For example, the ligands BMP-4 and BMP-6/7 have different preferred receptors.

The results of our study clearly demonstrate that BMP-3b protects immature mice from severe HI insult. Further studies are warranted to explore the precise mechanism of BMP-3b' beneficial effect and the therapeutic potential of using BMP-3b supplementation in neonatal HIE.

### AUTHOR CONTRIBUTIONS

YO performed the experiments (behavioral tests, etc.), analyzed the data, and prepared the figures. MT designed the study, performed the experiments (animal surgeries, etc.), and wrote the manuscript. ET performed the experiments (behavioral tests, etc.). JH designed the study, prepared the animals and critically revised the manuscript. MM critically revised the manuscript for important intellectual content. All authors gave their approval to the manuscript.

### FUNDING

This research was supported by the Intramural Research Fund for Cardiovascular Diseases of the National Cerebral and Cardiovascular Center of Japan.

### ACKNOWLEDGMENTS

We thank Dr. Se-Jin Lee for providing the BMP-3b/GDF10 KO mice. We thank Drs. Mariko Harada-Shiba and Kenji Kangawa for helpful discussions. We thank Ritsuko Maki, Mika Kitazume, and Manami Sone for their excellent technical assistance.


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

,

# Alterations of Both Dendrite Morphology and Weaker Electrical Responsiveness in the Cortex of Hip Area Occur Before Rearrangement of the Motor Map in Neonatal White Matter Injury Model

Yoshitomo Ueda<sup>1</sup> , Yoshio Bando<sup>2</sup> , Sachiyo Misumi <sup>1</sup> , Shino Ogawa1,3, Akimasa Ishida<sup>1</sup> Cha-Gyun Jung<sup>1</sup> , Takeshi Shimizu<sup>1</sup> and Hideki Hida<sup>1</sup> \*

#### Edited by:

*Masahiro Tsuji, Kyoto Women's University, Japan*

#### Reviewed by:

*Jacques-Olivier Coq, UMR7289 Institut de Neurosciences de la Timone (INT), France Lauren Jantzie, University of New Mexico, United States*

> \*Correspondence: *Hideki Hida hhida@med.nagoya-cu.ac.jp*

#### Specialty section:

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

Received: *05 January 2018* Accepted: *25 May 2018* Published: *19 June 2018*

#### Citation:

*Ueda Y, Bando Y, Misumi S, Ogawa S, Ishida A, Jung C-G, Shimizu T and Hida H (2018) Alterations of Both Dendrite Morphology and Weaker Electrical Responsiveness in the Cortex of Hip Area Occur Before Rearrangement of the Motor Map in Neonatal White Matter Injury Model. Front. Neurol. 9:443. doi: 10.3389/fneur.2018.00443* *<sup>1</sup> Department of Neurophysiology and Brain Science, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan, <sup>2</sup> Department of Functional Anatomy and Neuroscience, Asahikawa Medical University, Asahikawa, Japan, <sup>3</sup> Department of Obstetrics and Gynecology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan*

Hypoxia-ischemia (H-I) in rats at postnatal day 3 causes disorganization of oligodendrocyte development in layers II/III of the sensorimotor cortex without apparent neuronal loss, and shows mild hindlimb dysfunction with imbalanced motor coordination. However, the mechanisms by which mild motor dysfunction is induced without loss of cortical neurons are currently unclear. To reveal the mechanisms underlying mild motor dysfunction in neonatal H-I model, electrical responsiveness and dendrite morphology in the sensorimotor cortex were investigated at 10 weeks of age. Responses to intracortical microstimulation (ICMS) revealed that the cortical motor map was significantly changed in this model. The cortical area related to hip joint movement was reduced, and the area related to trunk movement was increased. Sholl analysis in Golgi staining revealed that layer I–III neurons on the H-I side had more dendrite branches compared with the contralateral side. To investigate whether changes in the motor map and morphology appeared at earlier stages, ICMS and Sholl analysis were also performed at 5 weeks of age. The minimal ICMS current to evoke twitches of the hip area was higher on the H-I side, while the motor map was unchanged. Golgi staining revealed more dendrite branches in layer I–III neurons on the H-I side. These results revealed that alterations of both dendrite morphology and ICMS threshold of the hip area occurred before the rearrangement of the motor map in the neonatal H-I model. They also suggest that altered dendritic morphology and altered ICMS responsiveness may be related to mild motor dysfunction in this model.

Keywords: hypoxia-ischemia in premature infants, white matter injury, intracortical microstimulation (ICMS), golgi staining, cortical layer I-III, hip area, dendritic branches

### INTRODUCTION

Although advances of perinatal medicine have improved the survival rate of preterm infants (1, 2), these infants often have neurological insults due to hypoxia-ischemia (H-I) accompanied with brain immaturity (3). A profound shift in the features of H-I over time has also been reported: a milder form of H-I, characterized by nondestructive lesions, is reportedly increasing in prevalence (4–7). However, there is a pressing need to expand current understanding of the mechanisms underlying the effects of brain insult and to develop treatments for the resulting behavioral and cognitive dysfunction in development.

Preterm infants have a higher risk of neonatal white matter injury (WMI) because late oligodendrocyte progenitor cells (OPC), which are abundant at gestational weeks 20–28 in humans (8), are particularly susceptible to H-I (9–11). Ischemiainduced neuroinflammation and prenatal inflammatory response are known to be related to neonatal WMI (12–15). Neonatal WMI causes neurodevelopmental deficits during development, including motor deficits (such as cerebral palsy), learning disorders, and behavioral difficulties (including attention deficit/hyperactivity disorder) (16–18). Various animal models of preterm infants have been reported, including models using sheep (19, 20), rabbit (21–23), piglets (24–26), and rodents (15, 27–32). Among the available rodent models, the Rice-Vannucci model (33) and its variations have been commonly used (27, 28, 30). Other models, created by unilateral uterine artery ligation of dams at embryonic day (E) 17 (34, 35) or transient bilateral occlusion of the uterine arteries at E18 (15, 29, 31), have also been reported.

We previously established a rat model of neonatal WMI produced by right carotid artery occlusion followed by under 6% oxygen for 1 h (36), based on the notion that late development of OPC (preOLs) in the immature brain is specifically associated with vulnerability to H-I (11). In this neonatal WMI model, actively proliferating OL progenitors are primarily damaged, with a decreased number of mature OL cells and hypomyelination in the sensorimotor cortex in adulthood, indicating that impaired motor coordination is induced by impaired myelination in layer I-IV rather than neuronal loss (37). This neonatal WMI model exhibited moderate motor deficits, especially in the hindlimbs, accompanied with disorganization of oligodendrocyte development in layers II–III of the sensorimotor cortex (38).

As neonatal WMI is a complex amalgam of destructive developmental disturbances in premature infants (5), changes in neuronal circuit formation and connectivity in the cerebral cortex may be affected by the close relationship between neuronal development and myelination. Although our model has revealed motor coordination dysfunction without the loss of cortical neurons (37, 38), it remains unclear how motor dysfunction is induced in this model.

To reveal the mechanisms underlying imbalanced motor coordination in the neonatal WMI model, we investigated electrical responsiveness, and dendrite morphology in the sensorimotor cortex in adulthood. Thus, intracortical microstimulation (ICMS) in the hindlimb motor cortex was performed to reveal electrophysiological responses, and morphological changes were examined in the motor cortex using Golgi staining.

We found that the cortical motor map was significantly changed in neonatal WMI model at 10 weeks of age: the cortical area related to hip joint movement was reduced while the area related to trunk movement was increased. In addition, we found that layer I–III neurons on the H-I side had more dendrite branches compared with the contralateral side in Golgi staining. ICMS and Golgi staining performed at 5 weeks of age revealed that the minimal ICMS current to evoke twitches was higher on the H-I side while the motor map was unchanged, and more dendrite branches in layer I–III neurons were shown in the H-I side, indicating that alterations of both dendrite morphology and ICMS threshold of the hip area occurred before the rearrangement of the motor map in the neonatal H-I model.

#### MATERIALS AND METHODS

#### Animals

Animal care was carried out according to the guidelines of the Institute for Experimental Animal Sciences, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan. All experimental procedures were approved by the committee of animal experimentation of Nagoya City University Medical School, and appropriate measures were taken to minimize the pain and discomfort of the animals used in the study. Until weaning, 10 male Wistar rat pups (Japan SLC, Japan) were reared with a foster mother. After weaning, 3–5 rats were housed together in each cage. A total of 48 rats from multiple different litters were euthanized for this study, including 17 rats as controls, and 31 rats as neonatal H-I models.

#### Neonatal WMI Model

A procedure involving H-I treatment was conducted to produce neonatal WMI model, as we confirmed many characteristics of the injury in the white matter in our previous papers (36–38). Rat pups at postnatal day 3 (P 3) were subjected to right common carotid artery cauterization under isoflurane (Pfizer, NY, USA) anesthesia \$(5% [v/v] induction, 1.0% [v/v]) and kept at 37◦C on a heat pad. After a 2-h recovery period with their dam, pups were exposed to 6% (v/v) O<sup>2</sup> hypoxia for 60 min in a container submerged in a 38◦C water bath.

As a control condition, we performed 6% hypoxia after a sham operation: skin incision, separation around the right common carotid artery and skin suture were performed before the hypoxia. We used this condition as controls for ICMS because we previously confirmed that there was no significant difference between "sham operation with normoxia" and "sham operation with 6% hypoxia" in histology (the numbers of neurons and microglia, the morphology of microglia, and the intensity of myelin basic protein staining) or behavior (motor function by motor deficit score) (38).

#### Intracerebral Microstimulation (ICMS)

We first performed ICMS at 10 weeks of age to investigate differences in the pattern of motor maps (n = 9 for neonatal WMI and n = 8 for controls). ICMS was then performed at 5 weeks of age to reveal the developmental pattern of the change (n = 11 for neonatal WMI and n = 9 for controls), using a procedure similar to methods described elsewhere (39–41) with some modifications (42).

Rats underwent anesthesia with a mixture of ketamine (60 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.), and dexamethasone (0.1 mg/kg) was then administered into the right gluteal fascia to prevent brain edema. Thirty minutes after the first injection, a mixture of ketamine (30 mg/kg) and xylazine (5 mg/kg) was injected into the right gluteal fascia to maintain constant anesthesia during cranial scraping. Supplemental injection of ketamine (40 mg/kg), or a mixture of ketamine (15 mg/kg) and xylazine (2.5 mg/kg), was occasionally used when animals exhibited vibrissae twitching, muscle twitching and/or tachypnea during recording, followed by a 10-min break from recording to keep the measurement consistent.

Rats were fixed on a stereotaxic apparatus (Narishige, Japan), and craniotomy was performed to expose the right sensorimotor cortex. The coordinates for the openings were 1.0–3.0 mm posterior and 1.0–3.0 mm lateral from the bregma, known as the sensorimotor area, including the hindlimb area. The exposed area was regularly spaced out over a 0.5-mm grid. Under a microscope, a glass-insulated tapered tungsten electrode (250µm shank diameter, impedance, 0.7 M; Alpha Omega #380-080607-11, GA, USA) was lowered perpendicularly into the cortex to 1,600µm below the cortical surface, a depth corresponding to layer V of the sensorimotor cortex. The electrode was then adjusted by ±200µm to find the appropriate location for measuring the spike pattern with an oscilloscope and loudspeaker. Thirty biphasic pulses (333 Hz, 200 µs pulse duration) produced by an electrical stimulator (Nihon Kohden, Japan) connected to an isolator (Nihon Kohden, Japan) were passed through the electrode every 2 s, and the response of the cortex was amplified using an amplifier (AM Systems, WA, the USA) connected to an oscilloscope. Starting at a current of 0 µA, intensity was increased in 10 µA steps until twitching of the hip joint, knee joint, foot joint, digit and trunk were confirmed by palpation and/or visual inspection. The current was then gradually decreased until the twitch was no longer detectable. This level was defined as the current threshold. If no twitches were evoked at 200 µA, the site was defined as "non-responsive."

ICMS data are represented as cortical maps with current thresholds in **Figure 1**, (**Supplemental Figure 1**, and as heat maps in **Figures 2A**, **4A**. Cortical maps were created using areas that responded to the stimulation and are shown as 5-colored code in the grid: in each animal, we noted for each stimulated square whether the stimulation resulted in twitching of the hip joint (red), knee joint (yellow), foot joint (light blue), and trunk (green), or the non-responsive square (gray). The generated heat maps represent the occurrence rate of each stimulated spot in the sensorimotor area, and were calculated separately for each experimental group: for each point, the number of animals in which the movement was elicited was taken over the number of times that spot was stimulated in each group. For example, if three animals in a group responded to cortical stimulation at a cortical spot when five of the animals in the group were stimulated at that point, the occurrence rate would be 0.6. To visualize the organization of the cortical spots with higher occurrence rates of the movements, a 4-grade white-black code was applied to the map.

#### Golgi Staining

We used an FD Rapid GolgiStain Kit (COSMO BIO, Japan) for Golgi staining according to the standard protocol (43). The rats were deeply anesthetized with sodium pentobarbital, and cervical dislocation was performed. Coronal brain sections of 2 mm thickness (1.0–3.0 mm posterior to the bregma, which was the same area used for ICMS) were obtained using a brain matrix and soaked in Golgi-Cox solution for 14 days followed by 30% sucrose for 3 days. The sections were sliced into 200µm, stained and photographed with BZ-X700 (Keyence, Japan) focusing on the motor cortex (same area for ICMS), particularly layers I–III.

Based on our previous paper (44), we used Sholl analysis to evaluate dendrite expansion from a cell body: the numbers of cross sections at every 20-µm from center of the cell body were counted in 27 Golgi-positive neurons per animal (neonatal WMI model: n = 4 at 5 weeks of age and n = 5 at 10 weeks of age; sham-operated control: n = 1).

#### Immunohistochemistry for Microglia

We performed immunohistochemistry at P 17 (n = 4) and P 28 (n = 4). Under deep anesthesia with pentobarbital (>50 mg/kg), rats were perfused with 4% paraformaldehyde in 0.1 M phosphate buffered saline (PBS). The brains were obtained, postfixed with the same fixative overnight, and cryoprotected with 30% sucrose. Coronal-sections (40µm) were prepared from 1– 3 mm posterior to the bregma that contains the sensorimotor cortex of the hindlimb area. After soaking with 0.25% Triton X-100 in PBS (PBS-T), blocking was performed with 10% normal goat serum (NGS) (Vector Labs, USA) for 60 min. The slices were reacted with anti-Iba1 polyclonal antibody (1:1,000; Wako, Japan) immersed in PBS-T containing 1% NGS at 4◦C overnight, followed by immersion goat anti-rabbit IgG conjugated with Alexa Fluor 594 (1:1,000; Abcam, UK). After the slices were embedded on a glass slide with mounting medium (Vector Labs, USA), slices were photographed with a fluorescence microscope (Axio Observer.Z1; Zeiss, Germany) and an AX70 microscope (Olympus, Japan).

#### Statistics

Mann–Whitney U-tests were used to compare the current threshold and the number of blocks between control and neonatal WMI rats in the ICMS experiment. Mann–Whitney Utests were also used to compare the number of cross sections between ipsilateral side and contralateral side in Scholl analysis. All data are shown as mean ± standard error of the mean (SEM).

#### RESULTS

#### Motor Map Change of the Neonatal WMI Model in Adulthood at 10 Weeks of Age

To examine neuronal responsiveness in the sensorimotor cortex, we performed ICMS focusing on the right H-I side compared with the right control side at 10 weeks of age (**Figures 1**, **2**). We

stimulated the hindlimb area of the sensorimotor cortex because gross hindlimb function exhibited more disruption than forelimb function in our previous study (37, 38).

Each cortical map in the control group (n = 8) and neonatal WMI group (n = 8) is shown in **Figure 1.** To clarify the map expansion of each group, we re-organized the map data into heat maps of the hip and the trunk (**Figure 2A**). The heat maps clearly revealed a decrease in the size of the hip area and enlargement of the trunk area in the WMI group. The number of blocks in the hip area and the trunk area confirmed a significant decrease in the hip area (control: 12.8 ± 1.3, n = 8; WMI: 7.5 ± 1.0, n = 8; p < 0.01) and a significant increase in the trunk area (control: 2.5 ± 0.5, n = 8; WMI: 5.9 ± 0.8, n = 8; p < 0.01) (**Figure 2B**). In contrast, the current threshold for evoking the movements remained unchanged in both the hip joint (control: 78.2 ± 4.3 µA; WMI: 87.6 ± 6.2 µA) and trunk muscle (control: 112 ± 8.57 µA; WMI: 114 ± 7.26 µA) (**Figure 2C**).

### Change of Dendrite Expansion in the Neonatal WMI Model in Adulthood

To investigate whether neuronal morphology changed at 10 weeks of age, when cortical map reorganization was altered, we performed Golgi staining focusing on the same sensorimotor cortex in neonatal WMI rats (n = 5) (**Figure 3**). To evaluate developmental changes in morphology, we used Sholl analysis and counted the number of cross sections every 20-µm from the center of the cell body (**Figure 3B**). Neurons in the right H-I hemisphere (ipsilateral side) appeared to exhibit denser expansion of dendrites in layers I–III (**Figure 3A**). A similar pattern of dendrite expansion was observed in layer V of the neonatal WMI model (data not shown), although accurate analysis was difficult due to the higher neuron density. Sholl analysis revealed that neurons on the H-I side had significantly more dendritic branches (**Figure 3B**; **Supplemental Table 1**). In contrast, sham-operated animals exhibited no differences between hemispheres (**Figure 3C**), showing an equivalent number to the contralateral side of the neonatal WMI model (data for the contralateral side in **Figure 3B**).

#### Electrophysiological and Morphological Changes in Earlier Developmental Stages at 5 Weeks of Age

To determine whether the electrophysiological and morphological changes described above were exhibited at earlier developmental stages, we also performed ICMS and Golgi staining at 5 weeks of age.

FIGURE 2 | Electrophysiological analysis of sensorimotor cortex by ICMS at 10 weeks of age. (A) A summary of the motor maps in Figure 1 is shown as a heat map that represents the occurrence rate of each stimulated spot: the number of animals in which the movement was elicited was taken over the number of times that spot was stimulated in each group (*n* = 8 for each group). Each number in the grid shows the rate of responsive individuals in each portion, displayed as monochrome heat maps. Note that the rate for the hip area decreased in the WMI model, whereas the rate for the trunk area increased. (B) The number of grids associated with the hip area was significantly reduced, while that of the trunk area significantly increased in neonatal WMI model animals. (C) The minimal current intensity (threshold) to evoke muscle twitch was unaltered between the groups with regard to both the hip and trunk area. \*\**p* < 0.01 by Mann–Whitney *U*-tests. Data are presented as mean ± SEM.

At 5 weeks of age, the map size appeared to be smaller than that at 10 weeks old (**Supplemental Figure 1**). The heat map indicated that the size of the hip area of the neonatal WMI group (n = 11) was unchanged compared with the control animals (n = 9) (**Figure 4A**). The number of blocks in the hip area in each animal was also unaltered between groups (control: 8.1 ± 1.4; WMI: 7.5 ± 0.6) (**Figure 4B**). In contrast, the threshold of the hip area was significantly higher in the neonatal WMI group (control, 107 ± 6.44 µA; WMI, 127 ± 5.50 µA; p < 0.05) (**Figure 4C**, **Supplemental Figure 1**). Sholl analysis of layers I–III in the motor area (n = 4) indicated that neurons in the right hemisphere had significantly more dendritic branches (**Figure 4D**, **Supplemental Table 1**).

It has been previously reported that microglia have a phagocytotic effect on the elimination of synapses during development (pruning) (45–48). To test the possibility that microglial activation was related to changes in cortical map reorganization (ICMS), Iba1 immunohistochemistry was performed at P 17 and P 28 (before 5 weeks old) because morphological changes of the dendrites (Golgi staining) were already observed at 5 weeks of age.

Iba-1 positive cells were detected in the parietal cortex of the neonatal WMI brain (**Supplemental Figure 2A**). There were many Iba-1 positive microglia with amoeboid bodies and thick processes on the H-I side at P 17 (n = 4) (**Supplemental Figure 2B**). However, Iba-1 positive cells were not detected in the neonatal WMI model at P 28 (n = 4) (**Supplemental Figure 2C**).

### DISCUSSION

To elucidate how motor dysfunction in neonatal WMI is induced without the loss of cortical neurons, we investigated electrical responsiveness and dendrite morphology in the sensorimotor cortex in adulthood (10 weeks of age), as well as earlier developmental stages (5 weeks of age) in WMI model. The results revealed that the motor map in the sensorimotor cortex was altered at 10 weeks of age but not 5 weeks of age. However, the threshold of the hip area was higher on the H-I side at 5 weeks of age, while it was similar at 10 weeks of age. In addition, an increased number of dendritic branches in layers I–III were shown in the H-I cortex as early as 5 weeks old, and this increase was maintained until 10 weeks old. Interestingly, activated microglia were detected on the H-I side at P 17, but not at P 28.

### Alteration of Motor Maps in the Neonatal WMI Model

It is known that the cortical motor map changes after brain insults such as intracerebral hemorrhage and spinal cord injury, but can be recuperated by rehabilitation (42) and motor training (40). Environmental enrichment is one method for improving behavioral performance, and has been found to promote cortical development (49). Thus, the motor map is closely related to motor function. These previous findings suggest that the alteration of the motor map at 10 weeks of age found in the present study (hip area reduction and trunk area expansion) may

Dendrite projections of 27 Golgi-positive neurons in the sham-operated right sensorimotor cortex were similar to those in the left cortex, which was equivalent to the left (contralateral control) side of the WMI group in B. \*\**p* < 0.01 by Mann-Whitney *U*-tests. Scale bar, 100µm in lower figures. Data are presented as mean ± SEM.

be related to the imbalanced motor coordination in the neonatal WMI model (38). However, alterations of the motor map cannot completely explain the motor imbalance in this model, because it was unchanged at 5 weeks of age in the present study.

A previous study reported that the motor map emerges after 4 weeks old and eventually expands, while gradual decreases of the current threshold are promoted by enriched environments in development (49). These findings indicate that the period after weaning is a critical window in the development of normal function acquisition. Therefore, the higher threshold we detected in the hip area of the WMI group at 5 weeks of age might indicate a developmental delay of functional acquisition, as the threshold became almost equivalent at 10 weeks of age.

### Alteration of Dendrite Morphology in the Surface Area of the Cortex

Although both excitatory and inhibitory neurons were maintained in the sensorimotor cortex of this model (37), the surface area was affected, exhibiting oligodendrocyte loss and hypomyelination (37, 38). However, it remains unclear how motor dysfunction is induced in this model. Changes in neuronal circuit formation and connectivity in the cerebral cortex could be induced, as neonatal WMI is a complex amalgam of destructive developmental disturbances in premature infants (5). Morphological changes of dendrites in layers II–III were also confirmed by detailed analysis with Golgi staining as early as P 28 in the present study. As the surface area in the sensorimotor cortex receives many afferent fibers, the altered dendritic morphology of Golgi-positive neurons in layers II–III, as revealed by Scholl analysis in this study, might change local neuronal circuits and connectivity, causing the motor imbalance observed in this model.

The current study was unable to provide a clear answer regarding the question of how this morphological change in layers II–III is related to the response to ICMS in this model. GABAergic neurons in the sensorimotor cortex should be taken into consideration, as inhibitory neurons are important for normal motor map function (49), and the motor map changes immediately with injection of a GABA receptor antagonist to the motor cortex (50). Although further studies will be required to understand the contribution of GABAergic neurons, it is important to consider the migration mechanisms and temporal excitatory effects of immature GABAergic neurons by P 14–16.

The retardation of OL differentiation (37, 38) is likely to affect cortical development, as OL-expressing proteins such as myelin associated glycoprotein, Nogo-A, and oligodendrocytemyelin glycoprotein changed neurite outgrowth during development after neonatal hypoxia (51). Thus, changes of dendrite morphology in layers II–III of sensorimotor cortex

at 5 weeks of age. Sholl analysis revealed that the number of cross sections increased over 60µm from cell bodies on the right side of the cortex, even at 5 weeks of age. \**p* < 0.05 and \*\**p* < 0.01 by Mann–Whitney *U*-tests. Data are presented as mean ± SEM.

in adulthood could be altered by oligodendrocyte loss and hypomyelination (37, 38).

#### Possible Effects of Iba-1 Positive Microglia

Strong expression of ED1-positive microglial/macrophage cells was detected in the cortex of the model at P 5 in our previous report (36). In this study, the appearance of Iba-1 positive microglia was confirmed in the sensorimotor cortex until P 17, but the cells were unremarkable on both sides of the cortex at P 28.

Previous studies have reported that microglia affect synaptic pruning during development (45–48), and that synapse maintenance and the stripping function of microglia are related to dendrite expansion (52–54). Thus, it is likely that transient appearance of microglial cells until P 28 affected dendritic changes in the cortex that were detected at 5 weeks of age in this study. Further studies will be necessary to examine the morphology during earlier periods of development, and to elucidate the relationship between microglia and dendrite expansion in more detail.

#### Limitations of the Study

It should be noted that only male pups were used in this study, as there are several reports about the sex-dependent differences of the response to H-I (55–58). Several points in the present study would be strengthened by further experiments as follows. As our discussion related to possible effects of Iba-1 positive microglia is highly speculative, co-staining of Iba-1 with different antigens or synaptic markers, detection of genes related to proinflammatory cytokines, and unbiased quantification of Iba-1 positive microglial number (59, 60) should be performed in future. In addition, small sample size of Golgi-staining of shamoperated animal and lack of true control (sham-operation with normoxia) will be solved in near future. As for the limitations in ICMS experiment, the exact motor map boundaries of the hip and trunk are not outlined in this study due to our large regularly space over a 0.5-mm grid and small sample size of our study. It might be possible to get the exact motor map boundaries using more precise ICMS experiment (35).

## CONCLUSION

The current results demonstrated that the neonatal WMI model exhibited alterations of neuronal morphology and cortical responsiveness. Our data also indicated that alterations of dendrite morphology and electrical responses in the cortex occurred before the rearrangement of the motor map in the neonatal WMI model.

## AUTHOR CONTRIBUTIONS

ICMS was performed my YU with assistance of AI and Golgi staining was performed by YB with hard assessment by YU and SM. The hypoxia-ischchemia model was made by SM with the help by SO, followed Iba1 staining by TS. C-GJ and TS are involved in helpful discussion and joined in writing paper. The draft of this paper was written by YU. HH organizes this study and wrote this paper as correspondence. This study is mainly supported by the grants to SM, YU, and HH with partial support of the grants to C-GJ and TS.

### FUNDING

This study was supported by Grants-in-Aid for Scientific Research in priority area (C) (# 26430020 to HH, #16K10100 to SM), and young area (B) (#26860851 to SM, #17K16303 to YU), and a Grant-in-Aid for Research Activity Start-up (#15H06538 to YU) from the Japan Society for the Promotion of Science (JSPS). This study was also supported by a Grant-in-Aid for Scientific Research on Innovative Areas (Adaptive Circuit Shift) to H.H.

#### ACKNOWLEDGMENTS

We thank Edanz Group Ltd. for editing a draft of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00443/full#supplementary-material

#### REFERENCES


Supplementary Figure 1 | Cortical motor map by ICMS at 5 weeks of age. Cortical maps at 5 weeks of age (*n* = 9 for control, *n* = 11 for neonatal WMI) are shown as a 5-colored code in the grid: twitch of the hip joint (red), knee joint (yellow), foot joint (light blue), and trunk (green), and non-responsive square (gray). The current threshold at each point is presented in each grid.

Supplementary Figure 2 | Iba1 immunostaining. (A) Iba1 immunoreactivity was upregulated on the right H-I side of the cortex at P 17 (*n* = 4). (B) Appearance of Iba-1 positive microglia was increased in the right hemisphere (R) compared with the left side (L). (C) At P 28 (*n* = 4), the upregulation of immunoreactivity was not seen in either hemisphere. Error bars show 1 mm (A), and 40µm (B).

Supplementary Table 1 | Sholl analysis.

development of oligodendrocyte progenitors in a model of periventricular leukomalacia. Dev Neurosci. (2013) 35:182–96. doi: 10.1159/000346682


injury in the very immature rat brain. Pediatr Res. (2003) 54:263–9. doi: 10.1203/01.PDR.0000072517.01207.87


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

# Umbilical Cord-Derived Mesenchymal Stromal Cells Contribute to Neuroprotection in Neonatal Cortical Neurons Damaged by Oxygen-Glucose Deprivation

Takeo Mukai 1,2, Arinobu Tojo1,2 and Tokiko Nagamura-Inoue<sup>2</sup> \*

<sup>1</sup> Division of Molecular of Therapy, Center for Advanced Medical Research, Institute of Medical Science, University of Tokyo, Tokyo, Japan, <sup>2</sup> Department of Cell Processing and Transfusion, Institute of Medical Science, University of Tokyo, Tokyo, Japan

#### Edited by:

Olivier Baud, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Cindy Van Velthoven, School of Medicine, Stanford University, United States Benjamin Guillet, Aix-Marseille Université, France

> \*Correspondence: Tokiko Nagamura-Inoue tokikoni@ims.u-tokyo.ac.jp

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 10 November 2017 Accepted: 31 May 2018 Published: 15 June 2018

#### Citation:

Mukai T, Tojo A and Nagamura-Inoue T (2018) Umbilical Cord-Derived Mesenchymal Stromal Cells Contribute to Neuroprotection in Neonatal Cortical Neurons Damaged by Oxygen-Glucose Deprivation. Front. Neurol. 9:466. doi: 10.3389/fneur.2018.00466 Several studies have reported that human umbilical cord-derived mesenchymal stromal cells (UC-MSCs) restore neurological damage in vivo through their secretion of paracrine factors. We previously found that UC-MSCs attenuate brain injury by secreting neurotrophic factors, such as brain-derived neurotrophic factor (BDNF) and hepatocyte growth factor (HGF). However, how these factors contribute to neuroprotection remains unknown. In this study, we aimed to investigate to what extent UC-MSC-derived HGF and BDNF contribute to neuroprotection using a Transwell co-culture system of neonatal cortical neurons damaged by oxygen-glucose deprivation. The influence of HGF and BDNF were determined by investigating neurons in both the presence and absence of UC-MSCs as these cells consistently secrete both factors and can be blocked by neutralizing antibodies. In the co-culture, UC-MSCs significantly improved neuronal injury, as indicated by an increase in immature neuron number, neurite outgrowth, and cell proliferation. Co-culture of damaged neurons with UC-MSCs also exhibited a reduction in the number of neurons displaying signs of apoptosis/necrosis. The neuroprotective actions of UC-MSCs were partially reverted by neutralizing antibodies. Together, our findings reveal that UC-MSC-secreted HGF and BDNF have neuroprotective effects on damaged neurons. Further studies should address the existence of other potential neurotrophic paracrine factors.

Keywords: mesenchymal stromal cell, umbilical cord, neonatal encephalopathy, cerebral palsy, brain derived neurotrophic factor, hepatocyte growth factor

#### INTRODUCTION

Mesenchymal stromal cells (MSCs) can be isolated from several sources, including bone marrow, cord blood, adipose tissue, the placenta, and the umbilical cord (UC) (1–5). Among various sources of MSCs, we focused on the UC for the following reasons: (1) abundant sources and ease of collection, storage, and transport; (2) no invasive process of collection; (3) little ethical controversy; (4) multipotency to differentiate into various cell types; (5) low immunogenicity with significant immunosuppressive ability; and (6) migration ability toward injured sites (5). Two major mechanisms have been postulated for the observed improvements following MSC treatment; namely, antiinflammatory and neurotrophic mechanisms. UC-MSCs have been reported to exert anti-inflammatory effects via contact with activated T cells and partly through indoleamine 2, 3 dioxygenase and prostaglandin E2 (6, 7). Indeed, it is thought that UC-MSCs exert anti-inflammatory actions on brain lesions in the acute stages of injury. On the other hand, the mechanisms which underlie the neurotrophic effects of UC-MSCs have not been fully elucidated.

We previously reported that intravenously administered UC-derived MSCs (UC-MSCs) attenuate intraventricular hemorrhage-induced injuries, and brain-derived neurotrophic factor (BDNF) and hepatocyte growth factor (HGF) concentration were elevated in serum and cerebrospinal fluid in some part of UC-MSCs administered mice (8). Therefore we aim to confirm the function of BDNF and HGF in vitro in this study.

Guo et al. (9) reported the paracrine effects of UC-MSCs on nerve regeneration, observing that UC-MSCs express neurotrophic factors and that UC-MSC-conditioned medium enhances Schwann cell viability and proliferation via increases in nerve growth factor and BDNF expression . However, since HGF and BDNF are not separately inhibited, there is a possibility that each individual of HGF or BDNF may not contribute to neurotrophic effect. In addition, neuroprotective effect, such as anti-apoptosis/necrosis effect, was not examined.

In this study, we focused on whether BDNF and HGF secreted by UC-MSCs exert neuroprotective effect in addition to the neurorestorative effect in vitro, and verified our previous results.

### MATERIALS AND METHODS

#### UC-MSC Preparation

This study was carried out in accordance with the recommendations of Ethics Committee of the Institute of Medical Science, the University of Tokyo, and the NTT Medical Center Hospital and Yamaguchi Hospital, Japan with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of the Institute of Medical Science, the University of Tokyo, and the NTT Medical Center Hospital and Yamaguchi Hospital, Japan. UC-MSCs were isolated from three individual donors using previously reported methods (8, 10). Briefly, the UCs were collected after informed consent was obtained from pregnant women planning to undergo cesarean sections. Frozen-thawed UC tissues were minced into fragments and underwent the improved explant culture method (11). Tissue fragments were cultured with α-minimal essential medium (αMEM; Wako Pure Chemical Industries, Ltd., Japan) supplemented with 10% fetal bovine serum and antibiotics-antimycotics (Antibiotic-Antimycotic, 100X; Life Technologies, USA) at 37◦C with 5% CO2. Fibroblast-like adherent cells that migrated from the UC tissue fragments were harvested using TrypLE Select (Life Technologies), and were defined as passage 1 UC-MSCs. The harvested cells underwent four passages, after which they were used for experimental analyses as well as previous report (8). UC-MSCs were preserved in cryoprotectant and thawed before use.

### Cortical Neuron Primary Cultures

All experiments were carried out in accordance with the Animal Experiment Committee of the Institute of Medical Science, the University of Tokyo. Cortical neurons from B6 Albino mice (B6N-Tyrc-Brd/BrdCrCrl, Charles River Laboratories International, Inc.) were prepared according to previous reports (12, 13). Briefly, embryonic day 16 fetuses (n = 21) were taken from euthanized, pregnant mice in sterile conditions. Fetal brains were removed and cortical tissues were dissected under a microscope. The meninges were then removed and cortical tissues were chopped into small pieces. Cells were dispersed followed by mechanical trituration using Neuron Dissociation Solutions (Wako Pure Chemical Industries, Ltd., Japan) and filtered through a 70µm pore-size cell strainer. Cells were then resuspended in neurobasal medium (GIBCO) supplemented with 2% B27 (Invitrogen) and plated onto Poly-L-Lysine Culture Dishes (BioCoatTM , Corning Inc. Japan). Cells were cultured in a humidified incubator at 37◦C with 5% CO2, and half of the medium was replaced with fresh solution every 3 days. To reduce contamination by glial cells, 10µM cytosine arabinofuranoside (Sigma-Aldrich) was added for 24 h on the 4th day of culture. We cultured cortical neurons for 7days, and OGD procedure was performed.

### Oxygen-Glucose Deprived Neurons Co-cultured With UC-MSCs

A model of neonatal cortical neurons injured by OGD was established as previously described (13, 14). For deprivation of glucose, primary cortical neurons were washed twice with phosphate-buffered saline (PBS) and cultured in glucose-free Dulbecco modified eagle medium (GIBCO). Cells were incubated in an anaerobic chamber (95% N2, 5% CO2) (ASTEC Co, Ltd., Japan) at 37◦C. The OGD condition was maintained for 4 h, after which cells were re-oxygenated in the original medium and placed in a normoxic chamber (37◦C, 5% CO2). After injury by OGD was completed, co-culture with UC-MSCs was started immediately. Neurons were co-cultured with UC-MSCs according to previously reported methods (10). Briefly, using a 24-well trans-well chamber (Corning, USA) equipped with an 8-µm filter membrane, cortical neurons were cultured in the bottom chamber, while UC-MSCs were plated in the upper chamber at 5 × 10<sup>4</sup> cells/well overnight for 24 h at 37◦C with 5% CO2. In experiments aimed to measure HGF and BDNF concentrations in the culture supernatant, αMEM without fetal bovine serum was used during co-culture.

**Abbreviations:** UC, umbilical cord; MSCs, mesenchymal stromal cells; OGD, oxygen - glucose deprivation; MAP-2, microtubule-associated protein 2; GAP-43, growth associated protein 43; BDNF, brain-derived neurotrophic factor; HGF, hepatocyte growth factor; DAPI, 4',6-diamidino-2-phenylindole; NAb, neutralizing antibody; FLICA, fluorescent-labeled inhibitor of caspases.

#### Multiplex Flow Cytometric Beads Assay

For HGF and BDNF inhibition, the following neutralizing antibodies (NAbs) were added to UC-MSC culture media in order to deplete HGF and BDNF, as previously reported (12, 15, 16): anti-HGF neutralizing antibody (ab10678, Abcam) and recombinant human TrkB Fc chimera protein (#688-TK; R&D Systems). Briefly, cells were treated with 0.5µg/mL of anti-HGF antibody or 1–2µg/mL recombinant human TrkB Fc chimera protein at plating; media was not changed during the course of the experiment. As a negative control, appropriate recombinant human IgG1 Fc (R&D Systems) was used. In order to measure the neutralized concentration of human-HGF and human-BDNF, the supernatant of UC-MSCs was analyzed using a multiplex flow cytometry beads assay (#111116 and #111362, HQ-Plex Kit; Bay Bioscience, Japan). Also in order to measure mouse-HGF and mouse-BDNF, the supernatant of cortical neurons was analyzed using a multiplex flow cytometry beads assay (#211230 and #211226, HQ-Plex Kit; Bay Bioscience, Japan). All samples were analyzed in triplicate according to the manufacturers' instructions. Bead fluorescence readings were done by a flow cytometry apparatus (BDTM FACS Canto II), and data were analyzed using FCAP Array ver.3.0.1 Software (BD Biosciences, CA, USA). Results are expressed in pg/ml.

#### Immunocytochemical Assessment of Cortical Neurons

The expression of neural protein markers and a mitotic marker was analyzed by immunocytochemistry. Briefly, neurons were fixed with 4% paraformaldehyde for 30 min at room temperature, blocked in 5% skim milk and 0.3% Triton-X 100 (Sigma-Aldrich, USA), and incubated overnight with primary antibodies at 4◦C. The primary antibodies included mouse antihuman microtubule-associated protein 2 (MAP-2; 1:200 dilution, Abcam), rabbit anti-growth associated protein-43 (GAP-43; 1:500 dilution, Abcam), rabbit anti-BrdU (1:200 dilution, Abcam), and mouse anti-phospho histone H3 (S10) (1:500 dilution, Abcam). Secondary antibodies used were donkey antimouse IgG heavy and light chain-specific (H&L) (Alexa Fluor <sup>R</sup> 488) (1:1,000; Abcam), and donkey anti-rabbit IgG H&L (Alexa Fluor <sup>R</sup> 594) (1:1,000; Abcam). Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich, Japan), and images were acquired using a fluorescence microscope (Nikon Eclipse Ti, Nikon Instruments Inc., Japan) and NIS-Elements microscope imaging software version 4.10 according to previously described imaging methods (10). For BrdU labeling, the cells were incubated in the 10µM BrdU (ab142567, Abcam) labeling solution for 24 h at 37◦C in a CO2 incubator before staining. To objectively enumerate phospho Histone H3- and BrdU-positive cells, Image J software version 1.49 was used. Positive cells were counted in five randomly-selected fields at a magnification of 200x using a microscope (Nikon Eclipse Ti, Nikon Instruments Inc., Japan), and the proportion of positive cells was calculated as the number of histone H3-positive cells/total number of cells ×100%. The length of neurites in each neuron was measured by tracing using Image J (17–20). Neurites were identified by immunofluorescence with MAP-2 and counts were made in at least three randomly selected microscopic fields (containing 50 cells).

#### Western Blotting

Proteins were extracted from the cells according to the manufacturer instructions as previously described (8, 10). Protein concentrations of the samples were measured using the RC DC Protein Assay kit (Bio-Rad) and equal amounts of the protein and sample loading buffer were boiled for 5 min and separated by sodium dodecylsulfate polyacrylamide gel electrophoresis, followed by transfer onto PVDF membranes (Immobilon-P Membrane, PVDF, Millipore). The membranes were blocked by 5% skim milk in Tris-NaCl-Tween buffer and incubated overnight at 4◦C with the primary antibodies, GAP-43 (mentioned above) and anti-beta-tubulin (Wako Pure Chemical Industries, Ltd., Japan), at the recommended dilutions. This was followed by incubation with secondary antibodies (horseradish peroxidase-conjugated anti-rabbit IgG or horseradish peroxidase-conjugated anti-mouse IgG) for 1 h at room temperature. PVDF membranes were imaged using an enhanced chemiluminescence system with Pierce ECL Western Blotting Substrate (Thermo Scientific).

### Cortical Neuron Proliferation Assay

To evaluate whether co-culture with UC-MSC attenuated the death of cortical neurons after OGD, cell proliferation was determined using the Cell Proliferation ELISA kit, BrdU (Roche, #11669915001) according to the manufacturer's instructions. Briefly, neuronal BrdU incorporation was quantitatively evaluated using OD450 reader measurements (Bio-Rad iMark Microplate Absorbance Reader Version 1.02.01).

### Fluorochrome-Labeled Inhibitor of Caspases (FLICA)

To evaluate caspase activity in cortical neurons after OGD, cells were incubated in fluorochrome-labeled inhibitor of caspases (FLICA) solution (Immunochemistry Technologies, FAM-FLICA <sup>R</sup> Poly Caspase Assay Kit) at 37◦C for 1 h. After three washes, cells were labeled with propidium iodide (PI) and immediately imaged using a fluorescence microscope. Cell nuclei were counter-stained with Hoechst 33342. PI labeling was used with FLICA to identify four populations of cells: living (FLICA–, PI–); early apoptotic (FLICA+, PI– ); late apoptotic (FLICA+, PI+); and necrotic (FLICA–, PI+). Unstained live cells were labeled with only Hoechst 33342, and necrotic cell membranes appeared compromised and stained with PI in red. Cells in the late apoptotic phase were dually stained with FAM-FLICA (green) and PI, and cells in the early apoptotic stage were stained only with FAM-FLICA. Apoptotic/necrotic cells were counted using a microscope (Nikon Eclipse Ti, Nikon Instruments Inc., Japan), and the proportion of apoptotic/necrotic cells /total number of cells ×100% was calculated (containing total number of cells ranging from 321 to 556 cells).

### Statistical Analysis

Values are expressed as mean ± standard deviation (SD) from three independent experiments. Differences between groups were analyzed with JMP 10.0.2 software (SAS Institute, USA). Groups were compared using one-way analyses of variance (ANOVAs), followed by Turkey's tests. P-values of 0.05 or less were regarded as statistically significant.

### RESULTS

### Constitutive Secretion of HGF and BDNF From UC-MSCs Co-cultured With Cortical Neurons After OGD

Primary cultures of neonatal cortical neurons exhibited typical morphology, showing a neurite outgrowth-forming network (**Figure 1A**) that could be visualized by positive MAP-2 labeling in green, which represented mature neurons. Neurites and neuronal clusters disappeared after OGD (**Figure 1B**), but co-culture with UC-MSCs restored mature neurons, long neurites, and cluster formations (**Figure 1C**). Using a multiplex flow cytometry bead assay to analyze HGF and BDNF concentrations in medium containing UC-MSCs confirmed that both factors were constitutively secreted from UC-MSCs, and that their concentrations varied by UC-MSC lot (n = 3, **Figure 1D**). Importantly, HGF and BDNF concentrations could be appropriately reduced by the addition of NAbs (**Figure 1E**). Using these methods, the following experiments were performed: (1) control, (2) neurons injured by OGD (OGD), (3) neurons injured by OGD co-cultured with UC-MSCs (OGD+MSC), (4) OGD + MSC with anti-HGF NAb (OGD+MSC+ HGF NAb, (5) OGD + MSC with anti-BDNF NAb (OGD+MSC+BDNF NAb), and (6) OGD + MSC with both anti-HGF NAb and anti-BDNF NAb (OGD+MSC+ HGF NAb + BDNF NAb).

#### UC-MSCs Exert Neuroprotective Effect on Cortical Neurons After OGD Injury

Under control conditions, both immature (GAP-43-positive; red) and mature (MAP-2-positive; green) neurons could be observed (**Figure 2A**). After OGD injury, neurites appeared diminished in length, and the number of cells positive for GAP-43 was reduced (**Figure 2B**). Co-culture with UC-MSCs maintained the mature neurons with long neurites to some extent, and increased the number of GAP-43-positive immature

FIGURE 1 | Constitutive secretion of HGF and BDNF from UC-MSCs co-cultured with cortical neurons after OGD. UC-MSCs co-cultured with cortical neurons post-OGD and stained with MAP-2. Mature neurons are stained in green and nuclei are counterstained with DAPI in blue. (A) Control, (B) OGD, (C) OGD + MSC (Scale bar = 100µm). (D) UC-MSC-secreted HGF and BDNF. (E) HGF NAb and recombinant Human TrkB Fc chimera were used for the inhibition of HGF and BDNF, respectively. Recombinant human IgG1 Fc was used as a negative control. NAb, neutralizing antibody.

(H) Quantitative analysis of neurite outgrowth. The data shown are representative of three independent experiments. \*\*p < 0.01 compared to the control group, ##p < 0.01, #p < 0.05 compared to the OGD group, and † p < 0.05 compared to the MSC group. NAb, neutralizing antibody.

neurons (**Figure 2C**). Addition of NAbs reduced the UC-MSCmediated improvement (**Figures 2D–F**). Quantitative analysis of GAP-43 expression revealed a significant decrease in the OGD compared to the control group, and that co-culture with UC-MSCs significantly improved GAP-43 expression relative to levels observed in the OGD group. On the other hand, addition of NAbs tended to decrease the expression of GAP-43 (**Figure 2G**). The length of neurites identified by immunofluorescence with MAP-2 was significantly shortened in the OGD relative to the control group, and co-culture with UC-MSCs significantly improved neurite outgrowth length. The addition of anti-HGF + BDNF NAbs significantly reduced neurite length after OGD (**Figure 2H**). Interestingly the effects of UC-MSCs on GAP-43 expression and neurite length were reverted partially by the addition of HGF and BDNF NAbs, and there was no synergistic effect by the addition of them.

To investigate the effect of HGF and BDNF on neurorestoration, cortical neurons were stained with the mitotic marker, anti-phospho histone H3 and counterstained with DAPI (**Figure 3A**). In the OGD group, the number of phospho histone H3 - positive cells decreased, whereas this

reduction was restored in neurons co-cultured with UC-MSCs (**Figures 3B,C**). Addition of NAbs, however, attenuated this recovery (**Figures 3D–F**). Quantitative analysis revealed a significantly higher number of proliferating phospho histone H3 -positive cells were seen in neurons co-cultured with UC-MSCs compared to neurons in the OGD group. whereas addition of NAbs to the OGD+MSC group decreased the amount of mitotic cells (**Figure 3G**). On the other hand, BrdU incorporation into neurons was reduced significantly in the OGD group. Co-culture with UC-MSCs significantly increased BrdU incorporation post-OGD (**Figure 3H**); however, the effect was not attenuated by the addition of NAbs. We also performed double staining of phospho histone H3 and BrdU (**Supplementary Figure 1**). Quantitative analysis revealed a tendency that higher number of proliferating phospho histone H3-positive cells and BrdU-positive cells were seen in neurons co-cultured with UC-MSCs compared to neurons in the OGD group, whereas the effect of NAbs to the OGD+MSC group wasn't observed.

Next, we examined the neuroprotective effect of HGF and BDNF on neuronal apoptosis/necrosis after OGD using FLICA labeling. We identified four populations of cells: living (FLICA–, PI–); early apoptotic (FLICA+, PI–); late apoptotic (FLICA+, PI+); and necrotic (FLICA–, PI+) (**Figures 4A–G**). Quantitative analysis of the ratio of apoptotic and necrotic cells to total cells revealed significantly more apoptotic/necrotic cells in the OGD group compared to the control group. This analysis also demonstrated that co-culture with UC-MSCs reduced the number of cortical neurons displaying signs of apoptosis and necrosis post-OGD. This improvement was attenuated by the addition of anti-BDNF and anti-HGF NAbs (**Figure 4H**).

### DISCUSSION

Recently, UC-MSCs have attracted attention for their potential in treating neurological disorders, as several studies using neurological disease models have reported improvements after UC-MSC transplantation, and clinical studies using UC-MSCs to treat traumatic brain injury and cerebral palsy have already been implemented (21–26). The neurotrophic effects of UC-MSCs can be characterized by two mechanisms of action; (1) neurogenic differentiation and cell replacement, and (2) secretion of neurotrophic factors. UC-MSCs can differentiate into neural cells expressing high levels of neural markers (10, 27, 28). However, intravenously administered UC-MSCs in hosts that are not immunocompromised are likely ultimately eliminated. Indeed, our previous study revealed that UC-MSCs injected in neonatal mice with intraventricular hemorrhage are eliminated after 3 weeks. Interestingly, human BDNF and HGF was detected in the serum and cerebrospinal fluid of these mice (8).

Consistent with a previous report (9), we confirmed constitutive secretion of HGF and BDNF from UC-MSCs, which were inhibited by the HGF NAb and recombinant human TrkB Fc chimeras, as described in previous reports (12, 15). However, the concentrations of secreted HGF and BDNF demonstrated variability that was dependent on UC-MSC lot. Lot-to-lot variation in these secreted neurotrophic factors is an important issue considering their potential for clinical application.

In this study, we successfully generated an in vitro OGD model of primary cortical -neurons, as indicated by shortened neurites, a reduction in the number of network-forming neurites, fewer developing neurons, decreased cell proliferation, and increased apoptosis/necrosis. Co-culture with UC-MSCs was sufficient to maintain MAP2-positive mature neurons showing extended neurites and cluster formations, whereas HGF and BDNF NAbs significantly attenuated the restorative effect of UC-MSCs on neurite elongation.

Firstly we investigated neurorestorative effect of UC-MSCs using phospho histone H3 and BrdU experiment. Higher number of phospho histone H3 and BrdU-positive mitotic cells were observed in UC-MSC co-cultured cells compared to cells in the OGD group, whereas the reverse effects of HGF and BDNF NAbs on cell mitosis were not significant in histone H3 and BrdU experiments. Next we confirmed neuroprotective effect of UC-MSCs on cortical neurons after OGD injury by apoptosis/necrosis assay. The results demonstrated that co-culture with UC-MSCs reduced the number of cortical neurons displaying signs of apoptosis and necrosis post-OGD and this improvement was attenuated by the addition of anti-BDNF and anti-HGF NAbs.

These results suggest that HGF and BDNF secreted from UC-MSCs may support neuroprotection through anti-apoptotic effect rather than neurorestoration, and that there is the possibility of other UC-MSC-secreted trophic factors.

HGF—a multi-functional growth factor originally reported as a potent mitogen for mature parenchymal hepatocytes in primary culture—plays an important role in tissue regeneration in the nervous system. It has been reported that HGF binds a tyrosine kinase receptor encoded by the human protooncogene, c-Met, and that HGF and c-Met are expressed in both the adult and fetal central nervous system (29, 30). Following activation of the tyrosine kinase receptor, the induction of several downstream pathways, such as the phosphatidylinositol 3-kinase/Akt, MAP-kinase, and signal transducers and activators of transcription 3 pathways, lead to neurorestorative, anti-apoptotic and neurogenic effects (29– 31). Indeed, Liu et al. (32) reported the neuroprotective effects of UC-MSCs infected with adenovirus-expressing HGF in a model of Parkinson's disease, suggesting that HGF acted via the promotion of damaged cell regeneration. These previous reports, together with findings of the current study, support the important role of HGF in neurogenesis and maintaining cell viability.

Similar to HGF, BDNF (the second neurotrophic factor to be characterized, after NGF and before neurotrophin-3) is also expressed in the adult and developing brain, and plays a key role in the proliferation, survival, and differentiation of neurons (33–35). BDNF binds to the tropomyosin-related kinase family of receptor tyrosine kinases, which activate the phosphatidylinositol 3-kinase/Akt and MAP-kinase pathways. BDNF also binds to the p75 neurotrophin receptor, which activates nuclear factor-kB—a protein complex important in inducing the activation of pro-survival and pro-differentiation genes (36). Consistent with the supportive role of BDNF described above, we found that BDNF inhibition attenuated the extent of neurorestoration observed in cortical neurons co-cultured with UC-MSCs and increased the amount of apoptotic/necrotic cells.

To exclude the possibility of mouse-BDNF and mouse-HGF secreted from damaged cortical neurons in this study, we measured the concentration of mouse-BDNF and mouse-HGF in the supernatant of the cortical neurons with or without OGD. The result showed that mouse-BDNF and mouse-HGF were very low below detection limit (data not shown), but they should be taken into consideration in the present study. In addition, even if it seems more likely that UC-MSCs secrete those neurotrophic factors, we cannot exclude the possibility that UC-MSCs induce selfsecretion of BDNF and HGF by injured neurons. These endogenous neurotrophic factors must be considered for future study.

FIGURE 4 | UC-MSCs restore apoptotic and necrotic neurons after OGD. (A) FLICA labeling showing the proportion of apoptotic and necrotic neurons after OGD. Four populations of cells can be identified: living (FLICA–, PI–); early apoptotic (FLICA+, PI–); late apoptotic (FLICA+, PI+); and necrotic (FLICA–, PI+). Cell nuclei are counterstained with Hoechst 33342. (B) Control, (C) OGD, (D) OGD + MSC, (E) OGD + MSC + HGF NAb, (F) OGD + MSC + BDNF NAb, and (G) OGD + MSC + HGF NAb + BDNF NAb (Scale bar = 100µm). (H) Ratio of apoptotic and necrotic cells to the total number of cells. \*p < 0.05 compared to the control group. NAb, neutralizing antibody.

In conclusion, although the presence of other UC-MSCsecreted factors likely exist, the current study showed that UC-MSCs exert their neuroprotective effects partially through secretion of BDNF and HGF by inhibiting the apoptosis/necrosis of injured neurons. Considering that UC-MSCs have been administered to treat several neurological disorders, including cerebral palsy, traumatic brain injury, and hereditary spinocerebellar ataxia (25, 26, 37), the vast potential that UC-MSCs have in the clinic encourage us to facilitate allogeneic third-party UC-MSC therapies for brain injuries.

# AUTHOR CONTRIBUTIONS

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

### FUNDING

This study was supported by Grants-in-Aids for Scientific Research from the Japan Agency for Medical Research and Development /Ministry of Health, Labor and Welfare (16Abk0104010h0015) and by the Ministry of Education, Culture, Sports, Science and Technology (JSPS KAKENHI Grant Number JP17J02535 and JSPS KAKENHI Grant Number JP26293251).

### ACKNOWLEDGMENTS

I would like to express my gratitude to Hajime Tsunoda (Department of Obstetrics, NTT Medical Center Tokyo Hospital), Satoru Yamaguchi (Department of Obstetrics, Yamaguchi Hospital), and the staff members at the NTT Medical Center and Yamaguchi Hospital for their assistance with the collection of UCs.

### REFERENCES


# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00466/full#supplementary-material

Supplementary Figure 1 | UC-MSCs support neuronal mitosis after OGD. Immunostaining showing mitotically proliferating neurons stained with phospho histone H3 (green), BrdU (red) and counterstained with DAPI (blue). (A) Control, (B) OGD, (C) OGD + MSC, (D) OGD + MSC + HGF NAb, (E) OGD + MSC + BDNF NAb, and (F) OGD + MSC + HGF NAb + BDNF NAb (Scale bar = 100µm). (G) Ratio of the number of phospho histone H3-positive cells and BrdU-positive cells to the total number of cells. ∗∗<sup>p</sup> <sup>&</sup>lt; 0.01, <sup>∗</sup><sup>p</sup> <sup>&</sup>lt; 0.05 compared to the control group. NAb, neutralizing antibody.

through activation of PI3K/AKT signaling in primary cortical neurons followed oxygen-glucose deprivation injury. PLoS ONE (2013) 8:e78514. doi: 10.1371/journal.pone.0078514


cerebral palsy: pilot study on the correlation of efficacy and hereditary factors. Cytotherapy (2015) 17:224–31. doi: 10.1016/j.jcyt.2014.09.010


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

# Brain Metabolism Alterations Induced by Pregnancy Swimming Decreases Neurological Impairments Following Neonatal Hypoxia-Ischemia in Very Immature Rats

Eduardo F. Sanches <sup>1</sup> \*, Yohan Van de Looij 1,2, Audrey Toulotte<sup>1</sup> , Analina R. da Silva<sup>2</sup> , Jacqueline Romero<sup>2</sup> and Stephane V. Sizonenko<sup>1</sup>

<sup>1</sup> Division of Child Development and Growth, Department of Pediatrics, University of Geneva, Geneva, Switzerland, <sup>2</sup> Laboratory for Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

#### Edited by:

Brigitte Vollmer, University of Southampton, United Kingdom

#### Reviewed by:

Gugu T. J. Kali, Stellenbosch University, South Africa Tina Bregant, Children's Hospital Ljubljana, Slovenia

> \*Correspondence: Eduardo F. Sanches ef.sanches@yahoo.com

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 21 January 2018 Accepted: 01 June 2018 Published: 25 June 2018

#### Citation:

Sanches EF, Van de Looij Y, Toulotte A, da Silva AR, Romero J and Sizonenko SV (2018) Brain Metabolism Alterations Induced by Pregnancy Swimming Decreases Neurological Impairments Following Neonatal Hypoxia-Ischemia in Very Immature Rats. Front. Neurol. 9:480. doi: 10.3389/fneur.2018.00480 Introduction: Prematurity, through brain injury and altered development is a major cause of neurological impairments and can result in motor, cognitive and behavioral deficits later in life. Presently, there are no well-established effective therapies for preterm brain injury and the search for new strategies is needed. Intra-uterine environment plays a decisive role in brain maturation and interventions using the gestational window have been shown to influence long-term health in the offspring. In this study, we investigated whether pregnancy swimming can prevent the neurochemical metabolic alterations and damage that result from postnatal hypoxic-ischemic brain injury (HI) in very immature rats.

Methods: Female pregnant Wistar rats were divided into swimming (SW) or sedentary (SE) groups. Following a period of adaptation before mating, swimming was performed during the entire gestation. At postnatal day (PND3), rat pups from SW and SE dams had right common carotid artery occluded, followed by systemic hypoxia. At PND4 (24 h after HI), the early neurochemical profile was measured by <sup>1</sup>H-magnetic resonance spectroscopy. Astrogliosis, apoptosis and neurotrophins protein expression were assessed in the cortex and hippocampus. From PND45, behavioral testing was performed. Diffusion tensor imaging and neurite orientation dispersion and density imaging were used to evaluate brain microstructure and the levels of proteins were quantified.

Results: Pregnancy swimming was able to prevent early metabolic changes induced by HI preserving the energetic balance, decreasing apoptotic cell death and astrogliosis as well as maintaining the levels of neurotrophins. At adult age, swimming preserved brain microstructure and improved the performance in the behavioral tests.

Conclusion: Our study points out that swimming during gestation in rats could prevent prematurity related brain damage in progeny with high translational potential and possibly interesting cost-benefits.

#### HIGHLIGHTS


Keywords: prematurity, hypoxia-ischemia, pregnancy swimming, neuroprotection, magnetic resonance imaging, brain

### INTRODUCTION

#### Prematurity and Neonatal Hypoxia-Ischemia

Preterm birth represents around 11% of all live births (∼15 million children) (1) and is one of the most important causes of perinatal mortality and morbidity. Despite the progress of neonatal medicine improving their survival rate, the incidence of premature babies has increased in most of the countries (2–4). Prematurity is linked to subcortical white and gray matter lesions and to impaired structural connectivity (5, 6), leading to lifelong neurodevelopmental disturbances (7–10).

Neonatal hypoxic-ischemic (HI) brain injury is a major public health problem leading to complications during and after birth (11), and is part of the etiology of cerebral palsy, neurodevelopmental deficits, learning disabilities, ADHD, autism and other diseases (4, 12). HI leads to a distinct neurological injury pattern depending at gestational age it occurs. In preterm infants, brain injury leads to a diffuse pattern of white matter damage with altered myelination, ventriculomegaly and reduced cortical development or to cystic periventricular leukomalacia whereas in full-term newborns, the gray matter areas are the primary regions injured (12). HI occurs due to a drop in the brain blood and/or oxygen flow (13) which compromises the oxidative metabolism, leading to a decrease in energy levels and increased glutamate release, leading to excito-oxidative injury cascade (11), metabolic failure, alterations in the neuron-glia coupling and cell death (14). Multimodal magnetic resonance techniques can be used to monitor metabolic and microstructural changes following HI (15). Localized <sup>1</sup>H-Magnetic resonance spectroscopy (MRS) has been used to follow biochemical changes in the pup rat brain following HI (16). In addition, diffusion tensor imaging (DTI) probes the brain microstructure and allows evaluation of microstructural alterations in the brain following HI (17).

The Rice-Vannucci rodent model is often used to mimic the pathological mechanisms as well as the functional consequences of hypoxia-ischemia, allowing a better comprehension of the HI pathophysiology and evaluating effects of therapeutic strategies [for a review, see (18)]. In terms of cerebral maturity, the 3-day-old rat corresponds to a preterm human baby birth at 24–28 weeks of gestation and is used to study the mechanisms of perinatal brain damage in this population defined as early preterm (19–24). HI in early preterm leads to disruption in cell development and in the cortical cytoarchitecture (23), inflammation, alterations in myelination and cognitive impairments (21, 24, 25). HI pathophysiology complexity enables multiple therapeutic targets and neuroprotective strategies can counteract one or, ideally, multiple pathways (26–28).

### Physical Exercise Benefits During Pregnancy

The importance of gestational interventions that improve maternal, perinatal, and neonatal health outcomes is recognized (10, 29). Exercise during pregnancy is considered beneficial to both mother and fetus and is recommended by the Colleges of Obstetricians and Gynecologists (30, 31). Several risk factors such as diabetes mellitus (32) and preeclampsia, commonly associated to premature delivery (33, 34) can be reduced by physical exercise (35, 36). Preclinical studies evidenced that pups born from exercised mothers had significantly higher brain, liver, heart and kidney weights compared to the controls, which suggests that regular exercise during pregnancy can improve placentary functioning and support fetal development (37, 38). Labonte-Lemoyne et al.suggested that babies born from exercised mothers were born with more mature brains (39).

Swimming during pregnancy is widely recommended for women (40) due to the low-impact effects of buoyancy (41), the excellent heat conductor capacity (42) and the beneficial effects on the cardiovascular system in the mother (40), and although some evidence demonstrates that urinary tract infections (UTI) could eventually occur, which could imply birth defects, literature shows no significant association between swimming during pregnancy and UTI (43). Besides, preclinical evidence show that pregnancy swimming can improve the intrauterine environment and improve brain maturation in the pups (38, 44, 45) by inducing hippocampal neurogenesis (46– 48), enhancing the brain antioxidant capacity (49), maintaining the ionic gradients as well as the levels of neurotrophins (38, 46, 50) and leading to an improvement in cognitive tasks (38, 48, 50). Although pregnancy swimming could have a positive impact over multiple pathways involved in HI injury, the knowledge about its potential beneficial effects are still limited. Thus, we hypothesize that pregnancy swimming can induce metabolic adaptations in the pup's brain that are sufficient to reduce a subsequent HI damage. Using a multimodal approach involving in vivo MR spectroscopy and ex vivo MR imaging techniques, biological and behavioral evaluation we assessed the potential neuroprotective effects of gestational swimming on HI brain injury in the immature rat brain.

### MATERIALS AND METHODS

#### Animals

The Geneva State Animal Ethics Committee and the Swiss Federal Veterinary Service approved this study under GE/132/15 license. Male and female Wistar rats were ordered from Charles River Laboratories (L'Arbresle, France). Animals were housed under standard laboratory conditions (12-h-light, 12-h-dark cycle and room temperature at 22 ± 1 ◦C). One week prior to mating, the females were distributed to Sedentary (SE) or Swimming (SW) group and acclimated to a black circular acrylic water tank (200 cm diameter) filled with warm water at 32 ± 1 ◦C (25 cm depth for SW group). A 100 cm diameter tank (made with the same black plastic material) and kept empty was used for the exposition of the SE group. Standard rat chow and water was provided ad libitum. Sedentary (SE) and Swimming (SW) females were mated (Gestational day 0—GD0) and kept two per cage until GD20. The timeline of the experiments is shown in **Figure 1**.

## Swimming Protocol

The swimming protocol consisted in a training period of 4 days swimming with increasing time exposures in the tank (5, 10, 15, and 20 min/day) previously to mating. Then it consisted of 20 min daily sessions from 1st to the 21st day of pregnancy (50). After each session, the animals were dried with a face towel and kept under an infrared lamp until completely dried. From GD 20 until delivery, they were housed individually in a clean standard cage. The control non-swimming animals were daily exposed to an empty circular open field (measures 100 cm diameter and 45 width) to be manipulated by the experimenters and exposed to a different environment (same material as the swimming tank). SE (n = 8) and SW (n = 12).

### Neonatal Hypoxia-Ischemia

At PND1 the newborn animals were counted and the litters were culled to between 8 and 12 pups to avoid differences in animal weights. Both sexes were used for the procedure in a rate of 50% each. At PND3 pups were submitted to mild to moderate hypoxic-ischemic injury as previously described (16, 21, 23). Briefly, under isoflurane anesthesia (4% induction and 1.5–2.0% maintenance), the right carotid artery was isolated from the vagus nerve and surrounding tissue and permanently occluded with 6.0 silk thread. The surgical access was closed with HistoacrylTM and Steri-stripTM. After a 30 min recovery period in a chamber at 37◦C with room air, the flux of room air was replaced by a 2 l/min of 6% O<sup>2</sup> at 37◦C during 30 min to induce hypoxia. Sham animals were anesthetized, had the incision without carotid occlusion or hypoxia. For all experiments, SE and SW litters were processed in parallel. In total, 4 groups were assessed: (1) Sedentary-Sham (SESH), (2) Sedentary-Hypoxic-Ischemic (SEHI), (3) Swimming-Sham (SWSH) and (4) Swimming-Hypoxic-Ischemic (SWHI).

### Magnetic Resonance

MR experiments were performed on an actively-shielded 9.4T/31 cm magnet (Agilent/Varian/Magnex) equipped with 12 cm gradient coils (400 mT/m, 120 µs) with a quadrature transceive 20 mm surface RF coil as previously described (16, 51).

#### [ 1 -H] MR Spectroscopy

For <sup>1</sup>H-MRS (24 h after injury), the rats were continuously anesthetized under a flow of 1.5–2% isoflurane in O2. Body temperature was kept at 37 ± 0.5◦C during the entire procedure. For a better characterization of the injury as well as to identify more precisely the effects of maternal swimming, animals were categorized according to the lesion severity using the presence of a hypersignal in the cortex on T2W images. At PND4, 24 h after injury, after automatic FASTMAP shimming, spectra acquisition on a voxel of interest of 1.5 × 1.5 × 2.5 mm<sup>3</sup> within the parietal cortex was performed for the 4 groups using an ultra-short echo time (TE/TR = 2.7/4,000 ms) SPECIAL spectroscopy method (52). Proton spectra were analyzed with LCModel (53) providing the neurochemical profile of the right injured hemisphere and in the right hemisphere for the controls, for SE and SW groups. The results provided the quantification of the following metabolite concentrations: aspartate (Asp), alanine (Ala), ascorbate (Asc), creatine (Cr), phosphorylcholine (PCho), phosphocreatine (PCr), γ-aminobutyric acid (GABA), glutamate (Glu), glutamine (Gln), glutathione (GSH), glycine (Gly), lactate (Lac), macromolecules (Mac), myoinositol (Ins), N-acetylaspartate (NAA), N-acetylaspartylglutamate (NAAG), phosphoethanolamine (PE) and taurine (Tau).

#### Diffusion Tensor Imaging (DTI)/Neurite Orientation Dispersion Index (NODDI)

At PND4 and PND60 (n = 4–6 animals/group per time point), rats were sacrificed and brains were paraformaldehyde-fixed for subsequent ex vivo MRI with a 2.5 mm diameter birdcage coil. A multi-b-value shell protocol was acquired using a spin-echo sequence (FOV = 21 × 16 mm<sup>2</sup> , matrix size = 128 × 92, 12 slices of 0.6 mm, 3 averages with TE/TR = 45/2,000 ms). 96 DWI were acquired, 15 b<sup>0</sup> images and 81 separated in 3 shells (noncollinear and uniformly distributed in each shell) with number of directions/b-value in s/mm<sup>2</sup> : 21/1750, 30/3,400 and 30/5,100, respectively. Acquired data were fitted using the NODDI toolbox (54). At PND4 and PND60, three different brain regions were identified: cortex (Cx), corpus callosum (CC) and external capsule (EC). DTI derived parameters [Axial diffusivity (AD), Radial diffusivity (RD), Mean diffusivity (AD) and Fractional anisotropy (AD)] as well as NODDI derived parameters (intraneurite volume faction (ficvf ), isotropic volume fraction (fiso) and orientation dispersion index (ODI) were averaged in the different regions assessed.

#### Behavioral Analysis

Given that HI lesion involve several regions, including sensorimotor cortex, and hippocampus, as from 45 days of age animals were tested in the Elevated Plus Maze (EPM), Open Field (OF), Novel Object Recognition (NOR) and Morris Water Maze (MWM). All animals performed the tasks in the above cited order. The apparatuses were thoroughly cleaned between every animal and the male rats were tested first. All behavioral procedures were performed between 9 a.m. and 4 p.m. The same investigators performed all experimental sessions in a controlled light, temperature and sound room. After each trial, the apparatuses were cleaned with a 70% ethanol solution (24).

#### Elevated Plus Maze (EPM)

The elevated plus maze, allowing to measure anxiety, is a device with two open arms (50 × 10 cm), surrounded by an edge of 0.5 cm and two closed arms (50 × 10 × 15 cm) and the central area measuring 10 cm<sup>2</sup> . The maze was elevated to a height of 70 cm. Each rat was placed at the center of the apparatus facing one enclosed arm. The test was video recorded for 5 min and using the ANY-Maze software (Stoelting Co., USA) the number of entries into open or closed arms and the total time spent in each arm was recorded. An entry was defined by placing the four paws into an arm (24).

#### Open Field (OF)

The test allows the observation of exploratory activity of animals in a novel environment. The apparatus consists of a circular wooden chamber (100 cm diameter × 30 cm high wall) with a floor divided into 21 fields. Using ANY-Maze software, the open field test was video recorded during 5 min. The latency to leave the central circle, number of crossings and rearings were considered as indicative of spontaneous motor activity.

#### Novel Object Recognition (NOR)

The novel object recognition task assesses declarative memory (55). In the first phase of the test, each animal was confronted with two different objects, placed in an open-field box (the same used for the open field test) and the time of object exploration was registered for 5 min. Following this phase, the rodent was removed from the open-field box and put in another separate box for a period of 5 min. In the second phase, each animal was exposed to two objects placed in the same open-field box: one familiar object, used in the first phase, and one novel object. The time spent exploring the novel object and the familiar object was measured. A discrimination index was calculated in the test session (second), as follows: the difference in exploration time divided by the total time spent exploring the two objects (B – A/B + A, where B is the new object and A is the familiar object) (56).

#### Morris Water Maze (MWM)

Spatial memory was tested in the Morris water maze task as previously described (24). Rats entered the pool facing the wall and from a start position designated as N, S, W or E. All rats accomplished four trials/day, on 5 consecutive days, with a 10 min inter-trial interval and every starting point was used in a different order each day. The latency to find the platform during each trial was measured as a learning index. During the five training days, the platform remained at the same location. A probe test (without the platform) was performed on the 6th day and parameters such as latency to cross the platform zone, time spent in platform quadrant, time spent in the opposite platform quadrant and total distance traveled were assessed using the ANY-Maze software.

## Protein Analysis

For the western blotting analysis, pups were sacrificed at either PND4 or PND60 and brain structures were quickly collected on ice and the right cortex and hippocampus (ipsilateral to the lesion) were dissected out and frozen in RIPA buffer (Cell Signaling, 9806S) at −20◦C. Structures were sonicated and the protein concentration was determined using a Bradford assay. Proteins (25 µg) were separated by SDS-PAGE, transferred on nitrocellulose membrane and analyzed by immunoblotting. The primary antibodies were diluted (1:1,000) in blocking solution containing 0.1% casein (Sigma-Aldrich, C8654). PND4 brains were analyzed for neurons (NeuN and DCX), astrocytes (GFAP), (GLT-1) and glutamine synthetase, oligodendrocytes progenitors (NG2), microglia/macrophages (CD11b and Iba-1), apoptosis (fractin and cleaved caspase 3), neurotrophic factors (VEGF and BDNF) and the BDNF receptor Tyrosin Kinase (Trk-B). For the PND60 assessment, the membranes were incubated with the primary antibodies: NeuN, GFAP, MBP, BDNF, VEGF, and Trk-B. After overnight incubation with the primary antibody, the following secondary antibodies (1:10,000) were applied: goat anti-mouse IgG conjugated with IRDye 680 (LI-COR, B70920-02), goat anti-rabbit IgG conjugated with IRDye 800 (LI-COR, 926- 32210) and donkey anti-guinea pig IgG conjugated with IRDye 800 (LI-COR, 926-32411). Protein bands were visualized using the Odyssey Infrared Imaging System (LI-COR). ImageStudioTM Lite (LI-COR) was used to measure the optical densities of the protein signals on scans. The relative optical density was calculated using the optical density of protein signals divided by the optical density of a loading control (actin or βIII-tubulin) and expressed as a percentage of values obtained compared to the SESH group (100%) (n = 6–8 animals/group). The list of the antibodies used is in **Table 1**.

#### Statistical Analysis

All statistical analysis was performed using SPSS 19.0 for Windows (SPSS Inc., Chicago, IL, USA). Data are presented as mean ± standard error of the mean (SEM). Non-parametric data was analyzed by Kruskall-Wallis followed by Mann-Whitney test for multiple comparisons. One-way ANOVA followed by Duncan's post-hoc was used to compare differences among the groups presenting normal distribution. The significance was accepted when p < 0.05.



#### RESULTS

Pregnant rats were weighed daily from GD1 to GD21 before swimming sessions or exposition to the open field, in the SE group. Animals in both swimming and sedentary groups gained weight during pregnancy [F(4,72) = 158.96; p < 0.05] with no significant differences between the groups [F(4,72) = 0.329; p = 0.858] (data not shown). On delivery, the swimming group had an average of 10 pups whereas the control group litters averaged 11 (no statistical difference was observed data not shown). The litters were sorted to have 50% rats of each sex distributed equally among the four groups. Pups weight was modulated by maternal swimming at PND14 [F(3, 72) = 4.74, p < 0.05], PND21 [F(3, 72) = 6.38, p < 0.05], PND45 [F(3, 72) = 3.06, p < 0.05] and PND60 [F(3, 72) = 4.04, p < 0.05], in which animals from the swimming groups (SW) had increased body weights compared to the sedentary (SE) ones. There was no effect of HI per se on this measure (data not shown).

### <sup>1</sup>HMR Spectroscopy

FASTMAP shimming (first-order and second-order correction of the magnetic field homogeneity) enabled to obtain a very good-quality of spectra in a volume of 12 µl in the parietal cortex. The average signal-to-noise ratio calculated on all acquired spectra was 13.9 ± 1.9. **Table 2** shows the concentration of the 18 metabolites assessed using the spectral analysis and absolute quantification by LCModel (53). Significant differences were observed in the concentration of Gln and the ratio Gln/Glu between SESH and SWSH groups, as well as trends to a decrease in NAAG (p = 0.07), PE (p = 0.06), Mac (p = 0.06), Glu+Gln (p = 0.07) in the SWSH group, evidencing the effect of maternal swimming on pup's brain metabolism. 24 h after HI, concentration of almost all the quantified metabolites decreased in the cortical tissue of the SEHI group compared to the SESH group including [PCho], [Cr], [PCr], [Glu], [GSH], [Ins], [NAA], [NAAG], [Tau], [Asc], [PE], [Mac], [Glu+Gln], [GPC+PCho], and [Cr+PCr]. Maternal swimming prevented the decrease of [PCho], [PCr], [Ins], [NAAG], [NAA+NAAG], and [Lac/NAA] in the SWHI group compared to the SWSH, evidencing preservation of the energetic metabolism induced by maternal swimming. Interestingly, no increase in [Lac] (marker of anaerobic metabolism) was observed in the injured groups pointing to a milder injury level compared to previous studies performed by the group (16, 57).

#### Western Blotting Expression of Cell Markers **Neurons**

No differences were observed in the expression of mature (NeuN) [F(3, 27) = 1.012, p = 0.405] nor migrating (DCX) neurons in the cortex [F(3, 27) = 1.950, p = 0.148] (**Figure 2A**) nor in the hippocampus [F(3, 27) = 2.428, p = 0.900; F(3, 27) = 1.715, p = 0.191] (**Figure 3A**).


Differences among the different groups (SESH, SEHI, SWSH and SWHI) 24 h post-HI (p < 0.05, \*Effect of injury - HI vs. SH, #SESH vs. SWSH. Ala, Alanine; +Asp, Aspartate; PCho, phosphocholine; Cr, creatine; PCr, phosphocreatine; GABA, gama aminobutyric acid; Gln, glutamine; Glu, glutamate; GSH, glutathione; Gly, glycine; Ins, Myoinositol; Lac, lactate; NAA, N-acetylaspartate; Tau, taurine; Asc, Ascorbate; NAAG, N-acetylaspartylglutamate; PE, phosphoethanolamine; Mac, macromolecules.

#### **Oligodendrocytes progenitors (NG2)**

No difference was observed in NG2 expression neither in the cortex [F(3, 27) = 0.936, P = 0.43] nor in the hippocampus [F(3, 26) = 0.55, P = 0.65; **Figures 2A**, **3A**].

#### **Microglia (Iba-1)**

No differences were observed in the protein expression of microglial cells (Iba-1) neither in the cortex [F(3, 27) = 0.419, P = 0.741] nor in the hippocampus [F(3, 27) = 0.638, P = 0.59; **Figures 2A**, **3A**].

#### **Astrocytes**

GFAP was increased in the cortex of the SEHI compared to SESH group [F(3, 28) = 3.52, P = 0.02] evidencing the early astrogliosis caused by HI and the protection offered by the maternal swimming (**Figure 2D**). No alteration was observed in the right hippocampus [F(3, 27) = 1.442, p = 0.255; **Figures 3D**]. Glutamate transporter 1 (GLT-1) was not altered in the cortex [F(3, 25) = 0.871, p = 0.471] nor in the hippocampus [F(3, 26) = 0.152, p = 0.927; **Figures 2D**, **3D**]. The enzyme glutamine synthetase was not altered in the cortex [F(3, 26) = 1.393, p = 0.270; **Figure 2D**]. In the hippocampus (**Figure 3D**) the enzyme expression was increased in the SWHI compared to the other groups [F(3,26) =4.65, p = 0.01].

#### Apoptosis and Inflammation

#### **Cleaved caspase 3 (ccasp3)**

As shown in **Figures 2B**, **3B**, in the cortex, the expression of cleaved caspase-3 was increased in the SEHI compared to all other groups [F(3, 29) = 3.63, p = 0.02] ipsilateral to injury at 24 h. In the hippocampus [F(3, 26) = 2.89, P = 0.05], ccasp3 was increased in the SEHI compared to the SESH group. Maternal swimming prevented the apoptotic cell death increase in the SWHI group in both structures.

#### **Fractin**

No differences were observed neither in the cortex [F(3, 27) = 1.348, p = 0.282] nor in the hippocampus [F(3, 27) = 0.618, p = 0.610; **Figures 2B**, **3B**].

#### **Macrophages/monocytes (CD11b)**

Despite the increase observed in CD11b expression in the SEHI group in the cortex of the group, no significant differences were observed [F(3, 25) = 0.525, p = 0.669; **Figure 2B**]. In the hippocampus, there was a significant increase in the protein expression in the SWHI group compared to SESH and SWSH groups [F(3, 27) = 3.243, p = 0.04] evidencing an early inflammatory reaction (**Figure 3B**).

inflammation (CD11b), (C) neurotrophins VEGF and BDNF and the TRK-B receptor and (D) astrogliosis (GFAP), glutamate receptor GLT-1 and glutamine synthetase enzyme in the four experimental groups: sedentary sham (SESH), sedentary hypoxic-ischemic (SEHI), swimming sham (SWSH) and swimming hypoxic-ischemic (SWHI). WB results are plotted normalized to SESH level expression (100%) (mean ± SEM). Significance testing was determined using one-way ANOVA followed by Duncan's post hoc and was performed on Actin or βIII-tubulin normalized data. \*HI vs. its respective SH group, Significance accepted when p < 0.05.

### Expression of Neurotrophins (VEGF and BDNF and the Receptor Tyrosin Kinase Receptor-B (Trk-B)

**Figures 2C**, **3C** show neurotrophins expression in cortex and hippocampus at PND4. No significant differences observed in VEGF expression in the cortex (**Figure 2C**). **Figure 2C** shows a significant decrease in BDNF expression in the cortex of HI groups (sedentary and swimming) [F(3, 25) = 6.37, P = 0.003]. However, Tyrosin kinase B (Trk-B) receptor expression was decreased only in this SEHI group in the structure [F(3, 24) = 3.20, p = 0.04] evidencing an effect due to swimming.

In the hippocampus, SWHI groups had an increase in VEGF expression compared to SESH and SEHI groups [F(3, 26) = 3.28, p = 0.03; **Figure 3D**]. BDNF expression was increased in the SWHI compared to SESH and SWSH groups [F(3, 27) = 4.94, P = 0.008; **Figure 3C**] and TRK-β protein expression [F(3, 27) = 3.23, p = 0.04] in the SWHI group (**Figure 3C**) compared to all other groups.

At PND60, no differences were observed in the expression of neurons (NeuN) in the cortex [F(3, 27) = 1.315, p = 0.293] and hippocampus [F(3, 26) = 1.915, p = 0.155], astrocytes (GFAP) in the cortex [F(3, 26) = 0.377, p = 0.770] and hippocampus [F(3, 27) = 0.829, p = 0.491], myelin (MBP) in the cortex [F(3, 24) = 0.215, p = 0.885] and hippocampus [F(3, 27) = 0.350, p = 0.789], BDNF in the cortex [F(3, 26) = 0.208, p = 0.890] and hippocampus [F(3, 26) = 0.409, p = 0.748] and Trk-B in the cortex [F(3, 27) = 0.843, p = 0.484] and hippocampus [F(3, 26) = 0.699, p = 0.562; **Figure 7**, upper panels]. The expression of VEGF was significantly increased in the hippocampus of the SEHI group compared to the other groups [F(3, 25) = 3.48, P = 0.03; **Figure 7**, upper panels]. No differences in VEGF in the cortex were observed [F(3, 26) = 0.620, p = 0.620].

#### Behavioral Testing

**Table 3** shows the Elevated Plus Maze (EPM) and Open Field (OF) analysis. In the EPM, rats from the SEHI group had a trend to spend more time in the closed arms (p = 0.06) than in the open arms indicative of anxiety (16 s compared to 8.7 s), however, the latencies were not statistically significant. Together, these results suggest an anxiogenic profile in SEHI rats

prevented by swimming. No increased locomotor or exploratory activity were observed in the OF in the number of neither crossings (horizontal) nor rearings (vertical) exploration that could indicate hyperactivity induced by HI. The cognitive capabilities were tested using non-spatial and spatial tests. The non-spatial testing consisted of the NOR test, based on the inherited exploratory behavior of novelty in rodents. We did not detect impairment in the non-spatial memory, as sham and HI animals explored equally both objects. When spatial memory was evaluated in the MWM, repeated measures ANOVA indicated a significant effect of groups [F(1, 67) = 7.66, p < 0.05] and in the days of training [F(1, 67) = 7.66, p < 0.05]; also, SWHI showed decreased escape latencies to find the platform on the 5th day of training [F(3, 44) = 3.10, p < 0.05] as well as on the latency to reach the platform location in the Probe trial (test day) [F(3, 44) = 3.01, p < 0.05] corresponding to learning impairments in the SEHI group, not observed in the SWHI group (**Figure 5**).

#### Microstructure Evaluation—DTI/NODDI

Direction encoded brain color maps of the rat pups are presented in **Figure 6**. The excellent SNR and resolution quality (70µm inplane) of these images allowed an accurate estimation of diffusion tensor derived parameters. No obvious visual differences were observed between the maps (i.e., thinner cortex in the injured hemisphere due to cortical loss following injury) at the intervals studied. At PND4, FA measurements are presented as mean values of two different brain regions (Cortex and External capsule) (**Figures 4A,B**). In the cortex fiso [F(3, 17) = 3.44, p = 0.04] was reduced in the SWHI group compared to the SWSH. In the external capsule, ficvf [F(3, 17) = 4.01, p = 0.03] was reduced in the SESH group compared to the other groups. No differences were observed in the FA or in the ODI in any of the structures at PND4.

At PND60, FA measurements are presented as mean values of right hemisphere cortex (C), external capsule (EC) and corpus

when p < 0.05.

other groups.

callosum (CC) (**Figures 6A–C**). In the cortex, SEHI group had decreased FA (Z = −2.75, p = 0.02), and increased fiso (Z = −2.052, p = 0.04) and ODI (Z = −2.196, p = 0.028) compared to SESH, evidencing the disruption in the cortical microstructure following HI. No differences were observed in the SWHI group compared neither to SWSH nor to SESHI which implies the neuroprotection of the tissue offered by swimming.

In the CC, no differences between SESH and SWSH were observed. SEHI animals had increased RD (Z = −1.92, p = 0.04) and a decrease in FA (Z = −1.89, p = 0.04) compared to SESH. SWHI had increase AD (Z = −1.93, p = 0.04) and FA (Z = −2.47, p = 0.01) and decreased ODI (Z = −2.04, p = 0.01) compared to SEHI, evidencing the protective effect of swimming on myelinated structures.

In the EC, RD (Z = −2.92, p = 0.03) and fiso (Z = −2.91, p = 0.04) were increased comparing SWSH and SESH groups. Decreased FA (Z = −2.82, p = 0.005) and increased ODI (Z = −2.65, p = 0.008) were observed comparing SEHI and SESH groups. When comparison was made between SWHI and SEHI, FA was increased (Z = −1.93, p = 0.04) and ODI decreased (Z = −193, p = 0.04) in the SWHI groups. No differences regarding microstructure were observed in the basal ganglia using NODDI derived parameters.

### DISCUSSION

In this study, we describe the effects of a gestational swimming protocol on preventing HI-induced early metabolic damage, brain microstructure and late behavioral outcomes. Despite the extensive results presented in the literature about the benefits of maternal exercise, there are gaps in knowledge about its effects and pathways that could lead to brain protection in the offspring. We have shown that at an early stage (24 post HI), <sup>1</sup>H-MRS showed that pregnancy maintained the brain energetic metabolism and limited neuronal damage. Western blotting analysis evidenced that swimming decreased the expression of proteins related to apoptotic cell death, astrogliosis and

FIGURE 5 | Water Maze performance during the 5 days of training (left upper panel). (Right upper panels) - performance on the probe trial. Data are expressed as mean ± SEM (n = 8–16). Lower panels show the representative plots of the Probe Trial. The results were analyzed by two-way ANOVA followed by Duncan's post-hoc test. Significance was accepted when p < 0.05. \*SEHI vs. SESH.

(MD), fractional anisotropy (FA) and NODDI estimates: intraneurite volume fraction (ficvf), cerebrospinal volume fraction (fiso) and orientation dispersion index (ODI) in the external capsule (A), cerebral cortex (B), and corpus callosum (C) for SESH, SEHI, SWSH and SWHI rats at P60. \*SEHI vs. SESH, #SESH vs. SWSH, §SEHI vs. SWHI; p < 0.05.

modulated neurotrophins, especially in the hippocampus. At the brain microstructural level DTI/NODDI showed that swimming caused preservation in the myelinated white matter areas. Also, gestational swimming reduced spatial memory impairments due to HI, which implies that early protection induced by swimming can confer long-term neuroprotection.

panels: representative immunoblots of cortex (left lower) and hippocampus (right lower). WB results are plotted normalized to the SESH group level expression (100%) (mean ± SEM). Significance testing was determined using one-way ANOVA followed by Duncan's post-hoc using Actin or βIII-tubulin as normalizer. \*SESH vs. SEHI. Significance accepted when p < 0.05.

#### TABLE 3 | Behavioral analysis at adult age.


Data are expressed as mean ± SEM (n = 8–16). The results were analyzed by one-way ANOVA followed by Duncan's post-hoc test. Significance was accepted when p < 0.05. No differences were observed.

#### Pregnancy Swimming Alters Brain Response to HI Measured by <sup>1</sup>HMRS

During hypoxia-ischemia insults there is a primary phase of energy failure (up to 24 h following injury) with a decrease in energetic brain metabolites such as ATP and PCr (58, 59), alterations in aminoacids and neurotransmitters, oxidative stress and osmoregulation failure (60). In our study, gestational swimming could limit the decrease in NAAG and total NAA (NAA+NAAG), which implies that gestational swimming can prevent neuronal damage following HI (61). Gestational swimming caused preservation of the energetic metabolism, observed by preservation of PCr concentrations. Interestingly, hypothermia, the clinical standard of HI care, has shown to increase ATP, phosphocreatine, and total NAA levels after HI (62). Also, recently it was shown that pregnancy swimming prevented the failure in the Na+/K+-ATPase caused by HI (50). The reduction of Tau and Ins in the ipsilateral cortex suggests loss of water homeostasis and alterations in glial osmolytes following HI. Both metabolites were reduced in the SEHI, and swimming could impede Ins decrease only in the SWHI group. In agreement with previous reports (16) we observed a decrease in concentration of metabolites related to cell membrane integrity (such as Mac and PE). However, swimming was not able to neither restore nor maintain the Mac and PE levels compared to SESH. The decrease in Cho observed in the SEHI group is attributed to impairments in cell membrane metabolism and to apoptosis (17, 63) and this phenomenon was reverted in the SWHI group, supporting that swimming is acting to decrease apoptosis following HIPND3. The glutamatergic neurotransmission system was altered as suggested by the decrease in the Glu and Tau, and swimming had no effect on these alterations. Contrarily to our expectations, due to the glutamine decrease, the ratio [Glu]/[Gln] was also decreased in the SWHI group, which could indicate an impairment in the Glu and Gln cycling between neurons and glia (16). However, the expression of the glutamine synthetase in the hippocampus (**Figure 3D**) can be interpreted as an attempt of the astrocytes to convert the excess of glutamate due to the HI into glutamine. Lac/NAA ratio reflects mitochondrial impairment and neuronal integrity and a high ratio in the first month after birth asphyxia predicts a poor 12–18 months neurodevelopmental outcome in clinical studies and has been suggested as a potential biomarker of outcome prognostic (64, 65). In our study, we observed an increase in Lac/NAA ratio in the SEHI group, prevented by gestational swimming in the SWHI. One feature of HI injury is the Lac accumulation as the consequence of the anaerobic metabolism following HI (12). We observed that the levels of Lac remained unaltered after HI, pointing to a less severe injury compared to a previous study from the group, in which this metabolite was increased 24 after injury (16). One possible interpretation for this result is that, as like most interventions that have been shown to have neuroprotective effects in HI models (62), swimming could show its effects when the lesion is not as severe, evidencing a limited recovery potential, as observed by Marcelino et al. (66).

### Pregnancy Swimming Decreased Apoptosis and Astrogliosis Following HI

In vivo MRS can detect the disturbances caused by HI in the energy metabolism that trigger a number of pathophysiological responses that ultimately lead to different types of cell death (67, 68). HI on PND3 is well characterized as having both necrotic and apoptotic cell death (21, 23, 69). There was an increase in cleaved caspase 3 (an indicator of apoptotic cell death) in the lysate of hippocampus and cortex of SEHI 24 h after injury. Kim et al. (70) reported (in healthy animals) no difference in DG neuronal apoptotic cell death. Leite et al. (71), using hippocampal slices submitted to oxygen glucose deprivation observed a reduction in the LDH (and decreased cell death) in animals whose mothers swam during pregnancy. Maternal swimming was able to prevent this increase in the SWHI group evidencing the anti-apoptotic effects of maternal swimming as suggested in the literature (72). In agreement, pre-conditioning induced by chronic swimming is also able to protect the brain from excitotoxic events in different models (46, 73–75) as well as modulating the expression of the apoptotic effector proteins such caspase 3 in in vivo experiments (76). Following HIPND3, Sizonenko et al. (23) correlated the acute reduced apparent diffusion coefficient and fractional anisotropy in the ipsilateral cortex to regions of neuronal death, radial glia disruption and astrogliosis. In the present study, the SEHI group had an increase in GFAP levels (an astrogliosis index) in the cortex and the pregnancy swimming was able to minimize astroglial reaction. This is supported by the preservation in the Ins observed by <sup>1</sup>HMRS. It is interesting to note the lack of data reporting the effects of exercise during pregnancy and evaluation of the astrocytes. Kim et al. (70), using a model of PVL reported a decrease in the GFAP immunoreactivity following a protocol of exercise. However, the protocol was performed in the pups, which makes the comparison more difficult.

One of the central hypothesis of the beneficial effects of gestational swimming is its ability to increase the production of neurotrophins (44, 45, 48). When evaluated 24 h following HI, we observed a decrease in BDNF in the cortex of HI groups (SE and SW). However, the receptor Trk-B was decreased only in the SEHI group. In the hippocampus, there was an increase in the neurotrophins (BDNF and VEGF) as well as in the TRK-B receptor in the SWHI group. BDNF controls the development, survival, and differentiation of the neurons through Trk-B. We can speculate that the hippocampus acts like a "sensor," identifying the injury and increasing the production of neurotrophins. In agreement, authors observed an increase in BDNF levels in the hippocampal formation of animals whose mother swam during pregnancy (38, 46, 50). At adult age, HI increased VEGF expression in the hippocampus (related to spatial memory) of the SEHI rats but not in the SWHI (46). The increase in VEGF expression in the hippocampus during chronic epilepsy in both humans and animal models has been associated with increased angiogenic processes and blood-brain barrier disruption, which could worsen the injury (77). The expression of BDNF and the Trk-B receptor were not altered at adult age. In agreement, Marcelino et al. (66) did not find differences in the levels of BDNF at adult age and attributed this to the time point of evaluation (66). Here, we show that protein expression alteration induced by swimming in the BDNF signaling following HI seems to be more important in the early phase of injury.

### Pregnancy Swimming Mitigates Cognitive Impairments and White Matter Injury Induced by HI

To assess the effects of pregnancy swimming over the functional impairments caused by HI at PND3, we used anxiety-related (elevated plus maze), locomotor (open field) and cognitive tests (NOR and Morris water maze). HI causes anxiety related alterations in the SEHI animals that were not observed in the SWHI animals, providing evidence of functional neuroprotection induced by gestational swimming. In agreement, Torabi et al. showed that maternal swimming prevented anxiety-related behavior in the offspring of morphine-dependent mothers (78). The motor function analysis did not reveal any gross motor deficit nor hyperactivity due to the hypoxia-ischemia model nor an improvement induced by gestational swimming. In agreement, literature has shown that hyperactivity in the open field (i.e., increased number of crossings) and in other motor tests (such as the asymmetrical use of the forelimbs in the cylinder test) using the same model are not altered by the HI PND3 which can point to the preservation of the cortico-spinal tract (50, 79). Since we did not detect a volumetric decrease in the ipsilateral hemisphere, it is reasonable to accept that the tissue injury was not sufficient to induce motor impairments. The degree of injury is highly correlated to the functional deficits and when the injury parameters are modified to obtain a more severe damage, motor alterations are observed (80–83). In agreement, Ueda et al. (81) have shown discrete motor impairments following HI attributed to a disorganization of oligodendrocyte development in layers II/III of the sensorimotor cortex (81).

At the functional level, one of the main consequences of neonatal hypoxia-ischemia is cognitive impairment, independently of the stage of brain maturation in which the injury occurs (83–85). In this context, extensive research has demonstrated that maternal exercise can potentially have positive effects on cognitive function in the offspring (45, 48, 50, 66). In our study, we did not observe non-spatial cognitive impairment (assessed in the NOR test) meaning preservation of areas in the perirhinal cortex (which plays the role of encoding information for the object discrimination performance) (86). The hippocampus, the most studied structure in the cognitive tests, seems to be highly correlated to the NOR test. Different degrees of injury (lesser than 70%) in the structure are unable to produce impairments in the test. However, MWM evidenced learning impairments in the SEHI animals, who presented greater latencies to find the platform in the last day of training and in the probe trial. Swimming had neuroprotective effect by preserving spatial memory following HI at adult age, in agreement with recent published data (50).

White matter injury is associated with a wide range of neurologic dysfunction (4, 87, 88), and can be the cause of spatial memory impairment observed in the SEHI group. Structural alterations can be observed through DTI derived parameters (median, axial and radial diffusivity and fractional anisotropy) that delineate white matter microstructural damage in animal models of perinatal brain injury in relation with altered myelination (17, 57, 89, 90). Typically, there is a reduction in MD values in the acute phase of ischemia, and in FA

#### REFERENCES


values in the subacute/chronic phase (91). In our study, in the early phase after injury (24 h post HI) we did not detect any substantial differences in the DTI derived parameters observed in the cortex as well as in the external capsule. However, at PND60 ipsilateral hemisphere of SEHI animals showed a decrease in FA in the assessed structures (cortex, external capsule and corpus callosum) not observed in SWHI groups. Also, ODI was increased in SEHI groups in cerebral cortex and external capsule. Such alterations were partially recovered in the SWHI group, providing evidence that the myelination long-term impairment after HI injury was partially protected by pregnancy swimming.

We demonstrate that exercise during pregnancy is able to modulate brain functioning and to adapt its metabolism in order to protect itself against HI-induced damage. This adaptation induced the inhibition of apoptotic cell death, astrogliosis and the preservation of the white matter structure, reducing behavioral outcomes. To define the relationship of diffuse white matter injury sparing, further analysis of the cell types and of the damage and repair mechanisms involved will be necessary. The findings of this work indicate that maternal swimming modulates several pathways related to the HI cascade, denoting that gestational interventions have the potential to induce long-term neuroprotective effects on biomarkers and should be examined in future human studies.

### AUTHOR CONTRIBUTIONS

ES: conception of the study, acquisition, analysis and interpretation of data, drafting the article. YVdeL: acquisiton, analysis and interpretation of data, drafting the article. AdS, JR, and AT: acquisition of data. SS: supervisor, conception, data analysis, critical revision of the article, final approval.

#### FUNDING

ES received a Swiss Excellence Scholarship for Foreign Scholars to perform the study in our laboratory. This study was supported by the Swiss National Fund N◦ 33CM30-124101/140334 and the Fondation pour Recherches Médicales, Geneva.


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

# Modulation of Microglial Activation by Adenosine A2a Receptor in Animal Models of Perinatal Brain Injury

Marina Colella1,2 \*, Manuela Zinni <sup>1</sup> , Julien Pansiot <sup>1</sup> , Michela Cassanello<sup>3</sup> , Jérôme Mairesse1,4, Luca Ramenghi <sup>5</sup> and Olivier Baud1,4

<sup>1</sup> Robert Debré Hospital, PROTECT, Inserm U1141, Paris, France, <sup>2</sup> Istituto G. Gaslini, Università di Genova, Genoa, Italy, <sup>3</sup> Laboratory for the Study of Inborn Errors of Metabolism, Istituto Giannina Gaslini, Genoa, Italy, <sup>4</sup> Division of Neonatology and Pediatric Intensive Care, Children's University Hospital of Geneva, University of Geneva, Geneva, Switzerland, <sup>5</sup> Neonatal Intensive Care Unit, Istituto Giannina Gaslini, Genoa, Italy

Neuroinflammation has a key role in the pathogenesis of perinatal brain injury. Caffeine, a nonspecific antagonist of adenosine receptors (ARs), is widely used to treat apnea of prematurity and has been linked to a decrease in the incidence of cerebral palsy in premature infants. The mechanisms explaining its neuroprotective effect have not yet been elucidated. The objective of this study was to characterize the expression of adenosine and ARs in two neonatal rat models of neuroinflammation and to determine the effect of A2aR blockade on microglial activation assessed through inflammatory cytokine gene expression. We have used two rat models of microglial activation: the gestational low protein diet (LPD) model, associated with chronic brain injury, and postnatal ibotenate intracerebral injections, responsible for acute excitotoxicity injury. Adenosine blood levels have been measured by Tandem Mass Spectrometry. The expression of ARs in vivo was assessed using qPCR and immunohistochemistry. In vivo models have been replicated in vitro on primary microglial cell cultures exposed to A2aR agonist CGS-21680 or antagonist SCH-58261. The effects of these treatments have been assessed on the M1/M2 cytokine expressions measured by RT-qPCR. LPD during pregnancy was associated with higher adenosine levels in pups at postnatal day 1 and 4. A2aR mRNA expression was significantly increased in both cortex and magnetically sorted microglial cells from LPD animals compared to controls. CD73 expression, responsible for extracellular production of brain adenosine, was significantly increased in LPD cortex and sorted microglia cells. Moreover, CD73 protein level was increased in ibotenate treated animals. In vitro experiments confirmed that LPD or control microglial cells exposed to ibotenate display an increased expression, at both protein and molecular levels, of A2aR and M1 markers (IL-1β, IL-6, iNOS, TNFα). This pro-inflammatory profile was significantly reduced by SCH-58261, which reduces M1 markers in both LPD and ibotenate-exposed cells, with no effect on control cells. In the same experimental conditions, a partial increased of M1 cytokines was observed in response to A2aR agonist CGS-21680. These results support the involvement of adenosine and particularly of its receptor A2aR in the regulation of microglia in two different animal models of neuroinflammation.

#### Edited by:

Alberto Spalice, Policlinico Umberto I, Italy

#### Reviewed by:

Renata Rizzo, Università degli Studi di Catania, Italy Francesco Fazio, I.R.C.C.S. Neuromed, Italy

> \*Correspondence: Marina Colella marina.colella@inserm.fr

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 26 January 2018 Accepted: 06 July 2018 Published: 11 September 2018

#### Citation:

Colella M, Zinni M, Pansiot J, Cassanello M, Mairesse J, Ramenghi L and Baud O (2018) Modulation of Microglial Activation by Adenosine A2a Receptor in Animal Models of Perinatal Brain Injury. Front. Neurol. 9:605. doi: 10.3389/fneur.2018.00605

Keywords: adenosine, inflammation, brain damage, fetal growth restriction, prematurity, microglia

### INTRODUCTION

Brain injury is one of the most important complication related to preterm birth (1). From 25 to 50% preterm infants display neurodevelopmental disabilities (2), with dramatic consequences in terms of cost and impact on quality of life.

Preclinical and clinical studies show that neuroinflammation plays a central role in the pathogenesis of perinatal brain damage (3–8). One of the first events following neuroinflammation is activation of microglia cells (9), that assume different phenotypes, conventionally classified as M1 (pro-inflammatory) and M2 (anti-inflammatory, reparative) (10, 11).

During the inflammatory process, extracellular adenosine, an ubiquitous molecule implicated in neuromodulation, reaches high concentrations capable of activating the adenosine receptors (ARs), denoted A1, A2a, A2b, and A3 (12, 13).

The most implicated adenosine receptor in neuroinflammation is A2aR (14). Its expression in microglia is usually low but increases following brain insults. In microglial cells, activation of A2aRs has facilitating effects on the release of cytokines (15) and on the change into amoeboid morphology (16). Conversely, A2aR antagonists suppress microglia activation, as described using in vitro (17, 18) and in vivo (18) studies.

To our knowledge, there are no data regarding the adenosine pathway and neuroinflammation in preterm infants, but nevertheless, the involvement of adenosine signaling in prematurity is suggested by the clinical use of caffeine. Indeed, caffeine, a non-specific antagonist of ARs widely used to treat apnea of prematurity, not only improves survival and reduces the duration of respiratory support, but also reduces the incidence of cerebral palsy and cognitive delay (19). Recently, a retrospective study demonstrated the existence of high blood levels of adenosine in premature infants (20), with the highest adenosine concentrations associated with the lowest birthweight.

These data suggest a possible link between caffeine action, adenosine plasma levels and an imbalance between the proand anti-inflammatory profiles in very preterm infants usually delivered following a perinatal inflammatory event. Whether a similar link exists for adverse neurological outcomes in preterm infants is not known and there is still little evidence relating to effects of caffeine on brain development, especially at the cellular and molecular levels (21).

Therefore, this study was aimed to characterize the synthesis and expression of adenosine and its receptors in two experimental animal models of neonatal neuroinflammation. The effect of A2aR blockade was also studied in vitro using a specific antagonist on microglial activation assessed through pro- and anti-inflammatory cytokine gene expressions.

#### MATERIALS AND METHODS

#### Animals and Models

All experiments were carried out according to INSERM ethical rules and approved by the institutional review board (Robert Debré ethics committee, Paris, France, approval number Big Project 01542.01). Sprague-Dawley rats (Janvier SAS, Le Genest-St-Isle, France) were housed in temperature-controlled rooms (24◦C), with 12 h light cycling and free access to chow and water ad libitum.

#### Low Protein Diet (LPD) Model

After mating, dams were randomly allocated to either isocaloric low-protein diet (LPD) (9% casein; as previously described (22, 23), SAFE-diets Augy, France) or control diet (CTL) (23% casein) during the gestational period. The control and LP diets are balanced for energy intake assuming equivalent consumption rates. At birth, dams were returned to standard diet. Sex, birthweight and postnatal growth rates were determined. Experiments research plan using this model is summarized in **Supplemental Figure 1**.

#### Ibotenate (IBO) Model

Ten µg IBO diluted in Phosphate Buffered Saline (PBS) was injected intracerebrally (i.c.) at postnatal day 5 (P5) to rat pups of both sexes as previously described (24). Experiments research plan using this model is summarized in **Supplemental Figure 2**.

The rat pups were killed and dissected at different postnatal days (P1, P4, and P5). Blood was collected by exsanguination on filter paper. Brains were collected, immediately snap frozen and stored at −80◦C or immediately dissociated for microglia cells isolation.

#### Antibodies and Reagents

Ibotenate (IBO, Tocris, Bristol, UK) was diluted in PBS to prepare a stock solution of 20 mg/ml. SCH-58261 (SCH; Sigma Aldrich, Lyon, France S4568) and CGS-21680 hydrochloride hydrate (CGS; Sigma Aldrich, Lyon, France, C141) were diluted in DMSO to a stock concentration of 10 mM. Primary antibodies: anti-A2aR antibody (rabbit polyclonal, ab3461); anti-CD73 antibody (rabbit polyclonal, ab175396); anti-ionized calcium-binding adaptor protein-1 antibody (anti-Iba1, goat polyclonal, ab5076) were all purchased from Abcam (France). Goat anti-rabbit and anti-mouse IgG conjugated to horseradish peroxidase were purchased from Sigma (Lyon, France). AlexaFluor <sup>R</sup> 488-conjugated anti-goat IgG and DAPI were from Life Technologies, while Cy3 conjugated anti-rabbit IgG was from Jackson Immuno Research Laboratories.

#### Microglia Cell Isolation and Primary Culture

Brains were collected from control and LPD animals at P1 and P4 removing the cerebellum and the olfactory bulbs. The tissues dissociation was performed using the Neural Tissue Dissociation Kit and the gentleMACS Octo-Dissociator with Heaters accordingly to the manufacturer's instruction (Miltenyi Biotec, Germany). CD11b positive cells were isolated from the resulting homogenates using an anti-CD11b (microglia marker) MicroBeads (Miltenyi Biotec, Germany) and multiMACSCell-24 separator (Miltenyi Biotec, Germany). After elution the sorted microglia cells were stored at −80 for RNA extraction. In a second set of experiments, microglia cells were magnetically

sorted from control and LPD animals at P4 and after elution pellet was isolated by centrifugation (300 g - 10 min). Following re-suspension in Dulbecco's modified Eagle's minimum essential medium/Nutrient mixture F-12 (DMEM/F-12, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin/streptomycin (P/S), cells were maintained in DMEM/F-12 supplemented with 10% FBS and 1% P/S at a concentration of 5 × 10<sup>5</sup> cells/ml in 12-well culture plates. The purity of isolated microglia cells was verified by Iba1 immunostaining (dilution 1/1,000). A medium change was performed after 24 h and cells were treated as follows after 48 h. Microglial cells were treated with SCH-58261 (A2aR antagonist) at 50 nM (17, 25) or CGS-21680 (A2aR agonist) at 10µM (26, 27) or DMSO. For the IBO model, cells sorted at P4 were treated with SCH, CGS or DMSO 20 min before adding ibotenate 300µM (28). After 6 h, cells were harvested and RNA extracted for gene expression analysis. For cytokine levels, supernatant (conditioned media) was collected after a longer exposure time (12 h) and stored at −80◦C until analysis.

### RNA Extraction, Retro-Transcription and Real-Time PCR

Total RNA was extracted from cortex using Qiazol reagent and RNeasy mini kit (Qiagen, France) and from microglia cells using the NucleoSpin RNA Plus extraction kit (Macherey-Nagel, France) according to the manufacturer's instructions. RNA yield and purity were determined by spectrophotometry absorption at 260 and 280 nm by means of a NanodropTM apparatus (Thermofisher Scientific, MA, USA). Five hundred ng of mRNA from cortex and 150 ng from microglia were used to perform reverse transcription (iScript TM cDNA synthesis kit, Biorad, France), respectively. qPCR measurements were performed in duplicate using SYBR Green Super-mix (Bio-Rad). The reaction conditions were as follows: 98◦C for 10 min (Polymerase activation), followed by 45 cycles at 95◦C for 5 min, 60◦C for 10 min and 72◦C for 10 min. The specificity of used primers was assessed with a melting curve analysis and the results were quantified using the relative standard curve methods. The relative mRNA expression for each target gene was calculated after normalization respect to the Rpl13 references gene. The primers sequences are available in **Supplemental Table 1**.

### Multiplex Cytokine Assay

Cytokines were measured using the Bio-Plex rat cytokine multiplex kit (Bio-Rad). Calibration curves from recombinant cytokine standards were prepared with serial dilutions in the same media as the culture supernatant (DMEM/F-12 supplemented with 10% FBS and 1% P/S). Standards and samples were analyzed in duplicate and blank values were subtracted from all readings. All assays were carried out directly in a 96-well filtration plate (Bio-Rad) at room temperature and protected from light. Briefly, wells were pre-wetted with culture supernatant, then beads together with either standard, sample, or blank were added in a final volume of 50 µl, and incubated together at room temperature for 30 min with continuous shaking. Beads were washed three times with 100 µl Bio-Plex wash buffer. A cocktail of biotinylated antibodies (25 µl/well) was added to the beads for a further 30-min incubation with continuous shaking. Beads were washed three times, then streptavidin-phycoerythrin was added for 10 min. Beads were again washed three times and resuspended in 125 µl assay buffer. The fluorescence intensity of the beads was measured using the Bio-Plex array reader. Bio-plex manager software with five-parametric-curve fitting was used for data analysis.

#### Immunofluorescence Assay and Quantification

For histological analysis after ibotenate i.c., injections, animals were anesthetized with pentobarbital and transcardially perfused with 4% paraformaldehyde in PBS. Brains were collected, postfixed in 4% paraformaldehyde overnight, cryoprotected, cut coronally in 10 µm-thick slices, and stained according to standard protocols. After three washings of the slices with PBS, the non-specific binding was blocked by incubating the tissue sections with PBS-Triton 0.5%-gelatin 0.2% for 45 min at room temperature. Incubation with primary antibodies (rabbit anti-CD73 1/1,000; goat anti-Iba1 1/1,000) was performed overnight at 4◦C in PBS-Triton 0.5%-gelatin 0.2%. After rinsing three times in PBS for 5 min each, sections were exposed (1 h, room temperature) to secondary species-specific antibodies (all at 1/1,000 dilution in PBS-Triton 0.5%-gelatin 0.2%) conjugated to Alexa Fluor <sup>R</sup> 488 or to Cy3. Nuclei were then labeled with the fluorescent DAPI dye (1/10,000 in PBS). Stained sections were mounted on microscope slides with Fluoromount-G (SouthernBiotech).

Primary microglia cells cultured in micro-slide 8-well chamber (Ibidi, Germany) and treated as reported above were fixed in 4% paraformaldehyde for 30 min at room temperature. Each well was incubated with a blocking solution (PBS with 1% BSA) for 1 h at room temperature and incubated overnight at 4 ◦C with goat anti-Iba1 (1/500) and anti-A2aAR (1/250). The following day, after rinsing three times in PBS for 5 min, cells were incubated with secondary antibodies coupled to the green and red fluorescence markers (1/500 dilution) for 1 h at room temperature. Nuclei were visualized by staining the cells with DAPI dye (1/10,000).

Cells were analyzed using a fluorescent microscope (Nikon Eclipse Ti-E) and images captured with a 20X objective (4 wells/group and 5 images/well). Fields used for quantitation were randomly selected throughout the dish and focused using phase contrast optics. Images from different emission specters were acquired separately using the same parameters and superimposed in the aftermath. For the analysis Image J software (Research Service Branch, National Institutes of Health, Bethesda, MD; http://rsb.info.nih.gov/ij/) was used. Images were converted to binary images using an automatic threshold function. The cells in each image were then defined by outlining a mask image (ROI). For LPD experiments, as the cells showed a significant difference in cell size due to microglial activation, the fluorescence power was calculated as the mean of all the pixel intensities of each individual cell. For the ibotenate experiment, the fluorescence power was calculated as an integrated density (i.e., the product of the

mean of all the pixel intensities of each individual cell and the ROI area). Cell size was calculated using Iba1-positive cells as the product of number of pixels in ROI and the conversion factor 0,103. Finally, the sums of the values for each condition were normalized to control values for the statistical analysis.

### Statistical Analysis

The graphs and the statistical analysis were performed with GraphPad Prism 5.0 (GraphPad Software, San Diego, CA). Appropriate statistical analyses were carried out with a two-sided alpha level of 0.05 or 95% confidence interval. For continuous data, descriptive analyses were carried out employing means and ± SEM, Parametric or non-parametric tests were used to compare quantitative variables (Student's t-test for independent samples for comparisons between two groups and one-way ANOVA with Newman Keuls post-test for comparisons of more than two groups). A p-value < 0.05 was considered statistically significant.

### RESULTS

#### The LPD Model Influences Adenosine Blood Levels and the Cerebral Expression of Enzymes and Receptors Implicated in Adenosine Metabolism

Prenatal LPD-induced malnutrition resulted in the absence of postnatal mortality but fetal growth restriction with body weight of the pups significantly lower from P0 to at least P4 compared to the controls (**Supplemental Figure 4**). LPD pups at P1 and P4 showed significantly higher adenosine blood levels compared to the control pups (0.822 ± 0.088µM in Controls vs. 1.850 ± 0.355µM in LPD at P1, 0.598 ± 0.051µM in Controls vs. 1.464 ± 0.215µM LPD at P4, see **Figure 1**).

In the pre-frontal cortex of LPD animals, the expression of adenosine deaminase (ADA), adenosine kinase (ADK), ectonucleoside triphosphate diphosphohydrolade-1 (Entpd1) and cluster of differentiation 73 (CD73), enzymes involved in the regulation of extracellular and intracellular adenosine levels, were significantly increased at P1, compared to controls (**Figure 2A**). The relative gene expression of intracellular enzymes ADA and ADK were comparable between LPD and controls at P4, whereas the two ectonucleotidases Entpd1 and CD73 are persistently increased at this age in the LPD group, indicating an increased extracellular production of adenosine. A2aR and A2bR, the two main ARs with pro-inflammatory functions, were significantly up-regulated at P1 and P4 in the cortex of LPD pups (**Figure 2B**), suggesting a pro-inflammatory status of the LPD brains. In contrast, no significant differences were observed at a transcriptional level for A1 and A3 receptors, except for an increased expression of A3R at P4 in the LPD group (**Figure 2B**).

### A2aR Expression Is Increased in Microglial Cells Sorted From the Rat Pups Subjected to Antenatal LPD

While most of the receptors display a similar pattern of expression, A2aR expression was found significantly increased in microglia cells sorted from LPD brains compared to controls, both at P1 and P4 (**Figure 3**). Finally, no statistically significant difference was observed in the expression of genes encoding for the enzymes involved in the adenosine extracellular metabolism, except for a transient and mild increase in the expression of CD73 at P1 in LPD animals.

The mean density of A2aR immunoreactivity in microglial cells after 48 h in culture was significantly higher in LPD compared to control group (**Figures 4A,B**). Moreover, Iba1 immunoreactivity was found increased and cell size reduced in LPD microglial cells, compared to control cells (**Figures 4C,D**).

### A2aR Antagonist Exposure Changes Microglial Reactivity in vitro

After 2 day, significant increases in gene expression of both M1 (IL-1β, IL-6, iNOS, TNFα) and M2 markers (IL-10, IL-4ra) were detected in microglial cells sorted from LPD rat pups, compared to control cells (**Figure 5**). SCH-58261, an A2aR antagonist, induced a significant reduction in the expression of M1 markers while no effect on M2 markers was detected in LPD microglial cells. Conversely, the A2aR agonist CGS-21680 was able to increase the mRNA levels of iNOS, TNFα and IL-4ra in LPD microglial cells. No substantial effect of either SCH-58261 or CGS-21680 was observed in control cells.

### A2aR Plays a Role in Excitotoxic-Induced Microglial Activation

In the IBO model assessed 24 h after ibotenate injection, 1/3 of Iba-positive cells surrounded the white matter lesion expressed CD73, an enzyme responsible for the extracellular production of adenosine in the brain (**Supplemental Figure 5**).

To assess the putative involvement of the adenosine pathway in neuro-inflammation induced by excitotoxicity,

FIGURE 2 | Comparisons of gene expression of enzymes involved in the production of adenosine (A) and its receptors (B) in animals exposed to LPD (black bars) and in control animals (white bars) in prefrontal cortex. Data are shown as relative expression of control values normalized to 1 (\*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001, using unpaired Student's t-test). ADA, adenosine deaminase; ADK, adenosine kinase; Entpd1, ectonucleoside triphosphate diphosphohydrolade-1; CD73, cluster of differentiation 73; A1R, adenosine A1 receptor; A2aR, adenosine A2a receptor; A2bR, adenosine A2b receptor; A3R, adenosine A3 receptor.

magnetically sorted microglial cells from P4 control pups were exposed to 300µM IBO for different lengths of time from 2 to 12 h. The exposure to IBO was unable to induce an increase in A1 and A2a receptor expression levels after 2-h exposure (**Figure 6**). In contrast, a significant and progressive increase in the gene expression of both receptors was observed after 6- and 12-h IBO exposure. CD73 was also found to be slightly but significantly increased, while no difference was observed regarding A2b and A3 receptors gene expression.

These results were confirmed at protein level using immunocytochemistry (**Figure 7**). IBO exposure for 6 h was associated with a significant increase in A2aR immunoreactivity in microglial cells, while cell size and the Iba1 immunoreactivity were found similar with and without IBO.

### A2aR Modulation Is Involved in the Regulation of M1-M2 Microglia Phenotype in Excitotoxic-Induced Inflammation

In our in vitro model of excitotoxic activation of primary microglial cells, highly significant increase in gene expressions of IL1β, IL6, iNOS, TNFα, and IL10 were was observed in response to 300µM IBO exposure for 6 h (**Figure 8**). This effect was significantly reduced by SCH-58261 pre-treatment, 20 min before IBO challenge. Interestingly, SCH-IBO treated microglia display a higher expression level of the M2 cytokine IL-4ra.

Pre-treatment with CGS-21680 induced down-regulation of TNFα and IL4ra and up-regulation of IL6 gene expression in IBO-treated cells, but had no effect on gene expressions of IL1β, iNOS and IL10.

In another set of experiments, conditioned culture media were collected after 12-h IBO exposure with or without SCH-58261 pre-treatment, and cytokine concentrations were assessed (**Figure 9**). While IBO induced higher IL1β and TNFα concentrations in the culture media, the A2aR antagonist SCH-58261 exposure was associated with a reduced cytokine production in response to excitotoxic challenge.

## DISCUSSION

This study strongly suggests that adenosine and the regulation of its receptor A2aR play a role in neonatal brain inflammation and microglial activation in rat.

Abundance of literature has demonstrated both in clinical and preclinical studies that neuroinflammation is a relevant component in the pathogenesis of prematurity-related brain injury (3, 4, 9, 29–32). Interestingly, adenosine exerts a role in

this context and is able to orchestrate the inflammatory response (33–35). However, its role in the neonatal brain has not yet been elucidated, even if caffeine, a non-specific adenosine antagonist, has shown an important neuroprotective role in premature infants.

The adenosine system appears to be involved in regulating inflammation in both acute (IBO) and chronic (LPD) neuroinflammation in vivo. The two animal models used in the present study display alterations that occur at a developmental stage of the rat brain that corresponds to the human brain at 28–32 gestation weeks (GW) (36). This window is recognized as a period of high vulnerability for the developing brain to either excitotoxic or inflammatory insults (37). Interestingly, these effects are exerted through the modulation of microglia reactivity, that, as previously reported (23, 38), characterizes the two animal models.

In conditions of inflammation, oxidative stress, excitotoxicity or cellular necrosis, the purinergic system is the first to

be involved (39). Indeed, under pathological conditions, extracellular ATP is produced by both neurons and glial cells (40) and is rapidly converted to ADP and AMP by Entpd1 and by CD73, which convert AMP to adenosine (41). Despite its very short half-life (few seconds), increased adenosine brain levels contributes to induction and modulation of neuro-inflammation (42).

To assess the adenosine implication in the LPD model, an adenosine assay on whole blood was performed. Since deliveries can span a 12-h and to avoid the stressful peak related to delivery, the blood samples were collected at P1, when LPD pups showed higher level of adenosine blood levels compared to controls. As described for human neonates (43), delivery is responsible for a physiological increase in adenosine blood levels in the newborn. Interestingly, adenosine blood levels remained higher in LPD at P4 in our study, when the effects of delivery have disappeared.

Remarkably, the increasing adenosine blood levels in LPD rats are similar to those found in premature infants (20) and suggest a pro-inflammatory condition that characterizes these babies

from birth until the first month of life. Similarly to neonates, LPD pups also showed a chronic inflammatory condition, as a result of maternal malnutrition, with a detrimental effect on neurodevelopment (23). In the brain, CD73, also known as Nt5e, is considered as the principal enzyme involved in the production of extracellular adenosine (44). The mRNA expression of the two ectonucleotidases Entpd1 and CD73, responsible for the final step of ATP catabolism into adenosine, is significantly increased in LPD both at P1 and P4. These results are in agreement with the study conducted by Chen et al. who reported that the activity of the ectonucleotidases is stimulated by inflammatory conditions (45).

Adenosine elicits its physiological responses by binding to and activating one or more of the four transmembrane ARs. Solid evidence demonstrated that the four receptors are all expressed in the brain (33) and our results confirm the expression of all ARs in the rat pup cortex. Interestingly, LPD animals displayed an increased expression of A2aR and A2bR, known to exert a pro-inflammatory action (14, 46), in the pre-frontal cortex both at P1 and P4. In the LPD model, which induces fetal growth restriction, the main alteration consists in disturbance of oligodendrocytes progenitor cells (OPC) maturation conducing to a deficit of myelination, that occurred in combination with a proinflammatory state evidenced by transcriptomic analysis performed in sorted microglia (23). Our in vitro results revealed that microglial cells sorted from rat pups subjected to LPD have abnormal reactivity with increased Iba1 staining and smaller size, when compared to control cells. Iba1 is constitutively expressed by microglia and is involved in the actin-crosslinking associated with membrane ruffling of microglial cells, an event essential for the morphological changes from quiescent ramified microglia to activated amoeboid microglia (47). Furthermore, the reduction in cell size has been shown to be strongly correlated to microglial activation (48, 49).

Regarding the potential role of adenosine in the modulation on microglia activity, we reported an increase in A2aR transcripts and protein levels in LPD-exposed microglia cells. A2aR has an important role in the control of inflammatory events (14) by regulating microglial reactivity, changing microglial

morphology (16), increasing the release of cytokines and prostaglandin E2 (15) and nitric oxide synthase activity. In addition to A2aR, CD73 has been shown to have an important role in modulation of microglia ramification and activation (51). These data are consistent with our findings and support a possible role of adenosine in the regulation of the inflammatory response following brain injury.

The results of in vitro studies conducted using an A2aR pharmacological approach clearly evidenced the involvement of A2aR in the regulation of inflammatory response in LPD model. These results are well supported by previous studies, which demonstrated that A2aR antagonists suppress microglia activation and IL-1β secretion in murine microglial cells exposed to an inflammatory stimulus induced by LPS (17, 18). A2aR gene disruption in mice showed a lower severity of inflammatory response and subsequent damage in different models of brain injury including ischemia/hypoxia and traumatic brain injury (50–52).

On the other hand, treatment with CGS-21680, an A2aR selective agonist, promotes the increase in M1 markers in LPD microglial cells suggesting that chronic inflammation causes microglial cells to be more susceptible to the pro-inflammatory effect of adenosine via A2a receptors.

The Ibotenate model closely mimics the pathological features observed in periventricular white matter (38). In this model

of acute brain injury and inflammation, both in vivo and in vitro findings support the involvement of A2aR. However, some features of microglial activation in vitro appear to be different from those observed in the LPD model, as microglial cells treated by ibotenate showed no difference in cell size and Iba1 expression. In conclusion this study gives evidence of the involvement of adenosine and in particular of its receptor A2aR in the regulation of microglia in two models of perinatal brain injury associated with neuro-inflammation. The present study focused only on the relation between adenosine formation and A2aR inactivation; it remains to be explored whether the other receptors might play a role in the regulation of microglia in the same models of perinatal brain injury. Indeed, all the 4 receptors subtypes are non-specifically inactivated by caffeine. Further studies may provide a functional role for caffeine and specific antagonists of the remaining ARs in the limitation of perinatal brain injury associated with neuro-inflammation. In summary, the present study suggests that A2aR, up-regulated as consequence of inflammation, can influence the microglia phenotype, building up a potential for A2aR antagonist as a therapeutic strategy for neonatal brain damage.

#### AUTHOR CONTRIBUTIONS

MC, JM, LR, and OB designed the study. MC, MZ, and JP performed experiments. MCa performed adenosine measurements. MC, JM, and OB wrote the paper. All authors revised and approved final version of the manuscript.

#### FUNDING

This study was supported by Inserm, France and by University of Genoa, Italy.

#### ACKNOWLEDGMENTS

We thank Audrey Toulotte-Aebi for the editing of the manuscript draft.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00605/full#supplementary-material

### REFERENCES


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

Copyright © 2018 Colella, Zinni, Pansiot, Cassanello, Mairesse, Ramenghi and Baud. 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.

# Intravenous Administration of Bone Marrow-Derived Mesenchymal Stem Cell, but not Adipose Tissue-Derived Stem Cell, Ameliorated the Neonatal Hypoxic-Ischemic Brain Injury by Changing Cerebral Inflammatory State in Rat

#### Edited by:

*Olivier Baud, Geneva University Hospitals (HUG), Switzerland*

#### Reviewed by:

*Sheffali Gulati, All India Institute of Medical Sciences, India Susan Cohen, Medical College of Wisconsin, United States*

#### \*Correspondence:

*Yoshiaki Sato yoshiaki@med.nagoya-u.ac.jp*

#### Specialty section:

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

Received: *22 January 2018* Accepted: *20 August 2018* Published: *11 September 2018*

#### Citation:

*Sugiyama Y, Sato Y, Kitase Y, Suzuki T, Kondo T, Mikrogeorgiou A, Horinouchi A, Maruyama S, Shimoyama Y, Tsuji M, Suzuki S, Yamamoto T and Hayakawa M (2018) Intravenous Administration of Bone Marrow-Derived Mesenchymal Stem Cell, but not Adipose Tissue-Derived Stem Cell, Ameliorated the Neonatal Hypoxic-Ischemic Brain Injury by Changing Cerebral Inflammatory State in Rat. Front. Neurol. 9:757. doi: 10.3389/fneur.2018.00757* Yuichiro Sugiyama<sup>1</sup> , Yoshiaki Sato<sup>1</sup> \*, Yuma Kitase<sup>1</sup> , Toshihiko Suzuki <sup>1</sup> , Taiki Kondo<sup>1</sup> , Alkisti Mikrogeorgiou<sup>1</sup> , Asuka Horinouchi <sup>2</sup> , Shoichi Maruyama<sup>2</sup> , Yoshie Shimoyama<sup>3</sup> , Masahiro Tsuji <sup>4</sup> , Satoshi Suzuki <sup>5</sup> , Tokunori Yamamoto5,6,7 and Masahiro Hayakawa<sup>1</sup>

*<sup>1</sup> Division of Neonatology, Center for Maternal-Neonatal Care, Nagoya University Hospital, Nagoya, Japan, <sup>2</sup> Department of Nephrology, Nagoya University Graduate School of Medicine, Nagoya, Japan, <sup>3</sup> Pathology and Clinical Laboratories, Nagoya University Hospital, Nagoya, Japan, <sup>4</sup> Department of Regenerative Medicine and Tissue Engineering, National Cerebral and Cardiovascular Center, Osaka, Japan, <sup>5</sup> Center for Advanced Medicine and Clinical Research, Nagoya University Graduate School of Medicine, Nagoya, Japan, <sup>6</sup> Department of Urology, Nagoya University Graduate School of Medicine, Nagoya, Japan, <sup>7</sup> Laboratory for Clinical Application of Adipose-Derived Regenerative Cells, Nagoya University Graduate School of Medicine, Nagoya, Japan*

Perinatal hypoxic-ischemic (HI) brain injury occurs in 1 in 1,000 live births and remains the main cause of neurological disability and death in term infants. Cytotherapy has recently emerged as a novel treatment for tissue injury. In particular, mesenchymal stem cells (MSCs) are thought to have therapeutic potential, but little is known about the differences according to their origin. In the current study, we investigated the therapeutic effects and safety of intravenous injection of allogeneic bone marrow-derived MSCs (BM-MSCs) and adipose-derived stem cells (ADSCs) in a rat model of HI brain injury. HI models were generated by ligating the left carotid artery of postnatal day 7 Wistar/ST rats and exposing them to 8% hypoxia for 60 min. Bone marrow and adipose tissue were harvested from adult green fluorescent protein transgenic Wistar rats, and cells were isolated and cultured to develop BM-MSCs and ADSCs. At passaging stages 2–3, 1 × 10<sup>5</sup> cells were intravenously injected into the external right jugular vein of the HI rats at 4 or 24 h after hypoxia. Brain damage was evaluated by counting the number of cells positive for active caspase-3 in the entire dentate gyrus. Microglial isotypes and serum cytokines/chemokines were also evaluated. Distribution of each cell type after intravenous injection was investigated pathologically and bio-optically by *ex vivo* imaging (IVIS®) with a fluorescent lipophilic tracer DiR. The mortality rate was higher in the ADSC group compared to the BM-MSC group, in pups injected with cells 4 h after hypoxia. The number of active caspase-3-positive cells significantly decreased in the BM-MSC group, and the percentage of M1 microglia (a proinflammatory isotype) was also lower in the BM-MSC vs control group in the penumbra of the cortex. Moreover, BM-MSC administration increased anti-inflammatory cytokine and growth factor levels, while ADSCs did not. Each injected cell type was mainly distributed in the lungs and liver, but ADSCs remained in the lungs longer. Pathologically, pulmonary embolisms and diffuse alveolar hemorrhages were seen in the ADSC group. These results indicated that injection of allogeneic BM-MSCs ameliorated neonatal HI brain injury, whereas ADSCs induced severe lung hemorrhage and higher mortality.

Keywords: neonatal encephalopathy, regenerative medicine, cytotherapy, M1 microglia, serum chemokine, multiplex, cell distribution

#### INTRODUCTION

Perinatal hypoxic-ischemic (HI) brain injury occurs in 1 in 1,000 live births and remains a main cause of neurological disabilities and death in term infants (1, 2). Therapeutic hypothermia is the only established treatment option; however, its effect is limited (3–5). Meanwhile, cytotherapy has been emerging as a novel therapy for HI. Recently, we demonstrated the beneficial effect of umbilical cord blood mononuclear cells in rat neonatal HI and mouse stroke models (6–8), and autologous umbilical cord blood cells therapy is now at the clinical trials stage(ClinicalTrials.gov: NCT02256618) (9). However, in situations where asphyxiated babies are born, there is a chance of failure of cord blood collection owing to the clinical staff being busy treating the mother and resuscitating the infant. Allogeneic cell transplantation should be considered as a treatment for such asphyxiated infants.

In animal models, neural stem cells have been shown to have a powerful effect on regeneration of damaged brain regions (10, 11). However, ethical issues discourage their clinical use because such treatments require a fetal brain to obtain the neural stem cells. As a result, other cell types, such as bone marrow-derived mesenchymal stem cells (BM-MSC) and adipose tissue-derived stem cells (ADSC), have been investigated and employed as alternatives. MSCs are thought to be a practical cell source, and many studies have demonstrated that they attenuate tissue damage in various inflammation and/or ischemic models (12–17). The beneficial effect of MSC transplantation is related to their potency to differentiate into multiple lineages (18). Their administration has been shown to improve the tissue environment via endocrine/paracrine effects (19, 20) and have immunosuppressive effects (21, 22).

In particular, ADSCs have some advantages over BM-MSC as adipose tissue is easy to collect under local anesthesia (liposuction), can be collected repeatedly (23), and greater numbers can be collected at one time. Moreover, ADSCs proliferate faster than BM-MSCs (24). Despite the numerous reports about the efficacy of ADSCs (25, 26) and their use in clinical trials (27), there is still no report on their use for neonatal HI. On the other hand, there is some debate about the safety of intravenous administration of ADSCs. Recent reports warn of embolism after intravenous ADSC administration (28), and increased coagulation activity after transplantation (29, 30) is thought to be an underlying cause. In addition, cellular distribution after intravenous administration of ADSCs has not been well-investigated, especially in the subacute/chronic phase, raising safety concerns, such as risk of pulmonary embolism. In the current study, we investigated the safety and efficacy of intravenous administration of ADSCs and BM-MSCs in a rat model of neonatal HI brain injury.

#### MATERIALS AND METHODS

#### Animals

This study was carried out in accordance with the Regulations on Animal Experiments in Nagoya University. The protocol was approved by the Institutional Review Board of Animal Experimentation of Nagoya University School of Medicine (Nagoya, Japan; Protocol No.: 24337-2012, 25170-2013, 26128- 2014). Wister/ST rats (SLC Inc., Shizuoka, Japan) were used for the HI model. MSCs were harvested from green fluorescent protein (GFP)-Transgenic (Tg) Wistar rats which were supplied by the National BioResource Project-Rat, Kyoto University (Kyoto, Japan). All rats were maintained under a 12 h light/12 h dark cycle (lights on from 9:00 AM to 9:00 PM) with ad libitum access to food and water. Every effort was made to reduce animal suffering.

#### Hypoxic-Ischemic Brain Injury Animal Model

HI rat models were made according to the method of Rice et al. (31) with minor modification as described in our previous reports (7, 32). On postnatal day 7 (P7), Wistar/ST male and female rat pups were anesthetized with isoflurane and their left common carotid artery was double-ligated with 5-0 surgical silk and cut between the ligatures. The anesthesia time never exceeded 10 min for each pup. After a 1 h rest with dam, they were exposed to 8% hypoxia at 37 C in an incubator for 60 min.

#### Cell Preparation

For preparation of BM-MSCs, 3- to 5-week-old female GFP-Tag Wistar/ST rats were anesthetized with isoflurane and their femurs

**Abbreviations:** ADSC, adipose tissue-derived stem cells; BM-MSC, bone marrowderived mesenchymal stem cells; CA, cornu ammonis; DAPI, 4',6-diamidino-2-phenylindole; DiR, 1,1-dioctadecyl-3,3,3,3-tetramethyl indotricarbocyanine iodide; EDTA, ethylenediaminetetraacetic acid; FBS, fetal bovine serum; GFP, green fluorescent protein; Tg, transgenic; HI, hypoxic-ischemic; Iba1, ionized calcium-binding adapter molecule 1; iNOS, inducible nitric oxide synthase; MEM, modified Eagle's medium; P7, postnatal day 7; PBS, phosphate-buffered saline; DIC, disseminated intravascular coagulation.

and tibias were removed aseptically. Then, heparinized saline was used to flush the marrow shafts using a 23-G needle, and the bone marrow suspension was harvested. After washing with 0.1 mM EDTA-saline, cells were resuspended in 5 mL of Minimal Essential Medium (MEM) alpha (Invitrogen, Carlsbad, CA, USA) with 2% albumin (Japan Blood Products, Tokyo, Japan). Mononuclear cells were isolated with Ficoll <sup>R</sup> -Paque PLUS (GE Healthcare Life Sciences, Uppsala, Sweden). To culture BM-MSCs, mononuclear cells were suspended in 5 mL MEM alpha with 20% FBS (Thermo Fisher Scientific, Waltham, MA, USA), and plated at 4–6 × 10<sup>6</sup> cells per 25-cm<sup>2</sup> flask and incubated at <sup>37</sup>◦C in a humidified atmosphere with 5% CO<sup>2</sup> for 1–2 weeks until the first passage. We selected these plastic-adherent cells as BM-MSCs. BM-MSCs were used for injection after the second or third passage.

ADSCs were also prepared from 3- to 5-week-old female GFP-Tag Wistar/ST Rats. Rats were gently killed by CO<sup>2</sup> asphyxiation, and adipose tissues were obtained from the fatty layer of the subcutaneous tissue. Generally, 2–4 g of adipose tissue was obtained from each rat. Adipose tissue was well-minced in MEM alpha (Gibco <sup>R</sup> ) and digested with 1 mg/mL collagenase type II solution (Invitrogen) with stirring for 1 h at 37◦C. The digested tissue was filtered using a 100-µm cell strainer. Then stromal vascular fraction was precipitated by centrifugation at 1,200 rpm for 5 min at room temperature then washed twice with MEM alpha containing FBS and antibiotics. Stromal vascular fraction cells were seeded (2 × 10<sup>6</sup> cells) in 225-cm<sup>2</sup> T-flasks and cultured in Dulbecco's MEM (Gibco <sup>R</sup> ) containing 20% FBS at 37◦C in a humidified atmosphere with 5% CO<sup>2</sup> and 95% air. Four to Five days later, unattached cells were removed, and the medium changed to Dulbecco's MEM containing 3% FBS. Cells were collected from culture flasks at 90% confluence using 0.05% trypsin-EDTA (Wako, Osaka, Japan) and reseeded at 1,000 cells/cm<sup>2</sup> to ensure optimal proliferation. ADSCs were used for injection after the second or third passage.

#### Intravenous Injection of Cells

Rats were set on an electric warmer plate to maintain proper body temperature and anesthetized with inhaled isoflurane. Then, the skin was cut to expose the right external jugular vein. ADSCs or BM-MSCs were injected slowly into the vein using a 35-G needle; cells were suspended in 0.1 mL phosphate-buffered saline (PBS) and kept on ice until being rewarmed to room temperature just before injection. To evaluate the treatment effect, each cell or vehicle was administered at 24 h after HI. To assess the mortality and pathological findings, cells and vehicle were given at 4 or 24 h after HI.

#### Cell Labeling With DIR

Injected cells were labeled with the fluorescent tracer 1, 1 dioctadecyl-3,3,3,3-tetramethyl indotricarbocyanine iodide (DiR; Caliper Life Sciences, Hopkinton, MA, USA) following to the manufacturer's protocol. Briefly, cells were incubated with DiR for 30 min at 37◦C, centrifuged for 5 min at 1,500 rpm at room temperature, and then rinsed twice with PBS. In all the cases, DiR-labeled cells were suspended in PBS, and 1 × 10<sup>5</sup> cells were injected within 2 h after labeling.

### Ex vivo Imaging and Analysis

To reduce fluorescent noise, all rats used for ex vivo imaging were fed an alfalfa-free diet (D10001, Research Diets Inc., New Brunswick, NJ, USA). DiR-labeled BM-MSCs (n = 18) or ADSCs (n = 18) were intravenously injected into neonatal HI rats 24 h after the hypoxic insult. Of these, three rats from each cell type-treatment group were sacrificed at 1 h, 1 d, 3 d, 7 d, 14 d, and 28 d after injection to collect brain, lungs, heart, liver, spleen, gut, kidney, and bladder for ex vivo imaging. The collected organs were imaged using IVIS <sup>R</sup> Spectrum (Caliper Life Sciences). Filter conditions and illuminations settings for DiR imaging were set an excitation/emission of 710/760 nm, high lamp level, medium binning, filter 1, and 1.0 sec exposure time. Grayscale and fluorescent images of each organ were analyzed using Living Image software version 4.3 (Xenogen). Regions of interest of each organ were automatically drawn over the signals on images and, if necessary, they were manually corrected according to the grayscale image. Quantification was made according to the method of Cho et al. (33) with modification. The distribution of each DiR-labeled cell in each organ was quantified as the average radiant efficiency (total photons/s/cm<sup>2</sup> /steradian) in the irradiance range (µW/cm<sup>2</sup> ): (photons/s/cm<sup>2</sup> /steradian)/(µW/cm<sup>2</sup> ). To reduce variability in measurements, the ratio of the average radiant efficiency of the organs to the background was calculated. The minimum detectable fluorescence required 1 × 10<sup>3</sup> cells as in a previous report (34).

#### Immunohistochemistry

Immunostaining of brain sections with anti-active caspase-3 was performed as previously described (35) with minor modifications. Briefly, rats were anesthetized with pentobarbital (Kyoritsu Seiyaku Co., Tokyo, Japan) and intracardially perfusion-fixed with 0.9% NaCl, followed by 4% paraformaldehyde in PBS. Then, brains were immersion-fixed in 4% paraformaldehyde in PBS at 4◦C for 24 h, dehydrated with a graded series of ethanol and xylene, embedded in paraffin, and cut into 5-µm-thick coronal sections. After deparaffinization and rehydration, antigen retrieval was performed by heating sections for 10 min in 10-mM citrate buffer (pH 6.0). Then, sections were blocked in PBS containing 0.1% Triton and 4% donkey serum and incubated overnight at 4◦C with rabbit anti-active caspase-3 (1:200; BD Pharmingen, Franklin Lakes, NJ, USA). Sections were subsequently incubated with a donkey antirabbit biotinylated secondary antibody (Vector Laboratories, Burlingame, CA, USA) for 1 h at room temperature. Endogenous peroxidase activity was blocked with 3% H2O<sup>2</sup> in PBS for 10 min and then an avidin-biotin-peroxidase complex (Vectastain ABC Elite kit; Vector Laboratories), followed by peroxidase detection for 10 min (0.12 mg/mL 3,3'-diaminobenzidine, 0.01% H2O2, and 0.04% NiCl2). For immunostaining of cells injected into the lungs, rats were euthanized by decapitation, and the lungs were removed and fixed with 4% paraformaldehyde in PBS. Subsequent immunostaining procedures were performed in the same way as for brain sections, except lung sections were incubated with a rabbit anti-GFP (1:200; MBL) primary antibody. To evaluate microglia, two brain sections per pup at

the hippocampal and basal ganglia level were used. After antigen retrieval and blocking of nonspecific binding, sections were incubated with anti-ionized calcium-binding adapter molecule (Iba) 1 (1:100; Wako) and anti-inducible nitric oxide synthase (iNOS; 1:40; Abcam, Cambridge, MA, USA) primary antibodies at 4◦C overnight. Sections were subsequently incubated with Alexa-548 and Alexa-488 for 1 h at room temperature, and mounted with ProLong Gold Antifade reagent containing DAPI (Thermo Fisher Scientific Inc.).

#### Cell Counting

For counting of anti-active caspase-3-positive cells, every 50th section at the hippocampus level (typically 5 sections) were stained using an anti-active caspase-3 primary antibody (1:1000). The hippocampal CA3 and dentate gyrus were outlined under low magnification (40×), and active caspase-3-positive cells in these areas were counted under high magnification (200×) using Stereo Investigator version 10 stereology software (MicroBrightField Europe EK, Magdeburg, Germany). The total number of cells was calculated using the following formula: N = ΣA × P, where N = the total number of cells, ΣA = the sum of the counted number of cells, and P = the inverse of the sampling fraction.

For evaluation of microglial M1 polarization, Iba1- (panmicroglia marker) and iNOS- (M1 microglia marker) positive cells were counted. M1 polarity was calculated by the percentage of iNOS/Iba1 double-positive cells in Iba1-positive cells (36). All positive cells were counted within a 200-µm<sup>2</sup> area in the hippocampus (CA3), basal ganglia, and upper/lower side of the penumbra of the cortex in the two sections.

#### Serum Cytokine, Chemokine, and Growth Factor Analyses

Blood samples were collected from the heart at sacrifice for immunohistochemical evaluation. To obtain serum samples, blood samples were immediately centrifuged and kept on ice until freezing. Serum samples were analyzed by MILLIPLEX <sup>R</sup> Multiplex Assays using Luminex <sup>R</sup> with a rat cytokine/chemokine panel (Merck Millipore, Billerica, MA, USA) according to the manufacture's protocol. The MILLIPLEX <sup>R</sup> plate was read with Luminex MagPix technology. Data was analyzed using xPONENT <sup>R</sup> software (Luminex, Austin, TX, USA).

#### Pathology

ADSCs or BM-MSCs (1 × 10<sup>4</sup> , 1 × 10<sup>5</sup> or 1 × 10<sup>6</sup> cells) were injected intravenously 4 h after HI insult. Organs of interest were excised 15 min after injection without perfusion then fixed in paraformaldehyde, embedded in paraffin, cut into 10 mm sections, and stained with hematoxylin-eosin. Pathological findings were determined by our pathologist (Y Shimoyama).

#### Statistical Analyses

The sample size was decided to be 5 to 10 in each group based on our previous studies (7, 32). Statistical analyses were performed using JMP11.0 software (SAS Institute, Cary, NC, USA). Mortality after injection was compared using a Fisher's exact test. For analyses of immunohistochemistry and serum cytokine/chemokine and growth factor levels, one-way analysis of variance was used, followed by a Dunnett's post-hoc test. Twogroup analyses of organ fluorescence by ex vivo imaging were compared using a Student's t-test. All data are expressed as the mean ± standard error of the mean. A P-value of less than 0.05 was considered statistically significant.

### RESULTS

### Mortality After Administration of ADSC or BM-MSC

To assess the safety of each cytotherapy, ADSCs or BM-MSCs (1 × 10<sup>5</sup> cells/0.1 mL PBS) or vehicle (0.1 mL PBS) were given 4 or 24 h after HI insult. The mortality rate within 24 h after administration was significantly higher in the ADSC group (64%) than the BM-MSC group (6%) when cells were given 4 h after hypoxia exposure (**Table 1**). However, there was no significant difference among ADSC, BM-MSC, and vehicle groups when cells were given 24 h after hypoxia exposure.

### Impact of ADSC and BM-MSC Administration on Apoptosis After HI

Twenty-four hours after HI, P7 rats were injected with 1×10<sup>5</sup> ADSCs (n = 8), BM-MSCs (n = 7), or vehicle (n = 8). The rats were sacrificed 24 h after injection, and the numbers of active caspase-3-positive cells in the CA3 area and entire dentate gyrus were counted. Photomicrographs of representative hippocampal sections are shown in **Figure 1**. The number of active caspase-3-positive cells significantly decreased in CA3 area and dentate gyrus by 76% (P < 0.05) and 59% (P < 0.05), respectively, in the BM-MSC group but not in the ADSC group (**Figure 2**).

TABLE 1 | Mortality within 24 h after administration of MSCs.


FIGURE 1 | Photomicrographs of anti-active caspase-3 staining of the hippocampus. Representative photomicrographs of the hippocampus 48 h after HI insult. Sections are from rats injected with vehicle (PBS), ADSCs, or BM-MSCs; bar = 500µm. Insets show higher magnification; bar = 50 µm.

### Microglial M1 Polarization After Administration of ADSCS and BM-MSCS

To evaluate the impact of ADSCs and BM-MSCs on microglial M1 polarization, we double-stained brain sections with anti-Iba1 and anti-iNOS antibodies. Then, the number of Iba1/iNOS double-positive cells and Iba1-positive/iNOS-negative cells was counted. Representative photomicrographs are shown in **Figure 3A**. The number of Iba1-positive cells in the BM-MSC group tended to decrease in the penumbra of the cortex, hippocampus, and basal ganglia but was not statistically significant (**Figure 3B**); this was not observed in the ADSC group.

In the BM-MSC group, the number of Iba1/iNOS doublepositive cells and the ratio of Iba1/iNOS double-positive to Iba1 positive cells were found to be significantly decreased in the penumbra of the cortex compared to the vehicle group (both, P < 0.05; **Figures 3A,B**); no significant change was found for the ADSC group. The same trend was seen in the hippocampus and basal ganglia. This indicates that intravenous injection of BM-MSCs, but not ADSCs, decreased the polarity of M1 microglia.

### Impact of ADSC and BM-MSC Administration on Serum Cytokines/Chemokines

To evaluate the serological effect of BM-MSCs or ADSCs, serum cytokine/chemokine levels 24 h after injecting ADSCs or BM-MSCs were analyzed by multiplex assay (MILLIPLEX <sup>R</sup> ). All cytokine/chemokine levels measured are listed in **Table 2**. BM-MSC administration significantly increased antiinflammatory cytokine interleukin (IL)-2 level compared to vehicle (**Figure 4A**). And anti-inflammatory cytokine IL-4 also tended to be increased, but not significantly (p = 0.068). Granulocyte colony stimulating factor (G-CSF) levels were also significantly elevated in the BM-MSC group compared with control (**Figure 4B**). BM-MSCs also significantly reduced the levels of several chemotactic chemokines, including CCL2 (monocyte chemoattractant protein-1), CCL3 (macrophage inflammatory protein-1a), CX3CL1 (Fractalkine), CXCL1 (human growth-regulated oncogene/keratinocyte chemoattractant), CXCL2 (monocyte inflammatory protein-2), CXCL3 (lipopolysaccharide-induced CXC chemokine), and CXCL10 (interferon-γ-induced protein-10) [**Figures 4C–I**]. In contrast, BM-MSCs significantly increased three inflammatory cytokines, IL-12p70, IL-17a, and tumor necrosis factor-α (**Figures 4J–L**). ADSC administration, however, made little impact on cytokines/chemokines or growth factor levels in serum (**Figures 4A-L**).

### Time-Course of Distribution and Fate of ADSCS and BM-MSCS After Intravenous Injection

The distribution and fate of ADSCs and BM-MSCs after injection were evaluated serially by ex vivo imaging. **Figure 5A** shows representative pictures of the time-course of distribution in ADSC and BM-MSC groups. Both ADSCs and BM-MSCs mainly distributed in the lungs and liver within 3 d (**Figures 5A,C,D**). After that time, BM-MSC levels gradually reduced in the lungs. On the other hand, ADSCs remained in the lungs longer, even up to 28 d after injection (**Figures 5A,C**). No significant fluorescence was detected in the brain (**Figures 5B,E**) or kidney (**Figure 5F**) in either group at any time point.

### Pathological Findings After Injection of ADSCS Or BM-MSCS 4 H After HI

Gross pathological observation after ADSC or BM-MSC injection revealed much more severe lung hemorrhaging in the ADSC group vs. the BM-MSC group (**Figures 6A–D**). The hemorrhages in the ADSC group were exacerbated as the number of cells injected increased. In rats given ADSC (1 × 10<sup>5</sup> cells; n = 2), the hemorrhages were diffuse and filled the lung surface (**Figure 6A**). In rats given fewer ADSCs (1 × 10<sup>4</sup> cells; n = 2), the hemorrhages were diffuse but not as extensively (**Figure 6B**). On the other hand, rats administered BM-MSCs (1 × 10<sup>5</sup> or 1 × 10<sup>4</sup> cells) showed sparse bleeding (**Figures 6C,D**).

Micropathologically, pulmonary embolism and alveolar hemorrhage were seen in both cell groups. Similarly, rats given ADSCs exhibited more severe pathology than BM-MSC rats,

and severity increased by injected cell number. In the ADSC group (1 × 10<sup>6</sup> cells; n = 2), pulmonary thrombosis by cells with large nuclei and fibrins were seen in some large vessels in the lung (**Figure 6E**) but not in the kidney or liver. This finding was compatible with pulmonary embolism. ADSC rats (1 × 10<sup>5</sup> cells; n = 2) also had diffuse alveolar hemorrhages and fibrin deposition in small vessels; no special findings were seen in other organs. Administration of 1 × 10<sup>5</sup> ADSCs per pup resulted in small vessel embolism and diffuse alveolar hemorrhage (**Figure 6F**). Immunohistopathology of lung tissue using anti-GFP showed that cells with large nuclei filling pulmonary vessels were GFP-positive (**Figures 6G,H**), which were injected cells derived from GFP-Tg rats. In contrast, BM-MSC rats (1 × 10<sup>5</sup> or 1 × 10<sup>6</sup> cells; n = 2) had cells with large nuclei in alveolar vessels, depending on the amount of injected cells, but hemorrhages and fibrin deposition were rare.

### DISCUSSION

In the present study, we showed that intravenous administration of BM-MSCs had a therapeutic effect on HI brain injury in rats that resulted in reduction of apoptotic cells in the hippocampus. BM-MSC administration 24 h after HI decreased serum chemokine levels and increased anti-inflammatory cytokines. In addition, BM-MSCs also significantly decreased proinflammatory M1 microglia. In contrast, ADSC injection did not exhibit any such therapeutic effects but induced severe lung hemorrhaging and pulmonary embolism, leading to high mortality.

Herein, intravenous injection of BM-MSCs reduced apoptosis induced by HI in the CA3 area and dentate gyrus of the hippocampus, the most vulnerable areas to HI insult in premature brain (37). This finding corresponds with those we recently reported using mononuclear cells derived from human umbilical cord blood cells (7) and dedifferentiated fat cells (32). Furthermore, BM-MSCs have been shown to exert a therapeutic effect through different administration routes, including intracranial (38, 39), intracardiac (40), and intranasal (38).

One of the possible mechanisms related to this positive effect was that BM-MSC administration decreased M1 microglia. Recently, it was shown that not only macrophages but also microglia are divided into two types, proinflammatory (M1) and anti-inflammatory [M2] (41, 42). In the present study, BM-MSCs reduced the number of M1 microglia. This microglial change is one of the emerging targets for the treatment of neuronal injury or degenerative diseases (36, 43, 44). In the neonatal HI model, Donega et al. (45) also showed that intranasal administration of BM-MSCs decreased the M1 microglia, in accordance with our previous data showing the same change with intravenous injection of umbilical cord blood cells (8).

To elucidate further BM-MSC therapeutic mechanisms and support the observed microglial change, we also evaluated the impact of ADSC and BM-MSC administration on serum cytokine/chemokine levels in the current study. For the first time in neonatal HI rat models, we demonstrated serological amelioration of chemokines and anti-inflammatory cytokines by administration of BM-MSCs. Several proinflammatory chemokines/cytokines known to activate microglia were markedly decreased in the BM-MSC group, including CCL3, CX3CL1, CXCL1, CXCL2, CXCL3, and CXCL10 (46, 47). Thus, chemokine reduction, especially CX3CL1, is likely one mechanism by which BM-MSCs exert their therapeutic effect and reduce the M1 phenotype of microglia (48, 49). On the other hand, CCL2 levels, which are thought to decrease M1 phenotype, were decreased in our model. However, this chemokine is also known to activate circulating inflammatory monocytes (46). Therefore, its reduction may reflect an immunosuppressive effect of BM-MSC injection (50). Moreover, anti-inflammatory cytokines IL-2 and IL-4 were increased with BM-MSC injection. These anti-inflammatory cytokines are known to change microglial polarity M1 to M2 phenotype and are considered to be promising neuroprotective agents/targets (51). BM-MSC also significantly increased serum granulocyte colony stimulating factor levels. Granulocyte colony stimulating

FIGURE 3 | Impact of BM-MSCs and ADSCs on microglial M1 polarization in the cortex, hippocampus, and basal ganglia. (A) Representative photomicrographs of the affected side cortex stained with anti-Iba-1 (pan-microglia marker) and anti-iNOS (M1 phenotype marker) 24 h after injection. Bar = 50µm. (B) Number of Iba-1/iNOS positive cells, and the ratio of iNOS to Iba-1 in the penumbra of the cortex, hippocampus, and basal ganglia. In the penumbra of the cortex, iNOS-positive cell and the iNOS/Iba-1 ratio were significantly decreased in BM-MSC rats (*n* = 8), but not in ADSC rats (*n* = 7), compared to those in the vehicle group (*n* = 7). These parameters in hippocampus basal ganglia, and the number of Iba-1 in penumbra of the cortex and hippocampus showed the same trend. \**P* < 0.05.

factor is also known as another agent that reduces M1 microglia (52). In adult stroke studies, many clinical trials using granulocyte colony stimulating factor are now undergoing (53). These chemokines/cytokines changes following BM-MSC administration support immunohistochemical findings on microglial status herein (i.e., reduced M1 phenotype).

TABLE 2 | Serum cytokine/ chemokine / growth factor analysis at 24 h after administration of MSCs.


*IL, interleukin; TNF, tumor necrosis factor; INF, interferon; CCL, C-C motif chemokinne ligand; CXCL, C-X-C motif ligand; LIX, LPS-induced CXC chemokine; MCP, monocyte chemoattractant protein; MIP, macrophage inflammatory protein; RANTES, regulated on activation, normal T cell expressed and secreted; CX3CL, C-X3-C motif chemokine; IP, interferon gamma-induced protein; GM-CSF, granulocyte macrophage colony stimulating factor; G-CSF, granulocyte-colony stimulating factor; EGF, epidermal growth factor; VEGF, vascular endothelial growth factor.*

Conversely, inflammatory cytokines, tumor necrosis factorα and IL-12p70 were increased by BM-MSC injection. It has been shown that injection of MSCs alone can increase serum inflammatory cytokines levels (54). As the distribution of BM-MSCs was more systemic than that of ADSCs in the present study, BM-MSCs may be more likely to elevate reaction products in serum, whereas ADSCs remained in the lungs and induced local reactions.

The most amazing finding in the present study was that ADSC administration did not elicit any therapeutic effects (no change in apoptosis, microglial polarity, or serum chemokine/cytokine levels) but increased mortality instead. Pathologically, injection of 1 × 10<sup>6</sup> ADSCs resulted in pulmonary embolisms with local inflammatory findings, and the embolisms contained many GFP-positive ADSCs. Even with a lower dose, (1 × 10<sup>5</sup> cells) ADSC injection resulted in alveolar hemorrhage. There are some reports of MSCs having procoagulant activity when intravenously injected (29, 30, 55). However, it is unknown why ADSCs are more likely to cause such a response. Shiratsuki et al. (30) reported that ADSCs, but not BM-MSCs markedly increased prothrombin time, indicating that ADSCs potentially enhance procoagulation activity. In addition, we showed that ADSCs remained in the lungs longer than BM-MSCs after injection. A longer stay in the lungs can exacerbate the negative features of ADSCs action. In the present study, severe pulmonary embolisms were seen but were not present in the kidneys. If the procoagulant effect of ADSCs is systemic, like disseminated intravascular coagulation (DIC), similar findings should also be seen in the kidneys. Therefore, the hypercoagulable condition is thought to be due to a local reaction. The longer time spent by ADSCs in the lung may partially explain the local reactions observed therein.

One possible reason why ADSCs remain longer in the lungs than BM-MSCs may be due to different expression of cell adhesion molecules. Yang et al. (56) reported that BM-MSCs express CD106 (vascular cell adhesion molecule-1), while ADSCs do not. Vascular cell adhesion molecule-1 plays an important role in cellular adhesion to vessels (57). Intravenously injected cells first encounter blood capillaries in the lungs. Therefore, the lack of such an adhesion molecule may cause a "rough landing" on the capillaries, resulting in inflammation and embolism.

CCL3 (D), CX3CL1 (E), CXCL1 (F), CXCL2 (G), CXCL3 (H), and CXCL10 (I). On the other hand, levels of inflammatory cytokines IL-12p70 (J), IL-17a (K), and tumor necrosis factor (L) were increased by BM-MSC administration, but not ADSC. \**P* < 0.05.

Moreover, another previous report revealed that allogenic BM-MSCs upregulate urokinase plasminogen activator expression in a mouse model of pulmonary embolism (58). Thus, BM-MSCs may inherently prevent emboli.

There are two limitations in the present study. One limitation is that there is no evaluation of M2 microglia. Considering the cytokine/chemokine result in the present study and previous publications (8, 45), changing microglial polarity from M1 into M2 is most plausible, but it is not shown in the present study. The other limitation is that each cell type was injected intravenously using simple preparations (i.e., suspended in 0.1 mL of PBS without any measures). A safe method of ADSC injection

has not yet been fully investigated, but there are reports on countermeasures against embolism. For example, Yukawa et al. (59) showed that co-administration of an antithrombin agent prevented lung entrapment of ADSCs in a rodent model. As another countermeasure, cell culture methods may be able to improve development of embolisms. Our low serum-cultured ADSCs (23) have been shown to effectively ameliorate kidney disease in a rodent model via intravenous administration (20). In our preliminary experiments, low serum-cultured ADSCs could be administered to the present HI model as safely as BM-MSCs without adding antithrombin agent (**Supplemental Table**).

## CONCLUSION

Intravenous injection of allogeneic BM-MSCs, but not ADSCs, ameliorated neonatal rat HI injury by reducing M1 microglia and suppressing expression of inflammatory cytokines/chemokines. In contrast, administration of ADSCs induced severe lung hemorrhage and higher mortality.

### AUTHOR CONTRIBUTIONS

YuS, YK, TS, TK, and AM were actively involved in animal experiments. SS supported cell preparation. YuS, YSat, SM, and TY conceptualized and designed the study. YuS, YSat, AH, SM, MT, TY, and MH interpreted the data. YShi performed pathological evaluation. YuS drafted the initial manuscript, and YSat, MT, and MH critically reviewed the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

### FUNDING

This work was supported in part by grants-in-aid for scientific research from the Ministry of Education, Science, Culture,

### REFERENCES


Sports, and Technology of Japan and from the Japan Society for the Promotion of Science (No. 25860908).

### ACKNOWLEDGMENTS

We are thankful to the National BioResource Project-Rat (http://www.anim.med.kyoto-u.ac.jp/NBR/) for providing rat strains (GFP-Tag Wistar/ST). The authors would like to thank Enago (www.enago.jp) for the English language review. We are also grateful for the skillful technical assistance of Ms. Kimi Watanabe, Ms. Eiko Aoki, Ms. Tokiko Nishino, Ms. Azusa Okamoto, and Ms. Tomoko Yamaguchi.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00757/full#supplementary-material


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

Copyright © 2018 Sugiyama, Sato, Kitase, Suzuki, Kondo, Mikrogeorgiou, Horinouchi, Maruyama, Shimoyama, Tsuji, Suzuki, Yamamoto and Hayakawa. 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.

# Administration of Bone Marrow-Derived Mononuclear Cells Contributed to the Reduction of Hypoxic-Ischemic Brain Injury in Neonatal Rats

Yoshiaki Sato<sup>1</sup> \* † , Kazuto Ueda1†, Taiki Kondo1†, Tetsuo Hattori <sup>1</sup> , Alkisti Mikrogeorgiou<sup>1</sup> , Yuichiro Sugiyama<sup>1</sup> , Toshihiko Suzuki <sup>1</sup> , Michiro Yamamoto<sup>2</sup> , Hitoshi Hirata<sup>2</sup> , Akihiro Hirakawa<sup>3</sup> , Keiko Nakanishi <sup>4</sup> , Masahiro Tsuji <sup>5</sup> and Masahiro Hayakawa<sup>1</sup>

#### Edited by:

Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Christiane Charriaut-Marlangue, Institut National de la Santé et de la Recherche Médicale (INSERM), France Tania Fowke, University of Auckland, New Zealand

#### \*Correspondence:

Yoshiaki Sato yoshiaki@med.nagoya-u.ac.jp

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 07 December 2017 Accepted: 02 November 2018 Published: 30 November 2018

#### Citation:

Sato Y, Ueda K, Kondo T, Hattori T, Mikrogeorgiou A, Sugiyama Y, Suzuki T, Yamamoto M, Hirata H, Hirakawa A, Nakanishi K, Tsuji M and Hayakawa M (2018) Administration of Bone Marrow-Derived Mononuclear Cells Contributed to the Reduction of Hypoxic-Ischemic Brain Injury in Neonatal Rats. Front. Neurol. 9:987. doi: 10.3389/fneur.2018.00987 <sup>1</sup> Division of Neonatology, Center for Maternal-Neonatal Care, Nagoya University Hospital, Nagoya, Japan, <sup>2</sup> Department of Hand Surgery, Nagoya University Graduate School of Medicine, Nagoya, Japan, <sup>3</sup> Department of Biostatistics and Bioinformatics, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan, <sup>4</sup> Department of Perinatology, Aichi Human Service Center, Institute for Developmental Research, Aichi, Japan, <sup>5</sup> Department of Regenerative Medicine and Tissue Engineering, National Cerebral and Cardiovascular Center, Osaka, Japan

#### Background/Objective: Perinatal hypoxic-ischemia (HI) causes neonatal death and permanent neurological deficits. Cell therapy using various cell sources has been recently identified as a novel therapy for perinatal HI. Among the available types of cell sources, bone marrow-derived mononuclear cells (BMMNCs) have unique features for clinical application. For example, stem cells can be collected after admission, thus enabling us to perform autologous transplantation. This study aimed to investigate whether the administration of BMMNCs ameliorated HI brain injury in a neonatal rat model.

Methods: Seven-day-old rats underwent left carotid artery ligation and were exposed to 8% oxygen for 60 min. BMMNCs were collected from the femurs and tibias of juvenile rats using the Ficoll–Hypaque technique and injected intravenously 24 h after the insult (1 × 10<sup>5</sup> cells). Active caspase-3, as an apoptosis marker, and ED1, as an activated microglia/macrophage marker, were evaluated immunohistochemically 48 h after the insult (vehicle, n = 9; BMMNC, n = 10). Behavioral assessments using the rotarod treadmill, gait analysis, and active avoidance tests were initiated 3 weeks after the insult (sham, n = 9, vehicle, n = 8; BMMNC, n = 8). After these behavioral tests (6 weeks after the insult), we evaluated the volumes of their hippocampi, cortices, thalami, striata, and globus pallidus.

Results: The mean cell densities of the sum of four parts that were positive for active caspase-3 significantly decreased in the BMMNC group (p < 0.05), whereas in the hippocampi, cortices, thalami, and striata cell densities decreased by 42, 60, 56, and 47%, respectively, although statistical significance was not attained. The number of ED1 positive cells for the sum of the four parts also significantly decreased in the BMMNC group compared to the vehicle group (p < 0.05), whereas in each of the four parts the decrease was 35, 39, 47, and 36%, respectively, although statistical significance was not attained. In gait analysis, the BMMNC normalized the contact area of the affected hind paw widened by HI. The volumes of the affected striata and globus pallidus were significantly larger in the BMMNC group than in the control group.

Conclusion: These results indicated that the injection of BMMNCs ameliorated HI brain injury in a neonatal rat model.

Keywords: infant, encephalopathy, cell therapy, regenerative medicine, cerebral palsy, mental retardation

### INTRODUCTION

Neonatal encephalopathy (NE) is a neurological syndrome that presents with clinical features consistent with that of brain disorders. Its most relevant clinical features are decreased consciousness and respiratory depression, abnormal strength and muscle tone, impaired feeding, and seizures (1). In developed countries, the incidence of NE is ∼1–6 per 1,000 live births (2). NE is frequently associated with acute hypoxic-ischemic insults, and 50–80% of cases are considered to have hypoxic-ischemic encephalopathy (HIE) (1). Despite the developments in perinatal medicine, perinatal asphyxia remains an important cause of neonatal death and permanent neurological deficits (3, 4), which currently has no effective treatments other than hypothermia. Previous randomized trials have shown that hypothermia reduces the risk of death or disability in infants with HIE (5). However, it has not been found effective in cases of severe HIE (6).

Stem cell therapy is expected to have applications in the treatment of central nervous system diseases (7), and several types of stem cells constitute potential sources of cell therapy for future clinical applications. We have previously demonstrated that intracerebroventricular injection consisting of neural stem/progenitor cells (NSPCs) with chondroitinase ABC, which digests glycosaminoglycan chains on chondroitin sulfate proteoglycans, reduces brain injury in a rat model of neonatal hypoxic-ischemia (HI) (8, 9). Additionally, Ji et al. (10) reported the effects of intranasal treatment with engrafted NSPCs. However, ethical and safety concerns hinder the use of NSPCs derived from the brain of a human fetus in clinical practice (11). Ethical considerations can be avoided through the use of stem cells originating from non-neural tissues such as umbilical cord blood cells (UCBCs), which are readily available and can be exploited for autologous transplantations. We have recently confirmed the beneficial effects of UCBCderived mononuclear cells in a rat model of neonatal HI and in a mouse model of stroke (12–14). The subsequent clinical trials are currently under way (ClinicalTrials.gov: NCT02256618). Although autologous UCBCs are a promising source for stem cell therapy against NE, their collection may be difficult during emergencies, such as precipitous delivery.

Bone marrow contains populations of multipotent precursors, including mesenchymal cells, that can differentiate into multiple cell types (15). The ability of bone marrow cells to differentiate into neurons and glia was recently demonstrated (16), as was their ability to cross the blood brain barrier and preferentially enter the brain upon intravenous infusion (17). Bone marrowderived mononuclear cells (BMMNCs) have been reported to reduce neurological impairments in a rat model of adult ischemic stroke (18). BMMNCs constitute a promising source for cellular therapy, since they can be rapidly collected and isolated from bone marrow after admission. Moreover, they are enriched with stem cells, allowing autologous application, and the feasibility and safety of their harvest and reinfusion has been previously demonstrated by a clinical trial in acute stroke patients (19). However, to our knowledge, no studies have studied the effect of BMMNCs in a rat model of neonatal HI. The present study investigated the effects of BMMNCs administration to neonatal HI rats.

### MATERIALS AND METHODS

#### Animals

All animal experimental protocols in the present study were approved by the Institutional Review Board of Nagoya University School of Medicine (Nagoya, Aichi Prefecture, Japan; permit No.: 23181, 24337, 25170, 26128). Only male Wistar/ST rat pups were acquired from Japan SLC Inc. (Shizuoka, Japan) to avoid the possible behavioral effects of the sexual cycle in female rats (20– 22). The rats were housed under a 12-h light/dark cycle (lights were kept on from 8:00 a.m. to 8:00 p.m.) with ad libitum access to food and water. The animal room and experimental space were maintained at 23◦C.

### BMMNC Preparation

For each experiment, bone marrow cells were collected from the femurs and tibias of one juvenile Wistar female rat at postnatal day 30 (P30). The bone ends were cut, and the marrow was flushed out using 5 ml of saline heparin solution (200 U heparin/100 ml saline) with a 22 or 23-gauge needle. After washing with 10 ml saline supplemented with 0.1 mM EDTA, the cells were suspended with DMEM/F-12 (Thermo Fisher Scientific, Waltham, Massachusetts, USA) with 2% albumin. Mononuclear cells were isolated using the Ficoll–Hypaque technique (Ficoll-Paque PLUS; GE Healthcare Bio-Sciences AB, Björkgatan, Uppsala, Sweden) and suspended in DMEM/F-12 and 1% fetal bovine serum (FBS) at a concentration of 1 × 10<sup>6</sup> cells/ml. The cells were administered immediately after collection. Three batches of cells were used, since the experiments were performed on three different dates: two for acute injury evaluation and one for behavioral tests.

#### HI Insult and BMMNC Administration

Hypoxic-ischemic brain damage was induced on P7 rats according to the method of Rice et al. (23) with minor modifications as described previously (13). Briefly, each pup was anesthetized with isoflurane inhalation (5% for induction and 1–2% for maintenance). The left carotid artery was doubly ligated and incised between the ligatures. After a 1-h rest with a dam, the pups were exposed to a hypoxic environment (8% <sup>O</sup><sup>2</sup> and 92% N2, at 37◦C for 60 min), after which they were returned to the dam in an animal room maintained at 23◦C. The pups in the treatment group (BMMNC group; n = 10 for acute injury evaluation and n = 10 for behavioral tests) were injected with BMMNCs (1 × 10<sup>5</sup> cells/0.1 ml) intravenously via the right external jugular vein under isoflurane inhalation anesthesia 24 h after the insult. A vehicle group (n = 10 for acute injury evaluation and n = 10 for behavioral tests) underwent ligation of the left carotid artery and hypoxia in the same manner and received an equivalent volume of DMEM/F-12 and 1% FBS. The sham group (n = 3 for acute injury evaluation and n = 9 for behavioral tests) underwent neither left carotid artery ligation nor hypoxia. In order to avoid hypothermia, the pups were kept on a water-bath (set to 37◦C) before and after the surgery and on a hot plate (set to 37◦C) during the surgical procedure.

### Histological and Immunohistochemical Procedures

Histological and immunohistochemical procedures were performed as previously described (24) with minor modifications. Briefly, rats were deeply anesthetized and intracardially perfusion-fixed with 0.9% NaCl followed by 4% paraformaldehyde in phosphate-buffered saline (PBS) 48 h after the insult. The brains were excised and immersion-fixed in the same solution at 4◦C overnight, after which they were dehydrated with a graded series of ethanol and xylene, embedded in paraffin, and cut into 5-µm-thick coronal sections. We then performed antigen retrieval by heating the sections for 10 min in 10-mM citrate buffer (pH 6.0) after deparaffinization and rehydration. Non-specific binding was blocked using 4% donkey serum in PBS. Then, sections were incubated overnight at 4◦C with rabbit anti-active caspase-3 (product number 559565; dilution, 1:200; BD Pharmingen, Franklin Lakes, NJ, USA), mouse anti-ED1 (product number MAB1435; dilution 1:300; Merck Millipore, Darmstadt, Germany), or anti-MAP2 (product number MAB3418; dilution 1:400; Merck Millipore) in PBS with 0.1% Triton. Subsequently, the sections were incubated with sufficient biotinylated secondary antibodies (Vector Laboratories, Burlingame, CA, USA) for 1 h at room temperature. Endogenous peroxidase activity was blocked with 3% H2O<sup>2</sup> in PBS for 10 min and then with avidinbiotin-peroxidase complex (Vectastain ABC Elite kit; Vector Laboratories). Peroxidase detection was then performed for 10 min (0.12 mg/mL 3,3-diaminobenzidine, 0.01% H2O2, and 0.04% NiCl2).

#### Cell Counting for Acute Injury Biomarkers

Cells positive for active caspase-3 and ED1 were counted in every 50th section throughout the cortex, striatum-pallidum, thalamus, and hippocampus, resulting in a total of four sections per animal, using the Stereo Investigator version 10 stereology software (MicroBrightField Europe EK, Magdeburg, Germany). A square (500 × 500µm) was placed on the cortex, the striatumpallidum, and thalamus, and the hippocampus was outlined (**Figure 1**). The area was measured in each section. The cortical ROI was placed in the lateral cortex to avoid areas of infarction. The positive cells were counted under high magnification (×400) inside each square and outlined hippocampus. Cell counts were expressed as densities. The evaluation was blinded regarding group allocation.

#### Behavioral Tests

All behavioral tests and evaluations were blinded regarding group allocation.

#### Rotarod Treadmill

A rotarod treadmill was used to evaluate the motor performance and coordination at P24 and P25. Each rat was placed on a rotating rod (Med Associates Inc., St. Albans, VT, USA), accelerating at 4–40 rpm over 5 min, and the time at which the rat fell from the rod was recorded (maximum 300 s). For 2 consecutive days, the tests were performed twice a day with a 2-h interval.

#### Gait Analysis

Gait assessment was performed at P31 using the CatWalk quantitative gait analysis system (Noldus Information Technology, Wageningen, The Netherlands) as previously described (13). The experimental rats ran across a glass walkway transversely, and the runs were recorded by a camera positioned below. If an animal failed to complete a run within 5 s, walked backward, or reared during the run, the process was repeated with each rat, and the average of three runs was calculated. The glass walkway was illuminated with beams of light in the dark atmosphere as the animals' paws could reflect light as they touched the glass floor. To calculate the paw-related parameters, each paw was labeled on the recorded video. In the experiment, the contact area (area of paw print) and

FIGURE 1 | Cell-counting area in the cortex, striatum-pallidum, thalamus, and hippocampus. A square (500 × 500µm) was placed on the cortex (A), the striatum-pallidum (B), and the thalamus (C). The hippocampus was outlined (D). Bar = 500µm.

maximal intensity (the maximal intensity of each paw in the run) were measured.

#### Active Avoidance Test

The active avoidance test was performed from P35 to P38, following the method described by Ichinohashi et al. (25) using the same equipment. Each rat underwent 20 daily sessions of a shuttle avoidance test for 4 consecutive days. The test was conducted in an automated shuttle box (Med Associates Inc., St. Albans, VT, USA), which was divided into two compartments with independently electrified stainless steel bars as a floor. Each session consisted of presenting a buzzer tone and light stimulation (conditioning stimulus, CS) and an electric shock (unconditioned stimulus, US). Both the CS and the US were presented for 5 s, and the US consisted of a positive half-wave constant current of 0.5 mA. Upon presentation of the CS, the rat could avoid the US by escaping to the other compartment of the shuttle box, thus switching off the CS. The interval between each trial varied from 10 to 90 s (30 s on average). The parameters were analyzed using the MED-PC IV software (Med Associates Inc.,). The avoidance proportion, i.e., the number of sessions in which the rat successfully switched off the alert and avoided electric shock, was evaluated each day.

#### Volume Measurement

After the behavioral tests, rats were deeply anesthetized, intracardially perfusion-fixed, and had their brains excised at P43 using the same histological procedures. The remaining volumes of the hippocampus, thalamus, cortex, striatum, and globus pallidus, were evaluated by staining every 100th section from the whole cerebrum (7–9 sections) with anti-MAP2 antibody. The volumes of each section were calculated according to the Cavalieri principle (Stereo Investigator Ver.10) using the following formula: V = PA × P × T, where V = the total volume, PA = the sum of area measurements, P = the inverse of the sampling fraction, and T = the section thickness. Section areas with missing regions were regarded as zero. The evaluation was blinded regarding group allocation.

#### Statistical Analysis

All continuous data were presented as mean ± standard error. The mean body weight gain and acute injury biomarker data in each region were compared between two groups (vehicle vs. BMMNC) using Student's t-test. The survival rate, body weight gain, behavioral data, and volume measurement for each region were compared among three groups (sham, vehicle, and BMMNC) using the Dunnett's test. In addition, a two-way ANOVA with the three experimental groups (sham, vehicle, and BMMNC) and four regions as the two independent variables was used to analyze the data pertaining to acute injury biomarker and volume measurement for the sum of the four regions. For acute injury biomarkers, only the vehicle and BMMNC groups were compared since the data of the sham group was prepared only for reference. A p-value < 0.05 was considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics 24.0.

### RESULTS

#### Safety Assessments

At 48 h post-insult 49 rats survived. Three rats died at P7, immediately after hypoxia and at P8 probably due to deep anesthesia. Twenty-two rats were sacrificed at P9 for immunohistological evaluation and 27 rats were maintained for behavioral evaluations. No significant differences in the survival rates were observed among the three groups at P18: 100% (9/9) survived in the sham group, 89% (8/9) in vehicle group, and 100% (9/9) in the BMMNC group. At 6 weeks the survival rates were 100% (9/9) in the sham group, 89% (8/9) in vehicle group, and 89% (8/9) in the BMMNC group.

In the vehicle and BMMNC groups one rat died at P17 and one at P21, respectively, which was probably due to HI-derived debilitation.

From P8 to P13, rats that received the HI insult (i.e., those in the vehicle and BMMNC groups), had a lower body weight gain compared to those in the sham group. After injection,

the body weight gain did not differ between the BMMNC and vehicle groups (**Figures 2A,B**). From P25 no differences were observed among the three groups regarding the body weight gain (**Figure 2B**).

#### Impact of BMMNC on the Expression of Acute Injury Biomarkers After HI

Twenty-four hours after BMMNCs administration (i.e., 48 h after the insult), active caspase-3 and ED1 were evaluated immunohistologically as an apoptosis and an activated microglia/macrophage marker, respectively, in the hippocampi, cortices, thalami, and striata (**Figure 1**). Brains without HI injury (n = 3) had very few/none positive cells in both markers: active caspase-3; 314 ± 15 cells/mm<sup>3</sup> in the hippocampus, 67 ± 67 cells/mm<sup>3</sup> in the cortex, 67 ± 67 cells/mm<sup>3</sup> in the thalamus, and 133 ± 67 cells/mm<sup>3</sup> in the striatum-pallidum, ED1; 2,349 ± 316 cells/mm<sup>3</sup> in the hippocampus, 0 ± 0 cells/mm<sup>3</sup> in the cortex, 867 ± 133 cells/mm<sup>3</sup> in the thalamus, and 1,133 ± 742 cells/mm<sup>3</sup> in the striatum-pallidum.

Representative photomicrographs of active caspase-3-positive cells in the hippocampus (**Figures 3A–C**), cortex (**Figures 3D–F**), thalamus (**Figures 3G–I**), and striatum-pallidum (**Figures 3J–L**) are shown, as well as representative photomicrographs of ED1-positive cells in the hippocampus (**Figures 4A–C**), cortex (**Figures 4D–F**), thalamus (**Figures 4G–I**), and striatum-pallidum (**Figures 4J–L**) are shown.

In the BMMNC group, a significant decrease was observed in the mean cell density for the sum of the four parts that were positive for active caspase-3 (p < 0.05, two-way ANOVA model

FIGURE 6 | Rotarod treadmill. Rotarod treadmill test was performed on P24–25. The endurance time was compared among all three groups. Rats in the BMMNC group and vehicle group fell down significantly earlier than in those in the sham group from the first to the third trial (\*\*p < 0.01, Dunnett's test). The endurance time was not significantly different between the BMMNC and vehicle groups in any trials. Open squares and dashed line: sham, n = 9; open circles and dotted line: vehicle, n = 8; closed circles and solid line: BMMNC, n = 8.

FIGURE 5 | Effect of BMMNC on the expression of acute injury biomarkers for apoptosis [active caspase-3, (A–D)] and activated microglia/macrophage [ED1, (E–H)]. The number of marker-positive cells was counted in the hippocampus, cortex, thalamus, and striatum-pallidum. The number of active caspase-3-positive cells for the sum of the four parts was significantly lower in the BMMNC group (n = 10) than in the vehicle group [n = 9, (A–D), two-way ANOVA]. The number of ED1-positive cells for the sum of the four parts was also significantly lower in the BMMNC group [(E–H), two-way ANOVA]. Data are presented as mean ± standard error of the mean.

including the two main effects of treatments and regions). On the other hand, in the hippocampi, cortices, thalami, and striata the mean cell density decreased by 42, 60, 56, and 47%, respectively, although the difference did not attain statistical significance (**Figures 5A–D**).

The number of ED1 positive cells for the sum of the four parts also decreased significantly in the BMMNC group compared with the vehicle group (p < 0.05, two-way ANOVA model including the two main effects of treatments and regions). The number of ED1 positive cells in the hippocampi, cortices, thalami, and striata decreased by 35, 39, 47, and 36%, respectively, in the BMMNC group, although the difference did not attain statistical significance (**Figures 5E–H**).

#### Impact of BMMNC on Behavior After HI

Two rats were excluded due to HI-derived debilitation and death.

#### Rotarod Treadmill

Motor coordination and motor learning were measured using a rotarod treadmill, 17–18 days after the insult (P24–25). The endurance times were shorter in the BMMNC and vehicle groups than those in the sham group in the first three trials. In the fourth trial (the second trial on day 2), no significant differences were observed among the three groups. Overall, no significant differences were observed between the BMMNC and vehicle groups during all trials (**Figure 6**).

#### Gait Analysis

Motor deficits were evaluated 24 days after the insult (P31) through gait analysis using the CatWalk system. No significant differences were observed among the three groups regarding the maximal intensity of each running paw in either hind or fore paws (**Figures 7A–B**), or the contact areas of the fore paws (**Figure 7C**). In contrast, the contact areas of the hind paws were

significantly larger in the vehicle group than in the sham group, but BMMNC normalized the contact area of the affected hind paw widened by HI (**Figure 7D**).

#### Active Avoidance Test

An active avoidance test was performed 28–31 days after the insult (P35–38). The mean avoidance proportion of each group was calculated for 4 consecutive days. The avoidance rates increased with time in all groups, although there were no significant differences among the three groups throughout this period (**Figure 8**).

#### Impact of BMMNC on Brain Volume After HI

The volumetric assessment was performed for the hippocampus, thalamus, cortex, striatum, and globus pallidus to assess the absolute tissue loss after HI. Sections throughout the whole cerebrum at P43 were evaluated after the behavioral tests. Three rats were excluded due to either HI-derived debilitation and death or sampling failure.

Representative photomicrographs stained for MAP2 are shown in **Figures 9A–C**. Two-way ANOVA showed a statistically significant treatment effect of BMMNC on brain volumes in the affected side after HI (p < 0.05). In the analyses of individual regions, the volume of every region was reduced in the vehicle group (**Figures 9D–H**), and the tissue volumes of ipsilateral

performed on P35–38. The avoidance rates increased with time in all groups. There were no significant differences among all three groups on any day (Dunnett's test). Open squares and dashed line: sham, n = 9; open circles and dotted line: vehicle, n = 8; closed circles and solid line: BMMNC, n = 8. Data represent the mean ± standard error of the mean.

striatum and globus pallidus were significantly larger (reduced tissue loss after HI) in the BMMNC group than in the vehicle group (**Figures 9D,E**). The volumes of hippocampus (**Figure 9F**), thalamus (**Figure 9G**), and cortex (**Figure 9H**) in the affected side appeared to be larger but they were not statistically significant (p = 0.11, 0.21, and 0.06, respectively).

#### DISCUSSION

In the present study, we demonstrated the safety of intravenous BMMNC administration and its therapeutic effect in reducing the expression of various acute injury biomarkers in the brain using a neonatal HIE rat model. We further showed improvements in the brain infarct volume and limb paralysis.

First, we showed that a single intravenous injection of BMMNC did not cause death to rat pups after HI insult. Cell therapy with BMMNC administration has been gradually developed for several diseases (26–28), although there were few reports regarding the safety of BMMNC intravenous administration to brain-injured rodents during the neonatal period. We previously reported intravenous deliveries of BMMNC with neonatal mice (29). We showed that the BMMNC administrated intravenously was relatively less confined to the lung, which can lead to serious complications, compared with mesenchymal stem cells. In the previous study, we focused on only the acute phase and did not further evaluate the safety. However, the present study demonstrated the safety of BMMNC administration from acute to chronic phase in neonatal rats. BMMNC can be further developed for clinical applications in neonatal HIE. BMMNC administration for cerebral palsy has been evaluated in some clinical trials (30, 31). It appeared safe for human use, but such data were mainly based on patients from early childhood to adolescence but not in the neonatal period. Here, we demonstrated a good survival rate for longer period as well as just after the administration, which supports the safety of BMMNC administration during the neonatal period.

Second, we showed that BMMNC administration attenuated the expression of several acute injury biomarkers. Active caspase-3 is known to play an important role in the apoptosis pathway (32, 33). Once neuronal damage occurs, activated microglia accumulate in the ischemic core and express ED1 (34). Franco et al. (35) has already reported that BMMNC suppresses apoptosis and microglia activation in adult rats. They showed a significant decrease of active caspase-3 positive cells and ED1 positive cells in the ischemic brain after BMMNC treatment. To the best of our knowledge, our study is the first to report anti-apoptotic effects of BMMNC in a neonatal HI rat model. It suggested that BMMNC exhibited therapeutic effects by preventing the acute inflammatory response for newborns as well as for adults.

Finally, we confirmed the therapeutic effects of BMMNC according to histological and motor functional aspects. BMMNC mitigated the loss of brain tissue after HI and reduced the RH/LH print area ratio, reflecting an improvement in gait. Previous studies have reported its beneficial effects on histological and behavioral outcomes using adult rodents (36–38), although

ipsilateral striatum (D) and globus pallidus (E) were significantly larger in the BMMNC group than in the vehicle group. No significant difference was found between the BMMNC and vehicle groups regarding the volumes of hippocampus (F), thalamus (G), and cortex (H). Data represent the mean ± standard error of the mean (sham, n = 9; vehicle, n = 8; BMMNC, n = 7) \*p < 0.05, Dunnett's test.

few studies reported this effect in newborn rats. Brenneman et al. reported that administration of autologous BMMNC significantly reduced the brain infarct volume in young and middle-aged rats (37). Vahidy et al. (39) conducted a systematic review and meta-analysis and reported that BMMNC had a beneficial effect on histological outcomes in animal ischemic stroke models. Clinical trials have already been performed, but their participants are limited to adults (40). Our study suggested the clinical benefits and application of BMMNC in newborns.

Several publications are currently available on stem cell therapy in the developing brain; Although most publications have reported a positive effect of that treatment (12, 13, 41– 43), Dalous et al. (44) have shown that an intraperitoneal injection consisting of human umbilical cord blood mononuclear cells increased the size of the brain lesion in an animal model of excitotoxic brain injury by revealing that the cells increased inflammatory cytokines, which were associated with the aggravation of the lesion. Since the cell administration can, by itself, increase the levels of inflammatory cytokines, it is difficult to interpret how that increase affects the lesion (45). Intravenous infusion of mesenchymal stem cells is known to induce an inflammatory response. We have previously shown that the use of mesenchymal stem cells increased the levels of several inflammatory cytokines, including tumor necrosis factorα, but still achieved a treatment effect (46). The same study revealed no exacerbation when administration was performed through an intravenous injection. Taken together, these results suggest that the response of cytokines and their effect on the lesion vary with the type of brain injury, the type of stem cells, and/or the administration route.

There are some limitations in the present study. There were no significant differences in some behavioral evaluations, i.e., rotarod treadmill and active avoidance. Iihoshi et al. (18) has reported the improvement of motor function in the treadmill test using a cerebral ischemic adult rat model. Their injected dose (1.0 × 10<sup>7</sup> cells/rat) was higher than ours even after considering the difference of body weight between adult and neonatal rats. In studies using other types of stem cells with the neonatal HI model, de Paula et al. showed that the treatment effect of the stem cells was dose-dependent (47), and van Velthoven et al. demonstrated that repeated administration exerted a better outcome (48). There is a possibility that our dose and/or times might not be enough to improve behavioral outcomes, including motor and cognitive functions. Some reports have demonstrated the therapeutic effect in the cylinder test, a test for locomotor function (14, 41), which was not performed in the present study. The beneficial effects of BMMNC administration may not appear in motor coordination and cognitive function but be limited to motor function. Further analysis is required to evaluate its mechanism and long-term effect. Another limitation of the present study was the fact that we did not evaluate the synergistic effect of BMMNCs and hypothermia. Since hypothermia is an established, standard therapy for moderate to severe HIE, BMMNCs monotherapy without concomitant hypothermia is unlikely to be performed in clinical practice. Evaluating the combination therapy is therefore necessary and should be addressed by further studies, as well as different types of stem cells (49).

The present study clarified the safety of intravenous BMMNC administration and its therapeutic efficacy in a neonatal HI rat model. Transplantation of autologous stem cells has multiple advantages; ethical issues and possible immune responses associated with allogenic transplantation can be avoided. Unlike UCBCs, BMMNC can be collected even after admitting/transfer to a neonatal intensive care unit. BMMNC might be a good

#### REFERENCES


candidate to treat NE. Further studies using different protocols (e.g., various doses, repeated administration, and combination with hypothermia or some other treatments) are needed to elucidate a more detailed mechanism.

#### CONCLUSIONS

A single intravenous injection of BMMNC 24 h after HI produced morphological and functional improvement. Our findings confirmed that the intravenous injection of BMMNC can be a novel treatment for HI brain injury.

### AUTHOR CONTRIBUTIONS

YoS, TK, AM, YuS, TS, and TH were actively involved in experiments. MY and HH were involved in behavioral experiments. YoS, TK, KN, MT, and MH conceptualized and designed this study. YoS, TK, KN, and MT interpreted the data. TK and KU drafted the initial manuscript, and YoS, AH, KN, MT, and MH critically reviewed the manuscript. AH conducted the statistical analysis. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

#### FUNDING

This work was supported by the JSPS KAKENHI (grant No. 23791220).

#### ACKNOWLEDGMENTS

We are grateful to Ms. Kimi Watanabe, Ms. Eiko Aoki, Ms. Tokiko Nishino, Ms. Azusa Okamoto, Ms. Yui Kanazawa, and Ms. Tomoko Yamaguchi for the skillful technical assistance.


hypoxic-ischemic brain injury in neonatal rats. Dev Neurosci. (2015) 37:95– 104. doi: 10.1159/000368396


clinical trial. BMC Pediatr. (2017) 17:104. doi: 10.1186/s12887-017- 0859-z


of human umbilical cord blood cells on brain damage and spatial memory deficits in neonatal hypoxia-ischemia. Neuroscience (2012) 210:431–41. doi: 10.1016/j.neuroscience.2012. 03.009


**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 is currently co-organizing a Research Topic with one of the authors MT, and confirms the absence of any other collaboration.

Copyright © 2018 Sato, Ueda, Kondo, Hattori, Mikrogeorgiou, Sugiyama, Suzuki, Yamamoto, Hirata, Hirakawa, Nakanishi, Tsuji and Hayakawa. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Pilot Study of Inhaled CO Therapy in Neonatal Hypoxia-Ischemia: Carboxyhemoglobin Concentrations and Brain Volumes

Martha Douglas-Escobar <sup>1</sup> , Monique Mendes <sup>2</sup> , Candace Rossignol <sup>3</sup> , Nikolay Bliznyuk <sup>4</sup> , Ariana Faraji <sup>3</sup> , Abdullah S. Ahmad<sup>2</sup> , Sylvain Doré2,5,6,7 \* and Michael D. Weiss <sup>3</sup> \*

<sup>1</sup> Department of Pediatrics, University of California, San Francisco, San Francisco, CA, United States, <sup>2</sup> Department of Anesthesiology, Center for Translational Research in Neurodegenerative, McKnight Brain Institutive, University of Florida, Gainesville, FL, United States, <sup>3</sup> Department of Pediatrics, University of Florida, Gainesville, FL, United States, <sup>4</sup> Department of Agricultural and Biological Egineering, University of Florida, Gainesville, FL, United States, <sup>5</sup> Department of Neurology, Center for Translational Research in Neurodegenerative, McKnight Brain Institutive, University of Florida, Gainesville, FL, United States, <sup>6</sup> Department of Psychiatry, Center for Translational Research in Neurodegenerative, McKnight Brain Institutive, University of Florida, Gainesville, FL, United States, <sup>7</sup> Department of Neuroscience, Center for Translational Research in Neurodegenerative, McKnight Brain Institutive, University of Florida, Gainesville, FL, United States

#### Edited by:

Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Robert Galinsky, Ritchie Centre, Australia Gunnar Naulaers, KU Leuven, Belgium

#### \*Correspondence:

Sylvain Doré sdore@anest.ufl.edu Michael D. Weiss mweiss@ufl.edu

#### Specialty section:

This article was submitted to Neonatology, a section of the journal Frontiers in Pediatrics

Received: 18 January 2018 Accepted: 13 April 2018 Published: 01 May 2018

#### Citation:

Douglas-Escobar M, Mendes M, Rossignol C, Bliznyuk N, Faraji A, Ahmad AS, Doré S and Weiss MD (2018) A Pilot Study of Inhaled CO Therapy in Neonatal Hypoxia-Ischemia: Carboxyhemoglobin Concentrations and Brain Volumes. Front. Pediatr. 6:120. doi: 10.3389/fped.2018.00120 Objective: The objective of this pilot study was to start evaluating the efficacy and the safety (i.e., carboxyhemoglobin concentration of carbon monoxide (CO)) as a putative neuroprotective therapy in neonates.

Study Design: Neonatal C57BL/6 mice were exposed to CO at a concentration of either 200 or 250 ppm for a period of 1 h. The pups were then sacrificed at 0, 10, 20, 60, 120, 180, and 240 min after exposure to either concentration of CO, and blood was collected for analysis of carboxyhemoglobin. Following the safety study, 7-day-old pups underwent a unilateral carotid ligation. After recovery, the pups were exposed to a humidified gas mixture of 8% oxygen and 92% nitrogen for 20 min in a hypoxia chamber. One hour after the hypoxia exposure, the pups were randomized to one of two groups: air (HI+A) or carbon monoxide (HI+CO). An inhaled dose of 250 ppm of CO was administered to the pups for 1 h per day for a period of 3 days. At 7 days post-injury, the pups were sacrificed and the brains analyzed for cortical and hippocampal volumes.

Results: CO exposure at 200 and 250 ppm produced a peak carboxyhemoglobin concentration of 21.52 ± 1.18% and 27.55 ± 3.58%, respectively. The carboxyhemoglobin concentrations decreased rapidly, reaching control concentrations by 60 min post exposure. At 14 days of age (7 days post injury), the HI+CO (treated with 1 h per day of 250 ppm of CO for 3 days post injury) had significant preservation of the ratio of ipsilateral to contralateral cortex (median 1.07, 25% 0.97, 75% 1.23, n = 10) compared the HI+A group (p < 0.05).

Conclusion: CO exposure of 250 ppm did not reach carboxyhemoglobin concentrations which would induce acute neurologic abnormalities and was effective in preserving cortical volumes following hypoxic-ischemic injury.

Keywords: babies, ischemic stroke, preclinical, therapeutic gas

**181**

### INTRODUCTION

Hypoxic-ischemic encephalopathy (HIE) is a serious birth complication due to systemic asphyxia (1). The incidence of HIE ranges from 1 to 8 per 1,000 live births in developed countries and as high as 26 per 1,000 live births in underdeveloped countries (2). Until recently, treatment of HIE consisted of supportive care including respiratory support, treatment of hypotension, careful monitoring of fluid and electrolytes, and treatment of seizures. In the last decade, research has shown that therapeutic hypothermia improves the neurological and neurodevelopmental outcome of a subgroup of infants with moderate HIE (3–6). Therapeutic hypothermia decreases mortality and improves the neurological and neurodevelopmental outcome of up to 53% of treated infants (3, 5–8). Neonates with mild HIE have been excluded from hypothermia trials due to earlier studies which showed that these neonates did not have long-term handicaps (9). However, emerging data has shown that neonates with mild HIE may be at risk for brain injury. Currently, there is not a specific neuroprotective therapy for neonates with mild HIE.

Contrary to the traditional view of carbon monoxide (CO) as a toxic agent (10), CO can be neuroprotective at lowdoses (11–13). Exogenous CO has anti-inflammatory, antiapoptotic and vasodilation effects that are cytoprotective (12–14). In vitro, CO preconditioning of neurons prevents apoptosis after induced excitotoxicity and oxidative stress (15). We previously showed that low doses of inhaled CO administered immediately after transient focal ischemia reduced cortical infarct volumes and improved neurological outcomes (16). Our laboratory also discovered that CO regulates the transcriptional factor Nrf2, a key factor in controlling the entire cell antioxidant system through the ARE-Nrf2-Keap1 pathway (17). Although CO has demonstrated great promise in adult animals, there is paucity of knowledge about the potential use of CO as a neuroprotective agent in neonatal animals.

As an initial step in investigating the therapeutic benefit of CO for neuroprotection following mild hypoxic-ischemic brain injury in neonates, we by examined the carboxyhemoglobin over time in neonatal mice exposed to concentration of 200 and 250 ppm of CO exposure for 1 h. These chosen doses were based on animal models of lung injury and our laboratory's results, which show neuroprotection for stroke at these doses (16, 18, 19). Next, a pilot study was performed to test if CO was neuroprotective in a neonatal mouse model of HI. The volumes of cortical and hippocampal injuries were examined in CO-treated HI pups compared with HI air pups and sham controls. We hypothesize that low-dose CO would be safe and preserve brain cortical and hippocampal volumes in a mouse model of neonatal HI.

### MATERIALS AND METHODS

All procedures and anesthetics on the animals were performed in accordance with University of Florida and NIH regulations governing the ethical care and handling of laboratory animals and were approved by the IACUC at the University of Florida.

#### Neonatal HI Mouse Model

The Rice-Vannucci neonatal HI model, which was validated to induce brain injury similar to that seen in neonates with HIE (20, 21). The rat model was modified and validated for use in mice (22) and we have previously established hypoxic-ischemic injury using this model (23–25). Surgical carotid ligation was performed under anesthesia (isoflurane induction 5% and maintenance 2–3%) for an average of 5–10 min. Briefly, the left carotid artery was isolated and ligated in 7-day-old pups. Following the procedure, all pups were allowed to recover for 1 h with the dam. The pups were then placed in a Billups-Rothenburg hypoxic chamber, which was perfused with a humidified gas mixture of 8% oxygen and 92% nitrogen for 20 min. A gel pad (Deltaphase Isothermal Pad, Braintree Scientific Inc., Braintree, MA) was used to ensure the animals maintained a normal temperature during the procedures and during perfusion with the hypoxic gas source. A group of sham-operated animals underwent anesthesia and a surgical incision but did not have their carotid artery ligated. To control for sex and litter variations between litters, each pup from a litter were randomized to one of three groups: Sham operated, hypoxia-ischemia+air (HI+A) or hypoxia-ischemia+carbon monoxide (HI+CO). Due to the randomization, the sex of the pups was not initially recorded. Recording of sex was performed midway through the experiments.

#### Temperature Monitoring

Prior to the hypoxia exposure, the pups had a baseline temperature taken with a Traceable infrared thermometer (Fisher Scientific, Waltham, MA). Immediately following the hypoxia exposure, the temperature was recorded. In addition, the pups that were exposed to CO had a baseline recording prior to exposure to CO and immediately at the completion of the CO exposure for the pups in the cortical and hippocampal studies. A single temperature measurement from the abdomen was performed prior to, during and after CO exposure.

#### Carboxyhemoglobin Concentration in Neonatal Pups

To analyze the carboxyhemoglobin concentration, the 7-dayold neonatal pups were exposed to two inhaled doses of CO at 200 and 250 ppm. These chosen doses were based on animal models of lung injury and our laboratory's preliminary results, which show neuroprotection for stroke at these doses (16, 18, 19). The pups were placed in the hypoxic chamber. Thermoregulation of the pups was obtained using a gel pad (Deltaphase Isothermal Pad, Braintree Scientific Inc., Braintree, MA) during CO exposure. The pups were exposed to the CO for 1 h. A gas analyzer attached to the outflow will strictly monitor the CO levels in the chamber. Following CO exposure, the pups were sacrificed at 0, 10, 20, 60, 120, 180, and 240 min after completion of CO and blood was collected for carboxyhemoglobin analysis.

#### Carboxyhemoglobin Measurements

An avoximeter was used to measure hemoglobin levels in the blood following CO exposure to evaluate the therapeutic window

**Abbreviations:** CO, carbon monoxide; HI, hypoxia-ischemia.

for CO administration. Carboxyhemoglobin was measured using the Avoximeter 4000, a whole blood CO-oximeter. The animals were exposed to 250 ppm of CO for 1 h while being kept warm with a space gel pad. They were sacrificed at various times after the exposure, and whole blood was collected by cardiac puncture using a heparinized syringe. The syringe, containing the sample, was connected to the cuvette and held at a 45◦ . The cuvette was filled by gently pressing the syringe plunger until the sample reached the vent patch. The cuvette, with the syringe still attached, was placed in the test chamber. After 10 s the hemoglobin (g/dL), % carboxyhemoglobin, % oxyhemoglobin, and % methemoglobin were displayed. In an effort to reduce animal numbers, the number of mice was decreased for the 250 ppm exposure group compared to the 200 ppm group since the measurements had little standard deviation. In addition, the number of pups was gradually decreased for the later time points since they approached control pup values.

### CO Therapy for HI

After the hypoxic exposure, the pups were allowed to recover with the dam for 1 h. The HI pups were then randomized to one of two groups: air (HI+A) or carbon monoxide (HI+CO). An inhaled dose of 200 ppm of CO was administered to the pups 1 h after completion of hypoxia using the Billups-Rothenburg chamber. The pups were thermoregulated using a gel pad during CO exposure. The pups were exposed to the CO for a period of 1 h. A gas analyzer attached to the outflow will strictly monitor the CO levels in the chamber. Seven days post-CO exposure, the pups were deeply anesthetized with 5% isoflurane and perfused with 4% PFA. The brains were then collected for volume analysis.

#### Cresyl Violet Staining

The cerebellum was removed and each brain was completely sectioned using a cryostat at a thickness of 30 µm/section (approximately 120 sections/brain). The fixed frozen sections were mounted on Superfrost Plus slides (Fisher Scientific, Waltham, MA) from −80◦C freezer and air-dry overnight at room temperature. The slides were hydrated in 70% ethanol followed by 50% ethanol and finally in distilled H2O for 3 min each, and then placed in 0.5% Cresyl Violet acetate (Electron Microscopy Sciences, Hatfield, PA) for 12 min. Each slide was then dehydrated in ascending ethanol and citrasolve, and coverslipped with permount.

#### Volume Analysis

A Zeiss axiophot equipped with a Microfire CCD camera (Optronics, Goleta, CA) was used for volume analysis. Real-time images were analyzed using the Stereologer 2000 version 2.1 (Stereology Resource Center, Chester, MD). The injured cortex and hippocampus were identified and outlined on every section. The procedure was then repeated for the uninjured cortex and hippocampus.

### Weights, Scoring Testing and Surface Righting

All three groups were weighed on the day of surgery and 3 days after the surgery following the last CO exposure. For each CO exposure, the pups were observed prior to and after. A scoring system was used with Level 1 = unconscious, Level 2 = conscious but weak, Level 3 = standing, Level 4 = alert, nursing well. The pup were placed on their back on a bench pad and held for 5 s. The pups were released and the time it took for them to return to a prone position and the direction of righting were recorded. The test was repeated 3 times. Testing occurred at 24 h post-injury. The CO exposed mice had received 2 doses of CO at the time of testing.

### Statistical Analysis

The injured cortex and hippocampus was standardized to the uninjured cortex. All statistical analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC) and graphs created using GraphPad Prism (GraphPad Software Inc., La Jolla, CA). One-way ANOVA models were considered for the standardized response data. For each of the two data sets, homogeneity of variance assumption was verified using Bartlett's test (pvalues greater than 0.5). The results from the normal theory (untransformed) responses were validated nonparametrically using the Kruskal–Wallis test. For the temperature analysis, a paired t-test was used to test the treatment effect; every subject serves as their own baseline. All data are represented as mean ± sd. A power analysis was not performed to determine sample size because this was a pilot project.

# RESULTS

### Carbon Monoxide Exposures

Exposure of P7 mice to 200 ppm CO for 1 h resulted in a mean carboxyhemoglobin concentration of 21.52% ± 1.18% (n = 6) after the 1 h exposure (time 0). The carboxyhemoglobin decreased to 10.83 ± 1.29% at 20 min (n = 6), 8.12 ± 1.31% at 40 min (n = 6), 4.73% ± 1.20% at 60 min (n = 4), 2.93 ± 1.21% at 90 min (n = 3), and 3.63 ± 1.36% at 120 min (n = 3) (**Figure 1A**).

Exposure of P7 mice to 250 ppm CO produced a higher mean blood carboxyhemoglobin concentration after the 1-h exposure (time 0) of 27.55 ± 3.58% (n = 4). The carboxyhemoglobin concentration was 18.28 ± 3.41% at 20 min (n = 4), 13.90 ± 0.42% at 40 min (n = 2), 6.62 ± 3.79% at 60 min (n = 4), 3.22 ± 2.21% at 120 min (n = 4), and 2.20 ± 0.70% at 180 min (n = 3) (**Figure 1B**). By 60 min post exposure, the concentrations of carboxyhemoglobin were similar to the control group which was not exposed to CO control group (3 ± 2%).

Comparing the two dosages of CO, 200 and 250 ppm, there was a significant increase in the concentration of carboxyhemoglobin at 0, 20, and 40 min at a concentration of 250 ppm compared to 200 ppm (p < 0.05). There was no difference after 40 min at the other time points examined. The concentrations of carboxyhemoglobion in both the 200 and 250 ppm exposed animals were similar to air control pups at 60 min.

Since the carboxyhemoglobin is based on a percentage which can be affected by the total hemoglobin, the hemoglobin was measured to verify that there was not a change postnatally during the period of time in which CO was administered. The hemoglobin concentration remained stable postnatally (**Figure 2**). All pups that were exposed to CO at 200 and 250 ppm survived the exposure.

#### Temperatures

The mean (±sd) temperature for the pups that underwent hypoxia (HI+A and HI+CO) were 33.28 ± 1.03◦C prior to hypoxia and 33.22 ± 0.68◦C immediately following hypoxia (p > 0.05). Prior to CO exposure, the mean temperature were 32.72 ± 0.96◦C and immediately after 34.7 ± 0.51◦C (p > 0.05).

#### Cortical Volumes

At 12 days of age (5 days post injury), the ratio of ipsilateral to contralateral cortical volumes of HI+A (median 0.93, 25% 0.88, 75% 1.03, n = 10) was decreased compared to sham pups (median 1.01, 25% 0.97, 75% 1.25, n = 5) (p < 0.05). The HI+CO (treated with 1 h per day of 250 ppm of CO for 3 days post-injury) had significant preservation of the ratio of ipsilateral to contralateral cortex (median 1.07, 25% 0.97, 75% 1.23, n = 13) compared the HI+A group (p < 0.05) (**Figures 3A** and **4**). All pups that were exposed to CO at 200 and 250 ppm survived the exposure.

#### Hippocampal Volumes

At 12 days of age (5 days post injury), similar to the cortical volumes, the ratio of ipsilateral to contralateral hippocampal volumes of the HI+A (median 0.76, 25% 0.52, 75% 0.84, n = 10) were decreased compared to the sham pups (median 1.0, 25% 0.88, 75% 1.14, n = 5) (p < 0.05). The HI+CO (treated with 1 h per day of 250 ppm of CO for 3 days post injury) preservation of the ratio of ipsilateral to contralateral hippocampus was not significantly different from the HI+A group (median 0.83, 25% 0.52, 75% 0.97, n = 13) (**Figures 3B** and **4**).

FIGURE 1 | Carboxyhemoblobin levels following exposure to CO in neonates. Mouse pups were exposed to CO for 1 h. Time 0 represents the sampling at the completion of the 1-h exposure. Blood concentrations of CO in mouse pups exposed to CO at 200 ppm (A) and 250 ppm (B) are shown over time after completion of the infusion. In the mice exposed to 200 ppm, the mean carboxyhemoglobin concentration was 21.5 ± 1.3% (n = 6) after the 1-h exposure (time 0). The carboxyhemoglobin concentration decreased to 10.8 ± 1.9% at 20 min (n = 6), 8.1 ± 1.8% at 40 min (n = 6), 4.7 ± 1.2% at 60 min (n = 4), 2.9 ± 1.2% at 90 min, and 3.6 ± 1.4% at 120 min (A). Exposure to 250 ppm of CO produced a higher mean blood carboxyhemoglobin concentration after the 1-h exposure (time 0) of 27.6 ± 3.6% (n = 4). The carboxyhemoglobin concentration was 17.4 ± 3.4% at 20 min (n = 4), 13.9 ± 0.4% at 40 min (n = 2), 6.7 ± 3.8% at 60 min (n = 4), 3.2 ± 2.2% at 120 min (n = 4), 2.2 ± 0.7% at 180 min (n = 3) (B). By 60 min post exposure, the concentrations of carboxyhemoglobin were similar to the control group that was not exposed to CO (control group 3.9 ± 2%).

FIGURE 3 | Cortical and hippocampal volumes. At 12 days of age (5 days post injury), the brain region volumes are represented as the ratio of the left (injured)/right (uninjured). (A) The cortical volumes of HI+A were decreased compared to sham pups (n = 8, \*p < 0.05). The HI+CO (treated with 1 h per day of 250 ppm of CO for 3 days post injury) had significant preservation of the ratio of ipsilateral to contralateral cortex compared the HI+A group (n = 10, #p < 0.05). (B) The hippocampal volumes of the HI+A (median 0.76, 25% 0.52, 75% 0.84, n = 9) were decreased compared to the sham pups (n = 5, p < 0.05). The HI+CO (treated with 1 h per day of 250 ppm of CO for 3 days post injury) preservation of the ratio of ipsilateral to contralateral hippocampus was not significantly different from the HI+RA group (n = 10).

### Weights, Subjective Testing and Surface Righting

The weights for the sham pups increased by 1.16 ± 0.17-fold from the first measure compared with an increase of 1.11 ± 0.18 fold in the HI+A group and 1.15 ± 0.21-fold in the HI+CO. There were no differences between the weights when the groups were compared (p > 0.05). The overall behavior of the pups were examined prior and after CO exposure, the CO exposed pups did not display any obvious behavioral differences compared to the Sham and HI+A group after CO exposure. All CO pups were subjectively observed to be Level 4 (alert, nursing well). In addition, during the CO exposure, the pups did not demonstrate any observable abnormal behavior or changes in behavior. The mean (±sd) time for righting was 1.41 ± 0.45 s in the sham pups (n = 17) compared with 1.22 ± 0.0.39 s in the HI+A pups (n = 19) and 1.18 ± 0.48 s in the HI+CO pups (n = 7). There were no differences between the groups (p > 0.05).

### DISCUSSION

The major findings are that, using our preclinical established protocol, 250 ppm CO exposure (1) did not lead to a carboxyhemoglobin concentration that produce acute neurologic changes and (2) was effective in preserving cortical volumes following mild hypoxic-ischemic injury. This is the first report to demonstrate that CO given after HI in a neonatal model preserves cortical tissue. The data is promising and CO should therefore be investigated further as a potential synergistic therapy to be combined with therapeutic hypothermia.

CO is traditionally thought of as an environmental pollutant. Is it safe as a possible therapy in neonates? It is generally understood that inhalation of CO leads to its preferential binding with hemoglobin binding at 250 times greater affinity than oxygen (26). CO bound to hemoglobin produces carboxyhemoglobin which is a stable complex of CO and hemoglobin. In human adults,

carboxyhemoglobin concentrations from 10 to 20% cause tightness across the forehead and possible headache, 20–30% cause headache, 30–40% cause severe headache, dizziness, and dim vision, 40–50% cause fainting, increased respiratory rate, and pulse, 50–60% cause coma with intermittent convulsions; 60–70% cause depressed cardiac function and depressed respiratory effort, and greater than 70% cause death (27). The highest average carboxyhemoglobin concentrations obtained in our experiments were 22% for 200 ppm and 27% for 250 ppm. The highest single carboxyhemoglobin concentration for the 250 ppm group was 31%. The average carboxyhemoglobin concentrations would produce only a mild headache in human patients if the same concentrations were obtained. This is assuming an exact translation to human subjects. However, the hemoglobin affinity for CO varies in mammalian species; thus, 1 h of inhalation of 250 ppm will increase carboxyhemoglobin to 15–20% in adult rats and hamsters, 10–12% in pigs and only 6–8% in healthy adult human subjects (28). These differences in carboxyhemoglobin among species relate primarily to the higher ventilation rates in smaller mammals (28). The concentration of 250 ppm of CO used for neuroprotection in our experimental design is also well below the reported fetal toxic threshold above 300 ppm of CO exposure when administered 24 h per day throughout pregnancy in mice (29). Fetal hemoglobin has a higher affinity for CO with a calculated ratio of 1.74 fetal vs. maternal % carboxyhemoglobin concentrations (29). Based on our results, CO, at 250 ppm in the neonatal mouse, produces carboxyhemoglobin concentrations within a non-toxic range. In addition to the monitoring of carboxyhemoglobin, we did not observe any changes in weight trends or behavior during or after exposure to CO. We performed a righting reflex which did not reveal differences between the groups. Given the mild injury, this test may not be sensitive enough to detect subtle differences between groups and we are currently performing long-term functional outcomes. It should be noted that the CO exposed group did not have a decrease in weight or a worsening performance compared to the other groups indicating that the treatment does not have a grossly negative impact on the pups.

The dose of 250 ppm was our target based on previous work from our laboratory demonstrating protection against transient focal ischemia in an adult mouse model. However, the design differed from our previous report in an adult mouse model of transient focal cerebral ischemia in which the 250 ppm was administered over 18 h (16). CO was administered for 1 h per day over 3 days. The design was based on a recent study in which CO was administered at 250 ppm for 1 h daily for 3 days prior to HI in a neonatal rat model (30). The study demonstrated a decrease in hippocampal apoptosis, an increase in the anti-apoptotic protein Bcl-2, and increased cytochrome c concentrations in CO-treated pups (30). Since the pups were exposed for 1 h and demonstrated an effect, we chose to emulate this design. Our study design differed in that we administered the CO post HI injury. This design was chosen to mimic the potential clinical scenario in which human neonates would be given the therapy post injury.

Similar to the work in adult mice with transient focal cerebral ischemia, there was preservation of cortical tissue in a neonatal mouse model of HI when given CO at 250 ppm (16). However, as noted above, the duration of administration was not as long—an 18-h single dose in the adult transient focal cerebral ischemia model vs. 1 h per day for 3 days in the neonatal HI model. The degree of cortical preservation was not as great as in the adult animal and this may relate to the dosing or the model. The injuries of the hypoxic-ischemic pups that were not treated with CO were not as significant as we expected. This may relate to the duration of exposure to hypoxia following the carotid artery ligation; the insult was relatively minor with only a 14% reduction in the cortical volume from the injury. It is, however, encouraging that CO could still preserve cortical volumes in the neonate following a minor HI insult. The finding is of clinical interest since neonates with mild HIE have been excluded from hypothermia trials due to earlier studies which showed that these neonates did not have long-term handicaps (defined as cerebral palsy, hearing or visual deficits, epilepsy or a score 3 SD below the population mean on the Stanford-Binet IQ test (9)). However, emerging data has shown that neonates with mild HIE may be at risk for brain injury. In a recent study, 50 neonates with mild HIE and who were not cooled underwent an MRI at 10 days of life (31). The MRI revealed injury in 40% of this mild group including near total injury in the basal ganglia and watershed areas of the cortex in 25% (31). Another study examined 104 neonates with a perinatal acidemia. These babies underwent a neurologic exam and 60 were found to have a mild encephalopathy. Of these 60, 12 (20%) experienced an abnormal short-term outcome (i.e., abnormal brain MRI, seizures, abnormal neurologic exam at discharge, need for gastronomy tube, death). Murray et al., in a prospective cohort study, examined the outcome of neonates with mild HIE at 5 years of age (32). Infants with mild HIE had significantly lower full, verbal, and performance IQs when compared to healthy control infants (32). Our data would suggest that CO may have therapeutic benefit in mild HIE.

Future experiments will expose the pups to longer durations of hypoxia post ligation to test if CO can have more of a significant impact on cortical volume preservation. Alternatively, the duration of exposure to CO may have to be increased.

The mechanism by which CO produces neuroprotection in the neonate following HI is currently unknown. CO protects against various insults similar to the pathophysiology of HIE through activation of anti-inflammatory, anti-apoptotic, and vasodilatory mechanisms (33–35). Our group has demonstrated that CO may act through the ARE-Nrf2-Keap1 pathway (17). This pathway is instrumental in regulating environmental stress by activating genes for antioxidants and detoxification. The pathway protects cells against inflammation. In its nonactivated state, Nrf2 is bound to Keap1; following exposure to stressors the Nrf2-Keap1 complex dissociates and Nrf2 moves to the nucleus. Activation of this pathway may lead to many neuroprotective downstream mechanisms. In our experiments, the concentrations of carboxyhemoglobin decreased rapidly after administration of CO for 1 h, reaching concentrations comparable to controls after 60 min. Although speculative, the limited CO exposure could have activated many downstream pathways which led to preservation of cortical tissue by limiting

inflammation and oxidative stress produced following hypoxiaischemia in neonates, and this is actively being pursued using our preclinical model. Using a similar strategy in the clinical arena could minimize side effects from CO while producing beneficial downstream neuroprotective cascades.

A limitation of our study was the inability to sample carboxyhemoglobin from the same pup over time. The size of the pups precluded multiple samplings and the pups were sacrificed at each time point that the carboxyhemoglobin concentration was measured to obtain adequate blood volume for analysis. In addition, the cardiorespiratory status was also not monitored during the administration of CO and the neurologic outcomes were preliminary. Future studies will address any physiologic disturbances during the administration of CO and perform intricate neurologic testing to understand the short and longterm outcomes following CO exposure.

In summary, we have demonstrated that exogenous administration of CO did not produce carboxyhemoglobin concentrations which are associated with detectable CNS dysfunction and preserves cortical volumes following neonatal HI. Future experiments will focus on the functional outcomes of pups exposed to CO following HI, examining possible synergy with hypothermia and using transgenic mice to dissect the mechanism of action of CO in modulating injury post-HI through the ARE-Nrf2-Keap1 pathway. If future experiments are promising, large animal models will need to be utilized before proceeding with human trials, as rodents do not have hemoglobin F (only embryonic hemoglobin that switches to

#### REFERENCES


adult hemoglobin at E17). Thus, the pharmacokinetic of carbon monoxide in neonates should be tested in large mammals to adjust for fetal hemoglobin prior to human application.

### AUTHOR CONTRIBUTIONS

MD-E, MW, and SD designed, provided funding and expertise for, and analyzed all experiments, wrote the manuscript, and trained all staff in performing behavioral and histochemical experiments and data analyses; CR and AA trained and assisted in behavioral and histochemical experiments and analyses and edited the manuscript. CR and MM performed the behavioral analyses and quantified the histology. All authors have accepted the final version of the manuscript.

#### FUNDING

This work was supported by a grant from the McKnight Brain Research Foundation, Brain and Spinal Cord Injury Research Trust Fund (SD, MW).

#### ACKNOWLEDGMENTS

We extend special thanks to CR for her technical assistance with behavioral testing, immune and histopathology, and data quantification, and to all Doré and Weiss lab members for their generous assistance. Publication of this article was funded in part by the University of Florida Open Access Publishing Fund.


J Respir Crit Care Med. (2004) **170**:613–20. doi: 10.1164/rccm.200401- 023OC


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

*Lauren L. Jantzie1,2\*, Akosua Y. Oppong3 , Fatu S. Conteh3 , Tracylyn R. Yellowhair1 , Joshua Kim3 , Gabrielle Fink3 , Adam R. Wolin3 , Frances J. Northington4 and Shenandoah Robinson3 \**

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Masanori Iwai, Kumamoto University Hospital, Japan Tina Bregant, Children's Hospital Ljubljana, Slovenia*

#### *\*Correspondence:*

*Lauren L. Jantzie ljantzie@salud.unm.edu; Shenandoah Robinson srobin81@jhmi.edu*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 09 December 2017 Accepted: 26 March 2018 Published: 13 April 2018*

#### *Citation:*

*Jantzie LL, Oppong AY, Conteh FS, Yellowhair TR, Kim J, Fink G, Wolin AR, Northington FJ and Robinson S (2018) Repetitive Neonatal Erythropoietin and Melatonin Combinatorial Treatment Provides Sustained Repair of Functional Deficits in a Rat Model of Cerebral Palsy. Front. Neurol. 9:233. doi: 10.3389/fneur.2018.00233*

*1Department of Pediatrics, University of New Mexico School of Medicine, University of New Mexico, Albuquerque, NM, United States, 2Department of Neurosciences, University of New Mexico School of Medicine, University of New Mexico, Albuquerque, NM, United States, 3Pediatric Neurosurgery, Johns Hopkins University, Baltimore, MD, United States, 4Neonatology, Department of Pediatrics, Johns Hopkins School of Medicine, Baltimore, MD, United States*

Cerebral palsy (CP) is the leading cause of motor impairment for children worldwide and results from perinatal brain injury (PBI). To test novel therapeutics to mitigate deficits from PBI, we developed a rat model of extreme preterm birth (<28 weeks of gestation) that mimics dual intrauterine injury from placental underperfusion and chorioamnionitis. We hypothesized that a sustained postnatal treatment regimen that combines the endogenous neuroreparative agents erythropoietin (EPO) and melatonin (MLT) would mitigate molecular, sensorimotor, and cognitive abnormalities in adults rats following prenatal injury. On embryonic day 18 (E18), a laparotomy was performed in pregnant Sprague– Dawley rats. Uterine artery occlusion was performed for 60 min to induce placental insufficiency *via* transient systemic hypoxia-ischemia, followed by intra-amniotic injections of lipopolysaccharide, and laparotomy closure. On postnatal day 1 (P1), approximately equivalent to 30 weeks of gestation, injured rats were randomized to an extended EPO + MLT treatment regimen, or vehicle (sterile saline) from P1 to P10. Behavioral assays were performed along an extended developmental time course (*n* = 6–29). Open field testing shows injured rats exhibit hypermobility and disinhibition and that combined neonatal EPO + MLT treatment repairs disinhibition in injured rats, while EPO alone does not. Furthermore, EPO + MLT normalizes hindlimb deficits, including reduced paw area and paw pressure at peak stance, and elevated percent shared stance after prenatal injury. Injured rats had fewer social interactions than shams, and EPO + MLT normalized social drive. Touchscreen operant chamber testing of visual discrimination and reversal shows that EPO + MLT at least partially normalizes theses complex cognitive tasks. Together, these data indicate EPO + MLT can potentially repair multiple sensorimotor, cognitive, and behavioral realms following PBI, using highly translatable and sophisticated developmental testing platforms.

Keywords: cerebral palsy, chorioamnionitis, hypoxia-ischemia, inflammation, social interaction, gait, touchscreen, cognition

# INTRODUCTION

Cerebral palsy (CP) is the leading cause of motor impairment for children worldwide and typically results from perinatal brain injury (PBI) (1, 2). While preterm birth is a common etiologic antecedent, motor impairment and associated deficits can also arise from other insults to the developing central nervous system (CNS), including trauma and stroke. Notably, the scope of PBI has shifted over recent decades as more preterm infants survive (3–5), and the proportion of children with more severe motor impairment has increased in the USA (6). Within subpopulations of neonates with PBI, multiple injury mechanisms have been implicated, and emerging evidence strongly suggests that each newborn suffers a unique vulnerability to CNS injury from a combination of (1) inflammation from prenatal infection and/or hypoxia-ischemia (HI); (2) individualized risk from genetic and/ or congenital predisposition and acquired prenatal exposures to drugs, toxins, and nutritional status; and (3) postnatal stresses, such as sepsis and surgery. Indeed, intrapartum events are implicated in the etiology of less than 12% of children with CP (7). Thus, there is an urgent need for safe, effective interventions for PBI, and subsequent CP and related deficits.

Infection and HI catalyze PBI by creating a toxic *in utero* and neural microenvironment that limits oxygen exchange and propagates inflammation during critical periods of neurodevelopment (8–15). Typically, infants with PBI present with injury to major white and gray matter structures that leads to reduced connectivity of developing cerebral networks. Subsequently, diverse functional deficits ensue with impairment in multiple motor, cognitive, and behavioral realms that precipitates poor educational progress during childhood (16–25). Chorioamnionitis (infection/inflammation of the amniotic fluid, membranes, and placenta) affects placental permeability and blood flow, facilitates HI and fetal transmission of inflammation, and is associated with a significant increase in systemic inflammation (26–29). Chorioamnionitis is common in both preterm and term infants (23, 24, 29–34). It affects approximately 40–80% of very preterm deliveries and 20–34% of deliveries at term (30, 33, 35). Chorioamnionitis is also recognized in as many as 42% of placentas from unremarkable pregnancies (29, 36). Notably, in term infants with HI encephalopathy, the presence of chorioamnionitis predicts decreased responsiveness to hypothermia treatment (8–10, 30–32, 37, 38) and magnesium sulfate (39). Because these current strategies for neonatal repair are less effective in the setting of chorioamnionitis, we sought to address this unmet need by testing promising neuroreparative agents using a preclinical model of chorioamnionitis.

Despite the wealth of epidemiological and clinical data related to chorioamnionitis and the development of motor deficits in children born preterm, little progress has been made in identifying interventions that mitigate the CNS injury that leads to CP. Indeed, ambulation, behavior, and cognition are complex tasks impacted by early CNS injury (40). To minimize deficits and optimize outcomes for children with CP, novel therapies are required to restore motor skills, sensation, behavior such as attention and social interaction, and cognition, including executive function. However, few novel therapies have directly addressed these complex and compound deficits, particularly the functional pillars of cognition and behavior with motor impairment. Here, we studied a combination therapy of the endogenous neuroreparative agents erythropoietin (EPO) and melatonin (MLT) in an established preclinical rat model that accurately encompasses the complete maternal-placental-fetal brain axis with intrauterine injury and recapitulates pathophysiology from extreme preterm birth. We chose a cocktail strategy to mitigate the multiple pathophysiological mechanisms that contribute to PBI, capitalize on innate CNS recovery, and respond to clinical recommendations on utility of single therapies (41–43). Furthermore, data from our labs and others, confirm combinatorial therapy with EPO + MLT may provide enhanced, synergistic neurorepair by (1) optimizing the genesis and survival of multiple neural cell lineages, including cells with high bioenergetic demands, such as oligodendrocytes, myelin sheaths, and ependyma with motile cilia, (2) normalizing excess calpain activity and its destruction of essential molecules during neurodevelopment, (3) reducing neuroinflammation and free radicals, and (4) limiting mitochondrial dysfunction and associated endoplasmic reticulum stress (44–55). Given this unique avenue for translation and targeted mechanisms of action, we tested the hypothesis that an extended postnatal EPO + MLT cocktail would mitigate gait, sensorimotor, cognitive, and behavioral changes associated with PBI, using highly translatable and sophisticated testing platforms that are similar to the ones used in humans, including digital gait analysis and touchscreen cognitive testing.

### MATERIALS AND METHODS

The Institutional Care and Use Committee at the University of New Mexico Health Sciences Center, Boston Children's Hospital and Johns Hopkins University approved all experimental procedures. For each experiment described, equal numbers of male and female pups were used, and data represents true *n* (individual pups) from at least two different dams per condition. Specifically, we adhered to accepted standards for rigorous study design and reporting to maximize the reproducibility and translational potential of our findings, as described by Landis et al. and in the ARRIVE guidelines (56–58). Animals of both sexes were randomized to experimental or sham control groups and EPO + MLT or vehicle treatments. All investigators were blinded to injury and treatment group during the conduct and analyses of each experiment. For each experiment, a power analysis was also performed to estimate the required sample size (G\*Power 3.1.9.3). For these calculations, we used published and preliminary data to define the expected means and SDs for each group, and we exceeded the calculated number needed in every experiment. Separate cohorts of rats were used for open field, gait and social interaction, and touchscreen assessments.

#### *In Utero* Injury: Chorioamnionitis

As placental structure and function is of significant clinical importance to neurologic sequelae in preterm survivors (59–61), we use a prenatal model of *in utero* transient systemic HI (TSHI) and intra-amniotic lipopolysaccharide (LPS) administration in pregnant rats (62–64). This approach capitalizes on an intact maternal–placental–fetal unit and is a model of PBI from extreme preterm birth (<28 weeks of gestation) that mimics dual intrauterine injury from placental underperfusion and chorioamnionitis (65). Briefly, under isoflurane anesthesia, a laparotomy is performed on embryonic day (E) 18. Uterine arteries are clamped for 60 min and followed by intra-amniotic injections of LPS (4 μg/sac; 0111:B4, Sigma, St. Louis, MO, USA) (62, 64, 65). Sham controls undergo anesthesia and laparotomy for 60 min without arterial clamping or LPS injections. Following closure of the laparotomy, dams receive narcotic pain medication, recover, and pups are born vaginally at E22, approximately equal to 30–32 weeks in human gestation. We have previously reported the effects of TSHI and LPS alone, and in concert, on CNS pathological hallmarks, functional motor outcomes, histologic placental injury, and expression of common pro-inflammatory cytokines (63, 64).

#### EPO and MLT Combination Therapy

Erythropoietin and MLT are endogenous, developmentally regulated molecules that are individually most effective for neurorepair when administered in extended dosing regimens (45, 48, 66–69). Rodents are born at a time equivalent to the human third trimester, with P9 approximately equivalent to term in human gestation (70). Accordingly, we used an established, clinically relevant dosing regimen (47, 48, 55, 71), in which pups on postnatal day (P) 1 from all injured litters were individually randomized to receive either EPO (2,000 U/kg, R&D Systems, Minneapolis, MN, USA) plus MLT (20 mg/kg, Sigma), or vehicle (sterile saline). Subsequently, EPO was then administered intraperitoneally once daily from P1 to P5 and MLT was administered once daily from P1 to P10, comparable to dosing regimens used in human neonatal trials. When EPO was administered in isolation, it was given from P1 to P5 at 2,000 U/kg/dose as previously published (47, 48, 54, 55, 71, 72). Prior work has shown that shams do not exhibit any negative effects from EPO and MLT treatment (55), and thus to conserve resources, shams received only vehicle.

#### Open Field

A circular open field arena (100 cm diameter) was placed in a quiet, well-lit room (130 lm), and was marked to divide the arena into three equally spaced, concentric circles labeled the center, neutral, and peripheral zones. At P28–P30, each rat was initially placed against the wall of the testing arena and allowed to explore for 15 min. Anymaze™ video-tracking software was used to record and measure open field behavior.

#### Gait Analysis

Computerized gait analysis was performed on P25–P26 as previously described (62, 71). Briefly, digital video of each rat running on a backlit transparent treadmill set at 30 cm/s was acquired with a high-speed camera and analyzed using Digigait software (Mouse Specifics, Framingham, MA, USA). Digigait software analyses identifies individual paw prints and allows calculation of multiple gait metrics and kinematic measurements based on the position, area, and timing of each step. *In utero* chorioamnionitis induces a global injury. Thus, data from right and left hindlimbs were combined for analysis.

### Social Interaction

A standard paradigm was used to identify impaired social interaction in rats at P30–P32 (73–75). Briefly, 1 h prior to testing, each rat was isolated in a clean cage. For social interaction testing, two rats of the same sex and treatment group, but from different litters, were placed in a dimly lit (30 lm) circular testing arena (100 cm diameter) and recorded for 10 min using Anymaze™ video-tracking software. Each pair was counted as one social unit. Two observers blinded to the treatment group independently reviewed the trials and scored periods of social interaction (trailing, sniffing, grooming, playing, etc.). Intraclass coefficient was calculated for interrater reliability of social scoring. Olfactory testing for social and food odors confirmed primary sensory deficits were not related to the impaired social interaction observed in the injured animals.

#### Touchscreen Testing

To better define deficits in cognitive realms, we use a touchscreen operant platform to test specific components of cognition and executive function commencing with mild food deprivation at P28, training at P35, initial testing at P42, and continuing through completion of the paradigms at approximately P90 (76–80). Briefly, using a separate cohort of rats, operant behavior was tested in a sound and light attenuating chamber (Med Associates, St. Albans, VT, USA). A pellet dispenser delivers 40 mg dustless pellets (Bioserv, Frenchtown, NJ, USA) into a magazine, and a houselight is located at one end of the chamber. The opposite end of the chamber houses a touch-sensitive screen (Conclusive Solutions, Sawbridgeworth, UK) overlaid by a black acrylic aperture plate, resulting in two separate touch areas for the rat to register a response. Stimulus presentation in the response windows and touches were controlled and recorded by KLimbic Software (Conclusive Solutions).

#### Pretraining

On P28, rats were first slowly reduced and then maintained at 85% free-feeding body weight. Rats were weighed and assessed for general health daily. The mild weight reduction was well tolerated. Prior to training, rats were acclimated to the 40 mg food pellet reward by provision of 25 pellets/rat in the home cage. Rats were then habituated to the operant chamber and to eating from the pellet magazine. Rats retrieving at least 48 pellets in 60 min were moved to a 4-stage training regimen. Rats first performed autoshaping, followed by three visual discrimination training sessions (76–79).

#### Discrimination and Reversal Learning

Following pretraining, all rats were tested on a pairwise discrimination-reversal paradigm during daily 60 min sessions. For discrimination learning, 2 novel, equiluminescent stimuli verified for rats, were presented in a spatially pseudo-randomized manner over 60-trial sessions (5-s inter-trial interval) (76–80). Responses at one stimulus yielded a reward, whereas responses at the other stimulus resulted in a 5 s time-out (singled by extinguishing the house light). Designation of initially reward stimulus was randomized across treatment. Stimuli remained on screen until a response was made. Rats were trained to an *a priori* criterion of greater than ≥80% correct responses for two consecutive days. Assessment of reversal learning began on the session after discrimination criterion was attained. For this test, the designation of stimuli as correct versus incorrect was reversed for each rat. Like discrimination, rats were tested on 60-trial daily sessions for reversal to an *a priori* criterion of ≥80% correct responses for two consecutive sessions. Errors on first presentation reversal trials were followed by correction trials which continued until a correct response was made, or the session ended. Failing criteria was set *a priori* at 21 sessions (days) for visual discrimination and 21 sessions (days) for reversal.

We recorded the following dependent measures during discrimination and reversal: total sessions, correct responses made, errors (incorrect responses made), correction errors (correction trials, reversal only), reaction time (time from stimulus presentation to touchscreen response), and magazine latency (time from touchscreen response to reward retrieval) (81). Discrimination performance was analyzed across all sessions required to reach criterion. To examine distinct phases of reversal (early perseverative and late learning) mediated by cortical and striatal subregions, respectively, we also analyzed errors and correction trials. Assuming a rat would achieve 50% correct by chance, perseveration was defined as sessions where performance was below 50% correct, and learning as performance from 50% correct to passing criterion, as previously described (81–83).

#### Statistical Analysis

Statistical analyses were performed using SPSS25 (IBM, Armonk, NY, USA). For all analyses, data are represented as mean ± SEM, with *p* < 0.05 considered significant. For analysis of sham, vehicle-treated injury and EPO + MLT-treated injury groups, all parametric variables were tested for normal distribution with the Shapiro–Wilk test with Levene's test to confirm homogeneity of variances. A two-way ANOVA was then performed with Bonferroni's *post hoc* correction for multiple comparisons. For non-parametric variables such as passing criteria in touchscreen testing, a Kruskal–Wallis test with Dunn's *post hoc* correction was performed.

# RESULTS

### EPO **+** MLT Mitigates Disinhibition Following Prenatal Injury

We assessed open field behavior to quantify activity and disinhibition. Compared to sham controls (*n* = 29), rats subjected to prenatal injury (*n* = 23) were much more mobile, which was particularly evident in the last 5 min of the 15-min observation period (**Figure 1A**). Interestingly, compared to vehicle-treated rats with prenatal injury, EPO + MLT normalized the hypermobility (*n* = 28, two-way ANOVA, *p* = 0.022), whereas EPO alone (*n* = 15) did not. After prenatal injury, adult rats were also disinhibited. Specifically, those with *in utero* injury showed a lack of environmental awareness by spending more time immobile in the arena center compared to shams (*p* = 0.03; **Figure 1B**). Similarly, prenatal injury had a significant effect on disinhibition, with sham rats exploring the peripheral zone for significantly longer periods compared to vehicle-treated injury rats (*p* = 0.001; **Figure 1C**), and spending less time in the neutral zone (*p* = 0.041; **Figure 1D**). Treatment with EPO + MLT, but not EPO alone, normalized total time spent in the peripheral (*p* = 0.024) and neutral zones (*p* = 0.038; **Figure 1**), consistent with typical rat behavior, appropriate anxiety and general avoidance of open areas.

## EPO **+** MLT Normalizes Hindlimb Deficits After Prenatal Injury

After observing that EPO + MLT normalized hyperlocomotion and disinhibition in adult rats following prenatal injury, we performed a detailed computerized digital analysis of gait to determine if EPO + MLT could improve motor performance. Compared to shams (*n* = 18), after prenatal injury vehicle-treated adult rats exhibit an abnormal gait, stance and paw placement, with decreased paw area contact (*n*= 21, *p*= 0.012) and decreased paw pressure (*p* = 0.031) suggestive of toe-walking, and reduced percent shared stance (*n* = 21, *p* = 0.006) (**Figure 2**) consistent with spastic gait patterns observed in ambulatory children with CP. Significantly, neonatal combination therapy with EPO + MLT (*n* = 7) reverses deficits in stance (*p* = 0.035) and paw placement (area and pressure both *p* < 0.001), consistent with an improved gait kinematic efficiency with combination treatment.

### EPO **+** MLT Attenuates Deficits in Social Interaction

To quantify social interaction, pairs of sex, injury, and treatmentmatched rats from different litters were observed and scored. The interrater reliability of social interaction scoring was 0.932. Sham (*n* = 18 rats in 9 pairs, *p* < 0.001) and injured rats treated with EPO + MLT (*n* = 8 rats in 4 pairs, *p* = 0.007) had significantly more social interactions during the observation period, including sniffing, playing, and grooming, compared to vehicle-treated rats with *in utero* injury (*n*= 14 rats in 7 pairs) (**Figure 3**). Significantly, EPO + MLT treatment ameliorated deficits in social drive and behavior.

## EPO **+** MLT Mitigates Deficits in Executive Function

To complement our assessment of gait, open field, and social behavior in adult rats with *in utero* injury, we completed a sophisticated assessment of visual discrimination and reversal learning in our animals to evaluate executive function. We began by validating the touchscreen platform in our model of *in utero*

chorioamnionitis and assessed whether adult rats following prenatal insult could perform visual discrimination. Importantly, rats in all three treatment groups were successful in completing all aspects of touchscreen habituation and training.

We first assessed cognitive performance on visual discrimination. Rats from each experimental group were able to perform VD, with 67% of sham (*n*= 27) and 20% of vehicle-treated injured rats (*n* = 20, *p* = 0.005) achieving passing criteria, compared to 58.3% of injured rats treated with EPO + MLT (*n* = 12, *p* = 0.11; **Figure 4A**). After assessing overall performance and pass rate, we then analyzed the number of errors throughout the visual discrimination paradigm as a more rigorous and granular metric of task performance. Notably, for those rats completing VD, a similar number of errors to achieve passing criteria was noted

across injury and treatment groups (**Figure 4B**). As expected, all rats had comparable reaction time and magazine latency (**Figures 4C,D**) throughout the VD paradigm.

Upon successful completion of VD, rats were evaluated for reversal learning. Vehicle-treated injured rats were significantly impaired and fewer passed the overall learning paradigm compared to sham and EPO + MLT-treated rats. Specifically, only 10% of vehicle-treated injured animals successfully passed criteria for VD and reversal (*p* = 0.046) compared to 55.5% of sham and 41.5% of injured animals treated with EPO + MLT (*p* = 0.07; **Figure 4E**). Notably, injured animals treated with vehicle required more correction trials compared to shams (*p*= 0.034). Injured rats treated with EPO + MLT showed a trend toward fewer correction trials (*p* = 0.077), compared to vehicle-treated rats (**Figure 4F**). Further analyses of the maladaptive learning in the reversal paradigm consistent with a lack of cognitive flexibility, indicated a trend for improved performance during both perseveration and learning phases of the reversal paradigm. These results show that touchscreen testing can be used in PBI to distinguish complex behavior related to executive function and learning and that EPO + MLT can at least partially reverse the reduced cognition present after prenatal injury.

#### DISCUSSION

In this investigation, we tested the efficacy of combined EPO + MLT for neurorepair of the deficits associated with CP using translatable outcome measures, with the goal of facilitating rapid transition to neonatal clinical trials. To begin, we capitalized on a preclinical platform and model of CP that accurately recapitulates the multi-faceted pathophysiology of early CNS injury, including an intact maternal-placental-fetal axis, and sustained deficits in adult animals in multiple functional domains. These investigations reflect recent clinical epidemiological progress that indicates most CP arises from prenatal injury and that only 12% of cases arise from intrapartum insults (7). Consistent with clinical data, the preclinical findings reported here reaffirm the concept that chorioamnionitis concomitant with placental insufficiency results in dynamic, multifactorial, and permanent changes to the CNS culminating in functional deficits in mature rats. Indeed, this prenatal insult causes significant chronic behavioral, social, executive function, and gait deficits in adult rats, similar to those observed in children with CP (46, 62, 71). This model that incorporates intrauterine chorioamnionitis is one of very few preclinical models to induce persistent gait and neurocognitive deficits in the mature CNS (40, 70).

Mechanistically, compelling evidence suggests EPO and MLT have significant promise as potential synergistic interventions for neonates at high risk of CP. EPO and MLT have multiple over-lapping and complementary mechanisms of action. As developmentally regulated growth molecules, both EPO and MLT enhance neuronal and oligodendroglial survival and differentiation after CNS injury, suppress toxic cell death pathways, reduce free radicals, and mitigate inflammation from neonatal CNS infection (44–55). Unlike EPO, which is produced predominantly by the kidney and neural cells after birth (84), MLT is an endogenous indoleamine that is produced by the pineal gland postnatally and classically reported to regulate circadian rhythms. It is a direct antioxidant and free radical scavenger and also has indirect actions to increase the production of antioxidant enzymes including glutathione peroxidase (GP) and superoxide dismutase (SOD) (85). Notably, preterm infants with PBI, including those initiated *in utero* by choriomanionitis, are known to have reduced levels of GP and SOD in both the brain and lung (86).

The combination of EPO and MLT also has direct antiinflammatory and immunomodulatory properties integral to their beneficial effects. Impaired regulation of immune responses is detrimental to multiple pregnancy outcomes, including preterm birth. It is plausible that both EPO and MLT link maternal, placental, and fetal physiological cell signaling through mechanisms of entrainment and direct biological actions (87). Interestingly, MLT is synthesized in higher concentrations within the placenta than the pineal gland (87, 88). Specifically, cyto- and syncytiotrophoblasts from the placenta contain two enzymes, serotonin *N*-acetyltransferase and *N*-acetylserotonin methyltransferase, which metabolize serotonin to MLT. Once in the circulation, MLT can increase phagocytosis, antigen presentation, and exert antioxidant effects (87, 89). Indeed, both EPO and MLT are known to affect Th1/Th2 ratio, Th17, neutrophils, and microglia, major cellular mediators of chorioamnionitis and PBI (87, 90). Through separate signal transduction, CNS inflammation actually reduces innate CNS EPO and MLT production, thereby diminishing endogenous neurorepair (91). In this context, similar to exogenous EPO therapy, exogenous MLT administration after birth may supplement innate levels by the pineal gland and replace a MLT deficit arising from premature separation from placental sources and/or induced inflammation from intrauterine infection or injury.

Erythropoietin and MLT also promote the genesis, survival, and differentiation of neural cells in the developing and mature CNS and reduce calpain-mediated injury. Previously, we have shown that sustained excess calpain activity is an important mechanism of injury in the immature CNS (45, 47, 48) and that extended EPO treatment mitigates calpain-mediated damage (48). Specifically, calpain degrades CNS molecules and proteins essential for the formation of cerebral circuits, including neurofilaments, myelin basic protein, and the potassium chloride cotransporter KCC2 (48). Therefore, through EPO and MLT together, it may be possible to cumulatively preserve more axon-myelin structural units, including those in major cerebral white matter and corticospinal tracts, by inhibiting detrimental protease expression, preserving structural connectivity, and restoring inhibitory neural networks. Indeed, it is through this action on structural and functional connectivity, neural conduction, and excitatory/inhibitory balance of fundamental circuitry that this combination of therapy likely improves motor and cognitive function into early adulthood (P90) following prenatal injury (46–48, 71).

Recently, it has been demonstrated that EPO has an additional novel mechanism in regulation of homeostatic plasticity and synaptic strength (92). Together, with previous reports on the modulation of inhibitory circuitry in brain regions key to higher order brain function and structural connectivity, this effect on synapses provides an additional novel molecular mechanism supporting the improvement in cognition and behavior shown here, and the normalization of the trajectory of brain development after perinatal injury (47, 66, 71, 72, 92–94). Similarly, complimenting the EPOR distribution on glia, and neurons, and the importance of receptor/ligand balance in the developing brain (55), MLT receptors, MT1 and MT2, are present in regions of the brain that are important to cognition and memory, including the hippocampus and frontal cortex and are similarly regulated by endogenous MLT (69, 87, 95, 96). In rodents, several studies have reported improved social behavior, anxiolytic, antidepressant, and memory-facilitating effects of exogenous MLT related to modulation of essential neurotransmitters and their receptors, including GABAergic, dopaminergic, glutamatergic, cholinergic, and noradrenergic transmission (69, 97–100). Consistent with our data, studies in mice confirm a MLT-induced decrease in open field hyperlocomotion (69). Indeed, our data also matches prior studies demonstrating that MLT administration in similar dosing ranges reverses ketamine-induced deficits in social interaction and memory impairment (69). Significantly, these investigators found that mitigation of social deficits only occurred with dose-dependent, repeated administration of MLT. Interestingly, in our studies, the combination of EPO + MLT was able to reverse abnormal open field behavior, whereas EPO alone was not. This highlights that, in specific microenvironments, a combination of agents may be more effective than one agent alone, and through repair of white matter, synapses, structural connectivity, neural network efficiency, and multiple neurotransmitter systems, EPO and MLT together may additively improve multiple pillars of cognition and behavior.

This is the first time to our knowledge that touchscreen platforms have been used with preclinical models of CP and PBI. Given that many children with CP have sensorimotor *and* cognitive deficits, there is a need to study strategies that address both functional domains. Touchscreen assessment is a highly translatable outcome measure also utilized in human trials and the Cambridge Neuropsychological Test Automated Battery is regularly used for neuropsychological testing of children and adults (76, 80). Our results demonstrate that EPO + MLT partially mitigates deficits of cognition, specifically executive function and reversal learning. These data support the clinical literature showing that very preterm children and adolescents are at high risk for executive function deficits that only become apparent with increasing cognitive demands. Specifically, compared to healthy born peers, preterm adolescents scored significantly lower in the most demanding levels of working memory, planning, cognitive flexibility, and verbal fluency tasks, despite no group differences being detected at lowest demand levels (101).

We also found *in utero* injury influenced exploratory behavior in mature animals, with EPO + MLT normalizing excess center mobility and resting time in an open field. These data are consistent with prior studies demonstrating improvements in exploratory behavior and resting time with MLT treatment, consistent with normalization of disinhibition, hyperlocomotion, and depression-like behavior (102). Indeed, prior reports suggest that MLT exerts a long-term effect on striatal dopamine content by enhancing monoamine synthesis (102, 103). While that investigation was performed in older rats, it is possible that normalized monoamine system development might improve motor coordination and cognition. Similarly, MLT also acts as a 5HT2A antagonist in the hippocampus, and through the regulation of 5HT release may also impact complex behaviors related to anxiety, behavioral inhibition, and locomotion (102, 104). Notably, children with CP often exhibit spasticity and difficulty with selective motor control. Selective motor control is regulated predominantly by descending serotonin pathways that innervate the lumbar spinal cord on E18 in rats (105), the same age as the prenatal injury used here. Importantly, signaling *via* 5HT2A upregulates spinal KCC2 levels (106). Serotonin signaling is also integral to cognitive flexibility (107). Thus, repairing myelination, axons, KCC2 levels and preserving homeostasis of serotonin signaling may be key mechanistic conversion points of EPO + MLT combination therapy and critical to minimizing spasticity, loss of selective motor control, and preserving cognition in children with CP from prematurity.

In conclusion, EPO + MLT are plausible targeted pharmacotherapies that specifically enhance neurorepair *via* novel and disease-specific molecular mechanisms. Receptor and non-receptor-mediated pathways underpin the multiple neuroprotective effects of MLT and EPO that include supporting mitochondrial function, and post-injury plasticity, and antioxidant, anti-apoptotic, and anti-inflammatory actions (108, 109). Together with normalization of social interaction, these data suggest a plausible treatment strategy to address multiple realms of cognition and behavior. In conjunction with data presented in this study, coupled with EPO and MLT's known safety profile, multiple beneficial mechanisms of action, ability to penetrate the brain and organelles, and ease of administration, EPO and MLT in combination merits thoughtful consideration of clinical trials for preterm infants with brain injury.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations and approval of The Institutional Care and Use Committee (IACUC) at the University of New Mexico Health Sciences Center, Boston Children's Hospital, and Johns Hopkins University.

#### AUTHOR CONTRIBUTIONS

Conception and design: SR and LJ. Acquisition of data: LJ, AO, FC, TY, JK, GF, and AW. Analysis and interpretation of data: LJ, SR, and FN. Drafting the article: SR and LJ. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Study supervision: SR and LJ. SR's previous institution and location at the beginning of this study: Department of Neurosurgery, Boston Children's Hospital, Harvard Medical School, Boston MA.

#### ACKNOWLEDGMENTS

The authors are grateful to Jonathan Brigman, Ph.D. for his expertise with touchscreen testing and interpretation. We are also grateful for the funding provided by the Cerebral Palsy Alliance Research Foundation (PG4116 to SR, FN, and LJ), the Genise Goldenson Fund (to SR and LJ), and the American Heart Association (17SDG33670850 to LJ).

#### REFERENCES


in the newborn infant? *J Pediatr* (2012) 160(4):544–52.e4. doi:10.1016/j. jpeds.2011.12.052


separation in rats and mice. *Nat Protoc* (2013) 8(10):2006–21. doi:10.1038/ nprot.2013.124


tail suspension test. *Eur J Pharmacol* (2010) 638(1–3):78–83. doi:10.1016/j. ejphar.2010.04.011


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

# Mild Intrauterine Hypoperfusion Leads to Lumbar and Cortical Hyperexcitability, Spasticity, and Muscle Dysfunctions in Rats: Implications for Prematurity

Jacques-Olivier Coq1,2 \*, Maxime Delcour 2†, Yuko Ogawa<sup>3</sup> , Julie Peyronnet <sup>1</sup> , Francis Castets <sup>4</sup> , Nathalie Turle-Lorenzo<sup>5</sup> , Valérie Montel <sup>6</sup> , Laurence Bodineau<sup>7</sup> , Phillipe Cardot <sup>7</sup> , Cécile Brocard<sup>1</sup> , Sylvie Liabeuf <sup>1</sup> , Bruno Bastide<sup>6</sup> , Marie-Hélène Canu<sup>6</sup> , Masahiro Tsuji <sup>3</sup> and Florence Cayetanot 1,7

<sup>1</sup> Centre National de la Recherche Scientifique, Institut de Neurosciences de la Timone, UMR 7289, Aix Marseille Université, Marseille, France, <sup>2</sup> Centre National de la Recherche Scientifique, Neurosciences Intégratives et Adaptatives, UMR 7260, Aix Marseille Université, Marseille, France, <sup>3</sup> Department of Regenerative Medicine and Tissue Engineering, National Cerebral and Cardiovascular Center, Suita, Japan, <sup>4</sup> Centre National de la Recherche Scientifique, Institut de Biologie du Développement de Marseille, UMR 7288, Aix-Marseille Université, Marseille, France, <sup>5</sup> FR 3512 Fédération 3C, Aix Marseille Université – Centre National de la Recherche Scientifique, Marseille, France, <sup>6</sup> EA 7369 ≪Activité Physique, Muscle et Santé≫ - URePSSS - Unité de Recherche Pluridisciplinaire Sport Santé Société, Université de Lille, Lille, France, <sup>7</sup> Institut National de la Santé et de la Recherche Médicale, UMR\_S1158 Neurophysiologie Respiratoire Expérimentale et Clinique, Sorbonne Université, Paris, France

Intrauterine ischemia-hypoxia is detrimental to the developing brain and leads to white matter injury (WMI), encephalopathy of prematurity (EP), and often to cerebral palsy (CP), but the related pathophysiological mechanisms remain unclear. In prior studies, we used mild intrauterine hypoperfusion (MIUH) in rats to successfully reproduce the diversity of clinical signs of EP, and some CP symptoms. Briefly, MIUH led to inflammatory processes, diffuse gray and WMI, minor locomotor deficits, musculoskeletal pathologies, neuroanatomical and functional disorganization of the primary somatosensory and motor cortices, delayed sensorimotor reflexes, spontaneous hyperactivity, deficits in sensory information processing, memory and learning impairments. In the present study, we investigated the early and long-lasting mechanisms of pathophysiology that may be responsible for the various symptoms induced by MIUH. We found early hyperreflexia, spasticity and reduced expression of KCC2 (a chloride cotransporter that regulates chloride homeostasis and cell excitability). Adult MIUH rats exhibited changes in muscle contractile properties and phenotype, enduring hyperreflexia and spasticity, as well as hyperexcitability in the sensorimotor cortex. Taken together, these results show that reduced expression of KCC2, lumbar hyperreflexia, spasticity, altered properties of the soleus muscle, as well as cortical hyperexcitability may likely interplay into a self-perpetuating cycle, leading to the emergence, and persistence of neurodevelopmental disorders (NDD) in EP and CP, such as sensorimotor impairments, and probably hyperactivity, attention, and learning disorders.

Keywords: neonatal hypoxia-ischemia, cerebral palsy, intrauterine growth retardation, white matter injury, KCC2

#### Edited by:

Rick Dijkhuizen, University Medical Center Utrecht, Netherlands

#### Reviewed by:

Lauren Jantzie, University of New Mexico, United States Jialing Liu, University of California, San Francisco, United States

#### \*Correspondence:

Jacques-Olivier Coq jacques-olivier.coq@univ-amu.fr

#### †Present Address:

Maxime Delcour, Département de Physiologie, Equipe de Recherche en Réadaptation Sensorimotrice, Faculté de Médecine, Université de Montréal, Montreal, QC, Canada

#### Specialty section:

This article was submitted to Stroke, a section of the journal Frontiers in Neurology

Received: 07 February 2018 Accepted: 22 May 2018 Published: 15 June 2018

#### Citation:

Coq J-O, Delcour M, Ogawa Y, Peyronnet J, Castets F, Turle-Lorenzo N, Montel V, Bodineau L, Cardot P, Brocard C, Liabeuf S, Bastide B, Canu M-H, Tsuji M and Cayetanot F (2018) Mild Intrauterine Hypoperfusion Leads to Lumbar and Cortical Hyperexcitability, Spasticity, and Muscle Dysfunctions in Rats: Implications for Prematurity. Front. Neurol. 9:423. doi: 10.3389/fneur.2018.00423

### INTRODUCTION

It is now well admitted that perinatal brain damage and neurodevelopmental disorders (NDD) are usually related to several conditions, such as neonatal encephalopathy, perinatal arterial ischemic stroke, systemic infections, and premature birth (1). In the worst cases, perinatal brain damage often leads to cerebral palsy (CP), which is a complex syndrome of various sensory, motor (including spasticity, contractures, and spams), and cognitive deficits and is considered the main cause of physical disability in children (2, 3). About 40% of extremely born preterm children (24–32 weeks of gestation) develop moderate to severe sensorimotor and/or cognitive impairments, while the rest of these very preterm children often exhibit minor motor, behavioral, and cognitive disorders. Premature birth occurs in 1/8 of deliveries, but the proportion of prematurity is steadily increasing since the early 1990s in developed countries (1, 4). With increasing prevalence, encephalopathy of prematurity (EP) is mainly characterized by gray matter dysmaturation and reduction, diffuse white matter injury (WMI) related to abnormal oligodendroglial precursor maturation leading to hypomyelination, minor to mild sensorimotor, behavioral and cognitive disorders, and often results in CP (1). Although the relations between the NDDs, brain damage, perinatal hypoxiaischemia, and neuroinflammation are not completely clear and understood (3, 5), we recently developed a rodent model of EP, based on prenatal ischemia (6) which in fact better corresponds to mild intrauterine hypoperfusion (MIUH) (7).

In a prior series of studies (8–10), we used prenatal ischemia or MIUH at embryonic day 17 (E17), considered to be equivalent to embryonic weeks 20–25 in humans (11, 12). MIUH at E17 led to myelination deficits in the corpus callosum and cingulum of rat neonates when assessed between birth and postnatal day 21 (P21) (13, 14). In rats examined at adulthood after MIUH at E17, hypomyelination and axonal degeneration persisted in the internal and external capsules, corpus callosum, fornix, pontocerebellar tract, and in white matter zones below the cingular and primary somatosensory cortices. No hypomyelination nor axonal degeneration were detected in white matter zones below the primary motor cortex or in the corticospinal tract. Massive astrogliosis was also observed in white matter associated with the somatosensory cortex, as well as enlargement of the lateral ventricules (7–10). Interestingly, the severity of hypomyelination in these adult rats correlated with the gradient of growth restriction at birth (9, 13, 14). The cerebral area in MIUH rats was reduced at striatal and hippocampal levels at P15 (7). We also found decreased neuronal densities in the somatosensory cortex, particularly inhibitory neurons, and decreased height of the cortical gray matter in adult MIUH rats. In contrast, there was no structural changes in the motor cortex (9). In addition, abnormal lamination of the parietal cortex, presumably due to premature disruption of the cortical subplate, was associated with gait disturbances in adult rats exposed to MIUH at E17-18 (12, 15). There was no obvious tissue damage, such as cystic and necrotic lesions or inflammatory cell infiltration in adult MIUH rats (7). As a neonatal index of sensorimotor reflexes, negative geotaxis was delayed in MIUH pups. P15 and adult MIUH rats exhibited spontaneous exploratory and motor hyperactivity. Adult MIUH rats displayed deficits in information encoding, and deficits in short and long term object memory tasks, but no impairments in spatial learning or working memory in watermaze tasks (7, 9, 10). These behavioral and cognitive deficits in our rodent model appear to recapitulate some symptoms commonly found in children with EP, such as attention-deficit with hyperactivity disorder (ADHD), and learning and memory deficits (4, 16, 17). In addition to neuroanatomical disturbances, the primary somatosensory maps representing the hind paw skin surfaces were topographically disrupted and disorganized in MIUH rats, compared to controls (10), likely indicative of reduced tactile abilities (18). In contrast, we found no changes in the neuroanatomical and functional organization of the primary motor cortex (10). Adult MIUH rats exhibited minor locomotor deficits on treadmill with mainly knee-ankle hyperextension compensated by hip hyperflexion, as well as increased variations in locomotor kinetics that appeared mainly related to a disorganization in the somatosensory cortex but not in the motor cortex (8, 10), as observed in children with CP (19). MIUH rats also displayed muscle weakness in their hind limbs but not in their forelimbs, mild myopathic and secondary joint changes in their hind limbs, indicative of mild signs of spasticity (7, 8) that still remain to confirm.

From the previous studies related to MIUH cited above, we wondered what may relate the various MIUH-induced events such as inflammation, WMI, sensorimotor network disorganization, minor locomotor impairments and the emergence of NDDs. We suppose that MIUH-induced intrauterine inflammation induces early and postnatal pathophysiological cascades that may involve KCC2. KCC2, a K-Cl cotransporter is the main chloride extrusion system in the central nervous system (CNS) and regulates chloride homeostasis and neuronal excitability (20–23). Another candidate in this pathophysiological cascade is the disruption of the neuromuscular interactions, especially early, and long-lasting changes in stretch reflex and muscle properties/phenotype. We hypothesized that early inflammation leads to decreased expression of KCC2, which in turn may drive spinal and cortical hyperexcitability, altered neuromuscular interactions and muscle properties, thus inducing a disorganization of the sensorimotor circuitry and locomotor impairments, and subsequent NDDs. The present study was aimed at investigating the early and longlasting mechanisms that may contribute to the disorganization of the CNS and to the subsequent emergence of locomotor impairments and NDDs in rats after MIUH. Such a better understanding of the early pathophysiological cascades that lead to EP and CP may allow us to develop new strategies of remediation and prevention.

#### MATERIALS AND METHODS

All experiments and animal use have been carried out in accordance with the guidelines laid down by NIH (NIH Publication #80-23) and EC Council Directive (2010/63/EEC). The research involving animals has been approved by the local ethics committees in Marseille (Comité d'éthique en Neurosciences INT-Marseille—CEEA #71, authorization #00265.02), Lille (Comité d'éthique Région Nord Pas-de-Calais— CEEA #75, authorization APAFIS#4732-2016031112395755), and Japan (Committee of the National Cerebral and Cardiovascular Center, Suita, Japan).

#### Intrauterine Arterial Strenosis Using Microcoils

In prior studies, the intrauterine artery stenosis was performed by using unilateral ligation at embryonic day 17 (E17) to produce prenatal ischemia and intrauterine growth retardation in rats (8–10, 13). We developed a new model of intrauterine ischemia/growth retardation based on the application of metalcoated coils (Samini Co. Ltd., Shizuoka, Japan) wrapped around the intrauterine arteries at E17 (7, 24). Briefly, under deep anesthesia with isoflurane, microcoils (inner diameter: 0.16 mm) were wrapped around each proximal artery of both ovarian sides by using the same procedure as described previously (5, 6), to produce optimized blood flow reduction or MIUH (**Figure 1**). This technique has the advantage to increase the number of MIUH pups and thus to reduce the number of used dams, compared to stenosis by ligation. The sham group was subjected to the same surgery as the MIUH group but without coil insertion. Pups were delivered by spontaneous labor and attributed to each group depending on the weight at birth. Like in previous studies, pups whose weight was below 5.5 g were considered growth retarded and part of the MIUH group.

#### Laser Speckle Blood Flowmetry

Temporal changes of intrauterine blood flow during the surgery were monitored by using a laser speckle flowmetry (Omegazone, Omegawave Inc., Tokyo, Japan) at two time points: before stenosis and 1 h after stenosis on both ovarian and vaginal sides (**Figure 1**) under isoflurane deep anesthesia. To quantify blood flow, regions of interest (ROI) on all fetuses or placentas were analyzed in 2 sham and 2 MIUH rats (**Figures 1A,C**).

#### Early in Vitro Post-activation Depression

To determine the early impact of MIUH on the functional reorganization and excitation/inhibition balance in rat pups, we assessed the alterations of the monosynaptic reflex loop using an in vitro whole spinal cord preparation at postnatal days P4–P6 (Sham, n = 11; MIUH, n = 6). The spinal cord below T8 was isolated from neonatal rats from P4 to P6, as previously described (21, 22) and transferred to the recording chamber perfused with an oxygenated (95% O2/5% CO2) aCSF composed of the following (in mM): 130 NaCl, 4 KCl, 3.75 CaCl2, 1.3 MgSO4, 0.58 NaH2PO4, 25 NaHCO3, and 10 glucose (pH 7.4; 32◦C). Extracellular recordings/stimulation were made at lumbar L5 ventral (VR5) and dorsal (DR5) roots by contact stainless steel electrodes insulated with vaseline. AC recordings from VR were amplified (×2,000) and bandpass filtered from 70 Hz to 3 kHz. Supramaximal stimulation of a lumbar dorsal root (DR5) elicited a monosynaptic response in the ipsilateral homonymous ventral root (VR5) in vitro, corresponding to the earliest component of motoneurons excitation. To determine the level of post-activation depression (PAD) at the different frequencies, we discarded responses to the first three stimulations required for the depression to occur. The responses were rectified and the areas under the curves were measured. The monosynaptic response was expressed as percentages relative to the mean response at 0.1 Hz in the same series of measurements (21).

### Postnatal KCC2 Western Blots

To detect the expression of KCC2 in the spinal cord, tissue was collected at P8 and frozen after removing the dorsal and ventral roots (Sham, n = 8; MIUH, n = 7). Samples were prepared in ice– cold lysis buffer containing 1% Igepal CA-630, 0.1% SDS, 10 mM sodium vanadate, 10 mM sodium fluoride, 10 mM sodium pyrophosphate, 1.8 mg.mL−<sup>1</sup> iodoacetamide supplemented with protease inhibitors cocktail (Complete-mini, Roche Life Science). After centrifugation step at 18,000 g for 30 min at 4◦C, the supernatant was collected, and protein concentrations were determined using DC protein assay (Biorad). Equal protein amounts (30 µg) from samples were separated by electrophoresis SDS-PAGE (4-15% CriterionTM TGX Stain-FreeTM Precast Gels, Biorad), transferred to a nitrocellulose membrane and incubated overnight at 4◦C with the affinitypurified rabbit anti-KCC2 polyclonal antibody (diluted 1:1,000; Merck-Millipore). The blot was then incubated 1 h at 24◦C with an immunoPure goat HRP-conjugated rabbit specific antibody (1:80,000; Thermo Scientific, in blocking solution of Tris buffered saline containing 5% fat-free milk powder). Bands were visualized by chemiluminescence (Merck-Millipore). Signal intensities were measured with the image analysis software Labview (BioRad). Equal amounts of protein samples were loaded, and we performed total protein normalization using stain-free imaging (Biorad), which makes proteins fluorescent directly in the gel and following transfer (see Supplemental Figure 1). The total density for each lane was measured from the blot and was used to calculate the normalizing factors. After normalization to total proteins of signal intensity for KCC2, we normalized by dividing each sample by the mean value of control samples.

### Adult in Vivo Post-activation Depression

The Hoffmann reflex is commonly used to assess primary (type Ia) afferents–mediated motor neuronal excitability (monosynaptic reflex loop) in individuals with spasticity (21, 25). The H-reflex was measured in P60 rats from both groups (Sham, n = 9; MIUH, n = 8) under deep and constant anesthesia, induced first with isoflurane and then with ketamine (100 mg.Kg−<sup>1</sup> i.p. induction) and maintained with Supplemental doses of ketamine (20 mg.Kg−<sup>1</sup> i.p.), as widely used (21, 23, 26). Rat temperature was maintained around 38◦C with a thermal pad controlled by rectal temperature probe. A transcutaneous pair of stainless stimulating needle electrodes was inserted adjacent to the tibial nerve about 1 cm above the ankle. For EMG, a pair of stainless recording electrodes was inserted into the flexor digitorum beneath the ankle and the reference electrode into the tail's skin. First, we stimulated the tibial nerve for 0.2 ms at 0.2 Hz with increasing current intensities until the

flow before (Pre) and 1 h after microcoil stenosis (Post) at both sides of the ovarian artery. (D) Blood flow changes during coils stenosis in fetuses in both dams shown in (C). The blood flow decreased 1 h after stenosis (Post) compared to that before stenosis (Pre) in most of the fetuses and placentas (not shown). The location of pups in the right (R) and left (L) horns of the uterus is denoted from 1 to the maximal number of pups in each side, with 1 at the closest location to the ovary and the last one at the closest location to the vagina. The region of interest (ROI) for blood flow measures of the fetuses is depicted in (A,C) by black open circles, with numbers that correspond to the locations of the fetus on each uterus horn.

Mmax stabilized, and determined the intensity required for a maximal H response. Tibial nerve was stimulated with trains of 20 stimulations at 0.2, 0.5, 1, 2, and 5 Hz with 2 min intervals between each train to elicit PAD. To determine the level of PAD at the different frequencies, we discarded responses to the first three stimulations required for the depression to occur. The M and H waves were rectified and the areas under the curves were measured. The H responses were expressed as percentages relative to the mean response at 0.2 Hz in the same series of measurements (21, 23).

#### In Situ Contractile Properties of the Soleus Muscle and Muscle Removal

At P28, the rats were deeply anesthetized with intraperitoneal injections of ketamine (50 mg.Kg−<sup>1</sup> ) and dexmedetomidine (Domitor, 0.25 mg.Kg−<sup>1</sup> ), prolonged if necessary by Supplemental doses. The dissection protocol was previously described (27). Briefly, all the muscles of the right hindlimb were denervated, except the soleus muscle, which was isolated from surrounding tissues. Then, the limb was immersed in a bath of paraffin oil thermostatically controlled (37◦C), and fixed with bars and pins. The soleus muscle was maintained in a horizontal position and its distal tendon was connected to a force transducer (Grass FT 10; Grass Instruments, West Warwick, RI, USA). The muscle length was adjusted to produce a maximal twitch peak tension (Pt). Contractions were induced by stimulation of the sciatic nerve (0.2-ms pulses) through bipolar platinum electrodes at twice the minimum voltage required to obtain the maximal twitch response. The following parameters were recorded: P<sup>t</sup> , time-to-peak (TTP), half-relaxation time (HRT); peak tetanic tension obtained at 100 Hz (P0). The fatigue index (FI) was calculated as the percent of the initial tension divided by the force at the end of the fatigue protocol in a series of 120 consecutive contractions (330 ms duration, 40 Hz, one train per second). At the end of the recording session, the muscle was removed for determination of the soleus muscle wet weight (MWW), frozen in liquid nitrogen and stored at −80◦C until electrophoretic analysis (see Supplemental Methods). This study was performed on 17 pups (Sham, n = 8; MIUH, n = 9).

#### Excitatory and Inhibitory Neurotransmission

To gain insights into the long-term impact of MIUH on the extraand intracellular levels of glutamate and GABA, we performed in vivo microdialysis within the sensorimotor cortex contralateral to the mapped side from P90 to 120. In other rats, Western blot analysis was used to quantify the intracellular amounts of transporters of both glutamate (vGLUT1) and GABA (vGAT) in sensorimotor cortex tissues collected contralateral to the mapped side (Sham, n = 13; MIUH, n = 16).

#### In Vivo Microdialysis

Carnegie Medecin microdialysis probes (CMA/11, Phymep, France) were implanted within the hindpaw representation of the right S1-M1 area using stereotaxic coordinates (A: −1/0, L: +2; H: 2); the left cortex was mapped. The chosen implantation site of the probe avoided large blood vessels; traces of blood were not found after probe withdrawal. The terminal ends of the probes were covered with a polycarbonate membrane. The membrane had a diameter of 0.24 mm, a length of 2 mm to allow to sample all cortical layers and a 6 KDa molecular mass cutoff. The membrane acted like a blood vessel, with extracellular molecules passing through it by diffusion gradient. A new probe was used for each rat. This procedure is detailed in Supplemental Methods.

#### Western Blotting and Quantification

The leg region from the right side of the sensorimotor cortex was collected and homogenized in TBS (Tris-HCl 50 mmol.L−<sup>1</sup> pH 7.4, NaCl 150 mmol.L−<sup>1</sup> ) containing protease inhibitors (complete EDTA free, Roche, Basel, Switzerland) using a Potter homogenizer and the ratio of 100 µg of brain pieces per mL of buffer. Homogenates were centrifuged at 500 g for 3 min at 4 ◦C. Supernatants were aliquoted and stored at −80◦C. Proteins present in each supernatant were resolved by SDS-PAGE and transferred onto nitrocellulose membranes (see Supplemental Methods and Supplemental Figure 2). Proteins of interest were then detected with specific antibodies using chemiluminescence (ECL; Pierce Biotechnology, Rockford, IL, USA). Western blot quantification was performed on scanned autoradiographies with Image J software (28). Integrative intensities minus background were plotted for each sample. Values were normalized to 1 for the highest value for one western blot.

#### DATA ANALYSIS

Data normality and homogeneity of variance were determined with the Shapiro test, Bartlett test and var.test by using R (The R Foundation for Statistical Computing, Wien, Austria). We then applied either parametric (two-tailed and paired ttests, and one-way ANOVAs with Tukey's post-hoc comparisons and two-way ANOVAs) or non-parametric (Mann-Whitney, Wilcoxon and ANOVA of Friedman) tests using either R or Prism (GraphPad Software, CA). Significance was set at p < 0.05. The investigators were blind to rearing conditions throughout the different experimental sessions until statistical comparisons were performed.

#### RESULTS

#### Coil Stenosis Induces a Mild Intrauterine Hypoperfusion

Our preliminary results showed that microcoil stenosis of the ovarian artery at both ovarian sides decreased the blood flow in fetuses to 17.4% and placentas to 15.3% of pre-stenosis level in average (**Figures 1C,D**). The level of blood supply was mostly the same across each fetus (arbitrary unit of blood flow measurement: 33.7 ± 0.8 – mean ± SEM) and each placenta (33.1 ± 1.4) before stenosis (**Figures 1C,D**), and compared to sham conditions (fetus: 34.9 ± 0.7; placenta: 30.9 ± 1.2; **Figures 1A,B**). Applied to the bilateral ovarian branches of the uterine artery at E17, our new technique of microcoil stenosis produced a consistent and reproducible MIUH across all the placentas and fetuses. The hypoperfusion apparently resolved within 3 days at E20 [data not shown, see (7)].

#### MIUH Leads to Early Hyperreflexia

In sham pups, the amplitude of the monosynaptic response decreased when the dorsal root was stimulated repeatedly. From P4 to P6, this effect became stronger with increasing stimulation frequency from 0.1 to 5 Hz (p < 0.0001; **Figure 2A**), similar to the in vitro PAD observed previously (21–23). In contrast, MIUH pups at P4–P6 showed a decrease in the response amplitude with increasing stimulation frequencies, but much lesser than in sham rats (p < 0.003). Compared to sham rats at each frequency, MIUH pups showed a significant reduction of the PAD at P4–P6 (**Figure 2A**), suggestive of an early increase of stretch reflex and hyperexcitability within the lumbar spinal cord and early signs of spasticity as well, as shown in pups after complete spinal cord section (21, 23).

#### MIUH Decreases the Early Amounts of KCC2

Protein analysis was performed on the whole lysate (total fraction) from the lumbar spinal cord at P8, followed by immunoblotting with specific antibodies against KCC2 protein. The monomeric and oligomeric forms of KCC2 were detected at 140 KDa and upper than 250 KDa, respectively (See Supplemental Figure 1). Compared to sham rats, the amount of KCC2 in the total fraction decreased in MIUH rats (p < 0.006; data not shown). More interestingly, the monomeric and oligomeric forms were reduced after MIUH, especially the latter (p < 0.04; **Figure 3**). Monomeric and oligomeric forms correspond to the inactive and active forms of KCC2, respectively (20). This result suggests a decrease in both the active and inactive forms of KCC2 after MIUH, especially the former, responsible for chloride homeostasis and cellular excitability (20, 21).

#### MIUH Leads to Enduring Hyperreflexia in Adulthood

To evaluate the persistence of spasticity in adulthood after MIUH, corresponding to hyperexcitability in the spinal circuitry, we assessed the changes in the Hoffmann reflex (H-reflex; **Figure 2B**) (21, 23, 29). In adult sham rats, the H-reflex was depressed by repeated nerve stimulation at frequencies increasing from 0.2 to 5 Hz (p < 0.0001; **Figure 2B**), corresponding to PAD; (21, 23), as observed above in MIUH rats at P4–P6. In contrast, the PAD was reduced in adult MIUH rats (p < 0.0001), as illustrated by the lack of significant differences of the monosynaptic H-reflex with increasing frequencies (**Figure 2B**). Surprisingly, the monosynaptic H-reflex even tended to increase at 0.5 and 1 Hz, compared to the reference at 0.1 Hz (**Figure 2B**). When we compared the H-reflex PAD between sham and MIUH rats, we found significant decreases with increasing frequencies from 0.2 to 5 Hz (**Figure 2B**), suggesting the enduring presence of hyperreflexia and spasticity in adult MIUH.

### MIUH Alters Adult Contractile Properties of the Soleus Muscle

At P28, no significant difference was observed in morphological parameters between sham and MIUH rats. The body weight (BW) was near 90 g in sham and MIUH groups. It should be noted that soleus muscle tended to be smaller in MIUH rats (muscle wet weight, MWW: −10%, p = 0.09, tendency; **Table 1**). However, when MWW was normalized to BW, the value was similar in both groups.

The mechanical properties of the soleus muscle observed in situ revealed that the soleus muscle phenotype was unchanged in MIUH rats. The rate of force development was similar, as was the

FIGURE 2 | Decreased post-activation depression (PAD) performed in vitro in pups from P4 to P6 and in vivo in adult rats that both were submitted to mild intrauterine hypoperfusion (MIUH) at E17. (A) Mean (± SEM) relative amplitudes of the monosynaptic reflex at different stimulation frequencies in MIUH pups, compared to sham rats. In sham rats, the monosynaptic amplitude decreased with increasing frequencies (i.e., PAD) from 0.1 Hz (set as reference = 100%) to 5 Hz [(ANOVA of Friedman, x<sup>2</sup> = 40.1, df = 4; p < 0.0001) and Wilcoxon's post-hoc comparisons:\*\*p < 0.01; \*\*\*p < 0.0001; ns, non-significant] while this amplitude decreased lesser in MIUH pups [(ANOVA of Friedman, x<sup>2</sup> = 21.1, df = 4; p < 0.003) and Wilcoxon's post-hoc comparisons:\*p < 0.05; ns, non-significant]. The comparison of the monosynaptic reflex amplitude between sham and MIUH rats showed significant reduction of the PAD at stimulation frequency (Wilcoxon: b, p < 0.01) from 0.1 Hz. (B) Depression of the H-reflex (PAD), expressed as percentages, over consecutive stimulations from 0.2 Hz (set as reference = 100%) to 5 Hz in sham rats [(ANOVA of Friedman, x<sup>2</sup> = 69.6, df = 4; p < 0.0001) and Wilcoxon's post-hoc comparisons:\*\*p < 0.01] while the H-reflex was increased at 0.5 and 1 Hz and did not reduce at higher frequencies of stimulation in MIUH rats [(ANOVA of Friedman, x<sup>2</sup> = 12.5, df = 4; p < 0.01) and Wilcoxon's post-hoc comparisons: ns, non-significant]. When we compared the H-reflex PAD between sham and MIUH rats, we found significant decreases of PAD at stimulation frequencies from 0.2 to 5 Hz (Wilcoxon: a, p < 0.05; b, p < 0.01) in the latter group.

ratio of subtetanic tension at 20 Hz relative to P<sup>0</sup> (P20/P0), which was about 50% in both groups (**Table 1**). This parameter is an indicator of muscle type; a low value (0.20–0.30) is characteristic of a fast muscle, whereas a high value (0.70-0.80) indicates a slow muscle. Interestingly, the interval from peak tension to half peak

tension (HRT) was reduced by 18% in MIUH rats (p < 0.001; **Table 1**).

form of KCC2 at P8. The mean quantity of oligomeric KKC2 was also decreased in MIUH rats relative to sham rats, but in a greater extent (U = 4; p < 0.04) than for the monomeric form in (B). \* p < 0.05; \*\* p < 0.01.

Single twitch tension (Pt) was decreased by 21% (p = 0.02) in MIUH rats (**Table 1**). This change may due in part to the muscle atrophy, since the decrease was only −14% (p = 0.07, tendency) when P<sup>t</sup> was normalized to MWW. In the same line, the maximal tetanic force P<sup>0</sup> tended to decrease (−9%; p = 0.07, tendency; **Table 1**), but remained unchanged when normalized to MWW (−12%; p: n.s.).

The muscle was resistant to fatigue, whatever the group (fatigue index: 86 ± 9% in sham group and 78 ± 26% in MIUH group). This high resistance may be due to the fact that muscles were composed mainly of 1 and 2A MHC isoforms (**Table 1**). In contrast to sham rats where it was almost absent, a low proportion of neonatal MHC isoform was still present in the TABLE 1 | Contractile and morphological properties of soleus muscles in sham and MIUH rats.


Pt, single twitch tension; P0, tetanic tension at 100 Hz; P20, tetanic tension at 20 Hz; FI, fatigue index; BW, body weight; MWW, muscle wet weight. \*, \*\*, \*\*\*A significant difference with respect to sham rats at p < 0.05, p < 0.01, p < 0.001, respectively. † A tendency (p < 0.1). Mean ± SD.

soleus muscle of MIUH rats at P28 (p < 0.01; **Table 1**). Thus, immature isoform of MHC persisted in the soleus of MIUH young-adult rats, compared to shams. It is worth noting that the proportion of MHC 2A isoform was 25% higher in MIUH group, while MHC 1 decreased by −17%. However, the overall variation was non-significant because of high inter-individual variations (**Table 1**). Such a transition from slow to fast isoform may explain the decrease in HRT after MIUH.

#### MIUH Induces Hyperexcitability in the Adult Sensorimotor Cortex

To assess the balance between excitation and inhibition in the hind paw sensorimotor cortical area, we analyzed the contents of dialysates in extracellular glutamate and GABA obtained from P90 to P120 during in vivo microdialysis determined by gradient HPLC coupled to laser detection. The extracellular concentration of glutamate was greater in adult MIUH rats than in sham rats (p < 0.04); whereas, the extracellular concentration of GABA did not differ significantly between the two groups of rats (**Figure 4A**).

To complete microdialysis analysis, we performed semi quantitative Western blots in other adult animals, using specific antibodies against vesicular glutamate transporter (vGlut1) and vesicular GABA transporter (vGAT), reliable markers of excitatory and inhibitory transmission, respectively. The amount of vGlut1 increased of about 40% in MIUH rats relative to controls (p < 0.0001; **Figures 4B,C**). In contrast, the amount of vGAT decreased of about 15% (p < 0.02; **Figures 4B,C**). Thus, MIUH induces an increase in glutamate release in adult rats while GABAergic levels were reduced, suggestive of hyperexcitability in the hind paw area of the sensorimotor cortex.

### DISCUSSION

This study is the first to show the early and enduring functional disorganization of neuromuscular interactions after MIUH in a rat model that recapitulates the diversity of the symptoms observed in children with encephalopathy of prematurity (6). MIUH in rats led to (1) reduced PAD at P4–P6, indicative of early hyperreflexia and spasticity, (2) an early reduction of the oligomeric (active) and monomeric (inactive) forms of KCC2, which regulates chloride homeostasis and cell excitability, (3) reduced PAD in adult rats, suggestive of enduring hyperreflexia and spasticity in adulthood, (4) changes in muscle contractile properties and phenotype in young-adults, and (5) cortical hyperexcitability in the adult sensorimotor cortex devoted to the hind limb representation.

#### MIUH, Hyperexcitability, and Neuroinflammation

The presence of spasticity, hypertonicity, contractures or other related clinical signs is relatively common in children with CP (2) or in patients with stroke or spinal cord injury (SCI) (30–32) and has been reproduced in animal models of SCI (21–23). MIUH induced a decrease in PAD at P4– P6, a reliable correlate of early hyperreflexia and spasticity, as described in SCI rats (21–23). It is widely accepted that at least two mechanisms are responsible for such hyperreflexia after SCI: increased excitability in motoneurons and a reduction of inhibition, the so-called disinhibition, within the lumbar spinal network (21, 22). Disinhibition of the myotatic reflex seems to be caused by reduced expression of the active or oligomeric form of KCC2, which abnormally increases the intracellular concentration of chloride ions and reverses the effect of IPSPs from hyperpolarization to depolarization (21–23). In the present study, MIUH decreased the expression of inactive and active forms of KCC2, which induced early hyperreflexia and spasticity, like after SCI in neonate rats (21, 22). These results confirm and extend our previous study in which gastrocnemius histopathology indicated signs of spasticity (8). In addition, we also found a reduction of PAD in adult MIUH rats that suggests the persistence of increased stretch reflex, muscle hyperreflexia and likely spasticity in adulthood. It is worth noting that early and enduring increased stretch reflex and spasticity may be at the origin and persistence of the minor locomotor impairments found in adult MIUH (8).

It is well admitted that preterm infants and children with CP exhibit neuroinflammation processes from in utero to postnatal life (1, 5, 11, 17). MIUH induced early changes marked by the upregulation of several proteins related to inflammation and ischemic injury in placenta and the downregulation of mRNAs associated with axon and astrocyte growth in the fetal brains (7). As putative candidate, calpains, are intracellular proteases activated by calcium influx during glutamateinduced excitotoxicity and regulate cellular homeostasis, neuronal activity or apoptosis during development and in the mature CNS. Appropriate calpain activity is crucial for typical neurodevelopment, including learning and memory (33). Excessive calpain activity degrades proteins important for neural function (32), such as KCC2 in the hippocampus and layer IV of the cerebral cortex after MIUH at E18 (34) or in the spinal network after SCI (35). Indeed, recent studies have shown the deleterious contribution of excess calpain activity to cleave KCC2, which becomes inactive and thus contribute to the excitation/inhibition imbalance toward hyperexcitability (36) and to upregulate the persistent sodium current in motoneurons (37), thus leading to the development of hyperreflexia and spasticity, as shown after SCI (37, 38). Post-mortem cerebral samples from human preterm infants with WMI showed a loss of KCC2 expression (39).

Therefore, we suppose that MIUH-induced intrauterine inflammation (7) comes along with overactivity of calpains, which cleaves KCC2. Reduced expression of inactive (monomeric) and active (oligomeric) forms of KCC2 at P8 leads to hyperexcitability in the lumbar spinal cord, which likely contributes to postnatal hyperreflexia and spasticity, probably at the origin of the minor locomotion disorders observed later at P30 and P65 in MIUH rats (8).

### Enduring Perturbations of Neuromuscular Interactions

To extend our results showing the early and enduring presence of hyperreflexia, spasticity and minor locomotor deficits, we investigated the contractile properties of the soleus muscle in young-adult MIUH rats to better understand the complex neuromuscular interplay after MIUH. At the muscle level, the most striking result was a drastic reduction in HRT in MIUH rats. Such a decrease has not been observed in plantarflexor muscles of children with spastic diplegia (40). The reduction of HRT suggests that following muscular contraction, the transport of calcium from the cytosol into the lumen of the sarcoplasmic reticulum by the sarco(endo)plasmic reticulum calcium ATPase (SERCA) is faster. This could be the result of a change in SERCA isoform expression toward fast isoforms (41). It could also be due to an increase in the activity of the SERCA pump. This latter hypothesis is sustained by the fact that an acute muscle ischemia results in an increase in the maximal SERCA activity (42). The modification of the HRT could also result from a proliferation of the sarcoplasmic reticulum. Indeed, a proliferation of the sarcoplasmic reticulum has been described in the soleus muscle of rats submitted to 2 and 4 weeks of hind limb unloading (43), as well as after denervation (44). Another parameter that influences the HRT is the muscle stiffness. Compliance accelerates relaxation in striated muscle by allowing myosin heads to move relative to binding sites on actin (45). However, this hypothesis is very unlikely since many papers have reported higher tissue stiffness in the triceps surae of patients with CP (46).

extracellular levels assessed by in vivo microdialysis in adult rats. Compared to sham rats, MIUH induced a significant increase in glutamate release (U = 8.0; p < 0.04) while the release in GABA did not differ between the two groups (U = 18.0; p = 0.4, n.s.). (B) Typical immunoblots for vGlut1, vesicular transporter of glutamate, and vGAT, vesicular transporter of GABA in the two groups of rats. (C) Plots of glutamate and GABA intracellular levels assessed by Western blots of vGlut1 and vGAT, respectively. Note that both measures indicate an increase in glutamate levels [t (1,37) <sup>=</sup> 4.8; <sup>p</sup> <sup>&</sup>lt; 0.0001] while GABA levels were reduced [<sup>t</sup> (1,21) <sup>=</sup> 2.4; <sup>p</sup> <sup>&</sup>lt; 0.02] after MIUH, compared to sham rats. \*p < 0.05; \*\*\*p < 0.001.

The other contractile kinetic parameters (TTP, P20/P0) were not modified. In adult rats, the P20/P<sup>0</sup> ratio is near 80% for the soleus muscle, characteristic of a slow-type muscle. A value of 50% as observed herein shows that muscle maturation is still not achieved at P28. In addition, the persistence of neonatal MHC isoform suggests that MIUH delays the decline in expression of the neonatal isoform that occurs during typical development (47). A delay of maturation of the growth-associated shift of fiber phenotype toward slow type has also been observed in rats submitted to hypoactivity through hind limb unloading (48) or after denervation (47). These results emphasize the deleterious impact of abnormal motor commands in the maturation of muscle properties and in the expression of MHC isoforms (49). In the same way, HRT has been shown to decrease after deafferentation in rats, demonstrating that abnormal afferent inputs from muscle proprioceptors has in return an impact on muscle kinetic properties (27). Hyperreflexia and increased stretch reflex may occur as a compensatory process to abnormal

somatosensory activity and muscle disuse (50). In fact, a reciprocal interplay exists between motoneuron firing, muscle properties/phenotype and proprioceptive feedback/reafference (27, 51) that contributes to the development and refinement of the sensorimotor circuitry.

In the hind limb area of adult sensorimotor cortex, MIUH induced an excitation/inhibition imbalance toward increased glutamate excitability while GABAergic inhibition was unchanged or slightly decreased in adult rats. Such a cortical imbalance may be the continuity and spread of the early disinhibition and hyperreflexia described at P8 in the spinal cord after MIUH. In fact, reduced KCC2 expression has been revealed in cortical layer IV at P8 and P28 in a comparable model of MIUH at E18 (52). This imbalance also corroborates our previous study in adult MIUH rats (9), in which we found a decrease in the inhibitory interneuron density in the primary somatosensory cortex and the degradation of both the neuronal properties and somatotopic maps, devoted to represent the hind limb skin surfaces. This degradation was mainly characterized by a reduction of the map size, enlarged and multiple receptive fields that encompassed several toes or pads simultaneously, leading to a somatotopic, and topographic disorganization (9). In addition, reduced KCC2 function/expression has also been evidenced to contribute to WMI (36), as previously observed in our MIUH model of EP (8–10). Since we found WMI below the somatosensory cortex, it is possible that early somatosensory inputs from the periphery were not spatiotemporally synchronized to induce appropriate spinal and cortical plasticity (6, 53). It is now widely accepted that WMI-induced changes in conduction velocity mediate abnormal transmission and integration of afferent and efferent information, leading to brain dysfunctions, as observed in several pathologies including autism and schizophrenia (54, 55). One can speculate that WMI below the somatosensory cortex may alter timing precision of somatosensory reafference arising from early spontaneous movements and locomotion to the immature sensorimotor circuitry, including the lumbar spinal cord, and sensorimotor cortex (**Figure 5**), associated with other subcortical structures (56). Because precision in timing of action potentials is fundamental in CNS plasticity (57, 58), early somatosensory inputs may have induced the postnatal and enduring disorganization within the sensorimotor circuitry after MIUH, described here and previously (6, 8–10, 59). Several studies have shown that abnormal processing and integration of afferent information in the somatosensory cortex is sufficient to drive disturbances in motor planning and execution (19, 60, 61).

Therefore, we speculate that early inflammation induced by MIUH at E17 led to reduced KCC2 expression and altered proliferation of oligodendrocyte precursors (1–3) that both contributed to postnatal WMI and lumbar hyperreflexia (**Figure 5**). Thus, postnatal WMI may induce sensorimotor circuitry disorganization, through inappropriate information synchrony and maladaptive plasticity (6, 53, 62, 63). In addition to lumbar hyperreflexia observed at P4–P6, such a disorganization of the sensorimotor circuitry from the spinal cord to the cortex may in turn produce abnormal motor commands, as suggested above in young-adult MIUH rats, which in turn may participate to induce and maintain altered muscle synergies, maturation and functional properties, as well as locomotor disturbances and musculoskeletal histopathology into a self-perpetuating cycle (**Figure 5**). In fact, we previously showed that postnatal movement restriction through transient hind limb immobilization during development induced musculoskeletal histopathology, locomotor disorders, and

muscle overactivities including spasticity that interplayed into a self-perpetuating cycle, which likely contributed to maintain or even aggravate these disturbances (59). Thus, minor gait impairments related to MIUH likely produced atypical somatosensory feedback/reference to the immature sensorimotor circuitry that in turn participated to the functional disorganization of the sensorimotor circuitry in a second selfperpetuating cycle, while MIUH-related neuroinflammation contributed to the structural disorganization of the sensorimotor circuitry, including WMI (**Figure 5**).

#### Functional Implications

Early inflammation appears to be crucial in the pathogenic cascades and to contribute in the primary and secondary brain injuries, as well as in repair or recovery after insult events. Immunomodulatory interventions targeting inflammation seem beneficial in preclinical models and might have translational potentials (1, 5, 38). It appears crucial to develop new strategies to reinstate excitation/inhibition balance within the sensorimotor circuitry as early as possible after the insult and inflammation cascade inception. As a promising lead, erythropoietin (EPO) restores typical KCC2 expression in the hippocampus and brain following MIUH at E18 in rats (34, 36), and in motoneurons after neonatal stroke (64). In addition to neuroprotective properties (65), EPO modulates excess calpain activity via calpastatin (34, 36), reduces caspase activation (66), restores oligodendrogenesis, survival and process extension (67) after MIUH, and also improves motor recovery and neuronal regeneration after SCI (68). In addition, EPO is already used in clinical trials for many brain diseases including stroke, and in a clinical trial for extremely preterm infants (PENUT Trial).

Finally, our preclinical model based on MIUH contributes to elucidating the putative involvement of WMI on alterations in neural activity and plasticity, function, and refinement of

#### REFERENCES


the CNS. Further studies are required to enlighten the role WMI and altered neuromuscular interplay in the emergence of encephalopathy of prematurity and CP symptoms, particularly the development of NDDs such as ADHD and learning deficits, and to develop new strategies of prevention and rehabilitation.

### AUTHOR CONTRIBUTIONS

J-OC planned and contributed to all the experiments and wrote the manuscript. MD and NT-L performed in vivo microdialysis. FrC performed western blotting in the sensorimotor cortex. LB, PC, FloC, CB, and SL performed western blotting of KCC2. VM, BB, and M-HC assessed the contractile properties. JP, FloC, and J-OC performed early and adult PAD. YO, MT, and J-OC assessed intrauterine blood flowmetry. J-OC, M-HC, and FloC performed all statistics. J-OC, MD, CB, SL, BB, MT, M-HC, and FloC contributed to write, edit and revise the manuscript.

#### ACKNOWLEDGMENTS

The authors are grateful to Drs H Bras, M Kerzoncuf, R Bos, and F Boubred for helpful discussions on the experiments. This work was supported by the Centre National de la Recherche Scientifique (CNRS), la Fondation Paralysie Cérébrale (previously Fondation Motrice), the Cerebral Palsy Alliance, Aix-Marseille Université, the National Cerebral and Cardiovascular Center and Ministère des Affaires Etrangères et du Développement International.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00423/full#supplementary-material


development and behavioral activity in rats. Biol Neonate (2001) 80:81–7. doi: 10.1159/000047125


**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 Coq, Delcour, Ogawa, Peyronnet, Castets, Turle-Lorenzo, Montel, Bodineau, Cardot, Brocard, Liabeuf, Bastide, Canu, Tsuji and Cayetanot. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Effects of Intrauterine Inflammation on Cortical Gray Matter of Near-Term Lambs

Vanesa Stojanovska1†, Anzari Atik 1†, Ilias Nitsos <sup>1</sup> , Béatrice Skiöld<sup>2</sup> , Samantha K. Barton1,3, Valerie A. Zahra<sup>1</sup> , Karyn Rodgers <sup>1</sup> , Stuart B. Hooper 1,4 , Graeme R. Polglase1,4‡ and Robert Galinsky 1,4 \* ‡

United Kingdom, <sup>4</sup> Department of Obstetrics and Gynaecology, Monash University, Clayton, VIC, Australia

#### Edited by:

Olivier Baud, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Roland H. Hentschel, Universitätsklinikum Freiburg, Germany Marina Colella, Institut National de la Santé et de la Recherche Médicale (INSERM), France

#### \*Correspondence:

Robert Galinsky robert.galinsky@hudson.org.au

†These authors have contributed equally to this work.

‡Co-senior authorship.

#### Specialty section:

This article was submitted to Neonatology, a section of the journal Frontiers in Pediatrics

Received: 09 January 2018 Accepted: 01 May 2018 Published: 15 June 2018

#### Citation:

Stojanovska V, Atik A, Nitsos I, Skiöld B, Barton SK, Zahra VA, Rodgers K, Hooper SB, Polglase GR and Galinsky R (2018) Effects of Intrauterine Inflammation on Cortical Gray Matter of Near-Term Lambs. Front. Pediatr. 6:145. doi: 10.3389/fped.2018.00145

Introduction: Ventilation causes cerebral white matter inflammation and injury, which is exacerbated by intrauterine inflammation. However, the effects on cortical gray matter are not well-known. Our aim was to examine the effect of ventilation on the cerebral cortex of near-term lambs exposed to intrauterine inflammation.

<sup>1</sup> The Ritchie Centre, Hudson Institute of Medical Research, Clayton, VIC, Australia, <sup>2</sup> Department of Women's and Children's Health, Karolinska Institute, Stockholm, Sweden, <sup>3</sup> Centre of Clinical Brain Sciences, University of Edinburgh, Edinburgh,

Method: Pregnant ewes at 119 ± 1 days gestation received an intra-amniotic injection of saline or lipopolysaccharide (LPS; 10 mg). Seven days later, lambs were randomized to either a high tidal volume injurious ventilation strategy (INJSAL N = 6, INJLPS N = 5) or a protective ventilation strategy (PROTSAL N = 5, PROTLPS N = 6). Respiratory parameters, heart rate and blood gases were monitored during the neonatal period. At post-mortem, the brain was collected and processed for immunohistochemical assessment. Neuronal density (NeuN), apoptotic cell death (caspase 8 and TUNEL), microglial density (Iba-1), astrocytic density (GFAP), and vascular protein extravasation (sheep serum) were assessed within the frontal, parietal, temporal and occipital lobes of the cerebral cortex.

Results: A significant reduction in the number of neurons in all cortical layers except 4 was observed in LPS-exposed lambs compared to controls (layer #1: p = 0.041; layers #2 + 3: p = 0.023; layers #5 + 6: p = 0.016). LPS treatment caused a significant increase in gray matter area, indicative of edema. LPS+ventilation did not cause apoptotic cell death in the gray matter. Astrogliosis was not observed following PROT or INJ ventilation, with or without LPS exposure. LPS exposure was associated with vascular protein extravasation.

Conclusion: Ventilation had little effect on gray matter inflammation and injury. Intrauterine inflammation reduced neuronal cell density, caused edema of the cortical gray matter, and blood vessel extravasation in the brain of near-term lambs.

Keywords: chorioamnionitis, intrauterine inflammation, gray matter, ventilation, neuronal loss

## INTRODUCTION

Preterm birth (defined as <37 weeks gestation) affects approximately 15 million births worldwide annually, and is a major cause of neonatal morbidity and mortality (1, 2). Preterm infants have underdeveloped lungs, thus, there is an increased requirement for respiratory support in the delivery room. Whilst respiratory support is a life-saving intervention, manual or automated resuscitation techniques are poorly controlled, and this can lead to injurious effects on the pulmonary, cardiovascular, and cerebral systems (3–6). Recent studies have shown that the initiation of respiratory support causes systemic inflammation and haemodynamic instability resulting in brain inflammation and injury (5, 7). This condition, termed "ventilation-induced brain injury" (VIBI), is associated with diffuse white matter gliosis and compromised blood brain barrier integrity (8). Importantly, improving ventilation in the delivery room can reduce markers of brain inflammation and injury (7, 9). However, these studies have focused extensively on the white matter. The acute effect of ventilation on cortical gray matter in preterm neonates remains unknown. This is a significant gap in research, given the vulnerability of gray matter to injury in the near-term brain (10). Injury to the cortical gray matter can contribute to cognitive and physical disability such as cerebral palsy, a neurodevelopmental disorder commonly associated with preterm birth and intrauterine infection and inflammation (11, 12).

Furthermore, approximately 60% of preterm neonates are exposed to chorioamnionitis and display an increased risk and severity of neurological complications (9, 13–17). There is a clear association between white matter injury following preterm birth and intrauterine infection and inflammation (7, 14, 18–22). We previously showed that improving the ventilation strategy in lambs exposed to intrauterine inflammation, via intra-amniotic LPS, did not reduce VIBI in the white matter (7). However, little is known about the effects on the cortical gray matter.

We aimed to assess the effects of ventilation on the cortical gray matter, and to determine whether intrauterine inflammation exacerbates cortical gray matter inflammation and injury. The second aim was to determine whether a protective ventilation strategy could mitigate markers of inflammation and injury in the cortical gray matter in preterm lambs, in the presence or absence of intrauterine inflammation.

### MATERIALS AND METHODS

#### Ethical Approval

The experimental protocol was conducted according to the guidelines established by the National Health and Medical Research Council of Australia and was approved by the Monash Medical Centre animal ethics committee (Animal Ethics #MMCA-2015-37).

#### Preterm Delivery and Ventilation

Ultrasound guided intra-amniotic (IA) injection of Lipopolysaccharide (LPS) (10 mg/d; from Escherichia coli 055:B5; Sigma-Aldrich, Australia) or saline was administered

At 126 ± 1 days, pregnant ewes were anesthetized with intravenous (IV) sodium thiopentane, and inhalation of isoflurane (1.5–3.0% in 100% oxygen, Bomac Animal Health, NSW, Australia), and underwent cesarean section. Lambs were exposed, and polyvinyl catheters containing heparinised saline were placed into a jugular vein and carotid artery, for infusion of analgesia and for withdrawing blood for bloodgas analysis. Lambs were intubated (cuffed 4–4.5 mm) then randomly assigned to receive either a "protective ventilation" strategy (PROTSAL, N = 5 or PROTLPS, N = 6) or an "injurious ventilation" strategy (INJSAL, N = 6 or INJLPS, N = 5) as described previously (9).

Briefly, the "protective ventilation' strategy included prophylactic surfactant (100 mg/kg, Curosurf, Chiesi Pharma, Italy), one sustained inflation for 30 s with a peak inflation pressure (PIP) of 35 cmH2O (Neopuff; Fisher and Paykel Healthcare, Panmure, Auckland, New Zealand), followed by ventilation (Babylog 8,000+; Dräger, Lübeck, Germany) using volume guarantee mode with a set tidal volume (VT) of 7 mL/kg, and a positive end expiratory pressure (PEEP) of 5 cmH2O for 90 min. The "injurious ventilation' strategy targeted a V<sup>T</sup> of 10–12 mL/kg for the first 15 min, with 0 PEEP, with a max PIP set at 50 cmH2O to prevent pneumothoraxes. At 15 min lambs in the injury group were placed on volume guarantee mode with a V<sup>T</sup> of 7 mL/kg, and a PEEP of 5 cmH2O for the remainder of the ventilation period. The fraction of inspired oxygen was initially set at 0.4 in both groups and then subsequently adjusted to maintain arterial oxygen saturation (SaO2) between 88 and 95%. Respiratory rate was adjusted to maintain partial pressure of carbon dioxide (PaCO2) between 45 and 55 mmHg. Lamb wellbeing was monitored by frequent arterial blood gas measurement (ABL30, Radiometer, Copenhagen, Denmark). Ventilator parameters, including PIP, mean airway pressure (PAW) and V<sup>T</sup> were recorded in real time (PowerLab; ADInstruments, Castle Hill, NSW, Australia). Physiological parameters including arterial oxygen saturation, heart rate (Massimo, Irvine, CA) and cerebral oxygenation (SctO2, by Near Infrared Spectroscopy: Casmed, USA) were similarly recorded.

#### Tissue Collection

After 120 min of ventilation, lambs were euthanized with an overdose of sodium pentobarbitone (100 mg/kg IV). Brains from the PROTSAL and INJSAL groups were transcardially perfused in situ with isotonic saline and 4% paraformaldehyde in 0.1 M phosphate buffer (PFA; pH 7.4) and left in fixative overnight. The brain was halved and then cut coronally into 5 mm thick blocks. Blocks of the right cerebral hemisphere were then further fixed in 4% PFA (4 days, 4◦C) and embedded in paraffin. Brains from the PROTLPS and INJLPS groups were excised, halved along the midline and the right cerebral hemisphere was immersion fixed in 4% PFA (overnight, 4◦C). The hemisphere was then cut coronally into 5 mm thick blocks and further fixed in 10% PFA (4 days; 4◦C).

#### Immunohistochemistry

Ten Micrometer sections from equivalent sites from each lobe (frontal, parietal, temporal and occipital) of the right cerebral hemisphere (four sections/animal) were reacted with the following antibodies: mouse anti-NeuN (1:200, MAB377; Millipore) to identify neural nuclei, rabbit anti-caspase 8 (1:100, orb10241; Biorbyt, USA) to label for apoptotic cell death, rabbit anti-Iba-1 (1:1,500, 019-19741; WAKO Pure Chemical Industries, Osaka, Japan) to identify microglia, rabbit anti-GFAP (1:1,000, Zo2334; DAKO, Carpinteria, CA) to identify astrocytes, and rabbit anti-serum albumin (1:1,000, Accurate Chemical and Scientific Corporation, USA) to assess blood brain barrier permeability. All sections were incubated with appropriate Biotinylated secondary antibodies (1:200) and reacted using the avidin-biotin complex elite kit (Vector Laboratories, Burlingame, CA). The colorimetric TUNEL system (Promega, Madison, WI) was also used to identify apoptotic cell death (24).

#### Imaging and Tissue Analysis

All brain sections were imaged using a slide scanner (Leica Brightfield Aperio ScanScope AT Turbo). Analyses were performed on coded slides (observer blinded to treatment) using either ImageScope (Aperio Technologies, California, USA; NeuN, sheep serum, TUNEL analyses) or ImageJ (NIH image, Bethasda, Maryland, USA; Iba-1 and GFAP analyses). Immunohistochemical analyses were performed on one section from each of the frontal, parietal, temporal and occipital lobes from each lamb; the same regions of gray matter were assessed in each animal. NeuN<sup>+</sup> neurons were counted in one field of view (FOV; 0.56 mm<sup>2</sup> ) in three gyri as previously described (25). Each field was further divided into four bins for the analysis of the different cortical layers (bin 1: cortical layer 1; bin 2: layers 2 + 3; bin 3: layer 4; bin 4: layers 5 + 6). Gray matter area was measured in NeuN labeled sections using Aperio ImageScope software. Gray matter area was determined by measuring the total combined area (mm<sup>2</sup> ) from the four bins used to count NeuN<sup>+</sup> neurons within each cortical layer. Caspase 8<sup>+</sup> and TUNEL<sup>+</sup> cells were counted throughout the cortical gray matter. Iba-1<sup>+</sup> resting (ramified) and activated (amoeboid) microglia (distinguished by morphology) and GFAPimmunoreactive astrocytes were counted in random fields of view throughout the cortical gray matter (12 fields/section/lamb; FOV: 0.140 mm<sup>2</sup> ). The total number of vessel profiles with serum albumin extravasation within the cortical gray matter were counted, and expressed as the total number of leaky vessels per entire cortical gray matter area.

#### Statistical Analysis

Physiological data was analyzed using a 2-way repeated measures ANOVA using treatment (LPS vs. SAL) and ventilation (INJ vs. PROT ventilation) as the variables. The histological data was analyzed using a 2-way ANOVA. Post-hoc analysis was conducted using the Holm-Sidak test. All data are presented as mean ± SEM, values of p < 0.05 were considered statistically significant.

### RESULTS

#### Ventilation and Oxygenation

Tidal volume was significantly higher from 4 to 15 min in INJSAL and INJLPS compared to PROTSAL and PROTLPS ( <sup>∗</sup>p < 0.05; **Figure 1A**). There was no difference from 15 min when INJ lambs were placed on the PROT ventilation strategy. Similarly, PIP was significantly higher in the INJ ventilated lambs compared to those receiving the PROT strategy from 3 to 15 min (∗p < 0.05; **Figure 1B**).

Overall, group mean PAW tended to be higher in INJLPS ( <sup>∗</sup><sup>p</sup> <sup>&</sup>lt; 0.05) and PROTLPS ( <sup>∗</sup>p < 0.05) lambs compared to PROTSAL lambs. PAW was significantly higher in INJLPS and PROTLPS lambs at 3–7 min compared to PROTSAL lambs (**Figure 1C**)**.** Heart rate transiently increased in INJSAL lambs at 10 min compared to the other groups (**Figure 1D**) but no other differences in heart rate were observed.

PaCO<sup>2</sup> was not different between groups throughout the ventilation procedure (**Figure 2A**). PaO<sup>2</sup> in the INJLPS group was significantly higher at 15 and 30 min compared to the INJSAL ventilation cohort (∗<sup>p</sup> <sup>&</sup>lt; 0.05; **Figure 2B**). Similarly, PaO<sup>2</sup> was significantly higher in the PROTLPS group at 15 compared to the INJSAL group at 30 min (∗<sup>p</sup> <sup>&</sup>lt; 0.05; **Figure 2B**). No other differences were observed. SaO<sup>2</sup> was not different between the groups throughout the ventilation procedure (**Figure 2C**). SctO<sup>2</sup> was lower in INJLPS lambs compared to the PROTLPS group at 1 min (∗p < 0.05; **Figure 2D**) but no other differences were observed.

#### Neuronal Density

The number of NeuN<sup>+</sup> neurons from all 6 cortical layers were counted (total 1 mm<sup>2</sup> area) (**Figures 3A–D**). A significant reduction in the number of neurons in all cortical layers except 4 was observed in LPS lambs compared to controls (layer #1: p = 0.041; layers #2 + 3: p = 0.023; layers #5 + 6: p = 0.016) (**Figure 3E**). No significant differences in the number of NeuN<sup>+</sup> neurons in cortical layer #4 were observed (**Figure 3E**). Furthermore, gray matter area was significantly higher in the LPS-treated groups (p < 0.0001) when compared to the saline-treated lambs, whilst ventilation had no effect (**Figure 3F**).

#### Apoptotic Cell Death

The number of Caspase 8<sup>+</sup> cells in the cortical gray matter was significantly reduced in the PROTLPS and INJLPS cohorts (p = 0.006; N = 5/6) (**Figures 4A–E**). No significant differences in the number of TUNEL<sup>+</sup> cells were observed between the groups (**Figure 4F**).

#### Microglia

To determine the effect of LPS and ventilation strategy on microglial activation, Iba-1<sup>+</sup> cells from the cortical gray matter were counted (total 1 mm<sup>2</sup> area) (**Figures 5A–D**). Microglia showing ramified dendritic processes were considered to be in a resting/quiescent state, and those displaying retractive and thickened processes and/or are amoeboid in shape are considered to be activated. No significant differences in the number of activated microglia were observed between groups (LPS: p = 0.87;

vent: p = 0.16; LPS × vent: p = 0.99) (**Figure 5E**). No significant differences in the number of resting microglia were observed between the treatment groups (LPS: p = 0.08; vent: p = 0.13; LPS × vent: p = 0.38) (**Figure 5F**).

#### Astrocyte Number

To determine whether PROT or INJ ventilation, or LPS exposure induces astrogliosis, GFAP<sup>+</sup> astrocytes from the cortical gray matter were counted (total 1 mm<sup>2</sup> area) (**Figures 6A–D**). No significant differences in astrocyte numbers were observed between all groups (LPS: p = 0.98; vent: p = 0.95; LPS × vent: p = 0.46) (**Figure 6E**).

### Blood Brain Barrier Permeability

The total number of blood vessel profiles with protein extravasation in the cortical gray matter were counted (**Figures 7A–D**). LPS treatment was associated with blood vessel protein extravasation (LPS: p = 0.006), whereas ventilation alone or LPS combined with ventilation were not associated with a significant increase in extravasation (vent: p = 0.594; LPS × vent: p = 0.483) (**Figure 7E**).

### DISCUSSION

Respiratory support is a life-sustaining intervention for preterm neonates that cannot breathe sufficiently or autonomously at birth, however, if delivered inappropriately, can cause preterm white matter injury (3, 5, 6, 9). In addition, ventilation after chorioamnionitis exacerbates white matter inflammation and injury (7, 9). However, the effect on the gray matter is not known. In this study, we found that LPS exposure, irrespective of ventilation strategy, caused gray matter edema, reduced neuronal cell density throughout the cortical layers and increased vascular protein extravasation. While injurious ventilation tended to increase activated microglial number, no discernible effect of ventilation strategy was observed. These findings suggest that improving ventilation of preterm infants exposed to chorioamnionitis may not reduce gray matter inflammation and injury.

We have previously demonstrated that injuriously high tidal volumes initiate white matter inflammation and injury via two mechanisms; systemic inflammation and haemodynamic instability (7, 9). In this study, V<sup>T</sup> was significantly higher in injuriously ventilated groups compared to gentle ventilation within the first 15 min. LPS-exposed lambs had a higher PAW compared to the PROTSAL cohort. This indicated better compliance initially, but this effect was limited to 4 min within the initial 15 min of ventilation. Furthermore, it should be noted that PaO<sup>2</sup> was significantly higher in the LPS-treated groups compared to lambs that received saline. Exposure to LPS/intrauterine inflammation causes thinning of the tissue-airspace barrier resulting in improved gas exchange initially after birth (26, 27). This results in improved PaO<sup>2</sup> at the same inspired oxygen.

White matter brain injury has been reported in preterm infants exposed to intrauterine infection and inflammation (7, 14, 28–30). Injurious high V<sup>T</sup> ventilation increases neuroinflammation, protein extravasation, and lipid peroxidation in the white matter of preterm lambs. Further, LPS exposure exacerbates periventricular white matter injury, irrespective of the ventilation strategy used (PROT or INJ). Indeed, LPS exposure upregulates IL-6 and IL-8 cytokine expression within the white matter, and is associated with increased microglial activation, astrogliosis, apoptosis, and protein extravasation (7). Whilst no significant differences in the number of activated microglia were observed, the immunohistochemical labeling shows clear structural remodeling of microglia (amoeboid morphology) adjacent to those with a resting, ramified phenotype in the gray matter, which would account for the high variability in the counts. Our results are in contrast to ventilation and LPS-induced white matter injury, in which the underlying mechanism of damage is thought to be severe microglial-induced inflammation. Nevertheless, our data suggest that injurious ventilation had little effect on acute cortical gray matter inflammation and injury in the late preterm brain. A recent study in preterm infants has shown cortical gray matter injury using MRI, and subsequent lower cognitive scores (31). Prolonged mechanical ventilation was associated

layers (except 4) (E). Gray matter area was significantly higher in the LPS-treated groups (p < 0.0001) when compared to the saline-treated lambs, whilst ventilation had no effect (F).Magnification<sup>=</sup> 2x, scale bar<sup>=</sup> <sup>20</sup>µm. PROTSAL, <sup>N</sup> <sup>=</sup> 5; PROTLPS, <sup>N</sup> <sup>=</sup> 6; INJSAL, <sup>N</sup> <sup>=</sup> 6; and INJLPS, <sup>N</sup> <sup>=</sup> 5.

with a higher rate of brain abnormalities on term equivalent age MRI. Thus, while the initial resuscitation strategy may not produce immediate effects within the cortical gray matter, prolonged ventilation appears to have adverse consequences for the preterm brain, which include but are not limited to cortical gray matter.

It is possible that VIBI within the cortical gray matter takes longer than 2 h to manifest. By contrast we have previously demonstrated that this time period is sufficient for the development of inflammation and injury within the white matter (7, 9). Further studies are required to determine whether gray matter inflammation and injury is present days after the initial resuscitation insult. Furthermore, it should be noted that infants that receive prolonged ventilatory support would have many confounding factors that may contribute to brain injury.

In this study, LPS treatment caused a significant reduction in the number of NeuN<sup>+</sup> neurons in all cortical layers, except layer 4. Neuronal loss and reduced gray matter volumes have been observed in preterm infants that received prolonged respiratory support and/or were exposed to intrauterine inflammation (32– 39). Whilst the mechanisms of neuronal death and changes in

gray matter volumes have not been elucidated, it is thought that they may have some overlap with white matter injury. Furthermore, we observed reduced apoptosis in the LPS-exposed cohorts. However, the decrease in apoptosis simply reflects the increased cortical gray matter area in the LPS-treated groups, indicative of edema. The larger area would reduce cell density if the same number of cells were present.

Neither ventilation strategy or LPS exposure led to significant changes in the number of astrocytes in the cortical gray matter. There are limited data regarding astrogliosis within the cortical gray matter following chorioamnionitis. However, similar to microglia, astrocytes can modulate inflammatory responses within the brain, particularly through their ability to participate in pro-inflammatory cytokine production (40). The lack of astrocytosis following ventilation or LPS exposure suggests other mechanisms mediate cortical neuronal injury. Moreover, LPSexposure was associated with an increase in blood vessel protein extravasation, irrespective of the ventilation strategy. Albumin is the most abundant protein found in blood serum and is not found outside of cerebral blood vessels. However, cerebral blood vessels may become leaky resulting in serum/protein extravasation. This is a known marker of compromised blood brain barrier integrity (7, 9).

This study has some limitations in translation, particularly due to the model being used. While preterm lambs at 126 days gestation have relatively immature lungs and require respiratory support in order to survive, their brains are relatively mature,

and more accurately reflect a near-term infant (41). Indeed, gray matter injury, as detected by MRI in term infants, is likely to be more prominent after severe insults, such as moderate-to-severe asphyxia/hypoxic injury (42). Thus, it may be that the insult from ventilation does not cause overt cortical injury in the latepreterm brain acutely after birth. We did not assess maturation of the neurons in this study. Previous studies in humans and sheep have shown maturational arrest of oligodendrocytes and neurons is a hallmark of preterm brain injury (43). However, it is unlikely that a brief ventilatory period would alter neural maturation. Finally, the ventilator insult, although pronounced, was only brief (15 min). Prolonged respiratory support has been associated with white and gray matter preterm brain injury (31).

Thus, it is possible that longer ventilation would be associated with greater cortical injury.

In summary, injurious ventilation had little effect on gray matter inflammation or injury. LPS-exposure increased neuronal loss, edema and protein extravasation within the cortical gray matter. Neuronal loss was not accompanied by increased gliosis, which is typically observed in white matter brain injury following mechanical ventilation or chorioamnionitis. These data suggest that the underlying mechanisms and/or timing of cortical gray matter injury following intrauterine inflammation and ventilation may differ from white matter injury.

### AUTHOR CONTRIBUTIONS

VS wrote manuscript, analyzed data; AA assisted with data collection analysis and contributed to manuscript writing; IN, BS, SKB, VZ, and KR assisted with animal work; SKB, VZ, and SBH revised the manuscript and contributed to data interpretation. GRP and RG were responsible for study conception, design, manuscript revision and oversight of the research.

#### FUNDING

This research was supported by National Institute of Health R01HD072848-01A1, a joint National Heart Foundation of

### REFERENCES


Australia and National Health and Medical Research Council (NHMRC) Career Development Fellowship (GRP: 1105526), an NHMRC-Australian Research Council Dementia Research Development Fellowship (SKB; 1110040) and NHMRC CJ Martin Early Career Fellowship (1090890; RG) NHMRC Principcal Research Fellowship (SBH: APP1058537) and the Victorian Government's Operational Infrastructure Support Program.

#### ACKNOWLEDGMENTS

The authors acknowledge the facilities and scientific and technical assistance of Monash Histology Platform, Department of Anatomy and Developmental Biology, Monash University.


**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 Stojanovska, Atik, Nitsos, Skiöld, Barton, Zahra, Rodgers, Hooper, Polglase and Galinsky. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary iron repletion following early-life Dietary iron Deficiency Does not correct regional Volumetric or Diffusion Tensor changes in the Developing Pig Brain

*Austin T. Mudd1,2, Joanne E. Fil1,2, Laura C. Knight1,3 and Ryan N. Dilger1,2,3,4,5\**

*1Piglet Nutrition & Cognition Laboratory, University of Illinois, Urbana, IL, United States, 2Neuroscience Program, University of Illinois, Urbana, IL, United States, 3Division of Nutrition Sciences, University of Illinois, Urbana, IL, United States, 4Beckman Institute for Advances Science and Technology, University of Illinois, Urbana, IL, United States, 5Department of Animal Sciences, University of Illinois, Urbana, IL, United States*

#### *Edited by:*

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Susan Cohen, Medical College of Wisconsin, United States Juliane Schneider, University of Lausanne, Switzerland*

> *\*Correspondence: Ryan N. Dilger rdilger2@illinois.edu*

#### *Specialty section:*

*This article was submitted to Neuropediatrics, a section of the journal Frontiers in Neurology*

*Received: 16 October 2017 Accepted: 20 December 2017 Published: 11 January 2018*

#### *Citation:*

*Mudd AT, Fil JE, Knight LC and Dilger RN (2018) Dietary Iron Repletion following Early-Life Dietary Iron Deficiency Does Not Correct Regional Volumetric or Diffusion Tensor Changes in the Developing Pig Brain. Front. Neurol. 8:735. doi: 10.3389/fneur.2017.00735*

Background: Iron deficiency is the most common micronutrient deficiency worldwide and children are at an increased risk due to the rapid growth occurring during early life. The developing brain is highly dynamic, requires iron for proper function, and is thus vulnerable to inadequate iron supplies. Iron deficiency early in life results in altered myelination, neurotransmitter synthesis, neuron morphology, and later-life cognitive function. However, it remains unclear if dietary iron repletion after a period of iron deficiency can recover structural deficits in the brain.

Method: Twenty-eight male pigs were provided either a control diet (CONT; *n* = 14; 23.5 mg Fe/L milk replacer) or an iron-deficient diet (ID; *n* = 14; 1.56 mg Fe/L milk replacer) for phase 1 of the study, from postnatal day (PND) 2 until 32. Twenty pigs (*n* = 10/diet from phase 1) were used in phase 2 of the study from PND 33 to 61, all pigs were provided a common iron sufficient diet, regardless of their early-life dietary iron status. All pigs remaining in the study were subjected to magnetic resonance imaging (MRI) at PND 32 and again at PND 61 using structural imaging sequences and diffusion tensor imaging (DTI) to assess volumetric and microstructural brain development, respectively. Data were analyzed using a two-way ANOVA to assess the main and interactive effects of early-life iron status and time.

results: An interactive effect was observed for absolute whole brain volumes, in which whole brain volumes of ID pigs were smaller at PND 32 but were not different than CONT pigs at PND 61. Analysis of brain region volumes relative to total brain volume indicated interactive effects (i.e., diet × day) in the cerebellum, olfactory bulb, and putamen-globus pallidus. Main effects of early-life iron status, regardless of imaging time point, were noted for decreased relative volumes of the left hippocampus, right hippocampus, thalamus, and increased relative white matter volume in ID pigs compared with CONT pigs. DTI indicated interactive effects for fractional anisotropy (FA) in the whole brain, left cortex, and right cortex. Main effects of early-life iron status, regardless of imaging time point, were observed for decreased FA values in the caudate, cerebellum, and internal capsule in ID pigs compared with CONT pigs. All comparisons described above were significant at *P* < 0.05.

conclusion: Results from this study indicate that dietary iron repletion is able to compensate for reduced absolute brain volumes early in life; however, microstructural changes and altered relative brain volumes persist despite iron repletion.

Keywords: neurodevelopment, iron deficiency, pig, iron repletion, myelination, pediatric nutrition

### INTRODUCTION

Iron is an essential micronutrient that is important for many physiological processes. Despite its importance in the body, iron deficiency remains the most pervasive micronutrient deficiency worldwide (1, 2). Moreover, iron deficiency appears to afflict all age groups, although it is more prevalent in children between birth and 5 years of age (3). Iron deficiency in early life is of great concern as this is a period during which brain development is highly dynamic and thus susceptible to alterations due to nutrition. Iron is specifically needed throughout neurodevelopment for neurotransmitter synthesis (4), myelination (5), and vascular development (6). Accordingly, reductions in dietary iron during the postnatal period greatly influence developmental trajectories. Research suggests that children who were iron deficient (ID) as infants exhibit delayed cognitive development later in life (7, 8), presumably due to structural alterations that developed in infancy and persisted despite iron repletion. These findings are supported by research in ID pigs, which suggests decreased spatial learning at 4 weeks of age (9) and 12 weeks of age (10). As such, there is a need to identify the structural changes that occur during periods of early-life iron deficiency and understand if any of these structural anomalies can be corrected by dietary iron repletion.

Iron is known to be important for neuronal growth and development. Research in 4-week-old pigs suggests that postnatal iron deficiency results in altered hippocampal gene expression for axon growth and guidance, as well as genes responsible for blood brain barrier integrity, angiogenesis, and hypoxia (11). Moreover, reductions in hippocampal iron content (9), alterations in hippocampal gray and white matter (12), and changes in hippocampal neurometabolites (12) all have been previously reported in 4-week-old ID pigs. In 12-week-old pigs that were provided ID diets from birth until 4 weeks of age followed by iron-replete diets until 12 weeks of age, altered hippocampal brain-derived neurotrophic factor (BDNF) profiles were observed, which suggests altered hippocampal plasticity despite iron repletion (13). In perinatally ID rodents, reduced hippocampal volumes and hippocampal subfield volumes have been reported (14). Analysis of rodent hippocampal cell cultures indicated decreased dendritic arborization and expansion as a result of induced iron deficiency (15). Moreover, this alteration in hippocampal dendritic morphology does not appear to be corrected in ID rats after receiving an iron-replete diet (16). Iron deficiency is also known to influence monoamine metabolism in subcortical brain regions, which may result in cognitive deficits later in life (17). While cellular and molecular mechanisms underlying the effects of iron deficiency on brain development have been extensively characterized in animal models, most of these approaches require invasive techniques.

The presence of iron in the brain is also required for proper myelination (5). Oligodendrocytes are rich in iron and are responsible for production of myelin in the brain (18). Iron is specifically needed for myelination because it serves as a cofactor for fatty acid synthesis (19), cholesterol utilization, and metabolic processes (18). Immunohistochemical analyses in rats that were exposed to perinatal iron deficiency indicated decreased myelin concentrations in the cerebellum and spinal cord (20) and a separate study reported reduced markers of oligodendrocyte activity (21). Additionally, postnatal iron deficiency in rats resulted in decreased concentrations of myelin basic protein and 2′,3′-cyclic nucleotide 3′-phosphohydrolase (CNPase), both of which increase as myelination occurs (22). The mechanisms that underlie changes in myelination due to iron deficiency appear to be well characterized; however, most of these markers also require invasive techniques for characterization. Moreover, most studies do not delineate which brain regions might be more susceptible to alterations in myelination due to iron deficiency.

To date, most studies focus on the influence of iron deficiency on hippocampal development; however, high concentrations of iron are found throughout the brain. Therefore, our study sought to characterize the global influence of iron deficiency on brain development and to assess if particular regions are more vulnerable at specific time points. Additionally, this study aimed to non-invasively characterize changes in brain development that occurred as a result of dietary iron deficiency up to 32 days of age, and if these changes could be corrected by 61 days of age, after a period of dietary iron repletion. In doing so, we used the pig because it is a superior model for assessing the influence of nutrition on neurodevelopment (23) and is a well-established model for early-life iron deficiency (9–13). This study is novel in that we employed a longitudinal design and non-invasive neuroimaging techniques to characterize the effects of iron deficiency at postnatal days (PNDs) 32 and 61. Findings from this study are poised to be clinically relevant as characterizations of neuroanatomical changes were assessed using non-invasive techniques, which are immediately translatable to infants in the first year of life. In this study, we hypothesized that dietary iron deficiency would alter brain development and a period of iron repletion may be able to recover particular aspects of brain development that were previously altered.

#### MATERIALS AND METHODS

#### Animal Care and Use

All animal and experimental procedures were in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals and approved by the University of Illinois at Urbana-Champaign Institutional Animal Care and Use Committee. Twenty-eight, naturally farrowed, intact male pigs were obtained from Carthage Veterinary Services and transferred to the University of Illinois Piglet Nutrition and Cognition Laboratory (PNCL) at PND 2. Per standard agricultural protocol, pigs were provided an intramuscular injection of a prophylactic antibiotic (0.1 mL of ceftiofur crystalline free acid such as Excede; Zoetis, Parsippany, NJ, USA) within 24 h of birth. Contrary to typical agricultural procedures, pigs on this study were not provided an iron dextran shot, as iron is the nutrient of interest in the experimental diet. Recent pig studies observed hippocampal transcriptome changes (24) and possible effects of iron overload (10) after iron dextran shot in the first few days of life, which further justify our decision to not provide iron dextran to any pigs. Upon arrival to PNCL on PND 2, pigs were stratified into one of the two experimental diets described below. Pigs were provided experimental milk replacer diets from PND 2 until PND 32 or 33 (phase 1), at which point both treatment groups were weaned onto the same series of weanling pig diet from PND 32 or 33 until PND 61 or 62 (phase 2).

For phase 1 of this study, 28 pigs were housed individually in custom pig rearing units (87.6 cm × 88.9 cm × 50.8 cm; *L* × *W* × *H*), which were composed of three acrylic walls, one stainless steel wall, and vinyl-coated metal flooring. Each caging unit was designed for pigs to see, hear, and smell, but not touch neighboring pigs. Pigs were allowed to physically interact with one another for approximately 15 min each day, and each pig was provided a toy for enrichment in their home-cage. Pig rearing environment was maintained on a 12 h light and dark cycle from 0800 to 2000 hours, with ambient temperature set at 26.6°C for the first 21 days of the study and gradually lowered to 22°C during the last 7 days of phase 1.

For phase 2 of this study, 20 pigs from phase 1 were transferred to the University of Illinois Veterinary Medicine Research Farm (VMRF) immediately following magnetic resonance imaging (MRI) at PND 32 or 33 and housed there until the end of the study. At VMRF, pigs were housed in floor pens (1.5 sq meters) and the rearing environment was maintained on a 12 h light and dark cycle from 0800 to 2000 hours, with ambient temperature set at approximately 22°C.

#### Dietary Treatments

For phase 1 of this study, pigs (*N* = 28, *n* = 14/phase 1 diet) were provided one of the two dietary treatments with varying iron contents. The control diet (CONT) was formulated to meet all of the nutrient requirements of the growing pig and was formulated to contain 117.5 mg Fe/kg milk replacer powder. The ID diet was based off of the CONT diet; however, iron was only formulated to be supplemented at 7.8 mg Fe/kg milk replacer powder. Additionally, both diets were formulated to contain arachidonic acid (ARA) (2.08 g ARA/kg milk replacer powder) and docosahexaenoic acid (DHA) (1.04 g DHA/kg milk replacer powder). Milk replacer was reconstituted fresh daily with 200 g of milk replacer powder per 800 g water. Thus, formulated iron concentrations in reconstituted pig milk replacers were CONT (23.5 mg Fe/L milk replacer) and ID (1.56 mg Fe/L milk replacer). All pigs were provided *ad libitum* access to liquid diets from PND 2 until PND 32 or 33.

For phase 2 of this study, all pigs (*N* = 20, *n* = 10/phase 1 diet) were weaned onto the same series of iron-adequate diets, regardless of their phase 1 dietary iron status. Pigs were provided *ad libitum* access to standard complex diets (major ingredients including corn, whey, and soybean meal) and standard agricultural feeding practices were followed by sequentially switching from stage 1 diets, to stage 2 diets, to stage 3 diets, on PND 32, 41, and 50, respectively. During this phase of the study, all diets were formulated to meet all nutrient requirements of growing pigs (25), including iron. No zinc oxide, copper sulfate, or in-feed antibiotics were included in any diets.

#### Magnetic Resonance Imaging

All pigs remaining in each phase underwent MRI procedures on PND 32 or 33 for phase 1 and again at PND 61 or 62 for phase 2, at the Beckman Institute Biomedical Imaging Center. For phase 1, 28 pigs (*n* = 14 per diet) were subjected to neuroimaging procedures and for phase 2, 20 pigs (*n* = 10 per phase 1 diet) were subjected to neuroimaging procedures. Imaging procedures were performed using a Siemens MAGNETOM Trio 3 T MRI, with a custom pig-specific 8-channel head coil at PND 32 and a human 8-channel head coil at PND 61. The pig neuroimaging protocol included three magnetization prepared rapid gradientecho (MPRAGE) sequences and diffusion tensor imaging (DTI) to assess brain macrostructure and microstructure. Upon arrival to the imaging facility, anesthesia was induced using an intramuscular injection of telazol (50.0 mg of tiletamine plus 50.0 mg of zolazepam reconstituted with 5.0 mL deionized water; Zoetis, Florham Park, NJ) administered at 0.07 mL/kg BW, and maintained with inhalation of isoflurane (98% O2, 2% isoflurane). Pigs were immobilized during all MRI procedures. Visual observation of each pig's well-being, as well as observations of heart rate, PO2 and percent of isoflurane were recorded every 5 min during the procedure. Total scan time for each pig was approximately 60 min. Upon completion of the scan, pig respiration and heart rate were monitored every 15 min until complete recovery from anesthesia. Imaging techniques are briefly described below.

#### Structural MRI Acquisition and Analysis

A T1-weighted magnetization-prepared rapid gradient echo (MPRAGE) sequence was used to obtain anatomic images of the pig brain, with a 0.7 mm isotropic voxel size. The following sequence-specific parameters were used to acquire T1-weighted MPRAGE data: repetition time (TR) = 1,900 ms; echo time (TE) = 2.49 ms; inversion time (TI) = 900 ms; 224 slices; field of view (FOV) = 180 mm; flip angle = 9°. Pig brains were manually extracted as previously described (26). All toolboxes described herein were available in SPM12, and Matlab R2015a was used for data processing. Once extracted, the "Coregister: Estimate & Reslice" toolbox was used to coregister individual brains to the Pig MRI Atlas (27). Next, the "Old Normalize: Estimate & Reslice" toolbox was used to transform individual pig brains into atlas space. The following parameters in Old Normalize were used for pig-specific data processing: template image (Pig MRI Atlas), bounding box (−30.1 −35 −28/30.1 44.8 31.5), voxel size (0.7). Finally, the "Deformations" toolbox was used to generate region of interest masks for volumetric assessment of individual brain regions in the pig MPRAGE space. In the deformation step, the estimated deformation from the Normalize step was used with a pushforward function, which was applied to all 19 regions of interest (defined by the Pig MRI Atlas), and the FOV image was defined as the individual pig's skull-stripped coregistered brain image. A seventh degree interpolation was used and a binary mask of each region was generated. The fslstats-V function (FSL 5.0) was then used to estimate the volume of each individual brain region. In order to account for differences in absolute whole brain volume, all regions of interest were also expressed as a percent of total brain volume (%TBV), using the following equation: (region of interest absolute volume)/(total brain absolute volume) × (100), within subject.

#### Diffusion Tensor Imaging

Diffusion tensor imaging was used to assess white matter maturation and axonal tract integrity using a *b*-value = 1,000 s/mm2 across 30 directions and a 2 mm isotropic voxel. Diffusionweighted echo-planar imaging (EPI) images were assessed in FSL for fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) using the previously described methods (26). Assessment was performed over the following regions of interest: caudate, corpus callosum, cerebellum, both hippocampi, internal capsule, left and right cortex, thalamus, DTI-generated white matter, and atlas-generated white matter using a customized pig analysis pipeline and the FSL software package. For the purposes of this analysis, the Pig Brain atlas, generated from the same species, and previously reported by Conrad and colleagues were used (27). The diffusion toolbox in FSL was used to generate values of AD, RD, MD, and FA.

Masks for each ROI from the atlas were non-linearly transformed into the MPRAGE space of each individual pig and a linear transform was then applied to transfer each ROI into DTI space. A threshold of 0.5 was applied to each ROI, and the data were dilated twice. For each individual ROI, an FA threshold of 0.15 was applied to ensure inclusion of only white matter in the region of interest despite the mask expansion.

#### Statistical Analysis

All researchers involved in this study (i.e., those performing daily procedures, data collection, and data analysis steps) remained blinded to dietary treatment identity until final data analyses had been completed. Data were analyzed by using the MIXED procedure of SAS 9.4 (SAS Institute, Cary, NC, USA). All longitudinal measures reported herein were obtained from pigs on both PND 32 and 61, thus all data were analyzed using a two-way repeated measures ANOVA (i.e., dietary iron status with postnatal day at time of neuroimaging acquisition as the repeated measure). Interactive effects were defined as an interaction between diet (CONT vs ID) and MRI day (PND 32 vs 61). Number of animals per treatment group was based on a power analysis using variability estimates from previous studies to detect differences with sufficient power of 80% at a significance of 0.05. Data were analyzed for outliers (defined as having a studentized residual with an absolute value greater than 3) and outliers were removed prior to statistical analysis. Significance was accepted at *P* ≤ 0.05. Data are presented as least-squares means with pooled standard errors of the mean (SEM).

## RESULTS

### Pig Health Measures

Pig health measures were assessed at the end of each phase, and all outcomes are presented as means ± SD. At the end of phase 1, body weights of ID and CONT pigs were 6.24 ± 1.26 and 10.78 ± 1.45 kg, respectively, and at the end of phase 2 body weights of ID and CONT pigs were 17.05 ± 2.62 and 24.95 ± 2.40 kg, respectively. Average daily liquid milk intake for ID and CONT pigs during phase 1 was 0.93 ± 0.18 and 1.74 ± 0.24 L, respectively, and for phase 2, the average daily feed intake (as-is basis) for ID and CONT pigs was 0.73 ± 0.10 and 0.92 ± 0.13 kg, respectively. Hematocrit concentrations for ID and CONT pigs at the end of phase 1 were 14 ± 3 and 32 ± 3% packed cell volume (PCV), respectively, while at the end of phase 2, hematocrit concentrations in ID and CONT pigs were 35 ± 3 and 36 ± 2%, respectively. Hemoglobin concentrations for ID and CONT pigs at the end of phase 1 were 4.8 ± 0.9 and 10.86 ± 1.1 g/dL, respectively, while at the end of phase 2, hemoglobin concentrations for ID and CONT pigs were 11.85 ± 1.15 and 12.13 g/dL, respectively.

#### Absolute Brain Volumes

An interactive effect of diet and MRI day was observed for whole brain volumes (*P*= 0.02), **Figure 1**. On PND 32, ID pigs exhibited

smaller (*P* < 0.01) brain volumes compared with CONT pigs, and by PND 61 brain volumes were larger (*P* < 0.001) than volumes observed at PND 32, but were not different (*P* = 0.12) between ID and CONT pigs. Interactive effects were also observed for absolute volumes of the caudate (*P* = 0.01), gray matter (*P* < 0.01), hypothalamus (*P* = 0.01), internal capsule (*P* = 0.03), lateral ventricle (*P* = 0.05), olfactory bulb (*P* = 0.01), putamen-globus pallidus (*P* < 0.01), and right cortex (*P* = 0.02), Table S1 in Supplementary Material. Notably, these interactions are likely confounded by the difference in pig whole brain volumes, thus relative brain volumes are analyzed below. Also of note, a main effect of MRI day (*P* < 0.001) was observed for absolute volume of every brain region, indicating that all brain regions increased in size from PND 32 to 61, regardless of dietary treatment.

#### Relative Brain Volumes

Due to observed differences in absolute whole brain volumes, all regions of interest were assessed relative to whole brain volume. Relative brain volume measures were generated by dividing the absolute volume for an individual region by the whole brain volume within subject and multiplying by 100, resulting in a percent of total brain volume measure. Interactive effects of diet and MRI day were observed for relative brain volumes in the cerebellum (*P* < 0.001), olfactory bulb (*P* = 0.03), and putamenglobus pallidus (*P* = 0.01), **Figure 2**. Relative brain volumes of the cerebellum indicated an increase (*P* < 0.01) in size from PND 32 to 61 in the CONT pigs and a decrease (*P* = 0.01) in relative size from PND 32 to 61 in the ID pigs. Analysis of the olfactory bulb indicated no change (*P* = 0.28) in relative size from PND 32 to 61 in the CONT pigs, whereas ID pigs exhibited an increase (*P* < 0.001) in relative size of the olfactory bulb from PND 32 to 61. The putamen-globus pallidus relative size decreased (*P* < 0.01) from PND 32 to 61 in the CONT pigs, but did not change (*P* = 0.60) during the same timeframe in ID pigs. Main effects of diet were also observed for relative volumes in the left hippocampus (*P* < 0.01), right hippocampus (*P* < 0.01), thalamus (*P* = 0.04), and white matter (*P* = 0.03), **Figure 3**. Notably, ID pigs exhibited decreased relative volumes in the left hippocampus, right hippocampus, and thalamus compared with CONT pigs. However, ID pigs exhibited increased white matter compared with CONT pigs. Main effects of MRI day were only observed for relative volumes of the caudate (*P* = 0.05), gray matter (*P* < 0.001), olfactory bulb (*P* < 0.001), and right cortex (*P* = 0.01), Table S2 in Supplementary Material.

#### Diffusion Tensor Measures

Diffusion measures of FA indicated interactive effects of diet and MRI day in the left cortex (*P* = 0.03), right cortex (*P* = 0.05), and DTI-generated whole brain (*P* < 0.001), **Figure 4**. In the left cortex, ID pigs exhibited decreased (*P* = 0.03) FA values compared with CONT pigs at PND 32. By PND 61, left cortex FA values in both dietary treatments were increased (*P* < 0.001)

PND, postnatal day. a,b,cMeans without a common letter differ, *P* < 0.05.

compared with FA values at PND 32, but FA values for ID pigs remained lower (*P* < 0.001) than CONT pigs at PND 61. FA values in the right cortex and whole brain indicated no difference (*P* = 0.36) between ID and CONT pigs at PND 32, but by PND 61 ID pigs exhibited decreased (*P* < 0.01) FA values compared with CONT pigs. Main effects of diet were observed for FA values in the caudate (*P* = 0.04), cerebellum (*P* < 0.01), and internal capsule (*P* = 0.04), in all instances ID pigs exhibited decreased FA values compared with CONT pigs, **Figure 5**. Main effects for MRI day were observed for cerebellum (*P* < 0.001), corpus callosum (*P* < 0.001), and internal capsule (*P* < 0.001), **Table 1**. To verify our whole-brain FA findings, we applied a white matter mask, generated from the pig brain atlas, to the FA data for each pig. Our findings were consistent with the DTI-generated whole brain measures indicating an interactive effect (*P* < 0.001), Figure S1 in Supplementary Material.

No interactive effects of diet and MRI day were observed for measures of RD. Main effects of diet were observed for RD in the left cortex (*P* = 0.03), Table S3 in Supplementary Material, right cortex (*P* < 0.001), and whole brain (*P* < 0.01), **Figure 6**. In all instances ID pigs exhibited increased RD values compared with CONT pigs. Main effects of MRI day were observed for RD measures in all analyzed regions (*P* < 0.05), except for the left hippocampus (*P* = 0.23), Table S3 in Supplementary Material. No interactive effects of diet and MRI day were observed for measures of MD. Main effects of diet were observed for MD in the right cortex (*P* < 0.01) and whole brain (*P* < 0.01), **Figure 6**. In both instances, ID pigs exhibited increased MD values compared with CONT pigs. Main effects of MRI day were observed for MD measures in all analyzed regions (*P* < 0.05), except for the cerebellum (*P* = 0.11) and left hippocampus (*P* = 0.11), Table S4 in Supplementary Material. No interactive effects of diet and MRI day were observed for measures of AD. Main effects of diet were observed for AD in the right cortex (*P* = 0.01) and whole brain (*P* < 0.01), **Figure 6**. In both instances, ID pigs exhibited increased AD values compared with CONT pigs. Main effects of MRI day were observed for all analyzed regions (*P* < 0.05), except for the cerebellum (*P* = 0.66), corpus callosum (*P* = 0.25), and left hippocampus (*P* = 0.05), Table S5 in Supplementary Material.

#### DISCUSSION

Iron deficiency influences structural brain development, which results in lasting functional deficits in animals and humans alike (19, 28, 29). However, it remains unclear which, if any, structural changes that result from early-life iron deficiency can be recovered by dietary iron repletion later in life. We used pig as a biomedical model to non-invasively assess the effects of early-life iron deficiency followed by dietary iron repletion on structural brain development. This study is novel because it utilizes noninvasive neuroimaging techniques to longitudinally characterize the effects of dietary iron status in the pig model, which provides

clinically relevant information to be used in infant populations. To our knowledge, this is the first pig neuroimaging study to report longitudinal DTI measures, thereby providing insight into the timing of maturational events in the pig brain. Comparison of neuroimaging in human infant brains and pig brains suggests that one week of pig brain growth is roughly equal to one month of human brain growth (23). Thus, pigs that were imaged at PND 32 were approximately equal to a four-month-old infant and pigs that were imaged at PND 61 were equivalent to a six-to-12 month-old infant. Results from this study indicate that pig whole brain volumes are influenced by dietary iron status at PND 32, but are able to recover with dietary iron repletion by PND 61. Despite this apparent compensatory volumetric growth, analysis of relative brain region volumes and DTI measures indicate lasting microstructural changes, which may underlie known functional deficits later in life.

#### Brain Volumes

Analysis of whole brain volumes revealed smaller brains in dietary ID pigs at PND 32 when compared with CONT pigs. Interestingly, after a 30-day period of dietary iron repletion, ID and CONT pig brain volumes were not different, thus suggesting compensatory brain volume growth in the ID pig brain. It has previously been reported that iron deficiency influences expression of genes related to axon guidance and expansion in the 4-week-old pig hippocampus (11). Thus, it is possible that neuron growth and expansion were inhibited during the period of dietary iron deficiency, but were no longer inhibited in ID pigs once they were provided an iron-replete diet. In pigs, volumetric wholebrain growth reaches the peak rate of development at 4 weeks of age (30), and it is possible that iron repletion during this dynamic period aided in compensatory volumetric expansion. These findings are in contrast to two ID pig studies in which whole brain volume at 4 weeks of age was not influenced by dietary iron status (9, 12). Notably, both of the previous pig studies provided an iron dextran shot to CONT pigs, whereas no pig in our study received an iron dextran shot. It is common agricultural practice to provide a newborn pig with 250 mg of iron dextran to prevent overt signs of iron deficiency prior to weaning because the sow's milk is largely deficient in iron. It is possible that providing a surfeit amount of supplemental iron via an iron dextran shot may have influenced aspects of brain growth in the previous pig studies, as this has been shown to influence the hippocampal transcriptome (24) and possibly induce an iron overload in young pigs (10). Discrepancies between the studies may also be a result of improved neuroimaging hardware and techniques that provided a more sensitive assessment of structural changes reported in the current study.

Due to the differences in absolute brain volumes between dietary treatment groups, individual brain regions were assessed as a percent of total brain volume. Our analyses indicated volumes of the cerebellum, olfactory bulb, and putamen-globus pallidus were influenced by dietary iron status over time. Interestingly, cerebellar volumes were not different at PND 32, but by PND

observed for FA values in the (A) caudate (*P* = 0.04), (B) cerebellum (*P* < 0.01), and (C) internal capsule (*P* = 0.04), in all instances ID pigs exhibited decreased FA values compared with CONT pigs. Abbreviations: CONT, control; ID, iron deficient; FA, fractional anisotropy. a,bMeans without a common letter differ, *P* < 0.05.


*CONT, control; ID, iron deficient; PND, postnatal day; FA, fractional anisotropy.*

*a,b,c,dMeans in a row without a common superscript letter differ, P* < *0.05.*

*e Data presented as mean and pooled standard error of the means (SEM) for each dietary treatment group. Main effects of dietary treatment (Diet; CONT vs ID) and postnatal MRI day (Day; PND 32 vs 61) and the interaction between Diet and Day are presented.*

*f Whole brain FA values were generated from FA diffusion masks with a threshold of 0.15 to ensure only white matter was measured.*

*g Whole brain FA values were generated from T1-weighted white matter segmentation that was applied to the diffusion data.*

61 dietary ID pigs exhibited smaller relative cerebellar volumes. Iron deficiency has been shown to delay cerebellar myelination in rodents (20), which may be the result of our observed volumetric differences. The putamen-globus pallidus region did not change over time in the ID pigs but decreased in relative size in CONT pigs from PND 32 to PND 61. It is unclear why there was no change in relative size in the ID pigs; however, this brain region is known to contain relatively high iron concentrations (31)

and thus may be susceptible to changes in dietary iron during development.

Interestingly, relative volumes of the olfactory bulb were increased, relative to CONT, in ID pigs at PND 61, corroborating findings from a recent pig study, which reported increased gray matter in the olfactory bulb of ID pigs at 4 weeks of age (12). It remains unclear why the olfactory bulb is influenced by dietary iron deficiency, but it may be possible that this is a form of compensatory growth to aid the pig in acquiring iron through other dietary means. Iron deficiency was first observed in pigs in the 1920s when gestating sows were allowed to forage in fields up until a few days before farrowing (32). Once sows were moved inside and raised on concrete flooring for farrowing, foraging in the ground was no longer possible and their vegetable-based diets were largely deficient in available iron. Researchers noticed that pigs from these sows started to exhibit signs of iron deficiency by 3 weeks of age, and were able to correct these symptoms by supplementing iron in the diets of sows and pigs. Although not explicitly stated by McGowan and Crichton, this study highlights the foraging behaviors of pigs which allow them to maintain adequate iron stores through ingestion of exogenous sources, presumably dirt. Thus, in a case of iron deficiency, we speculate that the olfactory bulb may be preferentially developed to aid in seeking sources of iron in the environment. However, this hypothesis is purely speculative and future work should seek to elucidate the physiological relevance of this observed phenomenon. Additionally, it would be interesting to understand if this is a species-specific observation, or one that is relevant to all animals experiencing iron deficiency.

Regardless of age, pigs on the ID diet exhibited decreased relative volumes of the left and right hippocampi, as well as the thalamus. The observed effects in the hippocampi support previous work, which indicated prenatal, postnatal, and perinatal iron deficiency in rodents resulted in reduced hippocampal subfield volumes (33). In a separate study, Ranade and colleagues observed reductions in hippocampal volume in rodents whose dams were perinatally ID, and it is hypothesized that the observed differences are due to reduced glial cell proliferation (14). Rodent hippocampal cell cultures have shown reductions in dendritic arborization and growth due to iron deficiency further suggesting that iron deficiency can influence morphology of the developing hippocampus (15). Additionally, a recent pig study suggests dietary iron deficiency decreases hippocampal expression of genes responsible axon guidance and growth (11). Early-life iron deficiency followed by dietary iron repletion resulted in altered hippocampal BDNF and tyrosine receptor kinase B (TRKB) expression in pigs at 12 weeks of age, thus indicating lasting effects of an early-life ID diet on hippocampal development (13). Relative volumes of the thalamus were also decreased in pigs that were provided an ID diet early in life, and these findings support research that suggests the thalamus is sensitive to dietary iron status (19). In a 4-week-old pig model of iron deficiency, voxel-based morphometric analysis indicated decreased white matter and increased rates of diffusion, suggesting decreased myelination, in the thalamus of ID pigs (12), further supporting the sensitivity of this region to dietary iron deficiency in pigs.

Results from our study also indicate that relative volumes of white matter were increased in dietary ID pigs. At first pass, this finding might seem opposite of what has been reported, as iron deficiency is known to reduce myelination, which certainly contributes to white matter volume and maturation. However, when focusing on the absolute volumes of white matter in the CONT and ID pigs, we observed no differences between the groups at PND 32 or 61, yet whole brain volumes changed over time. Moreover, the absolute volume of gray matter increased to a larger extent in the ID pigs from PND 32 to 61 than it did in CONT pigs, suggesting that the observed whole brain compensatory growth is likely due to disproportionate growth in gray matter-rich regions. Thus, the larger relative white matter volumes in ID pigs may be driven by smaller absolute brain volumes at PND 32. Our results are in contrast to a previous pig study, which indicated region-specific decreases in white matter concentrations and FA measures of ID pigs (12). However, the results in our study assess absolute volumes relative to whole brain volume, whereas the study by Leyshon and colleagues (12) used voxel-wise comparisons to assess regional differences in white matter, regardless of brain size. Together, these findings of iron deficiency influencing total and relative regional brain volumes may be related to later life cognitive deficits observed in pigs, rodents, and humans (19, 28, 29).

#### Diffusion Tensor Imaging

Diffusion tensor measures provide non-invasive measures of microstructural water movement and are related to axonal growth and myelinating events (34). FA is a diffusion measure that accounts for rate of water diffusion and direction of diffusion in tissue, and typically increases throughout development as myelination occurs and fiber coherence increases (35, 36). Our results indicate whole brain, left cortex, and right cortex FA values are influenced by dietary iron status over time. Measures of whole brain FA indicate both treatment groups exhibited increased measures from PND 32 to 61, indicating development occurred over time. No difference was observed between dietary treatments in whole brain FA value at PND 32, whereas decreased FA values were observed in ID pigs by PND 61 when compared with CONT pigs. It is interesting that at PND 61, whole brain volumes were not statistically different; however, diffusion measures suggest persistent microstructural changes despite a period of dietary iron repletion. These data support recent findings which indicate sensitivity of whole brain FA values in ID pigs (12), but it is important to point out that we did not observe differences until PND 61 (i.e., after providing iron-replete diets), whereas Leyshon and colleagues observed differences by 4 weeks of age. Analysis of left and right cortical FA values indicated increased FA values in both dietary treatments from PND 32 to 61; however, at both PND 32 and PND 61, ID pigs exhibited decreased measures compared with age-matched CONT pigs. These findings suggest that development occurred in both dietary treatment groups; however, dietary iron deficiency delayed cortical myelinating or tissue organization processes at each of the neuroimaging time points. A study of two-week-old infants suggests cortical FA values related to maternal dietary iron intake, suggesting sensitivity of cortical FA measures to dietary iron status (37). However, it should be noted that the study by Monk and colleagues (37) observed clusters of voxels where FA values inversely related to maternal reported iron intake. Our study did not assess maternal iron status nor FA values on a voxel-wise basis, instead our findings indicate decreased FA values averaged over left and right cortical regions in ID pigs. Comprehensive characterization of cortical diffusivity changes should be assessed in future studies to determine the normal trajectory of FA changes in pig brain development. Our results are novel because we comprehensively observe indications of decreased myelination and/or tissue organization despite observed compensatory volumetric growth after a period of dietary iron repletion. These data further suggest that particular aspects of neurodevelopment, might be more sensitive to the timing of early-life iron deficiency compared with other neurodevelopmental events such as brain volume growth.

Dietary iron deficiency also appeared to decrease FA measures in the cerebellum, caudate, and internal capsule, regardless of imaging time point. Indications of delayed myelination in the cerebellum of our pigs support observations in rodents, which suggest decreased myelination due to dietary iron deficiency (20). Because the decreased cerebellar FA values may indicate altered myelination, these findings may support our observations of decreased relative cerebellar volumes in ID pigs. A recent study in children suggests that increased iron content in the caudate positively relates to spatial IQ in 7- to-11-year-old children (38). Although our study does not report iron content of the caudate, decreased caudate FA values in ID pigs suggest the caudate is influenced by dietary iron status and these findings of early-life differences in caudate development may help to explain the observed differences noted in spatial IQ in children. Notably, pigs that were provided an ID diet for the first 4 weeks of life, followed by an iron–replete diet from 4-to-12 weeks, exhibited cognitive deficits even after iron repletion (10). Antonides and colleagues (10) observed decreased hippocampal iron content in the pigs at 12 weeks of age, which may have contributed to the cognitive deficits; however, our study suggests differences in the caudate may also underlie these cognitive deficits. The internal capsule is one of the earliest myelinating regions in infant development (39), and has been shown to be sensitive to dietary treatment in pigs (40, 41). A recent study observed decreased white matter, assessed through voxel-based morphometry, in the internal capsule of ID pigs (12). Therefore, the observed decreased FA values in the internal capsule of ID pigs support recent findings in an ID pig model and further suggest this brain region is highly sensitive to early-life nutrition. Notably, ID and ID-anemic infants were reported to exhibit delayed motor development (42), and it is known that both the internal capsule and cerebellum are involved in motor development. Thus, our diffusion data might suggest that delayed development as a result of iron deficiency in the internal capsule and cerebellum could underlie the observed motor development delays observed in infants.

Analysis of whole brain and right cortex MD, AD, and RD measures indicated increased rates of diffusion in the ID pigs compared with the CONT pigs, regardless of imaging time point. Leyshon and colleagues previously reported increased MD, AD, and RD in 4-week-old pigs; however, their observations were located in the hippocampi and thalamus (12), rather than the whole brain and right cortex as we observed. Extensive research in rodent models of early-life iron deficiency suggests lifelong alterations in myelin fatty acid profiles and gene expression for myelin basic protein, despite iron repletion later in life (19). Moreover, it is speculated that iron uptake may be influenced in preoligodendrocytes and oligodendrocytes (19), thereby attenuating myelination throughout development, which would support our findings, thus suggesting reduced myelination in particular brain regions. As described previously, absolute volumes of white matter were not different between treatment groups at PND 32 or PND 61, which might suggest that oligodendrocytes are present. However, as supported by our diffusion data, the mere presence of oligodendrocytes is not necessarily enough for myelination to proceed, as evidenced by lower FA values in previously-ID pigs at PND 61. Future work should seek to quantify iron content in the brain regions that are reported to be influenced by dietary iron deficiency and elucidate whether the physical presence of iron may be contributing to these observed changes in diffusion measures. Further characterization of brain iron content will help to illuminate the mechanisms though which dietary iron deficiency influences brain myelination, as well as characterization of oligodendrocyte maturation throughout the myelination process in the growing pig.

#### Longitudinal MRI Assessment

A notable strength of this study is the implementation of a longitudinal dietary intervention in which brain development was measured at two time points. To our knowledge, no other neuroimaging study has comprehensively assessed changes in volumetric and diffusion measures over time in the pig. In doing so, this study provides insight into the growth and development of specific brain regions, and can therefore be used as normative data to which other studies can be compared when using the pig as a biomedical model. Analysis of absolute brain volumes and regions within the brain indicated all brain regions increased in size from PND 32 to 61. When assessing regional brain growth relative to whole brain volumes, the proportion of gray matter in the brain decreased and the olfactory bulb exhibited increases as a proportion of total brain volume. Assessment of diffusion tensor measures indicated region-specific FA values increased from PND 32 to 61, which supports observations in human neurodevelopment (35, 36). Future studies should seek to characterize changes in DTI measures at multiple time points throughout pig development. In doing so, researchers will be able to better identify particular developmental processes occurring in the brain (i.e., myelination, reductions in radial glial cells, and alterations in neuron morphology) and relate these to specific observations in diffusivity parameters. These findings further support the use of the pig as model for human brain development and highlight sensitive changes in neurodevelopment across a short time span. By characterizing these changes over time, we are able to better identify clinically relevant, critical windows of neurodevelopment when early-life nutrition may have the greatest influence.

#### CONCLUSION

Using multiple imaging techniques, we were able to noninvasively characterize changes in brain development related to early-life dietary iron deficiency followed by dietary iron repletion. Measures of total brain volume at PND 61 suggest dietary iron repletion is able to compensate for delayed whole brain growth at PND 32; however, diffusion FA measures suggest dietary iron repletion was not able to correct myelination and tissue organization in specific brain regions by PND 61. Many previous studies have shown structural changes in brain development

#### REFERENCES


after a period of iron deficiency, and our study expands upon this work by identifying aspects of brain development, which do not appear to benefit from iron repletion later in life. Thus, these data highlight the importance of the critical window during which adequate dietary iron is necessary to ensure proper brain growth trajectories are established. Moreover, this research suggests possible heightened sensitivity of myelination to dietary iron status, and future work should seek to expand upon the mechanisms through which these changes develop and persist despite iron repletion.

#### ETHICS STATEMENT

All animal and experimental procedures were in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals and approved by the University of Illinois at Urbana-Champaign Institutional Animal Care and Use Committee.

#### AUTHOR CONTRIBUTIONS

AM, LK, and RD were involved in study design and implementation. All authors were involved in data acquisition, analysis, and interpretation. All authors read and approved the final version of this manuscript.

#### ACKNOWLEDGMENTS

The authors would like to thank Drs. Brian Berg and Rosaline Waworuntu for their expertise in study design. They also thank Mead Johnson Nutrition for formulating and providing the diets for this study. They thank Dr. Brad Sutton for his help in optimizing our data processing pipeline. Additionally, they thank Kristen Karkiewicz for managing the day-to-day pig rearing. Lastly, they thank Nancy Dodge, Holly Tracey, and Tracey Henigman from the Beckman Imaging Center for their help in MRI data acquisition.

#### FUNDING

Funding for this project was provided through the University of Illinois Division of Nutrition Sciences Margin of Excellence Award and the University of Illinois Campus Research Board. Dietary treatments were provided in kind by Mead Johnson Nutrition.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fneur.2017.00735/ full#supplementary-material.


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

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

# Protection of Oligodendrocytes Through neuronal Overexpression of the small gTPase ras in hyperoxiainduced neonatal Brain injury

*Meray Serdar <sup>1</sup> , Josephine Herz1 , Karina Kempe1 , Elke Winterhager <sup>2</sup> , Holger Jastrow2,3, Rolf Heumann4 , Ursula Felderhoff-Müser <sup>1</sup> \* and Ivo Bendix <sup>1</sup> \**

*1Department of Pediatrics I, Neonatology, University Hospital, University Duisburg-Essen, Essen, Germany, 2 Imaging Center Essen, EM Unit, University Hospital Essen, University Duisburg-Essen, Essen, Germany, 3 Institute of Anatomy, University Hospital Essen, University Duisburg-Essen, Essen, Germany, 4Biochemistry II, Molecular Neurobiochemistry, Faculty of Chemistry and Biochemistry, Ruhr University Bochum, Bochum, Germany*

#### *Edited by:*

*Olivier Baud, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Susan Cohen, Medical College of Wisconsin, United States Wei-Liang Chen, Medical College of Wisconsin, United States*

#### *\*Correspondence:*

*Ursula Felderhoff-Müser ursula.felderhoff@uk-essen.de; Ivo Bendix ivo.bendix@uk-essen.de*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 08 December 2017 Accepted: 06 March 2018 Published: 21 March 2018*

#### *Citation:*

*Serdar M, Herz J, Kempe K, Winterhager E, Jastrow H, Heumann R, Felderhoff-Müser U and Bendix I (2018) Protection of Oligodendrocytes Through Neuronal Overexpression of the Small GTPase Ras in Hyperoxia-Induced Neonatal Brain Injury. Front. Neurol. 9:175. doi: 10.3389/fneur.2018.00175*

Prematurely born infants are highly susceptible to various environmental factors, such as inflammation, drug exposure, and also high environmental oxygen concentrations. Hyperoxia induces perinatal brain injury affecting white and gray matter development. It is well known that mitogen-activated protein kinase signaling is involved in cell survival, proliferation, and differentiation. Therefore, we aim to elucidate cell-specific responses of neuronal overexpression of the small GTPase Ras on hyperoxia-mediated brain injury. Six-day-old (P6) *syn*Ras mice (neuronal Ras overexpression under the synapsin promoter) or wild-type littermates were kept under hyperoxia (80% oxygen) or room air (21% oxygen) for 24 h. Apoptosis was analyzed by Western blot of cleaved Caspase-3 and neuronal and oligodendrocyte degeneration *via* immunohistochemistry. Short-term differentiation capacity of oligodendrocytes was assessed by quantification of myelin basic protein expression at P11. Long-lasting changes of hyperoxia-induced alteration of myelin structures were evaluated *via* transmission electron microscopy in young adult animals (P42). Western blot analysis of active Caspase-3 demonstrates a significant upregulation in wild-type littermates exposed to hyperoxia whereas *syn-*Ras mice did not show any marked alteration of cleaved Caspase-3 protein levels. Immunohistochemistry revealed a protective effect of neuronal Ras overexpression on neuron and oligodendrocyte survival. Hyperoxia-induced hypomyelination in wild-type littermates was restored in *syn*Ras mice. These short-term protective effects through promotion of neuronal survival translated into long-lasting improvement of ultrastructural alterations of myelin sheaths in mice with neuronal overexpression of Ras compared with hyperoxic wild-type mice. Our data suggest that transgenic increase of neuronal Ras activity in the immature brain results in secondary protection of oligodendrocytes from hyperoxia-induced white matter brain injury.

Keywords: preterm birth, brain injury, hyperoxia, neuronal Ras, white matter injury, neuroprotection

### INTRODUCTION

Within the last decade, mortality of very preterm infants and critically ill-term born infants has decreased by 25% due to major advances in obstetrics and neonatal intensive care (1). However, perinatal brain injury is still a leading cause of death and disability in children (2). With approximately 5.5–11.4% of all live births, the number of prematurely born infants has increased in industrialized

**235**

countries (3), which may further rise due to increasing infertility treatments, multiple pregnancies, and higher maternal age (4). Due to considerable progress in perinatal management of highrisk infants, long-term survival has become an almost expected outcome. However, even in the absence of severe intracranial pathology such as intraventricular hemorrhage or periventricular leukomalacia affecting 10% of very preterm born children (5), diffuse white matter (WM) injury and reduction of cortical gray matter volume are observed in most survivors. This is often associated with cognitive impairment and behavioral problems such as attention deficit disorder, autism, and development of psychiatric disease in later life (6–9).

Compared with intrauterine conditions, preterm infants suffer from chronic or fluctuating exposure to supra-physiological oxygen levels involved in prematurity-related diseases such as retinopathy of prematurity and bronchopulmonary dysplasia (10, 11). However, there is mounting evidence from clinical and experimental studies that large amounts of oxygen in the neonatal period also disturb brain maturation involving oligodendrocyte degeneration and impaired differentiation leading to hypomyelination (12–14). Importantly, these short-term alterations of myelination translate into long-lasting structural WM alterations (15–19) associated with hyperactivity and coordination deficits at adolescent age (20), and cognitive impairment persisting into adulthood (19, 21).

Several cellular mechanisms of hyperoxia-induced preterm brain injury including apoptosis, autophagy, differences in modulation of reactive oxygen species, and inflammation have been identified (17, 21–26). At the molecular level, our previous work demonstrated the importance of the mitogen-activated protein kinase (MAPK) signaling pathway (22, 27). It was shown that hyperoxia reduces Ras GTPase activity and downstream signaling pathways (e.g., phosphorylation of the extracellular regulated kinase (ERK)) (27), which modulate important cellular responses such as differentiation and survival (28). Of note, using *syn*Ras-transgenic mice overexpressing constitutively activated Ras (29) in the neuronal compartment, we confirmed the important involvement of this pathway demonstrating protection against oxygen-induced cell death (27). However, long-term consequences for the development of WM injury and direct effects on hyperoxia-induced oligodendrocyte degeneration and hypomyelination are still unknown. Moreover, neuronal Ras activation in adult mice leads to an increase of dendritic spine density and synapses as well as increased neuronal activity (30). Taking into account that oligodendrocyte development is not only controlled intrinsically, e.g., by transcription factors but also by neuronal activity (31–33), modulation of neuronal activity and signaling pathways may determine oligodendrocyte responses and subsequent effects on myelination in oxygen-related injury of the immature brain.

Although our understanding of consequences of neonatal hyperoxia on WM development significantly improved within the last decade, the impact of a potential cross talk between the neuronal compartment and myelinating oligodendrocytes is still unknown. Given the important role of MAPK signaling in oxygen-induced preterm brain injury and the fact that neuronal activity modulates oligodendrocyte development, we aimed to elucidate the impact of a transgenic overexpression of constitutively activated Ras in neurons on hyperoxia-induced oligodendrocyte degeneration, myelination, and long-lasting ultrastructural alterations of the WM.

#### MATERIALS AND METHODS

#### Animals and Housing

Transgenic mice overexpressing the mutated human-Ras gene [Ha-Ras G12V (exchange of glycin through valin)] under the synapsin 1 promoter (*syn*Ras) were used. These mice reveal a constitutive activation of Ras in post-mitotic neurons starting at postnatal day 4 (P4). Transgenic males were mated with wild-type female C57BL/6 mice. Genotyping was performed by PCR as described previously (29). The following primers were used to identify the mutated human transgene G12V (360 bp); Fwd (*syn*Ras for) AAT TCC AGC TGA GCG CCG GT and Rev (*syn*Ras rev) GAC ACA CTC ATG AGA TGC CT. Animals were housed under 12-h light/dark cycle, food and water *ad libitum*. All animal experiments were approved and performed in accordance with the guidelines of the University Hospital Essen, Germany and with local government approval by the State Agency for Nature, Environment and Consumer Protection North Rhine-Westphalia.

#### Experimental Procedures

Due to the well-known delay in the brain growth spurt period, which corresponds to the most susceptible period of brain injury in neonatal rodents (34–36), 6-day-old mice were used. *Syn*Ras pups and their appropriate wild-type littermates were placed in an oxygen chamber containing 80% oxygen (OxyCycler, BioSpherix, USA) for 24 h (HO) together with their lactating dams. Control pups were kept under normoxic conditions (21% oxygen, NO) resulting in four study groups: wild-type hyperoxia (BL6 HO); *syn*Ras hyperoxia (*syn*Ras HO); wild-type normoxia (BL6 NO); and *syn*Ras normoxia (*syn*Ras NO). Pups were sacrificed at P7, P11, and P42 under deep anesthesia. In accordance with our previous studies (17, 21, 27, 37), 24 h is sufficient to modulate myelin basic protein (MBP) expression and acute cellular degeneration determined at P11 and P7, respectively, without an induction of acute injury to the dams (27, 37). Since differences in MBP expression following neonatal hyperoxia are less prominent few days after the peak of brain growth spurt period (P7–P10) initial myelination was assessed at term-equivalent age (P11) (34, 36) according to our previous work (15, 17, 19, 21, 37). Ultrastructural alterations of myelination were analyzed in young adult mice at P42 by transmission electron microscopy as described below. For protein analysis, mice were transcardially perfused with PBS and brain hemispheres were snap frozen in liquid nitrogen. For histological studies, pups were transcardially perfused with PBS followed by 4% paraformaldehyde (PFA, Sigma–Aldrich). Brains were postfixed in 4% PFA overnight at 4°C and embedded in paraffin. Based on our previous experience (17, 21), whole hemispheres (excluding cerebellum) were used for Western blot analysis of MBP and cleaved Caspase-3 expression.

#### Immunoblotting

Western blotting was performed as described previously (17, 21), with adaptions in epitope detection. Membranes were blocked with 5% non-fat dry milk in Tris-buffered saline, 0.1% Tween-20 (TBST, Sigma Aldrich) and incubated overnight (4°C) with the following primary antibodies: polyclonal rabbit anti cleaved Caspase-3 (1:1,000, Cell Signalling, Germany), monoclonal mouse anti-MBP antibody (1:10,000, Abcam, UK), monoclonal mouse anti-Olig2 (1:1000, Millipore, Germany), and polyclonal rabbit anti-glyceraldehyde 3-phosphate dehydrogenase antibody (1:1,000, Santa Cruz, Germany). Horseradish peroxidase-conjugated secondary antimouse (1:5,000) or anti-rabbit (1:2,000, both DAKO, Denmark) antibodies were used. All antibodies were diluted in 5% non-fat dry milk in TBST. Antibody binding was detected by using enhanced chemiluminescence (GE Healthcare Life Sciences, Germany). For visualization and densitometric analysis, ChemiDoc™ XRS+ imaging system and ImageLab software (Bio-Rad, Germany) were used. Representative images of full-length Western blot are depicted in Figure S1 in Supplementary Material.

### Immunohistochemistry and Confocal Microscopy

After deparaffinization, 10 µm coronal sections (-2.255 ± 0.6 mm from bregma) were rehydrated. Antigen retrieval was performed in a preheated 10 mM sodium citrate buffer (pH 6.0) for 30 min. After blocking with 1% bovine serum albumin, 0.3% cold fish skin gelatine in 0.1% Tween-20 TBS (all Sigma–Aldrich) slides was incubated with primary antibodies overnight at 4°C followed by appropriate secondary antibody incubation for 1 h at room temperature. Sections were counterstained with 4,6-diamidino-2-phenylindole (DAPI) (1 µg/ml, Invitrogen, Germany). Myelination was evaluated at P11 using the primary mouse anti-rat MBP antibody (1:100, SMI-99, Sternberger Monoclonals, USA). Degeneration of oligodendrocytes or neurons were detected at P7 *via* co-labeling of Olig2 (1:100, monoclonal mouse anti-Olig2, Millipore) or NeuN (1:200, polyclonal rabbit anti-NeuN, Millipore) with terminal deoxynucleotidyl transferase mediated biotinylated dUTP nick end labeling (TUNEL, *In Situ* cell death detection kit, FITC; Roche, Germany), performed according to the manufacturers' instructions. All primary antibodies were followed by appropriate secondary antibody staining (anti-mouse Alexa Fluor 488, anti-rabbit Alexa Fluor 555, anti-mouse Alexa Fluor 555; all 1:500, Invitrogen). Stained brain sections were analyzed by confocal microscopy (A1plus, Eclipse Ti, with NIS Elements AR software, Nikon, Germany) using 10× or 20× objectives. Three laser lines (laser diode: 405 nm; Ar laser: 514 nm; G-HeNe laser: 543 nm) and three different filters (450/50-405 LP, 515/20-540 LP, 585/65-640 LP) were used for image acquisition. Confocal *z*-stacks of 10 µm thickness (*z*-plane distance 3 µm) were converted into two-dimensional images using maximum intensity projections. Image acquisition and analysis were performed by an observer blinded to treatment and genotype. The quantification of each staining was performed with the NIS Elements AR software. For MBP stained sections, large scale images (stitching) of deep cortical WM were generated. Degenerating oligodendrocytes or neurons were quantified by counting triple positive cells (DAPI, TUNEL, Olig2/NeuN) in eight bilateral regions of interest (ROI) derived from two sections per animal (each ROI: 396,900 µm2 , retrosplenial and primary somatosensory cortex; cingulum, deep cortical WM; parafascicular nucleus and posterior complex of the thalamus; CA1 and CA2 of the hippocampus).

#### Electron Microscopy

Six-week-old transgenic *syn*Ras mice and their appropriate wildtype littermates exposed to 24 h hyperoxia at P6 were perfused with 37°C warm phosphate buffer solution followed by 2.5% glutaraldehyde in 0.1 M cacodylate buffer (CB). Brains were removed and immediately transferred into 12-well plates filled with 2.5% glutaraldehyde in 0.1 M CB. Brains were cut into 1 mm thick frontal slices for better penetration. One section clearly showing the corpus callosum was chosen from each specimen and trimmed in a way that the latter was preserved in a final ~1 mm × 1 mm × 1 mm tissue block. After fixation lasting for at least 12 h, blocks were washed in CB at room temperature for 3× 30 min, followed by 1% osmium tetroxide (PolyScience, USA) in CB for 3.5 h in the dark. After washing in CB, 30 and 50% ethanol (45 min each) was added followed by incubation in 1% uranylacetate (Electron Microscopy Science, USA) in 70% ethanol in the dark for 1 h. Blocks were further incubated in 80, 90, and 96% ethanol (45 min each) and 100% ethanol (3 × 20 min) followed by propylene oxide (2× 20 min, Sigma) and EPON® (Sigma) solutions in propylene oxide with increasing EPON® concentration (3:1, 1:1, 1:3; 60 min each, 100% EPON® overnight) at room temperature. Polymerization was performed at 60°C for 2 days. After trimming solid EPON® blocks were cut on a Reichert-Jung Ultracut® E ultramicrotome (Reichert AG, Austria) set to a thickness of 50 nm. Sections were mounted on 200 mesh hexagonal copper grids and treated with 1% uranylacetate for 4 min and lead citrate trihydrate (0.4% in ddH2O) for 3 min for contrast enhancement. A Zeiss transmission electron microscope (EM 902A; Zeiss, Germany) was used for final investigation at 80 kV at magnifications from 3,000× to 100,000×. Digital image acquisition was performed by a Mega View III slow-scan-CCD camera connected to a PC running ITEM® 5.0 software (Soft Imaging Systems, Germany); images were saved as uncompressed 16 bit TIFF files and further processed using Adobe® Photoshop® CS5. Alterations of the myelin sheath were analyzed by using the Image J software (Java Software).

To evaluate myelin deficits a total of 1,645 axons (BL6 HO) and 1,488 axons (*syn*Ras HO) in 60–65 fields of view at 10,000× magnification (61.44 µm2 each) derived from 4 mice per group were analyzed. The percentage of non-myelinated axons related to the total number of axons per field of view was quantified. Pathological alterations in the myelin sheath of myelinated axons were evaluated through quantification of the number of axons with myelin encapsulation and decompaction, and axons with increased adaxonal space. Only axons whose contour was completely within each photograph were used. Analysis was carried out by an individual blinded to group assignment.

#### Statistical Analysis

Data are presented as mean + SD. For statistical analysis, the GraphPad Prism 6.0 software package (GraphPad Software, USA) was used. After checking for normal distribution using D'Agostino–Pearson omnibus test either ordinal one-way analysis of variance with *post hoc* Bonferroni multiple comparisons test or Kruskal–Wallis with Dunn's multiple comparisons test were applied (**Figures 1**–**3**). For comparison of two groups (**Figure 4**) either Student's *t*-test (parametric) or in case of non-parametric data Mann–Whitney *U* test was used. *p*-Values less than 0.05 were considered as statistically significant.

#### RESULTS

#### Neonatal Hyperoxia-Induced Apoptosis Is Decreased by Neuronal Ras Activation

To evaluate the extent of brain injury following neonatal hyperoxia at P6, we determined cleaved Caspase-3 protein abundance immediately after 24 h hyperoxia. In line with previous findings (27), we detected a significant increase of cleaved Caspase-3 levels in protein lysates of wild-type mice following hyperoxia (BL/6 HO) compared with normoxic controls (BL/6 HO) and compared with *syn*Ras mice exposed to hyperoxia (*syn*Ras HO) (**Figure 1**). No alteration of cleaved Caspase-3 protein levels was determined in *syn*Ras animals at normoxia (*syn*Ras NO, **Figure 1**).

### Constitutive Neuronal Ras Activity Ameliorates Neuronal and Oligodendrocyte Degeneration in Neonatal Mice Following Hyperoxia

To investigate the effect of hyperoxia and neuronal Ras overexpression at a cellular level, indicated by TUNEL labeling, we analyzed the amount of degenerating neurons (**Figure 2A**) and oligodendrocytes (**Figure 2B**). Hyperoxia significantly increased

(NO)] or hyperoxia [24 h, 80% oxygen (HO)] at P6 in *syn*Ras mice or their respective wild-type littermates (BL/6). Data are represented as relative protein expression [cleaved Caspase-3/glyceraldehyde 3-phosphate dehydrogenase (GAPDH)] normalized to the control group (BL/6 NO) (\*\**p* < 0.01).

neuronal degeneration in cortical, hippocampal, and thalamic (TH) regions of wild-type mice, which was significantly reduced in *syn*Ras-transgenic mice (**Figure 2A**). For oligodendrocyte degeneration, we detected a similar increase in BL/6 wild-type mice (**Figure 2B**). Importantly, selective neuronal activation of Ras led to a significant decrease of oligodendrocyte degeneration in TH and WM structures (**Figure 2B**).

#### Hyperoxia-Induced Hypomyelination Is Prevented in *syn*Ras Mice

A major hallmark of oxygen-induced toxicity to the immature brain is characterized by hypomyelination (15, 19, 21). To clarify whether the unexpected decrease in oligodendrocyte degeneration in *syn*Ras mice translates into disturbed myelination, we next examined MBP expression at P11 after exposure to hyperoxia at P6. As shown in representative images in **Figure 3A**, hyperoxia resulted in a decrease of MBP expression in wild-type BL/6 mice which was restored in *syn*Ras mice (**Figure 3A**). Quantification of protein lysates by Western blot analysis isolated from whole hemispheres confirmed these qualitative observations and revealed significantly reduced levels of MBP expression in animals exposed to hyperoxia compared with normoxic control wild-type control mice (**Figure 3B**). Normoxic and hyperoxic *syn*Ras mice revealed similar levels of MBP expression compared

Figure 2 | Hyperoxia-induced neuronal and oligodendrocyte degeneration is ameliorated in *syn*Ras mice. Brain sections from P7 mice either exposed to normoxia [21% oxygen (NO)] or hyperoxia [24 h, 80% oxygen (HO)] at P6 in *syn*Ras mice or their respective wild-type littermates (BL/6) were analyzed. Neuronal (NeuN) (A) and oligodendrocyte (Olig2) (B) degeneration was determined in cortical (CX), hippocampal (HC), thalamic (TH), and white matter (WM) regions by double-labeling with TUNEL (\**p* < 0.05, \*\**p* < 0.01, and \*\*\**p* < 0.001).

hyperoxia-induced hypomyelination. Myelin basic protein (MBP) expression was analyzed 4 days post hyperoxia. (A) Representative immunohistochemical staining of the deep cortical white matter assessed by confocal microscopy (scale bar = 100 µm). (B) MBP and (C) Olig2 protein expression was analyzed by Western blot in protein lysates of hemispheres of P11 mice that were exposed to either normoxia [21% oxygen (NO)] or hyperoxia [24 h, 80% oxygen (HO)] at P6 in *syn*Ras mice or their respective wild-type littermates (BL/6). Data are represented as relative protein expression [MBP/glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and Olig2/GAPDH] normalized to the control group (BL/6 NO) (\*\**p* < 0.01 and \*\*\**p* < 0.001).

with normoxic BL/6 controls (**Figure 3B**). Of note, 4 days after hyperoxia Olig2 expression was not significantly altered across all experimental groups (**Figure 3C**).

### Neuronal Ras Activation Protects From Hyperoxia-Induced Long-Lasting Ultrastructural Myelination Abnormalities

Since our major focus of this study was to determine the impact of enhanced neuronal Ras activity in the context of hyperoxiainduced WM pathology and according to the fact that we did not observe any differences between normoxic control animals (BL/6 and *syn*Ras NO) across all previous analyses, we focused our analysis regarding long-term ultrastructural changes *via* electron microscopy to hyperoxic animals only. Assessing crosssections of the corpus callosum of 6-week-old mice after neonatal hyperoxia, we observed several myelin abnormalities including non-myelinated axons, axons with an increased adaxonal space as well as split sheaths with decompaction, myelin lamellae partially broken down into vesicular structures at the innermost region of the myelin sheath (internal loops) as well as focal myelin damage (**Figure 4A**). Comparing both injury groups, we detected a significant reduction regarding non-myelinated axons (**Figure 4B**), axons with an increased adaxonal space (**Figure 4C**), as well axons with signs of myelin encapsulation and decompaction (**Figure 4D**) in *syn*Ras following neonatal hyperoxia at P6.

### DISCUSSION

Oxygen-induced preterm birth-related brain injury is associated with subtle neurodegeneration and impaired WM development resulting in long-lasting decrease of fractional anisotropy and diffusivity in the WM associated with neurodevelopmental disturbances (17, 19–21, 38, 39). Myelination is the key for efficient nerve conduction velocity proposed to strengthen circuitry throughout the nervous system (40). However, recent evidence also suggests a substantial impact of neuronal activity on myelination (31, 33). Our results demonstrate that the selective neuronal activation of the small GTPase Ras protects against hyperoxia-mediated neuronal but also oligodendrocyte degeneration and hypomyelination. Importantly, these short-term effects translated into longlasting improvement of myelin integrity resulting in a reduced amount of non-myelinated axons, an increase in axon-lamination, and an elevated number of axons with compact myelin.

Small GTPases from the Ras family are highly evolutionary conserved, molecular switches involved in important cellular responses such as proliferation, differentiation, and survival (28). We previously showed that hyperoxia reduces Ras activity and that transgenic mice overexpressing Ras under the synapsin 1 promoter (*syn*Ras) to drive the selective overexpression of V12-Ha-Ras in post-mitotic neurons (29) are protected against hyperoxia-induced cellular degeneration (27). Nevertheless, the cellular target of protection remained unclear. In this study, we show that cell-specific neuronal modulation of Ras protects neurons from cellular degeneration. According to our previous work demonstrating that hyperoxia-induced cell death is associated with reduced expression Ras effector molecules promoting cell survival (e.g., ERK1/2) (27), results of this study can be most likely explained by increased Ras-mediated neuronal survival signaling in *syn*Ras mice. Our results of regional neuronal cell death analysis correspond to previous studies revealing increased cell death

in the cortex and the deep gray matter following hyperoxia (41, 42). Of note, we also detected an increased neuronal cell death in parts of the hippocampus (i.e., CA1 and CA2) which seems to contrast a previous report revealing no impact of hyperoxia on cellular degeneration in the dentate gyrus (42). This difference might be explained by the different substructures analyzed in the former study and by the fact that the dentate gyrus belongs to the archicortex which respond different to cellular stress compared with the neocortex (43). Furthermore, in contrast to the previous study, we performed cell-specific analysis by NeuN/ TUNEL co-staining to specifically analyze neuronal degeneration which is limited with a single TUNEL staining. Even though the importance of Ras in cell survival is widely accepted, for specific pathophysiological cases and cellular systems, it has been shown that Ras may also promote cell death (44). In this study, there are no indications for such detrimental effects probably due to developmental differences, i.e., transgenic *syn*Ras mice exposed to normoxia did not show an increased cellular degeneration compared with wild-type littermate controls. Therefore, the immature brain might be less sensitive to proapoptotic effects of Ras activation and thus potential harmful effects.

Surprisingly, in addition to protection from neuronal death, we observed a marked decrease in hyperoxia-induced oligodendrocyte degeneration in *syn*Ras-transgenic mice. Since in these mice, Ras activity is only enhanced in neurons, an indirect impact of the neuronal compartment on oligodendrocyte responses to hyperoxia can be assumed. These results clearly support the current concept of an intense communication between neurons/axons and oligodendrocytes/myelin (31, 33). Interestingly, elevated Ras signaling exclusively in oligodendrocytes has been shown to mediate opposite effects (45). Increased Ras activity in oligodendrocytes of healthy adult mice resulted in myelin decompaction (45) emphasizing the importance of cell-specific analysis and the divergent roles of Ras-mediated signaling dependent on the pathophysiological context, the cellular system, and developmental stage. Nevertheless, expression was restricted to mature oligodendrocytes. Novel transgenic mouse models directly targeting immature oligodendrocyte precursor cells (OPCs) would help to clearly define the interplay between neuronal responses to hyperoxia and developmental processes of myelination in neonatal subjects.

Protective effects on acute oligodendrocyte degeneration were accompanied by improved myelination 4 days after hyperoxia. Interestingly, no differences were observed for the abundance of Olig2 at this time point, indicating compensatory oligodendrocyte proliferation, which, however fails to restore myelination deficits (37). This might be well explained by impaired oligodendrocyte maturation and differentiation following hyperoxia as previously described (21, 37). Increased MBP expression at similar Olig2 levels in hyperoxic *syn*Ras suggests that neuronal Ras activation positively influences differentiation capacity of OPCs in the context of neonatal oxygen-induced toxicity.

In addition to subacute effects on oligodendrocyte survival and differentiation, neonatal hyperoxia causes long-lasting structural alterations on tensor imaging scans of adult brains, which is accompanied by significant cognitive deficits (21). In this study, we provide more detailed information about axon–myelin integrity with ultrastructural analysis of myelin sheaths in young adult animals exposed to neonatal hyperoxia. Here, we detected a considerable amount of non-myelinated axons in hyperoxic wild-type mice, which was significantly improved in *syn*Ras mice. Furthermore, the percentage of myelinated axons with abnormal myelin structures was significantly lower in *syn*Ras mice. While being in concordance with previous descriptions of hyperoxiarelated disturbed ultrastructural myelin integrity (18), our results add important new knowledge, because pathological alterations of the WM were improved solely by modulation of the neuronal compartment, represented by increased Ras activity.

Different mechanisms may account for our observations. First, according to the initial description by Heumann et al. *syn*Ras mice reveal an increased expression of neuropeptide Y (29), which has been shown to improve myelination in the neonatal brain *via* induction of neurotrophin 3 (46). Second, increased neuronal activity described for *syn*Ras mice displayed by enhanced glutamatergic transmission, and long-term potentiation (30) may explain our findings. Whereas the essential need of intact myelination for preservation and maintenance of axonal structure and function is without any doubt (47), there is compelling evidence for a substantial communication into the other direction, i.e., an activity-dependent signaling from neurons/ axons to oligodendrocytes/myelin (33). Accordingly, recent studies indicate that axonal action potentials activate myelinic NMDA receptors (48) resulting in impaired metabolic coupling between axons and oligodendrocytes (49). In addition to metabolic processes, neuronal activity is supposed to determine the release of BDNF (50), a growth factor particularly important for OPC development and maturation (51, 52). While much focus has been given to neuronal activity affecting OPC responses during physiological development, potential interactions in response to pathology in the developing brain are less explored. Considering the aforementioned studies, we speculate that hyperoxia modulates neuronal/axonal function and activity due to reduced Ras activity, which may be compensated by neuronal Ras overactivation thereby improving oligodendrocyte differentiation and long-term myelination capacity. A clear goal for future work will be to characterize neuronal/axonal function and activity in hyperoxic animals with constitutively expressed activated Ras in the neuronal compartment. This is further supported by a very recent report that the pattern of neuronal activity triggers distinct responses of OPC proliferation and differentiation (53).

Strengths of this study are well statistically powered analyses, long-term evaluation of myelination deficits, and the use of a cellspecific transgenic mouse model. A potential limitation might be that this mouse model did not allow for conditional transgene expression initiated at various time points. Nevertheless, as in *syn*Ras mice, constitutive Ras activation starts at postnatal day 4, physiological effects on embryonic neuronal development can be excluded. Furthermore, long-term analyses of neurodevelopmental behavior and of axonal function/integrity would have strengthened our hypothesis. However, previous experimental studies combined with clinical data provide clear evidence for a good correlation between alterations of WM development and long-term behavioral deficits as well as axonal integrity and function (14, 18–21, 39).

To conclude, this work demonstrated that hyperoxia-induced impairment of neurodevelopment does not solely rely on direct modulation of oligodendrocyte responses but is also affected by neuronal cell signaling with major impact on WM development. This work emphasizes the unmet need for cell-specific analysis in models of neonatal brain injury to identify more specific targets for therapeutic intervention.

### ETHICS STATEMENT

All animal experiments were approved and performed in accordance with the guidelines of the University Hospital Essen, Germany and with local government approval by the State Agency for Nature, Environment and Consumer Protection North Rhine-Westphalia.

### AUTHOR CONTRIBUTIONS

MS, IB, JH, KK, RH, and HJ designed and performed the experiments and analyzed the data. MS, JH, RH, EW, HJ, UF-M, and IB discussed the data. UF-M and IB initiated and organized the study. MS, JH, UF-M, and IB wrote the manuscript.

### ACKNOWLEDGMENTS

The authors thank Dorothea Schünke for her ambitious help and patience preparing sections for electron microscopy and Mandana Rizazad for ongoing excellent technical assistance.

### FUNDING

This work was supported by the Mercator Research Center Ruhr (# Pr-2011-0066, to RH, UF-M, IB), Karl-Heinz Frenzen- and C. D.-Stiftung (to JH, UF-M, IB). RH was funded by HORIZON 2020, No 686841. Furthermore, we acknowledge support by the Open Access Publication Fund of the University of Duisburg-Essen.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fneur.2018.00175/ full#supplementary-material.

## REFERENCES


glutathione synthesis. *Free Radic Biol Med* (1998) 25:936–42. doi:10.1016/ S0891-5849(98)00131-2


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

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

# *In Vivo* Multimodal Magnetic Resonance Imaging Changes After *N*-Methyl-d-Aspartate-Triggered Spasms in Infant Rats

*Minyoung Lee1,2, Mi-Sun Yum1,2\*, Dong-Cheol Woo2 , Woo-Hyun Shim2,3, Tae-Sung Ko1 and Libor Velíšek4,5,6*

*1Department of Pediatrics, University of Ulsan College of Medicine, Ulsan, South Korea, 2Asan Institute for Life Sciences, Asan Medical Center, Seoul, South Korea, 3Department of Radiology, University of Ulsan College of Medicine, Ulsan, South Korea, 4Department of Cell Biology and Anatomy, New York Medical College, Valhalla, NY, United States, 5Department of Pediatrics, New York Medical College, Valhalla, NY, United States, 6Department of Neurology, New York Medical College, Valhalla, NY, United States*

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Yohan Van De Looij, Université de Genève, Switzerland Joon Won Kang, Chungnam National University Medical School, South Korea*

> *\*Correspondence: Mi-Sun Yum yumyum99@daum.net*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 27 December 2017 Accepted: 28 March 2018 Published: 16 April 2018*

#### *Citation:*

*Lee M, Yum MS, Woo DC, Shim WH, Ko TS and Velíšek L (2018) In Vivo Multimodal Magnetic Resonance Imaging Changes After N-Methyl-d-Aspartate-Triggered Spasms in Infant Rats. Front. Neurol. 9:248. doi: 10.3389/fneur.2018.00248*

Objective: Despite the serious neurodevelopmental sequelae of epileptic encephalopathy during infancy, the pathomechanisms involved remain unclear. To find potential biomarkers that can reflect the pathogenesis of epileptic encephalopathy, we explored the neurometabolic and microstructural sequelae after infantile spasms using a rat model of infantile spasms and *in vivo* magnetic resonance imaging techniques.

Methods: Rats prenatally exposed to betamethasone were subjected to three rounds of intraperitoneal *N*-methyl-d-aspartate (NMDA) triggering of spasms or received saline injections (controls) on postnatal days (P) 12, 13, and 15. Chemical exchange saturation transfer imaging of glutamate (GluCEST) were performed at P15 and 22 and diffusion tensor imaging and additional spectroscopy (1H-MRI/MRS) of the cingulate cortex were serially done at P16, 23, and 30 and analyzed. Pathological analysis and western blotting were performed with rats sacrificed on P35.

Results: Within 24 h of the three rounds of NMDA-induced spasms, there was an acute increase in the GluCEST (%) in the cortex, hippocampus, and striatum. When focused on the cingulate cortex, mean diffusivity (MD) was significantly decreased during the acute period after multiple spasms with an increase in γ-aminobutyric acid (GABA), glutamate, and glutamine *N*-acetylaspartate-plus-*N*-acetylaspartylglutamate (tNAA), total choline, and total creatine. The juvenile rats also showed decreased MD on diffusion tensor imaging and significant decreases in taurine, tNAA, and macromolecules-plus-lipids in the cingulate cortex. Pathologically, there was a significant reduction in glial fibrillary acidic protein, myelin basic protein, and neuronal nuclei expression in the cingulate cortex of rats with NMDA-induced spasms.

Significance: These neurometabolic and microstructural alterations after NMDAtriggered spasms might be potential imaging biomarkers of epileptic encephalopathy.

Keywords: infantile spasms, MR spectroscopy, neurometabolites, cingulate cortex, animal models

#### INTRODUCTION

Infantile spasms are an epileptic encephalopathy that requires urgent therapeutic intervention to avoid severe consequences, such as mental retardation and developmental regression (1). Diverse etiologies, including both unidentified causes and established brain pathologies, can lead to infantile spasms with diffuse electrographic abnormalities (2), which probably impair common developmental processes such as synaptogenesis, myelination, and neuronal migration, causing mental retardation or refractory epilepsies in later life (3, 4). However, these neurodevelopmental consequences have not been clearly characterized (2).

To support the rationale for an aggressive treatment approach for infantile spasms, it is crucial to determine the changes produced by the spasms in the immature brain and assess the possibility for a restoration or reversal of these changes. In children with infantile spasms, evaluation of the effects of spasms on the brain is extremely complicated because of ethical problems and a relatively small number of patients with heterogeneous etiologies (5) and treatment protocols (6, 7).

From the bench side, it would be advantageous to have a model without pre-existing lesions and a model with easily operable spasms to identify the pathological effects of the experimental spasms on the brain. The rat model of *N*-methyl-d-aspartate (NMDA)-triggered spasms (4) after prenatal betamethasone priming is regarded as having a cryptogenic etiology and the spasms are reliably triggered by the injection of NMDA. The NMDAtriggered spasms can be provoked at postnatal day 12–15, a period in rodents relevant to human infancy (3, 4). The model also displays several behavioral changes after multiple spasms (3), corresponding to those of patients with infantile spasms, and has recently been independently verified (8).

In the present study, using this rat model of NMDA-induced infantile spasms, we performed MR imaging to evaluate the acute and chronic neurometabolic and microstructural changes after spasms as potential biomarkers of epileptic encephalopathy. To add longitudinal *in vivo* data to this model, we adopted the latest high-field proton MRI technologies including chemical exchange saturation transfer imaging of glutamate (GluCEST), diffusion tensor imaging, and spectroscopy (1 H-MRI/MRS). GluCEST imaging can map the level of glutamate *in vivo* at a high spatial resolution (9–11), and diffusion tensor imaging is used to reveal the *in vivo* connectivity of the nervous system (12, 13). 1 H-MRS allows noninvasive quantification of brain metabolites in specific brain areas and has been successfully used in the field of epilepsy research (14, 15). Specifically, the *in vivo* developmental changes in neurometabolites and connectivity were measured in the cingulate cortex, which has extensive connections with critical regions such as the basal ganglia, thalamus, brainstem, and hippocampus (16, 17). Additional histopathologic analyses and behavioral tests were performed to further elucidate the microstructural and neurochemical changes.

#### MATERIALS AND METHODS

#### Animals

Experiments were approved by the Institutional Animal Care and Use Committee of the Ulsan University College of Medicine and conformed to the Revised Guide for the Care and Use of Laboratory Animals [NIH GUIDE, 25(28), 1996] (18). Timed-pregnant Sprague-Dawley rats were purchased from an approved source (Orient Bio Inc., Seoul, Korea). The rats were housed individually in the animal facility during the remainder of their pregnancy with free access to standard rat chow and water on a regular 12-h light-dark cycle with the lights on at 08:00. On gestational day 15, pregnant rats received two injections of 0.4 mg/kg betamethasone (Sigma-Aldrich, St. Louis, MO, USA) at 08:30 and 18:30. Delivery occurred consistently on gestational day 22, which was considered postnatal day (P) 0 for the offspring.

Spasms were triggered by intraperitoneal injection of NMDA on P12 (6 mg/kg), P13 (10 mg/kg), and P15 (15 mg/kg); control groups received the same volume of saline (**Figure 1**). Immediately

after NMDA administration, the rats were observed for 90 min and only the animals confirmed to have had three bouts of spasms at the appropriate time points on each day (i.e., on P12, P13, and P15) were included in the analyses (4).

#### MR Imaging Studies

Animals were maintained under anesthesia with 1% isoflurane in a 1:2 mixture of O2:N2O with monitoring of their respiratory rate, electrocardiogram, and rectal temperature. GluCEST MR imaging was performed using a 7.0 T/160-mm small-animal imaging system (Bruker Pharmascan, Ettlingen, Germany) with a single-channel surface coil. Images were obtained using a 9.4 T/160-mm bore animal MRI system (Agilent Technologies, Santa Clara, CA, USA) for 1 H-MRS and diffusion tensor imaging. Radiofrequency excitation and signal detection were accomplished with a 72-mm quadrature volume coil and a twochannel phased-array coil, respectively. Axial slices corresponding to coronal images in the neuroanatomic axis were collected from the cervical spinal cord to the olfactory bulb.

GluCEST images were acquired from an axial slice (1-mm thick) that included the hippocampal region. GluCEST images were acquired using T2-weighted imaging (rapid acquisition with relaxation enhancement [RARE]) with a frequency selective saturation preparation pulse comprised a Gaussian pulse with a total duration of 1,000 ms (irradiation offset of 500.0 Hz and interpulse delay of 10 µs) at a B1 peak of 5.6 µT. Z-spectra were obtained from −5.0 ppm to +5.0 ppm with intervals of 0.33 ppm (total, 31 images, Figure S1 in Supplementary Material). The sequence parameters were as follows: repetition time/echo time (TR/TE) = 4,200/36.4 ms, field of view = 30 mm × 30 mm, slice thickness = 1 mm, matrix size = 96 × 96, RARE factor = 16, echo spacing = 6.066 ms, and average = 1. To confirm the linear GluCEST effect, a phantom consisting of test tubes with different concentrations of glutamate (pH 7.0) images at 7 T was also done (Figure S2 in Supplementary Material).

To measure the GluCEST value (%), each region of interest (cortex, hippocampus, striatum) was drawn manually on the T2weighted anatomical MR images without a frequency selective saturation preparation pulse, and the regions of interest were overlaid on the GluCEST maps. GluCEST contrast is measured as the asymmetry between an image obtained with saturation at the resonant frequency of exchangeable amine protons (+3 ppm downfield from water for glutamate) and an image with saturation equidistant upfield from water (–3 ppm), according to the following equation:

$$\text{GluCEST}(\%) = \frac{S\_{-3.0 \text{ ppm}} - S\_{+3.0 \text{ ppm}}}{S\_{-3.0 \text{ ppm}}} \star 100$$

where *S*−3.0ppm and *S*+3.0ppm are the magnetizations obtained with saturation at a specified offset from the water resonance of 4.7 ppm.

The B0/B1 maps on the same slices were acquired for B0 and B1 correction. The B0 map was calculated by linearly fitting the accumulated phase per pixel following phase unwrapping against the echo time differences from gradient echo images collected at TEs of =1.9 and 2.6 ms. B1 maps were calculated by using the double-angle method (flip angles 30° and 60°) and the linear correction for B1 was calculated as the ratio of the actual B1 to the expected value.

1 H-MRI/MRS images were obtained at the assigned times as follows (**Figure 1**): (1) 1 day after the last cluster of spasms (P16), (2) about 1 week after the spasms (P23), and (3) about 2 weeks after the spasms (P30). The MR spectra were acquired through a signal voxel (from bregma to −3.0 mm in a coronal section, 3 mm × 2 mm × 1.5 mm) in the cingulate cortex using a point-resolved spectroscopy (PRESS) sequence for 128 acquisitions with TR/TE = 5,000/13.4 ms. For quantification, unsuppressed water signals were also acquired from the same voxel (average = 8). All the MR spectra were processed with the linear combination analysis method (LC Model ver. 6.0, Los Angeles, CA, USA) to calculate the metabolite concentrations from a fit to the experimental spectrum, based on a simulated basis set. The following brain metabolites were included in the metabolite basis set: alanine (Ala), aspartate (Asp), creatine (Cr), γ-aminobutyric acid (GABA), glucose (Glc), glutamate (Glu), glutamine (Gln), glycerophosphorylcholine (GPC), phosphorylcholine (PCh), myo-inositol (mIns), lactate (Lac), phosphocreatine (PCr), *N*-acetylaspartate (NAA), *N*-acetylaspartylglutamate (NAAG), taurine (Tau), macromolecules (MMs), and lipids (Lip). The water-suppressed time domain data were analyzed between 0.2 and 4.0 ppm, without further T1 or T2 correction. Absolute metabolite concentrations (mmol/kg wet weight) were calculated using the unsuppressed water signal as an internal reference (assuming 80% brain water content) (19). The *in vivo* proton spectra were judged to have an acceptable value if the standard deviation of the fit for the metabolite was less than 20% (Cramer–Rao lower bounds). MR diffusion tensor images were acquired using a four-shot DT-echo planar imaging sequence (TR = 3.7 s, TE = 20 ms, B0 = 1,000 s/mm2 ) with a 10-ms interval (Δ) between the application of diffusion gradient pulses, a 4-ms diffusion gradient duration (δ), a gradient amplitude (G) of 46.52 mT/m, and the Jones 30 gradient scheme.

Postprocessing analysis was performed using Diffusion Toolkit software (http://trackvis.org/). The cingulate cortex of each rat was selected and the fractional anisotropy (FA) and mean diffusivity (MD) were calculated from the diffusion tensor parametric maps. Subsequently, repeated measures ANOVA and *t*-tests were conducted to test for the treatment effect of the different diffusion parameters.

#### Behavioral Testing

After NMDA-induced spasms, rats performed an open-field test on P19 and fear conditioning tests on P27–29 (**Figure 1**). Behavioral experiments were conducted in a standard behavioral testing room during the light phase (08:00–20:00 h) of a regular 12-h light–dark cycle.

#### Open-Field Test

The locomotive activity of controls and rats after NMDAtriggered spasms was assessed using an open-field test as previously described (20). Rats were placed into a black plastic box (60 cm × 60 cm2 field with a 30-cm high perimeter) for 5 min, and their activities were monitored (3, 21) using computerized motion-tracking software (SMART 3.0; Panlab S.L.U., Barcelona, Spain). The center was defined as the middle area 10 cm apart from each wall and the other area was defined as the periphery (i.e., 2,000 cm2 periphery and 1,600 cm2 center; allocation of the area = 5:4).

#### Fear Conditioning

The observation chamber (25 cm × 25 cm × 25 cm; Panlab s.l.u.) was constructed of aluminum (two side walls and ceiling) and Plexiglas (rear wall and hinged front door) and was situated in a soundproof box. The floor of the chamber consisted of 19 stainless steel rods (4 mm diameter) spaced 1.6 cm apart (center to center), which were connected to a shock generator and grid scrambler (Panlab s.l.u.). A tone (conditioned stimulus) was delivered by a speaker mounted on one side panel of the chamber and both the shock and tone deliveries were controlled by a computerized system.

Tone conditioning was conducted on P27. Modified from a previous report (22), the conditioning consisted of a 3-min baseline habituation in the experimental chambers, followed by exposure to five pairings of a tone (conditioned stimulus; 30 s, 85 dB, 2,000 Hz) with white-light illumination, each ending with a final footshock (unconditioned stimulus; 0.8 mA, 2.0 s) followed by a 20-s silent interval. Response to the context was monitored on P28 (24 h after the conditioning trial). The rats were returned to the chamber for 5 min, and freezing behavior was measured in response to the context. Fear response to the conditioned stimulus was conducted at 24 h after the contextual fear testing (P29). To reduce the influence of context on cued fear conditioning, tactile and visual cues were manipulated with the replacement of a wall and a floor. Following a 2-min period without a conditioned stimulus (first session, unconditioned), the rats were presented with five tone parings (85 dB, 2 kHz, 30 s, session Nos. 2, 4, 6, 8, and 10; 20 s between tone presentations, session Nos. 3, 5, 7, 9, and 11) without any foot shock. The behaviors of the rats were recorded and analyzed using the signal generated by a high-sensitivity weight transducer system (Panlab s.l.u.).

### Sample Preparation and Immunohistochemistry

The animals with three bouts of spasms on P12, P13, and P15 and the corresponding controls were transcardially perfused with 4% paraformaldehyde on P35 under deep anesthesia, and their brains were removed and cryoprotected. Twenty-micron serial coronal sections were cut on a cryocut microtome.

For immunohistochemistry, myelin basic protein (MBP) antibody (1:1,000; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), glial fibrillary acidic protein (GFAP) antibody (1:2,000; EMD Millipore, Billerica, MA, USA), or neuronal nuclei (NeuN) antibody (1:2,000; EMD Millipore) were used as primary antibodies. All slides were examined by light microscopy (Olympus BX-53; Olympus Co., Tokyo, Japan). Semi-quantification analysis of immunohistochemical staining was performed using ImageJ (NIH, Bethesda, MD). The corresponding area in the cingulate cortex cg2 subfield was analyzed to determine the NeuN- and GFAP-positive cell count and the stained area measurement of MBP. Three or five randomly selected areas (1.0 × 104 sq μm) on each section were used for NeuN- and GFAP-positive cell counting. The same areas of binarized images of MBP-stained sections were quantified from each animal.

### Western Blot Analysis

After deep anesthesia, bilateral cortical tissues from bregma to the posterior hippocampal area without the hippocampus (anterior posterior 0 to −5 mm) were collected from controls and rats with NMDA-triggered spasms on P35. Anti-MBP (1:2,000; Santa Cruz Biotechnology, Inc.), anti-NeuN (1:5,000; EMD Millipore), or anti-GFAP (1:5,000; EMD Millipore) antibodies were used. Normalization was performed by the development of parallel western blots probed with β-actin (1:20,000; Santa Cruz Biotechnology Inc.) antibody analyzed by densitometry using ImageJ software (NIH, Bethesda, MD).

### Statistics

The Mann–Whitney's *U* test was used for the comparison of two groups and Student's *t*-test was used for the data following normal distribution. The concentrations of metabolites were compared between the groups and time points using a linear mixed model. The absence of significant sex-based differences (*p* always > 0.10) was confirmed in advance, and both sexes were combined. The significance level was preset to *p* < 0.05. Statistical analyses were conducted using SAS® Version 9.3 (SAS Institute Inc., Cary, NC, USA) and SPSS 18.0 (IBM Corp., Armonk, NY, USA).

## RESULTS

### Mapping of the Glutamate Concentration at Acute Periods After Three Rounds of NMDA Spasms

There were significant increases in the GluCEST (%) levels of rats with three rounds of NMDA-induced spasms in the cortex and hippocampus at P15 (N = 14; cortex, 13.20 ± 0.74%, *p* = 0.011; hippocampus, 12.36 ± 0.91%, *p* = 0.016; striatum, 10.10 ± 0.96%, *p* = 0.062) compared with controls (N = 14, 9.71 ± 0.82%, 9.32 ± 0.81%, 7.23 ± 1.00% in each area). There was a developmental increase in the GluCEST (%) at P22 but no significant difference in the GluCEST (%) between the rats with spasms and controls (**Figure 2**).

### Neurochemical Alterations in the Cingulate Cortex After Multiple NMDA-Triggered Spasms

There was no significant morphological difference between the two groups (NMDA-induced spasms [*n* = 12] and controls [*n* = 10]) on visual evaluation of serial T2-weighted MR images. One day after the last NMDA injection (P16), GABA (*p* = 0.008), Glx (Glu + Gln, *p* = 0.024), tNAA (*p* = 0.006), tCho (*p* = 0.014), and tCr (*p* = 0.007) levels measured from cingulate cortex (**Figure 3A**) were significantly increased in rats with NMDAinduced spasms (**Figures 3B,E,G,H**). At 1 week (P23) after the

Figure 2 | (A) After three rounds of spasms on postnatal day (P) 12, 13, and 15, glutamate chemical exchange saturation transfer (GluCEST, %) levels were measured in the cortex, hippocampus, and striatum on P15. (B) There were significant increases in the GluCEST (%) of rats with three rounds of *N*-methyl-d-aspartate (NMDA)-induced spasms compared with controls in all measured areas. A developmental increase in the GluCEST (%) from P15 to P22 was noted, and there was no significant difference in the GluCEST (%) between rats with spasms and controls.

NMDA-induced spasms, there were also no significant differences in the concentrations of the neurochemicals measured in the cingulate cortex between the two groups. At 2 weeks (P30) after the NMDA-induced spasms, Tau (*p* = 0.016), tNAA (*p* = 0.026), and MM + Lips (*p* = 0.042) were significantly decreased compared with controls (P30; **Figures 3D–F**). The temporal changes in each neurometabolite were compared between the groups, and glutamine and glutamate (Glx, *F* = 3.30, DF1 = 2, DF2 = 34.7, *p* = 0.049), Tau (*F* = 10.52, DF1 = 2, DF2 = 32.5, *p*< 0.001), tNAA (NAA + NAAG, *F*= 6.57, DF1 = 2, DF2 = 15.7, *p* = 0.008), MMs and lipids (MM + Lips, *F* = 4.01, DF1 = 2, DF2 = 15.4, *p* = 0.039), and total choline (tCho, *p* = 0.006) were significantly different between the two groups (**Figures 3C–G**).

#### In Vivo Microstructural Alterations Following Multiple Bouts of NMDA-Triggered Spasms

There was no significant difference in the FA values at each time point between rats with spasms (N = 15) and controls (N = 15). The rats with multiple bouts of NMDA-triggered spasms showed a significantly decreased MD at P16 and P30 in both cingulate cortices (right/left cingulate cortex: P16, 6.64 × 10<sup>−</sup><sup>4</sup> /6.59 × 10<sup>−</sup><sup>4</sup> , *p*= 0.012/*p*= 0.001; P30, 6.05 × 10<sup>−</sup><sup>4</sup> /6.14 × 10<sup>−</sup><sup>4</sup> ; *p*= 0.036/*p*= 0.016) compared with controls (P16, 7.01 × 10<sup>−</sup><sup>4</sup> /7.06 × 10<sup>−</sup><sup>4</sup> ; P30, 6.48 × 10<sup>−</sup><sup>4</sup> /6.74 × 10<sup>−</sup><sup>4</sup> , **Figure 4**).

### Behavioral Changes After Multiple Bouts of NMDA-Triggered Spasms

To evaluate the behavioral changes that occurred after multiple bouts of NMDA-induced spasms, the results of open-field and fear conditioning tests were compared between the rats that had experienced three bouts of NMDA-triggered spasms and controls. In the open-field test on P19, the rats with NMDAinduced spasms (*n* = 23) traveled a significantly shorter distance through the peripheral area (*p* = 0.005) as well as in the total area (*p* = 0.006) compared with controls (*n* = 20; **Figure 5A**). Resting time in the peripheral area was significantly increased in rats with NMDA-induced spasms (*p* = 0.012) compared with controls. The time spent in slow and fast motion in the peripheral area was significantly decreased in the rats with NMDA-induced spasms compared with controls (*p* = 0.014 and *p* = 0.010, respectively; **Figure 5B**). The time spent in the center was equivalent in the two groups.

In fear conditioning tests on P27–29, there was no difference in tone conditioning on P27 or response to context on P28. However, on P29, the total freezing durations were significantly increased in rats with NMDA-induced spasms compared with controls (*p*= 0.037; **Figure 5C**). During the first session, freezing durations were not different between the two groups (*p* = 0.058), but freezing durations were significantly increased in rats with NMDAinduced spasms at conditioned sessions from 6 to 9, in particular (*p* = 0.007, 0.002, 0.047, and 0.022, respectively; **Figure 5D**).

#### Reduced GFAP, MBP, and NeuN Expression in the Juvenile Period After Multiple Bouts of NMDA-Triggered Spasms

Semi-quantitative analysis of immunohistochemical staining on P35 revealed a significant reduction in NeuN- and GFAP-positive cells (both *p* < 0.001) and MBP-positive area (*p* = 0.041) in rats with three bouts of NMDA-triggered spasms (on P12, P13, and P15, *n* = 13) compared with controls (*n* = 11) (**Figure 6**). The cortical protein expression levels of GFAP, MBP, and NeuN on P35 were quantified and compared between the rats with multiple spasms and controls. GFAP (*p* = 0.007), MBP (*p* = 0.027), and NeuN (*p* = 0.001) expression was significantly reduced in the cortex of rats with NMDA-induced spasms (*n* = 13) compared with controls (*n* = 13; **Figure 6**).

#### DISCUSSION

An early diagnosis of infantile spasms has been considered critical for improving the neurodevelopmental outcomes of affected patients (23). However, it is uncertain which part of the developing brain is affected by infantile spasms, as is the extent of any brain insult and whether it is reversible. Our present study focused on metabolic and microstructural brain insults after spasms in a rat model of infantile spasms using *in vivo* MR imaging. Behavioral tests and *ex vivo* histopathological processes were employed to support the results.

Considering the increased metabolism of brain tissue during seizures, neurochemical dysfunction might be the main

Figure 3 | Neurochemical changes in rats with *N*-methyl-d-aspartate (NMDA)-induced spasms using 1 H-MRS findings. (A) A single voxel position in the cingulate cortex (top, coronal section; bottom, horizontal section). The brain levels of GABA, glutamine + glutamate (Glx), tNAA, total choline, and tCr were significantly higher in rats with NMDA-induced spasms on P16 (B,C,E,G,H). (D–F) At 2 weeks, taurine (Tau), tNAA, and macromolecules + lipids (MM + lipids) were significantly decreased in rats with spasms. (C–G) The changes in Glx, Tau, tNAA, MM + lipids, and total choline with time were significantly different between the groups. (I) The examples of MR spectra from rats with or without 3 NMDA spasms on each time points, P16 and P30 clearly shows the neurometabolic changes after NMDA-induced spasms. \*indicates a significant difference between the groups at each time point (*p* < 0.05). Using a generalized linear mixed model, there were also significantly different interactions between the groups and time points for Glx, Tau, tNAA, MM + lipids, and total choline (each *p*-value is marked on the graph). All imaging was done on the same timepoints (P16, 23, and 30), although the data points are not aligned across groups to avoid overlapping of the two graphs.

freezing durations between the two groups. \*indicates a significant difference between the groups (\**p* < 0.05, Mann–Whitney *U* test).

pathophysiologic feature of the epileptic brain (24), especially during the acute stage. The GluCEST technique quantifies glutamate by measuring proton exchange between the amine protons of the glutamate and the water protons and the measurement is at least 100 times more sensitive than the traditional 1H MRS method (9, 25). The GluCEST is suggested as a surrogate marker of glutamate concentration in glial cells at the synaptic level (25) and glutamate contributes >90% of the GluCEST signal with <10% contribution from other metabolites in glial neuronal unit (10). Previous study with temporal lobe epilepsy patients also showed the high potential of GluCEST identifying the epileptogenic foci (26). Thus, we hypothesized that the area of hyperexcitation in rats with NMDA-induced spasms can be screened with GluCEST imaging with high spatial resolution during acute periods. GluCEST mapping showed significantly increased glutamate in cortex and hippocampus at P15 after multiple rounds of spasms (**Figure 2**), which suggests that the cortex is one of the major brain area of hyperexcitation by the NMDA-induced spasms.

Our *in vivo* serial neurochemical/microstructural analysis focused on the cingulate cortex in the rat after multiple bouts of spasms. The cingulate cortex is a region of the limbic system reciprocally linked to the hippocampus (17) and with extensive connections to the critical regions for spasm generation, including the basal ganglia, thalamus, and brainstem (16, 27). In the cingulate cortex, consistent with the findings of GluCEST imaging, acute elevations in GABA, Glx, tNAA, tCho, and tCr were noted in rats after multiple bouts of NMDA-induced spasms compared with controls (**Figure 3**). The glutamate plus glutamine levels likely represent the neuronal metabolic pool. *In vivo* human proton MRS studies (28, 29) have also reported alterations in GABA and glutamate in patients with epilepsy. In addition, choline (Cho) represents a precursor for the neurotransmitter acetylcholine (30) and creatinine is associated with energy metabolism (31, 32). The early elevation of markers of neuronal metabolism together with the excitatory neurotransmitters in the cingulate cortex may also reflect activation of the cingulate cortex during NMDA-induced spasms in this model.

With diffusion tensor imaging, we also identified an acute decrease in MD in the cingulate cortex after three bouts of spasms, which represent the early brain damage (33). MD is an inverse measure of the membrane density, is sensitive to cellularity, edema, and necrosis (34). MD decrease quickly after injury (first day) in most models studied, which is consistent with our finding at P16. The decrease was explained by the decrease in water diffusion associated with injury such as cellular edema, despite

of GFAP, MBP, and NeuN after multiple bouts of NMDA-induced spasms (\**p* < 0.05, Mann–Whitney *U* test).

there is still debate on the precise mechanism for the decrease in MD associated with injury. Previously postictal decrease of MD was reported in human (35), which is consistent with our finding and probably reflects neuronal swelling by excitotoxic injury in the brain areas involved in seizure.

In addition to these acute changes associated with spasms, *in vivo* MR imaging enables us to perform a longitudinal followup of neurometabolic and microstructural alterations in rats with NMDA-triggered spasms. NMDA is an excitatory amino acid that can cause neuronal death and glial activation is well observed in kainic acid animal model and tissues of patients with chronic epilepsy (36). However, gross morphological changes caused by NMDA-induced spasms in our model were not evident on brain MR imaging and a previous pathologic report of NMDA-induced seizures in young animals also reported no gross morphological changes on pathology (37). Consistently with this finding, FA, a gross measure of microstructural integrity marker that is less specific to the type of injury (33), is not affected in every time point of this study.

In the cingulate cortices of the juvenile rats with NMDAinduced spasms (P30), tNAA, Tau, and MM + lipids were significantly decreased with a significant reduction in MD. We also found decreased expression of neuronal, astrocyte, and myelination markers after NMDA-induced spasms in these rats. These findings suggest the different mechanism of injury during infancy compared to that of the adulthood. Instead of cell death or gliosis after excitatory injury, the developmental process of their brain, active protein synthesis and cell proliferation, may be involved. These pathologic changes are also in line with a significant reduction in neuronal markers, NAA, and Tau and markers of myelin production and substructural components such as MMs and lipids (31, 38), as well as the reduced MD reflecting decreased membrane density and cellularity (34, 39).

Previous research into epilepsy has already shown the typical developmental pattern of neurometabolites in both a rodent model (40) and in humans (32). NMDA-triggered spasms significantly impaired this time-dependent neurometabolic developmental pattern of Glx, Tau, tNAA, MM + lipids, and total choline (**Figure 3**). These combined alterations in neurometabolites and structural maturation markers in this animal model indicate the compromised development of the cingulate cortex following NMDA-triggered spasms. Also, the decrease of all proteins examined can be explained by a decrease in protein synthesis during the critical brain development after NMDA-triggered spasms.

Furthermore, decreased exploratory behaviors in the peripheral areas and increased freezing activities to conditioned sound and light stimuli were also observed in our study group that suffered from NMDA-induced spasms. Other behavioral disruptions were found in previous studies with similar models (21, 37). Considering the widespread reduction in neuronal metabolites in neurodevelopmental disorders (31) and a positive correlation between motor skill learning and myelination (41), the pathological changes in the brain might explain these behavioral changes.

In conclusion, NMDA-induced spasms during infancy lead to time-dependent neurochemical and microstructural changes in the cingulate cortex and subsequent pathologic changes during the juvenile period. These age-dependent alterations in neurometabolites after NMDA-triggered spasms should be further explored as potential biomarkers of outcomes in human infantile spasms or other epileptic encephalopathies.

#### ETHICS STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

### AUTHOR CONTRIBUTIONS

ML: manuscript writing, experiments, and statistics; MSY: main idea, manuscript writing and revision, and experiments guiding;

#### REFERENCES


DCW: manuscript revision and GluCEST analysis; WHS: DTI, FA, and MD analysis; TSK: manuscript revision; LV: manuscript revision.

#### ACKNOWLEDGMENTS

The authors gratefully acknowledge technical support from the Biomedical Imaging Infrastructure, Department of Radiology, Asan Medical Center.

### FUNDING

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT and Future Planning (NRF-2016R1C1B2010078) and a grant (2017-551) from Asan Institute for Life Sciences, Seoul, Korea. This article has not been published or presented in elsewhere.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fneur.2018.00248/ full#supplementary-material.

undiscovered lesioned areas in kainate model of epilepsy in rat. *Brain Struct Funct* (2011) 216(2):123–35. doi:10.1007/s00429-010-0299-0


disorders in children with infantile spasms? *Epilepsia* (2015) 56(6):856–63. doi:10.1111/epi.12997


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

# Fentanyl induces cerebellar internal granular cell layer apoptosis in healthy newborn Pigs

*Hemmen Sabir1,2, John Dingley3 , Emma Scull-Brown1 , Ela Chakkarapani1 and Marianne Thoresen1,4\**

*1Neonatal Neuroscience, School of Clinical Sciences, University of Bristol, St. Michael's Hospital, Bristol, United Kingdom, 2Department of Pediatrics I/Neonatology, University Hospital Essen, University Duisburg-Essen, Essen, Germany, 3Swansea University College of Medicine, Swansea, United Kingdom, 4Division of Physiology, Institute of Basic Medical Sciences, University of Oslo, Oslo, Norway*

Background: Opioids like fentanyl are regularly used in neonates for analgesia and sedation. So far, they have been reported to be safe and eligible to use. The cerebellum has become a focus of neurodevelopmental research within the last years, as it is known to play an important role in long-lasting motor, cognitive, and other behavioral changes. The cerebellar cortex is of major importance in the coordinative role of the cerebellum and highly vulnerable to injury and impaired growth.

#### *Edited by:*

*Olivier Baud, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Maryline Lecointre, INSERM UMR1245 Genomic and Personalized Medicine in Cancer and Neurological Disorders, France Bruno J. Gonzalez, Université de Rouen, France*

#### *\*Correspondence:*

*Marianne Thoresen marianne.thoresen@bristol.ac.uk*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 08 December 2017 Accepted: 16 April 2018 Published: 01 May 2018*

#### *Citation:*

*Sabir H, Dingley J, Scull-Brown E, Chakkarapani E and Thoresen M (2018) Fentanyl Induces Cerebellar Internal Granular Cell Layer Apoptosis in Healthy Newborn Pigs. Front. Neurol. 9:294. doi: 10.3389/fneur.2018.00294*

Objective: This study was performed to evaluate the apoptotic effect of intravenous fentanyl infusion on the cerebellum in healthy newborn pigs.

Methods: Thirteen healthy pigs (<median 12 h old) were randomized into (1) 24 h of intravenous fentanyl at normothermia (NTFe, *n* = 6) or (2) non-ventilated controls at normothermia (NTCTR, *n* = 7). Cerebellar sections were morphologically assessed after staining with hematoxylin–eosin. In addition, paired sections were immuno-stained for cell death [Cleaved caspase-3 and terminal deoxynucleotidyl transferase-mediated deoxyuridine-triphosphate nick-end labeling (TUNEL)], and positive cells were counted in defined areas of the internal granular cell layer. In total, cells in three cerebellar gyri were counted.

results: We found that there was an increase in cells with apoptotic morphology in the internal granular cell layer in the NTFe group. For quantification, we found a significant increase in cell death in group (1) [median (range) number of caspase-3-positive cell group (1) 8 (1–22) vs. group (2) 1 (1–6) and TUNEL-positive cells (1) 6 (1–10) vs. (2) 1 (0–4)]. In both groups, there was no difference in the number of Purkinje cells. Both groups had comparable and stable physiological parameters throughout the 24 h period.

conclusion: Twenty-four hours of continuous intravenous fentanyl infusion increased apoptosis in the internal granular cell layer in the cerebellum of healthy newborn pigs.

Keywords: newborn, brain, neurotoxicity, opioids, sedation

## INTRODUCTION

The appropriate development of the central nervous system (CNS) relies on the precise temporal– spatial pattern of complex molecular pathways guiding proliferation, migration, differentiation, and survival of neural cells (1). Interference with these pathways can disrupt physiological development patterns and may lead to permanent CNS impairments. Analgesics and sedative drugs are potent modulators of molecular pathways, and their use in preterm and term newborns has been associated with impaired long-term neurological outcomes (2). Control of pain and agitation is a fundamental component of neonatal intensive care. Opioids, especially morphine, are a commonly used analgesic in both preterm and term neonates and its use has been shown to be safe (3–5). However, high doses of morphine have also been shown to be associated with increased risk of brain injury in preterm infants (6). Fentanyl, a more potent opioid, has become an alternative to morphine in preterm and term infants. Small randomizedcontrolled trials have shown reduced stress responses in ventilated preterm infants receiving continuous fentanyl infusion, with no increased incidence of brain injury (7, 8). However, current clinical (2) and preclinical (9) data show that continuous fentanyl infusion may alter cerebellar development, leading to cerebellar hypoplasia and long-term impairments. In addition, perinatal opioid exposure has been shown to lead to cerebellar neuronal loss and cerebellar dysfunction (10).

The cerebellum has become a focus of neurodevelopmental research within the last years, as it is known to play an important role in long-lasting motor, cognitive and other behavioral changes (11–14). Input from the cerebellar cortex has a major role in the functioning of the cerebellum (coordination, precision, and accurate timing). The cerebellum is very vulnerable to injury and impaired growth (12). As the basic architecture of the cerebellar cortex is comparable between pigs and humans (15), we aimed to evaluate the apoptotic effect of continuous intravenous fentanyl on the cerebellum in healthy newborn pigs.

#### MATERIALS AND METHODS

#### Conduct of Experiment

All experiments were conducted according to the United Kingdom Home Office license guidelines and were approved by the University of Bristol Ethical Review Panel (Bristol, United Kingdom). The experimental setup is detailed in the larger experiment, where we reported the safety of 50% Xenon (Xe) ventilation in healthy newborn pigs, showing that ventilation with 50% Xe does not cause cellular injury to the newborn cerebrum (16). This study uses data from 13 healthy newborn pigs (median age 10 h, interquartile range 9–12 h) receiving intravenous fentanyl sedation, while being mechanically ventilated at normothermia (*n* = 6) or serving as control animals without special treatment (*n* = 7).

### Animal Preparation, Baseline Data, and Management of Pigs

All animals were handled as previously published (16). In brief, after initial intubation, insertion of umbilical arterial and venous catheters, continuous monitoring of mean arterial blood pressure and heart rate was enabled in the fentanyl treatment group (NTFe group) and pigs were subsequently extubated and selfventilating in air. Physiological parameters, mean arterial blood pressure and heart rate, were continuously recorded. Intensive care management was performed as previously described with 5 ml/kg/h intravenous maintenance fluid (5% dextrose/0.45% saline) in addition to being bottle fed with pig formula (Pig formula milk "Baby Lactal"; Peter Möller A/S, Oslo, Norway) at a rate of ~10 ml/ kg/h. Control pigs (NTCTR group) were self-ventilating in air and bottle fed every 2–3 h with pig formula to maintain a similar fluid intake. Blood sampling was undertaken from the inserted lines at preset time points, as well as when clinically indicated. Frequent temperature measurements were undertaken with a rectal probe (reusable YSI 400 series, CritiCool, MTRE, Yavne, Israel) inserted 6 cm into the rectum, and a skin probe (CritiCool, MTRE, Yavne, Israel) sited on the ear lobe. Both probes were calibrated before use within ±0.1°C, over a temperature range of 20–40°C, against a certified mercury-in-glass thermometer (BS593; Zeal, London, United Kingdom). Rectal temperature (*T*rec) was maintained at 38.5 ± 0.2°C using a servo-controlled (CritiCool, MTRE, Yavne, Israel) mat containing circulating water.

#### Fentanyl Sedation

After intubation and vascular umbilical cord access, continuous fentanyl infusion was started with a bolus of 10 µg/kg followed by maintenance infusion with 1 µg/kg/h. Thereafter, the fentanyl infusion was adjusted to achieve adequate sedation and tolerance of the central and continuous arterial blood pressure monitoring lines. Mean arterial blood pressure was higher than 40 mmHg in all pigs throughout the 24-h treatment period, providing an adequate cerebral blood flow for newborn pigs (16–18).

### Neuropathology Assessment

After 24 h of allocated treatment, all pigs were intubated and deeply anesthetized with isoflurane (16). Brains were slowly flushed with 0.9% saline through the common carotid arteries followed by perfusion fixation with 10% neutral buffered formalin and dissected out. The cerebellum was removed and the hemispheres divided. The right hemisphere was coronally cut into 5-mm blocks and paraffin embedded. Two representative blocks of the left cerebellum were chosen, best presenting the cortex and white matter regions of the cerebellum (**Figure 1A**). Hematoxylin and eosin (H&E)-stained 5-µm thick sections were assessed at

Figure 1 | Photomicrography of histological features. (A) Assessed area of the internal granular cell layer of three complete gyri of the anterior cerebellar lobe as highlighted in blue. (B,C) Representative images of the inner granular cell layer of the NTCTR and NTFe groups are shown. Arrows indicate cells with homogenous eosinophilic cytoplasm and pyknotic nuclei.

40× magnification. Three complete gyri of the anterior lobe of the cerebellum were assessed (**Figure 1A**), and cells were scored as apoptotic when showing typical morphology of apoptosis (19).

#### Immunohistochemistry

Immunohistochemical staining was performed as previously described (16). Briefly, slides were prepared from paraffinembedded sections. For quantification of apoptotic cells, two adjacent sections were stained with Cleaved caspase-3. Primary rabbit antibody against Cleaved caspase-3 [1:500, polyclonal rabbit anti-Cleaved caspase-3 (ASP175) Cell Signalling Technologies] was applied overnight at room temperature. In addition, for the assessment of DNA fragmentation, the adjacent sections were stained with terminal deoxynucleotidyl transferase-mediated deoxyuridine-triphosphate nick-end labeling (TUNEL). TUNEL staining was performed as instructed by the manufacturer (TUNEL AP, cat. no. 11772457001, Roche).

For each animal, three complete gyri were counted for Cleaved caspase-3 and TUNEL-positive cells at 40× magnification. Total cell counting was performed in three non-overlapping fields, each sized 2,000 µm × 200 µm representing the three gyri assessed by H&E staining, by three independent observers blinded to the randomization and to clinical details of the pigs.

#### Statistical Analysis

Statistical analysis was performed with SPSS version 22 (SPSS Inc., Chicago, IL, USA). The Wilcoxon test was used for the two-group comparison. To assess the effect of sex and age since birth in hours on the number of Cleaved caspase-3 and TUNELpositive cells, regression analysis was used. Two-sided testing with *p* < 0.05 was considered statistically significant. Data are presented as median (interquartile range).

#### RESULTS

#### Physiological Data

There were no significant differences in baseline physiological parameters between the NTFe and the NTCTR group (**Table 1**). Blood gases, blood glucose, and lactate values were within the normal range in all animals.

#### Histological Results

There was a notable difference in H&E-stained sections of the internal granular cell layer in the NTFe group compared with the NTCTR group (**Figures 1B,C**). Cells in the NTFe group appeared with nuclear condensation and fragmentation as seen when thought to undergo apoptotic cell death (**Figure 1C**). As previously described, the Purkinje cell layer showed no signs of apoptosis.

#### Immunohistochemistry

Immunohistochemistry showed a significant increase of Cleaved caspase-3 (*p* = 0.035) and TUNEL (*p* = 0.023) positive cells in the internal granular cell layer of pigs from the NTFe group compared with the NTCTR group, analyzed in the standardized area of tissue (**Tables 2** and **3**; **Figures 2** and **3**). Regression analysis Table 1 | Baseline parameters during 24 h treatment period.


*Median (interquartile range) baseline parameters (weight, sex, and age), heart rate, arterial blood pressure, transcutaneous oxygen saturation (tcSO2), pH, blood glucose, and lactate levels of all 13 animals in the two treatment groups and median fentanyl dosage for 6 pigs receiving iv fentanyl.*

*IQR, interquartile range.*

*n/a: Data were not applicable; these animals were not cardiovascularly monitored as they were non-instrumented control animals (NTCTR).*

Table 2 | Cleaved caspase-3-positive cell counting results.


*Total cell count for Cleaved caspase-3-positive cells per standardized area of tissue. Results are presented as median (interquartile range) number of Cleaved caspase-3 positive cells per three non-overlapping fields, each sized 2,000 µm* × *200 µm. In the analyzed area, there was a significant increase in Cleaved caspase-3-positive cells in the NTFe group in the inner granular cell layer in the two-group comparison. NTFe: 24 h iv fentanyl at normothermia. NTCTR: controls at normothermia.*

Table 3 | Terminal deoxynucleotidyl transferase-mediated deoxyuridine-

triphosphate nick-end labeling (TUNEL) cell counting results.


*Total cell count for TUNEL-positive cells per standardized area of tissue. Results are presented as median (interquartile range) number of TUNEL-positive cells per three non-overlapping fields, each sized 2,000 µm* × *200 µm. In the analyzed area, there was a significant increase in TUNEL-positive cells in the NTFe group in the inner granular cell layer in the two-group comparison.*

*NTFe: 24 h iv fentanyl at normothermia.*

*NTCTR: controls at normothermia.*

showed no effect of sex or age on Cleaved caspase-3 or TUNELpositive cells.

### DISCUSSION

This study shows that 24 h of a clinical dose of continuous intravenous fentanyl administration significantly increases apoptotic

Figure 2 | Representative images of the inner granular cell layer after Cleaved caspase-3 staining of the NTCTR and NTFe groups. Arrows indicate Cleaved caspase-3-positive cells.

nick-end labeling (TUNEL) staining of the NTCTR and NTFe groups. Arrows indicate TUNEL-positive cells.

cell death in the internal granular cell layer of the cerebellum in healthy newborn pigs. As previously reported, there was no increase of apoptosis, neither in the Purkinje cell layer in the cerebellum nor in other parts of the cerebrum in the same pigs (16). The main purpose of our previous paper (16) was to investigate whether, 50% inhaled Xe gas induces apoptosis in the healthy newborn pig brain—which it did not. Our clinical feasibility study of therapeutic hypothermia (TH) + Xe in asphyxiated term newborns therefore followed this (20, 21).

Preterm and term newborns undergo various painful procedures during their stay in the neonatal intensive care unit. In particular, term asphyxiated newborns, undergoing TH, often require unavoidable painful or stressful procedures such as intubation, mechanical ventilation, or catheterization and of course a reduced core temperature of 33.5°C. It has been shown that stress reduces the neuroprotective effect of TH (22), and therefore routine sedation is required during hypothermia treatment. Opioids have long been used for neonates undergoing painful procedures. Morphine, as the most commonly used opioid in neonates, has been shown to be safe in preterm (3, 5) and term asphyxiated neonates (4) in normal clinical dosages, without causing side effects like hypotension. In the Neurological Outcome and Preemptive Analgesics in Neonates trial, continuous morphine infusion did not increase vulnerability of ventilated preterm or term neonates to adverse neurological events and no relationship among morphine use, blood pressure variability, and intraventricular hemorrhage could be determined (6). However, additional doses of morphine were associated with an increased risk of brain injury (6). It has been shown robustly in different animal models that intrauterine and postnatal morphine exposure leads to altered brain function and reduced brain growth (23, 24). It might be that in children, standard outcome measures at 2 years of age, do not fully answer the question of long-term safety. Another explanation might be that the newborns in need of continuous opioid infusion are the sickest of preterm and term children, with many other risk factors for impaired neurological long-term outcome. Opioid analgesics act on different opioid receptors (μ-, δ-, or κ-type), which after activation, initiate multiple intracellular signaling cascades (25, 26). Of concern, these signaling pathways are implicated in various other biological processes, including the modulation of proliferation, survival, and differentiation of the neural stem cells, neurons, or glia cells (25, 27). These modulations might alter brain development, and therefore further detailed analysis of the developing brain is needed.

The cerebellum has become a focus of neurodevelopmental research within the last few years, as it is known to play an important role in long-lasting motor, cognitive, and other behavioral changes (11–14). The cerebellar cortex is of major importance to the main roles of the cerebellum (coordination, precision, and accurate timing) and highly vulnerable to injury and impaired growth (12). The basic architecture of the cerebellar cortex is comparable between pigs and humans (15), consisting of the internal granular cell layer, the Purkinje cell layer, and the superficial molecular layer (28). During the third trimester, a rapid cerebellar growth takes place (29). During normal development, the Purkinje neurons are the first neurons to be generated, and they are already mature during early fetal life (12). These Purkinje neurons are important, as they are the only efferent cells, projecting to the outside of the cerebellar cortex (30). The internal granular cell layer forms an important filter of information between mossy fiber inputs and the Purkinje cells (31). Around the time of birth and during postnatal life, the internal granular cell layer is highly active, as granular cells from the external granular cell layer migrate radially inward along the Bergmann glia to the internal granular cell layer (12, 30). During this migration phase, the internal granular cell layer is highly vulnerable. This has also been described in newborn pigs (32). Comparable to humans, the pig cerebellum is not fully mature at birth (32), and full maturation appears several months after birth (33, 34). However, the external granular cell layer is not visible in term born newborn pigs making the brain slightly more mature compared with human newborns. Whereas the granular cells play an important filter between incoming information *via* Mossy fibers and outgoing information *via* Purkinje cell axons, altered growth, and development will have long-lasting effects on cerebellar function. Strackx et al. have previously shown in a fetal sheep model that prenatal intra-amniotic injection of lipopolysaccharide, mimicking chorioamnionitis, leads to altered granule cells and astrocytes in the internal granular cell layer, without affecting Purkinje cells or cell layer volumes (35). Even though they found an increase in granule cells, they also showed that the Purkinje cells were not altered by intra-amniotic infection. As in the newborn pig, the Purkinje neurons in the fetal sheep are already present early during fetal development, not being vulnerable around the time of birth. This explains our finding of normal Purkinje cell counts in our experimental setup. However, in this study, we demonstrate acute apoptosis in the internal granular cell layer, most likely caused by the continuous fentanyl administration. Due to the present acute experimental setup and animal legislation, we are unable to undertake longterm survival studies showing possible long-lasting effects like cerebellar growth impairment and altered neuro-functional outcome. However, it has been shown in neonatal rodents that intrauterine (10) and postnatal morphine exposure alters cerebellar growth and Purkinje cell survival (9, 36, 37). Compared with large animal models (pigs or sheep), the rodent cerebellum develops and matures postnatally, and therefore Purkinje cells are highly vulnerable in rodents, explaining the mentioned results. The use of fentanyl in preterm and term infants has increased in the last years, even though little is known regarding its effect on brain development and maturation (38, 39). Fentanyl is a potent synthetic μ-opioid receptor agonist. Small randomizedcontrolled trials claimed to have shown its feasibility and safety during continuous infusion in preterm infants (7, 8). We show here that fentanyl increases apoptosis in the internal granular cell layer of healthy newborn pigs. In preterm infants, McPherson et al. have shown that high cumulative fentanyl doses in preterm infants correlate with a higher incidence of cerebellar injury and lower cerebellar diameter at term equivalent age assessed by magnetic resonance imaging (MRI) (2). Both studies raise concerns over cumulative fentanyl use in preterm and term neonates. In addition, Zwicker et al. showed that preterm infants exposed to high cumulative morphine exposures had impaired cerebellar growth in the neonatal period and poorer neurodevelopmental outcomes in early childhood (40). As the cerebellum has not been the focus of previous reports on the safety and outcome of morphine or fentanyl use in preterm and term neonates, ours and the before mentioned results raise new concerns regarding its use in this patient population. Due to enhanced MRI imaging techniques, the focus of researchers and clinicians on the developing cerebellum is growing. Disruption of normal cerebellar development due to cell death in the internal granular cell layer may have long-lasting neurobehavioral effects.

There are limitations to our study. First, median, and therefore cumulative fentanyl dosages were within the higher range of normal dosage in pigs. However, we used these high dosages, as healthy pigs were self-ventilated in addition to the set mechanical ventilatory rates under fentanyl sedation, requiring high dosages of fentanyl in the original paper (16). Even though our dosages were much higher than the ones normally used in neonates, we did not experience side effects like arterial hypotension or apnea. Therefore, we claim that the increased apoptosis is due to the cumulative fentanyl dose. Second, we did not analyze longterm outcome in our study, due to the acute experimental setup and animal legislation of the original study. From the findings in fetal sheep (35) or rodents (9, 10, 36), one would expect to find long-term deficits and cerebellar growth restriction in our pigs as compared with the other animal models. Third, we only performed a subgroup analysis with a limited number of animals. However, as our data robustly shows a significant increase of apoptosis in pigs exposed to high intravenous fentanyl exposure, we do not believe that enlarged group sizes would have shown different results. Last, further detailed caspase-3-dependent apoptotic pathway analyses would have strengthened our findings and might have led to the investigation of mechanisms of fentanyl induced apoptosis in our pigs. However, we were unable to perform these analyses at the current time point due to the retrospective character of the study.

Control of pain and agitation is a fundamental component of neonatal intensive care. TH is certainly stressful, even though being the only available standard treatment for neonatal encephalopathy (41). TH reduces the risk for death and adverse neurodevelopmental outcome in moderately asphyxiated newborns (42). There is increasing use of TH in mildly asphyxiated newborns (43). In these patients, high sedative and analgesic dosages of opioids will be needed to tolerate the stress of being cold compared with comatose patients. As these newborns will most likely not develop brain injury due to mild asphyxia, they are at high risk of cerebellar impairment due to the required use of fentanyl or morphine during TH. Careful patient selection and classification is needed to identify asphyxiated newborns developing moderate to severe encephalopathy and to prevent non-beneficial over-treatment of patients. In this study, we found that 24 h of intravenous fentanyl increased apoptosis in the internal granular cell layer in the cerebellum of healthy newborn pigs.

#### ETHICS STATEMENT

All experiments were conducted according to the United Kingdom Home Office license guidelines and were approved by the University of Bristol Ethical Review Panel (Bristol, United Kingdom).

### AUTHOR CONTRIBUTIONS

HS, JD, and MT have planned and designed the study; HS, ES-B, and JD have performed the animal experiments; HS and MT have analyzed the data; HS, JD, EC, and MT have written and corrected the manuscript.

### ACKNOWLEDGMENTS

We thank Elke Maes for technical help during the experiments and analysis.

### FUNDING

This study was supported by Sport Aiding Medical Research for Kids [SPARKS (UK)], the German Academic Exchange Service, the Laerdal Foundation for Acute Medicine (Norway), and the Norwegian Research Council.

## REFERENCES


encephalopathy: a feasibility study. *Pediatrics* (2014) 133(5):809–18. doi:10.1542/ peds.2013-0787


Treatment Recommendations. *Circulation* (2010) 122(16 Suppl 2):S516–38. doi:10.1161/CIRCULATIONAHA.110.971127


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

# Grafting Neural Stem and Progenitor Cells Into the Hippocampus of Juvenile, Irradiated Mice Normalizes Behavior Deficits

Yoshiaki Sato1,2, Noriko Shinjyo<sup>1</sup> , Machiko Sato1,3, Marie K. L. Nilsson<sup>4</sup> , Kazuhiro Osato1,5 , Changlian Zhu1,6, Marcela Pekna<sup>1</sup> , Hans G. Kuhn<sup>1</sup> and Klas Blomgren1,7,8 \*

<sup>1</sup> Center for Brain Repair and Rehabilitation, Institute of Neuroscience and Physiology, University of Gothenburg, Gothenburg, Sweden, <sup>2</sup> Division of Neonatology, Center for Maternal-Neonatal Care, Nagoya University Hospital, Nagoya, Japan, <sup>3</sup> Department of Obstetrics and Gynecology, Narita Hospital, Nagoya, Japan, <sup>4</sup> Institute of Neuroscience and Physiology, University of Gothenburg, Gothenburg, Sweden, <sup>5</sup> Department of Obstetrics and Gynecology, Mie University, Tsu, Japan, <sup>6</sup> Department of Pediatrics, The Third Affiliated Hospital of Zhengzhou University, Zhengzhou, China, <sup>7</sup> Department of Pediatric Oncology, Karolinska University Hospital, Stockholm, Sweden, <sup>8</sup> Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden

#### Edited by:

Olivier Baud, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Ivo Bendix, Universitätsklinikum Essen, Germany Ryan J. Felling, Johns Hopkins University, United States

> \*Correspondence: Klas Blomgren klas.blomgren@ki.se

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 31 January 2018 Accepted: 08 August 2018 Published: 11 September 2018

#### Citation:

Sato Y, Shinjyo N, Sato M, Nilsson MKL, Osato K, Zhu C, Pekna M, Kuhn HG and Blomgren K (2018) Grafting Neural Stem and Progenitor Cells Into the Hippocampus of Juvenile, Irradiated Mice Normalizes Behavior Deficits. Front. Neurol. 9:715. doi: 10.3389/fneur.2018.00715 The pool of neural stem and progenitor cells (NSPCs) in the dentate gyrus of the hippocampus is reduced by ionizing radiation. This explains, at least partly, the learning deficits observed in patients after radiotherapy, particularly in pediatric cases. An 8 Gy single irradiation dose was delivered to the whole brains of postnatal day 9 (P9) C57BL/6 mice, and BrdU-labeled, syngeneic NSPCs (1.0 × 10<sup>5</sup> cells/injection) were grafted into each hippocampus on P21. Three months later, behavior tests were performed. Irradiation impaired novelty-induced exploration, place learning, reversal learning, and sugar preference, and it altered the movement pattern. Grafting of NSPCs ameliorated or even normalized the observed deficits. Less than 4% of grafted cells survived and were found in the dentate gyrus 5 months later. The irradiation-induced loss of endogenous, undifferentiated NSPCs in the dentate gyrus was completely restored by grafted NSPCs in the dorsal, but not the ventral, blade. The grafted NSPCs did not exert appreciable effects on the endogenous NSPCs; however, more than half of the grafted NSPCs differentiated. These results point to novel strategies aimed at ameliorating the debilitating late effects of cranial radiotherapy, particularly in children.

Keywords: neural stem progenitor cells, irradiation, transplantation, grafting, learning deficits, developing brain, late effects

### INTRODUCTION

The second most common type of childhood cancer is brain tumors, amounting to nearly one-third of all childhood cancers. The survival rates of pediatric brain tumor patients has improved in recent decades, and currently more than 80% of them survive (1). Although neurosurgery techniques and chemotherapy regimens have improved, radiation therapy (RT) is still an essential treatment modality not only for malignant brain tumors, but also for central nervous system (CNS) leukemia/lymphoma. However, RT causes both increased mortality and morbidity, in survivors of pediatric brain tumors (2). Cognitive impairments, as well as perturbed growth and puberty are

**262**

some of the known late effects observed after RT. Moreover, it has been shown that the cognitive deficits are more severe in younger children after RT (3, 4), and the deficits increase progressively over time. It is unclear whether low doses of ionizing radiation (less than 0.5 Gy) administered to the CNS can cause cognitive impairment. One study claimed to find such effects (5), whereas a similar but larger study failed to do so (6). Ameliorating the late effects of RT will improve the quality of life of cancer survivors, whose prevalence is increasing, and particularly of pediatric cases, whose remaining life expectancy is long.

Irradiation (IR) causes injuries to many brain regions and cell types; however, the underlying pathogenesis is not clear. Neurogenesis persists throughout life in two regions, the subventricular zone (SVZ) and the dentate gyrus of the hippocampus. These regions harbor proliferating cells, and are therefore particularly susceptible to IR (7). Several reports suggest that injury to neural stem and progenitor cells (NSPCs) in the hippocampus can cause some of the late effects observed after IR (8–10), and the depletion of NSPCs induced by IR appears to be long-lasting, even after a single, moderate dose of IR (11, 12). The depletion is even aggravated over time (13). Currently, there are no established interventions after RT, but it was shown that memory training improved the attention and memory performance of children treated for medulloblastoma (14). Voluntary physical exercise increases the number of stem cells and enhances neurogenesis after IR of the young mouse brain, and at least partly normalizes IR-induced behavior changes (8, 15). In one study, human embryonic stem cells grafted into the hippocampus of adult immune-deficient rats improved their performance in a memory task after IR (16). On the other hand, grafting of syngeneic enteric neural stem cells (17) or syngeneic cerebral neural stem cells (18) into the hippocampus of irradiated mice caused local gliosis. Given the importance of inflammationrelated signaling in the brain under both normal and pathological conditions, the aim of the present study was to explore whether grafting NSPCs into the hippocampus of immune-competent mice could ameliorate the deficits observed after IR at a young age in the absence of immunosuppressive treatment.

### MATERIALS AND METHODS

#### Animals

All animal experimental protocols in the present study were approved by the Gothenburg committee of the Swedish Animal Welfare Agency (326-2009). C57BL/6J male pups with dams were purchased from Charles River (Sulzfeld, Germany) and maintained under a 12-h light/dark cycle with access to food and water ad libitum.

Six litters (6 pups/litter), in total 36 mice, were used for the study. Among them, 26 mice received IR on postnatal day 9 (P9). Three mice died during the IR procedure. Twenty-three irradiated mice were allocated into the NSPC (n = 13) or vehicle (n = 10) groups. All mice in the NSPC group were grafted with NSPCs and all in the vehicle group were injected with vehicle. However, 3 of 13 NSPC-grafted mice were excluded from analyses after histological evaluations as they were not grafted correctly (the cells were accidentally injected outside the hippocampus). All non-irradiated mice (n = 10) were allocated to the non-irradiated group and injected with vehicle. One IR NSPC mouse died during the IC chip insertion procedure, and one vehicle-treated mouse died before behavioral tests. Therefore, IntelliCage <sup>R</sup> was started with 13 NSPC-treated mice, 9 vehicletreated mice and 10 non-irradiated mice. In the "Introduction 2" section in IntelliCage <sup>R</sup> (**Supplementary Figure 2**), when doors were closed, 1 NSPC-treated mouse and 2 vehicle-treated mice had not been able to learn how to do a "nosepoke" by the end of the section, and could hence not drink water. Therefore, these three mice were excluded from the "First corner" section. As 2 vehicle-treated mice died between the end of the IntelliCage <sup>R</sup> and the movement pattern analysis, the subsequent two behavioral tests were performed with 12 NSPC-treated mice, 7 vehicle-treated mice and 10 non-irradiated mice. One non-irradiated mouse died, and the sections for histological evaluations were made from the remaining mice (12 NSPCtreated mice, 7 vehicle-treated mice and 9 non-irradiated mice).

## Experimental Procedures

IR was administered to P9 C57BL/6J mice, and NSPCs derived from the same mouse strain were injected into each hippocampus (in the right and left hemispheres) at P21. IR-Vehicle- treated mice received irradiation and were injected with vehicle instead of NSPCs. Non IR- Vehicle mice were only anesthetized, without irradiation, and received injections of vehicle (**Supplementary Figure 1**). Three months after grafting, the behavior of the mice was examined using three tests: IntelliCage <sup>R</sup> (P110-129), Movement pattern analysis (P162) and Sugar water (anhedonia) test (P166-170) (**Supplementary Figure 1**). After these behavioral tests, mice were sacrificed (P173, 5 months after grafting) (**Supplementary Figure 1**).

#### Irradiation Procedure

The IR procedure was performed as described earlier (18, 19). For IR, a linear accelerator (Varian Clinac 600C/D) with 4-MV nominal photon energy and a dose rate of 2.3 Gy/min was used. Nine-day-old mice were anesthetized with an intraperitoneal injection of tribromoethanol (Sigma-Aldrich, Stockholm, Sweden), placed in a prone position (head to gantry) on an expanded polystyrene bed. Both cerebral hemispheres of each animal were irradiated with a 2 × 2-cm asymmetrical radiation field. The source-to-skin distance was ∼99.5 cm. The head was covered with a 1-cm tissue equivalent. A single absorbed dose of 8 Gy was administered to each mouse. Dose variation within the target volume was estimated to be ±5%. The entire procedure was completed within 10 min. After IR, pups were returned to their biological dams until weaning. Sham control animals were anesthetized but not subjected to IR. Using the LQ-model (20) and an α:β ratio of 3 for late effects in the normal brain tissue, an acute exposure of 8 Gy is equivalent to approximately 18 Gy when delivered in repeated 2-Gy fractions (21). This represents a clinically relevant dose, equivalent to the total dose used in treatment protocols for prophylactic cranial IR in selected cases of childhood acute lymphatic leukemia.

### Culture of NSPCs

NSPCs were kindly provided by Prof. Fred H. Gage (22), which were isolated from the whole brain of adult mice, except the cerebellum or the olfactory bulbs. NSPCs were cultured as described previously (18). Briefly, NSPCs were cultured and expanded as a monolayer in Dulbecco's Modified Eagle medium (DMEM)/nutrient mixture F-12 (1:1) (Invitrogen, San Diego, CA, USA) containing 1% N<sup>2</sup> (Invitrogen), 20 ng/mL epidermal growth factor (Sigma-Aldrich, Saint Louis, MO, USA), 20 ng/mL basic fibroblast growth factor-2 (bFGF; BD Biosciences), and 5µg/mL heparin (Sigma-Aldrich). Two days before grafting, BrdU was added to the medium at a final concentration of 1.25µM. After collecting and washing, cells were suspended in DMEM containing 300 µg bFGF for grafting. More than 95% of the cells were viable as judged by a trypan blue exclusion test and more than 90% were BrdU-positive, as judged by immunostaining before grafting.

## NSPC Grafting

NSPC grafting was performed as described previously (18). Mice were mounted onto a stereotactic head holder (Kopf Instruments, Tujunga, CA, USA) under anesthesia with isoflurane (Isoba <sup>R</sup> vet; Schering-Plough Corp., NJ, USA; 5% for induction, and 2% to 3% for maintenance) in the flat skull position. In order to graft NSPCs into the hippocampus, a 5-µL 26 gauge syringe (Innovative Labor Systeme, Stuetzerbach, Germany) attached to the holder was inserted according to the following coordinates: body weight < 9 g: 0.45 × (distance from lambda to bregma) mm posterior and ± 1.2 mm lateral to bregma, 3.0 mm deep from the skull surface; body weight ≥ 9 g, 0.42 × (distance from lambda to bregma) mm posterior and ± 1.3 mm lateral to bregma, 3.2 mm deep from the skull surface. Before the needle was inserted, a small hole was drilled in the proper position according to the above coordinates. Then, 1 × 10<sup>5</sup> NSPCs in 2 µL DMEM were injected very slowly over a 2-min period, with a 4-min delay prior to removal of the syringe, and 2 min were allowed for syringe removal. No immunosuppressive drugs were administered.

### Tissue Preparation

Mice were deeply anesthetized at P173 and intracardially perfusion-fixed using 0.9% NaCl followed by buffered formaldehyde (Histofix, Histolab, Göteborg, Sweden). Brains were removed and immersion-fixed in the same solution at 4◦C for 24 h, and then, immersed in 30% sucrose for at least 2 days. Brains were cut sagittally at 30µm on a sliding microtome in dry ice.

### Microscopy and Immunohistochemistry

The following antibodies and final dilutions were used: anti-BrdU (1:500; AbD Seortec, Martinsried, Germany), goat anti-Sox2 (1:200; Santa Cruz Biotechnology, Santa Cruz, CA, USA), and rabbit anti-S100b (1:1000; Swant, Bellinzona, Switzerland). Immunoperoxidase detection of BrdU was performed as follows: free-floating sections were rinsed in Tris-buffered saline (TBS; 0.1 M Tris–HCl, pH 7.4/0.9% NaCl), and sections were then treated with 0.6% H2O2/TBS for 30 min, followed by incubation for 2 h in 50% formamide/2 × SSC (0.3 M NaCl, 0.03 M sodium citrate) at 65◦C, rinsed in 2 × SSC, incubated for 30 min in 2 N HCl at 37◦C, and rinsed in 0.1 M boric acid (pH 8.5). Incubation in TBS/3% donkey serum/0.1% Triton X (TBS++) for 30 min was followed by overnight incubation with mouse anti-BrdU. After rinsing in TBS, sections were incubated for 1 h with donkey anti-mouse-biotin (1:1,000 biotinylated donkey antimouse, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and then with avidin-biotin-peroxidase complex (Vectastain ABC Elite Kit, Vector Laboratories, Burlingame, CA, USA). This was followed by peroxidase detection for 5 min (0.25 mg/mL DAB, 0.01% H2O2, 0.04% NiCl). Immunoperoxidase detection of Sox2 was performed in the same manner with the proper secondary antibody, donkey anti-goat-biotin (1:1,000, Jackson ImmunoResearch Laboratories) as that of BrdU, except treatments for the formamide and HCl.

Triple immunofluorescence was performed as follows: freefloating sections were rinsed in TBS, and sections were incubated for 30 min in 2 N HCl at 37◦C and rinsed in 0.1 M boric acid (pH 8.5). After several rinses in TBS, sections were incubated in TBS++ for 30 min, followed by a primary antibody cocktail containing rat anti-BrdU, goat anti-Sox2 and rabbit anti-S100b, for 24 h at 4◦C. Sections were then rinsed in TBS, incubated with a cocktail of fluorochrome-labeled secondary antibodies for 2 h (1:1,000 donkey anti-rat Alexa 488, donkey anti-goat Alexa 546 and donkey anti-rabbit Alexa 647, Invitrogen Corporation, Carlsbad, CA, USA), rinsed again in TBS, and mounted on glass slides.

### Stereological Quantification of Cells

In each animal, every 12th section (typically 6 sections) containing a dorsal hippocampus was used to determine the total number of BrdU- and Sox2-positive cells in the granule cell layer (GCL) under light microscopy. These numbers were counted using stereology software (StereoInvestigator, version 6; MBF Bioscience, Williston, VT, USA). Cell counts were then multiplied with the series factor (12) to determine the total number of cells per GCL. In triple immunofluorescence, the percentage of each positive cell type (≥50 BrdU- or Sox2 -positive cells per animal) was assessed using a confocal microscope (Leica TCS SP2; Leica Microsystems, Wetzlar, Germany). Resulting percentages of each positive cell type were multiplied with the absolute number of BrdU- or Sox2-positive cells to calculate the absolute number of each cell type (23). GCL volume in P173 mice was calculated by measuring the area in every 12th section throughout the hippocampus, and the total sum of the areas traced for dorsal and ventral blades separately was multiplied by section thickness and series number to produce the entire GCL volume.

#### Behavioral Tests IntelliCage®

IntelliCage <sup>R</sup> is a novel system for unbiased automated monitoring of spontaneous and learning behavior of mice in their home cage environment. Each cage had four corners, and each corner contained two water bottles. A door was located in front of each water bottle, and the door opened when a mouse touched it with its nose. Each cage housed maximum of 11 mice. Mice from the same litters were kept together in the same IntelliCage. Cage 1: 4 NSPC-treated, 3 vehicle-treated and 4 non-irradiated mice. Cage 2: 5 NSPC-treated, 3 vehicle-treated and 2 nonirradiated mice. Cage 3: 2 NSPC-treated, 2 vehicle-treated and 2 non-irradiated mice. Cage 4: 1 NSPC-treated, 1 vehicle-treated and 2 non-irradiated mice. Microchips (DataMars, Bedano, Switzerland) for identification were implanted s.c. in the dorsal neck of the mice on P54 under isoflurane anesthesia.

This behavioral test consisted of five sections (**Supplementary Figure 2**). The objective of the first two sections was for the mice to learn how to drink water. In the first section, "Introduction 1," every mouse was permitted to drink in all corners, and cage doors were kept open to aid the mice in locating the water bottles. In the second section, "Introduction 2," the doors were closed, and the mice, therefore, had to touch them with their noses ("nosepoke") to open them and drink water. In the third section, "Corner training 1," one corner was randomly allocated to each mouse. The mouse could open the door only in its allocated corner. The number of times a mouse opened the door in its allocated corner and/or tried to open it in the other corners were automatically recorded. In the next section, "Corner training 2," the corner allocated in the previous section was changed randomly. Each mouse could access only its own allocated corner. In the final section, "Corner training 3," the allocated corner was changed again.

#### Movement Pattern Analysis

The method and variables used for movement pattern analysis have been described in detail elsewhere (24, 25). Briefly, on the day of the experiment, mice were individually introduced into an unfamiliar open field arena and videotaped for 10 min in their inactive phase (lights on). Four arenas were simultaneously videotaped from above with a single CCD Monochrome video camera. The camera was connected to an S-VHS videocassette recorder. The arenas were made of black Plexiglas (l × w × h = 46 × 33 × 35 cm, respectively) rubbed with sandpaper and indirectly illuminated by bright light to avoid reflexes and shadows. The arena floors were covered with light gray gravel that had been previously exposed to other mice. The gravel was used to contrast the black color of the mice, and to act as an absorbent. It was exposed to other mice to saturate it with smells from multiple other individuals in order not to distract them to or from certain areas of the arena more than other areas. After completion of the experiment, videotapes were analyzed with the video tracking program EthoVision 3.1 (Noldus Information Technologies, Wageningen, Netherlands). For each sampling occasion, the program provided data on the position of the mouse as well as the animal's body area viewed by the overhead camera. The analysis resulted in a track record which described the animals' behavioral pattern during the observation period. Video tracking was performed at a sampling frequency of 12.5 Hz. Variables were calculated in a middle zone, defined as the central part located ≥6 cm from the arena walls, as well as in a border zone (<6 cm from the arena walls). We evaluated the following variables: distance moved, number of stops made, and percent time spent in motion.

#### Sugar Water (Anhedonia) Test

This test spanned 4 days. Every mouse was housed individually, and normal water (Day 1 and 2) or 3% fructose water (Days 3 and 4) was supplied from the starting time of the night cycle after 6 h of water restriction. We measured the volume (weight) of water/sugar water each mouse drank per night. We then calculated the ratio of consumption (Day 3 + 4/Day 1 + 2) to evaluate preference for the fructose water. A decrease in fructose intake and preference over water is generally taken as a putative sign of anhedonia in rodents (26).

#### Statistical Analysis

All animal-related data are presented as mean ± standard error. ANOVA followed by Dunnett's post-hoc test was used to compare the three groups. A p-value of < 0.05 was considered to indicate a significant difference between compared mean values.

### RESULTS

#### Mortality and Body Weight Gain

Thirty-six mice were used in the present study, but three mice died during anesthesia/irradiation, and 3 mice were excluded from analyses due to grafting failure. 1 NSPCtreated mouse died during the insertion of a microchip during anesthesia, and three vehicle-treated mice died during the observation period. Therefore, the mortality rates of each group after IR were 1/10 in NSPC-treated mice, 3/10 in vehicle-treated mice, and 1/10 in vehicle-treated non-irradiated mice.

The body weight gains from the point of grafting (P21) to sacrifice (P173) was calculated. The gain was significantly lower in irradiated vehicle-treated mice than in non-irradiated mice (**Figure 1**, p < 0.01), whereas irradiated, NSPC-grafted mice were not different from controls.

### Grafting NSPCs Ameliorated the Negative Effects of IR on Behavior

To examine the effects of grafting on behavior after IR, a single dose of 8 Gy cranial IR was administered at P9, and NSPCs were grafted onto both hippocampi at P21. Beginning 3 months after grafting, we examined the mice's behavior using three tests: IntelliCage <sup>R</sup> , movement pattern analysis (Open Field test), and sugar water preference (anhedonia test).

IntelliCage <sup>R</sup> is a system for unbiased automated monitoring of spontaneous activity and learning of mice in their home cage environment (27–34). In the first session, we measured the time it took for each mouse to visit a corner, representing a combined measure of novelty-induced exploratory behavior and anxiety. Irradiated, vehicle-injected mice took four times longer to visit the first corner compared with non-irradiated mice (**Figure 2A**, p < 0.05), whereas irradiated, NSPC-grafted mice took an intermediate length of time (**Figure 2A**). The number of times that each mouse tried to open the door in non-allocated corners (errors) was examined. The number of errors in the second corner session (reversal learning) was significantly higher for irradiated, vehicle-injected mice than for irradiated, grafted mice (**Figure 2B**; Days 1, 2, and 4, p < 0.05) and non-irradiated mice (**Figure 2B**; Day 2, p < 0.01; Days 1, 3, and 4, p < 0.05). The accumulated number of errors during the second corner session

FIGURE 2 | Learning deficit after irradiation (IR) (behavioral evaluation with Intellicage®). (A) First visit in the first section (Introduction 1). The time it took for irradiated mice injected with vehicle (IR-vehicle), but not with neural stem and progenitor cells (NSPCs) (IR-NSPC), to visit a corner was significantly longer than that for non-irradiated mice (non-IR-vehicle; \*p < 0.05). (B) Number of errors (second corner). The number of times that each mouse tried to open the door in non-allocated corners (i.e., errors) was significantly higher for irradiated (IR-vehicle) than for irradiated and grafted (IR-NSPC) and non-irradiated (non IR-vehicle) mice (\*p < 0.05, \*\*p < 0.01). (C) Accumulated number of errors (second corner). The accumulated number of errors that each mouse made (trying to open a non-allocated door) in the IR-vehicle group was higher than that in the non IR-vehicle group, but that in the IR-NSPC group was almost identical to that in the non-IR-vehicle group. Open circle indicates the non-IR-vehicle group; closed square indicates the IR-vehicle group; and closed diamond indicates the IR-NSPC group. (D) Accumulated number of correct nosepokes (second corner). The accumulated number of correct nosepokes was almost identical between the three groups. Open circle indicates the non-IR-vehicle group; closed square indicates the IR-vehicle group; and closed diamond indicates the IR-NSPC group. Data represent the mean ± S.E.M.

(reversal learning), as registered when a mouse tried to open a door in a non-allocated corner, was higher for the IR + vehicle group than the other two groups (**Figure 2C**). Irradiated, vehicleinjected mice made several-fold more errors than non-irradiated or irradiated NSPC-grafted mice. Importantly, the number of correct door openings did not differ between groups (**Figure 2D**). Both this IR-induced learning deficit and the ameliorating effect of NSPC grafting were also apparent in the first and third corners (**Supplementary Figures 3A,B**). These results indicate that IR significantly impaired place learning and reversal learning, and grafting of NSPCs significantly ameliorated these deficits.

The effect of IR and NSPC grafting on movement pattern in the Open Field test was analyzed with respect to the distance moved, number of stops, and percent time in motion. IR significantly decreased the distance moved (**Figure 3A**, p < 0.05) and the number of stops made (**Figure 3B**, p < 0.05), and tended to decrease the time in motion (**Figure 3C**) in the middle of the arena. On the other hand, there were no significant differences in any variables in the border zone (**Figures 3D–F**). The irradiated mice appeared more "anxious" and avoided the middle of the arena, but this was normalized by the grafting of NSPCs. (**Figures 3A–C**, p < 0.05 vs. IR group).

We also applied the sugar water preference or anhedonia test. Irradiated mice treated with vehicle, but not NSPC-grafted mice, tended to consume less sugar water than did non-irradiated mice (**Figure 4**).

### The Fate of Grafted Cells and the Effect of Irradiation and/or NSPC Grafting on Morphological Change and Endogenous Neural Stem Cells

We evaluated the number and fate of the grafted cells, the GCL size after grafting, and the effect of grafted NSPCs on endogenous neural stem cells. Five months after IR, 7,222 ± 455 BrdU-positive grafted cells per brain survived in the GCL; i.e., 3.6% of the total number of cells injected. **Figure 5A** shows a representative image of the hippocampus 5 months after grafting.

IR induces apoptosis of endogenous NSPCs and thereby arrests subsequent growth of the GCL in irradiated developing and juvenile brains (7). NSPC grafting did not increase the GCL volume in irradiated brains, in the dorsal or ventral blades (**Figure 5B**). The GCL consists mainly of tightly packed granule neurons, and the grafted cells hence did not seem to replace the granule cells that failed to develop after IR.

We estimated how many grafted NSPCs remained undifferentiated, using Sox2 and S100β staining, assuming that Sox2+/S100β− cells represent undifferentiated neural stem cells (double positive cells were astrocytes, whereas double negative cells could be neurons or oligodendrocytes) (35). In the ventral blade of the GCL, 14% of the grafted cells were Sox2+/S100β−, whereas 58% were negative for both Sox2 and S100β. In the dorsal blade, however, 26% of grafted cells were Sox2+/S100β−, and only 25% were negative for both Sox2 and S100β (**Figure 5C**). The total number of undifferentiated (Sox2+/S100β−) cells were counted in each GCL; i.e., both grafted and endogenous undifferentiated NSPCs (**Figure 5D**). IR reduced the number of undifferentiated NSPCs to almost half that observed in non-IR hippocampi in both the dorsal and ventral blades (**Figure 5D**, p < 0.01). Interestingly, NSPC grafting restored the number of undifferentiated NSPCs to control levels but only in the dorsal blade (**Figure 5D**). In the ventral blade, grafting did not affect the number of undifferentiated neural stem cells (**Figure 5D**). The normalized number of undifferentiated cells in the dorsal blade could be the result of grafted cells populating the subgranular zone (SGZ), grafted cells stimulating the proliferation and/or survival of endogenous neural stem cells, or a combination of the two. To differentiate between endogenous and grafted undifferentiated neural stem cells, we counted the number of BrdU-negative, undifferentiated (BrdU–/Sox2+/S100β−) cells; i.e., the number of endogenous, undifferentiated cells (**Figure 5D**). The number of endogenous, undifferentiated cells in irradiated, grafted mice was virtually identical to that of irradiated, non-grafted mice, although there was a tendency toward higher numbers in the dorsal blade of grafted mice (**Figure 5D**). Therefore, the normalized number of NSPCs in the dorsal blade of irradiated, grafted mice seems to result mainly from the grafted cells populating the SGZ and remaining undifferentiated.

#### DISCUSSION

In the present study, we have demonstrated that grafted NSPCs can survive in the GCL for at least 5 months, that IR caused behavioral abnormalities, including place learning deficits, and that NSPC grafting into the hippocampus could ameliorate or normalize the behavioral abnormalities induced by IR.

In a previous study, we showed that the survival of grafted NSPCs was significantly impaired and that neuronal differentiation was lower in irradiated than in non-irradiated brains when grafting was performed 24 h after IR (53% in irradiated brains vs. 84% in non-irradiated brains), whereas there were no significant differences when grafting occurred 1 week (64 vs. 66%) or 6 weeks (29 vs. 38%) after IR (18). These findings indicate that we should not graft NSPCs in this acute phase, soon after IR. Grafting cells before IR was deemed not useful, considering that proliferating neural stem cells are very susceptible to IR (7, 11, 19, 36), and a large portion of the grafted NSPCs would be killed by IR and handling the dead and injured cells would add to the burden of the tissue. Based on these findings, we grafted the NSPCs after the acute phase and observed an ameliorating effect of grafting on CNS complications after IR. However, it remains unknown whether grafting before or in the acute phase after IR also would exert such an effect.

Since proliferating neural stem cells are highly susceptible to IR and since a reduced number of proliferating NSPCs in the GCL of the dentate gyrus of the hippocampus can lead to learning/behavioral abnormalities later in life, we hypothesized that exogenous NSPCs could ameliorate the side effects of IR therapy by compensating for the lost cells. In addition, the younger the brain, the higher the number of differentiated neuronal cells after grafting (18). NSPC grafting may therefore be more effective in developing brains than in adult brains.

decrease in irradiated, vehicle-injected and grafted mice compared with non-IR mice (\*\*p < 0.01). Volume of grafted mice was almost identical to that of vehicle-injected mice. (C) Phenotype of surviving cells. We evaluated the phenotype of surviving cells by staining for BrdU, Sox2, and S100β. Only 14% of BrdU-positive cells in the lower blade and 25% in the upper blade of the GCL were positive for Sox2 and negative for S100 β. (D) Total number of stem cells. We evaluated the number of stem cells by staining for Sox2 (positive) and S100β (negative). Total number of stem cells of irradiated vehicle-injected mice was significantly lower than that of non-irradiated mice. Grafted mice had almost the same number of stem cells in the upper blade as non-irradiated mice, but the number in the lower blade was reduced to that of irradiated vehicle-injected mice (\*\*P < 0.01, \*P < 0.05). We also evaluated the number of endogenous stem cells by staining for BrdU (negative), Sox2 (positive), and S100β (negative). Number of grafted mice was almost the same as that of irradiated vehicle-injected mice. Volume measurement and positive cell counting were conducted as described in the Materials and Methods section. Data represent the mean ± S.E.M.

In the present study, we showed that the NSPC grafting into the hippocampi of developing brains could ameliorate learning deficits and behavioral abnormalities. It is known that NSPCs secrete various neurotrophic factors and exert neuroprotective effects (37, 38). In several animal models of stem cell grafting for neurodegenerative diseases, including stroke, both morphological and functional recovery have been observed, even when only a few grafted cells survived. It was reported that grafted cells did not replace degenerated neuronal cells but provided trophic factors and/or suppressed the immune/inflammatory response that induces angiogenesis, neurogenesis, and/or neuroprotection (39). Grafted NSPCs or mesenchymal stem cells enhance proliferation of endogenous cells and neurogenesis in neurogenic regions including the hippocampus and SVZ (40). To elucidate the mechanism of the ameliorating effect of NSPC grafting after IR in the present study, we evaluated the trophic effect of the grafted NSPCs. Specifically, these effects were the volume of the GCL of the hippocampus, which is reduced after IR due to reduced growth (19, 31), and the number of endogenous neural stem cells, based on the number of Sox2-positive and S100β-negative cells and BrdU-negative cells. However, no trophic effects were observed in the present study, as judged by the virtually identical size of the GCL and number of endogenous neural stem cells in irradiated grafted mice were compared to those in irradiated vehicle-injected mice. On the other hand, in our previous study, 5 weeks after grafting at P21, more than 60% of surviving cells were neurons (18). The present study also showed that at 5 months after grafting, more than 50% of surviving cells in the ventral blade were negative for both Sox2 and S100β, which indicated that they adopted a neuronal or oligodendrocytic phenotype, although it was not confirmed with a neuronal marker like NeuN. Although the functionality of the surviving neurons differentiated from grafted NSPCs remains to be evaluated, it is possible that grafted cells replaced, to some extent, the IRinduced loss of NSPCs. This in turn may have led to amelioration of abnormalities after IR. Another possible explanation could be continuous production of growth and trophic factors by NSPCs, which over time leads to changes in connectivity and other functional aspects not related to the actual numbers of cells or their phenotypes.

Toward clinical applications of NSPC grafting for IR-induced learning and behavioral deficits, several issues remain to be addressed. One of these is grafting-induced astrogliosis. We previously showed that injection of NSPCs, but not the vehicle, into the hippocampus induced astrogliosis and reduced thickness of the dorsal blade of the GCL (18). Cell grafting can induce and promote glial scarring (17, 41) and upregulation of expression of chondroitin sulfate proteoglycans around the injection site (42). These cause impaired migration and neuritogenesis of grafted cells, leading to abortive host-graft integration. In a clinical application of NSPC grafting, some modification [e.g., inhibition of reactive astrocytes (43), enzymatic removal of chondroitin sulfate (44)], may be needed to circumvent this problem and enhance the therapeutic effect.

In the present study, we administered NSPCs directly into the hippocampus. In other studies of neuronal diseases, intravenous administration has also been used (45–47). Systemic administration may be advantageous, especially for diseases with multifocal and/or disseminated lesions such as multiple sclerosis. Although NSPCs administrated intravenously can migrate into lesions to some degree, the number of cells is limited, up to 13% (45), and NSPCs accumulate in organs without lesions (47). More recently, an intranasal route for administration of NSPCs has been developed (48–50). It was shown that intranasally administered stem cells can migrate into the brain. These alternative administration routes are less invasive than intracranial grafting but also less effective, and for clinical applications a balance between safety and efficacy must be established.

It may be argued that using postnatal day 9 mice corresponds to a very young human brain, developmentally (51), an age when cranial radiotherapy wouldn't be considered in a clinical setting, but the numbers of endogenous NSPCs and the rate of neurogenesis are higher at this relatively early age, so for the purpose of studying the effects of IR and NSPC grafting, it is useful for proof of principle, as demonstrated in earlier studies in both rats and mice (8, 11, 19). Another reason is that the survival of grafted cells is better in a younger brain, as noted in our previous study (18). Further studies are needed to show the treatment effect of grafting NSPCs with older mice.

In conclusion, our results indicate that NSPCs grafting into the hippocampus can ameliorate or normalize the impaired place learning as well as the altered movement pattern induced by IR. Although ethical, technical and other issues remain to be addressed prior to its clinical application, NSPC grafting is promising as a therapeutic technique. Further studies are needed to optimize the treatment and clarify the underlying mechanisms.

## AUTHOR CONTRIBUTIONS

YS, MS, and KO were actively involved in experiments. NS cultured NSPCs for experiments. MN and YS performed open field test. YS, CZ, HK, MP, and KB conceptualized and designed this study. YS, CZ, HK, and KB interpreted data in this study. YS drafted an initial manuscript, and NS, MS, KO, CZ, MP, HK, and KB revised it critically. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

### ACKNOWLEDGMENTS

We thank Dr. Niklas Karlsson for help with the use and interpretation of Intellicage <sup>R</sup> experiments, and Mrs. Rita Grandér for excellent technical assistance. This work was supported by the Swedish Childhood Cancer Foundation (Barncancerfonden), the Swedish Research Council (Vetenskapsrådet), the Swedish Cancer Foundatio (Cancerfonden), grants provided by the Stockholm County Council and the Västra Götaland Region (ALF projects), the Frimurare Barnhus Foundations of Gothenburg and Stockholm, the Märta and Gunnar V. Philipson Foundation, the Swedish Brain Foundation (Hjärnfonden), the Swedish Radiation Safety Authority, the Aina Wallström's and Mary-Ann Sjöblom's Foundation, and the Ulla and Rune Amlöv Foundations. The funding agencies had no influence on the study design.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur. 2018.00715/full#supplementary-material

# REFERENCES


causes gliosis and disrupts the granule cell layer. Cell Death Dis. (2013) 4:e591. doi: 10.1038/cddis.2013.92


model of amyotrophic lateral sclerosis. Cell Transplant. (2010) 19:537–48. doi: 10.3727/096368910X498269


**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 Sato, Shinjyo, Sato, Nilsson, Osato, Zhu, Pekna, Kuhn and Blomgren. 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.

# Carbamylated Erythropoietin Decreased Proliferation and Neurogenesis in the Subventricular Zone, but Not the Dentate Gyrus, After Irradiation to the Developing Rat Brain

Kazuhiro Osato1,2†, Yoshiaki Sato1,3†, Akari Osato1,2, Machiko Sato1,4, Changlian Zhu1,5 , Marcel Leist <sup>6</sup> , Hans G. Kuhn<sup>1</sup> and Klas Blomgren1,7 \*

#### Edited by:

Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland

#### Reviewed by:

Eduardo Farias Sanches, Universidade Federal do Rio Grande do Sul (UFRGS), Brazil Christof Dame, Charité Universitätsmedizin Berlin, Germany

#### \*Correspondence:

Klas Blomgren klas.blomgren@ki.se

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology

Received: 19 December 2017 Accepted: 13 August 2018 Published: 12 September 2018

#### Citation:

Osato K, Sato Y, Osato A, Sato M, Zhu C, Leist M, Kuhn HG and Blomgren K (2018) Carbamylated Erythropoietin Decreased Proliferation and Neurogenesis in the Subventricular Zone, but Not the Dentate Gyrus, After Irradiation to the Developing Rat Brain. Front. Neurol. 9:738. doi: 10.3389/fneur.2018.00738 <sup>1</sup> Center for Brain Repair and Rehabilitation, Institute of Neuroscience and Physiology, Gothenburg University, Gothenburg, Sweden, <sup>2</sup> Department of Obstetrics and Gynecology, Mie University, Tsu, Japan, <sup>3</sup> Division of Neonatology, Center for Maternal-Neonatal Care, Nagoya University Hospital, Nagoya, Japan, <sup>4</sup> Department of Obstetrics and Gynecology, Narita Hospital, Nagoya, Japan, <sup>5</sup> Department of Pediatrics, The Third Affiliated Hospital of Zhengzhou University, Zhengzhou, China, <sup>6</sup> Department of Biology, University of Konstanz, Konstanz, Germany, <sup>7</sup> Department of Women's and Children's Health, Karolinska Institutet, Department of Pediatric Hematology and Oncology, Karolinska University Hospital, Stockholm, Sweden

Cranial radiotherapy for pediatric brain tumors causes progressive, debilitating late effects, including cognitive decline. Erythropoietin (EPO) has been shown to be neuroprotective and to promote neuroregeneration. Carbamylated erythropoietin (CEPO) retains the protective properties of EPO but is not erythrogenic. To study the effects of CEPO on the developing brain exposed to radiotherapy, a single irradiation (IR) dose of 6 Gy was administered to the brains of postnatal day 9 (P9) rats, and CEPO (40 µg/kg s.c.) was injected on P8, P9, P11, P13, and P15. To examine proliferation, 5- Bromo-2-deoxyuridine (BrdU) was injected on P15, P16, and P17. CEPO administration did not affect BrdU incorporation in the granule cell layer (GCL) of the hippocampus or in the subventricular zone (SVZ) as quantified 7 days after the last BrdU injection, whereas IR decreased BrdU incorporation in the GCL and SVZ by 63% and 18%, respectively. CEPO did not affect BrdU incorporation in the GCL of irradiated brains, although it was reduced even further (to 31%) in the SVZ. To evaluate the effect of CEPO on neurogenesis, BrdU/doublecortin double-positive cells were quantified. CEPO did not affect neurogenesis in non-irradiated brains, whereas IR decreased neurogenesis by 58% in the dentate gyrus (DG) but did not affect it in the SVZ. In the DG, CEPO did not affect the rate of neurogenesis following IR, whereas in the SVZ, the rate decreased by 30% following IR compared with the rate in vehicle-treated rats. Neither CEPO nor IR changed the number of microglia. In summary, CEPO did not promote neurogenesis in non-irradiated or irradiated rat brains and even aggravated the decreased neurogenesis in the SVZ. This raises concerns regarding the use of EPO-related compounds following radiotherapy.

Keywords: radiotherapy, pediatric oncology, late effects, immature brain, neural stem cell

### INTRODUCTION

Nearly one-third of all pediatric malignancies are brain tumors, and their incidence has increased over the last decades (1– 3). Improved treatment protocols have considerably increased the survival of patients with such malignancies, with >80% of patients currently surviving their disease (4). Treatment strategies for pediatric malignancies are associated with late adverse effects such as perturbed growth, endocrine dysfunctions, learning difficulties, and cognitive decline (5). Children receiving radiotherapy to the CNS are at the greatest risk of cognitive decline (6), which increases with younger age at diagnosis (7, 8). Thus, ameliorating the late effects of CNS radiotherapy would greatly improve the quality of life of the increasing numbers of childhood cancer survivors. However, effective preventions, protective treatments, and rehabilitation strategies are currently unavailable in clinical practice and experimental research.

Therapeutic doses of ionizing irradiation (IR) result in increased apoptosis (9, 10) and decreased cell proliferation in neurogenic regions (11). Both the subventricular zone (SVZ) and the subgranular zone (SGZ) contain proliferating neural stem and progenitor cells and undergo neurogenesis throughout life (12–14). Thus, these regions are particularly susceptible to IR-induced apoptosis (15), and animal studies of hippocampal function have supported the hypothesis that this decrease in neurogenesis contributes to cognitive deficits experienced by patients after cranial radiation therapy (15).

Erythropoietin (EPO) has been shown to have neuroprotective properties in several animal models (16), including those of spinal cord injury (17), adult focal ischemia (18), neonatal hypoxia-ischemia (19), neonatal white matter injury (20, 21), traumatic brain injury (22), chronic autoimmune encephalomyelitis (23), and amyotrophic lateral sclerosis (24) as well as neonatal brain injury (25, 26). In addition to its neuroprotective effects, EPO affects other tissues, including the heart (27), kidneys (28), intestine (29), liver (30), and skin (31). EPO can exert protective effects via several different mechanisms, such as the attenuation of apoptosis (32), excitotoxicity (33), oxidative stress (29), and inflammation (34), as well as through angiogenesis (35). We have previously failed to demonstrate the neuroprotective effects of EPO or caspase inhibition through XIAP overexpression in neonatal mice exposed to IR (36). However, apoptosis inhibition through lithium treatment was found to reduce the morphological and behavioral damage caused by IR to the young mouse brain (9). Carbamylated erythropoietin (CEPO), in which all lysine residues are transformed to homocitrulline via carbamylation, does not bind to the classical EPO receptor but retains the tissue-protective properties without affecting erythrogenesis and hematocrit (37). CEPO binds to the β common receptor, which may be involved in its neuroprotective activity (38). Protective effects of CEPO have been shown to resemble those of EPO in various models (21, 39, 40). In addition, CEPO has been demonstrated to enhance proliferation in vitro (41) and in vivo (42) and to promote neurite outgrowth and neuronal spine formation (43).

In the present study, we tested the hypothesis that CEPO enhances proliferation in the neurogenic regions, thereby enhancing recovery after IR.

#### MATERIALS AND METHODS

#### Animals

All animal experimental protocols were approved by the Gothenburg Animal Ethics Committee of the Swedish Board of Agriculture (46-2007 and 326-09). Wistar rats were purchased from B & K Universal (Solna, Sweden).

#### Irradiation Procedure

IR was performed as previously described (44, 45). Briefly, a linear accelerator (Varian Clinac 600CD) with a 4 MV nominal photon energy and a dose rate of 2.3 Gy/min was used. Male rats (9-day-old) were anesthetized with the intraperitoneal administration of tribromoethanol (Sigma-Aldrich, Stockholm, Sweden) and placed in the prone position (head to gantry) on an expanded polystyrene bed. The left cerebral hemisphere of each rat was irradiated with an asymmetrical radiation field of 1 × 2 cm without divergence toward the right hemisphere. The distance from the source to the skin was approximately 99.5 cm. The head was covered with a tissue equivalent bolus material (1 cm). A single absorbed dose of 6 Gy was administered to each rat. The variation of dose within the target volume was estimated to be ±5%. The entire procedure was performed within 10 min. After IR, rats were returned to their biological dams until they were sacrificed. Sham control rats were anesthetized but did not receive IR. Using the LQ-model (46) and an α/β-ratio of 3 for late effects in the normal brain tissue, an acute exposure of 6 Gy is equivalent to approximately 12 Gy when delivered in daily 2 Gy fractions, which represents a clinically relevant dose, equivalent to that used in prophylactic cranial IR in children with acute lymphatic leukemia.

#### Carbamylated Erythropoietin Treatment and 5-Bromo-2-Deoxyuridine Injection

CEPO (40 µg/kg), prepared and characterized as previously described in detail (37), was subcutaneously administered on P8, P9, P11, P13, and P15. 5-bromo-2-deoxyuridine (BrdU) (1 mg/mL, dissolved in PBS; Roche, Mannheim, Germany; 50 mg/kg) was intraperitoneally injected on P15, P16, and P17. Rats were sacrificed on P24, and their brains were processed for immunohistochemistry. A schematic diagram of the experimental procedures is shown in **Figure 1**.

#### Tissue Preparation

On P24, rats were deeply anesthetized and subjected to intracardial perfusion with 0.9% NaCl followed by 5% buffered formaldehyde (Histofix, Histolab, Gothenburg, Sweden). Brains were removed and immersion-fixed in the same solution overnight at 4◦C and then immersed in 30% sucrose for at least 2 days. Brains were then coronally cut at 40µm on a

**Abbreviations:** EPO, erythropoietin; CEPO, carbamylated erythropoietin; IR, irradiation; BrdU, 5-bromo-2-deoxyuridine; GCL, granule cell layer; SVZ, subventricular zone; dcx, doublecortin; DG, dentate gyrus; SGZ, subgranular zone; ML, molecular layer.

sliding microtome in dry ice. Sections were stored at −20◦C in a cryoprotectant solution (50% glycerol/25% glycol/25% 0.1 M phosphate buffer).

#### Immunohistochemistry

Following antibodies and final dilutions were used: mouse anti-BrdU (1:500; 11170376, Roche, Mannheim, Germany), goat anti-doublecortin (DCX; 1:500; sc-8066, Santa Cruz Biotechnology, Santa Cruz, CA, USA), rabbit anti-Iba1 (1:1,000; 019-19741, WAKO Pure Chemical Industries, ltd., Osaka, Japan), donkey anti-mouse-Alexa488 (1:1,000; A21202, Invitrogen Corporation, Carlsbad, CA, USA), and donkey anti-goat-Alexa555 (1:1,000; A-21432, Invitrogen Corporation).

Immunoperoxidase detection of BrdU was performed as previously described (45). Briefly, sections were rinsed in Trisbuffered saline [TBS; 0.1 M Tris-HCl (pH 7.4) in 0.9% NaCl], treated with 0.6% H2O<sup>2</sup> in TBS for 30 min, incubated with 50% formamide/2 × SSC (0.3 M NaCl, 0.03 M sodium citrate) at 65◦C for 2 h, rinsed with 2 × SSC for 5 min, incubated with 2 N HCl at 37◦C for 30 min, and rinsed with 0.1 M boric acid (pH 8.5) for 10 min. Incubation with 3% donkey serum and 0.1% Triton X-100 in TBS (TBS++) for 30 min was followed by overnight incubation with mouse anti-BrdU. After several rinses in TBS 2 × 10 min, sections were incubated with donkey anti-mouse-biotin (1:1,000; Vector Laboratories, Burlingame, CA, USA) for 1 h and then in avidin–biotin–peroxidase complex (Vectastain ABC Elite kit, Vector Laboratories), followed by peroxidase detection for 5 min (0.25 mg/ml DAB, 0.01% H2O2, 0.04% NiCl). For Iba-1 staining, sections were treated with 0.6% H2O<sup>2</sup> in TBS for 30 min. After rinsing, sections were blocked with TBS++ and incubated with rabbit anti-Iba1 at +4 ◦C for 24 h in TBS++, followed by 1 h at room temperature with donkey anti-rabbit-biotin (1:1,000, Vector Laboratories).

Double immunofluorescence for BrdU and DCX was performed as follows: sections were incubated with 2 N HCl at 37◦C for 30 min and rinsed with 0.1 M boric acid (pH 8.5) for 10 min. After several rinses with TBS, sections were incubated with TBS++ for 30 min, followed by a primary antibody cocktail, including mouse anti-BrdU and goat antidoublecortin, in TBS++ at +4 ◦C for 24 h. The sections were then rinsed with TBS 3 × 10 min, incubated with a cocktail of fluorochrome-labeled secondary antibodies for 2 h, rinsed again with TBS, and mounted on glass slides.

## Cell Counting

BrdU-positive cells were counted throughout the SVZ and SGZ of the dentate gyrus (DG) of the hippocampus, and Iba1-positive cells were counted throughout the SVZ and granule cell layer (GCL), hilus, and molecular layer (ML). The numbers were expressed as densities, divided by the area of the SVZ, GCL, hilus, or ML or the length of the SGZ. All counts and measurements were performed using Stereo Investigator ver. 6 (MBF Bioscience, Williston, VT, USA).

#### Statistics

Results are presented as mean ± S.E.M. Statistical analysis was performed using one-way ANOVA followed by Fisher's post-hoc test using SPSS 17.0 (SPSS Inc., Chicago, IL, USA). p < 0.05 was considered significant.

## RESULTS

### Impact of CEPO on Proliferation After IR

We first examined the effects of IR and/or CEPO on proliferation in the SGZ of the DG and the SVZ. Rats were irradiated on P9 and injected with CEPO 1 day before IR; immediately before IR; and then 2, 4, and 6 days after IR (i.e., on P8, P9, P11, P13, and P15, respectively). To examine proliferation, BrdU was injected 6, 7, and 8 days after IR (i.e., on P15, P16, and P17, respectively), and rats were sacrificed on P24 (**Figure 1**). Representative photomicrographs of the SVZ and DG in both vehicle- and CEPO-injected rats with or without 6 Gy IR of the ipsilateral hemisphere stained for BrdU are shown in **Figures 2A**, **3A**, respectively. The number of BrdU-positive cells in the SGZ of control rats (vehicle treatment) decreased by >50% after IR (**Figure 2B**; p < 0.01). In CEPO-treated rats, the number of BrdU-positive cells in the SGZ decreased to the same degree after IR (**Figure 2B**; p < 0.01). The number of BrdU-positive cells

remained unaffected both in the non-irradiated and irradiated rats.

In vehicle-treated animals, the number of BrdU-positive cells in the SVZ did not decrease after IR (**Figure 3B**), whereas in CEPO-treated rats, it significantly reduced by 31% after IR (**Figure 3B**; p < 0.01). CEPO did not alter the number of BrdUpositive cells in non-irradiated rats. These results indicate that CEPO decreases proliferation in the SVZ after IR.

#### Impact of CEPO on Neurogenesis After IR

Next, we examined the effect of CEPO on post-IR neurogenesis. Rats were irradiated and then treated with CEPO, and BrdU was injected as described above. To examine neurogenesis, sections were stained with antibodies to BrdU and DCX, and the BrdU/DCX double-positive cells were counted. **Figure 4A**

presents photomicrographs of the SVZ and DG in both vehicleand CEPO-treated rats with or without 6 Gy IR to the ipsilateral hemisphere stained for BrdU and DCX.

The number of BrdU/DCX double-positive cells in the SGZ of vehicle-treated rats significantly decreased after IR (58%, p < 0.01; **Figure 4B**). After CEPO treatment, the number of BrdU/DCX double-positive cells in the SGZ decreased to the same degree after IR (**Figure 4B**; p < 0.01). In both non-irradiated and irradiated rats, the number of BrdU/DCX double-positive cells was not significantly affected by CEPO.

In contrast, in vehicle-treated rats, the number of BrdU/DCX double-positive cells in the SVZ did not decrease (**Figure 4C**), whereas in CEPO-treated rats, there was a 36% decrease after IR (**Figure 4C**; p < 0.05). CEPO administration after IR reduced neurogenesis by 30% in the SVZ compared with that in vehicle-treated rats (**Figure 4C**), although the same effect was not observed in non-irradiated rats. These

FIGURE 4 | Neurogenesis in the subgranular zone (SGZ) and the subventricular zone (SVZ). (A) Representative microphotographs of the SGZ stained for BrdU (green) and DCX (red) 15 days after IR. Bar = 25µm. (B) The number of BrdU/DCX double-positive cells in the SGZ of vehicle-treated rats after IR (n = 8) was significantly lower than that of non-irradiated vehicle-treated rats (n = 7) (\*\*p < 0.01). The number of BrdU/DCX double-positive cells in the DG of CEPO-treated rats after IR (n = 6) was significantly lower than that of non-irradiated CEPO-treated rats (n = 9) (\*\*p < 0.05). CEPO did not alter the number of BrdU/DCX double-positive cells in the SGZ, in non-irradiated rats, or in irradiated rats. (C) There was no significant difference between the number of BrdU/DCX double-positive cells in the SVZ of vehicle-treated rats after IR (n = 10) and that without IR (n = 8). The number of BrdU/DCX double-positive cells in the SVZ of CEPO-treated rats after IR (n = 9) was significantly lower than that of non-irradiated CEPO-treated rats (n = 9) (\*p < 0.05). Positive cells were counted as described in the Materials and methods section. Results are presented as mean ± S.E.M.

results indicate that CEPO decreases neurogenesis in the SVZ after IR.

non-irradiated and irradiated brains were affected by CEPO (**Figures 5B-E**).

#### Impact of CEPO on Microglia After IR

Rats were irradiated and treated with CEPO as described above and the number of Iba1-positive cells in the SVZ and DG (GCL, hilus, and ML) was counted 2 weeks after IR. Representative photomicrographs of the SVZ and DG in both vehicle- and CEPO-treated rats with and without 6 Gy IR stained for Iba1 are shown in **Figure 5A**. The number of Iba1-positive cells was not changed after IR in the SVZ, GCL, hilus, or the ML (**Figures 5B–E**). No areas of both

#### DISCUSSION

In the present study, we examined the post-IR effects of CEPO on proliferation and neurogenesis in neurogenic regions and demonstrated that CEPO decreased neurogenesis in the SVZ.

The SVZ and DG undergo a decrease in proliferation after IR, both in adult (47, 48) and immature and juvenile (15) rats. In the present study, we demonstrated an IR-induced decrease in the DG, but not in the SVZ, which is in line

with our earlier findings in the rat brain where proliferation in the SVZ, but not in the DG, was transiently decreased and subsequently recovered to some extent (36, 49). In neurogenic regions, neurogenesis is decreased after IR both in adult (50) and juvenile and immature (36, 51) brains. Although there is an association between reduced hippocampal neurogenesis and cognitive impairment (15), functional consequences of reduced neurogenesis in the SVZ beyond olfaction are not well studied.

It has been shown that CEPO enhances proliferation in vitro (41) and in vivo (42) and exerts neuroprotective effects in various models (39). However, in the present study, we did not observe a protective effect on the proliferation rates of the hippocampal SGZ. Moreover, post-IR proliferation in the SVZ was unexpectedly decreased following CEPO treatment. The mechanism underlying this negative effect remains unclear. CEPO, which retains the neuroprotective properties of EPO without exerting an erythropoietic effect, has the potential to be an ideal neuroprotective compound, given its limited toxicity. CEPO has been shown to be protective in various animal models of brain injury (39). CEPO improves renal function and survival in acute kidney injury models without raising hematocrit levels and blood pressure as substantially as EPO (52). Ma et al. indicated that CEPO exerted antiapoptotic activity in myocardial cells independent of JAK2/STAT5 signaling, which was involved in the effect of EPO (53). Also in an IR model, Erbayraktar et al. showed that CEPO reduced brain injury in adult rats after highly localized IR using a gamma-knife (54). However, the dose used in that study was 100 Gy, which is very high and necrotizing, thereby generating a different type of injury than that observed after 6 Gy. In contrast, because CEPO appeared to aggravate the negative effects on proliferation and neurogenesis after a moderate dose of IR, our findings raise concerns regarding the potential use of EPO-related compounds, particularly CEPO, during and after radiotherapy.

Microglia are antigen-presenting scavenger cells that can engulf invading microorganisms, remove deleterious debris, secrete growth factors to promote tissue repair, and maintain tissue homeostasis. However, they can also be transformed into cytotoxic cells (10, 55). It has been shown that the number of microglia is increased in animal models of neurological disorders in both mature (34) and immature (56) brains. However, after IR to the juvenile brain, the number of microglia decreases. In P9 rats, we observed a decrease at 7 days after 8 Gy, which was preceded by an initial increase 6 h after IR (57). During normal brain development, microglia proliferation peaks at P9 (58) and it is conceivable that IR may be deleterious to proliferating microglia, as supported by our findings (57). In the present study, 6 Gy IR to P9 rat brains did not significantly alter the number of microglia in the SVZ or in the hippocampus. This discrepancy between our present and previous findings may be attributed to the lower dose used in the present study (6 vs. 8 Gy). CEPO has been shown to ameliorate the increase in microglia numbers after injury (18, 21, 34); however, we did not find any effect of CEPO on microglia numbers after IR.

There are some limitations of the present study. First, we have not quantified also the number of mature neurons (BrdU+/NeuN+ cells), only the immature neurons (BrdU+/DCX+ cells), so although it is unlikely that the decrease in the SVZ in CEPO-treated rats after IR did not also result in a decrease in mature neurons, this remains to shown unequivocally. Second, for future studies it would be valuable to investigate the dose-response effect by using different doses of CEPO on neurogenesis,

#### REFERENCES

since high concentrations of EPO under hypoxic, but not normoxic conditions invoked apoptosis (59). Finally, the underlying mechanisms are unclear and remain to be elucidated.

### CONCLUSION

EPO has been shown in other paradigms to enhance neurogenesis, but CEPO, which retains the protective properties of EPO but lacks its erythrogenic effects, did not stimulate neurogenesis in non-irradiated or irradiated rat brains. It even aggravated the negative effects of IR on neurogenesis, which raises concerns regarding the use of EPO-related compounds after radiotherapy.

### AUTHOR CONTRIBUTIONS

KO, YS, AO, and MS were actively involved in the experiments. KO, YS, and KB conceived and designed the present study. KO, YS, CZ, ML, HGK, and KB interpreted the data. YS drafted an initial manuscript, which was critically revised by KO, AO, MS, CZ, ML, HGK, and KB. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

### FUNDING

This work was supported by the Swedish Childhood Cancer Foundation (Barncancerfonden), the Swedish Research Council (Vetenskapsrådet), the Swedish Cancer Foundation (Cancerfonden), governmental grants from the agreement concerning research and education of doctors (ALF) in Stockholm and Gothenburg, the Sahlgrenska Academy at the University of Gothenburg, the Sten A. Olsson's Foundation, the King Gustav V Jubilee Clinic Research Foundation (JK-fonden), the Frimurare Barnhus Foundation of Gothenburg, the Wilhelm and Martina Lundgren Foundation, the Gothenburg Medical Society, the Sahlgrenska Foundations (SU-fonden), the Aina Wallström and Mary-Ann Sjöblom Foundation, and the Ulla and Rune Amlöv Foundations.

### ACKNOWLEDGMENTS

We are grateful for the skillful technical assistance of Rita Grandér. The funding agencies had no influence on the study design. Parts of the manuscript have previously been published as part of conference proceedings from the 34th Annual Meeting of the Japan Neuroscience Society https://www.sciencedirect. com/journal/neuroscience-research/vol/71/suppl/S (60) and appropriate permission has been obtained from the copyright holder for the reuse of this material.


<sup>1.</sup> Dreifaldt AC, Carlberg M, Hardell L. Increasing incidence rates of childhood malignant diseases in Sweden during the period 1960–1998. Eur J Cancer (2004) 40:1351–60. doi: 10.1016/j.ejca.2004.02.003

population-based 22-year retrospective study. J Neurol. (2012) 259:1131–6. doi: 10.1007/s00415-011-6314-4


in periventricular leukomalacia. Exp Neurol. (2011) 230:227–39. doi: 10.1016/j.expneurol.2011.04.021


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

Copyright © 2018 Osato, Sato, Osato, Sato, Zhu, Leist, Kuhn and Blomgren. 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.

#### *Edited by:*

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Steven W. Levison, Rutgers University, The State University of New Jersey, United States Maria Teresa Ferretti, University of Zurich, Switzerland*

#### *\*Correspondence:*

*Arshed Nazmi nazmia@tcd.ie; Xiaoyang Wang xiaoyang.wang@fysiologi.gu.se*

#### *†Present address:*

*Arshed Nazmi, School of Biochemistry & Immunology, Trinity College Dublin, Trinity Biomedical Sciences Institute, Dublin, Ireland* 

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

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 05 December 2017 Accepted: 02 March 2018 Published: 19 March 2018*

#### *Citation:*

*Nazmi A, Albertsson A-M, Rocha-Ferreira E, Zhang X, Vontell R, Zelco A, Rutherford M, Zhu C, Nilsson G, Mallard C, Hagberg H, Lai JCY, Leavenworth JW and Wang X (2018) Lymphocytes Contribute to the Pathophysiology of Neonatal Brain Injury. Front. Neurol. 9:159. doi: 10.3389/fneur.2018.00159*

*Arshed Nazmi1 \*†‡, Anna-Maj Albertsson1‡, Eridan Rocha-Ferreira2 , Xiaoli Zhang1,3,4, Regina Vontell <sup>5</sup> , Aura Zelco1 , Mary Rutherford5 , Changlian Zhu3,4,6, Gisela Nilsson1 , Carina Mallard1 , Henrik Hagberg1,2,5, Jacqueline C. Y. Lai1 , Jianmei W. Leavenworth7,8 and Xiaoyang Wang1,3\**

*1Department of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, 2 Department of Clinical Sciences, Sahlgrenska University Hospital, Gothenburg, Sweden, 3Department of Pediatrics, The Third Affiliated Hospital of Zhengzhou University, Zhengzhou, China, 4 Henan Key Laboratory of Child Brain Injury, The Third Affiliated Hospital of Zhengzhou University, Zhengzhou, China, 5Department of Perinatal Imaging and Health, Centre for the Developing Brain, King's College London, St. Thomas' Hospital, London, United Kingdom, 6Department of Neuroscience and Physiology, Center for Brain Repair and Rehabilitation, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, 7Department of Neurosurgery, University of Alabama at Birmingham, Birmingham, AL, United States, 8Department of Microbiology, University of Alabama at Birmingham, Birmingham, AL, United States*

Background: Periventricular leukomalacia (PVL) is the most common form of preterm brain injury affecting the cerebral white matter. This type of injury involves a multiphase process and is induced by many factors, including hypoxia–ischemia (HI) and infection. Previous studies have suggested that lymphocytes play a significant role in the pathogenesis of brain injury, and the aim of this study was to determine the contribution of lymphocyte subsets to preterm brain injury.

Methods: Immunohistochemistry on brain sections from neonatal mice was performed to evaluate the extent of brain injury in wild-type and T cell and B cell-deficient neonatal mice (*Rag1*−/− mice) using a mouse model of HI-induced preterm brain injury. Flow cytometry was performed to determine the presence of different types of immune cells in mouse brains following HI. In addition, immunostaining for CD3 T cells and CD20 B cells was performed on postmortem preterm human infant brains with PVL.

results: Mature lymphocyte-deficient *Rag1*−*/*− mice showed protection from white matter loss compared to wild type mice as indicated by myelin basic protein immunostaining of mouse brains. CD3+ T cells and CD20+ B cells were observed in the postmortem preterm infant brains with PVL. Flow cytometry analysis of mouse brains after HI-induced injury showed increased frequency of CD3+ T, αβT and B cells at 7 days after HI in the ipsilateral (injured) hemisphere compared to the contralateral (control, uninjured) hemisphere.

conclusion: Lymphocytes were found in the injured brain after injury in both mice and humans, and lack of mature lymphocytes protected neonatal mice from HI-induced brain white matter injury. This finding provides insight into the pathology of perinatal brain injury and suggests new avenues for the development of therapeutic strategies.

Keywords: lymphocytes, preterm, brain damage, mouse models, hypoxia–ischemia, brain

Brain injury in premature infants born at less than 30 weeks gestational age is a significant clinical problem (1–3). Many important maturation processes occur during the last half of gestation, including the development of premyelinating oligodendrocytes (pre-OLs), axons, and neurons (4). These events are complex and rapid, and they are, therefore, vulnerable to endogenous and exogenous insults such as inflammation, decreased blood flow (ischemia), decreased oxygen flow (hypoxia), and free radical activity (1).

Periventricular leukomalacia (PVL) is generally thought to be the most common form of injury to the preterm brain. PVL is a distinct form of cerebral white matter injury that is characterized by the loss of pre-OLs and is associated with a high risk of neurodevelopmental impairment (1, 2, 5). Cerebral ischemia, maternal infections, and fetal systemic inflammation are the primary factors that initiate PVL through excitotoxicity and the production of free radicals (6). Preterm hypoxia– ischemia (HI) is thought to be one of the leading causes of brain injury secondary to maternal infection (7), and infection of the chorioamniotic membrane with pathogenic bacteria (chorioamnionitis) is considered to be a life-threatening risk factor for preterm infants because it can directly cause brain injury and can make the fetal brain more vulnerable to insults such as hypoxia (8, 9).

The exact role of inflammation in neonatal brain injury is still not fully understood, although it is known that sterile inflammation is associated with the recruitment of peripheral immune cells to the brain. Early infiltrating cells after insult to the brain include polymorphonuclear leukocytes and monocytes, and lymphocytes can also enter in both the neonatal (10–14) and adult brain following ischemia. In the damaged brain, these cells release proinflammatory molecules such as those involved in type 1/type 17 immune responses (15) and anti-inflammatory cytokines that can either aggravate or repair injury (16, 17).

Understanding the dynamics of postischemic inflammation is a prerequisite for therapeutic intervention in this fragile system in order to prevent harmful side effects of such interventions in sick neonates. Here, we hypothesize that lymphocytes can gain access into the brain after HI-induced brain injury and contribute to the development of neonatal brain injury. The purpose of this study was (1) to determine the presence of lymphocytes in the neonatal mouse brain in our established mouse model of HI-induced preterm brain injury using postnatal day (PND) 5 neonatal mice (15) and (2) to evaluate the contribution of mature T and B cells to neonatal brain injury.

### MATERIALS AND METHODS

#### Animals

C57Bl/6J wild-type (WT) and recombination-activating gene 1 (*Rag1*) mutant mice (*Rag1*<sup>−</sup>*/*−; B6.129S7-Rag1tm1Mom/J) were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and were bred in the animal facility at the University of Gothenburg (Experimental Biomedicine, University of Gothenburg). Mice were housed with a 12-h light/dark cycle and had free access to a standard laboratory chow diet (B&K, Solna, Sweden) and drinking water. All animal experiments were carried out in accordance with the recommendations of the Animal Ethical Committee of the University of Gothenburg, which approved the protocol (ethical number 5/2013 and 58-2016).

### HI Procedure

For flow cytometry and immunohistochemical staining, WT and *Rag 1*<sup>−</sup>/<sup>−</sup> mice of both sexes were exposed to HI at PND5, a developmental stage where the brain development in the animals corresponds to preterm infants. The period PND2–5 is when rodents have the highest percentage of pre-OLs and thus mimic the vulnerable developmental stages of preterm human infants (18). Briefly, mice were anesthetized with isoflurane (5.0% for induction and 1.5–3.0% for maintenance). The left common carotid artery was ligated, and the mice were returned to their dam and allowed to recover for 1 h. The mice were then placed in an incubator perfused with a humidified gas mixture (10% oxygen) at 36°C for 70 min. After HI, the pups were returned to their dam until they were sacrificed. For flow cytometry, WT mouse pups were sacrificed at 6 h, 3 days, or 7 days after HI, and for immunohistochemical staining, both WT and *Rag 1*<sup>−</sup>/<sup>−</sup> mice were sacrificed 7 days after HI.

#### Assessment of Brain Damage

Deeply anesthetized mouse pups were subjected to transcranial perfusion with saline and 5–10% buffered formaldehyde (Histofix; Histolab Products AB, Gothenburg, Sweden). Dissected mouse brains were embedded in paraffin and cut into 10-µm coronal sections. The extent of white matter and gray matter injury at the hippocampal level was analyzed after immunohistochemical staining for myelin basic protein (MBP) and microtubule-associated protein 2 (MAP-2), respectively. The quantitative measurements of the brain injury were done by manually outlining the MBP- and MAP-2-positive areas using Micro Image version 4.0 (Micro-macro AB, Gothenburg, Sweden) as previously described (9) by an investigator blinded to the treatment groups.

We measured the following three parameters for each brain: (1) the total MAP-2-positive area in both brain hemispheres, (2) the total MAP-2-positive area in the hippocampus area of both hemispheres, and (3) the MBP-positive area in the subcortical white matter in both hemispheres. The total tissue loss was calculated as: (Contralateral hemisphere − Ipsilateral hemisphere)/ (Contralateral hemisphere) × 100%. One brain section/mouse at the hippocampus level was evaluated. Our previous study using simple linear regression analysis comparing the total brain tissue loss (volume) with the tissue loss in one representative brain section (area) from the hippocampus level showed a significant positive linear correlation between the two methods (15), suggesting that a single representative brain section/mouse can be used to estimate tissue loss.

#### Human Postmortem Brains

Informed parental consent was acquired from all individual participants included in the study according to the World Medical Association's Declaration of Helsinki and the guidelines of the National Health Service UK. Ethics permission for the study was obtained from the National Research Ethics Service, Hammersmith and Queen Charlotte's and Chelsea Research Ethics Services, London, UK (ethics number 07/H0707/139). Three extremely preterm postmortem brains (<28 weeks gestational age) of vaginally delivered neonates with PVL, and three without PVL (serving as controls), were obtained from the Department of Perinatal Pathology, Imperial Health Care Trust, London, UK. Paraffin-embedded tissue sections from the frontal and parietal lobes (at the level of Ammon's horn) of the postmortem brains were used for immunohistochemistry staining. The primary cause of death of each case was assessed by a pathologist. The details of each case are summarized in **Table 1**.

#### Immunohistochemistry Staining

Immunohistochemistry procedures were performed as previously described (10). Mouse brain sections were immersed in xylene twice and then in graded ethanol (100, 95, and 70%) to remove the paraffin and rehydrate the brain sections. Antigen retrieval was then performed by boiling the sections in sodium citrate buffer (pH 6.0) for 10 min followed by 3% H2O2 for 10 min to block endogenous peroxidase activity and background staining. After antigen recovery and blocking, the sections were incubated overnight at 4°C with the mouse anti-MAP-2 (clone HM-2, Sigma-Aldrich, Stockholm, Sweden) or mouse anti-MBP (SMI94, Covance, NJ, USA) primary antibodies followed by incubation with the appropriate biotinylated secondary antibody (Vector Laboratories, Burlingame, CA, USA) for 1 h at room temperature. Visualization was performed using Vectastain ABC Elite with 3,3′-diaminobenzidine (DAB) enhanced with ammonium nickel sulfate, beta-d glucose, ammonium chloride, and beta-glucose oxidase (all from Sigma-Aldrich, Stockholm, Sweden).

The processing of the human postmortem brains and the immunohistochemical staining were performed as described previously (19, 20). Human brain sections were blocked in 5% goat serum (Vector Laboratories) for 20 min before being incubated overnight at 4°C in a solution of anti-human CD3 antibody (cat. Code A0452, clone F7.2.38, Agilent, Carpento, CA, USA) or anti-CD20 antibody (cat. MA5-13141, clone L26, Agilent). The next day, the sections were incubated with biotinylated goat anti-rabbit IgG secondary antibody (15 µg/ml; Vector Laboratories) in PBS for 1 h followed by avidin–biotin complex for 1 h (1:200 dilution; Vector Laboratories). The reactions were visualized with DAB (Sigma-Aldrich, Gillingham Dorset, UK) for 10 min. Finally, the sections were dehydrated, cleared in xylene, and coverslipped. As negative controls, we performed staining in the absence of the primary antibodies and instead used the isotype controls of the respective primary antibodies. The sections were examined under bright-field microscopy using a light microscope (DM6000 B; Leica Microsystems Ltd., Bucks, UK).

#### Mouse Brain Mononuclear Cell (MNC) Isolation and Flow Cytometry Analysis

Mice were sacrificed at 6 h, 3 days, or 7 days after HI for flow cytometric analysis. After a brief transcranial perfusion with saline, the brains were dissected out and divided into ipsilateral and contralateral hemispheres. To prepare single-cell suspensions, the dissected brains were cut into small pieces using a sharp razor blade and incubated with an enzyme mixture containing 0.01% papain, 0.01% DNase I (Worthington, NJ, USA), 0.1% Dispase II (Roche, Sweden), and 12.4 mM MgSO4 in Ca/Mg-free HBSS (Thermo Fisher, Sweden) for 20 min at 37°C with gentle triturations before incubation and after 10 min of incubation. The single-cell suspensions were centrifuged at 300 *g* for 5 min at 4°C, and the pellets were resuspended in 5 ml of 30% isotonic Percoll (Amersham Biosciences) that was overlaid onto 70% isotonic Percoll. The Percoll gradient was centrifuged at 1,000 *g* for 30 min at room temperature. MNCs were collected from the 70/30% Percoll interface and washed with ice-cold PBS containing 1% bovine serum albumin.


−*: no positive cells;* ±*: positive cells seen inside the blood vessels;* +*: positive cells seen;* + +*: positive cells easily seen.*

*GA, gestational age; PVL, periventricular leukomalacia; PVWMI, periventricular white-matter injury; AFI, amniotic fluid infection.*

In order to identify adaptive immune cells in brain samples, isolated MNCs (1 × 106 cells) from each sample were first incubated with anti-mouse CD16/32 antibody in 100 µl FACS buffer (PBS and 1% BSA) for 15 min at 4°C to block the Fc receptor and then stained with anti-CD3 (FITC, 145-2C11, BD Biosciences), anti-TCRαβ (APC Cy™7, clone H57-597, BD Biosciences), anti-TCR γδ (PE-Cy™7, clone GL3, eBioscience), anti-CD19 (PE, clone 1D3, BD Biosciences), anti-CD45 (V500, clone 30-F11, BD Biosciences), and anti-CD11b (APC, clone M1/70, eBioscience) antibodies for 30 min at 4°C. After staining, the samples were washed with 500 µl FACS buffer, and the pellets were resuspended in 350 µl of FACS buffer. Dead cells were labeled by adding 7-AAD to the final 10 min of antibody staining, and these cells were excluded from the analysis. All samples were immediately acquired on a BD FACSCantoII™ flow cytometer. A total of 5 × 105 events was acquired, and the data were analyzed with FlowJo software (Tree Star, Ashland, OR, USA).

Cells were first gated based on size and granularity, then doublets and dead cells were excluded. At the 6 h time point where CD45 and CD11b were included in the staining cocktail, CD3<sup>+</sup>, CD19<sup>+</sup>, CD3<sup>+</sup>, and αβTCR<sup>+</sup> lymphocyte population back gating on the CD45 and CD11b plot was performed to ensure that they were CD45 single-stained cells and that the identified cell populations were not the result of nonspecific binding of antibody to cells of myeloid origin with incomplete Fc blockage (Figures S1 and S2 in Supplementary Material).

#### Statistics

Data were analyzed by Student's paired *t*-test (for flow cytometry data) and Student's unpaired *t*-test (for the tissue loss evaluation data), and the results are presented as means ± SEM with significance set at *p* < 0.05. All statistical analyses were carried out using GraphPad Prism software (version 6.02).

#### RESULTS

#### CD3**+** T and CD20**+** B Cells Were Found in the Postmortem Preterm Infant Brains With PVL

The CD3 molecular complex is the coreceptor of the T cell receptor (TCR) and consists of five subunits—a, γ, δ, and two ε chains—and antibodies against the ε chains are often used as pan T cell markers. Immunohistochemistry on postmortem brain sections from preterm infants with PVL showed a number of CD3<sup>+</sup> T cells in the periventricular white matter and meninges (**Figures 1A,B**). There was a higher frequency in the meninges, including cells outside the blood vessels and cells in the inner layer of the blood vessels (**Figure 1B**; **Table 1**). Such positive staining for CD3 was not observed in control cases that did not have PVL (**Figure 1C**).

CD20<sup>+</sup> (a B cell-specific surface antigen) cells were present in the periventricular white matter in all PVL cases, albeit in small numbers (**Figure 1D**). B cells were also observed in the meninges of PVL cases (**Figure 1E**). In control cases that did not have PVL, the immunohistochemistry staining either did not reveal CD20 positive staining or the CD20<sup>+</sup> cells were only observed within the blood vessels (**Figure 1F**).

### CD3**+** T Cells Were Found in the Mouse Brain After HI-Induced Preterm Brain Injury

Hypoxia–ischemia triggers inflammatory processes leading to the infiltration of different immune cells at different times after injury. T cell recruitment and activation has been shown to play an important role in cerebral ischemia–reperfusion injury in adults (21, 22), but whether this is true for neonates remains unknown.

To study the presence of T cells in the brain, we used an established neonatal mouse model of HI-induced brain injury at a developmental age (PND5) where the brain development in the animals corresponds to that of preterm human infants (15), and we performed flow cytometry on MNCs isolated from the brains of mice subjected to HI. Significant increases in CD3<sup>+</sup> T cells in the ipsilateral (injured) hemispheres compared to the contralateral (uninjured) hemispheres were observed at 3 days (*p* = 0.05) and 7 days (*p* = 0.01), but not at 6 h, after HI (**Figures 2A–C**).

### Increase in B Cell Frequency in the Mouse Brain Long After HI

A recent study has implicated B cells as one of the causal factors of cognitive impairments after adult stroke (23), but the role of B cells in neonatal preterm brain injury has not been studied. We used staining for CD19, the coreceptor of the B cell receptor, and flow cytometry to determine whether B cells were present in the neonatal mouse brain after HI-induced injury. The frequency of B cells was only significantly higher in the ipsilateral hemisphere compared to the contralateral hemisphere at 7 days (*p* = 0.0095), and not at 6 h or 3 days after HI (**Figures 2A,B,D**; Figure S1 in Supplementary Material).

### Increase in **αβ**T Cell Frequencies in the Mouse Brain Long After HI

Based on the type of TCR expressed on their surface, T cells are divided into αβT and γδT cells. The majority of T cells in mammals express conventional αβ TCRs on their surface, through which they recognize peptide antigens presented on MHC molecules by antigen-presenting cells. Flow cytometric analysis of mouse brains at 6 h, 3 days, and 7 days post-HI showed increased frequency of αβT (CD3<sup>+</sup> TCRβ+) cells in the ipsilateral hemisphere compared to the contralateral hemisphere at 7 days post-HI (*p* = 0.036). No change in αβT ells was observed in the ipsilateral hemisphere at 6 h after HI (**Figures 3A–C**; Figure S2 in Supplementary Material).

Interestingly, A small population of CD3<sup>+</sup> TCRβ− cells was identified at 6 h, 3 days, and 7 days after HI. However, there was no significant difference between ipsilateral and contralateral frequency at any of the time points (**Figures 3A,B,D**; Figure S2 in Supplementary Material).

Neonates have reduced capacity for mounting conventional αβT cell responses due to the immature status of the αβT cells. However, γδT cells are already functionally competent during early development and are important in early life immunity. We thus evaluated the presence of γδT cells in the neonatal mouse brain at different time points after HI.

without (case 1) PVL. (D–F) CD20+ B cell immunohistochemical staining in the periventricular white matter (D,F) and meninges (E) of brain sections from postmortem preterm infant brains with (case 5) and without (case 3) PVL. Positively stained cells are indicated by arrowheads. The inserts show a higher magnification of positively stained cells.

Flow cytometry analysis of MNCs isolated from brains at 6 h, 3 days, and 7 days after HI showed that there was a small population of γδT cells in the brains of naive mice, as well as in both the ipsilateral and the contralateral hemispheres of HI mice at all time points examined. However, no difference in the percentage of γδT cells was observed between the ipsilateral and contralateral hemispheres at any time points examined or between the naive animals and the animals exposed to HI (data not shown).

### Lack of Mature T and B Cells Protected Against HI-Induced White-Matter Injury

The recombination-activating gene 1 (*Rag1*) and *Rag2* genes encode the V(D)J recombinases Rag1 and Rag2 that are responsible for the rearrangement of antigen-receptor genes during T cell and B cell development. *Rag1*<sup>−</sup>/<sup>−</sup> mice do not produce mature T or B cells (24), thus they are ideal for studying the effect of those immune cells in brain injury.

To understand the involvement of mature T and B cells in HI-induced preterm brain injury, HI was induced in *Rag1*<sup>−</sup>/<sup>−</sup> and WT mice. The extent of gray-matter injury (as indicated by MAP-2 staining, **Figure 4A**) and white-matter injury (as indicated by MBP staining, **Figure 4B**) was measured. At 7 days after HI (PND12) when myelination was formed and was visible by MBP staining, there was a significant reduction in total tissue loss in the subcortical corpus callosum white matter of *Rag1*−/− mice (11.90 ± 2.67%) compared to WT mice (23.43 ± 2.27%) (*p* = 0.0023) (**Figures 4C,D**). Accordingly, in the subcortical corpus callosum and in the cortex of the ipsilateral hemisphere of brain sections, *Rag1*<sup>−</sup>*/*<sup>−</sup> mice showed

B (D) cells within the live cell population in the ipsilateral (Ipsi) and contralateral (Contra) hemispheres of mouse brains at 6 h (*n* = 6), 3 days (*n* = 5), and 7 days (*n* = 8) post-HI. \**p* < 0.05, \*\**p* < 0.01 using Student's paired *t*-test. Data are presented as the mean ± SEM.

more organized and denser myelin structure compared to WT mice (**Figures 4B,C**) indicating reduced HI-induced white-matter injury in the Rag1<sup>−</sup>/<sup>−</sup> mice compared to the WT mice.

The extent of injury in the gray-matter hemispheres of WT and *Rag1*−*/*− mice was very similar (**Figure 4A**), and no significant difference in the total gray-matter tissue loss was observed in *Rag1*<sup>−</sup>*/*<sup>−</sup> compared to WT mouse brains (**Figure 4E**). Further analysis of the most severely injured part of the mouse brain showed that the total tissue loss in the hippocampus area was also not significantly different between the WT and *Rag1*<sup>−</sup>*/*<sup>−</sup> mice (58.1 ± 10.43% in WT mice versus 67.66 ± 6.998% in the *Rag1*<sup>−</sup>*/*<sup>−</sup> mice, *p* = 0.3543).

(*n* = 5), and 7 days (*n* = 8) post-HI. \*\**p* < 0.01 using Student's paired *t*-test. Data are presented as the mean ± SEM.

### DISCUSSION

The present study evaluated the potential contribution of lymphocytes to neonatal brain injury. We have provided evidence for the presence of T and B cells in the mouse brain following HI-induced neonatal brain injury. T and B cells were also present in the brains of postmortem human preterm infants with PVL; moreover, we showed that mice lacking mature T and B cells have reduced HI-induced white-matter injury.

Lymphocytes, particularly T and B lymphocytes, have been implicated in the pathogenesis of ischemic brain injury. CD4<sup>+</sup> and CD8<sup>+</sup> T cells have been shown to contribute to the

Student's unpaired *t*-test. Data are presented as the mean ± SEM.

inflammatory response, brain injury, and subsequent neurological deficits associated with adult experimental stroke (3, 21, 22). B cells have also been shown to be involved in the cognitive impairments secondary to ischemic stroke, and activated B cells infiltrate the infarct tissue adjacent to the lesion in the weeks after stroke where they undergo class switching (23). Infiltrating lymphocytes, especially T cells, can contribute to ischemic brain injury through direct damage to neurons *via* secretion of granules and cytokines, as well as through the activation of microglia, neutrophils, and brain endothelial cells (25). The circumventricular organs and perivascular spaces have been suggested as the probable route of peripheral immune cell infiltration (17, 26). Studies using both mouse and human poststroke autopsy samples have suggested the choroid plexus as the key cerebral invasion route for T cells after stroke (27).

T cells and other peripheral immune cells are found in the neonatal rodent brain following HI and persist for hours to months postinjury (12–14, 28). Consistent with these findings, our present study has shown that CD3<sup>+</sup> T cells were found in the neonatal mouse brain after HI injury, and the frequency of these cells gradually increased in the days following HI. We have also shown that αβT cells and B cells were present in the injured brain at relatively late stages of the immune response.

Due to their innate-like nature and site of residence, γδT cells are involved in early immune responses against pathogenic insults in tissues (29, 30). In our recent publication on the contribution of γδT cells to neonatal brain injury, we found that γδT cells were present in the injured hemisphere as early as 6 h after injury, and mice deficient in γδT cells were protected from HI-induced (28) and sepsis-induced (11) brain injury. Different from the adult mice (31, 32), the protection we observed in the neonatal mice was IL-17 and IL-22 independent (28).

In the present study, using flow cytometry, we were able to detect a small population of γδT cells in the mouse brain, but no significant differences were observed between the naive and HI-injured brain nor between the ipsilateral and contralateral hemispheres at any of the time points examined. This might be due to the fact that γδT cells were observed in the brain meninges in both the neonatal and adult mice (28, 33); therefore, using whole brain homogenate to run flow cytometry is not a sensitive method for detecting the differences in this relatively small γδT cell population.

In contrast to the γδT cells, we found that the frequency of αβT cells increased in the injured neonatal brain at late time points (7 days) after HI. This is in line with previous findings in rodent models of neonatal HI brain injury and in adult experimental stroke (12–14). Conventional αβT cells start to infiltrate the injured brain as early as 3 days after stroke in mice (3). A similar study showed CD4+ T cell infiltration in the infarct region starting 7 days after ischemia, and the number of T cells peaked 14 days postinjury (34). Taken together, our findings and those of others suggest that conventional αβT cells are involved in immune responses at late time points after brain injury in both neonatal and adult mice.

We observed the presence of CD3+ TCRβ− cells in the neonatal HI brain at all assessed time points. These cells could be natural killer T (NKT) cells or CD4 and CD8 double-negative T cells. NKT cells have been shown to play a role in ischemic reperfusion brain injury, and especially in the associated systemic bacterial infection (35). However, the role of CD4 and CD8 double-negative T cells in adult and neonatal ischemia injury has not been studied until now. Despite non-significant changes in this population in the current study, future studies investigating the role of these CD3<sup>+</sup> TCRβ− cells in neonatal HI brain injury could provide further understanding of the role of lymphocytes in this injury model.

Using flow cytometry, we observed the presence of a small but clear B cell population in the mouse brain after HI. The frequency of these cells was significantly increased in the ipsilateral hemisphere 7 days after injury when compared to the contralateral hemisphere. In adults, activated B cells infiltrate the infarct tissue in the weeks following stroke, and lack of B cells prevents the appearance of delayed cognitive deficits caused by ischemia (23). In contrast, a recent study has shown that B cells do not have a major pathophysiological role in acute ischemic stroke in mice (36). To better understand the role of B cells in neonatal brain injury, further studies are needed.

Using immunohistochemistry, we have identified the presence of CD3+ T cells, and to a smaller extent B cells, in the postmortem brains of preterm infants. Despite the limitation of the total number of clinical cases used in this study, the findings were consistent and correlated with the mouse data. Therefore, the human postmortem preterm samples served as a valuable complimentary addition and strengthened the conclusions of the current study. Together with our other recent publication (28), we provide evidence for the presence of lymphocytes in the human postmortem brain of preterm infants with PVL. These results demonstrate a clear requirement for further studies using clinical cases to better characterize the role of these immune cells in neonatal brain injury.

Our *Rag1*<sup>−</sup>*/*<sup>−</sup> data demonstrated that the lack of T and B cells protects against HI-induced preterm brain injury. *Rag1*<sup>−</sup>*/*<sup>−</sup> mice showed reduced white matter brain tissue loss compared to WT mice at 7 days after HI. This finding is consistent with previous adult stroke studies, where mice lacking both T and B cells have significant reductions in brain injury (21, 22). Despite the potential contribution of innate immune cells in *Rag1*<sup>−</sup>/<sup>−</sup> mice to the protection against brain injury (37–39), our studies and others support the critical role of T and B cells in ischemic brain injury in both immature and mature developmental stages.

The protective effect of T and B cells deficiency is only observed in the white matter and not in the gray matter of HI mice. This is an interesting finding and warrants further investigations in the future.

The lack of molecular mechanisms behind the functional regulation of T and B lymphocytes, as well as long-term anatomical and functional outcomes, are some of the limitations of the current study. Furthermore, brain injury could have been more thoroughly evaluated by a full stereological assessment.

In summary, our study provides a temporal analysis of the presence of different populations of lymphocytes in a neonatal mouse model of preterm brain injury. Together with our other most recent publications, we have provided a description of the immune response (15) and the presence of lymphocytes (28), especially γδT cells (11, 28), in the mouse and human brain after preterm brain injury. Our current observation that mature lymphocyte deficiency protects neonatal mice from HI-induced brain white matter injury provides new evidence for the mechanism of neonatal brain injury that will help in the development of therapeutic strategies for such conditions.

#### ETHICS STATEMENT

National Research Ethics Service, Hammersmith and Queen Charlotte's and Chelsea Research Ethics Services, London, UK (ethics number 07/H0707/139). Animal Ethical Committee of the University of Gothenburg, Sweden (ethical number 5/2013 and 58-2016).

### AUTHOR CONTRIBUTIONS

XW, JWL, AN, and A-MA conceived and designed the study; AN, A-MA, XZ, ER-F, JL, AZ, and RV performed the experiments and analyzed the data; AN drafted the manuscript. XW, JL, ER-F, JWL, and CM revised the manuscript. MR, CZ, CM, HH and GN revising it critically for important intellectual content, and provide approval for publication of the content. All authors contributed to data interpretation and approved the submitted version.

#### ACKNOWLEDGMENTS

This work was supported by the Swedish Research Council (VR 2013-2475 and VR 2015-06276 to XW), Swedish governmental grants to researchers in the public health service (ALFGBG-429801 to XW), the Gothenburg Medical Society (011/14 to XW), the Chinese Scholarship Council (201407040032 to XZ), the Frimurare Barnhus Foundation (to AN), Wilhelm & Martina Lundgren's Foundation (to AN and XZ), the National Natural of Science Foundation of China (U1704281 to CZ, 81771418 & U1604165 to XW), the Department of Science and Technology of Henan Province (171100310200 to CZ), and the Science and Technology Bureau of Zhengzhou (131PCXTD621 to CZ).

### REFERENCES


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fneur.2018.00159/ full#supplementary-material.

Figure S1 | Gating strategy of CD19 B cells in the mouse brain after hypoxia– ischemia (HI). (A) Representative flow cytometry plots showing the gating strategy of CD3− CD19+ events in the brain at 6 h after HI. (B) The CD3− CD19<sup>+</sup> population backgated on CD11b versus CD45 plot to ensure a true population was detected and not myeloid cells with insufficient blocking of Fc receptors. (C) Fluorescent minus one (FMO) controls for CD19 staining.

Figure S2 | Gating strategy of T cells in the mouse brain after hypoxia– ischemia (HI). (A) Representative flow cytometry plots showing the gating strategy of CD3+ TCRβ+ events in the brain at 6 h after HI. (B) The CD3+ TCRβ+ population backgated on CD11b versus CD45 plot to ensure a true population was detected and not myeloid cells with insufficient blocking of Fc receptors. (C) Fluorescent minus one (FMO) controls for CD3 and TCRβ staining.


in crescentic GN. *J Am Soc Nephrol* (2012) 23(9):1486–95. doi:10.1681/ASN. 2012010040


**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 Nazmi, Albertsson, Rocha-Ferreira, Zhang, Vontell, Zelco, Rutherford, Zhu, Nilsson, Mallard, Hagberg, Lai, Leavenworth and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Sujith S. Pereira1,2, Stephen T. Kempley1,2, David F. Wertheim3 , Ajay K. Sinha1,2, Joan K. Morris <sup>4</sup> and Divyen K. Shah1,5\**

*1Neonatal Unit, Royal London Hospital, Barts Health NHS Trust, London, United Kingdom, 2Centre for Genomics and Child Health, Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom, 3 Faculty of Science, Engineering and Computing, Kingston University, Kingston upon Thames, United Kingdom, 4Centre for Environmental and Preventive Medicine, Wolfson Institute of Preventive Medicine, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom, 5Centre for Neuroscience and Trauma, Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom*

#### *Edited by:*

*Stephane Vladimir Sizonenko, Geneva University Hospitals (HUG), Switzerland*

#### *Reviewed by:*

*Sarah A. Kelley, Johns Hopkins University, United States Gunnar Naulaers, KU Leuven, Belgium*

*\*Correspondence:*

*Divyen K. Shah d.shah@qmul.ac.uk*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 27 November 2017 Accepted: 07 February 2018 Published: 26 February 2018*

#### *Citation:*

*Pereira SS, Kempley ST, Wertheim DF, Sinha AK, Morris JK and Shah DK (2018) Investigation of EEG Activity Compared with Mean Arterial Blood Pressure in Extremely Preterm Infants. Front. Neurol. 9:87. doi: 10.3389/fneur.2018.00087*

Background: Cerebral electrical activity in extremely preterm infants is affected by various factors including blood gas and circulatory parameters.

Objective: To investigate whether continuously measured invasive mean arterial blood pressure (BP) is associated with electroencephalographic (EEG) discontinuity in extremely preterm infants.

study design: This prospective observational study examined 51 newborn infants born <29 weeks gestation in the first 3 days after birth. A single channel of raw EEG was used to quantify discontinuity. Mean BP was acquired using continuous invasive measurement and Doppler ultrasound was used to measure left ventricular output (LVO) and common carotid artery blood flow (CCAF).

results: Median gestation and birthweight were 25.6 weeks and 760 g, respectively. Mean discontinuity reduced significantly between days 1 and 3. EEG discontinuity was significantly related to gestation, pH and BP. LVO and CCAF were not associated with EEG discontinuity.

conclusion: Continuously measured invasive mean arterial BP was found to have a negative relationship with EEG discontinuity; increasing BP was associated with lower EEG discontinuity. This did not appear to be mediated by surrogates of systemic or cerebral blood flow. Infants receiving inotropic support had significantly increased EEG discontinuity on the first day after birth.

#### Keywords: EEG, electroencephalogram, EEG continuity, blood pressure, preterm infant

**Abbreviations:** aEEG, amplitude-integrated electroencephalography; BP, blood pressure; CCAF, common carotid artery blood flow; EEG, electroencephalography; GA, gestational age; IQR, interquartile range; LVO, left ventricular output; PaCO2, partial pressure of carbon dioxide in blood; PDA, patent ductus arteriosus; UAC, umbilical arterial catheter.

## INTRODUCTION

Infants born extremely preterm are at risk of cerebral injury (1) and the underlying mechanisms are poorly understood. Cerebral electrical activity in extremely preterm infants is influenced by multiple factors, for example administration of certain drugs and partial pressure of carbon dioxide in blood (PaCO2) can affect cerebral electrical activity in the days immediately after birth (2–4). Cerebral electrical activity has been previously examined in this group of infants in association with respiratory and metabolic indicators (5–7), circulatory measurements (3, 4, 8, 9), morphine (10–12), and inotropic therapy (13).

Circulatory parameters such as cerebral perfusion and blood pressure (BP) may affect cerebral electrical activity (3, 8, 9). It is still unclear as to whether mean BP could be associated with cerebral electrical activity in the extremely preterm. Studies have investigated the relationship between BP, left ventricular output (LVO), and cerebral electrical activity (8, 9, 14) but have been limited by patient numbers (9, 14) and the use of non-invasive BP measurements (8). There is a paucity of large prospective studies jointly examining the relationship between various clinical measurements including blood gas levels, the use of sedation and circulatory indicators with cerebral electrical activity in a recent cohort of extremely premature newborn infants.

We thus investigated mean arterial BP and EEG activity in the first 3 days in extremely preterm infants. In addition, we took account of pH, PaCO2, lactate, morphine administration, and a measure of cerebral perfusion.

#### MATERIALS AND METHODS

#### The Study Population

Infants, both inborn and outborn, were eligible to participate if they were born at less than 29 weeks gestation, recruited within 12 h of age on the Neonatal Unit at the Royal London Hospital between February 2013 and April 2015. Formal exclusion criteria for this study included the presence of major congenital malformations and infants who did not have invasive arterial lines. This study received approval from the Research Ethics Committee (reference 12/LO/1553). Written parental consent was obtained prior to the start of the study.

### Amplitude-Integrated Electroencephalography (aEEG) Monitoring

Amplitude-integrated electroencephalography activity was recorded for 72 h in the majority of the infants using a 2-channel BRM3 monitor (BrainZ Instruments, Natus Medical Incorporated, ON, Canada) which provides a digital raw EEG signal output as well as the aEEG. After preparation of the scalp using NuPrep™ gel (Nuprep, D O Weaver & Co., Aurora, CO, USA) to reduce skin impedance, neonatal hydrogel electrodes (Neonatal Sensors, Natus Medical Incorporated, ON, Canada) were placed on the frontoparietal regions (C3-P3, C4-P4) bilaterally according to the international 10–20 system (15, 16). A 2-h artifact and seizure free electroencephalogram trace, confirmed by two observers (Sujith S. Pereira and Divyen K. Shah), recorded before and after measurement of the carotid artery blood flow was chosen for analysis.

#### EEG Discontinuity

Single channel cross-cerebral (P3-P4) raw EEG data were exported to Microsoft Excel® and continuity was analysed in 1-min epochs with software that we developed using MATLAB (The MathWorks, Inc., MA, USA) using a similar approach to that previously described (17). The system detected an interval if the absolute amplitude of the raw EEG was less than 20 µV with respect to the baseline for at least 6 s. The 20 µV threshold was chosen to reflect the fact that the EEG from preterm newborns is represented by more high voltage low frequency wave forms in contrast to full term newborns (18) and to help reduce any effects of background noise. The threshold level was thus chosen so as to reliably identify EEG bursts and distinguish them from background noise artefact in view of visual assessment of raw EEG characteristics. The 6 s criterion for defining an interval was chosen in order to exclude quiescent periods that are normally associated with tracé alternans. P3-P4 raw EEG was analysed, as C3P3 and C4P4 tend to be more susceptible to artifact due to the shorter inter-electrode distance. For each recording the mean of the total interval length per epoch, the discontinuity value, was calculated and expressed in seconds; this can also easily be expressed as a discontinuity proportion since the epoch length is constant.

#### BP Monitoring

As invasive BP monitoring is considered the gold standard, only infants with invasive arterial lines were included in this study. Umbilical arterial catheters (UAC) were inserted aiming for the tip of the catheter to be maintained between 6th and 10th thoracic vertebral levels. Following insertion of the UAC, patency of the line was maintained by continuous infusion of heparinised saline. GE Healthcare medical systems monitor (Carescape Monitor B850) were used to trace the heart rate, BP, oxygen saturations levels, and respiratory rate. BP calibration was performed with the transducer being held in the mid axillary line at the start of the study and every 24 h thereafter. The UAC was only used after ensuring that the line was free from air bubbles, it sampled and flushed well and produced a good arterial waveform tracing. If the UAC was malpositioned or blocked, a peripheral arterial line was inserted and used after the above-mentioned precautions were taken. The heart rate, systolic, diastolic, and mean BP were monitored and downloaded every 10 s for the first week. A 2-h artefact free period of BP data, before measurement of common carotid artery blood flow (CCAF), was chosen for analysis.

#### CCAF Measurement

Doppler ultrasound with a 7–15 MHz linear array probe (L15-7io Broadband compact linear array probe, Philips iE33, Bothwell, WA, USA) was used to measure the right CCAF volume on days 1 and 3. CCAF was used as a marker of cerebral blood flow using previously established methods (19) that have indicated good repeatability and reproducibility. An average of 5 right common carotid artery diameter and velocity time integral measurements were taken to calculate the blood flow volumes performed by one rater (SSP) after training. The right common carotid artery was used as it is furthest away from the ductus arteriosus and is less likely to be influenced by a patent ductus arteriosus (PDA) compared with the left common carotid artery. Whilst performing this examination, the presence or absence of a PDA on color-flow Doppler was also recorded.

#### LVO Measurement

Doppler ultrasound with a 4–12 MHz sector array probe (S12-4, cardiac ultrasound probe, Philips iE33, Bothwell, WA, USA) was used to measure the LVO immediately after measuring the CCAF using methods that have been well established (20) on days 1 and 3. This method of estimation of LVO has been found to have good correlation with that measured using phase contrast MRI (21).

Mean arterial BP (averaged over a 2-h epoch) was compared with EEG discontinuity over the same epoch. Prior to the start of the study, care was taken to ensure that, for every infant, time was synchronised accurately to the minute across all the equipment used in the study. The relationship between EEG discontinuity to LVO, CCAF, and BP could thus be explored.

#### Inotropic Support

A written policy for initiation of inotropic therapy was available at the cotside. Typically, infants were given a 10 ml/kg bolus of 0.9% saline and were then commenced on a dopamine infusion as necessary. Further inotropic agents were chosen based on the results of functional echocardiography that was performed on all infants in this study.

#### Blood Gas Parameters

Blood gas parameters such as pH, PaCO2 and lactate values were chosen from single measurements that were closest to the measurements of CCAF and LVO on days 1 and 3.

#### Statistical Methods

Data were tested for consistency with a normal distribution. Skewed data underwent logarithmic transformation for analysis. Effects on EEG discontinuity were analysed using independent samples *t*-tests for categorical variables, and Pearson's correlation for continuous variables. For factors showing significant effects on discontinuity in these analyses (*p* < 0.05), a mixed effects multiple regression analysis was performed to identify predictors of EEG discontinuity, retaining gestation rather than birthweight as a measure of maturity. All statistical analyses were performed using SPSS v22 (Chicago, IL, USA) and Stata Release 12 (StataCorp LLC, College Station, TX, USA).

### RESULTS

### Patient Characteristics

Of 134 cases assessed for eligibility, 59 were recruited to the study (**Figure 1**). Fifty-one infants had invasive mean BP monitoring on day 1 and 41 infants on day 3. The clinical characteristics of recruited infants are shown in **Table 1**. The median [interquartile range (IQR)] age of days 1 and 3 scans were 18 (13–22) h and 74 (67–79) h, respectively. Sedation using morphine was administered in 16 infants on days 1 and 17 infants on day 3. One infant

#### Table 1 | Patient and clinical characteristics.


*Where not specified, all figures are expressed as median (interquartile range). SI conversion factors: to convert PaCO2 to mmol/L, multiply values by 0.133. SI conversion factors: to convert lactate to mmol/L, multiply values by 0.111. a n* = *14.*

*bPaired t-test or chi-squared test.*

who received anticonvulsants in the first 72 h in view of suspected clinical seizures was excluded from analysis.

#### EEG

Analysable aEEG traces were obtained from 51 (100%) infants on days 1 and 36 (88%) infants on day 3. Infants with poor quality EEG signal and high impedance on day 3 were not included for analysis. From day 1 to day 3, EEG discontinuity decreased significantly (**Table 1**). EEG discontinuity decreased with increasing gestational age (GA) and this relation was more on day 1 than on day 3 (**Figure 2A**). Acidosis on day 1, higher PaCO2 and lactate, were related to lower voltage and a more discontinuous EEG (**Figures 2B–D**).

Morphine administration was significantly associated with increased mean discontinuity on both days. For those infants not receiving morphine, compared with those on morphine, median (IQR) mean discontinuity values were: 21 (13–25) vs. 36 (27–41) s (*p* < 0.001) on day 1 and 13 (9–21) vs. 23 (13–40) s (*p* = 0.022) on day 3.

Infants receiving inotropes had significantly (*p* < 0.001) suppressed mean discontinuity on day 1 only; on day 1 the proportion of infants studied who received inotrope treatment was 53%. For infants not receiving inotropes compared with infants on inotropes, median (IQR) mean discontinuity values were: 16 (12–24) vs. 29 (21–37) s. Inotrope administration did not appear to be associated with significant EEG discontinuity change on day 3 where a lower proportion (34%) of infants received inotropic support.

Continuously measured invasive mean arterial BP showed a significant relationship with EEG discontinuity; higher BP associated with lower EEG discontinuity on both day 1 and day 3 (**Figures 2E,F**). There was no correlation between LVO and CCAF and EEG discontinuity on day 1 or 3.

Using mixed effects multiple regression analysis (**Table 2**), we found that factors influencing mean EEG discontinuity include gestation (β = 3.57, *p* = 0.001), PaCO2 (β = 9.48, *p* = 0.009), lactate (β = 4.24, *p* = 0.028), morphine (β = 9.85, *p* < 0.001), and invasive mean arterial BP (β = −1.04, *p* < 0.001).

#### DISCUSSION

The most unwell and immature infants would be expected to have the lowest EEG continuity; furthermore administration of inotropic support may be an indication of the degree to which the infant was unwell. However, additionally, this study found gestation, PaCO2, lactate, morphine administration and invasive mean arterial BP were significantly associated with EEG discontinuity in extremely preterm infants during the first 3 days after birth.

EEG discontinuity was related to GA in agreement with published data (22–26). We observed that acidosis and hypercapnia were associated with increased EEG discontinuity as previously reported (6, 7, 27–29). The suppression of EEG caused by hypercapnia may be exerted through changes in pH. Hypercapnia is associated with altered neuronal nuclear enzyme activity and a reduction in ATP and phosphocreatinine levels that reflect energy metabolism in animal models (30). As energy is required for maintenance of electrical activity in the brain (31), the resulting neuronal hyperpolarisation during hypercapnoea was associated with a reduction in the steepness, amplitude and duration of excitatory postsynaptic potentials (32).

In our study as in others (10, 12), morphine therapy was significantly associated with suppression of the EEG activity even though only 31% of infants on day 1 and 41% of infants on day 3 received morphine. The increase in EEG discontinuity noted with the administration of inotropes on day 1 may be due to prior hypotension triggering inotropic support, as there was no effect seen on day 3, by which time the BP levels would have stabilised; the proportion of infants receiving inotropic support fell from day 1 to day 3.

Mean invasive arterial BP was found to have a significantly negative relationship to EEG discontinuity on both day 1 and day 3 in this large cohort of infants. In contrast to other studies, our study was prospective with all infants having invasive BP monitoring, with continuous BP data being extracted every 10 s for the first week. West et al. (3) reported BP (non-invasive and invasive) data, acquired every minute, from 40 preterm infants at 12 and

Table 2 | Mixed effects multiple regression analysis between mean EEG discontinuity and clinical parameters.


24 h to be related to aEEG continuity at 12 and 24 h after birth. Infants in the lowest quartile for BP, which was below 31 mmHg, had lower aEEG continuity. EEG abnormalities are predictive of adverse long-term neurodevelopment in this group of infants (33). Victor et al. (9) showed EEG continuity to be normal in infants whose mean BP was above 30 mmHg. Our study has shown that increasing BP was associated with increased cerebral electrical activity.

There are several possible mechanisms by which EEG discontinuity could increase with lower BP levels, before reductions in cerebral perfusion affect cellular energy status. This could be postulated to be part of an intrinsic cerebral protective response to hypotension, with lower electrocortical activity reducing neuronal oxygen demand. In response to hypoxia–ischaemia, a neuroprotective adenosine mediated suppression of EEG has been reported in animal models (34, 35).

In our study, there was no consistent effect of blood flow parameters on EEG discontinuity that other studies have previously reported (8, 14). CCAF was not related to EEG discontinuity both on days 1 and 3. This would suggest that the relationship between systemic BP and EEG discontinuity is not simply mediated by alterations in systemic blood flow transmitted to the cerebral circulation. The median LVO measures were comparable to previously published data from hypotensive preterm neonates, but slightly lower than in more mature babies with higher BP and gestation (36).

Limitations of this study include that cerebral scanning and cardiac blood flow measurements were only carried out as a single measurement on one occasion on days 1 and 3 in the majority of infants.

#### CONCLUSION

Our study suggests that continuously measured invasive mean arterial BP is negatively associated with EEG discontinuity and

#### REFERENCES


this does not appear to be mediated by systemic or cerebral blood flow parameters. Infants receiving inotropic support on the first day after birth had increased EEG discontinuity in comparison with those not receiving such support.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Research Ethics Committee (reference 12/LO/1553) with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the London-Surrey Borders Research Ethics Committee and NHS National Research Ethics Service.

#### AUTHOR CONTRIBUTIONS

SP: participated in conceptualising and designing the study, designed the data collection sheets, acquired and performed initial analysis of the data, drafted the initial manuscript, and approved the final manuscript as submitted. SK: conceptualised and designed the study, performed analysis of the data, reviewed and revised the manuscript, and approved the final manuscript as submitted. DW: designed the study, analysed the electroencephalograms, reviewed the manuscript, and approved the final manuscript as submitted. AS and DS: designed the study, performed analysis of the data, critically reviewed and revised the manuscript, and approved the final manuscript as submitted. JM: performed statistical analysis of the data, critically reviewed and revised the manuscript, and approved the final manuscript as submitted. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

#### ACKNOWLEDGMENTS

We thank the parents and infants who participated in this study and also all the doctors and nursing staff of the neonatal intensive care unit at the Royal London Hospital.

#### FUNDING

This study was supported by fundraising by parents of our neonatal patients, administered by Barts Charity (registered charity number 212563). Grant reference number 420/2189.


6. Granot S, Meledin I, Richardson J, Friger M, Shany E. Influence of respiratory acidosis and blood glucose on cerebral activity of premature infants. *Pediatr Neurol* (2012) 47(1):19–24. doi:10.1016/j.pediatrneurol.2012.03.018

7. Wikström S, Lundin F, Ley D, Pupp IH, Fellman V, Rosén I, et al. Carbon dioxide and glucose affect electrocortical background in extremely preterm infants. *Pediatrics* (2011) 127(4):e1028–34. doi:10.1542/peds.2010-2755

8. Shah D, Paradisis M, Bowen JR. Relationship between systemic blood flow, blood pressure, inotropes, and aEEG in the first 48h of life in extremely preterm infants. *Pediatr Res* (2013) 74:314–20. doi:10.1038/pr.2013.104


maturation aspects. *Neurophysiol Clin* (2007) 37(5):311–23. doi:10.1016/j. neucli.2007.10.008


**Conflict of Interest Statement:** David Wertheim is an inventor on a patent US5181520 "Method and apparatus for analysing an electro-encephalogram." The other authors have no conflicts of interest to disclose.

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

*Lei Xia1†, Mingjie Chen2†, Dan Bi1 , Juan Song1 , Xiaoli Zhang1 , Yangong Wang2 , Dengna Zhu1 , Qing Shang3 , Falin Xu1 , Xiaoyang Wang1,4, Qinghe Xing2,5\* and Changlian Zhu1,6\**

*1Henan Key Laboratory of Child Brain Injury, Third Affiliated Hospital of Zhengzhou University, Zhengzhou, China, 2 Institute of Biomedical Science, Children's Hospital, Fudan University, Shanghai, China, 3Department of Pediatrics, Zhengzhou Children's Hospital, Zhengzhou, China, 4Perinatal Center, Institute of Neuroscience and Physiology, University of Gothenburg, Gothenburg, Sweden, 5Shanghai Center for Women and Children's Health, Shanghai, China, 6Center for Brain Repair and Rehabilitation, Institute of Neuroscience and Physiology, University of Gothenburg, Gothenburg, Sweden*

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Brahim Tabarki, University of Sousse, Tunisia Hiroko Morisaki, Sakakibara Heart Institute, Japan*

#### *\*Correspondence:*

*Qinghe Xing xingqinghe@hotmail.com; Changlian Zhu zhuc@zzu.edu.cn*

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

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 13 November 2017 Accepted: 08 March 2018 Published: 22 March 2018*

#### *Citation:*

*Xia L, Chen M, Bi D, Song J, Zhang X, Wang Y, Zhu D, Shang Q, Xu F, Wang X, Xing Q and Zhu C (2018) Combined Analysis of Interleukin-10 Gene Polymorphisms and Protein Expression in Children With Cerebral Palsy. Front. Neurol. 9:182. doi: 10.3389/fneur.2018.00182*

Background: Interleukin-10 (IL-10) is an important anti-inflammatory and immunosuppressive cytokine, and it has indispensable functions in both the onset and development of inflammatory disorders. The association between persistent inflammation and the development of cerebral palsy (CP) has attracted much attention.

Objective: The purpose of this study was to investigate whether *IL-10* gene polymorphisms and plasma protein expression are associated with CP and to analyze the role of IL-10 in CP.

Methods: A total of 282 CP patients and 197 healthy controls were genotyped for *IL-10* polymorphisms (rs1554286, rs1518111, rs3024490, rs1800871, and rs1800896). Among them, 95 CP patients and 93 healthy controls were selected for plasma IL-10 measurement.

results: The differences in the rs3024490 (*p* = 0.033) and rs1800871 (*p* = 0.033) allele frequencies of *IL-10* were determined between CP patients and controls. The frequencies of allele and genotype between CP patients with spastic tetraplegia and normal controls of *IL-10* polymorphisms showed significant differences for rs1554286, rs151811, rs3024490, rs1800871, and rs1800896 (*pallele* = 0.015, 0.009, 0.006, 0.003, and 0.006, *pgenotype* = 0.039, 0.018, 0.027, 0.012, and 0.03, respectively). The plasma IL-10 protein level in CP patients was higher than normal controls (9.13 ± 0.77 vs*.* 6.73 ± 0.63 pg/ml, *p* = 0.017). IL-10 polymorphisms and protein association analysis showed that the TT genotype had higher plasma IL-10 protein levels compared to the GG + GT genotype at rs3024490 (11.14 ± 7.27 vs. 7.44 ± 6.95 pg/ml, *p* = 0.045, respectively) in CP cases.

conclusion: These findings provide an important contribution toward explaining the pleiotropic role of IL-10 in the complex etiology of CP.

Keywords: cerebral palsy, cytokine, inflammation, interleukin-10, single nucleotide polymorphisms

#### INTRODUCTION

Cerebral palsy (CP) comprises a group of disorders affecting movement and posture due to non-progressive lesions or abnormalities in the immature brain, and these can lead to costly and life-long disability (1, 2). Studies have identified a number of risk factors for the development of CP. During the last decade, intrauterine infection and inflammation have been recognized as the most common causes of preterm delivery and white-matter brain injury such as periventricular leukomalacia (PVL) and subsequent development of CP (3–6). It is also known that the activation of inflammation plays a key role in the mechanisms of intrauterine infection-triggered brain damage (7, 8).

It is well known that the pro-inflammatory cytokine level in the amniotic fluid or in neonatal blood is directly related to the risk of developing CP (3, 9), and high levels of expression of TNFα, IL-1β, and IL-6 increase the risk for development of brain lesions with PVL (7, 10). IL-17 has also been reported to act synergistically with TNF and IL-1 and to play a notable role in inducing and mediating pro-inflammatory responses (11–13). These pro-inflammatory cytokines are responsible for initiating inflammation in response to tissue damage, while anti-inflammatory cytokines are released to limit the sustained or excessive inflammatory response (14). IL-10 is a main anti-inflammatory cytokine and has been suggested to play a crucial role in neuronal homeostasis and cell survival (15, 16). IL-10 plays a protective role in microglial cultures after a pro-inflammatory insult (17) and in rat pups born to dams infected with *Escherichia coli* (18). However, several studies have questioned the perception of IL-10 solely as an immunosuppressive cytokine because IL-10 can also stimulate immune responses by promoting the proliferation and cytotoxic activity of natural killer cells and CD8+ T-cells (19) as well as the survival, proliferation, differentiation, MHC class II expression, and antibody production of B-cells (20, 21). Increasing evidence indicates that IL-10 is involved in both the onset and development of inflammatory diseases. In a study of very low birth weight children suspected of having late-onset sepsis, the expression of the anti-inflammatory cytokines IL-4 and IL-10 was significantly elevated (22). Together, these studies indicate that the biological activity of IL-10 is variable and results in protective or pathogenic effects at different stages of disease and that IL-10 might play an important role in the pathogenesis of CP.

The protein expression of cytokines is regulated by genetic variants in the cytokine genes (15, 23), and accumulating evidence suggests that inherited cytokine single nucleotide polymorphisms (SNPs) contribute to increased risk of CP (24, 25). In following with this, it has been reported that SNPs within the coding and promoter regions of the *IL-10* gene can affect the expression and secretion of this cytokine (26, 27). Considering the potential role of *IL-10* in the etiology of CP, the present research sought to evaluate the possible association of *IL-10* SNPs and IL-10 plasma levels with susceptibility to CP in a Chinese population.

#### MATERIALS AND METHODS

#### Subjects

A total of 282 CP patients and 197 healthy controls were enrolled (**Table 1**). We collected all CP patients between 1 July 2010 and 31 May 2012 from the 3rd Affiliated Hospital of Zhengzhou University and Zhengzhou Children's Hospital, which included 98 girls (34.8%) and 184 boys (65.2%) with a mean age ± SD of 16.2 ± 12.7 months. The diagnosis of CP patients was made by child neurologists according to the guidelines proposed by the Surveillance of CP in Europe network (28) through clinical examination or medical records, including brain imaging. The healthy controls [42 girls (21.3%) and 155 boys (78.7%) with a mean age ± SD of 24.0 ± 16.4 months] were enrolled during their physical examination from the same hospitals. Of these, 95 CP patients (56 boys and 39 girls) and 93 healthy controls (78 boys and 15 girls) were selected for the assay of IL-10 protein level (**Table 2**), and the mean age ± SD was 20.8 ± 14.4 and 21.6 ± 13.8 months, respectively. All subjects were Han Chinese as reported by their parents. All of the subjects, including both CP patients and controls, with myopathy, metabolic anomalies, or infections were excluded because of the genetic and familial factors that are associated with CP. Approval for the study was obtained from the ethics committee of Zhengzhou University in accordance with the Helsinki Declaration (201002006). Written

Table 1 | Sample description for gene polymorphism analysis.


*CP, cerebral palsy; M, male; F, female; HIE, hypoxia ischemia encephalopathy; PVL, periventricular leukomalacia; MR, mental retardation; PROM, premature rupture of membrane; TPL, threatened premature labor; PIH, pregnancy-induced hypertension.*



*CP, cerebral palsy; M, male; F, female; HIE, hypoxia ischemia encephalopathy; PVL, periventricular leukomalacia; MR, mental retardation; PROM, premature rupture of membrane; TPL, threatened premature labor; PIH, pregnancy-induced hypertension.*

informed consent was obtained from at least one of the parents after fully explaining the procedure.

The clinical data were stratified by the type of CP, birth asphyxia, and complications such as PVL and hypoxic–ischemic encephalopathy. The criteria for these risk factors were as previously described (28).

#### Sample Collection

The blood samples were taken by skilled nurses. EDTA was routinely used as the anti-coagulant in the study. The samples were separated by centrifugation (1,500 × *g* for 15 min) at 21°C within 2 h after being collected. The plasma was obtained from the supernatant, and DNA was obtained from the remaining blood components in the same sample. All the fractions were stored at −80°C until use. All methods were performed in accordance with the relevant guidelines and regulations.

#### Polymorphism Selection

Five SNPs (rs1554286, rs1518111, rs3024490, rs1800871, and rs1800896) of the *IL-10* gene whose minor allele frequencies in the Chinese Han population are more than 0.1 were selected from the dbSNP database1 and the HapMap human SNP database.2 The rs1554286 (intron 3), rs1518111 (intron 2), and rs3024490 (intron 1) SNPs are located in the coding region of *IL-10*, and the rs1800871 and rs1800896 SNPs are located in the upstream promoter region.

#### Genotyping

After collecting the plasma, the AxyPrep Blood Genomic DNA Miniprep Kit (Axygen Biosciences, Union City, CA, USA) was used for preparing genomic DNA from the remaining blood components by following the manufacturer's protocol. After the amplification of SNP-spanning fragments by multiplex PCR, the SNPs were genotyped on the Sequenom MassARRAY SNP genotyping platform (Sequenom, San Diego, CA, USA). The probes and primers were designed by the SEQUENOM online tools.3 The genotype results were analyzed by a person who was blinded to the clinical data.

#### Cytokine Protein Level Measurement

The plasma samples for IL-10 protein analysis were thawed completely at room temperature, mixed by vortexing, and centrifuged at 1,500 × *g* for 15 min to collect the supernatants for the subsequent cytokine assay. A Milliplex Human Cytokine/ Chemokine kit (IL-10, IL-17 IL-6, IL-8, TNF-α, and IFN-γ) was used with the Multiplex Cytokine Assay (Millipore, Billerica, MA, USA). Quality controls were performed in parallel between the plates using reagents provided in the kits. Data were acquired on a Luminex 200IS System (Luminex Corporation, Austin, TX, USA). There was a 6.94% plate variation in this assay, and the detection limit was 0.3 pg/mL. The samples with cytokine levels below the limit of detection were excluded. IL-10 levels were expressed as picograms per milliliter.

#### Statistical Analysis

All gene analyses, including Hardy–Weinberg equilibrium tests, comparison of allele and genotype frequency, calculation of odd ratios and 95% confidence interval (95% CIs), estimation of pairwise linkage disequilibrium (LD), and haplotype association analysis were conducted with the SHEsis online software platform.4 The program SNPSpD,5 which takes marker LD information into consideration, was used to correct for multiple testing performed on each individual SNP. For IL-10 analysis, Student's unpaired *t*-test was used. The Mann–Whitney *U*-test was used for the data with unequal variances. Statistical analyses were performed with SPSS (version 19.0) and Graphpad Prism 6.0 (Graphpad, La Jolla, CA, USA). All reported *p*-values were two-tailed, and statistical significance was set at *p* < 0.05.

#### RESULTS

#### Association of *IL-10* Gene Polymorphisms With the Total Group With CP

To examine the association between CP and the polymorphisms in the *IL-10* gene, we detected genotype and allele frequencies of five SNPs in the 282 CP patients and the 197 healthy controls. The frequencies and analytic results for the SNPs are presented in **Table 3**. The genotypic distribution of the five selected SNPs of *IL-10* in the normal controls was in Hardy–Weinberg equilibrium. In the analysis of allele frequencies, differences between

<sup>1</sup>www.ncbi.nlm.nih.gov/SNP (accessed December 18, 2014).

<sup>2</sup>www.hapmap.org (accessed March 10, 2015).

<sup>3</sup>http://agenabio.com/ (accessed June 5, 2015).

<sup>4</sup>http://analysis.bio-x.cn/ (accessed September 20, 2015).

<sup>5</sup>http://gump.qimr.edu.au/general/daleN/matSpD/ (accessed December 15, 2015)


*p*′*-Value, the p-value after the SNPSpD correction; The significance was set at p*′*<0.05 after correction; OR, odds ratio; CI, confidence interval.*

the total CP patients (*n* = 282) and normal controls (*n* = 197) for rs3024490 (*p* = 0.011, after SNPSpD correction, *p* = 0.033) and rs1800871 (*p* = 0.011, after SNPSpD correction, *p* = 0.033) were observed. There was also strong LD between rs3024490 and rs1800871 (*r*<sup>2</sup> = 0.955) (**Table 4**), which makes it reasonable to have the same *p*-value for rs3024490 and rs1800871.

#### Association of *IL-10* Gene Polymorphisms With the Subgroups of CP

Because CP is a complex syndrome with multiple etiological factors involved in disease pathogenesis, we performed subgroup analysis of *IL-10* SNPs according to subtypes of CP, birth complications, and maternal factors. The frequencies of allele and genotype were different between CP patients with spastic tetraplegia (*n* = 123) and normal controls (*n* = 197) for rs1554286, rs151811, rs3024490, rs1800871, and rs1800896 (*pallele* = 0.015, 0.009, 0.006, 0.003, and 0.006, respectively, after the SNPSpD correction), but no other significant differences were found when comparing allele and genotype distributions of these five SNPs between other CP subgroups and controls (**Table 5**).

#### Cytokine Analysis

The plasma IL-10 protein level was increased significantly in CP patients compared to normal controls (9.13 ± 7.07 vs. 6.79 ± 5.63 pg/ml, respectively, *p* = 0.017) (**Figure 1A**). The subgroup analysis showed that the plasma IL-10 level was also increased in the CP patients with spastic tetraplegia compared to the controls (9.86 ± 6.00 vs. 6.79 ± 5.63 pg/ml, respectively, *p* = 0.044) (**Figure 1B**). There were no significant differences in plasma IL-10 protein levels in other subgroup analyses.

#### Association of the SNPs With Cytokine Production

To confirm whether the IL-10 gene is linked with CP susceptibility, we further analyzed the relationship between genotypes of two SNPs associated with CP and plasma IL-10 protein levels. The results showed that the TT genotype had higher plasma IL-10 protein levels compared to the GG + GT genotype at rs3024490 Table 4 | The linkage disequilibrium among the SNPs in *IL-10.*


*The standardized D*′ *values are shown above the diagonal, and the r2 values are shown below the diagonal.*

(11.14 ± 7.27 vs. 7.44 ± 6.95 pg/ml, respectively, *p* = 0.045) (**Figure 2A**), and rs1800871 showed a tendency of high IL-10 protein level in the TT genotype compared to the CC + CT genotype (10.82 ± 7.33 vs. 7.62 ± 6.96 pg/ml, respectively, *p* = 0.079) in CP cases (**Figure 2B**), whereas no association was found for either SNP in controls. Although IL-10 is produced by constitutive or inducible expression, these results indicated that the genotypes of rs3024490 might play a more important role in the induced expression of IL-10.

#### DISCUSSION

It has been accepted that inflammation plays an important role in the brain during the perinatal period and might contribute to the development of CP (23). Several studies have reported that increased pro-inflammatory cytokines such as IL-1β, IL-6, TNFα, and IL-8 in cord blood or amniotic fluid are associated with CP (23, 29), and IL-6 protein levels are higher in the blood (28) and cerebrospinal fluid of CP patients (30). Moreover, the balance between the levels of pro-inflammatory and anti-inflammatory cytokines determines the effect of the inflammatory responses.

IL-10 function is complicated and can have both stimulatory and inhibitory effects on different types of immune responses. IL-10 can strongly inhibit the secretion of IL-2/IFN-γ by TH1 cells, and it inhibits the secretion of IL-1, IL-6, IL-12, and tumor necrosis factor in macrophages and dendritic cells in order to reduce tissue damage (31, 32). IL-10, as a potential anti-inflammatory cytokine, creates favorable conditions for the persistence


*p΄-Value, the p-value after the SNPSpD correction; The significance was set at p΄<0.05 after correction; OR, odds ratio; CI, confidence interval.*

Figure 1 | IL-10 concentration in cerebral palsy (CP) patients and controls. (A) The scatter plot of plasma IL-10 levels in CP patients and controls. (B) The scatter plot of plasma IL-10 in tetraplegia CP patients and controls (\**p* < 0.05).

of microbes and chronic diseases (33), both of which are involved in clinical perinatal brain damage (15). It has been reported that the upregulation of IL-10 after injury might have antiinflammatory effects in specific regions of the immature brain in the postnatal rat brain infection model (34), and anatomical characteristics have been associated with functional findings in athetotic-type and spastic-type CP using an atlas-based analysis (35). In the current study, we found that IL-10 protein levels were increased in CP patients, especially in spastic tetraplegia, the most severe subtype of CP patients. The predominantly negative immune regulatory functions of IL-10 can play a critical role in mediating T-cell functional exhaustion, which makes these cells detrimental in chronic infection (36). Prolonged microglial reactivity and increased cytokine expression have been noted in animal models of traumatic brain injury, and altered inflammatory responses have been shown to persist for at least 7 years after brain damage in CP patients (37). This suggests that one reason for why the plasma IL-10 level increases in CP patients is the inability for the body to clear antigen, which leads to repetitive antigen stimulation and IL-10 induction in a manner similar to chronic infection. In other words, the high plasma IL-10 level might be an attempt by the immune response to counterbalance the expression of pro-inflammatory cytokines (38). Our present results support the hypothesis that inflammation triggers an inadequate immunological response in preterm infants with a consequently increased risk of CP. The cytokines, as indicators of inflammation, usually have a half-life of only a few hours, whereas previous studies (3, 7, 9, 10, 23, 29) and our study have shown persistent inflammation in CP patients. Therefore, it is possible that the inflammatory process in the brain is continuously being activated after perinatal brain injury, and future studies should focus on the dynamic changes in neuroinflammation and how these changes relate to the progression of CP.

Gene polymorphism studies suggest the involvement of as yet unidentified linkages between allelic variants and the pathogenesis of disease (39), and SNPs in genes encoding cytokines and their receptors have been implicated in both increased and decreased risk of perinatal brain injury (23). Studies have suggested that *IL-6*, *IL-8*, and *TNF-α* SNPs (40, 41) might predispose to CP because an increased risk for CP has been shown to be positively associated with the increased production of these proteins, but nothing has been known about the relationship between *IL-10* SNPs and CP.

The human *IL-10* gene is located on chromosome 1q31–1q32 and is composed of four introns and five exons (42). Three SNPs in the *IL-10* intron region (rs1554286, rs1518111, and rs3024490) have been reported to be associated with risk for ischemic stroke and idiopathic recurrent miscarriage (43–45), and two SNPs in the *IL-10* promoter region (rs1800871 and rs1800896) have been identified as highly polymorphic risk factors for systemic sclerosis (46) and Alzheimer's disease (47). These five SNPs are in strong LD, and this is believed to have biological significance (39). In the current study, all five SNPs were associated with tetraplegia CP patients. In addition, reports have also shown that the upregulation of IL-10 after injury might have anti-inflammatory effects in distinct anatomic sites in the postnatal rat brain infection model (34), and anatomical characteristics have been delineated with functional findings in athetotic-type and spastic-type CP using the atlas-based analysis (35). In this study, we found the TT genotype frequency in the CP group to be higher than the control group and the TT genotype of rs3024490 and rs1800871 seems to be related to higher circulating levels of IL-10 protein in CP patients but not in controls. These results suggest that the genotype TT of rs3024490 associated an increased IL-10 expression might be due to the different reactivity of genetic variants to risk factors for CP (48, 49). Therefore, we hypothesize that cytokines affect different regions of the developing brain and that the subtypes of CP are determined by the functional genotypes of *IL-10* in the development of CP. Thus this group of children should be paid more attention to, especially in early life before clinical symptoms of CP begin to appear, which is exactly where the significance of the present study lies. However, we need to expand the sample size in subsequent experiments in order to test this hypothesis.

Cerebral palsy is a heterogeneous disease with complex interactions between genetic influences and environmental influences. Understanding the genes and gene–environment interactions in CP and the underlying cytokine-related mechanisms might lead to the development of new preventive and therapeutic strategies for CP. In summary, this study suggests

#### REFERENCES


that *IL-10* SNPs are strongly associated with CP through the regulation of *IL-10* gene function that leads to altered IL-10 production. Future work should focus on the interactions between genetic and cytokine-related mechanisms in the pathogenesis of CP. A better understanding of the potentially beneficial and detrimental roles of IL-10 in CP might not only help improve our understanding of CP pathogenesis, but might also help develop novel strategies for the prevention of CP.

#### ETHICS STATEMENT

Approval for the study was obtained from the ethics committee of Zhengzhou University (201002006) in accordance with the Helsinki declaration.

#### AUTHOR CONTRIBUTIONS

CZ, QX, and XW conceived the project. LX, DB, MC, JS, XZ, YW, DZ, QS, and FX conducted the experiments. DB, LX, XW, QX, and CZ wrote the manuscript. All of the authors discussed the results and commented on the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (grants 31611130035, U1604165 U1704281, 81771418), VINNMER–Marie Curie international qualification (VINNOVA, 2015-04780), the Swedish Research Council (VR 2015-06276), Swedish governmental grants to researchers in the public health service (ALFGBG-717791), the Fourth Round of Shanghai Three-year Action Plan on Public Health Discipline and Talent Program: Women and Children's Health (No. 15GWZK0401), the 973 Projects (2011CB710801), and the Shanghai Municipal Commission of Science and Technology Program (14DJ1400101).


after an excitotoxic injury to the postnatal rat brain. *J Neuropathol Exp Neurol* (2009) 68(4):391–403. doi:10.1097/NEN.0b013e31819dca30


**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 Xia, Chen, Bi, Song, Zhang, Wang, Zhu, Shang, Xu, Wang, Xing and Zhu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner 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.*

# Cerebral Lactate Concentration in Neonatal Hypoxic-Ischemic Encephalopathy: In Relation to Time, Characteristic of Injury, and Serum Lactate Concentration

*Tai-Wei Wu1 \*, Benita Tamrazi2 , Kai-Hsiang Hsu1,3, Eugenia Ho4 , Aaron J. Reitman5 , Matthew Borzage1 , Stefan Blüml2,6 and Jessica L. Wisnowski1,2,6*

*1Department of Pediatrics, Keck School of Medicine, Fetal and Neonatal Institute, Division of Neonatology, Children's Hospital Los Angeles, University of Southern California, Los Angeles, CA, United States, 2Department of Radiology, Children's Hospital Los Angeles, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States, 3Division of Neonatology, Department of Pediatrics, Chang Gung Memorial Hospital Linkou Branch, Taoyuan, Taiwan, 4Department of Neurology, Children's Hospital Los Angeles, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States, 5Division of Neonatology, Department of Pediatrics, LAC* + *USC Medical Center, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States, 6Rudi Schulte Research Institute, Santa Barbara, CA, United States*

#### *Edited by:*

*Masahiro Tsuji, National Cerebral and Cardiovascular Center, Japan*

#### *Reviewed by:*

*Tomoki Arichi, King's College London, United Kingdom Eric S. Peeples, University of Nebraska Medical Center, United States*

> *\*Correspondence: Tai-Wei Wu twu@chla.usc.edu*

#### *Specialty section:*

*This article was submitted to Pediatric Neurology, a section of the journal Frontiers in Neurology*

*Received: 07 February 2018 Accepted: 16 April 2018 Published: 11 May 2018*

#### *Citation:*

*Wu T-W, Tamrazi B, Hsu K-H, Ho E, Reitman AJ, Borzage M, Blüml S and Wisnowski JL (2018) Cerebral Lactate Concentration in Neonatal Hypoxic-Ischemic Encephalopathy: In Relation to Time, Characteristic of Injury, and Serum Lactate Concentration. Front. Neurol. 9:293. doi: 10.3389/fneur.2018.00293*

Background: Cerebral lactate concentration can remain detectable in neonatal hypoxic-ischemic encephalopathy (HIE) after hemodynamic stability. The temporal resolution of regional cerebral lactate concentration in relation to the severity or area of injury is unclear. Furthermore, the interplay between serum and cerebral lactate in neonatal HIE has not been well defined. The study aims to describe cerebral lactate concentration in neonatal HIE in relation to time, injury, and serum lactate.

Design/methods: Fifty-two newborns with HIE undergoing therapeutic hypothermia (TH) were enrolled. Magnetic resonance imaging and spectroscopy (MRI + MR spectroscopy) were performed during and after TH at 54.6 ± 15.0 and 156 ± 57.6 h of life, respectively. Severity and predominant pattern of injury was scored radiographically. Single-voxel 1 H MR spectra were acquired using short-echo (35 ms) PRESS sequence localized to the basal ganglia (BG), thalamus (Thal), gray matter (GM), and white matter. Cerebral lactate concentration was quantified by LCModel software. Serum and cerebral lactate concentrations were plotted based on age at time of measurement. Multiple comparisons of regional cerebral lactate concentration based on severity and predominant pattern of injury were performed. Spearman's Rho was computed to determine correlation between serum lactate and cerebral lactate concentration at the respective regions of interest.

Results: Overall, serum lactate concentration decreased over time. Cerebral lactate concentration remained low for less severe injury and decreased over time for more severe injury. Cerebral lactate remained detectable even after TH. During TH, there was a significant higher concentration of cerebral lactate at the areas of injury and also when injury was more severe. However, these differences were no longer observed after TH. There was a weak correlation between serum lactate and cerebral lactate concentration at the BG (*r*<sup>s</sup> = 0.3, *p* = 0.04) and Thal (*r*<sup>s</sup> = 0.35, *p* = 0.02). However, in infants with moderate–severe brain injury, a very strong correlation exists between

**307**

serum lactate and cerebral lactate concentration at the BG (*r*<sup>s</sup> = 0.7, *p* = 0.03), Thal (*r*<sup>s</sup> = 0.9 *p* = 0.001), and GM (*r*<sup>s</sup> = 0.6, *p* = 0.04) regions.

Conclusion: Cerebral lactate is most significantly different between regions and severity of injury during TH. There is a moderate correlation between serum and cerebral lactate concentration measured in the deep gray nuclei during TH. Differences in injury and altered regional cerebral metabolism may account for these differences.

Keywords: lactate, cerebral lactate, magnetic resonance spectroscopy, neonatal asphyxia, hypoxic-ischemic encephalopathy

### INTRODUCTION

Neonatal hypoxic-ischemic encephalopathy (HIE) affects about 1–6 per 1,000 live births (1). Therapeutic hypothermia (TH) is the first empirically supported therapy for neuroprotection in neonates with HIE. Still, even with TH, 40–50% continue to suffer adverse outcomes (2, 3). The pathophysiology of HIE involves a complex cascade of cellular and molecular processes, which centers on a decline in oxidative metabolism (4, 5). The goal of TH is to mitigate the ischemic cascade and restore energy homeostasis.

Energy metabolism can be monitored *in vivo* in brain tissue by way of MR spectroscopy (MRS). Serial MRS has been employed extensively in laboratory studies to interrogate temporal changes in the concentrations of key metabolites, including lactate, a byproduct of anaerobic metabolism and a biomarker for impaired oxidative metabolism (6). However, MRS is technically demanding in the clinical setting and even more so when performed during TH (7). For this reason, very little is known about the temporal evolution of cerebral lactate concentrations during and after TH in neonates with HIE. Furthermore, neuroimaging studies have established two different patterns of brain injury in neonates with HIE: a central pattern, characterized by injury to the deep gray [thalamus, basal ganglia (BG)] and perirolandic cortex, and a peripheral pattern, characterized by injury to the parasagittal cortex and underlying white matter (WM) (8, 9). It is not known whether regional cerebral lactate concentrations mirror the pattern or severity of brain injury during and/or after TH.

Finally, it should be noted that cerebral lactate may be locally produced in brain tissue or transported into and out of the brain *via* the blood–brain barrier (10). Moreover, the neonatal brain is characterized by a high capacity for lactate transport into and out of the brain (11, 12). Elevations in serum lactate are ubiquitous at birth in neonates with HIE and may even transiently increase before normalizing in the hours to days that follow (13, 14). Furthermore, prolonged elevation in serum lactate is associated with severe encephalopathy and seizure burden (15). Considering the high capacity for lactate transport into the neonatal brain and the underlying association between serum lactate and brain injury, it is important to assess the relation between serum lactate and cerebral lactate in neonates with HIE.

The objectives of this descriptive study were: (1) to determine the temporal resolution of serum and cerebral lactate concentrations in neonates with HIE; (2) to determine whether regional cerebral lactate concentration differs with respect to injury severity or the predominant pattern of brain injury (normal, peripheral/ watershed, central/deep gray); and (3) to explore the relationship between serum lactate and cerebral lactate in neonates with HIE undergoing TH. We hypothesize that cerebral lactate is elevated specifically in regions of injury and decreases with time after injury. Furthermore, we hypothesize that serum lactate is more strongly associated with cerebral lactate concentrations in regions of high metabolic demand and marked susceptibility to acute injury in neonates with HIE.

#### MATERIALS AND METHODS

Infants diagnosed with HIE who were admitted to the Children's Hospital Los Angeles Newborn and Infant Critical Care unit for TH were prospectively enrolled into an observational research study from April 2012 to July 2017. After obtaining written parental permission, patients underwent a research MRI during TH, as well as a standard clinical MRI after re-warming. Serum lactate concentrations and other clinical data were abstracted from medical records. Neurodevelopmental follow-up is ongoing. The study was approved by the Institutional Review Board at Children's Hospital Los Angeles.

#### Patient Sample

Inclusion criteria for TH at our institution is as follows: gestational age of at least 35 weeks, birthweight >1,800 g, admitted within 6 h of age with the history of cord blood gas or first hour blood gas of a pH of ≤7.0 or a base deficit of ≥16 mmol/L. In cases where pH fell between 7.01 and 7.15 or the base deficit was between 10 and 15.9 mmol/L, the following additional criteria must be met: (1) history of an acute perinatal event and (2) a 10-min Apgar score of ≤5 or the need for assisted ventilation at birth for more than a 10-min duration. Patients who met the above criteria underwent TH if they had moderate or severe encephalopathy based on the modified Sarnat score (16), or if they presented with clinical seizures. Whole-body TH was initiated within 6 h of age and a target rectal temperature (33.5–34.5°C) was maintained for 72 h by Cincinnati Sub-Zero Blanketrol III (Gentherm, Cincinnati, OH, USA). At the end of 72 h of whole-body TH, the infants were actively rewarmed to normothermia by incremental increase of rectal temperature at a velocity of 0.5°C per hour up to 36.5 C.

Infants with congenital anomalies, metabolic disorders, early onset sepsis, or perinatal stroke have been retrospectively excluded from all studies from this cohort. Furthermore, we excluded infants who were clinically unstable to undergo research MRI during TH.

### Serum Lactate Collection

Serum lactate concentration was measured every 6–8 h from initiation of TH until completion of rewarming, based on unit TH protocol. Blood samples were obtained from free flowing arterial lines and lactate concentration analyzed as point of care testing using the Epoc® blood analysis system (Alere Inc., Waltham, MA, USA), which has a measurement range of 2.7–180.2 mg/dL. When serum lactate concentrations were found to be greater than the measurable range of 180.2 mg/dL, the value of 180 mg/dL was used for data analysis. For the final analyses relating serum lactate to brain lactate, we focused on serum lactate measured at two time points: on admission (Lactate 1) and at time of first MRI scan (Lactate 2). Mean time-lapse between Lactate 2 sampling and MRS acquisition was 2.9 ± 2.0 h.

### Magnetic Resonance Imaging and Spectroscopy

All patients underwent MRI during TH and after re-warming. TH was maintained for the duration of transport to the MRI utilizing an external battery pack (7). During the MRI, extension tubing was passed through the waveguide and used to connect the MRI-compatible cooling blanket to the Blanketrol, which remain docked in the MRI control room. Rectal temperature was monitored using an MRI-compatible temperature probe (Philips Medical, Best, The Netherlands) and blanket temperature was manually adjusted to maintain core body temperature within therapeutic range (33.5 ± 0.5°C). There were no adverse events associated with transport or MRI.

As standard clinical practice, a series of standardized MR images (T1-, T2-, and diffusion-weighted) were acquired and reviewed by a board-certified pediatric neuroradiologist for brain injury (see below). MRS data were acquired using a single-voxel pointresolved spectroscopy sequence (PRESS; TE 35 ms, TR 2,000 ms, 128 signal averages, voxel size ~3 cm3 ), localized to the right BG, left thalamus, medial cortical gray matter (GM), and left parietal WM based on axial, sagittal, and coronal T2-weighed images (note that the WM voxel was added after the 10th neonate). The regions of interest were selected *a priori* because they are known areas of vulnerability to hypoxia-ischemia in neonates and corresponded to the central (thalamus/BG/perirolandic cortex) and peripheral/ watershed (WM) patterns of brain injury in neonatal HIE (9).

Furthermore, the rationality for the left thalamus and right BG stemmed from our desire to focus on the dominant hemisphere with regard to language and motor functions (left THAL) while also obtaining data from both hemispheres (right BG). Lactate was quantitated from each MRS voxel using LCModel (V6.3-1L) (17), consistent with prior studies in our laboratory (18). For quantitation, the unsuppressed water signal was used as a concentration reference, with tissue water content estimated at a standardized value (86%, i.e., 47.8M) based on published reference data (19). To ensure reliability, spectra of poor quality were excluded *a priori* based on stringent thresholds for line width (<0.05 ppm, i.e., <6.4 Hz) and signal to noise ratio (≥10). Spectra were not excluded based on Cramer–Rao lower bound as this would have biased results toward higher lactate concentrations.

### MRI Injury Classification

A pediatric neuroradiologist (Benita Tamrazi) blinded to clinical course and outcome of the subjects, classified patients with regard to the degree of injury and predominant pattern as observed on the post-cooling MRI, using a previously described scoring system (9). This system relies on acute and subacute signal abnormalities in the BG/thalamus (BG-T) region (score 0–4) and watershed region (score 0–5). Based on the scores, each patient was classified with regard to: (1) pattern of injury (normal, BG-T, or watershed pattern) and (2) severity of injury [normal to mild (nml/mild) or moderate to severe (mod/severe)]. The post-cooling MRI was scored instead of the on-cooling MRI in order to ensure highest sensitivity to detection of injury, as diffusion changes may not become fully apparent during the first 24–48 h of injury (20, 21).

## Data Analysis

Statistical analysis was performed using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA, USA). Data normality was tested using the D'Agostino & Pearson omnibus normality test. Data are presented as mean ± SD or median (interquartile range, IQR) depending on normality. Both cerebral and serum lactate concentration were plotted according to age at time of scan. Paired *t*-test or non-parametric Wilcoxon matched-pairs signed rank test was used to compare respective regional cerebral lactate concentration during vs. after TH. In order to maintain a fair comparison, we excluded cerebral lactate values if the second MRS was performed after 7 days of age. One-way ANOVA or Kruskal–Wallis test with Tukey's or Dunn's multiple comparisons test was used to compare regional cerebral lactate concentrations in nml-mild vs. moderate–severe brain injury and in BG-T vs. watershed pattern of injury. Non-parametric Spearman rank-order correlation was computed to assess the relationship between serum and regional cerebral lactate.

# RESULTS

## Patient Characteristics

From April 2012 to July 2017, 52 infants were enrolled into the prospective study and each underwent research MRS during and after TH. Four patients were excluded from analysis due to congenital anomaly (1), gestational age at birth <35 weeks (1), and poor quality on MR spectra (2). Forty-eight subjects (29 male) with mean gestational age of 39 ± 1.8 weeks and birth weight of 3.3 ± 0.6 kg were included in data analysis. The mean age at which the first (during TH) and second MRI (after TH) were performed was 54.6 ± 15 and 156 ± 57.6 h of life, respectively. Based on modified Sarnat exam (16) on admission, 39 had moderate and 9 had severe encephalopathy, respectively. Based on radiographic scoring of injury severity, 37 were normal to mild (nml/mild) and 11 were moderate to severe (mod/severe). Of note, 7 out of the 9 infants with severe encephalopathy based on Sarnat exam had mod–severe injury on MRI. With regard to pattern of injury: 28 were normal, 8 were BG-Thal pattern, and 12 were watershed pattern. Thirteen infants received dopamine or dobutamine infusion during the first 3 days of life. Only two patients received vasopressor–inotrope infusion at doses ≥10 g/kg/min. None received more than one form of vasopressor–inotrope. Markers of other end-organ injury included the highest value of creatinine, aspartate aminotransferase (AST), and alanine aminotransferase (ALT) within the first 72 h of postnatal life. Median (IQR) creatinine was 0.82 (0.7–1.1) mg/dL, AST 141.5 (87.3–282.3) U/L, and ALT 49 (34–105) U/L, respectively. Only two infants had creatinine level >1.5. Twenty out of the 48 infants had some degree of hepatic dysfunction, defined as AST > 200 U/L and/or ALT > 100 U/L. Of the 48 patients, two died and two required G-tube feedings at discharge.

#### Temporal Resolution of Serum Lactate

Overall, 720 serum lactate measurements were obtained for our cohort. The median age at which serum Lactate 1 was obtained was 5.02 (range: 4.1–6.5) hours of life. Serum lactate peaked on or near admission (0–8 h) and trended toward normal over time,

Figure 1 | Scatter plot of serum lactate measured over time. All serum lactate values in black circles were plotted over time (*n* = 780) with timing of first MRI indicated by fluorescent green vertical line. The red horizontal line marks the upper limit of normal serum lactate level (19 mg/dL). By 72–96 h of life, 96% percent of the serum lactate values are within reference range.

as evidenced by the increasing percentages of normal (defined as <19 mg/dL) serum lactate concentrations: 24.6% at 0–24 h; 59.5 at 24–48 h; 77.2% at 48–72 h, and 95.9% at 72–96 h, **Figure 1**.

#### Temporal Resolution of Cerebral Lactate

Overall, cerebral lactate concentration was not significantly different *during* vs. *after* TH [1.05 (0.85–1.41) vs. 1.16 (0.94–1.37) mmol/kg, *p* = 0.16]. When examining the cerebral lactate concentration in respective regions (i.e., BG lactate during TH vs. BG lactate after TH) during and after TH, there was also no significant difference. However, when cohort is dichotomized to severity of injury, there was a significant difference in cerebral lactate during vs. after TH. For the nml/mild injury, there was a small, albeit significant *increase* in cerebral lactate concentration from during TH to after TH [1.00 (0.83–1.2) vs. 1.20 (0.94–1.33) mmol/kg, *p*= 0.005]. By contrast, for the mod/severe group, there was a small, but significant *decrease* in cerebral lactate concentration [1.49 (1.0–2.54) vs. 1.3 (0.97–1.9), *p* = 0.028]. In general, as shown in **Figure 2**, cerebral lactate remained low, across the first 2 weeks of life in the nml/mild group (**Figure 2A**); however, cerebral lactate was markedly elevated and then, on average, decreased over time among the mod/severe group (**Figure 2B**).

### Regional Cerebral Lactate in Relation to Severity and Pattern of Injury

There were no significant differences among regional cerebral lactate concentrate in the nml/mild injury group (all adjusted *p* > 0.3). Similarly, there was no significant difference among regional cerebral lactate concentration in the mod/severe group (all adjusted *p* > 0.4). However, when comparing regional cerebral lactate concentration between severity groups *during* TH, we found that the mod/severe group had significantly higher lactate concentration than the nml/mild group at the BG (1.8 ± 1.0 vs. 1.0 ± 0.3, *p* < 0.001) and Thal (1.9 ± 1.3 vs. 1.0 ± 0.3, *p* < 0.001), **Figure 3A**.

When examining cerebral lactate levels in regards to pattern of injury (defined radiographically as normal pattern, BG-T pattern, and WS pattern), significantly higher lactate levels were found in

Figure 2 | (A) Cerebral lactate over time in normal/mild injury. (B) Cerebral lactate over time in moderate-severe injury. A marked difference between cerebral lactate clearance in infants with nml/mild injury (A) vs. infants with mod/severe injury (B). Although by 2 weeks of life, cerebral lactate trends toward 1 mmol/kg, it remains detectable.

the BG and Thal regions in the BG-T pattern (BG 2.0 ± 1.0 and Thal 2.3 ± 1.3 mmol/kg) compared to respective regions in the WS (BG 1.0 ± 0.5 and Thal 1.0 ± 0.4 mmol/kg) and normal pattern (BG 1.0 ± 0.3 and Thal 1.1 ± 0.2 mmol/kg), all adjusted *p* < 0.005 (**Figure 3B**). In short, significant differences in lactate levels were only detected at the BG and Thal regions when comparing by severity *or* predominant pattern of injury *during* TH. In contrast, *after TH*, there was no significant difference in cerebral lactate concentration at *all* brain regions when comparing infants with regard to injury severity (i.e., nml/mild vs. mod/severe injury, all *p*> 0.06) or pattern (i.e., BG-T vs. WS pattern of injury, all *p*> 0.1).

#### Serum Lactate and Cerebral Lactate Correlation

Across the full sample, there was no significant correlation between serum Lactate 1 and cerebral lactate in either the BG,

Figure 4 | (A) Scatter plot of serum Lactate 2 vs. cerebral lactate, all patients. There is a weak correlation between serum Lactate 2 and cerebral lactate concentration, only at regions basal ganglia (BG) and Thal, shown as closed triangle and closed circle, respectively. There is no significant correlation between serum lactate and cerebral lactate at regions gray matter (GM) and white matter (WM), denoted as open square and diamond in the graph. (B) Scatter plot of serum Lactate 2 vs. cerebral lactate in infants with moderate–severe injury. For subjects with moderate–severe injury, there is a strong correlation between serum Lactate 2 and cerebral lactate at the BG, Thal, and GM region, which are denoted here in closed triangle, closed circle, and closed square, respectively. There is no significant correlation between serum lactate and WM lactate, denoted as open diamond in the graph.

GM, or WM regions (*p* > 0.05); however, there was a weak positive association in the Thal (*r*<sup>s</sup> = 0.4, *p* = 0.006). Likewise, there was a weak positive association between serum Lactate 2 and cerebral lactate concentration at the BG (*r*<sup>s</sup> = 0.3, *p* = 0.04) and Thal (*r*<sup>s</sup> = 0.35, *p* = 0.02), but no correlation between Lactate 2 and cerebral lactate in the GM and WM regions (both *p* > 0.05), **Figure 4A**. When limiting our analysis to infants with moderate–severe injury (*n* = 11), the correlation between serum Lactate 2 and cerebral lactate concentration at the BG, Thal, and GM regions became strong and significant (*r*<sup>s</sup> = 0.7, *p* = 0.03, *r*<sup>s</sup> = 0.9, *p* = 0.001, and *r*<sup>s</sup> = 0.6, *p* = 0.04, respectively), **Figure 4B**.

#### DISCUSSION

In this prospective observational study of 48 infants with HIE, we investigated the changes in serum and cerebral lactate concentration over time, during and after TH. We also described the differences in cerebral lactate concentration in relation to severity and predominant pattern of injury. Finally, we examined the relationship between serum lactate and cerebral lactate concentrations.

As expected, serum lactate was elevated on admission and then normalized over time. By contrast, we did not observe a significant effect of time on cerebral lactate measures (during TH vs. after re-warming), when considered across the entire sample. However, there were markedly different trends in cerebral lactate concentration among the neonates with nml/mild MRIs as compared to those with mod/severe injury. Specifically, cerebral lactate concentration increased slightly after TH for the nml/mild subgroup, but decreased slightly after TH for the mod/severe subgroup. There was also a significant difference in regional cerebral lactate concentration relative to the severity and pattern of injury during TH. These significant differences, however, disappeared after TH. Serum Lactate 2 was modestly associated with BG and Thal lactate; the association became stronger in the subgroup of infants with mod/severe injury.

Lactate is generated by way of redox conversion of pyruvate by lactate dehydrogenase, which is necessary to generate ATP and NAD+ in the setting of low oxygen or insufficient mitochondrial metabolism (22). As such, serum lactate is often interpreted clinically as a surrogate marker for tissue hypoxia and/or ischemia (23, 24). As expected, in this cohort, serum lactate was markedly elevated immediately after birth and then trended toward normal. However, it is notable that by day 3 and 4 of life, 23 and 5% of the sample still demonstrated elevated lactate concentrations (≥20 mg/dL), respectively. This persistent elevation in serum lactate could be attributed to hemodynamic instability during the rewarming phase of TH (13); however, none of the infants remained on vasopressor inotropes during that period. Instead, the observation of a persistently and mildly abnormal concentrations of serum lactate even after restoration of adequate organ perfusion and oxygenation is more likely to be secondary to other mechanisms (25, 26). One possibility is that impaired cellular energy metabolism in injured organs led to an ongoing net flux of lactate into the bloodstream from injured tissues. Another possibility is that stress-induced epinephrine secretion, which stimulates Na-K ATPase and thereby can induce an increase in aerobic glycolysis in the skeletal muscle (26), contributed to elevated serum lactate. Finally, a mismatch between increased astrocytic production of lactate and reduced uptake from neurons may also contribute to elevated extracellular lactate in brain injury (27). Further research is needed to better elucidate the complex interplay between lactate production, uptake, oxidation across the CNS, systemic circulation, and other end organs.

The temporal evolution of cerebral lactate differed by injury severity. The nml/mild injury cohort demonstrated a small but significant increase in cerebral lactate concentration from TH to rewarming. It is important to note that the cerebral lactate concentration for this cohort was relatively low to start with [1.00 (0.83–1.2) mmol/kg]; thus, the clinical significance of this increase is unknown. Nevertheless, this finding is consistent with data from animal models whereby TH attenuates the anaerobic metabolism of glucose to lactate (28).

In contrast, cerebral lactate concentrations were significantly elevated during TH in the mod/severe group (range: 1.5–5 mmol/ kg), which then decreased slightly from TH to after re-warming. The initial elevation in cerebral lactate during TH is likely related to neuronal injury, reflecting either a failure in neuronal mitochondrial metabolism and concomitant increase in anaerobic metabolism or an uncoupling of glial-neuronal shuttling (aerobic glycolysis) and a concomitant increase in extracellular lactate (27). However, given the high capacity for lactate transport into and out of the newborn brain, it is possible that over time, excessive lactate is transported out of the brain (29) if not otherwise metabolized (25, 30).

To illustrate the interplay between serum lactate and cerebral lactate over time in the setting of severe HIE and BGT brain injury, we provide a time–resolution plot from one of our patients in (**Figure 5**). Briefly, this infant was born following uterine rupture and presented with severe encephalopathy on exam without laboratory evidence of kidney injury or hepatic dysfunction. Serum lactate was markedly elevated early but normalized over time. Lactate in the thalamus was markedly elevated during TH, and then decreased slightly after re-warming. By contrast, BG and GM lactate were moderately elevated during TH and then slightly increased after re-warming. Most notably, although cerebral lactate levels on day 24 were lower than during TH, they remained elevated. Given that this patient lacked evidence of any severe extracerebral end-organ injury, the cerebral lactate levels can be attributed to evolving brain injury. This patient died at 5 months of age.

Cerebral lactate concentrations were also higher in accordance to the radiographic scoring of injury, which was determined by a pediatric neuroradiologist blinded to both MRS findings and clinical characteristics. Cerebral lactate concentrations were significantly higher in the BG and Thal regions of infants with moderate–severe injury compared to normal–mild injury. Likewise, cerebral lactate was significantly higher in the BG and Thal regions in neonates with a predominantly BG-T pattern of injury.

Figure 5 | Time-resolution of serum and cerebral lactate in an infant with severe brain injury without any significant end organ injury. Serum lactate rapidly normalized and was <20 mg/dL when the first cerebral lactate was markedly elevated. A repeat scan performed at more than 3 weeks of age reveals lower but still elevated cerebral lactate, especially in the basal ganglia (BG) and thalamus (Thal) regions. Apparent diffusion coefficient map below the *x*-axis reveals progression of injury from the early scan during therapeutic hypothermia (TH) to post-TH, and then chronic injury seen on day 24.5. Cerebral lactate concentration at the BG, Thal, and gray matter (GM) region are denoted here in closed triangle, closed circle, and open square, respectively.

Wu et al. Systemic and Cerebral Lactate in Neonatal HIE

Finally, we observed modest associations between serum lactate and cerebral lactate measured in the deep gray nuclei. It has been widely hypothesized that acute, near-total asphyxia events are associated with injury to the thalamus, BG, and perirolandic cortex while partial prolonged events are associated with injury to the parasagittal cortex and underlying WM (31, 32). In line with this, we observed a modest association between serum Lactate 1 and thalamic lactate. Furthermore, we observed associations between serum Lactate 2 and lactate levels in the thalamus and BG. This not only supports the hypothesis that serum lactate, a marker of global hypoxia-ischemia, is most closely associated with injury to the deep gray nuclei, but also suggests that persistent elevations in serum lactate may arise, at least in part, from ongoing flux of lactate from the CNS into the bloodstream.

In the pre-therapeutic hypothermia era, Hanrahan et al. examined cerebral lactate by 1 H MRS and found that lactate in the BG persisted beyond 1 month after birth in those with poor outcome, while no lactate was detected in infants with normal development or in normal controls (33). We observed a similar pattern, albeit, an earlier decline in cerebral lactate among neonates with moderate to severe injury. Importantly, elevated cerebral lactate appears despite an increase in cerebral blood flow during the subacute phase of injury (34, 35) as well as adult stroke (36). This implies that the initial cerebral lactate peak may occur during the acute phase of injury due to anaerobic metabolism, while the persistence of increased cerebral lactate may be the result of a perturbed cerebral energy metabolism, even under aerobic conditions.

The finding that serum Lactate 2 was more strongly associated with cerebral lactate in the mod/severe injury group than the sample as a whole may be a reflection of the clinical heterogeneity of this cohort. As shown in **Figure 4B**, this association not only reflects infants with high serum and cerebral lactate but also infants with relatively *low* serum and cerebral lactate. It is unclear if the infants with relatively low serum and cerebral lactate concentration suffered a more subacute injury or if the injury timing was weeks prior to birth. This cohort of mod/severe injury reflects the myriad of injury patterns that is often seen clinically, where severe encephalopathy does not equate to severe end organ injury or vice versa, as previously illustrated (**Figure 5**). Furthermore, the strong correlation implies a fixed ratio between serum and cerebral lactate. It is unclear if this finding represents an equilibrium between systemic and cerebral lactate. Again, the current study was not designed to determine direction of lactate flux. Future studies that make use of carbon tagging are needed to further elucidate the direction of cerebral lactate flux in the infants with moderate–severe brain injury.

There are some limitations to the study. We were unable to assess the relationship between serum lactate and cerebral lactate after

#### REFERENCES


rewarming. Serum lactate values have often normalized by then and no further clinical trending of lactate values was warranted. Also, we were only able to transport clinically stable infants to the MRI suite during TH. Hence, the study cohort is subject to selection bias toward a moderate disease severity. Indeed, 39 out of the 48 infants were moderately encephalopathic on Sarnat exam.

#### CONCLUSION

Serum lactate was elevated on admission and then normalized over time. The temporal evolution of cerebral lactate differed by injury severity. Serum lactate correlated modestly with cerebral lactate measured from the deep gray nuclei (thalamus and BG). Further studies are needed to study the effects of altered lactate metabolism in neonatal HIE.

#### ETHICS STATEMENT

The research study was approved by the institution review board of the Human Subjects Protection Program at the Children's Hospital Los Angeles. Written consent from the parent(s) were obtained by the researchers prior to all subject enrollment.

### AUTHOR CONTRIBUTIONS

T-WW designed the research project, enrolled subjects, collected and analyzed data, and constructed the manuscript. K-HH collected and analyzed the data and helped with the improvement of the manuscript. AR was instrumental in subject recruitment, data analysis, and the proofreading of the manuscript. BT is a neuroradiologist who scored all the MR images according to the Barkovich scoring system for neonatal asphyxia. She also contributed to modification of the manuscript, specifically the Section "Materials and Methods." MB was involved in the organization of MRS data, in designing of this study, and proofreading of the manuscript. EH was instrumental in subject recruitment, composition of the manuscript. SB was involved in the design of the research, collected, analyzed the data, and was instrumental in MR spectroscopy analysis. JW designed the research, contributed to generation of hypothesis and subsequent methodology, collected and analyzed data, and improved the manuscript.

### FUNDING

The study was supported by funding from the Rudi Schulte Research Institute (SB and JW) and the American Society for Pediatric Neuroradiology (JW).


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