# GLUTAMATE-RELATED BIOMARKERS FOR NEUROPSYCHIATRIC DISORDERS

EDITED BY : Hsien-Yuan Lane, Chieh-Hsin Lin and Kenji Hashimoto PUBLISHED IN : Frontiers in Psychiatry

### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-379-1 DOI 10.3389/978-2-88963-379-1

### About Frontiers

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

### Frontiers Journal Series

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

### Dedication to Quality

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

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

### What are Frontiers Research Topics?

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

# GLUTAMATE-RELATED BIOMARKERS FOR NEUROPSYCHIATRIC DISORDERS

Topic Editors: Hsien-Yuan Lane, China Medical University, Taiwan Chieh-Hsin Lin, Kaohsiung Chang Gung Memorial Hospital, Taiwan Kenji Hashimoto, Chiba University, Japan

Citation: Lane, H.-Y., Lin, C.-H., Hashimoto, K., eds. (2020). Glutamate-Related Biomarkers for Neuropsychiatric Disorders. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-379-1

# Table of Contents


Ming-Chyi Huang, Lian-Yu Chen, Hu-Ming Chang, Xiao-Yu Liang, Chih-Ken Chen, Wan-Ju Cheng and Ke Xu

*50 Blood Levels of Glutamate and Glutamine in Recent Onset and Chronic Schizophrenia*

Caroline Madeira, Flavio V. Alheira, Marilia A. Calcia, Thuany C. S. Silva, Filippe M. Tannos, Charles Vargas-Lopes, Melissa Fisher, Nelson Goldenstein, Marco Antonio Brasil, Sophia Vinogradov, Sergio T. Ferreira and Rogerio Panizzutti


Mary-Anne B. MacKay, Maryana Kravtsenyuk, Rejish Thomas, Nicholas D. Mitchell, Serdar M. Dursun and Glen B. Baker

*86 The Effects of Acupuncture on Glutamatergic Neurotransmission in Depression, Anxiety, Schizophrenia, and Alzheimer's Disease: A Review of the Literature*

Cheng-Hao Tu, Iona MacDonald and Yi-Hung Chen

### *98 Targeting mGlu Receptors for Optimization of Antipsychotic Activity and Disease-Modifying Effect in Schizophrenia*

Ferdinando Nicoletti, Rosamaria Orlando, Luisa Di Menna, Milena Cannella, Serena Notartomaso, Giada Mascio, Luisa Iacovelli, Francesco Matrisciano, Francesco Fazio, Filippo Caraci, Agata Copani, Giuseppe Battaglia and Valeria Bruno

### *108 Autism Associated With Anti-NMDAR Encephalitis: Glutamate-Related Therapy*

Ruu-Fen Tzang, Chuan-Hsin Chang, Yue-Cune Chang and Hsien-Yuan Lane

# Editorial: Glutamate-Related Biomarkers for Neuropsychiatric Disorders

Chieh-Hsin Lin1,2,3, Kenji Hashimoto<sup>4</sup> and Hsien-Yuan Lane5,6,7\*

<sup>1</sup> Department of Psychiatry, Kaohsiung Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Kaohsiung, Taiwan, <sup>2</sup> Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan, <sup>3</sup> School of Medicine, Chang Gung University, Taoyuan, Taiwan, <sup>4</sup> Division of Clinical Neuroscience, Chiba University Center for Forensic Mental Health, Chiba, Japan, <sup>5</sup> Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan, <sup>6</sup> Department of Psychiatry & Brain Disease Research Center, China Medical University Hospital, Taichung, Taiwan, <sup>7</sup> Department of Psychology, College of Medical and Health Sciences, Asia University, Taichung, Taiwan

Keywords: glutamate, NMDA, schizophrenia, depression, Alzheimer's disease, autoimmune encephalitis, autism, addiction

Editorial on the Research Topic

### Glutamate-Related Biomarkers for Neuropsychiatric Disorders

### Edited by:

Ming D. Li, Zhejiang University, China

### Reviewed by: Glen Bryan Baker,

University of Alberta, Canada

\*Correspondence: Hsien-Yuan Lane hylane@gmail.com

### Specialty section:

This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry

Received: 24 September 2019 Accepted: 15 November 2019 Published: 05 December 2019

### Citation:

Lin C-H, Hashimoto K and Lane H-Y (2019) Editorial: Glutamate-Related Biomarkers for Neuropsychiatric Disorders. Front. Psychiatry 10:904. doi: 10.3389/fpsyt.2019.00904 The glutamate receptor plays an important role in synaptic plasticity, learning, memory, and the pathogenesis and pathophysiology of numerous neuropsychiatric disorders such as schizophrenia, depression, addiction, autism, autoimmune encephalitis, and neurodegenerative diseases such as Alzheimer's disease.

For example, both ionotropic (especially NMDA and AMPA) and metabotropic (mGLU) receptors were reported to be involved in schizophrenia. In details, taking NMDA receptors (NMDARs) as an example, the receptors are widely distributed in most major organs (including heart, ovary, kidney, gastrointestinal system, lung, and etc.) and tissues (including ganglia cells, nerve fibers, blood vessels, enteroendocrine cells, liver, mast cells, inflammatory cells and etc.). The NMDAR has binding sites not only for glutamate or aspartate, but also a separate coagonist site for the endogenous ligands, D-serine, D-alanine, and glycine. Occupancy of the coagonist site can increase the frequency of opening of the channels activated by NMDAR agonists, facilitating excitatory neurotransmission in the brain. In fact, the binding of both glycine (or D-serine, D-alanine) and glutamate is required to open the NMDAR channel ionophore.

Some of the NMDARs isolated from peripheral tissues have been cloned and sequenced. These sequences correspond with NMDARs that have been cloned in the CNS. Further physiological and pharmacological experiments support the hypothesis that NMDARs in the periphery have similar properties to those in the CNS or expressed in host cells transfected with cloned subunits. Physiological studies with agonists and antagonists of the NR1 subunit of the NMDAR in the pig ileum have shown that these receptors are similar to those characterized in the CNS.

Direct measurement of physiological changes and NMDAR-related biomarkers in human brains is nearly impossible because of invasive testing procedures and ethical issues. Thus, developing accessible peripheral or brain image biomarkers become more important for mental illness. Peripheral gene expression may be a useful surrogate for gene expression in the CNS when the relevant gene is expressed in both. Lymphocytes or white cells have been suggested to be a neural

**5**

probe because numerous studies showed similarities between receptor expression and mechanisms of transduction processes of cells in the nervous system (e.g. neurons and glia) and lymphocytes. Blood-derived RNA has also become a convenient alternative to traditional tissue biopsy-derived RNA. Therefore, peripheral gene expressions as well as other peripheral biomarkers might have potential to be surrogate CNS biomarkers for disorders, their outcomes, and treatment responses in the CNS.

Among the 12 articles of this research topic Glutamate-Related Biomarkers for Neuropsychiatric Disorders, the first paper Elevated Glutamate and Glutamine Levels in the Cerebrospinal Fluid of Patients With Probable Alzheimer's Disease and Depression, by Madeira et al., reports increased CSF levels of glutamate and glutamine in Alzheimer's disease and major depressive disorder compared to healthy controls and to patients with normal pressure hydrocephalus. Importantly, lower and Innotest amyloid tau index (IATI) scores were correlated with higher glutamate and glutamine levels in their study cohort. Further, in subjects without dementia (Mini-Mental State Examination [MMSE] above 23), lower MMSE scores were associated with higher glutamate and glutamine levels.

Accumulating evidence suggests that dysfunctions in in the glutamate-glutamine cycle also play a role in the pathophysiology of schizophrenia. Madeira et al. demonstrated altered blood levels of glutamate and glutamine in patients with schizophrenia in their another study, Blood Levels of Glutamate and Glutamine in Recent Onset and Chronic Schizophrenia. They indicate that circulating glutamine/glutamate ratios rise at the onset of illness but fall with progression of the disorder.

In a brief review, Astrocytic Regulation of Glutamate Transmission in Schizophrenia, Mei et al. updated findings on the crucial role of glutamate uptake and astrocyte-derived Dserine in the pathophysiology of schizophrenia. Neuroimaging and neurochemical biomarkers based on glutamate transmission and metabolism have also shown translational potential. Further elucidation of neuron-astrocyte interaction during the aforementioned processes will facilitate the development of novel diagnostic and therapeutic approaches for schizophrenia and related disorders.

Not only ionotropic but also metabotropic glutamate (mGlu) receptors are regarded as novel drug targets for the treatment of schizophrenia. Nicoletti et al. reviewed the development of selective ligands of individual mGlu receptors as new therapy for schizophrenia in the article Targeting mGlu Receptors for Optimization of Antipsychotic Activity and Disease-Modifying Effect in Schizophrenia. Subtype-selective mGlu receptor ligands thereby offer the potential of a precision medicine based pharmacological approach for various domains of schizophrenia. Moreover, some mGlu receptor ligands may exert a positive influence on neuroinflammation by modulating microglial function or other mechanisms.

While glutamate is the agonist of the NMDAR, D-serine appears to be themost potent coagonist.The reviewarticle (D-Serine: Potential Therapeutic Agent and/or Biomarker in Schizophrenia and Depression?) by Mackay et al. indicates that neurons are more relevant to D-serine action than originally thought. Interestingly, D-serine appears to have antipsychotic and antidepressive properties. D-serine may be also a biomarker for antidepressant response to ketamine. Preclinical and clinical studies reveal that Dserine may be a potential cognitive enhancer, deserving further study.

Besides D-serine and D-alanine, another endogenous Damino acid in the mammalian brain, D-aspartate, can influence NMDAR-mediated transmission. Errico et al. in their review, The Emerging Role of Altered D-Aspartate Metabolism in Schizophrenia: New Insights From Preclinical Models and Human Studies, indicated that increased levels of the NMDAR agonist, D-aspartate, impact on a series of functional and structural phenotypes relevant to schizophrenia. Moreover, these findings suggest a possible role of dysregulated embryonic D-aspartate metabolism in schizophrenia pathogenesis, and consequently propose the potential of D-aspartate as a novel therapy or prevention for this neurodevelopmental brain disorder.

D-amino acid oxidase activator (DAOA, also known as G72) protein regulates D-serine neurotransmission. Previous study found that the peripheral G72 protein expression is distinctively higher in patients with schizophrenia than in healthy individuals. Lin and Lin in their study Combination of G72 Genetic Variation and G72 Protein Level to Detect Schizophrenia: Machine Learning Approaches, demonstrated that G72 protein alone, even without adding G72 SNPs, can discriminate schizophrenia patients from healthy subjects, with adequate power. They also suggest a combination of logistic regression and naive Bayes models to establish an optimal model to predict schizophrenia.

Dysfunctional glutamatergic neurotransmission is implicated in the etiology of not only schizophrenia but also major depressive disorder. In the review article Glutamatergic Dysfunction and Glutamatergic Compounds for Major Psychiatric Disorders: Evidence From Clinical Neuroimaging Studies, Li et al. indicated that ketamine, an NMDAR antagonist, has displayed its rapid antidepressant efficacy for treatment-resistant depression. Many compounds which modulate the glutamatergic transmission have also shown potential in the treatment of major psychiatric disorders, via approaches such as magnetic resonance spectroscopy, positron emission tomography/single-photon emission computed tomography, and paired-pulse transcranial magnetic stimulation.

On the other hand, ketamine is also a common drug of abuse, while disrupted oxytocin system is involved in the development of addiction. In the original study Decreased Blood Levels of Oxytocin in Ketamine-Dependent Patients During Early Abstinence, Huang et al. found that the oxytocin level was significantly lower in ketamine-dependent patients and the level did not normalize after early abstinence. Lower oxytocin might be also related with anxious phenotype of ketamine dependence. These results suggest the potential of therapeutic use of oxytocin for treating ketamine dependence.

Previous studies have demonstrated that plasma asymmetric dimethylarginine (ADMA) was increased in patients with schizophrenia and correlated with cognitive dysfunction. In the study Treatment Responses of Cognitive Function and Plasma Asymmetric Dimethylarginine to Atypical Antipsychotic in Patients With Schizophrenia, Yu et al. found that atypical antipsychotic treatment in acutely exacerbated schizophrenia patients improved psychiatric symptoms and cognitive function (especially working memory and attention), in parallel with decreased plasma ADMA levels, implying that plasma ADMA may be a potential indicator of cognitive recovery in schizophrenia.

In the review, Autism Associated With Anti-NMDAR Encephalitis: Glutamate-Related Therapy, Tzang et al. addressed the role of autoimmune dysfunction in autism. The anti-NMDAR autoantibody can lead to dysfunctional glutamate neurotransmission in the brain that manifests as various psychiatric symptoms such as psychosis and personality changes. Glutamate-related therapy primarily normalizes glutamate neurotransmission and can be a new adjuvant intervention alongside antipsychotics for treating autoimmune autism.

Acupuncture has been regarded as a complementary therapy for neuropsychiatry disorders. In the article, The Effects of Acupuncture on Glutamatergic Neurotransmission in Depression, Anxiety, Schizophrenia, and Alzheimer's Disease: A Review of the Literature Review, Tu et al. reported the current evidence for the treatment efficacy of acupuncture in depression, anxiety, schizophrenia, and Alzheimer's disease. Further, they discuss the ways in which acupuncture treatment can potentially modulate glutamate receptors and excitatory amino acid transporters.

Overall, to promote the development of glutamate-related biomarkers for neuropsychiatric disorders, this research topic gathers a comprehensive body of review and original articles to update the current state of this increasingly important theme. Further basic, preclinical and clinical studies are needed to elucidate the possible mechanisms of glutamate-related dysfunction in neuropsychiatric disorders and to establish clinical applications for diagnoses and treatments of these disorders.

### AUTHOR CONTRIBUTIONS

C-HL and H-YL wrote the first draft of the manuscript, and KH provided opinions on it. All authors read and approved the submitted version.

### FUNDING

This work was partly supported by Ministry of Science and Technology, Taiwan (MOST 107-2632-B-039-001; 108-2314-B-039-002; 108-2622-B-039-001-CC2; 108-2628-B-182A-002), National Health Research Institutes (NHRI-EX108-10731NI; NHRI-EX108-10816NC), China Medical University Hospital, Taiwan (DMR-105-076), Chang Gung Memorial Hospital, Taiwan (CMRPG8G1391), and Taiwan Ministry of Health and Welfare Clinical Trial and Research Center of Excellence (MOHW108-TDU-B-212-133004).

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

Copyright © 2019 Lin, Hashimoto and Lane. 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.

# Elevated Glutamate and Glutamine Levels in the Cerebrospinal Fluid of Patients With Probable Alzheimer's Disease and Depression

Caroline Madeira<sup>1</sup> , Charles Vargas-Lopes <sup>1</sup> , Carlos Otávio Brandão<sup>2</sup> , Taylor Reis <sup>2</sup> , Jerson Laks <sup>2</sup> , Rogerio Panizzutti 1,2 \* and Sergio T. Ferreira3,4 \*

1 Institute of Biomedical Sciences, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, <sup>2</sup> Institute of Psychiatry, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, <sup>3</sup> Institute of Medical Biochemistry Leopoldo de Meis, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, <sup>4</sup> Institute of Biophysics Carlos Chagas Filho, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

### Edited by:

Hsien-Yuan Lane, China Medical University, Taiwan

### Reviewed by:

Kenji Hashimoto, Chiba University, Japan Chih-Chiang Chiu, Taipei City Psychiatric Center, Taiwan Cheng-Sheng Chen, Kaohsiung Medical University Hospital, Taiwan

### \*Correspondence:

Rogerio Panizzutti rogerio@icb.ufrj.br Sergio T. Ferreira ferreira@bioqmed.ufrj.br

### Specialty section:

This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry

Received: 19 June 2018 Accepted: 17 October 2018 Published: 06 November 2018

### Citation:

Madeira C, Vargas-Lopes C, Brandão CO, Reis T, Laks J, Panizzutti R and Ferreira ST (2018) Elevated Glutamate and Glutamine Levels in the Cerebrospinal Fluid of Patients With Probable Alzheimer's Disease and Depression. Front. Psychiatry 9:561. doi: 10.3389/fpsyt.2018.00561 Recent evidence suggests that Alzheimer's disease (AD) and depression share common mechanisms of pathogenesis. In particular, deregulation of glutamate-mediated excitatory signaling may play a role in brain dysfunction in both AD and depression. We have investigated levels of glutamate and its precursor glutamine in the cerebrospinal fluid (CSF) of patients with a diagnosis of probable AD or major depression compared to healthy controls and patients with hydrocephalus. Patients with probable AD or major depression showed significantly increased CSF levels of glutamate and glutamine compared to healthy controls or hydrocephalus patients. Furthermore, CSF glutamate and glutamine levels were inversely correlated to the amyloid tau index, a biomarker for AD. Results suggest that glutamate and glutamine should be further explored as potential CSF biomarkers for AD and depression.

Keywords: Alzheimer's disease, depression, glutamate, glutamine, cerebrospinal fluid, innotest amyloid tau index

# INTRODUCTION

Epidemiological and clinical studies suggest an association between Alzheimer's disease (AD) and depression (1, 2), the latter being a risk factor for development of AD and other forms of dementia (3–5). About half of the patients with major depression show cognitive impairment that can be persistent and last even after remission of the acute phase of symptoms (6). This could mean cognitive impairment precedes or predisposes to depression or, alternatively, that depression produces persistent cognitive deficits (6, 7). Recent studies have further suggested that similar pathogenic mechanisms may underlie cognitive changes in depression and in AD. Indeed, we have demonstrated that soluble oligomers of the amyloid-β peptide (AβOs), neurotoxins that accumulate in the AD brain and are thought to cause synapse failure and memory loss in AD (8, 9), induce depressive-like behavior in mice (10, 11), providing insight into molecular/cellular mechanisms potentially connecting both disorders.

Glutamate is the major excitatory neurotransmitter in the mammalian brain. However, glutamate at high concentrations in the synaptic cleft is toxic and may result in neuronal death, a phenomenon generally termed excitotoxicity (12). Excitotoxicity mediated by aberrant activation of glutamate receptors, notably of N-methyl-D-aspartate (NMDA) receptors, has been related to the neuropathology of AD (13–16). Consistent with a role of abnormal NMDA receptor function in

**8**

AD, memantine, an NMDA receptor blocker, is one of the few drugs in clinical use for treatment of moderate to severe AD (17, 18). Interestingly, NMDA receptor antagonists, in particular ketamine, also produce rapid and consistent antidepressant action in subjects with major depression (19, 20).

To prevent excitotoxicity after physiological neurotransmission, glutamate is rapidly removed from the synaptic cleft and converted into glutamine by glutamine synthetase in glial cells (21). Glutamine is then transported back to the presynaptic neuron, where it is converted to glutamate by glutaminase (22). Dysfunction in the glutamate-glutamine cycle could contribute to excitotoxicity mediated by glutamate (22).

Changes in glutamate and glutamine may occur in AD and depression, previous studies of glutamate and glutamine levels in the cerebrospinal fluid (CSF) present controversial findings. Some studies have reported that CSF glutamate levels in AD patients were higher than in the controls (23, 24) or MCI (25), whereas others studies have found decreased (26, 27) or unchanged glutamate levels in AD patients compared to healthy controls (28, 29). Studies that measured CSF glutamine levels have also reported controversial results, with studies reporting increased (30), decreased (28, 31) or no change in glutamine levels in AD patients compared with controls (23, 24, 26).

Controversial findings have also been observed in CSF glutamate and glutamine levels of patients with depression. One study found an increase in CSF glutamine levels (32), whereas another study showed a decrease in CSF glutamate levels in patients with depression compared to controls (33). Moreover, recent studies reported no differences in CSF glutamate and glutamine levels in patients with depression compared to controls (34, 35).

Here, we investigated whether changes in brain levels of glutamate and glutamine are present in major depression and AD. We have studied glutamate and glutamine levels in the CSF of age-matched patients with probable AD or major depression compared to two control groups: healthy subjects and patients with an unrelated neurological condition (normal pressure hydrocephalus). We further investigated the correlation between glutamate and glutamine levels, Mini-Mental State Examination (MMSE) scores, and the Innotest amyloid tau index, a biomarker for AD (36, 37).

## MATERIALS AND METHODS

### Human Subjects

The current study was approved by the Committee of Research Ethics involving human subjects of the Institute of Psychiatry of Federal University of Rio de Janeiro (protocol # 32liv2/07). Subjects provided written informed consent before study admission.

All subjects underwent an evaluation comprising full medical history, physical and neurological examination, laboratory tests and neuropsychological assessments. The complete work up is detailed elsewhere (38). The severity of dementia was classified according to the Clinical Dementia Rating (CDR) (39). Mini-Mental State Examination (MMSE) (40) was additionally used to evaluate cognitive state. Exclusion criteria included more than 10 packs/year of cigarette smoking, alcohol abuse or other current or previous psychiatric or clinical disorder. The probable AD group included 21 subjects recruited from the AD Center of the Institute of Psychiatry of the Federal University of Rio de Janeiro. Subjects were diagnosed according to National Institute of Neurological and Communicative Disorders and Stroke (NINCDS), Alzheimer's Disease and Related Disorders Association (ADRDA) and Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria (41). The depression group included 9 subjects with major depression, recruited from the Institute of Psychiatry of the Federal University of Rio de Janeiro and diagnosed according to DSM-IV criteria. The severity of depression was measured by the Brazilian version of the Hamilton Depression Scale (HAM-D) (42–44). Only patients in the first episode of major depression were included in the study. The healthy control group included 10 subjects without any clinical disease or neuropsychiatric disorder. We also studied an additional control group including 9 patients with hydrocephalus, diagnosed according to International Classification of Diseases (ICD-10) (45). Both healthy and hydrocephalus subjects were recruited at Neurolife Laboratory, a private clinic specialized in CSF analysis in the city of Rio de Janeiro. Selected characteristics of each studied group are presented in **Table 1**. Detailed demographics of individual subjects are presented in **Table S1**.

Psychotropic medications used by probable AD patients were: rivastigmine (47.6%; n = 10), risperidone (38.1%; n = 8), memantine (28.6%; n = 6), donepezil (23.8%; n = 5), clonazepam (19.0%; n = 4), citalopram (4.8%; n = 1), trazodone (4.8%; n = 1), biperiden (4.8%; n = 1), escitalopram (4.8%; n = 1), and mirtazapine (4.8%; n = 1). Two (9.5%) patients with probable AD were not taking any medication at the time of the study. Medications used by major depression patients were: citalopram (11.1%; n = 1), clonazepam (33.3%; n = 3), fluoxetine (22.2%; n = 2), desvenlafaxine (11.1%; n = 1), paroxetine (11.1%; n = 1), buspirone (11.1%; n = 1), sertraline (11.1%; n = 1), venlafaxine (11.1%; n = 1). One (11.1%) patient with major depression was not taking any medication at the time of the study.

## CSF Collection

CSF samples were collected by lumbar puncture in the L3-4 or L4-5 interspace at Neurolife Laboratories (Rio de Janeiro, Brazil) and immediately stored at −80◦C. All lumbar punctures were performed between 10 am and noon to limit potential circadian fluctuation in CSF content.

**Abbreviations:** AD, Alzheimer's disease; ADRDA, Alzheimer's Disease and Related Disorders Association; ANOVA, Analysis of variance; Aβ, amyloidβ peptide; AβO, amyloid-β oligomer; CDR, Clinical Dementia Rating; CSF, cerebrospinal fluid; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders; GLX, glutamate plus glutamine concentrations; HPLC, high performance liquid chromatography; IATI, Innotest amyloid tau index; ICD-10, International Classification of Diseases; MMSE, Mini-Mental State Examination; MRS, magnetic resonance spectroscopy; NINCDS, National Institute of Neurological and Communicative Disorders and Stroke; NMDA, N-methyl-D-aspartate; P-tau181, phosphorylated tau; T-tau, total tau.

### TABLE 1 | Characteristics of study subjects.


Values are presented as means (standard deviation). Statistical significance of differences between groups was assessed by one-way ANOVA, F (p-value), followed by Bonferroni adjustment for multiple comparisons, except for sex distribution, which was assessed by the Chi-Square Test, X<sup>2</sup> (p-value). Asterisks indicate statistically significant differences. AD, Alzheimer's disease; MMSE, Mini Mental State Examination; IATI, Innotest amyloid tau index; HAM-D, Hamilton Depression Scale; N.A., Not applicable or not available.

<sup>a</sup> AD significantly different from control, hydrocephalus and depression (P = 0.0001).

<sup>b</sup> AD significantly different from control, hydrocephalus and depression (P = 0.0001).

<sup>c</sup> AD significantly different from control and hydrocephalus (P = 0.0001).

<sup>d</sup> Depression significantly different from control and hydrocephalus (P = 0.0001).

<sup>e</sup> AD significantly different from control and hydrocephalus (P = 0.008 and P = 0.012, respectively).

<sup>f</sup> AD significantly different from control (P = 0.0001).

<sup>g</sup> Depression significantly different from control (P = 0.0001).

<sup>h</sup> Hydrocephalus significantly different from control (P = 0.018).

### Glutamate and Glutamine Measurements in CSF

Glutamate and glutamine levels in CSF were measured by high performance liquid chromatography (HPLC) as previously described (46–48).

### Determination of the Amyloid Tau Index

CSF concentrations of Aβ1-42, total tau (T-tau) and phosphorylated tau (P-tau181) were measured using commercially available enzyme-linked immunosorbent assays (ELISA INNOTEST p-tau-181, INNOTEST htau, INNOTEST β-amyloid (1-42) kits; Innogenetics, Gent, Belgium) according to manufacturer's instructions. The Innotest amyloid tau index (IATI) is a score that combines CSF levels of Aβ1-42 and T-tau (36, 37) and was calculated as:

IATI = Aβ1−42/(240+1.18∗T-tau).

### Statistical Analysis

Results are presented as means (±S.D.) unless otherwise indicated. Statistical significances between groups were determined by one-way analysis of variance (ANOVA) followed by Bonferroni test for multiple comparisons. Sex distribution was studied using Chi-Square. Interactions between years of education, age, disease duration and levels of glutamate or glutamine were studied using correlation analysis and were demonstrated using Pearson's correlation coefficient (r).

### RESULTS

Patient groups did not differ in age, but the sex distribution was significantly different between groups (**Table 1**). However, logistic regression analysis did not reveal significant interactions between sex and levels of glutamate or glutamine (χ <sup>2</sup> = 0.21, p = 0.64 for glutamate; χ <sup>2</sup> = 0.93, p = 0.33 for glutamine). Groups also differed in terms of numbers of years of education (**Table 1**). However, no significant correlations were found between years of education and levels of glutamate or glutamine (r = −0.19, p = 0.18 for glutamate; r = −0.02, p = 0.87 for glutamine). Moreover, glutamate and glutamine levels were not significantly correlated to age (r = 0.04, p = 0.78 for glutamate; r = 0.22, p = 0.12 for glutamine) or disease duration (r = 0.22, p = 0.23 for glutamate; r = 0.03, p = 0.86 for glutamine). None of the medications in use by the patients showed a significant effect on glutamate and glutamine levels (**Table S2**).

Mean CSF glutamate levels were significantly higher in patients with probable AD compared to healthy controls and hydrocephalus patients (F = 50.8, p < 0.0001) (**Table 1**; **Figure 1A**). Interestingly, CSF glutamate levels in the major depression group were similar to the levels found in the probable AD group (**Table 1**, **Figure 1A**). The sensitivity and specificity were 95.2 and 100%, respectively (using a cutoff of 13.63 µmol glutamate/l) for the diagnosis of probable AD compared to healthy controls (AUC = 0.99, p < 0.0001). The corresponding ROC curve is presented in **Figure S1**.

Mean glutamine levels were significantly higher in patients with probable AD than in healthy controls and in the hydrocephalus group (F = 5.92, p = 0.002). Moreover, the mean CSF glutamine level in the major depression group was similar to the mean level found in the group of patients with probable AD (**Table 1**; **Figure 1B**).

The glutamate/glutamine ratio, an index of glutamineglutamate cycle in the brain (49), was significantly higher in all three patient groups (probable AD, major depression

and hydrocephalus) compared to healthy controls (**Table 1**; **Figure 1C**). Thus, while measurements of CSF glutamate alone robustly separated probable AD and depressive patients from controls and hydrocephalus patients, the glutamate/glutamine ratio was elevated in all three disorders compared to controls, suggesting it was a less specific biomarker for AD and depression.

As expected, the mean MMSE score was significantly lower in patients with probable AD than in healthy controls, major depression and hydrocephalus (**Table 1**). Interestingly, in subjects without dementia (MMSE above 23) (50), lower MMSE scores were significantly associated with higher CSF glutamate (r = −0.51, p = 0.006) (**Figure 2A**) and glutamine levels (r = −0.47, p = 0.013) (**Figure 2B**).

As also expected, the mean CSF Innotest amyloid tau index (IATI) was significantly lower in the AD group compared to the other three groups (**Table 1**). Remarkably, CSF glutamate levels were significantly and inversely correlated to the IATI across the four subject groups (r = −0.45, p = 0.002) (**Figure 2C**). Glutamine levels were also inversely correlated to the IATI, although the correlation was less robust and significant (r = −0.31, p = 0.04) (**Figure 2D**).

Finally, CSF glutamate and glutamine levels were analyzed in subject groups separated by their clinical dementia rating (CDR) scores. Individuals with CDR 0.5, 1, 2, and 3 showed significantly elevated mean glutamate levels compared to non-cognitively impaired individuals (CDR 0) (**Figure 3A**). On the other hand, only the group with CDR 2 exhibited significantly higher glutamine than the CDR 0 group (**Figure 3B**).

## DISCUSSION

We report increased CSF levels of glutamate and glutamine in AD and major depression compared to healthy controls and to patients with normal pressure hydrocephalus. Significantly, lower MMSE and IATI scores were correlated with higher glutamate and glutamine levels in our study cohort.

The current findings are consistent with previous studies reporting increased glutamate in the CSF of patients with probable AD (23, 24). In addition, CSF glutamate levels were found to be significantly elevated in patients with AD in comparison to patients with mild cognitive impairment (25). However, other groups have reported decreased (26, 27) or unchanged glutamate levels in patients with probable AD compared to healthy controls (28, 29). Moreover, our finding of increased CSF glutamine in AD patients is in accord with a previous study (30) but differs from other studies that reported a decrease (28, 31) or no change in glutamine levels in patients with probable AD compared to healthy controls (23, 24, 26). It is noteworthy that the criteria for diagnosis of probable AD in early studies included only clinical features (51), which could lead to greater heterogeneity amongst individuals included in the studies and, hence, to greater variability in results from CSF analysis. More recent revisions in diagnostic and research criteria have advocated the use of biomarkers as closely as possible to the pathological criteria for AD so as to increase the level of certainty in the diagnosis for research purposes (52, 53). All subjects with probable AD studied here were positive for the biomarker amyloid tau index (IATI), thus showing clear evidence of both

amyloid-β and tau neuropathology. Interestingly, glutamate and glutamine levels were inversely correlated with individual IATI values across all groups of subjects. We further note that we have previously reported measurements of CSF levels of D-serine, Lserine and glycine in the same patient cohort investigated in the current study (48). We found that D-serine levels in patients with probable AD were significantly higher than in healthy controls, but there were no differences in L-serine and glycine levels.

We further found elevated CSF glutamate and glutamine in major depression patients, similar to levels found in the AD group and significantly higher than the levels found in both healthy control individuals and hydrocephalus patients. Previous studies have measured glutamate and glutamine in the CSF of patients with major depression, and they report controversial findings. Levine et al. (32) reported increased glutamine levels in the CSF of depressed patients compared to controls, whereas Frye et al. (33) showed decreased glutamate levels in depression. More recent studies reported no differences in glutamate or glutamine levels in the CSF of depressed patients compared to controls (34, 35). Of note, Levine et al. (32), Frye et al. (33) and Garakani et al. (34) evaluated middle-aged adult patients, while here we studied older patients. On the other hand, Hashimoto et al. (35) also studied older patients, but the MMSE scores average of their study was slightly higher than in our study. More studies are thus warranted to clarify whether glutamate and glutamine are altered in individuals affected by depression and belonging to different age groups. Increased CSF glutamate may be related to the neurodegenerative process thought to occur in major depression. Evidence indicates that inflammation leading to neurodegeneration plays an important role in depression (54– 56). Glutamate excitotoxicity may be involved in this cascade as inflammatory mediators increase glutamate release and decrease glutamate uptake in the central nervous system (57–59).

CSF glutamate and glutamine levels were similar in agematched patients with probable AD or depression, suggesting another shared mechanism of pathogenesis in these two disorders. Accordingly, a prospective study in a large cohort suggested that depression in the older adult is a prodrome rather than a risk factor for AD (60). Given the central role of glutamatergic neurotransmission in synaptic plasticity, learning and memory (61), as well as in regulation of mood (62), the impact of altered glutamate and glutamine levels on neuropathological mechanisms connecting dementia and depression warrants further investigation. Notably, evidence from animals models indicates that AD and depression involves shared mechanisms, such as hyperactivation of NMDA receptors through increased D-serine (48), inflammatory process (10, 11) and serotonergic signaling (11).

What is the significance of increased glutamate and glutamine levels in the CSF of AD patients? The increase in CSF glutamine may be due to increased activity of aspartate aminotransferase, an enzyme that forms glutamine, in the AD brain (30). A possible mechanism to explain the increase in CSF glutamate involves the build-up of Aβ oligomers (AβOs) in the AD brain. AβOs

are increasingly recognized as proximal neurotoxins in AD (8, 9, 63, 64) and oligomer levels increase in AD brains. Of note, we have previously demonstrated that AβOs cause an increase in extracellular levels of glutamate in hippocampal neurons (65), and intracerebroventricular injection of AβOs induces both depressive-like behavior and cognitive deficits in mice (10, 11, 66, 67). It may thus be that AβO-instigated increases in brain glutamate levels underlie, at least in part, cognitive and mood alterations in AD.

Interestingly, we found that glutamate and (albeit somewhat less strongly) glutamine levels are significantly and inversely associated with the MMSE score in subjects without dementia (with MMSE above 23). This implies that alterations in CSF glutamate levels sensitively correlate with sub-clinical deterioration in cognitive performance. Thus, CSF measurements of glutamate/glutamine may serve as a biomarker of subtle cognitive changes that, while still within the range of normality, may reveal underlying mechanisms of pathogenesis potentially leading to future dementia. This could be of major clinical utility in terms of detecting pre-clinical dementia, with clear implications for inclusion of subjects in clinical trials and initiation of preventive/treatment strategies prior to overt cognitive deterioration.

A limitation of the present study was the absence of males in the group of patients with major depression. We note, however, that no sex differences in glutamate and glutamine levels were found in the other three patient groups studied. Additionally, the groups of patients with probable AD and major depression showed a clear trend (albeit not statistically significant) toward lower education than the control and hydrocephalus groups. However, analysis of differences in glutamate levels using education as a covariate confirmed that differences between patient groups remained significant (data not shown). Moreover, the cohort we have studied was of modest size, and it may not be representative of the population as a whole. However, the highly significant differences we have found between groups strongly suggest that the differences may hold in larger patient groups. Nevertheless, despite the current statistically robust results, we acknowledge that this is an initial investigation and, thus, a larger study is warranted to confirm and extend the validity of our findings.

In conclusion, glutamate and glutamine levels are increased in the CSF of patients with probable AD. Significantly increased glutamate levels were detected in CDR 0.5 patients, raising the possibility that determination of CSF glutamate levels might constitute an additional biomarker for preclinical cognitive deterioration or early stages of dementia. Finally, our finding that glutamate and glutamine levels were also increased in the CSF of older patients with major depression suggests that these amino acids may be involved in shared mechanisms of pathogenesis between AD and major depression.

### AUTHOR CONTRIBUTIONS

All authors certify that they have participated sufficiently in the work to take public responsibility for the content. RP and SF participated in the conception and design of study. CM, RP, and SF wrote the manuscript. CM and CV-L performed the analysis and interpretation of the data. CB, TR, and JL conduced analysis of patients.

# ACKNOWLEDGMENTS

We thank A. C. Rangel and A. Fantinatti for administrative and technical support. We also thank Dr. L. Scoriels for review of the manuscript. This research was supported by grants from DECIT/SCTIE/MS, SESDC, Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ) to RP and SF; Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) to RP, SF, and JL; Instituto Nacional de Ciência e Tecnologia de Biologia Estrutural e Bioimagem

(INBEB) to RP; and National Institute for Translational Neuroscience to SF. RP was recipient of a Long-Term Fellowship from the Human Frontier Science Program. CM and CV-L were supported by pre-doctoral fellowships from FAPERJ and CNPq.

### REFERENCES


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyt. 2018.00561/full#supplementary-material


Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. (2011) 7:263–9. doi: 10.1016/j.jalz.2011.03.005


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

Copyright © 2018 Madeira, Vargas-Lopes, Brandão, Reis, Laks, Panizzutti and Ferreira. 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.

# Astrocytic Regulation of Glutamate Transmission in Schizophrenia

### Yu-Ying Mei 1,2, Dong Chuan Wu1,2 and Ning Zhou1,2 \*

*<sup>1</sup> Translational Medicine Research Center, China Medical University Hospital, Taichung, Taiwan, <sup>2</sup> Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan*

According to the glutamate hypothesis of schizophrenia, the abnormality of glutamate transmission induced by hypofunction of NMDA receptors (NMDARs) is causally associated with the positive and negative symptoms of schizophrenia. However, the underlying mechanisms responsible for the changes in glutamate transmission in schizophrenia are not fully understood. Astrocytes, the major regulatory glia in the brain, modulate not only glutamate metabolism but also glutamate transmission. Here we review the recent progress in understanding the role of astrocytes in schizophrenia. We focus on the astrocytic mechanisms of (i) glutamate synthesis via the glutamate-glutamine cycle, (ii) glutamate clearance by excitatory amino acid transporters (EAATs), (iii) D-serine release to activate NMDARs, and (iv) glutamatergic target engagement biomarkers. Abnormality in these processes is highly correlated with schizophrenia phenotypes. These findings will shed light upon further investigation of pathogenesis as well as improvement of biomarkers and therapies for schizophrenia.

### Edited by:

*Kenji Hashimoto, Chiba University, Japan*

### Reviewed by:

*Jennifer C. Felger, Emory University, United States Kurt Leroy Hoffman, Autonomous University of Tlaxcala, Mexico*

> \*Correspondence: *Ning Zhou ningzhou.nrsc@gmail.com*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *12 July 2018* Accepted: *12 October 2018* Published: *06 November 2018*

### Citation:

*Mei Y-Y, Wu DC and Zhou N (2018) Astrocytic Regulation of Glutamate Transmission in Schizophrenia. Front. Psychiatry 9:544. doi: 10.3389/fpsyt.2018.00544* Keywords: schizophrenia, glutamate, NMDA receptors, excitatory amino acid transporters, glutamate-glutamine cycle, D-serine

# INTRODUCTION

Dysregulation of glutamatergic neurotransmission is critically implicated in the pathophysiology of psychotic disorders. In the central nervous system (CNS), glutamate is the principal excitatory neurotransmitter and mediates the fast excitatory transmission by activation of ionotropic glutamate receptors, including AMPA, kainate, and NMDA receptors (NMDARs). Astrocytes affect glutamatergic transmission in several important ways, including glutamate biosynthesis, glutamate-glutamine cycle, glutamate uptake, and releasing glutamate and D-serine as gliotransmitters. This review focuses on recent advances in astrocyte-mediated dysregulation of glutamate transmission in schizophrenia.

# THE GLUTAMATE HYPOTHESIS OF SCHIZOPHRENIA

The "glutamate hypothesis of schizophrenia" proposes that schizophrenia symptoms and cognitive impairment are due to hypofunction of NMDARs and excessive glutamate release, especially in brain areas including prefrontal cortex and hippocampus (1). This theory was initiated from the observation that NMDAR antagonists, like phencyclidine and ketamine, could evoke negative symptoms and cognitive dysfunction resembling schizophrenic phenotypes in healthy subjects (2). In schizophrenic subjects, subanesthetic doses of ketamine could exacerbate psychotic and cognitive symptoms (3, 4). Interestingly, in encephalitis patients, the first-episode psychosis is associated with the presence of anti-NMDAR antibodies that could cause a reduction of

**16**

surface expression of NMDARs (5, 6). In animal studies, suppression of NMDAR function by pharmacological or genetic approaches leads to schizophrenia-like behaviors (7, 8). Schizophrenia is also associated with dysregulation of some genes and/or proteins involved in glutamate transmission (9). Genomewide association studies have reported that the NMDAR subunits encoding genes, GRIN2A and GRIN2B, are schizophreniarelated genes (9, 10). Two de novo mutations in GRIN2A were found in sporadic schizophrenia patients (11). The single nucleotide polymorphisms of genes related to D-serine synthesis and metabolism, such as genes encoding serine racemase (SR), D-Amino Acid Oxidase (DAAO), and the DAAO activator G72, are also associated with schizophrenia (12).

The hypofunction of NMDAR causes excessive glutamate release, hyper-glutamatergic functions, and hypermetabolism (13). However, it remains largely unknown how NMDAR hypofunction leads to elevated glutamate levels and overactivation of non-NMDA glutamate receptors. One possibility is that interneurons are more sensitive to NMDAR blockade than principal glutamatergic neurons, and interneuron inhibition leads to an overall effect of disinhibition and hyperexcitation of glutamatergic output (**Figure 1**) (1, 14, 15). The increased glutamate levels have been documented by brain imaging studies and the source of excessive glutamate is mainly from presynaptic release (16). This notion was strengthened by the effect of drugs that specifically reduce presynaptic glutamate release, like the mGluR2/3 agonists, which significantly reverse ketamineevoked glutamate elevation and cerebral blood volume increase in the hippocampus (13). High extracellular glutamate levels lead to hypermetabolism, structural disorganization, and eventually hippocampal volume reduction, and these dysfunctions are correlated with disease progression from prodromal symptoms to psychosis (17). Nevertheless, many mechanisms that cause an overall increase in glutamate tone could contribute synergistically to schizophrenia pathogenesis.

# SYNAPTIC AND EXTRASYNAPTIC NMDARS

NMDARs are distributed at both synaptic and extrasynaptic sites with different subunit compositions. Their activation requires the binding of both glutamate and D-serine/glycine as a co-agonist. Synaptic NMDARs are activated by presynaptic glutamate release at low frequency, while intensive synaptic activation could lead to glutamate spillover and activation of extrasynaptic NMDARs (18). Synaptic NMDARs are important for synaptic plasticity and learning and memory, which is largely controlled by D-serine concentration within the synaptic cleft depending on astrocyte coverage and function (19, 20). Moreover, extrasynaptic NMDARs are in close proximity to synapse-enveloping astrocytes and are a preferential target for astrocyte-released glutamate, D-serine, and glycine (**Figure 2**). Activation of extrasynaptic NMDARs could generate two forms of responses. First, ambient glutamate could produce a tonic NMDAR current, whose amplitude is directly modulated by glial EAATs (21, 22). The tonic NMDAR currents contribute to neuronal excitability and their existence has been identified in multiple neuron types in the prefrontal cortex, hippocampus, and cerebellum (23–25). Second, activation of extrasynaptic NMDARs could produce "slow inward currents" (SICs), which are infrequent phasic events characterized by slow activation and decay kinetics, large amplitude, and insensitivity to sodium channel blockers. It is believed that SICs are due to astrocyteoriginated glutamate and the co-occurrence of SICs in adjacent neurons could lead to synchronized neuronal activity (26). It is also possible that SICs are generated by the phasic release of D-serine/glycine, which potentiates extrasynaptic NMDARs that are already tonically activated by low concentrations of ambient glutamate (18). While accumulating evidence has demonstrated a critical role of deficient synaptic NMDARs in schizophrenia (27, 28), alterations in the expression and function of extrasynaptic NMDARs are less understood. It was recently reported that the tonic NMDAR conductance was significantly larger and more sensitive to NMDAR blockers in interneurons compared with pyramidal neurons in the hippocampus (29). However, it remains to be determined whether extrasynaptic NMDAR deficiency occurs under schizophrenic conditions.

# ASTROCYTES AND GLUTAMATE TRANSMISSION

Regulation of glutamate transmission could occur at several steps, including synthesis, transportation, release, and clearance of glutamate, as well as activation of glutamate receptors by endogenous agonists and co-agonists. **Figure 2** summarizes some key mechanisms underlying the impact of astrocytes in glutamate transmission in the CNS. Astrocytes are the primary locus for the biosynthesis of glutamate from glucose. Through the tricarboxylic acid cycle, the glycolysis product pyruvate is converted into α-ketoglutarate, which is then catalyzed into glutamate by aspartate aminotransferase (30). Glutamate is catalyzed by glutamine synthetase (GS), which is exclusively expressed in astrocytes in the brain, to form glutamine that is further transferred to neurons. In glutamatergic neurons, glutamate is converted from glutamine by phosphateactivated glutaminase (30). After packaged into synaptic vesicles, glutamate is released as the neurotransmitter to active postsynaptic AMPA/kainate receptors to mediate fast excitatory synaptic transmission, and to activate NMDARs to depolarize the postsynaptic membrane potential and to allow Ca2<sup>+</sup> influx. The clearance of excessive glutamate from the synaptic cleft occurs rapidly, largely due to the function of four subtypes of EAATs (EAAT1-4), among which the EAAT1 (also termed glutamate-aspartate transporter, GLAST) and EAAT2 (also termed glutamate transporter 1, GLT-1) are mainly expressed in astrocytes (31). Once glutamate is taken up into astrocytes, it is converted into glutamine and re-enters the glutamateglutamine cycle. Besides their key roles in the biosynthesis of glutamate and maintenance of the glutamate-glutamine cycle, astrocytes can directly release glutamate as a gliotransmitters by either Ca2+-dependent vesicular exocytosis or by non-vesicular release (32, 33). Astrocytes are also a major source of glycine

and D-serine to potentiate NMDAR responses. Glycine is stored in astrocytes with high concentrations (3–6 mM) and can be released via the reversed operation of glycine transporters and other Ca2+-independent pathways (34–36). The synthesis of Dserine is initiated from the glycolytic intermediate metabolite 3 phosphoglycerate, which is synthesized into L-serine in several steps. D-serine is converted from L-serine by the action of SR and is released by Ca2+-dependent and/or -independent machineries from astrocytes (19, 37) [but see (38)]. Notably, these above processes might interact with one another; for example, the increased glutamate synthesis leads to elevated glutamate release (39), and activation of glutamate transports could evoke rapid glutamine release as a feedback mechanism (40).

# EAATS

The efficiency of glutamate clearance directly affects extracellular resting glutamate levels and kinetics of glutamate-mediated synaptic responses. Among the total four subtypes of brainexpressed EAATs, EAAT1, and EAAT2 are mainly expressed in astrocytes, and EAAT2 is the dominant transporter in adults and is responsible for about 90% of total glutamate uptake (41, 42). The EAATs are mainly responsible for clearing synaptic glutamate and shaping postsynaptic responses. Deficits in EAAT functions cause the persistently increased glutamate levels, which represent one of the most remarkable changes in the schizophrenic brain. Accumulating evidence has reported abnormal mRNA and/or protein expression levels of EAATs in the prefrontal cortex, hippocampus, and anterior cingulate cortex in the postmortem tissue from schizophrenia patients. While many studies have revealed an overall reduction of EAAT2 levels in schizophrenia, some other studies reported unchanged or increased EAAT2 levels in certain brain regions [for review see (43)]. In a recent study using the laser-capture microdissection technique to separately harvest glutamatergic relay neurons and astrocytes from the mediodorsal nucleus of

alanine-serine-cysteine transporter; xCT, cystine-glutamate antiporter; GLAST, glutamate aspartate transporter; GLT-1, glutamate transporter 1; GlyT-1, glycine transporter 1; mGluR, metabotropic glutamate receptor; IP3R, inositol trisphosphate receptor.

thalamus, it was reported that changes of EAAT levels could be cell-specific: mRNA expression of EAAT1 was decreased in astrocytes, whereas EAAT2 was increased in excitatory relay neurons in schizophrenia (44). The upregulation of EAATs in neurons is possibly due to a compensatory response to the reduced EAAT levels in astrocytes. Notably, the mRNA levels of EAAT2b were remarkably increased in anterior cingulate cortex pyramidal neurons and thalamus relay neurons in schizophrenia (44, 45). EAAT2b is a splicing variant of EAAT2. It contains the PDZ domain-binding motif and possibly interacts with neuronal scaffolding proteins like PSD-95 (45–47). The EAAT2b proteins are localized predominantly at the cell surface and their distribution can be regulated by intracellular Ca2<sup>+</sup> and CaMKII (48), suggesting that EAAT2b might exhibit neuronal specific functions.

The abnormal EAAT levels are highly associated with schizophrenia phenotypes. Mice lacking EAAT1 displayed schizophrenia-like behavioral changes: they showed higher locomotor activity in the novel environment but not in the home cage, and had increased sensitivity to the locomotor hyperactivity induced by NMDA antagonists (49). The locomotor hyperactivity of EAAT1 knockout mice was reversed by either the antipsychotic haloperidol or the mGlu2/3 agonist LY379268 (49). These mice also exhibited abnormalities in behavioral tests that measure the negative and cognitive symptoms of schizophrenia, including poor nesting behaviors, lesser preference for novel social stimulus albeit normal overall social interaction, a significant reduction in acoustic startle amplitude, and impaired learning in an instrumental visual discrimination task, suggesting that EAAT1 dysfunction could generate certain behavioral changes resembling schizophrenia phenotypes (49, 50). Knockout of EAAT2 caused lethal spontaneous seizures, high susceptibility to brain injury, and short life-span in the homozygous mice (51). The heterozygous EAAT2 knockout mice had less severe abnormalities, and they exhibited increased locomotion activity in a new environment and greater freezing responses in the fear-conditioning test compared to the wildtype (52). The EAAT1/2 double knockout mice showed multiple brain defects resembling developmental defects found in schizophrenia (53). In a recent study, the inducible EAAT2 knockout mice were generated in a temporal controlled, cell-specific manner. The study demonstrated that knockout during adolescence caused 60–80% reduction of EAAT2 selectively in astrocytes, and the animals showed no lethal seizures or neuronal loss. Interestingly, these animals exhibited pathological repetitive behaviors, like excessive self-grooming and tic-like head shakes, whereas they had no obvious anxiety or social abnormality (54).

It remains unknown how EAAT abnormality leads to schizophrenia-relevant behaviors. In hippocampus, pharmacological inhibition of EAAT functions increased the peak amplitude of NMDAR-mediated EPSCs and prolonged their decay kinetics, but had no changes in AMPA receptormediated EPSCs, possibly due to the fast desensitization of AMPA receptor currents (55). It has been generally considered that EAATs could limit glutamate spillover to neighboring synapses and enhance synapse independence, which is related to synaptic plasticity as well as learning and memory (31). Reduced expression of glial EAATs will not only impair these important physiological functions but also increase the susceptibility of the brain to injury and cell death (51). Neuronal EAATs might have much lower efficiency than the astrocytic transports: glutamate is quickly converted into glutamine by GS, resulting in low free glutamate concentrations in astrocytes; whereas neurons are liable to accumulate intracellular glutamate concentration and thereby lead to poor operation of neuronal EAATs. Moreover, the action of EAATs will produce a non-stoichiometrically coupled anion current (56, 57), which contributes to the regulation of neuronal membrane potentials and transmitter release (58). In the zebrafish photoreceptor synapse, the EAAT2b isoform mediates a large-conductance Cl<sup>−</sup> current, potentially capable of affecting resting membrane potentials (59). Nevertheless, these mechanisms remain to be determined in schizophrenia and may vary depending on the brain region and disease progression.

### THE GLUTAMATE-GLUTAMINE CYCLE

Schizophrenia is associated with the abnormal glutamateglutamine cycle. Altered levels of brain metabolites including glutamate, GABA, and glutamine have been well-documented in human studies and these alterations can be detected in different brain regions and the cerebrospinal fluid (CSF) (60). Researchers using proton magnetic resonance spectroscopy (1H-MRS) have reported abnormal glutamine and/or glutamate levels (some studies report glutamate+glutamine levels in combination depending on the <sup>1</sup>H-MRS approach and field strength), especially in the brain regions involving the medial temporal lobe (including hippocampus), basal ganglia, and thalamus of schizophrenia patients (61). An elevation of the glutamate+glutamine level was also observed in hippocampus of healthy humans receiving ketamine administration (62). The CSF glutamine/glutamate ratio is higher in the first-episode patients compared with normal controls (63). GS, the key enzyme involved in the glutamate-glutamine cycle, exhibits lower protein levels in spite of increased expression of mRNA in schizophrenia; but the results were less consistent among studies [for review see (43)]. Elevated glutamine levels are also directly related to psychotic symptoms and neuropsychological tests (64, 65). Although changed glutamine levels are supposedly related to functional alterations of astrocytes, it is notable that the <sup>1</sup>H-MRS and CSF data cannot provide a precise measure of intracellular vs. extracellular substrates and further evidence is needed to establish the causality.

Disturbance of the glutamate-glutamine cycle and downregulation of GS have also been reported in human epilepsy (66, 67). Reduced GS activity will build up intracellular glutamate in astrocytes, leading to reduced uptake capacities of EAATs. Moreover, glutamine produced by astrocytic GS is also one of the major sources for maintaining synaptic vesicle content of GABA in inhibitory interneurons (68). Pharmacological inhibition of GS reduced inhibitory postsynaptic currents in hippocampal pyramidal neurons and was reversed by exogenous application of glutamine (68). Similar dysfunction of the inhibitory synaptic response was also found in neurons near reactive astrocytes, which showed profound decreases in GS expression levels (69). These effects increased network hyperexcitability and produced spontaneous recurrent seizures (70). The loss of interneurons and deficient cortical GABA synthesis are also one of the most robust pathologies of schizophrenia (71–73). The reduced parvalbumin interneurons are associated with impaired oscillatory activity in hippocampus and imbalance of cortical excitation-inhibition (72, 74). Impaired parvalbumin interneuron activity by selective deletion of ErbB4 or dopamine D2 receptors produced schizophrenia-like behaviors in mice (75, 76). Dysfunction of neuron-astrocyte interaction has been suggested in schizophrenia models (77). However, it remains unclear whether there is any causative relationship between the dysfunction of the glutamate-glutamine cycle and interneuron deficiency.

# D-SERINE

D-serine plays an important role in the dysregulation of NMDAR functions in schizophrenia (78). This was supported by several lines of evidence, including the reduced serum and CSF D-serine levels (79, 80), a genetic association of SR and DAAO (the enzyme responsible for the degradation of D-serine) polymorphisms in schizophrenia patients (12, 81), the reversal effects of D-serine on schizophrenia-like behaviors in animals (82), and the beneficial effects of clinical D-serine-targeting therapies in patients (83, 84).

Recent studies revealed that the abnormality in astrocytereleased D-serine plays a critical role in the DISC1-associated pathological processes. The DISC1 gene was identified from patients with familial mental disorders and has been considered a critical susceptibility gene for schizophrenia (85, 86). The density of DISC1-expressing astrocytes is largely reduced in the dentate gyrus of hippocampus in schizophrenia patients compared with healthy controls (87). Selective expression of mutant DISC1 in astrocytes decreased protein levels of SR by increasing its ubiquitination (88). Astrocytes expressing mutant DISC1 caused less elaborated dendritic arborization and decreased the density of excitatory synapses in co-cultured normal neurons (89). The downregulation of SR resulted in a significant decline in D-serine levels, an enhanced locomotor activity, and impaired responses to pre-pulse inhibition in the mutant mice (88). Mice carrying the astrocytic specific mutation of DISC1 also exhibited increased anxiety, attenuated social interaction and preference for social novelty, and impaired cognitive behaviors (90). These behavioral abnormalitiesin DISC1 mutant mice were reversed by treatments of D-serine (88, 90), indicating a strong correlation with NMDAR hypofunction.

The α7 subunit-containing nicotinic acetylcholine receptor (α7 nAChR) has been considered a strong genetic contribution to schizophrenia (91). It has been reported that activation of α7 nAChR contributes to dopamine release and induction of longterm potentiation (91). Interestingly, α7 nAChRs are expressed not only in neurons but also in astrocytes (92–94). ACh could induce the intracellular Ca2<sup>+</sup> elevation in astrocytes by activation of α7 nAChRs and facilitate the synthesis and release of D-serine (94–97). Papouin et al. recently discovered that astrocytic α7 nAChRs activation could drive the vesicular release of D-serine from astrocytes to generate a wakefulness-dependent D-serine oscillation in z (98). The wakefulness and activity but not the circadian rhythms of the animal promote D-serine fluctuations over the 24 h period, during which D-serine was accumulated during the wakefulness and declined during sleep. Such an oscillation of the D-serine level is associated with NMDAR activity and the learning and memory behavior of the animal, consistent with the involvement of NMDARs in the cognitive deficits in schizophrenia. Astrocytes appear to play a central role since D-serine accumulation during the wakefulness is impaired by disruption of astrocytic vesicular exocytosis machinery or by selective deletion of α7 nAChRs from astrocytes rather than neurons. More importantly, EVP-6124, an α7 nAChR modulator tested in schizophrenia clinical trials, is able to promote D-serine release and enhance NMDAR activity (98).

# POTENTIAL BIOMARKERS

Several types of translational biomarkers have been developed to assess prognosis and monitor disease progression and treatment in schizophrenia. First, functional magnetic imaging (fMRI) blood-oxygenation-level dependent response (BOLD) has been applied to test brain physiology following acute NMDAR antagonist administration. NMDAR antagonists evoke robust changes in relative cerebral blood volume and local metabolism, which is largely due to the energy consumption for increased glutamate release and uptake (13, 99). The second approach applies <sup>1</sup>H-MRS to measure glutamine and/or glutamate levels in the brain (16, 100). As described above, abnormal glutamine and/or glutamate indices were found in brain regions of individuals with schizophrenia compared with healthy volunteers (61). The third method utilizes taskbased fMRI to evaluate BOLD responses in hippocampus and dorsolateral prefrontal cortex (101). The most adapted task is item-specific encoding task, which is designed to assess contributions of specific encoding and retrieval processes to episodic memory (102). Schizophrenia patients with more severe negative symptoms exhibited poor BOLD responses during the encoding and retrieval of episodic memory (101, 103). And fourth, positron emission tomography (PET) technique has been used to evaluate glutamatergic, GABAergic, and monoamine neurotransmission in various brain functions and have established the association between these responses and cognitive functions in schizophrenia (104, 105). Finally, the severity of schizophrenia symptoms is associated with abnormal serum levels of several glutamate transmission related factors, including brain-derived neurotrophic factor (BDNF), vascular endothelial growth factor (VEGF), D-serine, G72, and multiple inflammatory factors (106–108). A recent clinical study by Javitt et al. compared the first three types of glutamatergic targetengagement biomarkers in healthy volunteers receiving ketamine or placebo infusions (109). The data revealed a significant increase in the NMDAR antagonist-evoked BOLD responses. The <sup>1</sup>H-MRS results showed a smaller but significant change in glutamate+glutamine levels. However, the data from taskbased fMRI failed to yield a significant difference between the ketamine and placebo group (109). These observations indicate the fMRI BOLD as a particularly relevant biomarker for NMDAR antagonist-induced responses. The potential application of one or combined biomarkers for more refined prognostic classification of schizophrenia remains to be further assessed.

# FUTURE TREATMENTS

Based on the glutamate hypothesis, many drug targets have been proposed and some are under investigation. For example, drugs targeting the co-agonist binding site of NMDARs are of the primary interest [for review see (110)]. Direct enhancement of NMDAR functions with glycine and D-serine has been used in multiple clinical trials, among which one recent study showed improved negative symptoms in individuals at clinical high risk of schizophrenia (111). GlyT1 inhibitors, like sarcosine or Bitopertin, which are supposed to increase synaptic glycine levels, have been found effective in several trials (112). Other treatments to increase D-serine/glycine levels, like the DAAO inhibitor sodium benzoate, have demonstrated beneficial effects as an addon drug (83, 84). The mGluR2/3 agonists have been shown to inhibit excessive glutamate release in animal studies, and the tested drug LY2140023 has produced significant improvement in patients (113). Moreover, the α7 nAChR has attracted much attention; several drugs targeting this receptor are under clinical investigation (110). Besides these, more specific regulation of certain proteins in the glutamate transmission, such as selective NMDAR subtypes or EAAT isoforms, might represent potential targets for future drug development.

### CONCLUSIONS

The glutamate hypothesis has substantially advanced our understanding of the pathogenesis of schizophrenia. Although glutamate transmission is mediated by ionotropic glutamate receptors and involves other pre- and post-synaptic components at excitatory synapses, astrocytes regulate glutamate metabolism and shape glutamate transmission in several important aspects. This review has focused on some recent studies that explicitly report the important role of glutamate uptake, glutamateglutamine cycle, and astrocyte-derived D-serine in the etiology of schizophrenia. Neuroimaging and neurochemical biomarkers based on glutamate transmission and metabolism have also

### REFERENCES


demonstrated translational utility. Further understanding of neuron-astrocyte interaction during these processes will be critical to the development of diagnostic and therapeutic avenues for schizophrenia and psychotic disorders.

### AUTHOR CONTRIBUTIONS

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

### ACKNOWLEDGMENTS

This study was supported by grants to NZ and DW from the Ministry of Science and Technology (MOST 107-2320-B-039-060-MY3 and MOST 107-2320-B-039-061-MY3) and China Medical University Hospital (DMR-106-113).


neurosecretory neurons in hypertensive rats. J Physiol. (2017) 595:4647–61. doi: 10.1113/JP274327


pyramidal cell population in schizophrenia. Transl Psychiatry (2015) 5:e579. doi: 10.1038/tp.2015.74


parameters in schizophrenia. Schizophr Res. (2007) 90:41–51. doi: 10.1016/j.schres.2006.10.010


hippocampal NMDA receptors to wakefulness. Neuron (2017) 94:840–54.e7. doi: 10.1016/j.neuron.2017.04.021


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

# The Emerging Role of Altered D-Aspartate Metabolism in Schizophrenia: New Insights From Preclinical Models and Human Studies

Francesco Errico<sup>1</sup> \*, Tommaso Nuzzo<sup>2</sup> , Massimo Carella<sup>2</sup> , Alessandro Bertolino<sup>3</sup> and Alessandro Usiello4,5 \*

<sup>1</sup> Department of Agricultural Sciences, University of Naples "Federico II", Portici, Italy, <sup>2</sup> Translational Neuroscience Unit, IRCCS Casa Sollievo Della Sofferenza, San Giovanni Rotondo, Italy, <sup>3</sup> Group of Psychiatric Neuroscience, Department of Basic Medical Science, Neuroscience and Sense Organs, Aldo Moro University, Bari, Italy, <sup>4</sup> Laboratory of Behavioural Neuroscience, Ceinge Biotecnologie Avanzate, Naples, Italy, <sup>5</sup> Department of Environmental, Biological and Pharmaceutical Sciences and Technologies, Università Degli Studi Della Campania "Luigi Vanvitelli", Caserta, Italy

### Edited by:

Hsien-Yuan Lane, China Medical University, Taiwan

### Reviewed by:

Chih-Min Liu, National Taiwan University Hospital, Taiwan Marek Schwendt, University of Florida, United States Maarten Van Den Buuse, La Trobe University, Australia

### \*Correspondence:

Francesco Errico francesco.errico@unina.it Alessandro Usiello usiello@ceinge.unina.it

### Specialty section:

This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry

Received: 20 July 2018 Accepted: 17 October 2018 Published: 06 November 2018

### Citation:

Errico F, Nuzzo T, Carella M, Bertolino A and Usiello A (2018) The Emerging Role of Altered D-Aspartate Metabolism in Schizophrenia: New Insights From Preclinical Models and Human Studies. Front. Psychiatry 9:559. doi: 10.3389/fpsyt.2018.00559 Besides D-serine, another D-amino acid with endogenous occurrence in the mammalian brain, D-aspartate, has been recently shown to influence NMDA receptor (NMDAR) mediated transmission. D-aspartate is present in the brain at extracellular level in nanomolar concentrations, binds to the agonist site of NMDARs and activates this subclass of glutamate receptors. Along with its direct effect on NMDARs, D-aspartate can also evoke considerable L-glutamate release in specific brain areas through the presynaptic activation of NMDA, AMPA/kainate and mGlu5 receptors. D-aspartate is enriched in the embryonic brain of rodents and humans and its concentration strongly decreases after birth, due to the post-natal expression of the catabolising enzyme D-aspartate oxidase (DDO). Based on the hypothesis of NMDAR hypofunction in schizophrenia pathogenesis, recent preclinical and clinical studies suggested a relationship between perturbation of D-aspartate metabolism and this psychiatric disorder. Consistently, neurophysiological and behavioral characterization of Ddo knockout (Ddo−/−) and D-aspartate-treated mice highlighted that abnormally higher endogenous D-aspartate levels significantly increase NMDAR-mediated synaptic plasticity, neuronal spine density and memory. Remarkably, increased D-aspartate levels influence schizophrenia-like phenotypes in rodents, as indicated by improved fronto-hippocampal connectivity, attenuated prepulse inhibition deficits and reduced activation of neuronal circuitry induced by phencyclidine exposure. In healthy humans, a genetic polymorphism associated with reduced prefrontal DDO gene expression predicts changes in prefrontal phenotypes including greater gray matter volume and enhanced functional activity during working memory. Moreover, neurochemical detections in post-mortem brain of schizophrenia-affected patients have shown significantly reduced D-aspartate content in prefrontal regions, associated with increased DDO mRNA expression or DDO enzymatic activity. Overall, these findings suggest a possible involvement of dysregulated embryonic D-aspartate metabolism in schizophrenia pathophysiology and, in turn, highlight the potential use of free D-aspartate supplementation as a new add-on therapy for treating the cognitive symptoms of this mental illness.

Keywords: D-aspartate, D-serine, schizophrenia, NMDA receptor, D-aspartate oxidase, mouse models

# INTRODUCTION

Although L-amino acids are mostly used for protein synthesis and metabolic processes in eukaryotes, some free amino acids are present in a substantial amount in D configuration in mammalian tissues, including humans. In particular, free D-aspartate (D-Asp), and D-serine (D-Ser) are enriched in the brain where they emerge in an age- and region-dependent manner (1–5). Since its discovery at the beginning of the 90s (6), D-Ser has been extensively studied in the mammalian brain. Nowadays, it is well known that D-Ser is an endogenous N-Methyl D-Aspartate (NMDA) receptor (NMDAR) co-agonist that regulates the activation of glutamatergic excitatory synapses (7–10), thus influencing different NMDAR-dependent functions, such as brain development (11), synaptic transmission and plasticity (12–17), and behaviors (18–20). In this line, alteration of D-Ser metabolism has been linked to pathological conditions associated with NMDAR dysfunction, including schizophrenia (21–23). Unlike D-Ser, the neurobiological role of D-Asp has been most significantly studied only from the last decade, so it is still poorly defined. Nonetheless, we know today that D-Asp is able to directly modulate NMDAR-mediated transmission and functions in rodents, and dysregulation of its metabolism occurs in the brain of schizophrenia patients. In this review, we will summarize the evidence collected in both animal models and humans that have led to hypothesize a role for this atypical amino acid in schizophrenia pathophysiology.

### PRESENCE OF FREE D-ASPARTATE AND ITS METABOLIC REGULATION IN THE MAMMALIAN BRAIN

The presence of free D-Asp in the mammalian brain was reported for the first time by Dunlop et al. in the mid-80s (1). This pioneering observation has been followed by many other studies in rodents and humans that overall showed a transient emergence of D-Asp in the brain, characterized by a peak of concentration during developmental stages and a drastic fall after birth (1– 4, 24, 25). Among these studies, HPLC detections obtained by Hashimoto et al. revealed that the amount of D-Asp in the human prefrontal cortex (PFC) homogenates even exceeds that of the corresponding L-form at gestational week 14 (3). Consistent with this finding, in the rat brain D-Asp appears at embryonic day 12 (E12) selectively in non-proliferating neuroblasts but not in mitotic cells of the ventrocaudal forebrain, midbrain, and hindbrain (24). In the ventral side of the forebrain, D-Asp emerges in cell bodies of migrating neuroblasts and then shifts to axons once these cells have reached the outer layer of neural epithelium. Between E14 and E20, D-Asp concentration increases and spreads to the whole brain. In particular, in the cerebral cortex, D-Asp is present in both the cell bodies and projections of neuroblasts (24). In another work, Snyder et al. revealed that D-Asp is present in a substantial amount at birth in the forebrain regions, like the cortex and part of the midbrain (25). At post-natal day 2, D-Asp extends to the hindbrain and cerebellum and is enriched in the subventricular zone and the cortical plate of the cerebral cortex, the CA1-CA3 area and the dentate gyrus of the hippocampus, and the external granular layer of the cerebellum. In this last brain region, D-Asp staining is associated with granule cells that have not yet migrated to their definitive location (25). One week after birth, D-Asp decreases and almost disappears in one-month-old brains (26). Notably, in both prenatal and postnatal phases, D-Asp seems to be present exclusively in neurons but not in glial cells (25, 26).

The peculiar spatiotemporal occurrence of D-Asp in the embryonic brain suggests that this atypical molecule is mainly produced through a de novo biosynthetic mechanism (**Figure 1**). Accordingly, a time-dependent accumulation of D-Asp was demonstrated in PC12 cells cultured in D-Asp-free medium (27, 28). Further evidence resulted from primary cultures of rat embryonic neurons in which [14C]-D-Asp was produced starting from [14C]-L-Asp precursor (25). A pyridoxal 5'-phosphate (PLP)-dependent glutamate-oxalacetate transaminase 1-like 1 (Got1l1) was identified as the enzyme that converts L-Asp to D-Asp in the adult mouse brain (29). However, subsequent findings have demonstrated that knockout of Got1l1 gene does not produce any change in cerebral D-Asp levels of mutant mice (30). Interestingly, two studies have shown that serine racemase (SR), the only known enzyme responsible for D-Ser biosynthesis in the mammalian brain (31), could be involved also in D-Asp production, at least in some brain areas (32, 33). Indeed, Sr knockout mice display considerably decreased D-Asp levels in the forebrain regions (around 50–60 %) but not in the cerebellum (32, 33). Remarkably, overexpression of SR in rat PC12 cells resulted in intracellular D-Asp increase while genetic ablation of Sr in these cells did not produce changes in this D-amino acid levels (33), further supporting that besides SR an additional enzymatic activity for D-Asp biosynthesis is present in mammals.

While D-Asp biosynthetic pathway is still unclear, the enzyme responsible for D-Asp degradation, D-aspartate oxidase (DDO, or DAPOX EC 1.4.3.1), has been well known for a long time (34). DDO is a peroxisomal flavoenzyme (35, 36) that oxidizes D-Asp in presence of H2O and O2, producing α-oxaloacetate, H2O2, and NH<sup>+</sup> 4 (**Figure 1**) (37–39). DDO is inactive toward D-Ser (40) that is degraded by its homologous flavoenzyme, D-amino acid

a methyl donor. See the main text for further descriptions.

oxidase (DAAO, EC 1.4.3.3) (41–43). A specific tripeptide signal at the C-terminus of DDO targets this enzyme to peroxisomes (40, 44), where the toxic H2O<sup>2</sup> produced by D-Asp oxidation can be safely removed by peroxisome-resident catalases (45).

Consistent with a primary role for DDO activity in controlling its substrate concentration, it has been reported that spatial and temporal expression of this enzyme is reciprocal to D-Asp (26, 35, 46). During mouse lifespan, the postnatal decrease of cerebral D-Asp is mirrored by a time-dependent increase of Ddo gene expression, as Ddo mRNA levels are very low from E14 to 1 week of age and then strongly rise during the following weeks until adult phase (46). Interestingly, the postnatal increase of Ddo transcript levels is temporally associated with concomitant dramatic demethylation of the putative promoter region of Ddo gene. In agreement with an influence of DNA methylation on Ddo mRNA expression, it has been shown that treatment with the DNA demethylating agent, azacitidine, triggers the expression of Ddo transcript in embryonic cortical neurons, which do not express Ddo gene in physiological conditions (46). Notably, the temporal pattern of Ddo mRNA expression reported in the mouse brain matches closely with the substantial increase in DDO enzymatic activity found in the rat brain during the first weeks of life (35). Like D-Asp, also DDO is prominently localized in neurons, and only marginally expressed in glial population (36). Overall, reciprocal spatiotemporal localization of DDO and D-Asp suggests that during adulthood this enzyme is necessary to remove its endogenous substrate from brain regions where D-Asp is no more required and, indeed, could be functionally detrimental and neurotoxic for neuronal activity and survival, as found in the brain of elderly constitutive knockout mice for Ddo gene (Ddo−/−) (46–48). The substantial and persistent accumulation of <sup>D</sup>-Asp in the brain of Ddo−/<sup>−</sup> mice further suggests that DDO is the only enzyme that catalyzes the endogenous degradation of D-Asp, throughout the entire animal lifespan (47, 49, 50).

# PHARMACOLOGICAL FEATURES OF D-ASPARTATE

Pioneering neuropharmacological studies performed in the second half of the 80s, aimed at finding molecules with agonistic or antagonistic activity for L-glutamate (L-Glu) receptors, highlighted that D-Asp is able to bind to the L-Glu site of NMDARs (51–55). In particular, one of these works revealed that the efficiency of D-Asp in displacing the NMDAR antagonist, D-AP5, in rat brain membranes is the same as NMDA and around 10 times lower than L-Glu (54). Many years later, a series of electrophysiological observations demonstrated that D-Asp not only binds to NMDARs but also activates these receptors. Indeed, local application of D-Asp on adult mouse brain slices triggers inward currents in both hippocampal CA1 pyramidal neurons and striatal GABAergic medium spiny neurons, which are strongly reduced by the competitive and non-competitive NMDAR antagonists, D-AP5 and MK-801, respectively (56, 57). D-Asp activates NMDARs through the binding to each of the GluN2 subunits (58), and induces the transient increase of intracellular Ca2<sup>+</sup> (47). Notably, a small but significant percentage of the currents triggered by D-Asp application still endure even after application of high doses of MK-801 (47, 56, 58). The persistence of D-Asp-mediated currents under conditions of NMDAR blockade suggests that this D-amino acid may be able to affect, at least in part, additional receptor complexes. Accordingly, D-Asp can also stimulate metabotropic mGlu5 receptors in mouse and rat brain (59–61).

The endogenous occurrence of D-Asp in the brain and its main ability to activate NMDARs are consistent with an involvement of this D-amino acid in the in vivo modulation of glutamatergic synaptic functioning. In agreement with this idea, early studies have suggested that D-Asp can transit through the extracellular space by transporter mechanisms enabling its release and reuptake. Accordingly, experiments using different ex vivo preparations (tissue slices, cells or synaptosomes) demonstrated that D-Asp is released through exocytotic processes mediated by vesicular Ca2<sup>+</sup> (25, 62–65). <sup>D</sup>-Asp is contained in membrane-bound organelles, identified as secretory granules (in PC12 cells) (64) or synaptic vesicles (in synaptosome preparation) (66), in which this D-amino acid is actively stored through still unknown transporter systems. In line with in vitro data, Usiello et al. confirmed the existence of Ca2+-dependent <sup>D</sup>-Asp release in the cortex of living mice. Indeed, in microdialysis studies performed in the PFC of freely moving animals they showed that D-Asp is present in the extracellular space at nanomolar concentration (46, 61) and when dialysates were collected in Ca2+-free artificial cerebrospinal fluid, extracellular concentration of D-Asp dropped below the limit of HPLC detection. Extracellular D-Asp has been also recently demonstrated in the brain of domestic chicks, where it decreases in an age-dependent manner and is transiently induced following high K<sup>+</sup> stimuli (67). The recent evidence that <sup>D</sup>-Asp occurs also in the extracellular space confirms the idea that this Damino acid has the potential to bind to NMDARs also in vivo and, in turn, to affect the glutamatergic neurotransmission dependent on this receptor subclass. In addition to Ca2+-dependent efflux system, further studies reported that D-Asp may also be released through spontaneous processes (25, 68, 69) or mechanisms involving L-Glu exchange (70, 71).

Different studies suggest that intracellular uptake of D-Asp may depend by L-Glu/L-Asp transporter system, which moves excitatory amino acids against their concentration gradient using the Na+/K<sup>+</sup> electrochemical gradient. Interestingly, while on one side this carrier system recognizes only the L-enantiomer of Glu, on the other it is able to take up both Asp enantiomers with approximately the same efficiency (72). In line with an involvement of L-Glu/L-Asp transporters in the reuptake of D-Asp, autoradiographic and immunostaining experiments have shown that D-Asp preloading on rodent hippocampal and cerebellar slices causes a strong labeling in glutamatergic axons and in surrounding glial processes (73–75). No direct evidence is so far available about the existence of selective D-Asp reuptake mechanisms in the in vivo brain. Nonetheless, the drastic and fast reduction of cortical extracellular D-Asp after experimental removal of Ca2<sup>+</sup> from the artificial cerebrospinal fluid reported in microdialysis studies (46, 61), seems to suggest the presence of an active and efficient mechanism of D-Asp clearance in the mammalian brain.

### PRECLINICAL MODELS WITH A NON-PHYSIOLOGICAL INCREASE OF D-ASPARTATE CONTENT IN THE BRAIN

The generation of Ddo knockout mice (Ddo−/−) in two independent laboratories (49, 50) has been a turning point for the comprehension of the in vivo role of D-Asp and its catabolizing enzyme. Both Ddo−/<sup>−</sup> mouse lines display higher <sup>D</sup>-Asp levels in the brain and peripheral organs, but comparable L-Asp (49, 50) and L-Glu content (50), compared to their respective controls. In more detail, HPLC detection in homogenates deriving from the hippocampus, striatum, cortex, cerebellum and olfactory bulbs showed ∼10- to 20-fold increase in <sup>D</sup>-Asp levels in Ddo−/<sup>−</sup>

mice, compared to wild-type animals (46, 47, 56, 57). Along with D-Asp, also the levels of the N-methyl derivative of D-Asp, NMDA, were increased in Ddo−/<sup>−</sup> brain homogenates (47, 49). In this regard, biochemical findings indicated that cerebral NMDA could be generated through the transfer of a methyl group from S-adenosyl methionine to D-aspartate catalyzed by a methyltransferase activity (**Figure 1**) (76). In line with the rise of total D-Asp levels in brain homogenates, recent evidence on PFC dialysates has indicated that Ddo−/<sup>−</sup> mice display significantly increased D-Asp concentration also in the extracellular fraction, compared to Ddo+/<sup>+</sup> littermates (a rise of about five times, ∼100 nM in Ddo−/<sup>−</sup> vs. ∼20 nM in Ddo+/+) (46).

Besides Ddo gene targeting, another experimental approach has been widely used in the last years to raise the cerebral levels of D-Asp. This approach consists of the oral administration of D-Asp to animals for periods of one-two months until the age of two-three months. In relation to the brain region analyzed, such a treatment produces an approximately 2- to 5-fold increase in D-Asp content in homogenate samples, compared to the same areas of untreated mice (56–58, 61). Such rise is substantially lesser than that produced by Ddo gene targeting since the DDO enzyme expressed in the brain of adult D-Asp-treated mice is able to limit the further increase of this D-amino acid levels. In line with the effect of Ddo gene ablation, also the exogenous administration of D-Asp is able to induce a significant extracellular increase of this molecule in the PFC of freely moving mice, in both chronic and acute conditions (61). In particular, chronic oral administration of D-Asp for 1 month produces an increase in the same order of magnitude as Ddo knockout (∼80 nM in treated mice vs. ∼20 nM in untreated mice). On the other hand, acute D-Asp injection (at the dose of 500 mg/kg) is able to trigger a transient and rapid elevation of D-Asp levels already 20 min post-administration (from ∼20 to ∼500 nM). Consistent with the observations of Usiello et al. (61), a very recent study showed that intragastric infusion of D-Asp to rats reaches the hippocampus in 15 min via blood circulation (77). Taken together, these findings have a potential translational relevance as they highlight the ability of D-Asp to cross the blood-brain barrier, as described so far only for the D-stereoisomers of Ser and proline (78, 79). Strikingly, both chronic oral administration and acute injection of D-Asp are able to evoke a significant prefrontal efflux of L-Glu (48, 61) through the presynaptic activation of NMDA, mGlu5, and AMPA/kainate receptors triggered by this D-amino acid (61). Overall, these data indicate that D-Asp could influence glutamatergic neurotransmission not only through its direct binding to the L-Glu site of postsynaptic NMDARs but also through its ability to evoke the presynaptic release of endogenous L-Glu in selective brain regions.

## D-ASPARTATE AFFECTS FUNCTIONAL AND STRUCTURAL NEURONAL PROPERTIES DEPENDENT ON NMDA RECEPTORS ACTIVATION

The results described so far suggest that D-Asp stimulates also in vivo the glutamatergic transmission in adult rodents. In this regard, electrophysiological evidence obtained in the last 10 years has revealed that higher D-Asp content enhances NMDARdependent early-phase and late-phase long-term potentiation (E-LTP and L-LTP, respectively) in the CA1 area of the hippocampus of both adult Ddo−/<sup>−</sup> and chronically <sup>D</sup>-Asptreated mice (47, 56, 58, 80). Interestingly, the induction of E-LTP protocol, which causes a decaying LTP in Ddo+/<sup>+</sup> and untreated mice, is sufficient to maintain an enduring L-LTP (lasting for more than 160 min following tetanic stimulation) in Ddo−/<sup>−</sup> and <sup>D</sup>-Asp-treated mice (80). In both models, <sup>D</sup>-Asp-dependent L-LTP is insensitive to rapamycin administration but is suppressed by cytochalasin D (80), a potent inhibitor of actin polymerization. Moreover, chronic treatment with D-Asp increases the frequency of NMDAR-mediated miniature excitatory post-synaptic currents in pyramidal neurons of the medial PFC layer II/III (80). Remarkably, in line with the enhanced glutamatergic transmission, mice chronically treated with D-Asp display stronger metabolic activity in frontohippocampal areas, measured by basal cerebral blood volumeweighted functional magnetic resonance imaging (fMRI) (80). In this regard, increased synchronization between the hippocampus and cortical regions has been also recently produced through D-Asp gavage in awake rats subjected to blood oxygen level dependent (BOLD) fMRI (77). Changes in synaptic functioning are commonly associated with structural synaptic variations in dendritic morphology (81, 82). This relationship is observed also in the brain of Ddo−/<sup>−</sup> and <sup>D</sup>-Asp-treated mice, in which facilitated induction of late-phase synaptic plasticity is mirrored by increased dendritic length and spine density, and greater dendritic arborisation in pyramidal neurons of the PFC and hippocampus (80). These in vivo results are inferred also from very recent in vitro findings showing that the exogenous application of D-Asp to rat hippocampal slices produces a rapid genesis of middle size spines of CA1 neurons via an actin-sensitive mechanism, as early as 2 h after treatment (77). Consistent with structural and functional enhancement in cortico-hippocampal synaptic plasticity, Ddo−/<sup>−</sup> and <sup>D</sup>-Asp treated mice display significant improvements in the cognitive domain of spatial memory when tested in the hiddenplatform version of the Morris water maze and contextual fear conditioning (47, 57), two behavioral tasks involving the hippocampal activation of NMDARs (83, 84). Similar behavioral observations were reported also in D-Asp-treated rats (85).

However, if on the one hand NMDARs promote synaptic strength and connectivity, on the other they can produce neuronal death if their stimulation is abnormally intense or temporarily too long (86). In line with detrimental NMDARrelated effects, persistent increase of D-Asp levels results in precocious decay of basal glutamatergic transmission, synaptic plasticity and hippocampal reference memory in 13/14-monthold Ddo−/<sup>−</sup> (47), mirrored by loss of excitatory glutamatergic synapses and reduction of synaptic GluN1 and GluN2B subunits (87). In addition, a recent study also revealed that the lack of DDO leads to severe neuroinflammation processes and cell death in an age-dependent manner (46). In line with the results obtained in elderly Ddo−/<sup>−</sup> animals, also aged C57BL/6J mice treated with D-Asp (at the dose of 20 mM) for 12 months display similar deficits in hippocampal NMDAR-mediated synaptic plasticity despite such prolonged administration results in only two-fold D-Asp increase in this brain region (58). However, it should be noted that the interruption of D-Asp treatment for 3 weeks is sufficient to fully rescue the NMDAR-dependent LTP deficits and that such a long administration schedule does not produce locomotor and anxiety-like alterations in aged animals that, on the contrary, display a significant cognitive improvement in cue-dependent fear conditioning paradigm (58). Overall, these data point out that the "Yin and Yang" behavior of NMDARs (86) can be recapitulated in different stages of life by deficient DDO activity in the mouse brain.

### D-ASPARTATE AFFECTS SCHIZOPHRENIA-RELATED FEATURES AND ACTIVATION OF NEURONAL CIRCUITS IN THE RODENT BRAIN

Accumulating evidence supports the hypothesis that a developmental hypofunction of NMDARs is a causative factor in the etiology and pathophysiology of schizophrenia (88– 91). In the wake of the glutamatergic model of schizophrenia, several clinical, pharmacological, imaging, and genetic findings suggest today that dysfunctional metabolism of the NMDAR co-agonist, D-Ser, may produce NMDAR-mediated alterations leading to the manifestation of this mental illness (21–23, 92).

In the last years, a number of preclinical observations have shown that also D-Asp may contribute to influence some NMDAR-dependent phenotypes related to schizophrenia. For instance, behavioral studies performed in adult Ddo−/<sup>−</sup> and D-Asp-treated mice have revealed that chronic D-Asp elevation significantly reduces the prepulse inhibition (PPI) deficit induced by psychotomimetic drugs like amphetamine and MK-801 without affecting basal properties of sensorimotor filtering (57). In addition, Ddo−/<sup>−</sup> mice display a substantial reduction of motor hyperactivity, ataxia and PPI disruption triggered by acute administration of phencyclidine (PCP) (93, 94), the drug that better than any other models schizophrenia symptoms in both humans and rodents (22, 95–97). In line with behavioral observations, increased levels of <sup>D</sup>-Asp in Ddo−/<sup>−</sup> mice are also able to counteract the dysfunctional cortico-limbothalamic activation induced by PCP, as measured by fMRI (94). Moreover, consistent with increased dendritic length and spine density, resting-state fMRI has shown greater corticohippocampal connectivity in the brain of Ddo−/<sup>−</sup> mice (94). Increased functional connectivity between the hippocampus and cortex has been recently found also in the rat brain following intragastric administration of D-Asp (77). Since several clinical and preclinical studies suggest the occurrence of corticohippocampal dysconnectivity in schizophrenia (98, 99), the in vivo imaging data obtained in animal models with higher D-Asp levels highlight a potential translational relevance for D-Asp in treating this mental illness. Consistently, increased levels of D-Asp are also able to prevent corticostriatal long-term depression (57), a synaptic feature reported also in mice chronically treated with the typical antipsychotic haloperidol (100).

### REDUCED D-ASPARTATE CONCENTRATION IN THE HUMAN PREFRONTAL CORTEX OF SCHIZOPHRENIA POST-MORTEM BRAIN

Based on the results obtained in preclinical research, recent studies have focused the attention on humans in order to assess D-Asp metabolism and its impact on phenotypes relevant to schizophrenia. So far, two different studies have measured D-Asp levels in two different cohorts of post-mortem brain samples deriving from patients with schizophrenia and corresponding non-psychiatric controls (101, 102). The first study, performed on a small number of samples (7–10 subjects/diagnosis), revealed reduced D-Asp levels (about 40%) in the post-mortem PFC of schizophrenia-affected patients (101), linked to significantly increased DDO mRNA levels in the same brain area (94). However, neither methylation changes in the putative DDO promoter nor gross aberrations in this gene, including insertion, deletion, frameshift, or nonsense mutations were found in the same samples (94). Also the second study, performed on a larger number of samples (20 subjects/diagnosis/brain region), reported a significant D-Asp reduction (about 30%) selectively in the dorsolateral PFC (DLPFC) but not in the hippocampus of patients, compared to the respective brain regions of nonpsychiatric subjects (102). Interestingly, the biochemical analysis pointed out that reduced content of D-Asp in the DLPFC of schizophrenia-affected subjects is associated with an aberrant increase in DDO enzymatic activity (102). However, differently from the first cohort of samples (101), DDO gene expression, as well as DNA methylation status, was comparable between diagnoses (102, 103). Moreover, western blotting analysis of SR, regarded as an enzyme involved in D-Asp biosynthesis (32, 33), revealed no changes between schizophrenia-affected patients and healthy individuals (102).

Furthermore, to assess the effect of putative alteration of D-Asp metabolism in the human brain, another study reported the association among DDO gene variations, prefrontal DDO mRNA expression and structural/functional prefrontal phenotypes relevant to schizophrenia (80). Within this study, an in silico analysis performed on 268 post-mortem brains of non-psychiatric subjects (deriving from BrainCloud bank, http://braincloud.jhmi.edu) revealed that the C allele of the single nucleotide polymorphism (SNP), rs3757351, mapping in an intronic region of DDO gene, is associated with reduced prefrontal expression of DDO transcript, compared to the T allele and, thus, may hypothetically predict increased endogenous D-Asp levels in the PFC. This result led authors to perform in vivo imaging analyses whose results disclosed that healthy individuals bearing the C allele also display increased prefrontal gray matter volume and greater prefrontal activity, compared to the subjects with the T allele, when they were subjected to 1- and 2-Back working memory tasks.

Besides the neurochemical and functional studies reported above in post-mortem schizophrenia brain, a recent preclinical research by Usiello et al. revealed that the modulation of D-Asp metabolism could be instrumental for the mechanism of action by which a common second-generation antipsychotic, olanzapine, influences glutamatergic cortical neurotransmission (61). In particular, they found that olanzapine, differently from other typical and atypical antipsychotics, inhibits the enzymatic activity of DDO in vitro. Moreover, in agreement to in vitro results, chronic administration of this antipsychotic is able to increase the extracellular levels of D-Asp and L-Glu in the PFC of freely moving Ddo+/<sup>+</sup> mice, but not in Ddo−/<sup>−</sup> animals (61).

# CONCLUSIONS AND FUTURE PERSPECTIVES

The findings described in the present review highlight that increased levels of the endogenous NMDAR agonist, D-Asp, impact on a series of functional, and structural phenotypes relevant to schizophrenia. Despite the knowledge on the neurobiological role of D-Asp and its potential involvement in schizophrenia is significantly enhanced in the last years, many issues remain still unsolved or deserve further clarification. For instance, it seems clear today that D-Asp is synthesized in the brain where it is subjected to release and reuptake mechanisms that enable its neuromodulatory activity, but it is still unclear what are the specific enzymes and transport systems responsible for such processes. Is there a selective enzymatic activity responsible for D-Asp production in the mammalian brain? Moreover, the influence of SR upon the endogenous Daspartate level still waits to be clarified in the human brain. If SR is actually involved in D-Asp biosynthesis in the human brain, to what extent it contributes to the generation of this D-amino acid and in which spatiotemporal window? Are there relevant genetic SR variants associated to dysregulation of cerebral D-Asp levels? These questions may help to elucidate the effective significance of altered D-Asp metabolism in schizophrenia since SR is in the list of the largest genome-wide association study as a susceptibility gene for schizophrenia (92). The assessment of a D-Asp involvement in schizophrenia will necessarily undergo the verification of altered D-Asp amount in further schizophrenia brains. Indeed, the results of a dysregulated D-Asp metabolism in post-mortem brain samples so far achieved are encouraging but refer only to a limited number of samples, analyzed by the same research team (101, 102). Therefore, this issue deserves to be further assessed also by other laboratories and in a larger number of post-mortem brain tissues as well as in other biological materials such as the peripheral blood and cerebrospinal fluid of patients (FE, AB and AU personal communication).

Another important aim that should be pursued in future studies concerns the evaluation of the potential therapeutic value of D-Asp in schizophrenia treatment. In support of its translational application, D-Asp is nowadays approved for the use in humans, and commercialized as a dietary supplement. Accordingly, recent studies have shown that D-Asp treatment has no toxicological consequences or influences on hormonal activity of the hypothalamic-pituitary-gonadal axis and the mass of skeletal muscle (104–108). Besides D-Asp administration, future therapeutic strategies may be disclosed by the identification of compounds with inhibitory activity against DDO and, thus, able to indirectly enhance the availability of cerebral D-Asp. Novel DDO inhibitors have been so far functionally assayed in vitro by Homma et al. (109) and await further in vivo characterization to test their clinical validity.

Finally, future research needs to provide clear evidence of the potential role of embryonic D-Asp metabolism in the pathophysiological mechanisms leading to schizophrenia. In this regard, it is very important to underline that the etiology of this psychiatric disorder is likely to involve genetic and environmental risk factors emerging during the neurodevelopmental stages (110–113), in coincidence with the peak of D-Asp, and affecting several processes controlled by NMDARs, such as neurogenesis, survival, migration and formation of brain circuits (114–116). Therefore, we speculate that dysfunctional D-Asp metabolism

# REFERENCES


occurring during neurodevelopment may affect early critical processes dependent on NMDARs and, in turn, contribute to schizophrenia vulnerability. The recent generation of a genetic mouse model with prenatal reduction of D-Asp levels (FE and AU personal communication) may aid to disclose the importance of the transient occurrence of D-Asp in developmental brain processes and, in turn, the potential involvement of dysregulated D-Asp metabolism in a neurodevelopmental psychiatric disorder like schizophrenia.

### AUTHOR CONTRIBUTIONS

FE and AU conceived and wrote the manuscript. TN, MC and AB critically reviewed the manuscript.


pheochromocytoma MPT1 cells. Arch Biochem Biophys. (2002) 404:92–7. doi: 10.1016/S0003-9861(02)00241-2


underlying bidirectional plasticity. Nat Neurosci. (2004) 7:1104–12. doi: 10.1038/nn1311


**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 Errico, Nuzzo, Carella, Bertolino and Usiello. 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.

# Combination of G72 Genetic Variation and G72 Protein Level to Detect Schizophrenia: Machine Learning Approaches

Eugene Lin1,2,3†, Chieh-Hsin Lin3,4,5†, Yi-Lun Lai <sup>3</sup> , Chiung-Hsien Huang<sup>6</sup> , Yu-Jhen Huang<sup>7</sup> and Hsien-Yuan Lane3,7,8,9 \*

*<sup>1</sup> Department of Electrical & Computer Engineering, University of Washington, Seattle, WA, United States, <sup>2</sup> Department of Biostatistics, University of Washington, Seattle, WA, United States, <sup>3</sup> Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan, <sup>4</sup> Department of Psychiatry, Kaohsiung Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Kaohsiung, Taiwan, <sup>5</sup> School of Medicine, Chang Gung University, Taoyuan, Taiwan, <sup>6</sup> Department of Medicine Research, China Medical University Hospital, Taichung, Taiwan, <sup>7</sup> Department of Psychiatry, China Medical University Hospital, Taichung, Taiwan, <sup>8</sup> Brain Disease Research Center, China Medical University Hospital, Taichung, Taiwan, <sup>9</sup> Department of Psychology, College of Medical and Health Sciences, Asia University, Taichung, Taiwan*

### Edited by:

*Chad A. Bousman, University of Calgary, Canada*

### Reviewed by:

*Jingyu Liu, Mind Research Network (MRN), United States Tianmei Si, Peking University Sixth Hospital, China*

> \*Correspondence: *Hsien-Yuan Lane hylane@gmail.com*

*†These authors have contributed equally to this work*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *16 August 2018* Accepted: *18 October 2018* Published: *06 November 2018*

### Citation:

*Lin E, Lin C-H, Lai Y-L, Huang C-H, Huang Y-J and Lane H-Y (2018) Combination of G72 Genetic Variation and G72 Protein Level to Detect Schizophrenia: Machine Learning Approaches. Front. Psychiatry 9:566. doi: 10.3389/fpsyt.2018.00566* The *D-amino acid oxidase activator* (*DAOA*, also known as *G72*) gene is a strong schizophrenia susceptibility gene. Higher G72 protein levels have been implicated in patients with schizophrenia. The current study aimed to differentiate patients with schizophrenia from healthy individuals using *G72* single nucleotide polymorphisms (SNPs) and G72 protein levels by leveraging computational artificial intelligence and machine learning tools. A total of 149 subjects with 89 patients with schizophrenia and 60 healthy controls were recruited. Two *G72* genotypes (including rs1421292 and rs2391191) and G72 protein levels were measured with the peripheral blood. We utilized three machine learning algorithms (including logistic regression, naive Bayes, and C4.5 decision tree) to build the optimal predictive model for distinguishing schizophrenia patients from healthy controls. The naive Bayes model using two factors, including *G72* rs1421292 and G72 protein, appeared to be the best model for disease susceptibility (sensitivity = 0.7969, specificity = 0.9372, area under the receiver operating characteristic curve (AUC) = 0.9356). However, a model integrating *G72* rs1421292 only slightly increased the discriminative power than a model with G72 protein alone (sensitivity = 0.7941, specificity = 0.9503, AUC = 0.9324). Among the three models with G72 protein alone, the naive Bayes with G72 protein alone had the best specificity (0.9503), while logistic regression with G72 protein alone was the most sensitive (0.8765). The findings remained similar after adjusting for age and gender. This study suggests that G72 protein alone, without incorporating the two *G72* SNPs, may have been suitable enough to identify schizophrenia patients. We also recommend applying both naive Bayes and logistic regression models for the best specificity and sensitivity, respectively. Larger-scale studies are warranted to confirm the findings.

Keywords: artificial intelligence, D-amino acid oxidase activator, G72, machine learning algorithm, schizophrenia, single nucleotide polymorphism

# INTRODUCTION

Schizophrenia is a severe mental disorder characterized by symptoms such as delusions, hallucinations, blunted affect, impaired executive function, reduced motivation, and disorganized communication (1). The prevalence of schizophrenia is around 1% worldwide, and the social and economic costs of schizophrenia are enormous (2, 3). Converging evidence from genome-wide linkage studies, genetic case-control association studies, and genome-wide association studies indicate that several potential candidate genes are associated with schizophrenia (4). More and more genetic studies have employed novel computational tools such as naive Bayes to conduct gene discovery and detect new gene loci associated with schizophrenia (5). Identification of susceptibility genes for schizophrenia will help in early detection and prevention of high-risk individuals, as well as in developing novel therapies (6). The D-amino-acid oxidase activator (DAOA, also named G72) gene is one of the candidate genes.

The G72 gene, located on chromosome 13q3, exists in exclusively four primate species (7). Furthermore, the G72 gene encodes the protein that has been shown to function as a putative activator of D-amino acid oxidase (DAO), located in peroxisomes (7) and a mitochondrial protein (8). In vitro studies also demonstrate that the G72 protein binds to and activates DAO, which is capable of oxidizing Damino acids such as D-serine, an agonist of the N-methyl-D-aspartate receptor (NMDAR) (7, 9). The agonist activity at NMDAR may have particular relevance to a novel drug target for treatment of schizophrenia (10–16). One hypothesis of schizophrenia is that individuals who overproduce the G72 protein have lower D-amino acid levels and reduced NMDAR activity, predisposing them to schizophrenia (17, 18). A study suggests that the plasma G72 protein levels may be distinctively higher in patients with schizophrenia than healthy individuals (18). Of note, G72 protein levels are very similar between the medicated patients and the drugfree patients, implying that antipsychotic treatment does not influence G72 levels in plasma (18). In addition, G72 transgenic mice studies indicate a role of G72 in modulating behaviors relevant to schizophrenia (19–21). The G72 gene was also reported to predispose to schizophrenia in French Canadian (7), Russian (7), Chinese (22–24), German (25), and Ashkenazi (26) populations in single nucleotide polymorphism (SNP)-based studies.

A pilot study (18) modeled disease susceptibility to schizophrenia with plasma G72 protein levels using logistic regression. The current larger-sized study compared three artificial intelligence and machine learning techniques (including logistic regression, naive Bayes, and C4.5 decision tree) in predicting schizophrenia using G72 protein levels plus G72 SNPs. These three artificial intelligence and machine learning algorithms were chosen because they are well-known techniques with distinctively representational models; regression models for logistic regression (27), probabilistic models for naive Bayes (28), and decision tree models for the C4.5 algorithm (29).

# MATERIALS AND METHODS

### Study Population

This study was approved by the institutional review board of China Medical University Hospital, Taiwan, and carried out in accordance with the Declaration of Helsinki. Consecutive patients were screened and recruited from the psychiatric treatment programs of China Medical University Hospital, which is a major medical center in Taiwan. The patient population is similar to that of other mental health facilities. After complete description of the study to the subjects, written informed consents were obtained in line with the institutional review board guidelines. The study subjects were partially original to a previous study (18); the same 60 healthy individuals, but with more schizophrenia patients.

In the cohort, both patients and controls were Han Chinese aged 18–50 years, who were physically and neurologically healthy and had normal laboratory assessments (including urine/blood routine and biochemical tests). Both patients and controls were evaluated by the research psychiatrists using the Structured Clinical Interview for DSM-IV (SCID) for diagnosis. All patients had a DSM-IV diagnosis of schizophrenia. Patients with Axis I diagnosis other than schizophrenia, or any Axis II diagnosis were not included. All healthy volunteers were free of any Axis I or II psychiatric disorder. To exclude potential confounding effects, all participants were non-smokers and had no DSM-IV diagnosis of substance (including alcohol) abuse or dependence.

Drug history was ascertained by interviewing the patients and family members or caregivers, contacting other health care providers, and reviewing chart. Healthy controls had no history of exposure to psychotropic agents. Among schizophrenia patients, some patients were psychotropic-free for 3 months or longer and the other patients were stabilized on antipsychotics (risperidone, zotepine, haloperidol, quetiapine, amisulpride, sulpiride, flupentixol, olanzapine, ziprasidone, chlorpromazine, or paliperidone) for at least 3 months (18). The G72 protein level was not correlated with the medications administrated by patients (18).

### Laboratory Assessments: Genotyping

DNA was isolated from blood samples using MasterPure DNA purification kit following the manufacturer's instructions (EPICENTRE, Madison, Wisconsin, USA). To extract DNA, we used 200 µl of blood which was further solved in 100 µl of distilled water (30). The extracted DNA was diluted to the concentration of 50 ng/µl determined by the absorbance at 260 nM (ND-1000 UV-Vis spectrophotometer, Thermo Fisher Scientific Inc.). Four standard DNA samples with known genotypes were used for quality control (31).

All SNP genotyping was performed using the Taqman SNP genotyping assay (ABI: Applied Biosystems Inc., Foster City, CA, USA). The primers and probes of SNPs were provided by the ABI Company. The PCR reaction was conducted in 15 µl reaction volume which contained 0.4 µl DNA sample (50 ng), 7.5 µl Master mix (Roche), and 0.4 µl 40x primer pairs and probes. The samples were pre-incubated at 95◦C for 10 min to activate the Hot-Start DNA polymerase and to denature DNA, following by 40 amplification cycles of 92◦C denaturation for 15 s and 60◦C for 60 s. The probe fluorescence signal detection was performed using the ABI Prism 7500 Real-Time PCR System.

### Laboratory Assessments: Western Blotting

The plasma G72 protein expression levels were examined by western blotting (18). Ten milliliter of blood was collected into EDTA-containing blood collection tubes by personnel trained in phlebotomy using sterile technique. The blood specimens were processed immediately by centrifugation at 500 g. After centrifugation, plasma was quickly dissected and immediately stored at −80◦C until western blotting.

For western blotting, 100 µl plasma was depleted using ProteoPrep <sup>R</sup> Blue Albumin and IgG Depletion Kit (Sigma). The low-abundant protein fractions were collected to 100 µl. Then, 10 µl of the fractions were mixed with 4X sample buffer (500 mM Tris-HCl (pH 6.8), 16% SDS, 80% glycerol, 400 mM DTT, and 0.08% bromophenol blue) and separated on 12% SDS-PAGE. Proteins in the gels were transferred to 0.45µm polyvinylidene difluoride (PVDF) membrane (Millipore). The membranes were placed in 5% nonfat dry milk in TBST (20 mM Tris-HCl pH 7.6, 500 mM sodium chloride, 0.1% Tween 20) for 1 h at room temperature, then incubated with goat anti-G72 antibody (G72(N15):sc-46118, Santa Cruz Biotechnology) diluted by 1:1,000 in TBST overnight at 4◦C. The membranes were washed for 3 times in TBST and incubated for 2 h with a HRP-linked anti-goat IgG secondary antibody (sc-2030, Santa Cruz Biotechnology) diluted by 1:5,000 in TBST. After 3 washes in TBST, the blots were visualized with an ECL Advance Western Blotting Detection Kit (RPN2135, GE Healthcare). The stained membranes were photographed on ImageQuant LAS 4000 mini (GE Healthcare) and quantified using ImageQuantTM TL 7.0 software (GE Healthcare) by measuring the relative intensity from each band and normalized to the G72 recombinant protein (20 ng) signals. All western blot analyses were repeated for two times.

## Machine Learning Algorithms

Machine learning algorithm is a procedure for choosing the best hypothesis from a set of alternatives that fit a set of observations (27, 32). The advantages of machine learning algorithms, including nonlinearity, fault tolerance, and real-time operation, make them suitable for complex applications (33). The current study employed three families of machine learning algorithms, including logistic regression, naive Bayes, and C4.5 decision tree. Logistic regression analysis, the standard method for clinical classification (27), was used as a basis for comparison. The analyses were performed using the Waikato Environment for Knowledge Analysis (WEKA) software (27).

The naive Bayes classifier assumes that the presence or absence of a particular feature is unrelated to the presence or absence of any other feature (27). It calculates the probability that a given instance belongs to a certain class ("schizophrenia" or "control" in this study) by using Bayes' theorem.

The C4.5 decision tree is a model which builds decision trees top-down and prunes them using the concept of information entropy (27). The tree is first constructed by finding the root node (SNP or protein level) that is most discriminative for differentiating a disease status from "control." The best single feature test is decided by the information gain from choosing a feature (SNP or protein level) to split the data into subsets. Here, we used the default parameters of WEKA, such as 0.25 for the confidence factor and 2 for the minimum number of instances per leaf node (34).

# Evaluation of the Predictive Performance

The repeated 10-fold cross-validation method was used to investigate the generalization of the predictive models produced by the aforementioned algorithms (34–36). To measure the performance of the predictive models, we used the receiver operating characteristic (ROC) method and calculated the area under curve (AUC) to compare the performances of different predictive models (34, 35). AUC is a better performance metric than accuracy; the higher AUC means the better performance (36).

### Statistical Analysis

We analyzed the categorical data using the chi-square test. Differences for continuous variables were compared using the Student's t-test (37). Genotype frequencies were evaluated for Hardy-Weinberg equilibrium using a χ 2 goodness-of-fit test. The criterion for significance was set at P < 0.05 for all tests. Data are presented as mean ± standard deviation (SD).

# RESULTS

## Findings From the Unmatched Sample

The participants were 60 unrelated healthy individuals and 89 schizophrenia patients. As shown in **Table 1**, there was no significant difference in gender distribution between the two groups. The mean age (37.8 ± 10.5) of schizophrenia patients was older than that of healthy controls (32.8 ± 9.9, P = 0.004). The mean level of G72 protein in the plasma of schizophrenia patients was markedly higher than that of healthy controls (4.057 ± 2.594 ng/µL vs. 1.147 ± 0.574 ng/µL, respectively, P < 0.0001) (**Table 1**). The genotype frequencies for both rs1421292 and rs2391191 of the G72 gene were in Hardy–Weinberg equilibrium (P = 0.39 and 0.27, respectively).

### AUC After Adding rs1421292 Was Only Slightly Better Than That From G72 Protein Alone

**Table 2** summarizes the results from the naive Bayes algorithm. We generated five models (Models 1–5) with various combinations of three factors (rs2391191, rs1421292, and G72 protein levels). Among the five models, Model 2 with rs1421292 and G72 protein levels had the best AUC. Its AUC, sensitivity, and specificity were 0.9356, 0.7969, and 0.9372, respectively (**Table 2**). However, the AUC value after adding rs1421292 was only slightly better than that of G72 protein alone by 0.32% (Model 2 vs. Model 5).

We then employed the C4.5 decision tree algorithm with the same three factors (**Table 3**). Among the five models, Model 7 with rs1421292 and G72 protein levels had the best AUC. Its AUC, sensitivity, and specificity were 0.8525, 0.8202, and 0.8843,



*PANSS: Positive and Negative Syndrome Scale.*

*<sup>a</sup>Chi-square test for the categorical data; Student's t-test for continuous variables.*

TABLE 2 | Five naive Bayes models for differentiating schizophrenia patients from unmatched healthy individuals.


respectively (**Table 3**). The AUC value after adding rs1421292 was only slightly better than that of G72 protein alone by 0.19% (Model 7 vs. Model 10).

We finally tested the same factors with logistic regression (**Table 4**). The AUC, sensitivity, and specificity for the best logistic regression model (Model 12, applying rs1421292, and G72 protein levels) were 0.9272, 0.8576, and 0.8923, respectively. The AUC value only slightly increased by 0.97% after adding rs1421292 (Model 12 vs. Model 15).

### Of the G72 Protein Models, Naive Bayes Was Specific; And Logistic Regression, Sensitive

Among all the 15 models (Models 1–15) with unmatched schizophrenia patients and healthy controls, the naive Bayes (Model 2) with rs1421292 and G72 protein levels had the highest AUC. Of the three models with G72 protein alone (Models 5, 10, and 15), the naive Bayes (Model 5) had the best specificity (0.9503) and logistic regression (Model 15) had the best sensitivity (0.8765).

We further tested the relationship between G72 genotypes and G72 protein levels. The distribution of the two SNPs [for example, the numbers of TT (n = 57), TA (n = 65), and AA (n = 27) TABLE 3 | Five C4.5 decision tree models for differentiating schizophrenia patients from unmatched healthy individuals.


TABLE 4 | Five logistic regression models for differentiating schizophrenia patients from unmatched healthy individuals.


carriers in rs1421292] was illustrated in **Table 5**. As shown in **Table 5**, the G72 protein levels were marginally higher in the subjects with the TT or TA genotype than the AA homozygotes for rs1421292 (3.051 ± 2.588 vs. 2.137 ± 1.819; P = 0.084). There was no association between genetic variances of rs2391191 and G72 protein levels.

### Findings From the Matched Sample

Next, we selected 66 patients from the schizophrenia group to match better with healthy controls by age. The demographic characteristics of age and gender-matched schizophrenia patients and healthy controls are shown in **Table 6**. There was no significant difference in gender and age distributions between the two groups. The G72 levels in the plasma of schizophrenia patients were markedly higher than that of the matched healthy controls (4.188 ± 2.772 ng/µL and 1.147 ± 0.574 ng/µL, respectively, P < 0.0001) (**Table 6**).

### The Findings From the Matched Sample Were Similar to Those From the Unmatched Sample

**Table 7** shows the analytic results of schizophrenia patients and matched healthy controls. The findings from the matched sample were similar to those from the unmatched sample. Among the three models (Models 16–18) with G72 protein alone, the naive Bayes model (Model 16) performed best in specificity (0.966), and logistic regression (Model 18) had the best sensitivity (0.8483) (**Table 7**).



*<sup>a</sup>P value for comparing the subjects of the AA genotype with those of the TT genotype.*

*<sup>b</sup>P value for comparing the subjects of the AA genotype with those of the TA genotype.*

*<sup>c</sup>P value for comparing the subjects of the AA genotype with those of the TT or TA genotype.*

*<sup>d</sup>P value for comparing the subjects of the AA genotype with those of the GG genotype.*

*<sup>e</sup>P value for comparing the subjects of the AA genotype with those of the AG genotype.*

*<sup>f</sup>P value for comparing the subjects of the AA genotype with those of the GG or AG genotype.*

TABLE 6 | Demographic characteristics of schizophrenia patients and matched healthy individuals.


*PANSS: Positive and Negative Syndrome Scale.*

*<sup>a</sup>Chi-square test for the categorical data; Student's t-test for continuous variables.*

TABLE 7 | The models of naive Bayes, C4.5 decision tree, and Logistic regression for differentiating schizophrenia patients from matched healthy individuals using G72 protein.


### DISCUSSION

To our knowledge, this is the first study to examine the relationships between schizophrenia and G72 SNPs plus plasma G72 protein levels. We compared three machine learning algorithms, including logistic regression, naive Bayes, and C4.5 decision tree, in differentiating schizophrenia patients from healthy individuals. The results showed that the naive Bayes with G72 rs1421292 SNP and G72 protein levels (Model 2) performed best among all models (Models 1–15). The combination of G72 rs1421292 SNP and G72 protein levels was also the best model using the C4.5 decision tree (Model 7) and logistic regression (Model 12). These results were consistent with another finding of this study; that is, G72 rs1421292 SNP was marginally associated with G72 protein levels (**Table 5**). The proposed procedures can be implemented using the publicly available software WEKA (27) and thus can be widely used in genomic studies.

However, the AUC value after adding rs1421292 was only slightly better than that of G72 protein alone by an increase of 0.32% (Model 2 vs. Model 5) and 0.97% (Model 12 vs. Model 15), respectively. Hence, the present study suggests that G72 protein alone may have been feasible enough in AUC. Moreover, among the three models with G72 protein alone, logistic regression performed best in sensitivity, and the naive Bayes model was the most specific. This finding remained similar in the matched sample. We therefore recommend a combination model using logistic regression (for sensitivity) and naive Bayes (for specificity).

The common SNPs, such as rs2391191 and rs1421292, of the G72 gene have received considerable attention. These two SNPs were shown to be associated with schizophrenia (7, 23– 25); however, the findings are discordant. The rs2391191 SNP was reported to predispose to schizophrenia in Chinese (23, 24) and German (25) subjects, but not in French Canadian (7), United States (38), Scottish (22), Chinese (22, 39), and Taiwanese (40) populations. On the other hand, the rs1421292 SNP was found to be associated with schizophrenia in French Canadian (7), Russian (7), and German (25) subjects, but not in Japanese sample (41), UK population (42) and a mix of different races (including 84% Caucasian and 9% African American) in the United States (38). Moreover, two recent genome-wide association studies (GWAS) have been conducted to identify susceptible genetic loci affecting schizophrenia in mainly European (43) and Chinese (44) populations, respectively. However, no association of schizophrenia with SNPs in the G72 gene was found in these two large GWAS. Furthermore, by utilizing expression quantitative trait loci (eQTL) analyses, one of these GWAS implicated that there was no genetic risk regulating gene expression of G72 effect in brain or blood when eQTL analyses were used to explain associations with schizophrenia (43). The current study showed that the models which combined rs2391191 and rs1421292 (Models 4, 9, and 14) were not as good as other models (**Tables 2**–**4**). Moreover, adding rs2391191 or rs1421292 could not increase the AUC significantly than G72 protein alone. In agreement with several previous studies (7, 22, 38–44), the current study with a small sample size didn't demonstrate an association between the two G72 SNPs and schizophrenia.

In our previous study (18) on G72 protein levels, the severity of disease, the medications administrated by patients, as well as illness duration of the medicated patients did not influence the G72 protein level. In addition, the G72 protein level was significantly associated with schizophrenia in multivariate logistic regression analyses (18). The G72 protein level was also higher in drug-free or medicated schizophrenia patients than in healthy controls (18). The current larger-sized study extended the previous study by combining G72 protein levels with G72 SNPs as well as by leveraging the state-of-the-art artificial intelligence and machine learning algorithms. The current results implicated that the relevance of G72 protein levels to schizophrenia is much more significant than that of SNPs.

This study has several limitations. First, we chose only two SNPs of G72 for the current study because they seem to be two most commonly used SNPs. Whether other SNPs of G72 (25) could contribute more in the models of predicting schizophrenia remains unknown. Second, the findings of the current study came from a single population. More studies are necessary to testify whether the findings could be replicated in non-Taiwanese subjects (45, 46). Third, the small sample size does not allow us to draw definite conclusions (46). In the future, large-scale prospective studies in other ethnicities are warranted to reconfirm the potential

### REFERENCES


of G72 protein level and G72 SNPs as the biomarkers for schizophrenia.

# CONCLUSIONS

In conclusion, this preliminary study tested and compared numerous models using machine learning algorithms for predicting schizophrenia. The findings suggest that the models with G72 protein alone, without adding G72 SNPs, may have good enough power to discriminate patients with schizophrenia from healthy individuals. We also propose a combination of logistic regression and naive Bayes models to build a both sensitive and specific model to predict schizophrenia. Independent replications with larger-scale studies in other racial populations are needed to confirm the role of the G72 SNPs and G72 protein found in the current study.

## AUTHOR CONTRIBUTIONS

EL, C-HL, and H-YL designed the study. EL analyzed the data. EL, C-HL, and H-YL drafted and revised the manuscript. Y-LL, Y-JH, and C-HH conducted the laboratory experiments. All authors provided the final approval of the version to be published.

# FUNDING

This work was supported by National Health Research Institutes, Taiwan (NHRI-EX107-10731NI), Ministry of Science and Technology in Taiwan (MOST 107-2314-B-039−039), Taiwan Ministry of Health and Welfare Clinical Trial, and Research Center of Excellence (MOHW107-TDU-B-212-123004), China Medical University and Hospital (DMR-103-085; CMU107- BC-4).

gene G72 in mitochondrial function. Mol Psychiatry (2008) 13:685–96. doi: 10.1038/sj.mp.4002052


**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 Lin, Lin, Lai, Huang, Huang and Lane. 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.

# Decreased Blood Levels of Oxytocin in Ketamine-Dependent Patients During Early Abstinence

Ming-Chyi Huang1,2,3, Lian-Yu Chen1,4, Hu-Ming Chang<sup>1</sup> , Xiao-Yu Liang<sup>5</sup> , Chih-Ken Chen6,7 \*, Wan-Ju Cheng8,9 \* and Ke Xu<sup>5</sup>

*<sup>1</sup> Department of Psychiatry, Taipei City Psychiatric Center, Taipei City Hospital, Taipei, Taiwan, <sup>2</sup> Department of Psychiatry, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan, <sup>3</sup> Psychiatric Research Center, Taipei Medical University Hospital, Taipei, Taiwan, <sup>4</sup> Institute of Epidemiology and Preventive Medicine, National Taiwan University, Taipei, Taiwan, <sup>5</sup> Department of Psychiatry, Yale University School of Medicine, New Heaven, CT, United States, <sup>6</sup> Department of Psychiatry, Chang Gung Memorial Hospital, Keelung, Taiwan, <sup>7</sup> Chang Gung University School of Medicine, Taoyuan, Taiwan, <sup>8</sup> Department of Psychiatry, China Medical University Hospital, Taichung, Taiwan, <sup>9</sup> Department of Public Health, China Medical University, Taichung, Taiwan*

### Edited by:

*Kenji Hashimoto, Chiba University, Japan*

### Reviewed by:

*Chun Yang, Huazhong University of Science and Technology, China Kai Zhang, Chiba University, Japan*

\*Correspondence:

*Chih-Ken Chen kenchen@cgmh.org.tw Wan-Ju Cheng s871065@gmail.com*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *10 October 2018* Accepted: *07 November 2018* Published: *26 November 2018*

### Citation:

*Huang M-C, Chen L-Y, Chang H-M, Liang X-Y, Chen C-K, Cheng W-J and Xu K (2018) Decreased Blood Levels of Oxytocin in Ketamine-Dependent Patients During Early Abstinence. Front. Psychiatry 9:633. doi: 10.3389/fpsyt.2018.00633* Background: Ketamine, an N-methyl-D-aspartate (NMDA) receptor antagonist, is a common drug of abuse worldwide. Existing evidence suggest a disruption of oxytocin system involves in the development of addiction. In this study, we aimed to investigate the role of oxytocin in ketamine addiction by measuring the blood oxytocin levels in ketamine-dependent (KD) patients.

Methods: Sixty-five KD patients and 65 controls were enrolled. Fasting plasma levels of oxytocin were determined at baseline and 1 and 2 weeks after ketamine withdrawal. Ketamine use variables, Beck Depression Inventory, Beck Anxiety Inventory (BAI), Visual Analog Scale for craving, and Childhood Trauma Questionnaire-short form were assessed in KD patients.

Results: KD patients had significantly lower levels of oxytocin at baseline compared to controls (5.89 ± 2.13 vs. 9.53 ± 4.17 ng/mL, *P* < 0.001). Oxytocin levels increased after one (6.74 ± 2.63, *P* < 0.002) and 2 weeks (6.89 ± 2.69, *P* = 0.01) of withdrawal in KD patient despite the levels were still lower than controls (*P* = 0.001 and 0.002, respectively). The clinical variables did not correlate with baseline oxytocin levels except BAI scores, which showed a negative correlation with the levels (*r* = −0.263; *P* = 0.039).

Conclusion: We found a distinctively reduced oxytocin level in KD patients and the level did not normalize after early abstinence. Lower oxytocin might be associated with anxious phenotype of ketamine dependence. These results suggest that oxytocin system dysregulated following chronic ketamine abuse and might provide insight in evaluating the potential therapeutic use of oxytocin for treating ketamine dependence.

Keywords: oxytocin, ketamine, dependence, withdrawal, NMDA antagonist

# INTRODUCTION

Ketamine hydrochloride, a non-competitive antagonist of Nmethyl-D-aspartate (NMDA) receptor, has long been used as a short-acting anesthetic agent in humans and veterinary medicine. However, recreational ketamine use has been increased significantly worldwide (1), particularly in East and South-East Asia (2), and constitutes a major challenge to public health (3) because of the association with multi-organ harms (4). Ketamine has become one of the most abused drugs in Taiwan (5). As treatment for ketamine dependence has not yet been available (6), knowledge of the underlying mechanisms that modulates the reward system may provide insight in managing individuals suffering from ketamine dependence.

Emerging evidence suggests that oxytocin is involved in the neuroadaptive processes associated with the development of addiction (7). Oxytocin is a polypeptide hormone synthesized in the magnocellular neurons located within the supraoptic (SON), paraventricular (PVN), and accessory nuclei of the hypothalamus and released by the posterior pituitary gland, acting predominantly upon the oxytocin receptor, which is widely distributed throughout brain circuits related to reward, learning, memory, stress, and addiction (8). Animal studies have demonstrated that oxytocin interferes the development of tolerance to morphine (9) and alcohol (10). It may also have the effects of reducing self-administrative behavior of heroin (9), alcohol (11), methamphetamine (12), and cocaine (13). In reinstatement models of drug relapse, oxytocin was found to suppress the reinstatement of extinguished drugseeking behavior (13–16). The administration of carbetocin, an oxytocin receptor agonist, prevents reinstatement of conditioned reinforcing effects of morphine (16) and alcohol (14). Further evidence supporting the role of oxytocin in the involvement of addiction is that a negative association of plasma oxytocin levels with novelty-seeking, one of the indicators of addiction, was observed in heroin users (17).

Oxytocin has also been implicated in the regulation of withdrawal symptoms in both animals and humans (18). For example, oxytocin administration attenuates alcohol withdrawal convulsions (19) and blocks the somatic symptoms of nicotine withdrawal (20). In addition, oxytocin has also been reported to ameliorate the severity of acute or protracted opioid withdrawal (9, 16). In human studies, oxytocin was shown to be superior to placebo in reducing alcohol withdrawal symptoms in individuals undergoing alcohol detoxification (21). These observations collectively suggest that oxytocin dysregulation can occur after repeated exposure of drugs of abuse and play a role in the physical and behavioral effects associated with acute drug withdrawal.

Secretion of oxytocin is modulated by several mechanisms, including the regulation via NMDA receptors which are expressed in magnocellular cells of PVN and SON hypothalamic nuclei (22). Oxytocin neurosecretory cells in the nuclei receive a dense glutamatergic innervation, which plays a major role in the control of oxytocin release (23). It has been reported that intraperitoneal injection of a NMDA receptor antagonist (MK-801) decreases plasma concentrations of oxytocin in rats induced by osmotic stimulation (24). MK-801 also impaired the oxytocin responses to emotional stimuli (25). Supporting the link between NMDA and the oxytocinergic systems, one study in zebrafish demonstrated that oxytocin reversed MK-801 induced behavioral deficits in social interaction (26). Therefore, ketamine, as an NMDA antagonist, when being used chronically may alter the expression of oxytocin. To test this hypothesis, this study aimed to explore the difference of oxytocin plasma levels between treatment-seeking ketamine-dependent (KD) patients with the most recent ketamine use in the preceding 24 h and healthy controls. We also explored whether oxytocin level was normalized in the early stage of abstinence in KD patients. Finally, the correlations between oxytocin levels and ketamine use variables (frequency and quantity, craving scores, severity of dependence), depression and anxiety symptoms, and experiences of childhood trauma, which have been associated previously with oxytocin concentrations (27, 28), were also examined.

## MATERIALS AND METHODS

## Participants

This study was carried out in accordance with the recommendations of Institutional Review Board of Taipei City Psychiatric Center of Taipei City Hospital (IRB No: TCHIRB-1030408), 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 Institutional Review Board of Taipei City Psychiatric Center of Taipei City Hospital. The study was conducted in Taipei City Psychiatric Center of Taipei City Hospital after obtaining the approval of its Institutional Review Board. Treatment-seeking KD patients who voluntarily admitted to an inpatient ward from January 2014 to December 2015 were consecutively screened for the eligibility of enrollment in this study. The inclusion criteria were as follows: (1) age between 18 and 60 years; (2) fulfilling DSM-IV-TR criteria for ketamine dependence as verified by two board-certified psychiatrists; (3) last ketamine use within 24 h prior to admission, validated by self-report and positive in urine toxicology test for ketamine; (4) an ability to read Chinese and provide informed consent. The exclusion criteria were: (1) other substance use disorder (including abuse and dependence) in the past year except nicotine; (2) history of schizophrenia, bipolar disorder, or major depressive disorder, or having been treated with antipsychotics, mood stabilizers (including lithium, valproic acid, carbamazepine, and quetiapine), or antidepressants; (3) history of systemic medical illnesses such as hypertension, metabolic disorders (e.g., diabetes mellitus), or renal or liver diseases; (4) history of head injury, loss of consciousness, or neurological disorders; (5) inability or refusal to provide urine sample. Healthy controls were enrolled from the physical check-up unit in the hospital with inclusion criteria as (1) age between 18 and 60 years; (2) no other substance use disorder (including abuse and dependence) in the past year, except nicotine; (3) no history of a major psychiatric disorder (including schizophrenia, schizoaffective disorder, bipolar disorder, major depressive disorder, or organic mental disorders) screened using the Mini-International Neuropsychiatric Interview (MINI) (29) by a trained psychologist; (4) no known systemic or neurological diseases such as hypertension, metabolic disorders (e.g., diabetes mellitus), or renal or liver diseases; (5) an ability to read Chinese and provide informed consent. All the participants were given a comprehensive description of the study and then recruited after giving written informed consent.

The ward provides inpatient services in a controlled environment for individuals with substance use disorders and conducts various treatment programs, such as individual psychotherapy, group psychotherapy, family counseling, and vocational training, to help patients manage withdrawal symptoms, achieve initial abstinence, and prevent relapse. Every patient received regular drug testing to ensure their status of abstinence. Oral diazepam 20 mg/day were given on the first day of admission, and then gradually tapered off before day 5 for the treatment of withdrawal symptoms, such as anxiety, shaking, sweating, palpitations, and sleep impairment, as has been suggested previously (30–32).

### Clinical Assessment

The patients underwent the following assessments in the next morning of admission after the blood sample collection was performed.


### Laboratory Assays

Blood sample was collected once in control participants and 3 times in KD patients on the day following admission and after 1 and 2 weeks of ketamine abstinence. The instructions about the process were given on admission and repeated on the evening prior to the day of blood withdrawal. Venous blood samples were collected and placed in ice-cold vacutainer tubes containing EDTA (1 mg/ml of blood) anti-coagulant at 8:00–9:00 a.m. after an overnight fasting. The plasma was separated by centrifugation at 3,000 × g for 15 min at 4◦C, and aliquots of plasma were immediately frozen and stored at −80◦C until assayed. The range of storage time for plasma samples was from 1 to 1.5 years.

Plasma oxytocin levels were determined using enzyme immunosorbent assay (Catalog number: EKE-051-01, Phoenix Pharmaceuticals, Inc., Burlingame, California, USA). Each plasma sample was assayed in duplicates and the mean of the two measurements was used for analysis. The intra-essay and interassay coefficients of variation was 9 and 12%, respectively. The detection range of oxytocin assay were 0–100 ng/ml (provided by the manufacturer). There was no significant cross-reactivity or interference between oxytocin and the analogs observed.

## Statistical Analyses

In descriptive statistics, we used Chi-square test for categorical variables and independent t-test for numerical to compare demographic variables between KD and control groups.

To examine whether oxytocin differ between healthy controls and KD groups by time, we first performed a two-way ANOVA to compare mean differences among groups, adjusting for age and gender. Here, we assumed oxytocin level with no changes in week 1 and week 2 from baseline for controls.

$$\gamma\_{ijk} = \mu + \mathfrak{r}\_i + \beta\_j + \epsilon\_{ijk} \begin{cases} i = 1, 2 \\ j = 1, 2, 3 \\ k = 1, \dots, 195 \end{cases}$$

µ: overall mean effect

$$\tau\_{i} \text{: the effect of case and control } \begin{cases} case\\ control \end{cases}$$

$$\beta\_{j} \text{: the effect of the } j^{\text{th}} \text{ level of Time } \begin{cases} baseline\\ week1 \\ week2 \end{cases}$$

ǫi j k: is a random error component. Pairwise t-test was used to test the alterations of oxytocin levels after 1 or 2 weeks of ketamine discontinuation in the KD group. Bivariate correlation analysis was used to estimate the correlations between the clinical variables, such as ketamine use variables, BDI and BAI scores, and total scores or number of types of childhood trauma assessed by CTQ-SF, and laboratory measures. SAS 9.4 (SAS Institute, Cary, NC, USA) was used for all the analyses. Significance level was set at P < 0.05.

# RESULTS

### Sample Characteristics

A total of 130 participants, including 65 KD patients and 65 age- and gender-matched controls were enrolled. The


TABLE 1 | Sociodemographic and clinical characteristics in treatment-seeking ketamine-dependent (KD) patients and controls.

*BDI, Beck Depression Inventory; BAI, Beck Anxiety Inventory; CTQ-SF, Childhood Trauma Questionnaire-short form; SDS, Severity of Dependence Scale; VAS, Visual Analog Scale for craving.*

demographic and clinical characteristics are shown in **Table 1.** The proportion of tobacco smokers was significantly higher in KD group than control group (P < 0.001). All of KD patients had administered ketamine for an extended period of time (7.9 ± 4.3 years), with relatively high doses (average and maximum daily dose: 3.5 ± 2.7 and 7.2 ± 6.1 g, respectively) and high frequency of use (24.3 ± 10.9 days) in the past 1 month. All of them reported the main route of ingestion as snorting without injection. The demographic and clinical assessment data were comparable between those receiving blood withdrawal at week 1 (N = 38) and those who did not (N = 27).

### The Oxytocin Levels in KD Patients

Two-way ANOVA showed that oxytocin level was significantly lower in KD compared to control group. We found no significant difference of oxytocin level by time within KD groups (shown in **Figure 1**).

In pairwise t-tests, we found that KD patients had significantly lower levels of oxytocin at baseline compared to controls (5.89 ± 2.13 vs. 9.53 ± 4.17 ng/mL, P < 0.001). Thirty-eight out of 65 patients with KD provided blood samples after 1 week of ketamine abstinence whereas among them 32 patients provided blood after 2 weeks' abstinence. We found oxytocin levels increased after one (6.74 ± 2.63, P < 0.002) and 2 weeks (6.89 ± 2.69, P = 0.01) following ketamine discontinuation comparing with baseline oxytocin level; however, the oxytocin in week 1 and week 2 among KD patients were still

significantly lower than controls (P = 0.001 and 0.002, respectively).

### The Correlation of Oxytocin Levels With Ketamine Use Variables, Depression or Anxiety Symptoms, and Childhood Trauma Exposure

The baseline oxytocin levels were not significantly correlated with ketamine use variables, including total years of ketamine use, daily dose and using days in the past 1 month, or VAS of ketamine craving, depression severity, and total scores or number of types of childhood trauma in CTQ-SF (P-values: all >0.05) (shown in **Table 2**). There was a marginal significant correlation between oxytocin levels and SDS scores (r = −0.253; P = 0.058), suggesting a trend of lower oxytocin levels in patients with higher dependence severity. Noteworthily, anxiety symptoms were negatively correlated with oxytocin levels at baseline (r = −0.263; P = 0.039).

# DISCUSSION

To our knowledge, no studies to date have explored the possible association of oxytocin with chronic NMDA antagonism. Specifically, we found a significantly decreased blood level of oxytocin in patients with ketamine dependence compared to healthy controls. More interestingly, the reduction of oxytocin did not normalize after 2 weeks of withdrawal. Of note, patients with greater anxiety severity were associated with a lower oxytocin level in blood. Our preliminary findings highlight a possibility that chronic ketamine exposure is associated with a down-regulation of oxytocin and such abnormal oxytocin regulation was persistent in early abstinence.

Existing evidence indicates that oxytocin system interferes with different stages of addictive behavior across the addiction cycle (18) by reducing drug consumption, tolerance, and



*BDI, Beck Depression Inventory; BAI, Beck Anxiety Inventory; CTQ-SF, Childhood Trauma Questionnaire-short form; SDS, Severity of Dependence Scale; VAS, Visual Analog Scale for craving. Bold highlights statistically significant values (p* < *0.05).*

reinforcing effects, suppressing both acute and protracted withdrawal, and inhibiting relapse (9–16). The possible counteracting effect of oxytocin toward addictive behaviors has also been demonstrated in recent clinical studies that proposed oxytocin might serve as a novel treatment for drug abuse (38). Intranasal oxytocin administration has been shown to be beneficial in treating individuals with alcohol (39), cannabis (40), and cocaine use disorders (41), particularly through the effect of craving reduction. Therefore, it is likely that chronic drug abuse causes a neuroadaptaional down-regulation of oxytocin, and administration of oxytocin may compensate the abnormalities and further modulate the neurobehavioral response (7, 18). In light of this point, our results also support a potential role of oxytocin in the mechanisms underlying the development of ketamine addiction. Previous studies showed that NMDA receptors are involved in the release and secretion of oxytocin (22–24). Whilst NMDA receptor activation may enhance oxytocin release, NMDA antagonist treatment has been shown to suppress the release (42) and decrease the blood level of oxytocin (24). In agreement, our patients using ketamine, an NMDA antagonist, chronically similarly displayed a reduced blood level of oxytocin.

We found oxytocin levels in KD patients increased significantly after 1 week of ketamine discontinuation despite the levels were still deficient even after 2 weeks of withdrawal compared to controls. The alterations of oxytocinergic system have been suggested to contribute to withdrawal symptoms of various substances of abuse. For example, Zanos et al. found oxytocin involves the emotional impairment of morphine abstinence in mice and oxytocin administration attenuated the negative emotional consequences (16). Similarly, oxytocin administration was reported to decrease the alcohol withdrawalinduced convulsion (19), somatic symptoms of nicotine withdrawal (20), and morphine withdrawal-related hypothermia and body weight loss (9). In line with the potential advantage of oxytocin in blocking withdrawal, some clinical evidence also showed that in alcohol-dependent individuals undergoing alcohol detoxification, oxytocin treatment was associated with a superior effect over placebo on alcohol withdrawal symptoms blockade and less symptom-triggered benzodiazepine doses needed for withdrawal management (21). Although we did not measure the severity of withdrawal sequentially, it is possible that the clinical improvement of withdrawal symptoms in KD patients was associated with an elevation of oxytocin during withdrawal. Whether there would be normalization after sustained abstinence should be examined in future studies.

In our study, patients with greater anxiety appeared to have lower oxytocin levels. This finding is in line with the current literature revealing that oxytocin is a profound anxiolytic and anti-stress factor of the brain (43). Behavioral data from the oxytocin gene-deleted mice displayed more anxietyrelated behaviors and higher stress hormone (corticosterone) secretion after a stressor challenge compared to their wildtype counterparts (44). It is likely that the reduction in oxytocin may contribute to an increased level of anxiety in patients using ketamine chronically. In agreement with this notion, previous evidence demonstrated that oxytocin treatment reduces the anxiety response in rats during cocaine withdrawal (45). However, one recent study found in heroin users the plasma oxytocin levels were instead positively correlated with anxiety scores during acute withdrawal (46). The reasons for the contradictory findings are not clear. Oxytocin is thought to regulate the homeostasis of stress response mechanisms of hypothalamus-pituitary-adrenal (HPA) axis, which becomes dysfunctional following chronic drug abuse and in turn leads to an increased level of anxiety. The interactions with HPA axis are likely to have a critical role in oxytocin's effect to attenuate heightened anxiety during drug withdrawal (16, 18). Moreover, some evidence showed that different classes of drugs of abuse exhibited dissimilar effects on HPA axis activity at doses that reinforce addictive behaviors (47). For example, while ketamine may inhibit the HPA axis activity (47), chronic alcohol consumption produces a potent activation of the HPA axis in humans (48) and active heroin users showed a normalized HPA axis response (49). Given the interplay between oxytocin system and HPA axis, it is likely the expression pattern of oxytocin might differ between various drugs of abuse.

Findings from this study should be interpreted in light of several limitations. First, we only examined the oxytocin following 2 weeks of ketamine withdrawal and did not follow the depressive or anxiety symptoms after baseline. Given the persistent nature of addiction, future studies are required to examine the alterations oxytocin after a longer term of abstinence as well as how it contributes to the emotional disturbances that are commonly associated with addiction. Second, compared to patients with KD, only a small proportion of our controls were smokers and as result the differences of the biomarker levels between groups might be biased. Despite that preclinical studies suggest nicotine can modulate the oxytocin signaling (20, 50), no human studies to date have reported the association of oxytocin with tobacco smoking. In addition, we did not find a significant correlation between oxytocin levels and smoking amount in patients with KD (P = 0.22). Therefore, the effect of tobacco smoking on oxytocin levels might be limited. Third, using plasma oxytocin as a surrogate for central oxytocin function is still controversial because peripheral concentrations may also be contributed by peripheral organs including the gastrointestinal tract, heart, and reproductive tract (51). Despite an affirmative conclusion is still lacking, one recent meta-analysis showed a positive association may exist between central and peripheral oxytocin concentrations (52). Last, apart from the effect of chronic ketamine administration on the oxytocin expression, other individual factors may also contribute to the differences in oxytocin system, such as genetic variation, early life adversities, or stressful social experiences, and thus confound the results (53). We were unable to confirm whether the oxytocin expression in KD patients had been different from normal controls before they ingested ketamine, or a causal relationship between heavy ketamine use and deficient oxytocin expression.

In summary, we found that oxytocin levels were distinctively reduced in KD patients compared to controls and the reduction was persistent throughout the first 2 weeks of baseline was observed. In addition, a lower level of oxytocin was associated with greater anxiety symptoms. These results suggest that oxytocin down-regulation may contribute to the neuroadaptational mechanisms of ketamine addiction and a lower oxytocin might be associated with the anxious phenotype of ketamine addiction. Our data may further provide insight into the neurochemical changes of dysregulated oxytocin signaling

### REFERENCES


after chronic NMDA antagonism. Whether there will be a delayed normalization of oxytocin expression after a longerterm of abstinence should be further testified. Given the growing interests in the utility of oxytocin-based therapeutic approaches to drug and alcohol addiction (7, 18), it might be worth examining the relevance of oxyctocin administration to ketamine dependence treatment in future studies.

### AUTHOR CONTRIBUTIONS

M-CH designed the study and wrote the manuscript. L-YC and H-MC recruited participants and assisted in data interpretation. X-YL and W-JC did the statistical analysis and drafted the results section. C-KC assisted with the design of the analysis and provided statistical consultancy. KX revised the final version. W-JC incorporated edits from co-authors and was responsible for communication.

### FUNDING

This study was supported by grants from Ministry of Science and Technology, Taiwan (103-2628-B-532-001-MY3 and 106- 2314-B-532-005-MY3), Chang Gung Memorial Hospital, Taiwan (CMRPG2D0261, CMRPG8C1291, and CMRPG8D0481), Taipei City Hospital, Taiwan (10701-62-029 and TPCH-106-055), and National Center for Post-Traumatic Stress Disorder, United States.


Alcohol Alcohol (1987) 22:71–4. doi: 10.1093/oxfordjournals.alcalc. a044668


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

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

# Blood Levels of Glutamate and Glutamine in Recent Onset and Chronic Schizophrenia

Caroline Madeira<sup>1</sup> , Flavio V. Alheira<sup>2</sup> , Marilia A. Calcia<sup>2</sup> , Thuany C. S. Silva<sup>2</sup> , Filippe M. Tannos <sup>2</sup> , Charles Vargas-Lopes <sup>1</sup> , Melissa Fisher 3,4, Nelson Goldenstein<sup>2</sup> , Marco Antonio Brasil <sup>2</sup> , Sophia Vinogradov 3,4, Sergio T. Ferreira5,6 \* and Rogerio Panizzutti 1,2 \*

1 Instituto de Ciências Biomédicas, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, <sup>2</sup> Serviço de Psiquiatria e Psicologia Médica, Hospital Universitário Clementino Fraga Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, <sup>3</sup> Department of Psychiatry, School of Medicine, University of California, San Francisco, San Francisco, CA, United States, <sup>4</sup> Department of Psychiatry, University of Minnesota, Minneapolis, MN, United States, <sup>5</sup> Instituto de Bioquímica Médica Leopoldo de Meis, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, <sup>6</sup> Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil

### Edited by:

Hsien-Yuan Lane, China Medical University, Taiwan

### Reviewed by:

Svetlana Ivanova, Mental Health Research Institute, Tomsk NRMC, Russia Stefania Schiavone, University of Foggia, Italy

### \*Correspondence:

Sergio T. Ferreira ferreira@bioqmed.ufrj.br Rogerio Panizzutti rogerio@icb.ufrj.br

### Specialty section:

This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry

Received: 30 September 2018 Accepted: 05 December 2018 Published: 19 December 2018

### Citation:

Madeira C, Alheira FV, Calcia MA, Silva TCS, Tannos FM, Vargas-Lopes C, Fisher M, Goldenstein N, Brasil MA, Vinogradov S, Ferreira ST and Panizzutti R (2018) Blood Levels of Glutamate and Glutamine in Recent Onset and Chronic Schizophrenia. Front. Psychiatry 9:713. doi: 10.3389/fpsyt.2018.00713 Converging evidence indicates that dysfunctions in glutamatergic neurotransmission and in the glutamate-glutamine cycle play a role in the pathophysiology of schizophrenia. Here, we investigated glutamate and glutamine levels in the blood of patients with recent onset schizophrenia or chronic schizophrenia compared to healthy controls. Compared with healthy controls, patients with recent onset schizophrenia showed increased glutamine/glutamate ratio, while patients with chronic schizophrenia showed decreased glutamine/glutamate ratio. Results indicate that circulating glutamate and glutamine levels exhibit a dual behavior in schizophrenia, with an increase of glutamine/glutamate ratio at the onset of schizophrenia followed by a decrease with progression of the disorder. Further studies are warranted to elucidate the mechanisms and consequences of changes in circulating glutamate and glutamine in schizophrenia.

Keywords: recent onset schizophrenia, chronic schizophrenia, glutamate, glutamine, blood

# INTRODUCTION

Schizophrenia is a heterogeneous disorder characterized by a wide range of symptoms, including positive (hallucinations, delusions, disorganized thinking) and negative symptoms (affective flattening, alogia, apathy), and cognitive impairments (in memory and executive functions) (1, 2). Considerable evidence indicates that dysfunctional glutamatergic neurotransmission is involved in the pathophysiology of schizophrenia (3, 4). Among different glutamate receptor subtypes, hypofunction of N-methyl-D-aspartate (NMDA) receptors has been particularly implicated in schizophrenia. This comes from observations that administration of the NMDA receptor antagonists, phencyclidine (PCP) and ketamine, induces schizophrenia-like symptomatology in healthy individuals and exacerbates symptoms in patients with schizophrenia (5–7). Additionally, dysfunctions in metabotropic glutamate receptors (mGluRs), which can modulate NMDA receptor function, have also been implicated in schizophrenia (8, 9). Increased levels of mGluR1protein and mRNA have been reported in schizophrenia patients (10, 11). mGluR2/3 agonists improved both positive and negative symptoms of schizophrenia patients (12) and reduced cognitive deficits induced by ketamine in healthy volunteers (13). Finally, a polymorphism in the mGluR7 gene was associated with schizophrenia (14).

Glutamatergic neurotransmission is initiated by the release of glutamate into the synaptic cleft, where it binds to and activates glutamate receptors at the postsynaptic terminal membrane. Glutamate is subsequently removed from the synaptic cleft by astrocytes and converted to glutamine by glutamine synthetase (15). Glutamine can then be transported into the presynaptic neuron and reconverted to glutamate by phosphate-activated glutaminase (16, 17). Abnormalities in the glutamate-glutamine cycle have been associated with schizophrenia and may contribute to the dysfunction in glutamatergic neurotransmission (18, 19). Evidence indicates that glutamine synthetase protein levels are decreased in schizophrenia (19, 20). On the other hand, other studies found increased expression and enzymatic activity of the phosphateactivated glutaminase in schizophrenia (18, 21).

Changes in glutamate levels in the central nervous system (CNS) may reflect changes in blood levels and vice-versa, since a positive correlation has been reported between glutamate levels in the blood and in the CNS (22, 23). Altered glutamate and glutamine levels in the blood have been reported in schizophrenia, but results are not consistent. For example, previous studies have found increased blood levels of glutamate in patients with chronic schizophrenia compared with healthy controls (24, 25), but no difference was observed in acute schizophrenia (26). Interestingly, one study reported decreased blood glutamate levels in first episode psychosis (27). Blood glutamine levels were reported to be decreased in patients with schizophrenia (28), although another study found no difference in blood glutamine levels of individuals with schizophrenia compared with healthy controls (26).

The above findings indicate that changes in peripheral glutamate and glutamine may occur in schizophrenia, but the direction of change appears to depend on the duration of the disorder. We now report measurements of glutamate and glutamine levels in the blood of patients with recent onset schizophrenia or chronic schizophrenia compared to healthy controls.

# MATERIALS AND METHODS

## Participants

Thirty-two patients with recent onset (within 5 years of illness onset) and 56 patients with chronic schizophrenia were recruited from community mental health centers and outpatient clinics in the Bay Area of San Francisco, USA, to participate in randomized controlled trials of computerized cognitive training (ClinicalTrials.gov NCT00312962 and NCT00694889) (29–31). Blood collection was performed prior to the beginning of cognitive training. Fifty-three healthy control subjects were recruited from the same local community in the USA. In Brazil, 67 patients with chronic schizophrenia were recruited at the Institute of Psychiatry of the Federal University of Rio de Janeiro (IPUB-UFRJ) and the Psychiatry Center of Rio de Janeiro (CPRJ). Seventy-five healthy control subjects were recruited from the same local community in Brazil. Ethical approval for this study was obtained from the Committee for Ethics in Human Research from each participant institution. All subjects or their legal guardian provided written informed consent for study participation.

Patients with schizophrenia were diagnosed using the Structured Clinical Interview (SCID-I) for DSM-IV and evaluated using the Brief Psychiatry Rating Scale (BPRS). We used the BPRS because it is briefer and less time consuming than the Positive and Negative Syndrome Scale (PANSS), and there is a strong linear association between the two scales (32). Exclusion criteria were comorbid psychiatric diagnosis, substance abuse, intellectual disability, renal disease, epilepsy, and history of trauma or tumor in the central nervous system. Healthy controls had good general physical health and no Axis I or Axis II psychiatric disorder (SCID-NP), and no current or previous use of psychotropic medication.

### Determination of Glutamate and Glutamine Levels

Available samples comprised serum from the two clinical trial cohorts in the USA and plasma from the cross-sectional study in Brazil. Importantly, we note that a recent meta-analysis showed that altered levels of glutamate are found in both plasma and serum of patients with schizophrenia (33). Blood samples from patients or healthy controls were drawn before lunch, around noon, and stored at −80◦C until analysis. To extract free amino acids, trichloroacetic acid (TCA; 5% final concentration) was added to each sample, followed by centrifugation for 5 min. at 14,500 rpm at room temperature, separation of the supernatant and removal of TCA by extraction with saturated ether. Derivatization was performed with o-phtalaldehyde and N-tert-butyloxycarbonyl-L-cysteine (34). Separation and detection of derivatized amino acids was performed by highperformance liquid chromatography (34, 35) with a linear gradient of acetonitrile using a C18 reverse phase column. The investigators (CM and CVL) were blind to the group identity of the samples.

# Statistical Analysis

Data are presented as means ± SD (standard deviation). The distributions of glutamate and glutamine concentrations were evaluated for missing values and normalcy. We adjusted outliers by winsorization (36) replacing them with the next value plus or minus 10% (depending on whether the outlier is positive or negative). For glutamate, two outliers were adjusted in the healthy control group and two outliers were adjusted in the recent onset schizophrenia group. For the glutamine/glutamate ratio, two outliers were adjusted in the healthy control group and two outliers were adjusted in the recent onset schizophrenia group. In the chronic schizophrenia cohort from the USA, for glutamate we adjusted one outlier in the healthy control group and two outliers in the chronic schizophrenia group. For glutamine, we adjusted three outliers in the chronic schizophrenia group. For the glutamine/glutamate ratio, we adjusted one outlier in the healthy control group and three outliers in the chronic schizophrenia group. In the chronic schizophrenia cohort from Brazil, for glutamate we adjusted three outliers in the healthy control group and four outliers in the chronic schizophrenia group. For glutamine, we adjusted one outlier in the healthy control


Comparisons between the two groups were performed using χ 2 test for sex and smoking status, and t-test for age. Values are shown as means ± standard deviation (range). BPRS, Brief Psychiatric Rating Scale; CPZ, chlorpromazine; N/A, not applicable. Typical antipsychotics included flufenazine, clorpromazine, amoxapine, haloperidol, perphenazine, thioridazine, and thiothixene. Atypical antipsychotics included clozapine, olanzapine, quetiapine, risperidone, ziprasidone, and aripiprazol.

TABLE 2 | Blood levels of glutamate and glutamine in recent onset schizophrenia (USA cohort) adjusted for age, sex, and smoking status as covariates.


C.I., confidence interval.

group and two outliers in the chronic schizophrenia group. For the glutamine/glutamate ratio, we adjusted three outliers in the healthy control group and three outliers in the chronic schizophrenia group. Statistical significances of differences between groups were determined by unpaired t-test with Welch's correction. Differences between groups for gender and smoking status were analyzed using chi-square. Differences in glutamate or glutamine levels between healthy controls and patients with schizophrenia (recent onset or chronic) groups were tested using ANCOVA with age, sex or smoking status as covariates as appropriate.

### RESULTS

### Blood Glutamate and Glutamine Levels in Patients With Recent Onset Schizophrenia

We first asked whether circulating levels of glutamate and glutamine were altered in a cohort of 32 patients with recent onset schizophrenia (5 or less years of disorder) compared with 38 healthy controls. The mean age of patients with recent onset schizophrenia was 2 years older than the mean age of the healthy controls and this difference was significant at group level (t = 2.5, p = 0.01) (**Table 1**). However, no significant correlations were observed between age and blood levels of glutamate or glutamine in either healthy controls or recent onset schizophrenia groups (**Figure S1**). Sex distribution was different between groups, with the schizophrenia group including more men (X <sup>2</sup> = 8.07, p = 0.005). No significant difference in glutamine levels was observed between male and female healthy controls (**Figure S2**). In contrast, glutamate levels were significantly higher in males than in female patients with recent onset schizophrenia (151.7 ± 15.0 vs. 83.4 ± 16.8 µmol/L, respectively; t = 2.28; p = 0.03; **Figure S2**).

The proportion between smokers and non-smokers was also different between the two groups (X <sup>2</sup> = 8.94, p = 0.003). The group of patients with recent onset schizophrenia included more smokers than the healthy control group (28.6 vs. 2.7% smokers, respectively). We therefore introduced age, sex and smoking status as covariates in the analysis and found that blood glutamate and glutamine levels were ∼30% lower in patients with recent onset schizophrenia compared to healthy controls (**Table 2**). Further, the glutamine/glutamate ratio was significantly higher in patients with recent onset schizophrenia than in healthy controls (**Table 2**).


TABLE 3 | Blood levels of glutamate and glutamine in chronic schizophrenia (USA cohort) using smoking status as covariate.

C.I., confidence interval.

In patients with recent onset schizophrenia, glutamate and glutamine levels were not significantly correlated to total BPRS score (r = 0.22, p = 0.22 for glutamate; r = −0.33, p = 0.06 for glutamine), to the BPRS score for negative symptoms (r = 0.19, p = 0.28 for glutamate; r = −0.21, p = 0.24 for glutamine), or to the BPRS score for positive symptoms (r = 0.16, p = 0.39 for glutamate; r = −0.27, p = 0.13 for glutamine). Moreover, no significant correlation was observed between use of antipsychotic medications (expressed as chlorpromazine equivalents) and glutamate or glutamine levels (r = −0.09, p = 0.66 for glutamate; r = 0.17, p = 0.43, for glutamine; **Figure S3**).

### Blood Glutamate and Glutamine Levels in Patients With Chronic Schizophrenia

Next, we asked whether patients with chronic schizophrenia showed altered glutamate and glutamine levels compared to healthy controls. Patients with chronic schizophrenia and healthy controls had similar age and sex distributions, but patients with schizophrenia were more likely to smoke (X <sup>2</sup> = 5.93, p = 0.015) (**Table 1**). The smoking status was thus included as a covariate in the analysis (**Table 3**). Patients with chronic schizophrenia showed a trend of increase in blood glutamate levels compared to healthy controls (F = 3.50; p = 0.066). On the other hand, glutamine levels were significantly lower in patients with chronic schizophrenia than in healthy controls (F = 19.82; p = 0.0001). As a result, the glutamine/glutamate ratio was lower in patients with chronic schizophrenia than in healthy controls (F = 32.05, p = 0.0001).

No significant correlation was observed between age and levels of glutamate or glutamine in healthy controls (**Figure S4**). In contrast, age was significantly correlated with glutamate (r = 0.30, p = 0.02) and glutamine levels (r = −0.27, p = 0.04) in patients with chronic schizophrenia (**Figure S4**).

We evaluated the correlation between blood amino acid levels and the current use of medications and found that glutamate and glutamine levels were significantly and negatively correlated to the daily use of antipsychotic medications (measured in chlorpromazine equivalents) (r = 0.28, p = 0.04 for glutamate; r = −0.30, p = 0.02, for glutamine; **Figure 1**). We further evaluated the correlation between blood glutamate and glutamine levels and the severity of symptoms in patients with chronic schizophrenia. Glutamate and glutamine levels were correlated to total BPRS score (r = −0.35, p = 0.01 for glutamate; r = 0.26, p = 0.05 for glutamine), but not to the BPRS score for positive symptoms (r = −0.25, p = 0.06 for glutamate; r = 0.05, p = 0.72 for glutamine) or to the BPRS score for negative symptoms (r = −0.07, p = 0.63 for glutamate; r = 0.1, p = 0.48 for glutamine).

### Glutamate and Glutamine Levels in the Blood of Patients With Chronic Schizophrenia From Brazil

To validate and extend the above described findings (obtained in cohorts from the USA), we recruited a cohort of patients with chronic schizophrenia from Brazil and compared blood glutamate and glutamine levels with age- and sex-matched healthy controls. Although patient and control groups were quite similar in terms of age and sex, the group of patients with chronic schizophrenia included significantly more smokers than the healthy controls group (X <sup>2</sup> = 13.24, p = 0.0001) (**Table 4**). Thus, we included smoking status as a covariate in the analysis (**Table 5**). Results showed that patients with chronic schizophrenia had significantly increased blood glutamate levels compared to healthy controls (F = 16.3, p = 0.0001). Moreover, blood glutamine levels were significantly lower in patients with chronic schizophrenia than in healthy controls (F = 55.6, p = 0.0001). As a result, the glutamine/glutamate ratio in the blood was significantly lower in patients with chronic schizophrenia than in healthy controls (F = 27.0, p = 0.0001).

Glutamine levels were negatively correlated to age in healthy controls (r = −0.31, p = 0.01), but not in chronic schizophrenia patients (r = 0.01, p = 0.94). Glutamate levels were not correlated to age in both groups (healthy controls: r = 0.14, p = 0.24; chronic schizophrenia: r = 0.01, p = 0.92). Glutamate and glutamine levels were significantly correlated to the current use of antipsychotic medications (expressed as chlorpromazine equivalents) (r = −0.49, p = 0.0001 for glutamate; r = 0.35, p = 0.02, for glutamine; **Figure 2**). We compared levels of glutamate and glutamine in patients using typical antipsychotics vs. those using atypical antipsychotics. The glutamine/glutamate ratio was significantly lower in patients using typical antipsychotics than on those using atypical antipsychotics (0.57 vs. 1.25, F = 4.23, p = 0.046) (**Table S1**).

TABLE 4 | Characteristics of study subjects with chronic schizophrenia from the Brazilian cohort.


Comparisons between groups were performed using χ 2 test for sex and smoking status, and t-test for age. Values are shown as means ± standard deviation (range). p-value in bold indicates statistically significant difference. BPRS, Brief Psychiatric Rating Scale; CPZ, chlorpromazine; N/A, not applicable. Typical antipsychotics included flufenazine, clorpromazine, amoxapine, haloperidol, perphenazine, thioridazine, and thiothixene. Atypical antipsychotics included clozapine, olanzapine, quetiapine, risperidone, ziprasidone, and aripiprazol.

We further evaluated whether glutamate and glutamine levels were associated with the severity of symptoms in patients with schizophrenia, measured by their scores on BPRS. Glutamate and glutamine levels were not significantly correlated to total BPRS score (r = −0.07, p = 0.63 for glutamate; r = −0.04, p = 0.76 for glutamine), to the BPRS score for negative symptoms (r = −0.27, p = 0.06 for glutamate; r = 0.19, p = 0.18 for glutamine), or to the BPRS score for positive symptoms (r = 0.03, p = 0.84 for glutamate; r = −0.15, p = 0.31 for glutamine).

## DISCUSSION

The main findings of the present study are that, compared with healthy controls, patients with recent onset schizophrenia showed increased glutamine/glutamate ratio in the blood, while patients with chronic schizophrenia showed decreased glutamine/glutamate ratio. These findings are in agreement with one study that found decreased blood glutamate levels in first episode psychosis (27) and several studies including a meta-analysis that found increased levels of glutamate in the blood in patients with chronic schizophrenia compared with healthy controls (24, 25, 33). The evidence thus indicates that the glutamine/glutamate ratio is increased at the onset of schizophrenia but decreases with the progression of the disorder. Interestingly, the current use of antipsychotic medications (measured in chlorpromazine equivalents) was associated to the levels of glutamate and glutamine in chronic schizophrenia patients. However, the two cohorts showed opposite patterns of correlation: in Brazil more medication was associated to less glutamate and more glutamine, while in the USA patients taking more antipsychotic medications showed more glutamate and less glutamine. It is noteworthy that there were more patients in the Brazilian cohort taking typical antipsychotic (56%) than in the USA cohort (18%). Interestingly we found that patients using typical antipsychotic have lower glutamine/glutamate ratio than patients using atypical antipsychotics, indicating that different antipsychotic may have different effect on glutamate and glutamine levels. Accordingly, studies in patients and in animal models have been showing that some antipsychotic medication affects the release of glutamate in the brain, while others have no effect (37–39).


TABLE 5 | Blood levels of glutamate and glutamine in chronic schizophrenia (Brazil cohort) using smoking status as covariate.

C.I., confidence interval.

The increase in glutamate and the decrease in glutamine are in agreement with changes in the enzymes of the glutamateglutamine cycle previously found in chronic schizophrenia. Studies in post-mortem brain tissue from patients with chronic schizophrenia found decreased levels of the glutamine synthetase protein (19, 20) and increased glutaminase expression and enzymatic activity (18, 21). The consequent decrease in glutamine synthesis and increase in glutamate formation by glutaminase may also explain the significant decrease in glutamine/glutamate ratio observed in patients with chronic schizophrenia compared to healthy controls.

It is important to consider whether the observed changes in glutamate and glutamine in schizophrenia patients could reflect processes occurring peripherally. Although most studies to date have shown positive correlations between blood and cerebrospinal fluid (CSF) levels of glutamate and glutamine (22, 40–43), other studies have reported lack of such correlations (37, 44). Moreover, one cannot exclude the possibility that differences in dietary habits in patients with schizophrenia may account for differences observed here. Patients with schizophrenia often have less healthy diets, consuming less fruit, and vegetables or more calories than the general population (45). Moreover, secondgeneration antipsychotics may induce altered eating behaviors, including increased susceptibility to hunger and changes in appetite perception (46).

Importantly, studies of glutamate and glutamine levels in the CSF of patients with schizophrenia suggest that our findings in the periphery reflect what is happening in the CNS (41, 43). While patients with chronic schizophrenia showed a significant increase in CSF glutamate levels compared to healthy controls, no difference was observed in patients with recent onset schizophrenia (41). More recently, Hashimoto et al. observed higher glutamine/glutamate ratio in the CSF in patients with recent onset schizophrenia when compared to normal controls (43). Although this argues against a peripheral source for the glutamate and glutamine changes we observed, we have no direct evidence to rule out this possibility.

A limitation of the present study is the absence of a longitudinal follow up of the patients to examine possible changes in glutamate and glutamine levels over time. Moreover, although the differences found in our cross-sectional study were highly significant, the cohorts studied were of modest size and they may not be representative of the population with schizophrenia as a whole.

In conclusion, we found increased glutamine/glutamate ratio in patients with recent onset schizophrenia while in chronic schizophrenia the glutamine/glutamate ratio was decreased, as compared to healthy controls. Given the putative role of dysfunctional glutamatergic neurotransmission in the pathophysiology of schizophrenia, further studies are warranted to elucidate the underlying mechanisms and consequences of the alterations reported here.

## AUTHOR CONTRIBUTIONS

All authors certify that they have participated sufficiently in the work to take public responsibility for the content. RP and SF participated in the conception and design of study. CM, RP, and SF wrote the manuscript. CM, CV-L, and RP performed the analysis and interpretation of the data. FA, MC, TS, FT, MF, NG, MB, and SV conducted analysis of patients.

### ACKNOWLEDGMENTS

We thank A. C. Rangel and A. Fantinatti for administrative and technical support. This research was supported by

### REFERENCES


grants from DECIT/SCTIE/MS, SESDC, Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) (RP and SF); Instituto Nacional de Ciência e Tecnologia de Biologia Estrutural e Bioimagem (INBEB) (RP); and National Institute for Translational Neuroscience (SF). RP was recipient of a Long-Term Fellowship from the Human Frontier Science Program. CM and CV-L were supported by pre-doctoral fellowships from FAPERJ and CNPq.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyt. 2018.00713/full#supplementary-material


cortex of freely moving mice by inhibiting D-aspartate oxidase activity. Sci Rep. (2017) 7:46288. doi: 10.1038/srep46288


**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 Madeira, Alheira, Calcia, Silva, Tannos, Vargas-Lopes, Fisher, Goldenstein, Brasil, Vinogradov, Ferreira and Panizzutti. 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.

# Treatment Responses of Cognitive Function and Plasma Asymmetric Dimethylarginine to Atypical Antipsychotic in Patients With Schizophrenia

Zhi-min Yu1†, Ying Zhao2†, Jin-qiong Zhan3†, Tao Luo<sup>1</sup> , Jian-wen Xiong<sup>1</sup> , Bin Yu<sup>1</sup> , Bo Wei <sup>1</sup> \* and Yuan-jian Yang1,3 \*

### Edited by:

*Hsien-Yuan Lane, China Medical University, Taiwan*

### Reviewed by:

*Rogerio Panizzutti, Universidade Federal do Rio de Janeiro, Brazil Chih-Min Liu, National Taiwan University Hospital, Taiwan*

### \*Correspondence:

*Bo Wei jxmh\_wb@163.com Yuan-jian Yang yuanjimyang@yeah.net*

*†These authors have contributed equally to this work*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *02 November 2018* Accepted: *13 December 2018* Published: *10 January 2019*

### Citation:

*Yu Z, Zhao Y, Zhan J, Luo T, Xiong J, Yu B, Wei B and Yang Y (2019) Treatment Responses of Cognitive Function and Plasma Asymmetric Dimethylarginine to Atypical Antipsychotic in Patients With Schizophrenia. Front. Psychiatry 9:733. doi: 10.3389/fpsyt.2018.00733* *<sup>1</sup> Department of Psychiatry, Jiangxi Mental Hospital/Affiliated Mental Hospital of Nanchang University, Nanchang, China, <sup>2</sup> Department of Pharmacy, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, <sup>3</sup> Biological Psychiatry Laboratory, Jiangxi Mental Hospital/Affiliated Mental Hospital of Nanchang University, Nanchang, China*

Cognitive deficits represent a core feature of schizophrenia. Previous studies have demonstrated that plasma asymmetric dimethylarginine (ADMA) was increased in patients with schizophrenia and correlated with cognitive impairments. Atypical antipsychotics can produce cognitive benefits in schizophrenia patients. In this study, we conducted a prospective observation trial to explore whether plasma ADMA may serve as an indicator for evaluating cognitive improvements induced by atypical antipsychotics in patients with schizophrenia. A total of 41 schizophrenia patients with acute exacerbation were enrolled and 29 patients completed this study. These recruited patients were drug-naive or had no exposure to antipsychotics for at least 3 months. Thirty healthy individuals were recruited as a control group. Positive and Negative Syndrome Scale (PANSS) and a neuropsychological battery were used to evaluate schizophrenic symptoms and cognitive function, respectively. Plasma ADMA was measured by high-performance liquid chromatography (HPLC). We found that schizophrenia patients with acute exacerbation had significantly poorer cognitive performances and higher plasma ADMA levels than control individuals (*p* < 0.05). After 2 months of atypical antipsychotic treatment, patients showed significant improvements in processing speed, working memory, attention, and executive function (all *p* < 0.01). Plasma ADMA levels in patients after treatment were significantly decreased compared to baseline (2.42 ± 0.84 vs. 1.55 ± 0.34 µmol/L; *t* = 6.491, *p* < 0.001). Correlation analysis reveals that there is a significant correlation of the decrease in ADMA with improvements in working memory (*r* = −0.413, *p* = 0.026) and attention (*r* = −0.417, *p* = 0.025). Collectively, our results suggest that atypical antipsychotics improve cognitive function in schizophrenia patients with acute exacerbation, in parallel with decreased plasma ADMA levels. Plasma ADMA levels may be an indicator of cognitive recovery in schizophrenia.

Keywords: schizophrenia, cognitive function, asymmetric dimethylarginine (ADMA), atypical antipsychotic, plasma

# INTRODUCTION

Schizophrenia is a chronic mental disorder with a prevalence of ∼1% in the population worldwide. Along with positive and negative symptoms, deficits of cognitive function are also deemed as a core feature of schizophrenia (1, 2). A large number of cognitive areas including processing speed, attention, visual memory, verbal learning, working memory, and execute capacity have been reported to be impaired in patients with schizophrenia (3, 4). They are present before the onset of psychosis and may lead to functional disability, including defects in work and social contact, independent living, as well as skill acquirement (5, 6). Thus, the treatment of cognitive deficits has great benefits for the clinical outcome of patients with schizophrenia.

Antipsychotic drug treatment has been recommended to be a key component of schizophrenia treatment algorithms (7). First-generation antipsychotics (FGAs) were discovered in the 1950s and have therapeutic efficacy for positive symptoms, including delusions and hallucinations (8). However, longterm treatment with FGAs can cause a side-effect of cognitive deterioration (9). Since second-generation drugs (SGAs), known as atypical antipsychotics, have been developed and introduced clinically in the 1970s, they have been more frequently used for the management of schizophrenia in recent decades. In addition to positive symptoms, atypical antipsychotics also have significant impacts on negative symptoms and cognitive deficits associated with schizophrenia (9, 10). Numerous double-blind, random-controlled clinical studies have shown that, in compared to FGAs, treatment of atypical antipsychotics could greatly promote cognitive capability in patients with schizophrenia (9– 11). Overall cognitive ability and specific domains of cognition such as processing speed, memory, and attention were improved after the usage of atypical antipsychotics (10, 12).

Asymmetric dimethylarginine (ADMA) is an endogenous competitive inhibitor of nitric oxide synthase (NOS). It is physiologically generated during the methylation of protein arginine residues and released during proteolysis (13). Previous studies have demonstrated that plasma ADMA concentrations were increased in patients with schizophrenia relative to healthy individuals, but no association was found between plasma ADMA and the scores of psychiatric rating scales (14–16). We also found that plasma ADMA levels were elevated in schizophrenia patients and increased ADMA correlated with cognitive impairment (17). Zincir et al. reported that treatment with antipsychotics in patients for 2 months could reduce plasma ADMA levels, while no correlation was observed between the improvement of psychiatric symptoms and the change of plasma ADMA (16). However, whether there is a relationship between the reduction of plasma ADMA and cognitive improvement in patients treated with atypical antipsychotics remains unknown.

In this study, we hypothesized that peripheral ADMA levels may be a potential indicator of cognitive recovery in schizophrenia. To test this hypothesis, we investigated whether (1) eight-week atypical antipsychotic monotherapy improves cognition in schizophrenia patients with acute exacerbation; (2) peripheral ADMA is decreased at the endpoint (8th-week) compared to baseline; and (3) improvement of cognition is correlated with change of ADMA levels.

# MATERIALS AND METHODS

### Subjects

This is an open-label prospective observation study. All patients in this study were recruited from Jiangxi Mental Hospital. The patients of schizophrenia were diagnosed by two trained research psychiatrists using modified sections of the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV). The age of recruited patients was between 18 and 50 years old and the Positive and Negative Syndrome Scale (PANSS) score need to be between 60 and 120 at the baseline interview. All the patients who were recruited in this study had acute exacerbation for their psychotic symptoms and were admitted into acute ward for assessment and treatment. All these patients were antipsychotic naive or had not taken any antipsychotic drug over the past 3 months. In order to reassure the drug compliance, only patients who stayed well for at least 8-weeks were included in the present study. Healthy individuals were recruited from the local community as controls, matched with the patients by age, gender, education, and body mass index (BMI). All participants were Han Chinese ethnicity.

Patients with neurological disorders, mental retardation, and drug/alcohol abuse or addiction were excluded from this study. Some diseases including metabolic syndrome, diabetes, hyperlipidemia, coronary artery disease, inflammatory disease, liver damage, and renal failure may lead to changes in plasma ADMA (18–23). Thus, besides any additional axis I or axis II DSM-IV diagnosis, the exclusion criteria also included acute or chronic infections, current pregnancy, autoimmune, endocrine, and neoplastic diseases, allergic asthma, and other acute physical disorders, including heart or brain infarction within the past 3 months. Both patient and control groups were assessed with a medical history, physical examination, electrocardiogram, and laboratory tests. They were screened for acute infectious diseases by measuring body temperature, erythrocyte sedimentation rate, serum creatinine, C-reactive protein (CRP), and urinary culture. Subjects with abnormal parameters were excluded, for example urea (Normal range: 2.14∼7.85 mmol/L), creatinine (Normal range: 60∼130 µmol/L), SGOT (Normal range: 0∼40 U/L), SGPT (Normal range: 0∼40 U/L), CRP (Normal range: 0∼10 mg/L), cholesterol (Normal range: 3.38∼6.47 mmol/L), triglycerides (Normal range: 0.4∼1.81 mmol/L), systolic and diastolic blood pressure (Normal range: 140/90-90/60 mmHg), and BMI (Normal range: 18.5∼24).

All procedures for this study were reviewed and approved by the Institutional Review Board at Jiangxi Mental Hospital. The research was carried out in accordance with the Declaration of Helsinki and other relevant national and international regulations. A written informed consent was obtained from each subject, or his or her legal guardians.

### Evaluations of Clinical Symptom and Cognitive Function

PANSS was used to assess the severity of the psychotic symptoms of patients. The evaluation was conducted by two psychiatrists who had both attended a training session of PANSS use before the study began. The inter-observer correlation coefficient for the PANSS total score was >0.80.

The cognitive function of subjects was measured by a comprehensive battery of neurocognitive tests (17, 24, 25). These tests are commonly used in China, and their clinical reliability and validity have been evaluated in Chinese populations. The battery consists of seven tests and can be grouped into six cognitive domains: processing speed (Trail making test part A: TMT-A; Brief assessment of cognition in schizophrenia-symbol coding: BACS-symbol coding), working memory (Wechsler memory scale-3rd edition-spatial span: WMS-III spatial span), visual memory (Brief visual-spatial memory test-revised: BVMT-R), verbal learning (Hopkins verbal learning test-revised: HVLT-R), attention (Continuous performance test-identical pairs: CPT-IP), and executive function (Stroop color-word test).

TMT-A: In this test, you are given a pencil and a piece of paper with numbered circles, and you are asked to draw a line and connect the circles in the sequence of their numbers. You should draw the lines as quickly and as accurately as you can, since the time taken for the drawing is used for grading.

BACS-symbol coding: In this test, you are given 133 pairs of digits and symbols, and asked to copy the specific symbol as soon as you are shown its paired number. You should do this quickly as the symbols you finish correctly within 120 s are used for the grading.

WMS-III space span: In this test you are faced with a board with ten cubes on it spaced irregularly. An administrator will firstly show you combinations of the cubes in different ways and orders, forward and backward, and then ask you to recall the combinations. On each level of combination, you are given two trials. In the end you are graded in line with your recalled trials.

BVMT-R: In this test you will watch six geometric figures, which will appear three times, each time for 10 s. After that you are asked to draw the figures on a piece of paper in the layout as you watched. The more figures you draw correctly, the better you are rated.

HVLT-R: In this test, you will be presented with twelve Chinese words, which are listed in three categories. The list will be shown three times, followed by a delay time of 25–30 min. After that, you are asked to recall and speak out the words. Your performance is graded in accordance with the words you correctly recalled.

CPT-IP: In this test, you will see digital numbers of 2, 3, and 4 digits flashing on a computer screen. You are asked to immediately click the mouse after you see the same number flash on the screen in a row. In this test, you may hit 90 times in total, and possibly trigger false alarms 90 times in total. You will randomly respond about 270 times in total.

Stroop color-word test: In this test you're presented with three pages: firstly a word page, on which words of colors are black in color, secondly a color page, on which several rows of Xs are in different colors and thirdly a word-color page, on which the same words of colors on the first page are colored using the colors on the second page, though they are not colored in the color of themselves. In each trial you are presentedwith 100 words and you are asked to read them as fast as you can in a cycle time of 45 s. Your correct readings of the words are numbered for your grading.

## Plasma ADMA Level Analysis by High-Performance Liquid Chromatography (HPLC)

Peripheral blood was collected from an antecubital vein between 07:00 and 09:00 a.m. following an overnight fasting at baseline and at the endpoint of the study. The blood samples were centrifuged at 3,000 rpm for 5 min at 4◦C. The plasma was separated and stored at −80◦C until analysis.

The concentration of plasma ADMA was measured by HPLC as described in our previous study (17). Briefly, we added 5 mg of 5-sulphosalici-licacid (SSA) to 1 mL of plasma and allowed them to mix for 10 min at −20◦C. The sample was then centrifuged at 7,000 rpm for 5 min at temperature of 4◦C. After that, 10 microliters of the supernatant was mixed with 100 µL of derivatization reagent (prepared by dissolving 10 mg of ophthaldialdehyde in 0.5 mL of methanol, and adding 2 mL of 0.4 M borate buffer (pH 10.0) and 30 µL of 2-mercaptoethanol). The mixture was subsequently injected into the chromatographic system to determine the level of ADMA. The variability of this method was <7%, and the detection limit of the assay was 0.1µm.

# Statistical Analysis

Demographic and clinical variables were compared between control and patient groups using Student's t-test or analysis of variance (ANOVA) for quantitative variables and chi-squared test for qualitative variables. An analysis of covariance (ANCOVA) was performed to compare the scores of cognitive tests between two groups, controlling for gender, age, years of education, and BMI by using these variables as covariates. To compare the levels of ADMA between the patient and control groups, ANCOVA was performed using gender, age, years of education, and BMI as covariates to minimize the potential effect of these factors on the expression of ADMA levels. A paired t-test was performed to analyze plasma concentrations of ADMA between two time points (at baseline and 8th-week after atypical antipsychotic treatment) in the same patients. A Non-parametric test (Kruskal-Wallis test) was used to test whether there was a difference in cognitive function and ADMA level among different atypical antipsychotics at the endpoint. Spearman's correlation analysis was used to examine the correlation between change of cognitive test scores and ADMA level. All statistical tests were two-tailed, and the significance level was set at p < 0.05.

# RESULTS

## Demographic and Clinical Characteristics and Cognitive Performance

A total of 41 schizophrenia patients with acute exacerbation consented to the study and received baseline assessments. During the study session, these patients only received one type of atypical antipsychotics. The type of atypical antipsychotic was chosen according to the demographic and clinical symptom characteristics of patients. Patients using clozapine met the criteria of refractory schizophrenia. Benzene hydrochloride was only used when patients developed a marked extrapyramidal response. Of these, 12 patients dropped out for the following reasons: four patients were discharged from hospital before the endpoint of study; five patients did not respond to atypical antipsychotic monotherapy and required a combined treatment with other antipsychotic drugs or modified electric convulsive treatment (MECT); three patients suffered severe impairment of liver function and need to receive liver-protecting therapy. All data reported here are based on the remaining 29 patients who completed this study.

Baseline demographic and clinical characteristics for the patient and control groups are shown in **Table 1**. There was no significant difference in age, gender, years of education, or BMI between the two groups (all p > 0.05). Cognitive tests revealed that schizophrenia patients with acute exacerbation demonstrated significantly poorer cognitive performance than control individuals except for the BVMT-R controlling for age, gender, years of education, and BMI (all p < 0.05). We then measured the levels of plasma ADMA in both patient and control groups. The data of 29 patients and 30 healthy controls were included in the analysis of baseline ADMA levels. As previously reported (16, 17), patients with schizophrenia showed higher levels of plasma ADMA relative to the controls (2.42 ± 0.84 vs. 1.41 ± 0.14 µmol/L; F = 42.947, p < 0.001). ANCOVA analysis revealed that the difference between two groups still existed (F = 10.098, p < 0.001) when age, gender, years of education, and BMI were used as covariates.

### Clinical Efficacy and Neurocognitive Effect of Atypical Antipsychotic in Patients

The primary endpoint was the change from baseline to 8-weeks in PANSS and cognitive test scores. The number of patients treated with risperidone, olanzapine, clozapine, and aripiprazole was 10, 5, 7, and 7, respectively. The final median dose of risperidone, olanzapine, clozapine, and aripiprazole was 3.12 ± 1.05, 16.27 ± 5.33, 292.75 ± 115.63, and 14.66 ± 3.81 mg/day, respectively. **Table 2** shows the clinical efficacy and neurocognitive effect of atypical antipsychotics in schizophrenia patients with acute exacerbation. A significant improvement in PANSS total score and each subscore was observed when the patients completed this study (Paired t-tests; all p < 0.01). According to the proposed criteria for symptomatic remission in patients with schizophrenia (26), the rate of symptomatic remission at the endpoint was 82.75% (24/29). Additionally, patients given an 8-week period of atypical antipsychotic treatment showed significant improvements in the test scores of TAM-A, BACS-SC, WMS-III, CPT-IP, and Stroop color-word test (all p < 0.01). Furthermore, Kruskal-Wallis test showed that there was no difference in PANSS total score, each subscore, or the cognitive tests mentioned above, among risperidone, olanzapine, clozapine, and aripiprazole-treated groups at the endpoint (p > 0.05).

# Plasma ADMA Level After Atypical Antipsychotic Treatment

Together with clinical improvement, plasma ADMA levels after 8-week atypical antipsychotic monotherapy were also significantly decreased compared with baseline (2.42 ± 0.84 vs. TABLE 1 | Baseline demographic and clinical characteristics for patient and control groups.


*BMI, body mass index; PANSS, Positive and Negative Syndrome Scale; TMT-A, trail making task part A; BACS-SC, brief assessment of cognition in schizophrenia-symbol coding; CPT-IP, continuous performance test-identical pairs; WMS-III-SS, Wechsler memory scale-3rd edition-spatial span; HVLT-R, Hopkins verbal learning test-revised; BVMT-R, brief visual-spatial memory test-revised. NA, not applicable.*

1.55 ± 0.34 µmol/L; t = 6.491, p < 0.001) (**Figure 1**). There was no significant difference in plasma ADMA levels between the patients after atypical antipsychotic treatment and the controls (1.55 ± 0.34 vs. 1.41 ± 0.14 µmol/L; p > 0.05). Furthermore, all these atypical antipsychotics could lead to a significant decrease in the levels of plasma ADMA (p < 0.05), and there was no significant difference between groups in plasma ADMA levels at the study endpoint (Kruskal-Wallis test, χ <sup>2</sup> = 4.255, p = 0.235) (**Table 3**).

### Relationship Between Cognitive Improvement and Change of Plasma ADMA Levels

Spearman's correlation analysis was conducted to explore whether the change in plasma ADMA levels was correlated with clinical improvements. Similar to a previous study (16), we did not find any relationship between the reduction of plasma ADMA and the change of PANSS total score (r = 0.028, p = 0.886), positive subscore (r = −0.021, p = 0.913), negative subscore (r = −0.044, p = 0.821), or general psychopathology (r = 0.055, p = 0.776). Then we further analyzed the relationship between the change of plasma ADMA levels and cognitive improvement. As shown in **Figure 2**, there is a significant correlation between the decrease of ADMA and the change in WMS-III (r = −0.413,



p = 0.026) or CPT subscore (r = −0.417, p = 0.025), indicating that the decrease in plasma ADMA was correlated with the improvements of working memory and attention in patients treated with atypical antipsychotic.

### DISCUSSION

This is a prospective, open-label, 8-week observational study, aiming to explore the efficacy of atypical antipsychotics on cognitive deficits in patients with schizophrenia. We found that schizophrenia patients with acute exacerbation displayed multi-faceted cognitive impairments in comparison with healthy individuals. Biologically, plasma levels of ADMA were much higher in patients than in controls. 8-week treatment with atypical antipsychotic significantly improved psychiatric


symptoms and cognitive function, as well as leading to decreased plasma ADMA levels in patients. The decrease in ADMA levels is significantly associated with cognitive improvement of patients.

Cognitive deficits are a core feature of schizophrenia. Consistent with previous reports (17, 24, 25), our present study reveals that schizophrenia patients with acute exacerbation exhibited a wide range of cognitive impairments across multiple domains including processing speed, working memory, attention, verbal learning, and executive capability. Treatment with risperidone, olanzapine or aripiprazole could improve specific cognitive domains with similar global clinical efficacy in schizophrenia patients (10, 12). Results from animal experiments also showed that these atypical antipsychotics had positive benefits for cognitive deficits in a rat model of schizophrenia (27–29). In the present study, the recruited patients received medication of risperidone, olanzapine, clozapine or aripiprazole, and they showed significant improvements in cognitive domains of processing speed, working memory, attention, and executive function after 2 months of treatment. These results demonstrate the efficacy of atypical antipsychotics in cognitive impairments of schizophrenia patients with acute exacerbation.

Nitric oxide (NO) is a gas messenger that exerts multiple biological effects in the central nervous system. Deficits of NO signaling are shown to be linked with the pathogenesis of schizophrenia (30, 31). For example, Reif et al. reported that regulatory polymorphisms of NOS contributed to the genetic risk for schizophrenia (32). A decrease in nitrergic neurons was found in striatum of schizophrenia patients (33). The levels of NO and its metabolites were decreased in serum or plasma of patients with schizophrenia (14, 34). As an endogenous competitive inhibitor of NOS, ADMA was increased in the plasma of schizophrenia patients (16, 17). Treating schizophrenia patients with antipsychotic for 2 months led to a significant reduction in plasma ADMA levels (16). Consistent with this report, our present study showed that the levels of plasma ADMA in patients after 8-week atypical antipsychotic treatment were significantly decreased compared to that in controls. No difference was found in ADMA levels among patientstreated with risperidone, olanzapine, clozapine or aripiprazole at the study endpoint, demonstrating similar efficacy of these antipsychotics in the regulation of plasma ADMA levels in patients.

NO plays an essential role in synaptic plasticity and cognition (35, 36). Inhibition of endogenous NO generation impairs

synaptic plasticity and learning and memory (37–39). As an endogenous inhibitor of NOS, ADMA decreases the levels of NO by inhibition of NOS activity (13). Our previous study found that increased plasma ADMA was associated with cognitive deficits in patients with schizophrenia (17). The results of our present study showed that atypical antipsychotic treatment could significantly decrease plasma ADMA levels and parallelly improve cognitive function in schizophrenia patients with acute exacerbation. A series of studies have demonstrated the efficacy of NO on cognitive impairments in animal models of schizophrenia (31). Specifically, central supplement of Snitroso-N-acetylpenicillamine (SNAP), an NO donor, attenuated the impaired spatial memory in MK-801-treated mice (40). Intraperitoneal injection of GT 1061, a novel nitrate that enhance NO release, reversed MK-801-induced reversal learning deficits in rats (41). Treatment with sodium nitroprusside attenuated the attention impairment produced by amphetamine in the prepulse inhibition test in mice (42). In view of the role of NO in schizophrenia-related cognitive deficits and inhibition of endogenous ADMA on NO signaling, we postulated that atypical antipsychotic may exert a beneficial effect on cognitive function by upregulating NO signaling in patients with schizophrenia. However, further investigations are needed to address this postulation.

An interesting finding of this study is that improvement of cognitive function was associated with decreased ADMA levels, in patients treated with atypical antipsychotic. This finding indicates that the plasma ADMA level may be a potential indicator for evaluating the outcome for cognitive deficits after atypical antipsychotic treatment. Although underproduction of NO has been shown to contribute to the pathology of schizophrenia and decreased levels of plasma NO have been found in schizophrenia patients (14, 31), NO is not suitable to be a biomarker because it has a very short period of existence in plasma. ADMA levels are relatively stable and can be accurately measured in plasma (13). Thus, it may be more suitable to serve as a peripheral biomarker than NO. However, due to the small sample size of this study, further large-scale clinical studies are needed to confirm the value of plasma ADMA as an indicator for cognitive evaluation in schizophrenia.

Despite the suggestive results, some limitations of the present study should be noted. First, this is an open-label prospective observational study and selection bias may have been introduced. Generalization of these findings beyond the population should be cautious. Second, we measured the level of ADMA in plasma. Whether plasma ADMA can reflect a similar change in the brain is still uncertain. Third, each type of atypical antipsychotics may produce cognitive benefits in specific cognitive domains (12). However, the sample size of this study is small and limits further analysis of the correlation between improvement of specific cognitive domain and ADMA decrease stratified by type of atypical antipsychotics. Thus, large-scale clinical studies are needed to replicate and verify these findings. Fourth, Spearman's correlation analysis reveals a relationship between cognitive improvement and ADMA decrease in patients with schizophrenia. However, a causal relationship between the two variables cannot be drawn. Further research using animal experiments is needed to address this issue. Fifth, the symptomatic remission rate found in our present study (82.75%) exceeded reported rates (16 to 78%) (43). Perhaps the type of antipsychotics and the criteria for data collection could partially explain the difference. All the patients recruited in this study were treated with atypical antipsychotics, which might improve the treatment compliance of patients and thus resulted in a higher remission rate. In addition, as patients who did not respond to atypical antipsychotic monotherapy and required combined treatment with other antipsychotic or MECT were considered as dropout and were not included in the data analysis, this might also have contributed to the high statistical value for theremission rate. The sample characteristics of this study limit the generalization of the results. Collection of data from patients treated with different types of antipsychotics

(both typical and atypical antipsychotics), as well as patients with combined antipsychotic treatment, is needed to obtain a more exact value for the remission rate. Finally, we did not examine the pathway of ADMA/NOS/NO in the present study. NOS activity and nitrate and nitrite levels should be measured to verify that there is a corresponding change in NO levels.

In conclusion, our present study involved a prospective, open-label, 8-week observational trial to explore the effects of atypical antipsychotics on cognitive function and plasma ADMA levels in patients with schizophrenia. The results showed that atypical antipsychotic treatment in schizophrenia patients with acute exacerbation improved psychiatric symptoms and cognitive function, particularly working memory and attention, in parallel with decreased plasma ADMA levels. The change in ADMA was negatively correlated with the improvement of cognition, suggesting that plasma ADMA may be a potential indicator of cognitive recovery in schizophrenia.

## REFERENCES


# AUTHOR CONTRIBUTIONS

ZY, YZ, JZ, TL, JX, and BY were responsible for the clinical data collection and laboratory experiments. YY and BW were responsible for the study design, statistical analysis, and manuscript preparation. All authors have contributed to and approved the final manuscript.

## FUNDING

This work was funded by the National Natural Science Foundation of China (No. 81560232, 81600939, and 81760254) and the Natural Science Foundation of Jiangxi Province of China (No. 20161BAB205193 and 20171BAB205020).

### ACKNOWLEDGMENTS

We are deeply grateful to all participants.


**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 Yu, Zhao, Zhan, Luo, Xiong, Yu, Wei and Yang. 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.

# Glutamatergic Dysfunction and Glutamatergic Compounds for Major Psychiatric Disorders: Evidence From Clinical Neuroimaging Studies

Cheng-Ta Li 1,2,3,4 \*, Kai-Chun Yang1,3 and Wei-Chen Lin1,3

*<sup>1</sup> Department of Psychiatry, Taipei Veterans General Hospital, Taipei, Taiwan, <sup>2</sup> Institute of Brain Science and Brain Research Center, National Yang-Ming University, Taipei, Taiwan, <sup>3</sup> Division of Psychiatry, Faculty of Medicine, National Yang-Ming University, Taipei, Taiwan, <sup>4</sup> Institute of Cognitive Neuroscience, National Central University, Jhongli, Taiwan*

Excessive glutamate release has been linked to stress and many neurodegenerative diseases. Evidence indicates abnormalities of glutamatergic neurotransmission or glutamatergic dysfunction as playing an important role in the development of many major psychiatric disorders (e.g., schizophrenia, bipolar disorder, and major depressive disorder). Recently, ketamine, an *N*-methyl-d-aspartate antagonist, has been demonstrated to have promisingly rapid antidepressant efficacy for treatment-resistant depression. Many compounds that target the glutamate system have also become available that possess potential in the treatment of major psychiatric disorders. In this review, we update evidence from recent human studies that directly or indirectly measured glutamatergic neurotransmission and function in major psychiatric disorders using modalities such as magnetic resonance spectroscopy, positron emission tomography/single-photon emission computed tomography, and paired-pulse transcranial magnetic stimulation. The newer generation of antidepressants that target the glutamatergic system developed in human clinical studies is also reviewed.

Keywords: glutamate, major psychiatric disorders, NMDA antagonist, antidepressant, neuroimaging

### INTRODUCTION

### Overview of Glutamate and Risks of Neuropsychiatric Disorders

Glutamate is the most abundant excitatory neurotransmitter in the human brain and has critical roles in multiple brain functions and synaptic plasticity, such as long-term potentiation. However, excessive glutamate release can be toxic to the brain and has been linked to many neurodegenerative diseases, such as Alzheimer's disease, amyotrophic lateral sclerosis, and Huntington's disease (1). Glutamate excitotoxicity has been associated with exposure to severe stress, and excessive glutamate release and uptake have been identified in brain regions such as the frontal cortex and hippocampus of rats exposed to various forms of stress (2, 3). A growing body of evidence also indicates that abnormalities of glutamatergic neurotransmission play an important role in the development of many major psychiatric disorders (e.g., schizophrenia, bipolar disorder [BD], and major depressive disorder [MDD], including treatment-resistant depression [TRD]) (4, 5). In this review, we first update evidence from recent neuroimaging human studies that pinpoints glutamatergic dysfunction in the pathophysiology of major psychiatric disorders.

Edited by:

*Hsien-Yuan Lane, China Medical University, Taiwan*

### Reviewed by:

*Shigeyuki Chaki, Taisho Pharmaceutical, Japan Chun Yang, Huazhong University of Science and Technology, China*

\*Correspondence:

*Cheng-Ta Li ctil2@vghtpe.gov.tw*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *03 September 2018* Accepted: *21 December 2018* Published: *24 January 2019*

### Citation:

*Li C-T, Yang K-C and Lin W-C (2019) Glutamatergic Dysfunction and Glutamatergic Compounds for Major Psychiatric Disorders: Evidence From Clinical Neuroimaging Studies. Front. Psychiatry 9:767. doi: 10.3389/fpsyt.2018.00767*

**66**

### Glutamate Receptors and Subunits

Glutamate receptors can be divided into two categories: ionotropic glutamate receptors (iGluRs) and metabotropic glutamate receptors (mGluRs) (6). iGluRs with an ion channel pore activate when glutamate binds to receptors, whereas mGluRs activate ion channels on the plasma membrane indirectly through a signaling cascade. iGluRs have three subtypes of receptors—N-methyl-D-aspartate (NMDA) receptors, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, and kainate receptors—based on the chemical that binds to them more selectively than glutamate. Mammalian mGluRs are categorized into three groups: group 1 (mGluR1 and mGluR5), group 2 (mGluR2 and mGluR3), and group 3 (mGluR4, mGluR6, mGluR7, and mGluR8). Activation of the NMDA receptor (NMDAR) requires a glutamate binding to its NR2 subunits and a glycine binding to its NR1 subunits. Then, a nonspecific cation channel is opened, enabling Ca2<sup>+</sup> and Na<sup>+</sup> to enter and K<sup>+</sup> to exit the cell (7).

### Neuroimaging Techniques in Assessing Glutamate-Related Function in Human Brains

To date, it remains difficult to measure glutamatergic neurotransmission in human brains. However, some techniques have been developed to explore glutamatergic neurotransmission by calculating glutamate levels, such as magnetic resonance spectroscopy (MRS), or by measuring glutamate-related function, including positron emission tomography (PET)/singlephoton emission computed tomography (SPECT) and paired-pulse transcranial magnetic stimulation (ppTMS).

MRS is a specialized technique that directly quantifies brain molecules, including glutamate. This noninvasive and ionizing radiation-free neuroimaging technique is associated with magnetic resonance imaging (MRI), and both are used to acquire signal from hydrogen protons (or other nuclei, including carbon and nitrogen).

Typically, MRS is used to measure signals within a predefined region of interest (i.e., a voxel), although the technique is still being updated to acquire biochemical signals from the whole brain (8). Using <sup>1</sup>H-MRS, and depending on the chemical environment, each proton may be visualized at a specific chemical shift (peak position along the chemical shift axis). Glutamate (Glu) and glutamine can thus be marked by a series of resonance peaks between 2.2 and 2.4 ppm, which means that these metabolites overlap and are often referred to in combination as Glx. Some editing techniques (e.g., echo time averaging) enable differentiation of glutamate from glutamine (9). It has been reported that field strengths of 4 T or more are better for the separation of glutamate from glutamine (10). However, the information of glutamate derived by MRS does not exactly reflect glutamatergic neurotransmission of the neurons but rather the total level/concentration in the given voxel.

PET enables measurement of brain molecules by detecting gamma rays originating from the annihilation between electrons and positrons emitting from radionuclides. iGluRs are ligandgated ion channels that mediate excitatory neurotransmission in human brains and can be divided into at least three groups: NMDA, AMPA, and kainate receptors (11). Several radioligands have been developed to successfully image different subtypes of NMDARs in vivo (11). Moreover, human studies have been conducted using PET imaging with <sup>11</sup>C-ABP688, a radioligand for mGluR5, to evaluate ketamine-induced glutamate release both in healthy subjects (12) and in patients with MDD (13). However, the application of this PET imaging paradigm to measure glutamate receptors in clinical settings might be limited by the availability of the radioligand and the need for arterial input function for quantification. In addition, because PET/SPECT measures specific molecules, the findings do not represent exact glutamatergic neurotransmission. Alternatively, <sup>18</sup>F-fluorodeoxyglucose (18F-FDG)-PET, a clinical imaging tool widely used to measure brain glucose uptake with favorable signal-to-noise ratio in most brain regions, has been proposed to be a proxy measure of glutamatergic neurotransmission (14, 15). The rationale is that glutamate is produced in neurons from glucose-derived tricarboxylic acid cycle intermediates and branched-chain amino acids. The reuptake of glutamate from the synaptic cleft is coupled with Na+/K+-ATPase activation and glucose use (16). With neuronal depolarization and glutamate being released into the synaptic cleft from presynaptic vesicles, the process requires energy and is dependent on the use of glucose. Interestingly, ketamine-induced increased glucose uptake in patients with MDD (14) is in line with decreased <sup>11</sup>C-ABP688 binding in similar brain regions (13). Thus, <sup>18</sup>F-FDG-PET might serve as a promising tool to evaluate glutamatergic neurotransmission.

ppTMS is a noninvasive technique that manipulates the strength and stimulus intervals between two pulses to measure cortical inhibition and excitation in humans (17, 18). It can be used to examine at least two different corticocortical inhibitory processes in the human motor cortex that are mediated by different subtypes of GABAergic receptors: shortinterval cortical inhibition and long-interval cortical inhibition (19). Moreover, ppTMS can also be used to examine a corticocortical excitatory process, intracortical facilitation (ICF), when a subthreshold pulse precedes a test pulse by 8–30 ms (17, 20). The resulting facilitation of the motor-evoked potential response has been found to be mediated mainly by glutamatergic neurotransmission. When a glutamate antagonist, riluzole, is used, ICF can be suppressed without influencing cortical inhibition (21). Such findings indicate that the neurotransmitter glutamate is involved in facilitatory mechanisms of the motor cortex. Compared with the aforementioned techniques, ppTMS measurement such as ICF is more likely to reflect functional glutamatergic neurotransmission in the testing cortical region but not levels of the subtypes of glutamate receptors. In addition, I-wave facilitation is another ppTMS measurement that reflects glutamatergic activity of a different neuron population to ICF and could be mediated by non-NMDA receptors (22).

# Schizophrenia and Glutamatergic Dysfunction

Schizophrenia is a major psychiatric disorder characterized by prominent psychotic symptoms and abnormal social behaviors. Despite most current antipsychotics being dopamine antagonists or acting on dopamine receptors, alterations


TABLE 1 | Summary of major neuroimaging findings to support glutamatergic dysfunction in major psychiatric disorders.

*BG, Basal ganglia;ACC, anterior cingulate cortex; MTL, medial temporal lobe; THA, thalamus; PFC, prefrontal cortex; OFC, orbital frontal cortex; HIPPO, hippocampus; AMG, amygdala; PC, parietal cortex;OC, occipital cortex.*

\**Supported by at least one meta-analysis.*

in glutamatergic neurotransmission could be critical to the pathophysiology of schizophrenia. For example, administration of the NMDAR antagonist phencyclidine or ketamine could induce a schizophrenia-like state in human subjects (23, 24), supporting the hypothesis that glutamatergic dysfunction plays a crucial role in the pathophysiology of schizophrenia. Furthermore, group I mGluRs are heavily expressed in basal ganglia that contain high densities of dopamine receptors (25), and at least two independent studies have identified several deleterious single-nucleotide polymorphisms (SNPs) in the human gene encoding mGluR subtype I in patients with schizophrenia (26). Despite inconsistency, some postmortem studies have also revealed that iGluRs and mGluRs are abnormally expressed in human subjects with schizophrenia. For example, iGluR-AMPA receptors and kainate receptors were decreased in expression in the schizophrenic hippocampus, and the iGluR-NMDAR subunit NR1 might be abnormally expressed in some cortical regions in schizophrenia (27), whereas higher mRNA levels for group I mGluRs were found in the prefrontal cortex (Brodmann area 9) in patients with schizophrenia (28).

A large meta-analysis of <sup>1</sup>H-MRS studies identified 59 studies that included 1,686 patients and 1,451 healthy control subjects (**Table 1**) (29). By adopting a random-effects, inverseweighted variance model to calculate the pooled effect size, the investigators found that, in schizophrenia, there were significant elevations in glutamate in the basal ganglia (Hedges' g = 0.63; 95% confidence interval [CI], 0.15–1.11), glutamine in the thalamus (Hedges' g = 0.56; 95% CI, 0.02–1.09), and Glx in the basal ganglia (Hedges' g = 0.39; 95% CI, 0.09–0.70) and the medial temporal lobe (Hedges' g = 0.32; 95% CI, 0.12–0.52). No regions exhibited a reduction in glutamate metabolites in schizophrenia. A systemic review pinpoints that increased Glx in many cortical regions, including prefrontal cortex, temporal cortex, parietal cortex, and occipital cortex, as well as increased glutamine in thalamus, prefrontal cortex, and anterior cingulate cortex (30). By contrast, a recent systemic review and metaanalysis of <sup>1</sup>H-MRS studies on antipsychotic-naïve/free patients with schizophrenia included 21 studies and noted no changes in glutamate-related metabolites (31). Because brain glutamate levels may be confounded by drug use and related to response to antipsychotics, Egerton et al. investigated glutamate levels (Glu/Cr) in the anterior cingulate cortex and thalamus in antipsychotic-naïve or minimally medicated patients with firstepisode psychosis, and they found that higher levels of glutamate in the anterior cingulate cortex were associated with more severe psychotic symptoms at presentation and a lower likelihood of being in remission 4 weeks after amisulpride treatment (32).

Development of the glutamate hypothesis of schizophrenia was initially based on the effects of phencyclidine, which acts primarily as an NMDAR antagonist. Because presynaptic dopamine release is under the control of inhibitory GABAergic neurons that are activated by NMDARs, previous PET/SPECT studies that used D2/D3 receptor ligands (e.g., <sup>11</sup>C-raclopride and <sup>123</sup>I-iodobenzamide) in schizophrenia provided evidence for glutamate dysfunction in schizophrenia. A review study including D2/D3 receptor PET/SPECT studies revealed that increased dopamine uptake in the striatum and putamen was observed in schizophrenia, indicating an effect of NMDA blockade on striatal dopamine release (33). Not only findings in the basal ganglia, early studies measured cerebral blood flow following ketamine infusion in patients with schizophrenia by 15O-H2O PET and found increased blood flow in anterior cingulate cortex (34), which was found to correlate the ketamineinduced psychosis-effects in healthy control subjects (35). A study using 18F-FDG PET to study glucose metabolism after ketamine also revealed increased metabolism in frontal cortex and anterior cingulate cortex (36). In addition, repeated ketamine administration had been found to have increased dopamine D1 receptor binding in the dorsolateral prefrontal cortex by using 11C-NNC112 (37). As for studies using radioligands to image glutamatergic receptors directly, 123I-CNS-1261 as a SPECT ligand acts on NMDA receptors had been used in healthy subjects to study the binding of ketamine and the results found ketamine led to a global reduction in the binding signals (38). One study found the reduction of 123I-CNS-1261 binding after ketamine was greatest in thalamus, basal ganglia, and frontal cortex, which mainly correlated with negative symptoms (39). While marked reduction of NMDA receptor bindings had been reported in patients treated with schizophrenia, a study directly applying this compound in medication-free patients with schizophrenia found significant reductions of NMDA receptor binding in left hippocampus (40).

ICF of ppTMS reflects glutamatergic neurotransmission in the motor cortex. A meta-analysis including ppTMS studies from 1990 to 2012 found no changes of ICF, but decreased short-interval intracortical inhibition, in schizophrenia (41). By contrast, later research with small sample sizes revealed significant increases of ICF in first-episode schizophrenia compared with healthy control subjects (42). In addition, Iwave facilitation was found to be increased in patients with schizophrenia (43). Future studies with larger sample sizes controlling chronicity of illness course and the use of medications are still needed.

## Bipolar Disorder and Glutamatergic Dysfunction

BD is a major psychiatric disorder characterized by prominent mood fluctuation, including episodes of depression and periods of abnormally elevated mood. Postmortem studies of patients with BD revealed reduced expression of NMDAR subunit NR1 in the prefrontal cortex (44) and reduced expression of several NMDA, AMPA, and kainite receptor subunits in the medial temporal cortex (45), although other postmortem research found that mGluRs seem to be less involved in the anterior cingulum of patients with BD (46). However, genome-wide association studies (GWAS) and SNP results have also provided genetic evidence that glutamate signaling is implicated in the pathophysiology of BD (47, 48).

Previous neuroimaging studies in patients with BD (**Table 1**) have provided evidence that glutamate dysfunction plays a crucial role in the pathophysiology of BD. A meta-analysis of <sup>1</sup>H-MRS studies from 1980 to 2010 on brain glutamate and glutamine in BD showed that patients with BD had widespread increased Glx, including in the prefrontal cortex, compared with healthy control subjects, although no significant difference in Glu/Cr was noted (49). The finding of increased Glx in the frontal cortex was replicated in a subsequent meta-analysis (50). However, the results to date have been inconsistent, and many factors, including mood status and medication, might affect glutamate levels. For example, a recent systemic review examined the effects of lithium, a drug commonly used for mood control, by summarizing results of 26 <sup>1</sup>H-MRS studies (51). The investigators found inconclusive results regarding glutamate levels and the influence of lithium treatment.

Few PET/SPECT studies have used radioligands specifically to target glutamate receptors in BD. However, ketamine, an NMDA antagonist, has been found to have rapid antidepressant effects in depressed patients with BD and has been studied in combination with <sup>18</sup>F-FDG-PET before and after ketamine infusion. That study found that brain glucose metabolism changes in the right ventral striatum of basal ganglia were significantly correlated with depression improvement (52). Because <sup>18</sup>F-FDG PET in combination with glutamatergic agents could serve as a proxy for glutamate neurotransmission, the findings suggest that ketamine improved BD depression by promoting glutamatergic neurotransmission in brain regions involved in mood control. However, questions remain. For example, whether glutamate dysregulation is the fundamental cause of the pathophysiology of BD and whether glutamatergic agents can reverse the brain abnormalities of BD. We compared brain glucose metabolism in subtypes of BD by using <sup>18</sup>F-FDG-PET and found that patients with BD type I (compared with patients with BD type II) had significantly lower glucose metabolism in the bilateral anterior cingulum, insula, striatum, and part of the prefrontal cortex and higher glucose metabolism in some limbic structures (53). Frontolimbic dysregulation seems to play a critical role in the pathophysiology of BD, because the unaffected siblings of patients with BD have such abnormalities in a minor form that can be detected by resting-state functional MRI but not by <sup>18</sup>F-FDG-PET (54). Whether these abnormal findings are a result of glutamate dysregulation warrants further investigation. Furthermore, ppTMS research that specifically examines ICF is still lacking, despite such research demonstrating cortical inhibitory deficits in patients with BD (55).

## Major Depressive Disorder/Treatment-Resistant Depression and Glutamatergic Dysfunction

MDD is a severe psychiatric disorder characterized by episodes of depression and anhedonia. MDD is considered a severe illness because of a tendency for the illness to become chronic and a high prevalence of TRD. Evidence has revealed a pivotal role of glutamatergic neurotransmission in the pathophysiology of MDD.

Postmortem studies have revealed that expression of mGluR2/3 receptors in the anterior cingulate cortex was significantly reduced in patients with MDD (56), whereas another study found no significant difference in the anterior cingulate cortex between patients with MDD and healthy control subjects in expression of mGluR2/3 or mGluR5 (46). Furthermore, iGluRs are abnormally expressed in human subjects with MDD. For example, the expression of NMDAR subunit NR2 in the prefrontal cortex was reduced in patients with MDD (57). A study using postmortem brains of patients with MDD further demonstrated that elevated expression levels of the majority of mGluR and iGluR genes were found in the dorsolateral prefrontal cortex, and the genetic expression differences occurred mostly in female subjects (58). A metaanalysis of three large MDD GWASs (4,346 subjects with MDD vs. 4,430 control subjects) found that genes involved in glutamatergic synaptic neurotransmission were significantly associated with MDD (59).

Regarding neuroimaging findings in clinical patients (**Table 1**), <sup>1</sup>H-MRS studies in patients with MDD have revealed decreased glutamate and glutamine levels in the dorsolateral and other parts of the prefrontal cortex and increased glutamate levels in the occipital cortex (60). A meta-analysis of <sup>1</sup>H-MRS studies demonstrated that decreased Glx levels with absolute values in the prefrontal cortex were correlated with treatment severity (i.e., number of failed antidepressant treatments),

indicating that the severity of glutamatergic dysregulation could be related to the severity of illness (61). Another meta-analysis noted that glutamate and Glx concentrations were lower in the anterior cingulate cortex in patients with MDD than in control subjects (62).

A further piece of evidence comes from the surprisingly rapid antidepressant response to low-dose ketamine in the treatment of TRD. Structural and functional abnormalities in the prefrontal cortex have been found to be prominent in patients with TRD (63, 64), and intravenous low-dose ketamine (0.2–0.5 mg/kg) was revealed to reverse the prefrontal abnormalities and frontolimbic dysregulation of the human brain in 1 h (14). We applied <sup>18</sup>F-FDG-PET before and after intravenous injection of 0.5 mg/kg ketamine, 0.2 mg/kg ketamine, and placebo and found that prefrontal cortical function increased only in the low-dose ketamine groups, whereas the activation of prefrontal function correlated well with the deactivation of limbic function in the amygdala and hippocampus (**Figure 1**) (14). Moreover, recent PET research using <sup>11</sup>C-ABP688, a radioligand for mGluR5, revealed a significant ketamine-induced reduction in mGluR5 availability as reflected by decreased <sup>11</sup>C-ABP688 binding in all subjects, which persisted for more than 24 h (13). In addition, the changes of <sup>11</sup>C-ABP688 binding were correlated with the rapid antidepressant effect of ketamine. However, because ketamine is an NMDA antagonist (one type of iGluR), future studies directly investigating iGluRs in response to low-dose ketamine are warranted.

Although some ppTMS studies have found no significant differences in ICF between patients with MDD and healthy subjects (41, 65), few studies have differentiated MDD from TRD. In addition, medication could obviously confound the ppTMS findings, given that medicated euthymic patients with MDD exhibited much increased ICF compared with healthy control subjects and unmedicated patients with MDD (65). Furthermore, patients with MDD with a comorbidity of generalized anxiety disorder are more likely to have poorer responses to antidepressants than those without (66). Our team demonstrated that patients with generalized anxiety disorder in unmedicated status had decreased ICF in bilateral motor cortices, suggesting that patients with more impaired glutamatergic neurotransmission in cortical regions may have worse treatment outcomes (5). A study targeting at child and adolescence depressives (9–17 years old) and found depressed patients had significantly increased ICF (67). Future studies that control medications and more specifically target TRD over MDD without a history of antidepressant resistance are necessary.

## Glutamate Compounds That Target Glutamatergic Dysfunction

As aforementioned, glutamate receptors can be divided into iGluRs and mGluRs (6). iGluRs include NMDARs, AMPA receptors, and kainate receptors, whereas mGluRs can be categorized into group 1, group 2 (mGluR2 and mGluR3), and group 3. More recently, selective glutamate positive allosteric modulators (PAMs) have been developed that enhance glutamate receptor function in the presence of endogenous agonists without having adverse effects resulting from intrinsic activity (e.g., NMDAR PAMs) (68). By contrast, glutamate mGluR2 or mGluR5 (basimglurant) negative allosteric modulators (NAMs) may inhibit glutamate receptor function and have been tried in MDD studies, although the results are mixed (69). Therefore, given that glutamatergic dysfunction plays a crucial role in the major psychiatric disorders discussed herein, compounds aimed at activating or inhibiting the aforementioned receptors or indirectly modulating functions of receptors are of great research interest. **Table 2** lists glutamatergic compounds that have been TABLE 2 | Glutamatergic compounds in the treatment of major psychiatric disorders.


used to treat major psychiatric disorders in human clinical trials (searched in the ClinicalTrials.gov database, accessed on September 20, 2018).

The most attractive compounds are NMDA antagonists because a growing body of evidence has pinpointed glutamatergic dysfunction in the pathophysiology of TRD and demonstrated that the glutamatergic synapses present multiple targets for development of novel antidepressants. For example, ketamine and its S-enantiomer (esketamine) are NMDA antagonists and, when used in a low-dose range, have exhibited rapid antidepressant properties for TRD (70). In a low-dose range, tolerability seems to be acceptable with transient elevation of blood pressure and mild and self-limited psychotomimetic effects (70). Several other compounds, such as dextromethorphan, memantine, traxoprodil, AZD6765, and riluzole, among others (see **Table 2** and **Figure 2**), have similar pharmacological properties and have great potential in treating MDD and BD depression. Using a small sample size (n = 14 completed the study), researchers in an open-label study found that dextromethorphan/quinidine (Nuedexta; Avanir Pharmaceuticals, Inc., Aliso Viejo, CA, USA) could decrease depression scores with acceptable tolerability in patients with TRD (71). Traxoprodil, an NR2B subunit-selective NMDAR antagonist, in combination with paroxetine was found to decrease depression scores for patients with TRD (n = 30) (72). A single intravenous dose of AZD6765 (a low-trapping NMDA channel blocker) was also found to have rapid but short-lived antidepressant effects in a small trial (n = 22) (73). However, longer-duration and larger studies are required to prove clinical efficacy because not all NMDA antagonists possess rapid antidepressant efficacy. For example, an NMDA antagonist, memantine (74), had been shown to lack rapid antidepressant effects. Likewise, riluzole was found to have limited effects for ketamine nonresponders (75). Notably, sarcosine, an NMDA enhancer, had been found to improve depression-like behavior in rodent models and depression in humans (76). Some iGluR NMDA-related compounds are also used for treating schizophrenia (**Table 2**). For example, D-serine, a naturally occurring NMDAR glycine site agonist, was found to have significant effects on auditory mismatch negativity that correlated significantly with change in symptoms of schizophrenia in a small double-blind crossover trial (n = 16) (77). Researchers in a randomized, multicenter, double-blind, placebo-controlled study investigated adjunctive RO4917838 (bitopertin), a selective GlyT1-mediated glycine reuptake inhibitor, in patients with schizophrenia with suboptimally controlled symptoms (n = 1,772) and found that the antipsychotic effects were small (mean difference vs. placebo in score, −1.37) and were demonstrated in only one of six active treatment arms (78). Some trials have been withdrawn or terminated (**Table 2**), including a trial using EVT-101 for MDD, trials using Org 24448 for MDD and schizophrenia, a trial using diazoxide for MDD, and a trial using ceftriaxone for schizophrenia. The reasons may include prominent side effects or a lack of clinical efficacy, so the final outcomes of these clinical trials must be awaited.

Glutamatergic compounds are also used for treating BD depression. For example, researchers in a randomized, doubleblind, placebo-controlled study investigating a single intravenous infusion of ketamine (0.5 mg/kg) for BD depression found that depressive symptoms and suicidal ideation significantly improved within 40 min (79). In addition, a small proof-ofconcept study investigated the effects of ketamine and Dcycloserine in patients with BD, in which subjects received openlabel ketamine hydrochloride (0.5 mg/kg intravenously over 60 min) followed by 8 weeks of adjunctive D-cycloserine (titrated up to 1000 mg/d from a starting dose of 250 mg over 3 weeks) (80). The investigators found that four of seven subjects met remission criteria at 8 weeks. However, a randomized controlled study of riluzole monotherapy (50–200 mg/d) for BD depression was terminated early (**Table 2**) because of a high number of subject withdrawals and no significant antidepressant effects of riluzole (81).

Although most iGluR NMDA compounds seem to be used for treating MDD (**Table 2**), iGluR-AMPA and mGluR compounds are generally used for treating schizophrenia (**Table 2**). For example, CX516 (an AMPA receptor-positive modulator and also the first ampakine) was used for cognitive enhancement in schizophrenia; however, the results appear to be disappointing because CX516 was not effective for cognition or for symptoms of schizophrenia when added to clozapine, olanzapine, or risperidone (82). N-acetylcysteine is a widely available dietary supplement that may restore glutamate to its correct levels in the brain and is used for treating cognitive deficits in schizophrenia (**Table 2**). LY2140023 as an mGluR2/3 agonist was originally found to have significant antipsychotic effects in patients with schizophrenia early in disease or in those previously treated with D2 drugs (83), but investigators terminated another study because LY2140023 failed to achieve significant effects on the overall symptoms of schizophrenia (**Table 2**).

Hypofunction of NMDARs has been suggested to play an important role in the pathophysiology of schizophrenia, and glutamate PAMs may be effective for treating schizophrenia and related cognitive deficits. However, to date, the results have been inconsistent. For example, authors of a recent metaanalysis that included 17 randomized, placebo-controlled studies (n = 1,391) found that glutamate PAMs were not superior to placebo in improving cognitive function in schizophrenia (84). In addition, results of mGluR2 activators for schizophrenia seem to be inconsistent. For example, Eli Lilly's mGluR2/3 agonist (LY2140023) failed to meet the primary endpoints in a phase II trial (85), and another phase III trial was stopped mainly owing to the lack of efficacy; however, JNJ-40411813 (a PAM being developed by Janssen Pharmaceutica NV Beerse, Belgium, and Addex Therapeutics, Geneva, Switzerland) has been shown to have effects on negative symptoms in patients with schizophrenia in a phase II trial (85). Another mGluR2 receptor PAM (ADZ8529; AstraZeneca, Cambridge, UK) failed to separate from placebo in total, negative, and positive symptoms of schizophrenia in a phase II trial (86).

A recent paper published in JAMA Psychiatry in which researchers evaluated basimglurant (mGluR5 NAM) for patients with MDD with inadequate antidepressant responses in the current episode found that the primary endpoint (mean change in clinician-rated depression score from baseline to endpoint) was not met, but an antidepressant effect on patient-rated measures was found across secondary endpoints (87). However, an NAM targeting mGluR2/3 (decoglurant) was withdrawn, mainly owing to disappointing antidepressant efficacy (**Table 2**).

# CONCLUSION

Although most techniques today indirectly measure glutamatergic neurotransmission in vivo, accumulating evidence has revealed that glutamatergic dysfunction plays a crucial role in major psychiatric disorders, such as schizophrenia, BD, and MDD (including TRD). Ketamine, esketamine, and many other pharmacological compounds targeting the glutamate system are available for human trials of major psychiatric disorders. However, longer-duration and larger studies are required to prove their clinical efficacy in different psychiatric diseases.

### AUTHOR CONTRIBUTIONS

C-TL: conceived and designed the study; C-TL, K-CY, and W-CL: performed the analysis and review; C-TL, K-CY, and W-CL: wrote the paper; C-TL, K-CY, and W-CL: approved the paper.

### FUNDING

This study was sponsored by grants from the Ministry of Science and Technology (MOST 106-2314-B-075-034-MY3). The aforementioned organization had no influence on the study design, data collection, analysis, interpretation of results, writing of the report, or ultimate decision to submit the paper for publication.

### REFERENCES


### ACKNOWLEDGMENTS

We would like to thank all of the research assistants, physicians, and personnel who helped us with this study.


robust preclinical data but inconclusive clinical efficacy. Biol Psychiatry (2018) 83:955–62. doi: 10.1016/j.biopsych.2018.03.001


depression: a randomized controlled add-on trial. Biol Psychiatry (2012) 71:939–46. doi: 10.1016/j.biopsych.2011.12.010


**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 © 2019 Li, Yang and Lin. 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.

# D-Serine: Potential Therapeutic Agent and/or Biomarker in Schizophrenia and Depression?

Mary-Anne B. MacKay 1,2, Maryana Kravtsenyuk 1,2, Rejish Thomas <sup>1</sup> , Nicholas D. Mitchell 1,2, Serdar M. Dursun1,2 and Glen B. Baker 1,2 \*

*<sup>1</sup> Neurochemical Research Unit and Bebensee Schizophrenia Research Unit, Department of Psychiatry, University of Alberta, Edmonton, AB, Canada, <sup>2</sup> Neuroscience and Mental Health Institute, University of Alberta, Edmonton, AB, Canada*

D-Serine is a potent co-agonist at the NMDA glutamate receptor and has been the object of many preclinical studies to ascertain the nature of its metabolism, its regional and cellular distribution in the brain, its physiological functions and its possible clinical relevance. The enzymes involved in its formation and catabolism are serine racemase (SR) and D-amino acid oxidase (DAAO), respectively, and manipulations of the activity of those enzymes have been useful in developing animal models of schizophrenia and in providing clues to the development of potential new antipsychotic strategies. Clinical studies have been conducted in schizophrenia patients to evaluate body fluid levels of D-serine and/or to use D-serine alone or in combination with antipsychotics to determine its effectiveness as a therapeutic agent. D-serine has also been used in combination with DAAO inhibitors in preclinical investigations, and interesting results have been obtained. Genetic studies and postmortem brain studies have also been conducted on D-serine and the enzymes involved in its metabolism. It is also of considerable interest that in recent years clinical and preclinical investigations have suggested that D-serine may also have antidepressant properties. Clinical studies have also shown that D-serine may be a biomarker for antidepressant response to ketamine. Relevant to both schizophrenia and depression, preclinical and clinical studies with D-serine indicate that it may be effective in reducing cognitive dysfunction.

### Edited by:

*Kenji Hashimoto, Chiba University, Japan*

### Reviewed by:

*Tianmei Si, Peking University Sixth Hospital, China Yilang Tang, Emory University, United States*

> \*Correspondence: *Glen B. Baker glen.baker@ualberta.ca*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *28 October 2018* Accepted: *15 January 2019* Published: *06 February 2019*

### Citation:

*MacKay M-AB, Kravtsenyuk M, Thomas R, Mitchell ND, Dursun SM and Baker GB (2019) D-Serine: Potential Therapeutic Agent and/or Biomarker in Schizophrenia and Depression? Front. Psychiatry 10:25. doi: 10.3389/fpsyt.2019.00025* Keywords: D-serine, D-amino acids, schizophrenia, depression, serine racemase, D-amino acid oxidase

# INTRODUCTION

Several amino acids have a chiral center and thus can exist as D- and L-isomers. For many years, it was thought that only the L-isomers of these amino acids existed in mammalian tissue. However, it was discovered in the 1990s that relatively large quantities of free D-serine exist in the mammalian brain (1–3), although at lower concentrations than L-serine (1–5). Free D-aspartate and D-alanine (**Figure 1**) are also present in brain at levels much lower than those of D-serine and of their respective L-isomers, but still measureable (2, 3, 5, 7–13). Interestingly, it has been reported that all three of these D-amino acids may contribute to brain function (14–22) and may be useful as adjuncts in the therapy of schizophrenia (4, 7, 10, 17, 20–22).

The focus of this review is on D-serine and its possible involvement with both schizophrenia and depression. D-Serine is a potent coagonist at the N-methyl-D-aspartate (NMDA) glutamate receptor and appears to have a major modulatory role in NMDA receptor-mediated neurotransmission, neurotoxicity, synaptic plasticity, and cell migration (5, 8, 15, 18–23).

Considerable research now indicates that it may be a potential therapeutic agent and/or biomarker in both schizophrenia and major depressive disorder, as will be discussed in this review paper.

# METHODS

Searches were done in PubMed and Web of Science covering the period 1990–2018 and the key phrases "D-serine and neuropsychiatric disorders," D-serine and schizophrenia," "Dserine and depression," "D-amino acids in neuropsychiatric disorders," and "D-serine and ketamine" were used in the searches. Only papers in English were used in the preparation of this review. The references obtained were screened by the authors to determine which would be best to put in this paper.

# D-SERINE AS A POSSIBLE BIOMARKER AND/OR THERAPEUTIC AGENT IN SCHIZOPHRENIA

There is now a large body of evidence supporting hypofunction of NMDA glutamate receptors in schizophrenia (24–27). Because D-serine is such a potent coagonist at the NMDA receptor, there has been a great deal of interest in its role in the brain. D-Serine is present in glia (mainly astrocytes) and neurons. It has been proposed as both a glial transmitter (28, 29) and a neurotransmitter (30), and this has resulted in considerable controversy [see (29, 30) for an interesting discussion of the relevant importance of glia and neurons in the actions of D-serine]. Wolosker etal. (30) have proposed that astrocytes synthesize L-serine which then shuttles to neurons to be converted to D-serine.

The NMDA glutamate receptor requires not only glutamate but a coagonist in order to be activated. For many years, it was thought that glycine was the coagonist and the site at which it acts on the NMDA receptor is termed the glycine binding site. There is now considerable evidence, including a regional distribution more closely resembling that of NMDA receptors than is the case with glycine (4) and a stronger affinity than glycine for the glycine binding site on the NR1 subunit of the NMDA receptor (31, 32), indicating that D-serine may be more important in this regard. Because of the large body of evidence indicating hypofunctioning of the NMDA glutamate receptor in schizophrenia, D-serine has become of great interest to many researchers, and it has been proposed as the primary NMDA receptor coagonist in the forebrain and hippocampus (14). Glycine and D-serine appear to act at different NMDA receptor populations, D-serine at synaptic receptors and glycine at extrasynaptic receptors (33). It has been proposed that synaptic NMDA receptors are neuroprotective and that extrasynaptic receptors may promote cell death (34).

A number of preclinical studies in rodents have demonstrated that lowering brain levels of D-serine by reducing the activity of serine racemase (SR), the enzyme responsible for catalyzing formation of D-serine from L-serine (e.g., **Figure 1**), can produce symptoms reminiscent of clinical symptoms in schizophrenia: stereotypies, cognitive deficits, disruption of prepulse inhibition (measure of sensorimotor gating), persistent latent inhibition (measures inhibitory learning and cognitive flexibility), and deficits in social interaction (6, 19–22, 35). It has also been reported that SR knockout mice show a reduction of basal NMDA receptor activity and reduced arborization and spine density in dendrites (36). Chronic D-serine reverses expression of activityrelated cytoskeleton-associated protein (Arc) and causes partial rescue of dendritic abnormalities in the same model (36).

Spatial and reversal memory deficits in Sprague-Dawley rats treated with the NMDA receptor antagonist phencyclidine (PCP) at different developmental stages can be reversed with Dserine administration (37). Hagiwara etal. (38) conducted an experiment with SR-inhibited mice where D-serine was given as a supplement in the preadolescent phase, and this had some benefit in preventing adult onset psychosis, suggesting the possibility of using D-serine supplementation in early intervention in humans. Fujita etal. (39) reported that juvenile and adolescent rodents that had been exposed prenatally to maternal immune activation showed reduced expression of hippocampal NMDA receptor subunits and onset of cognitive deficits as adults; supplementing their drinking water with D-serine from P28 to P56 reduced cognitive deficits.

For further information about SR and D-amino acid oxidase (DAAO), see the next section.

## FORMATION AND CATABOLISM OF D-SERINE AND RELEVANCE OF THIS METABOLISM TO SCHIZOPHRENIA

D-Serine formation from L-serine is catalyzed via the enzyme SR and its catabolism is catalyzed by DAAO (40–46) (**Figure 2**).

In the brain, SR and D-serine are found in the same regions, i.e., high levels in forebrain areas such as cortex and hippocampus and very much lower levels in the cerebellum and brain stem (35). Activity of SR can be modulated through various α-amino-3-hydroxy-5-methylisoxazole-4-propionic

acid (AMPA) receptor-dependent, metabotropic mGluR2/3/5 receptor-dependent mechanisms, divalent cations and the adenosine triphosphate (ATP) pathway, suggesting potential targets by which D-serine concentrations could be enhanced (47–49). AMPA-induced postsynaptic membrane depolarization builds up intracellular calcium concentrations and stimulates SR with the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) (48). Ma etal. (50) reported SR interactions with stargazin and the scaffolding protein PSD-95 and suggested that these proteins regulate NMDA receptor-AMPA receptor cross-talk in neurons. Lin etal. (51) proposed an association of D-serine with PSD-95 and NMDA receptors in postsynaptic neurons and with stability of glutamatergic synapses during development of cortical synapses. Glutamate-receptorinteracting-protein (GRIP) forms a complex with mGluR2/3 receptors, and receptor activation changes the conformation of SR, altering its function and decreasing production of D-serine (49).

Single nucleotide polymorphisms in the gene encoding protein interacting with C-kinase 1 (PICK1), a component of protein kinase C signaling, have been reported to be associated with a higher risk of schizophrenia, and it has been suggested that this link could be mediated through the interaction with SR (52). Ma etal. (53) proposed that pathogenic disruption of Disrupted-In-Schizophrenia-1 (DISC1)-SR binding can produce schizophrenia-like behavior by depleting D-serine levels. DISC1, the perturbation of which has been implicated in the pathophysiology of a number of mental disorders, including schizophrenia, major depressive disorder and bipolar disorder (54), binds to SR and stabilizes it. Using a mouse model of schizophrenia, Ma etal. (53) found that mutant DISC1 results in SR degradation and a D-serine deficiency. Svane etal. (55) administered D-serine to male and female rats and measured expression of nitric oxide synthase1 adaptor protein (NOS1AP) (overexpressed in cortex of patients with schizophrenia), D2 receptors and DISC1 and found that it affects expression of these three genes in a sex-specific manner. Mustafa etal. (56) reported that NO S-nitrosylates SR, mediating feedback inhibition of formation of D-serine. Interaction of SR with Colga 3, a member of the Colgin subfamily A, may also be important since Colga 3 decreases the ubquitylation of SR, resulting in protection of the SR from degradation by the ubiquitin-proteasomal system (57). In their SR knockout mouse model (58), Balu and Coyle (59) observed reduced binding of cyclic adenosine monophosphate (cAMP)-responsive element binding (CREB) to the promoter regions of genes for three molecules implicated in the pathophysiology of schizophrenia, namely brain-derived neurotrophic factor (BDNF), microRNA-132, and Arc (60). In a recent review article, Wolosker has provided a useful table of regulators of SR activity and D-serine production (61).

In contrast to the distribution of SR and D-serine in the brain, DAAO is most abundant in the cerebellum and brain stem and at much lower concentrations in pre-frontal cortex, hippocampus, and substantia nigra (13, 35). It has been reported that the activity of DAAO in postmortem brain tissue from humans with schizophrenia is increased over that of controls (62, 63). Labrie etal. (41) found that genetic loss of DAAO reverses the schizophrenia-like phenotypes in mice displaying these behaviors because of a mutation in the NR1 subunit of the NMDA receptor. DAAO has been proposed to be regulated by the protein product of gene G72 (64–67), and variations in pLG72 have been associated with schizophrenia (68–72). There has been some discussion in the literature about the various functions of pLG72 in the brain and the molecular details of its interaction with DAAO, and these matters have been reviewed comprehensively by Pollegiani etal. (73). Hashimoto etal. (74) found that administering a DAAO inhibitor and Dserine simultaneously reduces prepulse inhibition seen in mice given dizocilpine, a non-competitive NMDA receptor antagonist. Other researchers have also co-administered DAAO inhibitors in conjunction with D-serine in preclinical studies and suggested that DAAO inhibitors could be useful in schizophrenia by reducing the required dose for D-serine (40, 74–77).

## CLINICAL STUDIES WITH D-SERINE IN SCHIZOPHRENIA

Levels of D-serine have been reported to be decreased in cerebrospinal fluid (CSF) and blood of schizophrenia patients (18, 78–84). Hashimoto etal. (84) reported that serum levels of D-serine in schizophrenia patients are lower than in controls, while D-serine serum levels in mood-stabilized bipolar disorder patients are higher than those in controls, and they felt that this situation may reduce misdiagnosis between these disorders. Further, plasma levels of D-serine and the D-/L-serine ratio were increased significantly by clozapine in schizophrenia responders (85).

Clinical studies have been conducted in schizophrenia with D-serine alone and as an adjunct to antipsychotics. In a 4-week open-label study it was reported that D-serine at high dose (60 mg/kg/day) resulted in improved neurocognitive function (86). The same group did follow-up studies on D-serine (87, 88). In one investigation they did a double-blind, placebo-controlled, parallel group randomized clinical trial on negative symptoms in at-risk individuals, and D-serine was given at a dose of 60 mg/kg/day in divided daily doses for 16-weeks; it was concluded that D-serine should be useful for treatment of prodromal symptoms. In another double-blind study, they found that in schizophrenia patients treated with D-serine at 60 mg/kg/day, there was a significant improvement in mismatch negativity (MMN) (auditory mismatch), a neurophysiological biomarker for NMDA receptor activity. Ermilov etal. (89) compared high dose (3 g/day) D-serine with high dose olanzapine in treatmentresistant schizophrenia patients and concluded that a subgroup of patients could be maintained on D-serine. Interestingly, doses of D-serine as high as 4 g per day have been reported to cause no adverse effects (18).

Several clinical studies with D-serine have used it as a potential adjunctive drug (30–120 mg/kg/day) to antipsychotics. When added to non-clozapine antipsychotics in schizophrenia patients, D-serine has been reported to improve negative, positive, and cognitive symptoms (90–94) and to be relatively free of side effects even at high doses. A double-blind placebocontrolled study by Heresco-Levy etal. (91) examined the effect of supplemental D-serine in a group of patients stabilized on olanzapine or risperidone; at the end of 6-weeks of treatment, the D-serine intervention group demonstrated an improvement in cognitive, positive, and negative symptom domains and a significant alteration in depressive symptoms. Simultaneous measurement of serum amino acid concentrations did not detect any fluctuations of levels of any amino acids except D-serine. Meta-analyses of clinical trials comparing therapy with glycine modulatory site agonists, including D-serine, concluded that all of these NMDA receptor agonists significantly ameliorate symptoms in multiple domains, including cognitive and affective symptoms, when added to atypical antipsychotics except for clozapine (92–94). However, negative studies on D-serine adjunctive therapy in schizophrenia have also been reported in the literature (95–98). There have been inconsistent results as to the therapeutic benefit of D-serine used at 30 mg/kg/d to improve the negative and cognitive symptoms of the illness, with more consistent improvements found at doses of 60 mg/kg/d or higher. The higher doses of D-serine may be needed to adequately potentiate NMDA receptor-mediated activation of the receptor and to also achieve adequate serum levels and a subsequent predictive increase in brain concentrations.

It is interesting that D-serine addition to clozapine does not increase the efficacy of clozapine (99); this may be because clozapine releases D-serine and glutamate (100) and may have agonist or partial agonist activity at NMDA receptors (99–101). It is also possible that individuals on clozapine do not respond to Dserine since they are more often older and/or treatment-resistant (99).

One of the problems of treating symptoms of schizophrenia with D-serine is the fact that it is metabolized rapidly by DAAO, reducing its bioavailability and requiring administration of high doses, which could lead to peripheral neuropathies. There are also safety concerns that high concentrations of D-serine can cause potential nephrotoxicity related to its metabolism by DAAO as has been reported in rats that have developed acute tubular necrosis associated with higher doses of D-serine (102). However, in clinical studies that have administered Dserine (4-weeks duration) in doses up to 120 mg/kg/day, there have been no significant side effects (including nephrotoxicity) reported (86). In this particular study, one patient who received 120 mg/kg/d of D-serine did show 2+ proteinuria without glycosuria during the last week of treatment with no change in creatinine which completely resolved following the discontinuation of D-serine. In a 16-week intervention study by the same group using 60 mg/kg/d in clinically high-risk individuals, there were two patients who were discontinued in relation to abnormal renal values associated with treatment, with all other patients' renal abnormalities reported being resolved with continued treatment (87). The long-term side effects of D-serine beyond a 16-week treatment period are currently unknown. Additional studies that include longer time intervals should be conducted on this aspect of D-serine to ensure patient safety.

Studies on DAAO inhibitors indicate that they have limited effects on brain D-serine levels in mice, although there is some disagreement in the literature on this [see Guercio and Panizzutti (103) for a discussion]. However, as mentioned in the above section on enzymes involved in metabolism of D-serine, several researchers have now co-administered DAAO inhibitors in conjunction with D-serine in preclinical studies, and their findings suggest that DAAO inhibitors could be useful clinically in schizophrenia for reducing the required dose of D-serine and thus also reducing potential side effects associated with the administration of high D-serine (74–77). Sodium benzoate, a frequently used food preservative, is also a DAAO inhibitor and has been reported to have beneficial effects in schizophrenia when added to antipsychotics (104).

# D-SERINE AND MAJOR DEPRESSIVE DISORDER

As indicated above, there is now a strong body of evidence indicating the possible involvement of D-serine in the etiology of schizophrenia and suggesting its potential as an antipsychotic or an adjunct to existing antipsychotics. It has thus been very interesting to see research results indicating that it may also have antidepressant effects and/or be a potential biomarker for depression and response to the antidepressant effects of ketamine. The possible mechanisms involved in the antidepressant actions of D-serine are the subject of another paper in this volume and of a paper by Chan etal. (105) and will not be discussed here in detail.

As mentioned above, in clinical studies investigating D-serine as an adjunct to antipsychotics, improvement in the affective symptoms of these patients was also often observed (e.g., 89– 94). D-serine has now been reported to have antidepressant-like effects in rodent models of depression (105–107). Malkesman etal. (106) studied the behavioral effects in mice of a single, acute i.p. dose of D-serine in several tests of antidepressants, including the forced swim test, the female urine sniffing test following serotonin depletion and the learned helplessness paradigm and in mice lacking NR1 expression in excitatory neurons in the forebrain and found that D-serine gave a positive response in each of the behavioral tests but that the same behavioral tests conducted in mice lacking NR1 expression did not respond to D-serine. Otte etal. (107) studied transgenic mice in which SR was overexpressed using paradigms of anxiety, depression, and cognition. These studies were done in the absence and presence of D-serine in the drinking water. D-Serine administration resulted in a reversal of the findings in these tests, suggesting a reduced proneness toward depression-related behavior. Wei etal. (108) studied the effects of D-serine and ketamine on rats in the forced swim test and proposed that D-serine produces antidepressant-like effects through the same mechanisms as ketamine. Wang etal. (109) found that chronic social defeat stress (CSDS) in mice induces expression of the D-serine transporter through epigenetic activation and decreases levels of D-serine in the hippocampus, leading to depression-like behavior.

The DAAO inhibitor sodium benzoate has been reported to have beneficial effects in a depressed patient (110) and acute administration of D-serine to healthy subjects has been reported to reduce subjective feelings of depression and anxiety as measured by Visual Analog Scales (111). Ishiwata etal. (112) reported that CSF levels of D-serine in depressed patients correlate negatively with severity of depression. In a comprehensive study on 70 depressed patients, Hashimoto etal. (113) found increased serum levels of both serine enantiomers compared to controls, while serum levels of glycine, glutamate, and glutamine did not differ between the depressed patients and controls. In contrast, Mitani etal. (114) had reported that plasma levels of D- and L-serine in major depressive disorder (MDD) patients were the same as those in healthy controls and levels of glutamate, glutamine, and glycine were higher in MDD patients than in controls. Hashimoto etal. (113) suggested that differences in severity of the depression between the two studies might account for some differences. Because of their findings of increased serum levels of D- and L-serine in MDD patients and a higher ratio of L-serine to glycine levels in the MDD patients, Hashimoto etal. (113) suggested abnormalities in synthesis and catabolism of serine enantiomers in MDD.

It is now well-accepted that intravenous administration of the NMDA receptor antagonist ketamine can produce a rapid improvement of symptoms of depression (115, 116). As noted above, several studies indicate that D-serine, an NMDA receptor coagonist, can also produce antidepressant effects. Malkesman etal. (106) state that the two drugs may activate common, convergent downstream targets, resulting in similar effects on protein expression and producing comparable changes in synaptic plasticity and dendritic remodeling. There are some interesting interactions between D-serine and ketamine reported. There is now evidence indicating that D-serine may be a predictive biomarker for antidepressant response to ketamine, with low plasma D-serine levels predicting a response to (R,S)-ketamine (117). Singh etal. (118) examined the effect of (R)- and (S)-ketamine on Alanine, Serine, Cysteine Transporter 2 (ASCT2)-mediated transport of Dserine in adrenal pheochromocytoma PC-12 and human neural astrocytoma 1321N1 cells, and primary neuronal cells in culture and reported that (S)-ketamine decreased cellular export by selectively inhibiting ASCT2; the authors suggested that this interaction might represent a source of dissociative effects seen with (R,S)-ketamine. Singh etal. (118) also incubated PC-12 cells with a variety of ketamine metabolites and determined the IC<sup>50</sup> values associated with attenuation of intracellular D-serine and proposed to use the findings to help in the design of more efficient modulators of Dserine.

### D-SERINE AND COGNITION

Serious cognitive deficits can occur in a number of psychiatric disorders, including both schizophrenia and MDD. It has been reported that depletion of D-serine levels in brain diminishes long-term potentiation (LTP), which is associated with learning and memory (19, 20). Genetic inactivation of SR in a mouse model and an acute stress model reduced brain levels of D-serine and produced cognitive deficits (6, 119). Intraperitoneal injection of D-serine results in improved social memory in rats (120) and improved recognition and working memory in mice (121). In a rodent model study following development of offspring after prenatal maternal infection [using poly(I:C)], it was reported that supplementing offspring at juvenile and adolescent stages with D-serine reduced cognitive deficits in adulthood (39). Panizzutti etal. (122) studied the association between serum D-serine levels and the results of intensive cognitive training in schizophrenia patients and found that in those patients receiving this training increased D-serine levels were positively correlated with improved global cognition and verbal learning, but such associations were not apparent with glycine. Despite the interesting findings mentioned above, D-serine is not used clinically as a cognitive enhancer; in a recent review, Guercio and Panizzutti (103) have discussed various factors, including pharmacokinetics and possible side effects that must be studied in more detail in order to increase the efficacy of D-serine.

# SUMMARY

In contrast to the usual situation with D-amino acids, Dserine is abundant in the brain and appears to have important neuromodulatory roles. D-Serine is more potent than glycine as a coagonist at the NMDA receptor, has a regional distribution in the brain that is similar to that of NMDA receptors and appears to be more closely associated with synaptic NMDA receptors than glycine (which is more closely associated with non-synaptic NMDA receptors). There has been considerable controversy about the concentration and function of D-serine in glia vs. neurons, but evidence in recent years indicates that neurons are more relevant to D-serine action than originally thought. Synthesis and catabolism of D-serine are catalyzed by SR and DAAO, respectively. Regulation of these enzymes by various factors is quite complex. Decreased levels of Dserine in schizophrenia have been reported in several studies. Increasing D-serine levels in the brain may be an effective adjunct to antipsychotic treatment. D-Serine administered alone or in combination with usual antipsychotics may be useful in treating schizophrenia, but the high doses required may cause peripheral neuropathies, although these have not been evident thus far using doses up to 4 g/day. Nonetheless, more studies at longer time intervals should be conducted on this aspect of D-serine to ensure patient safety. Several studies have been conducted with D-serine in combination with a DAAO inhibitor, with generally promising results obtained. Based on preclinical and clinical findings, D-serine also appears to have antidepressant properties. Several interesting pharmacodynamic and pharmacokinetic interactions between D-serine and ketamine have been reported, and, interestingly, evidence suggests that low plasma levels of D-serine may predict positive antidepressant response to ketamine. Several animal model and clinical studies also indicate that D-serine may be effective in reducing cognitive deficits, but that further study is necessary before considering it an effective cognitive enhancer for routine use in humans.

## REFERENCES


# AUTHOR CONTRIBUTIONS

GB and M-AM wrote the first draft. All authors (GB, M-AM, MK, RT, NDM, SMD) contributed to the literature search and the editing of the manuscript.

# FUNDING

Funding for GB's research has been provided by CIHR (MOP-86712) and the University of Alberta (FoMD TRIP research grant).

# ACKNOWLEDGMENTS

The authors are grateful to the Canadian Institutes for Health Research, University of Alberta, and Alberta Health Services for research funding and to Trudy Valliere for assistance in typing this manuscript and Patricia Kent for preparing the figures.


glutamate receptor interacting protein and mediation of neuronal migration. Proc Nat Acad Sci USA. (2005) 6:2105–10. doi: 10.1073/pnas.0409723102


function during verbal fluency. Hum Brain Mapp. (2012) 33:143–53. doi: 10.1002/hbm.21198


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

# The Effects of Acupuncture on Glutamatergic Neurotransmission in Depression, Anxiety, Schizophrenia, and Alzheimer's Disease: A Review of the Literature

### Cheng-Hao Tu<sup>1</sup> , Iona MacDonald<sup>1</sup> and Yi-Hung Chen1,2,3 \*

*<sup>1</sup> Graduate Institute of Acupuncture Science, China Medical University, Taichung, Taiwan, <sup>2</sup> Department of Photonics and Communication Engineering, Asia University, Taichung, Taiwan, <sup>3</sup> Chinese Medicine Research Center, China Medical University, Taichung, Taiwan*

### Edited by:

*Kenji Hashimoto, Chiba University, Japan*

### Reviewed by:

*Marek Schwendt, University of Florida, United States Li-Hong Kong, Hubei University of Chinese Medicine, China Tuya Bao, Beijing University of Chinese Medicine, China*

> \*Correspondence: *Yi-Hung Chen yihungchen@mail.cmu.edu.tw*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *01 September 2018* Accepted: *10 January 2019* Published: *12 February 2019*

### Citation:

*Tu C-H, MacDonald I and Chen Y-H (2019) The Effects of Acupuncture on Glutamatergic Neurotransmission in Depression, Anxiety, Schizophrenia, and Alzheimer's Disease: A Review of the Literature. Front. Psychiatry 10:14. doi: 10.3389/fpsyt.2019.00014* Neuropsychiatric disorders, including depression, anxiety, schizophrenia, and Alzheimer's disease (AD), are diseases that are directly or indirectly associated with cerebral dysfunction and contribute significantly to disability in adult populations worldwide. Important limitations surround the currently available pharmacologic agents for neuropsychiatric disorders and, moreover, many patients fail to respond to these therapies. Acupuncture might be a complementary therapy for neuropsychiatry disorders. In this review, we investigate the current evidence for the treatment efficacy of acupuncture in depression, anxiety, schizophrenia, and AD. Secondly, we review recent advances in understanding of the dysregulated glutamate system underlying the pathophysiology of these disorders. Finally, we discuss the ways in which acupuncture treatment can potentially modulate glutamate receptors and excitatory amino acid transporters. We conclude that the treatment effects of acupuncture may be underpinned by its intervention in the dysregulated glutamate system. Further preclinical and clinical studies are needed to clarify the possible mechanisms of acupuncture in these neuropsychiatric disorders and to establish protocols for treatment guidelines.

Keywords: acupuncture, glutamate, neuropsychiatric disorders, Alzheimer's disease, depression, anxiety, schizophrenia

# INTRODUCTION

Neuropsychiatry focuses on illness relating to altered cognition, mood, or behavior caused by cerebral dysfunction with neuronal pathological changes (e.g., dysregulated neurotransmitter systems or tissue damage), or abnormal physiological conditions (e.g., hyper/hypoglycemia or hypoxia). These changes in health conditions have profound effects upon individuals and society (1). Neuropsychiatric disorders contribute to over 10% of disability worldwide, exceeding the morbidity rates associated with cardiovascular disease or cancer (2). In developed-market economies, 25% of all disability has been attributed to neuropsychiatric disorders (2). In 2016, ∼18% of adults aged ≥18 years in the United States of America had any mental illness in the past year and ∼4% had a serious mental illness in that period (3, 4). A serious mental illness is defined as a diagnosable mental, behavioral, or emotional disorder (e.g., major depressive disorder [MDD], schizophrenia) causing serious functional impairment and substantially interfering with or limiting one or more major life activities, such as maintaining interpersonal relationships, activities of daily living, self-care, employment, and recreation (3, 4). Similarly, a systematic review of epidemiological data from 16 European countries estimated in 2005 that 27% of the adult population (18–65 years of age) had experienced at least one mental disorder (e.g., substance use, psychosis, depression, anxiety, or eating disorder) in the past 12 months (5). Despite the high burden of psychiatric illness, only a subset of these people receive the mental health services that they need. For instance, according to data from the World Health Organization (WHO) European Region, 3 out of 4 people with MDD are inadequately treated (6). The majority of psychiatric disorders are mild or moderate, but if left untreated, the evidence suggests that they can develop into more serious illness (2).

Important limitations surround the currently available treatments for neuropsychiatric disorders. For instance, first-line pharmacotherapy for MDD typically consists of a selective serotonin reuptake inhibitor, a serotonin and norepinephrine reuptake inhibitor, or the norepinephrinedopamine reuptake inhibitor bupropion, alone or in combination with psychotherapy. These therapies are associated with low remission rates and high dropout rates (7). As for the treatment of schizophrenia, the typical and atypical antipsychotics are mostly effective for the positive symptoms (hallucinations, delusions); cognitive and negative symptoms (deficits in working memory and attention, negative affect, and anhedonia) are largely unresponsive to current pharmacologic therapies. Moreover, serious side effects limit the use of some otherwise effective medications (8).

## ACUPUNCTURE IN NEUROPSYCHIATRY TREATMENT

Acupuncture has long been used in Chinese medicine to treat numerous neuropsychiatric conditions, from acute delirium to post-stroke spasticity. The two forms of acupuncture manipulation that are used clinically are manual acupuncture (MA) and electroacupuncture (EA) (9). Traditional acupuncturists commonly use MA, whereby they insert the acupuncture needle into the acupoint to a certain depth and rotate it by hand. In EA, the needles are connected to an electrical stimulator that delivers a stimulating current to the acupoints. Another method that is also described as EA is the positioning of a surface electrode on the skin at the acupoint, without insertion of acupuncture needles. For this review, the evidence on the use of acupuncture in neuropsychiatry treatment is limited to investigations using MA and EA, with acupuncture needle insertion.

Numerous clinical reports from various sources, including the non-Western scientific literature, attest to the efficacy of acupuncture in depression (10–13), anxiety disorders (14, 15), schizophrenia (16–19), and Alzheimer's disease (AD) (20–23). Although much of this evidence is widely acknowledged to be of varying quality, many reports attest to the efficacy and safety of acupuncture treatment (24–27). In experienced hands, acupuncture is a safe therapy with a low risk of adverse events. Serious and potentially life-threatening acupuncturerelated complications, including transmission of infections, pneumothorax, cardiovascular lesions, and hemorrhage or hematomas in the central nervous system (CNS), are very rarely reported (28). In a large study from Germany that included 2.2 million acupuncture sessions in 229,230 patients, the overall incidence of acupuncture-related adverse events was 8.6%, among which 2.2% of the patients required medical treatment (29). The vast majority of adverse events were due to minor bleeding/hematomas (6.1%), pain (1.7%), or vegetative symptoms such as vertigo, or nausea (0.7%). A review of three Chinese trials involving nearly 2,000 treatments identified instances of subcutaneous hematoma, bleeding and needle site pain, and reported that elderly people seem to be at greater risk of such adverse events (30).

# Acupuncture for the Treatment of Depression

The effectiveness of acupuncture in depression has been extensively investigated with various sets of acupuncture points and treatment parameters (e.g., duration, frequency, and number of treatment sessions). A recent Cochrane systematic review that included 64 studies (7,104 participants) examined the effectiveness of acupuncture for the treatment of depression (10). The evidence indicated that acupuncture treatment may moderately or slightly lower the severity of depression compared with treatment as usual and control acupuncture (invasive, non-invasive sham controls), respectively, although the quality of the evidence was judged as being mostly low or very low. Other reviews have also described evidence in support of MA and EA as generally beneficial, safe and well-tolerated as monotherapy in MDD and post-stroke depression, but the evidence is insufficient to support the use of acupuncture in combination with antidepressants (31, 32). Two Chinese studies examined the effectiveness of acupuncture and moxibustion (a treatment involving burning of the dried Chinese herb mugwort or Artemesia vulgaris to apply heat onto or very close to an acupoint) in relieving psychological distress in 163 patients with depression and sought to determine whether gender-related differences exist in response to acupuncture and moxibustion (12, 13). The authors reported that acupuncture and moxibustion can significantly improve distress at even as late as 3 months after the completion of treatment, and that the level of efficacy is higher among females than males. Hence, acupuncture treatment in depression may improve depressive symptoms of depression and endure for some months.

Moreover, sleep status may improve in patients with depression after acupuncture treatment. A recent meta-analysis that included 18 randomized clinical trials (RCTs) involved 1,678 adults given acupuncture for depression-related insomnia and found significant improvements in sleep quality with acupuncture compared with Western medicine (11). When acupuncture was given as an adjunctive therapy with Western medicine, both depression and insomnia were improved (11). However, in another RCT involving 150 patients with residual insomnia associated with MDD, traditional acupuncture needling produced only mild treatment effects that were similar to those of minimal acupuncture and placebo acupuncture (33). There were no significant group-by-time interactions during the 5-week post-treatment period. Thus, the psychological effect of acupuncture might play an important role in the treatment of acupuncture in depression-related insomnia.

### Acupuncture for the Treatment of Anxiety

The effectiveness of acupuncture in anxiety has been widely investigated, using various sets of acupuncture points and treatment parameters. A narrative review published by the British Acupuncture Council reported that regular acupuncture and EA treatments improved anxiety symptoms (34). However, significant differences between the protocols used in regular acupuncture and EA made it hard to rule out a general beneficial or possible placebo effect. A more recent systematic literature review that included 32 English-language clinical and preclinical studies published between 2000 and 2010 also reported significant, positive results with acupuncture treatment for anxiety (14). Although the quality of these studies was variable, the authors suggested that patients who are resistant to conventional interventions (e.g., cognitive behavioral therapy) may prefer acupuncture treatment. Thus, acupuncture treatment may have positive effects on the symptoms of anxiety.

Acupuncture may also be of benefit in anxiety-related insomnia. A Canadian study reported that acupuncture significantly improved sleep quality in patients with anxiety and insomnia (15). At the end of acupuncture treatment, urine 6-sulfatoxymelatonin (a metabolite of melatonin) levels were normalized and several polysomnographic measures, as well as self-reported fatigue, sleepiness, anxiety, and level of depression, were significantly improved. Combined with the evidence of treatment efficacy in depression-related insomnia, acupuncture may have broader utility in neuropsychiatric disorders with impaired quality of sleep.

Interestingly, acupuncture treatment may also reduce anxiety levels in medical conditions other than neuropsychiatry. An early study reported that auricular acupuncture reduced state anxiety in patients waiting for surgery by a significantly greater extent than either body or sham acupuncture (35). Another study has reported that both body and auricular acupuncture effectively reduce preoperative anxiety; self-rated anxiety scores were significantly reduced from baseline in both groups (36). Possible underlying mechanisms remain unclear. One plausible explanation is that acupuncture regulates the autonomic nervous system. For instance, acupuncture appears to modulate heart rate variability, a non-invasive indicator of changes in autonomic state. In patients with mild depression or anxiety, verum acupuncture but not sham acupuncture was associated with significant reductions from baseline in mean resting heart rate at 5 and 15 min after needle application, with a trend toward an increase in high frequency (HF; 0.15–0.4 Hz) and a decrease in low frequency (LF; 0.04–0.15 Hz) spectral power (37). These results suggest that verum acupuncture modulates autonomic activity in response to alterations of internal and external environments, and thus reduces overall anxiety in patients with depression or anxiety.

However, it should be recognized that the interpretation of results from clinical trials investigating acupuncture interventions in depression and anxiety disorders is complicated by different interventions, different comparators used against acupuncture interventions, and the small sample sizes in many trials (38). These shortcomings prevent any accurate assessment of acupuncture for these conditions or a true comparison of the relative effectiveness of different treatment regimens. Moreover, the data are difficult to interpret from those studies where needling at specific and non-specific points have yielded similar outcomes (38).

The evidence consulted for depression and anxiety is summarized in **Table 1**.

## Acupuncture for the Treatment of Schizophrenia

Compared with depression and anxiety, relatively few studies have addressed the efficacy of acupuncture in schizophrenia. A meta-analysis of 13 RCTs including 954 patients, all from China, provides positive evidence for the effectiveness of acupuncture (with or without EA or moxibustion) in treating the symptoms of schizophrenia (19). Some of the RCTs reported that acupuncture plus drug therapy significantly improved auditory hallucinations, positive symptoms and response rates compared with antipsychotics alone or in combination with sham EA. In a recent systematic review of data from 26 studies (1,181 participants) reporting limited evidence for the use of adjunctive acupuncture therapy in the treatment of positive, negative and cognitive symptoms, the authors point out the importance of differences between quantitative and qualitative changes (16). The limited evidence for treatment efficacy may be partly due to the fact that positive symptoms still exist, but patients are suffering less. Other evidence suggests that individualized acupuncture is beneficial for patients with schizophrenia as an adjunctive treatment with routine care (18). After completing individualized acupuncture sessions, patients reported improvements in symptoms of schizophrenia, side effects of medication, energy, motivation, sleep, addictions and other associated physical problems. A case study has reported that positive and negative symptoms can be improved for up to 3 months after add-on acupuncture treatment (17). A more recent case study describes how add-on acupuncture treatment improved general psychopathology and negative symptoms but not positive symptoms (39). Thus, individualized addon acupuncture treatment may help to alleviate symptoms of schizophrenia.

Interestingly, the review by van den Noort et al. also describes beneficial effects with add-on acupuncture in accompanying sleep disorders (16). Several of the studies in that review reported improvements in subjective and objective sleep measurements. Acupuncture treatment appears to have similar effects to but is safer than zopiclone, a prescription medication for sleep disorder. Thus, acupuncture treatment may improve sleep dysregulation in schizophrenia.


### TABLE 1 | Overview of characteristics of included studies evaluating acupuncture for depression and anxiety.

*(Continued)*

### TABLE 1 | Continued


*(Continued)*

### TABLE 1 | Continued


TABLE 2 | Overview of characteristics of included studies evaluating acupuncture for schizophrenia.


As with the clinical evidence for acupuncture in depression and anxiety disorders, the clinical evidence is limited for the effectiveness of acupuncture as a treatment for schizophrenia. The meta-analysis performed by Lee and colleagues found that all of the included studies were limited by low methodological quality, the low overall number, and the small sample sizes (19). Moreover, as all of the studies were conducted in China, international trials are needed to investigate whether the effects can be replicated in other ethnicities.

The evidence consulted for schizophrenia is summarized in **Table 2**.

# Acupuncture for the Treatment of Alzheimer's Disease

Another disorder that commonly exhibits neuropsychiatric symptoms is dementia. Neuropsychiatric symptoms are amongst the earliest signs and symptoms of neurocognitive disorders and incipient cognitive decline, and can be challenging to treat (40). Evidence shows that acupuncture can increase a patient's verbal and motor skills and improve mood and cognitive function. In an early Chinese trial, 38 patients with senile dementia were treated with acupuncture and acupoint-injection with aceglutamide (23). After treatment, symptoms had improved in 16 patients. In a small pilot study, 1 month of acupuncture treatment improved cognitive function in 8 patients with mild-to-moderate AD (22). Mini-Mental State Examination subscores assessing verbal orientation and motor coordination, as well as overall scores, were significantly improved from baseline. In a US study that included 11 patients with dementia (10 with AD and 1 with vascular dementia), depression and anxiety scores improved significantly after acupuncture treatment (21). These early pilot studies indicated that acupuncture treatment may be of benefit for neuropsychiatric symptoms in dementia, especially in AD.

More recently, a larger clinical trial involving 87 patients with mild-to-moderate AD reported that acupuncture may significantly improve cognitive function (20). When used as monotherapy, acupuncture treatment was associated with significantly greater decreases from baseline in Alzheimer's disease Assessment Scale-Cognitive (ADAS-cog) scores compared with donepezil. The improvement in cognitive function was observed for up to 12 weeks after the end of acupuncture treatment. Moreover, no patients discontinued treatment because of adverse events in the acupuncture group, but four patients did so in the donepezil group. In addition, a small-scale functional magnetic resonance imaging study explored the relationships between de qi sensations (a special needling sensation evoked by acupuncture) induced by different needling depths of acupuncture and their differential effects on the reorganizations of whole-brain networks in 12 patients with mild cognitive impairment (41). The results show that as compared with superficial needling, acupuncture with deep needling induces stronger, wider-ranging de qi sensations and enhances nodal centrality, primarily in the abnormal regions of the brain implicated in mild cognitive impairment. Hence, acupuncture treatment in mild-to-moderate AD appears to be not only beneficial but also safe, with the capacity to improve dysfunctional neural mechanisms involved in mild cognitive impairment.

The evidence consulted for AD is summarized in **Table 3**.

### GLUTAMATE AND GLYCINE IN CNS DISORDERS

In the past decade, increasing evidence implicates the important role of glutamate in the pathophysiology of many CNS disorders (42, 43). L-Glutamate, the most common excitatory neurotransmitter in the CNS, is involved in synaptic plasticity and cognition (42, 44, 45). Glutamate receptors, synaptic receptors located primarily on the membranes of neuronal cells, are responsible for the glutamate-mediated postsynaptic excitation of neural cells and are important for neural communication, memory formation, learning, and regulation. Neural glutamate signaling is accommodated by two receptor families: ionotropic glutamate receptors (iGluRs; e.g., the Nmethyl-D-aspartate [NMDA] receptor and the α-amino-3 hydroxy-5-methyl-4-isoxazolepropionic acid [AMPA] receptor) and metabotropic glutamate receptors (mGluRs) (46). The iGluRs produce excitatory glutamate-evoked currents, whereas the mGluRs are G protein-coupled receptors that control cellular processes through the G protein signaling cascades. Evidence implicates iGluRs, and/or mGluRs, as potential drug targets in neuropsychiatric disorders, including depression, anxiety, schizophrenia, and AD (46, 47).

One of the iGluRs is the NMDA receptor, which is activated by glutamate as well as glycine as a co-agonist. Its activation allows cations to flow through the cell membrane. The NMDA receptor is very important for controlling synaptic plasticity and memory function (48). Recent studies indicate that the dysregulation of the NMDA receptor may play an important role in schizophrenia (49, 50) and depression (51). Conversely, altered AMPA receptor functioning is a feature of many CNS disorders, including amyotrophic lateral sclerosis (ALS), ischemia, traumatic brain injury, epilepsies, and AD (47). The AMPA receptor is responsible for mediating most of the fast synaptic transmission in the CNS (47). Modulation of AMPA receptor numbers explains much of the plasticity of excitatory transmission in the brain, whereby increasing or decreasing AMPA receptors alters synaptic strength, which may be linked to neuropsychiatric disorders such as schizophrenia and depression (52, 53).

The release of glutamate in the synaptic cleft at a particular concentration is maintained by either glutamine synthetase or excitatory amino acid proteins, i.e., excitatory amino acid transporters (EAATs), which re-uptake excessive glutamate from the synapse. A high density of EAATs near the synapse ensures quick removal or transportation of any unbound glutamate. Of the five different membrane-bound transporters, EAAT2 performs more than 90% of the clearance of extracellular glutamate into crude synaptosomes to prevent neuronal excitotoxicity and hyperexcitability. When this process is impaired, it can allow large amounts of glutamate to spill out from the synapse, which may be a pathophysiological mechanism in CNS disorders (54) including schizophrenia, cognitive deficits, dementia and AD, and other disorders. Evidence suggests that EAAT2 activation is a promising therapeutic approach in many neuropsychiatric disease models (45).

Throughout the CNS, glycine acts as a co-agonist with Lglutamate at NMDA receptors. Glycine fluxes are regulated by two specific glycine transporters: GlyT1 and GlyT2. Whereas, GlyT2 is expressed in the spinal cord, brainstem and cerebellum, GlyT1 is expressed in these regions as well as in the forebrain areas such as the cortex, hippocampus,

### TABLE 3 | Overview of characteristics of included studies evaluating acupuncture for Alzheimer's disease.


septum, and thalamus (55). GlyT2 is expressed by glycinergic nerve endings in rat spinal cord, while GlyT1 appears to be preferentially expressed by glial cells. Preclinical investigations suggest that GlyT2 is predominantly responsible for glycine uptake at glycinergic synapses, and that GlyT1 is involved in monitoring glycine concentration surrounding NMDA receptor-expressing synapses. In rats, GlyT1 inhibition potentiates NMDA receptor activity and affects NMDA receptor-dependent long-term potentiation (55). It is possible to modulate the function of the NMDA receptor by varying the availability of the glycine co-agonist. This has potential in the treatment of schizophrenia: the NMDA hypofunction hypothesis of schizophrenia postulates that increasing glutamatergic transmission via the NMDA receptors and inhibiting GlyT1 on glial cells enhances NMDA receptor neurotransmission by slowing the process of removal of glycine from the synapse and thus elevates synaptic glycine levels (55, 56).

### Glutamate, Depression, and Anxiety

The monoamine hypothesis of depression contends that the underlying pathophysiology is due to depleted levels of serotonin, norepinephrine, and/or dopamine in the CNS (57). However, no direct evidence supports a primary dysfunction of a specific monoamine system in patients with MDDs. Moreover, not only do many patients fail to respond to monoamine antidepressants, but residual symptoms, relapses and recurrences are common even with adequate dosing of these medications (58) and it can take up to several days or weeks for core depressive symptoms to begin to lift after monoamine antidepressants have elevated synaptic monoamine levels (57–59). MDD pathogenesis appears to involve something else beyond the monoamine system.

Glutamatergic modulation shows potential in antidepressant treatment. A single subanesthetic intravenous dose of the NMDA receptor antagonist ketamine acts rapidly in treatmentresistant depression within hours of administration, with effects that are typically sustained for 7–14 days (60). Moreover, the ketamine metabolite (2R,6R)-hydroxynorketamine [(2R,6R)- HNK] appears to have the antidepressant effects of ketamine and lacks the psychiatric, psychotomimetic, cardiovascular, neurological, and other side effects associated with acute dosing of ketamine in patients with depression (52). Unlike ketamine, (2R,6R)-HNK is not a NMDA receptor antagonist but is associated with the modulation of the AMPA receptor. Thus, both NMDA and AMPA receptors may be involved in the pathophysiological changes of depression and potentially represent new targets for the development of rapid-acting antidepressants.

The glutamatergic system also plays a major role in the pathogenesis of anxiety (61–63). Long-term administration of various antidepressant agents including selective serotonin reuptake inhibitors (SSRIs), serotonin and norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants (TCAs), and monoamine oxidase inhibitors (MAOIs) lowers glutamatergic activity in some regions, such as the hippocampus (61, 64), while the acute administration of NMDA receptor antagonists produces anti-anxiety and antidepressant effects in preclinical and clinical models (43, 62). Lamotrigine, used in epilepsy and bipolar depression, inhibits glutamate release and has proven efficacy in certain symptoms of posttraumatic stress disorder (PTSD), namely, re-experiencing and avoidance/numbing (65). Similarly, topiramate, which acts partly as an AMPA/kainate blocker, decreases re-experiencing symptoms in PTSD (66).

## Glutamate Receptor Dysregulation Contributes to Schizophrenia

Animal models of schizophrenia have implicated glutamate receptor dysregulation and other proteins relating to glutamate transmission, including EAAT2 (54). The glutamate hypothesis of schizophrenia is supported by the finding that the NMDA receptor antagonist phencyclidine blocks glutamate-activated postsynaptic currents and induces schizophrenia-like symptoms, including psychosis and cognitive impairment (50). Moreover, phencyclidine impairs prepulse inhibition (PPI) of the startle reflex, a simple form of information processing that is consistently reduced in schizophrenia. In rats treated with ceftriaxone and acute phencyclidine, PPI was more impaired compared with rats given either treatment alone (67). A broad-spectrum cephalosporin antibiotic, ceftriaxone stimulates EAAT2 expression and lessens neurotoxicity by inhibiting neuronal cell death associated with glutamate excitotoxicity (68). Upregulation of EAAT2 expression by ceftriaxone is thought to be via presynaptic activation of the mGluR, as mGluR2/3 agonist treatment prevented the PPI impairment associated with ceftriaxone-induced upregulation of EAAT2 in rats (69). EAAT2 overexpression causing PPI impairment may be due to glutamate spillover, which mGluR2/3 relies on for activation due to its perisynaptic localization (54). In patients with schizophrenia, oral clozapine 25–35 mg/kg/day for 3 weeks downregulated astrocytic EAAT2 levels and increased extracellular glutamate levels (70).

Interestingly, expression of the presynaptic protein synaptophysin, which is involved in neurotransmitter release, is significantly increased in the same anatomical areas where EAAT2 levels are downregulated by clozapine (71). This suggests that glutamate release may assist clozapine in potentiation of the excitatory synapse. More support for glutamate receptor dysregulation in schizophrenia is seen with the selective GlyT1 inhibitor sarcosine (N-methylglycine), which has shown promise in the treatment of cognitive impairment in patients with chronic schizophrenia (56, 72).

## The Dysregulated Glutamate System and AD

It is well-known that AD is associated with reductions in glutamate transporter capacity and protein expression, as well as a selective loss of the vesicular glutamate transporter (73– 75). Recent evidence further points to impairment of EAAT2 function in AD, with findings of significantly lower levels of EAAT2 gene expression in the cortex and decreased EAAT2 immunoreactivity in the motor cortex of patients with AD (76). Although EAAT2b expression did not vary significantly according to disease severity, significantly upregulated levels of exon-skipping variant mRNA expression have been found, which reduces wild-type EAAT2 protein expression in primary astrocytes and inhibits glutamate transport (76). The correlation of EAAT2 expression with increasing neurodegeneration, in combination with the ability of exon-skipping variants to reduce glutamate reuptake, suggests that increased glutamate levels may propagate excitotoxic processes implicated in AD pathogenesis.

Moreover, iGluRs may also contribute to the excitotoxic processes implicated in AD pathogenesis. It is well-established that the NMDA receptor plays an important role in excitotoxicity (77). Hyperactivity of the NMDA receptor may result in a flooding of cations (e.g., Ca2+) into the neuron leading to degeneration of the dendritic spines or even the death of the neuron. In patients with AD, treatment with the uncompetitive NMDA receptor antagonist memantine can benefit the cognitive symptoms of AD (78). Notably, in preclinical models of AD, AMPA-mediated transmission, and altered synapse morphology correlated with cognitive decline (47). An initial loss of dendritic spines and synapses is accompanied by a concomitant increase in presynaptic release probability, as the neuronal circuit attempts to compensate for synaptic dysfunction and loss (47). This evidence suggests that increased synaptic glutamate levels induced by a reduction in reuptake may trigger iGluR hyperactivity and lead to the excitotoxic processes relating to the cognitive symptoms of AD.

## ACUPUNCTURE MODULATES GLUTAMATE NEUROTRANSMISSION

To date, very little neuroscience research has explored the effects of acupuncture on the role of glutamate in neuropsychiatric disorders. While studies of acupuncture analgesia indicate that acupuncture stimulation may modulate levels of expression of glutamate expression and its receptor, as well as EAAT expression (79, 80), no existing studies have reported on acupunctureinduced modulation of the glycine transporter or other upstream regulatory mechanisms (e.g., D-amino acid oxidase and the amino acid transporter system).

Research has reported that both high- and low-frequency EA significantly decreases upregulated levels of NMDA receptor 1 and 2A and AMPA receptor 1 expression in the spinal cord in an inflammatory pain animal model (81). These findings are corroborated by other research reporting that EA (10 Hz) inhibited phosphorylation of NMDA receptor 1 in spinal cord and alleviated pain in a rat model of inflammatory pain (82). At the supraspinal level, alternating high- and low-frequency EA decreased levels of NMDA receptor 1 and c-fos expression in the rostral ventromedial medulla in an animal model of visceral pain (83). Thus, the expression of glutamate and NMDA receptors may be modulated by acupuncture stimulation in the CNS.

Acupuncture also modulates EAAT2 expression. A recent study examined the effect of acupuncture on depressive behaviors and EAAT2 in rats subjected to chronic unpredictable mild stress (44). Both acupuncture therapy and drug treatment with the glutamate reuptake enhancer riluzole significantly increased sucrose consumption in the sucrose preference test paradigm. This increase in sucrose consumption was associated with an elevated food intake and shortened latency in the novelty-suppressed feeding test paradigm. The amelioration of depressive behavioral actions was consistent with increasing numbers of EAAT2-positive cells and protein expression in the hippocampus and prefrontal cortex. EAAT2 mRNA expression was also increased in the prefrontal cortex, but there was no change in the hippocampus. Moreover, the antidepressant effect was observed later with acupuncture than with riluzole, indicating that repeated acupuncture stimulation may be needed to accumulate EAAT2 expression. Thus, acupuncture-mediated modulation of EAAT2 expression may ameliorate depression.

## CONCLUSIONS

In summary, evidence indicates that acupuncture treatment may be of benefit in several neuropsychiatric disorders, including depression, anxiety, schizophrenia, and AD. The pathophysiology of these disorders may be associated with glutamate dysregulation, marked by a high rate of glutamate release and elevated expression of glutamate receptors and glutamate transporters in the CNS. The ability of acupuncture stimulation to modulate glutamate receptor and EAAT expression suggests that the treatment effects of acupuncture are underpinned by its intervention in the dysregulated glutamate system. Further preclinical and clinical studies are needed to clarify the possible mechanisms of acupuncture in these neuropsychiatric disorders and to establish protocols for treatment guidelines.

# AUTHOR CONTRIBUTIONS

C-HT and IM: collecting and analyzing literature and writing the manuscript; Y-HC: designing and coordinating the study as well as writing the manuscript.

### ACKNOWLEDGMENTS

This work was supported by grants from the Ministry of Science and Technology, Taiwan (MOST 107-2320-B-039−031 -MY3, 106-2320-B-039−019, 107-2314-B-039-058-MY2, and 106-2314- B-039-012) and China Medical University (CMU 105-S-25). This work was also supported by the Chinese Medicine Research Center, China Medical University from the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project under the Ministry of Education (MOE) in Taiwan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

### REFERENCES


in persons with dementia: results of a feasibility and effectiveness study. Alzheimers Care Today (2001) 2:28–41. doi: 10.1016/j.ctim.2007.07.005


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

Copyright © 2019 Tu, MacDonald and Chen. 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.

# Targeting mGlu Receptors for Optimization of Antipsychotic Activity and Disease-Modifying Effect in Schizophrenia

Ferdinando Nicoletti 1,2 \*, Rosamaria Orlando<sup>1</sup> , Luisa Di Menna<sup>2</sup> , Milena Cannella<sup>2</sup> , Serena Notartomaso<sup>2</sup> , Giada Mascio<sup>2</sup> , Luisa Iacovelli <sup>1</sup> , Francesco Matrisciano<sup>3</sup> , Francesco Fazio<sup>2</sup> , Filippo Caraci 4,5, Agata Copani 4,6, Giuseppe Battaglia<sup>2</sup> and Valeria Bruno1,2

*<sup>1</sup> Department of Physiology and Pharmacology, Sapienza University of Rome, Rome, Italy, <sup>2</sup> IRCCS Neuromed, Pozzilli, Italy, <sup>3</sup> Department of Psychiatry, The Psychiatric Institute, College of Medicine, University of Illinois, Chicago, IL, United States, <sup>4</sup> Department of Drug Sciences, University of Catania, Catania, Italy, <sup>5</sup> Oasi Research Institute (IRCCS), Troina, Italy, <sup>6</sup> Institute of Biostructure and Bioimaging, National Research Council, Catania, Italy*

### Edited by:

*Hsien-Yuan Lane, China Medical University, Taiwan*

### Reviewed by:

*Joshua T. Kantrowitz, Columbia University, United States Alessandro Usiello, Second University of Naples, Italy*

\*Correspondence: *Ferdinando Nicoletti ferdinandonicoletti@hotmail.com*

### Specialty section:

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

Received: *13 November 2018* Accepted: *23 January 2019* Published: *14 February 2019*

### Citation:

*Nicoletti F, Orlando R, Di Menna L, Cannella M, Notartomaso S, Mascio G, Iacovelli L, Matrisciano F, Fazio F, Caraci F, Copani A, Battaglia G and Bruno V (2019) Targeting mGlu Receptors for Optimization of Antipsychotic Activity and Disease-Modifying Effect in Schizophrenia. Front. Psychiatry 10:49. doi: 10.3389/fpsyt.2019.00049* Metabotropic glutamate (mGlu) receptors are considered as candidate drug targets for the treatment of schizophrenia. These receptors form a family of eight subtypes (mGlu1 to −8), of which mGlu1 and −5 are coupled to Gq/11, and all other subtypes are coupled to Gi/o. Here, we discuss the possibility that selective ligands of individual mGlu receptor subtypes may be effective in controlling the core symptoms of schizophrenia, and, in some cases, may impact mechanisms underlying the progression of the disorder. Recent evidence indicates that activation of mGlu1 receptors inhibits dopamine release in the meso-striatal system. Hence, selective positive allosteric modulators (PAMs) of mGlu1 receptors hold promise for the treatment of positive symptoms of schizophrenia. mGlu5 receptors are widely expressed in the CNS and regulate the activity of cells that are involved in the pathophysiology of schizophrenia, such as cortical GABAergic interneurons and microglial cells. mGlu5 receptor PAMs are under development for the treatment of schizophrenia and cater the potential to act as disease modifiers by restraining neuroinflammation. mGlu2 receptors have attracted considerable interest because they negatively modulate 5-HT2A serotonin receptor signaling in the cerebral cortex. Both mGlu2 receptor PAMs and orthosteric mGlu2/3 receptor agonists display antipsychotic-like activity in animal models, and the latter drugs are inactive in mice lacking mGlu2 receptors. So far, mGlu3 receptors have been left apart as drug targets for schizophrenia. However, activation of mGlu3 receptors boosts mGlu5 receptor signaling, supports neuronal survival, and drives microglial cells toward an antiinflammatory phenotype. This strongly encourages research of mGlu3 receptors in schizophrenia. Finally, preclical studies suggest that mGlu4 receptors might be targeted by novel antipsychotic drugs, whereas studies of mGlu7 and mGlu8 receptors in animal models of psychosis are still at their infancy.

Keywords: metabotropic glutamate receptors, schizophrenia, positive allosteric modulator, development of cortical interneurons, receptor cross-talk

**98**

# BACKGROUND

In spite of the continuous development of "monoaminergic" antipsychotic agents, drug treatment of schizophrenia remains suboptimal. Current second-generation antipsychotic drugs include drugs with different affinity for dopamine and serotonin receptor subtypes, such as clozapine, olanzapine, quetiapine, asenapine, ziprasidone, lurasidone, risperidone, paliperidone, iloperidone, aripiprazole, cariprazine, and brexpiprazole. These drugs display good therapeutic efficacy against positive symptoms of schizophrenia and their use may slow the progression of brain atrophy (as opposed to first generation antipsychotics) (1). However, a significant proportion of patients affected by schizophrenia is refractory to medication, and a prophylactic use of these agents (i.e., their use prior to the first episode of psychosis) is limited by class-related adverse effects, such as sedation, weight gain, and anticholinergic effects. In addition, none of the first- or second-generation antipsychotics is effective in improving cognitive dysfunction associated with schizophrenia, with the possible exception of clozapine, which is considered as a second- or third-line drug because of serious safety concerns.

Thus, there is an urgent need for new safer antipsychotic agents acting at new targets that lie at the core of the pathophysiology of schizophrenia. A new Molecular Psychiatry article entitled "mGluR5 hypofunction is integral to glutamatergic dysregulation in schizophrenia" (2) is one point of arrival of years of extensive research linking metabotropic glutamate (mGlu) receptors to the pathophysiology and treatment of schizophrenia. This article shows that mGlu5 receptor signaling is blunted in the dorsolateral prefrontal cortex (DLPFC) of individuals affected by schizophrenia (2).

mGlu5 is one of the eight mGlu receptor subtypes that are traditionally subdivided into three groups on the basis of their amino acid sequence, pharmacological profile, and signal transduction mechanisms in heterologous expression systems. Group I includes mGlu1 and mGlu5 receptors, which are coupled to Gq/<sup>11</sup> proteins. Group II (mGlu2 and mGlu3) and group III (mGlu4, mGlu6, mGlu7, and mGlu8) mGlu receptors are all coupled to Gi/o proteins (3). This classification is universally accepted but incomplete from a functional standpoint. Native mGlu receptors activate multiple transduction pathways in a cell-and context-dependent fashion, and numerous subtypeselective biased ligands have been developed. For example, mGlu5 receptors are physically and functionally linked to NMDA receptors in most of the CNS synapses, but there are biasedpositive allosteric modulators (PAMs) of mGlu5 receptors that enhance mGlu5 receptor function without recruiting NMDA receptors (4, 5). There are also dogmas in the mGlu field that have been recently challenged. For example, mGlu3 receptors are generally considered as presynaptic receptors that function to restrain neurotransmitter release. In contrast, recent evidence indicates that mGlu3 receptors are also present in postsynaptic elements, where they boost mGlu5 receptor signaling (6). mGlu3 and mGlu5 receptors synergize to induce long-term depression (LTD) in the mouse PFC, and mGlu3-saturating LTD requires the activation of mGlu5 receptors (6). This receptor-receptor interaction may be relevant to the pathophysiology and treatment of schizophrenia (see below).

mGlu receptors were discovered in the mid eighties (7, 8), but no mGlu receptor ligands are currently available for clinical use in spite of more than 30 years of extensive research. In our opinion, schizophrenia and Parkinson's disease are the two disorders in which the mGlu research will first translate into the clinic. In section mGlu2 receptors and schizophrenia: to be or no to be? we will comment on a series of clinical studies in which pomaglumetad, the prodrug of the mGlu2/3 receptor agonist, LY404039, has been tested in patients affected by schizophrenia. Although pomaglumetad is no longer under clinical development, these studies attracted more and more interest on mGlu receptors in schizophrenia. It is generally believed that the activity of LY404039 in preclinical models and in subgroups of patients affected by schizophrenia is mediated by the activation of mGlu2 receptors. Here we will challenge the ≪ mGlu2-centric ≫ hypothesis of schizophrenia suggesting that mGlu3 receptors might be at least as valuable as mGlu2 receptors as condadate drug targets for schizophrenia. mGlu2 and mGlu3 receptors show important differences in the expression and function in the tripartite synapsis (pre- and postsynaptic elements, and astrocytes), and modulate the activity of microglial cells in an opposite fashion. We will also discuss the rationale behind the development of mGlu5 receptor PAMs and the attractive possibility that mGlu3 receptor agonists or PAMs might produce disease modifying effects in schizophrenia. Finally, we will comment on the attractive possibility that activation of mGlu1 receptors represents a valuable strategy in the treatment of positive symptoms of schizophrenia, and we will conclude with the discussion of a new set of evidence suggesting that mGlu4 receptors are also candidate drug targets for schizophrenia.

# METHODS

We searched for the following terms on Pubmed: metabotropic glutamate receptors and schizophrenia, metabotropic glutamate receptors and psychosis, metabotropic glutamate receptors and neurodeveloment, metabotropic glutamate receptors and CNS development, metabotropic glutamate receptors and network activity.

### TARGETING mGlu5 AND mGlu3 RECEPTORS TO CORRECT NEURODEVELOPMENTAL ALTERATIONS ASSOCIATED WITH SCHIZOPHRENIA

A large body of evidence suggests that schizophrenia is a neurodevelopmental disorder in which cortical interneurons become dysfunctional as a result of genetic alterations or environmental challenges occurring in the perinatal period. There are at least four processes that characterize the development of cortical GABAergic interneurons: (i) the biochemical specification into different cell types, of which those containing parvalbumin (PV) or somatostatin (SSt)

are the most numerous; (ii) the correct matching between interneurons and pyramidal neurons (9); (iii) the GABA shift from excitatory into inhibitory driven by the expression of the potassium-chloride symport, KCC2 in mature neurons (10); and (iv) the formation of perineuronal nets (PNNs), which surround PV<sup>+</sup> interneurons at the offset of the critical temporal windows of developmental plasticity (11, 12). The molecular events that drive these processes are only partially elucidated. Schizophrenia is associated with a dysfunction of PV<sup>+</sup> (basket and chandelier) and other populations of interneurons, with a resulting defect in network oscillations underlying cognitive functions (13) (**Figure 1**). In addition, several lines of evidence suggest that formation of PNNs is altered in the PFC of patients affected by schizophrenia (14–16). What may link mGlu3 and mGlu5 receptors to these processes is the evidence that the expression of both receptors is high in the early postnatal development, and then progressively declines to reach adult levels after weaning (17). One of the first observations associated with the discovery of mGlu receptors was that mGlu receptor-mediated polyphosphoinositide (PI) hydrolysis (a biochemical process leading to intracellular Ca2<sup>+</sup> mobilization and activation of protein kinase C) is dramatically high in the cerebral cortex and other brain regions during the early 7–9 days of postnatal life, and declines afterwards (8). This robust PI response is mediated by mGlu5 receptors, which are heavily expressed early after birth even in cells that lack mGlu5 receptors in the adult life, such as cerebellar Purkinje cells (18). An unexpected finding was that a large component of mGlu5-mediated PI hydrolysis in the developing PFC was lost in mGlu3−/<sup>−</sup> mice, and, therefore, required the endogenous activation of mGlu3 receptors (6). In contrast, mGlu5 receptor-mediated PI hydrolysis was intact in mGlu2−/<sup>−</sup> mice (6). Interestingly, expression of mGlu2 receptors in the cerebral cortex is low in the early postnatal life and increases afterwards (as opposed to expression of mGlu3 and mGlu5 receptors) (17).

Which (if any) of the neurodevelopmental processes taking place early after birth is/are driven by the combined activation of mGlu3 and mGlu5 receptors? This question is starting to be investigated. A reduction in the transcript and protein levels of PV, SSt, GAD-65, GAD-67, Reelin, and AMPA and NMDA receptor subunits has been reported in the PFC of adult mGlu5−/<sup>−</sup> mice (19). In mGlu3−/<sup>−</sup> mice we found changes in biochemical markers of cortical GABAergic interneurons during the first 4 weeks of postnatal development, which were not seen in

FIGURE 1 | Hypothetical model highlighting the potential role of individual mGlu receptor subtypes in the pathophysiology and treatment of schizophrenia. Presynaptic mGlu2 receptors negatively modulate 5-HT2A receptor signaling, and their activation restrains glutamate release at synapses between thalamo-cortical neurons and cortical pyramidal neurons (see chapter 6). mGlu5 receptors are found in both pyramidal neurons and interneurons (shown here) and functionally interact with NMDA receptors. mGlu5 receptor PAMs are under development for the treatment of schizophrenia (see chapter 4). Recent evidence indicates that mGlu3 receptor activation boost mGlu5 receptor signaling, and, therefore, selective mGlu3 receptor agonists or PAMs should be tested in preclinical models predictive of antipsychotic activity (chapter 3). The functional cross-talk between mGlu3 and mGlu5 receptors has been shown in pyramidal neurons (6). We speculate that the two receptors also interact in GABAergic interneurons, but this remains to be demonstrated. mGlu1 receptors and type-4 muscarinic cholinergic receptors (M4) interact in stimulating the production of endocannabinoids (ECBs) in striatal projection neurons. ECBs activate presynaptic CB1 receptors, thereby reducing dopamine (DA) release from meso-striatal terminals. Hence, mGlu1 receptor PAMs might be effective in improving positive symptoms in schizophrenia (chapter 5). Finally, mGlu4 receptor agonists/PAMs are effective in animal models that are predictive of antipsychotic activity. mGlu4 receptors might act to reduce glutamate release from excitatory nerve endings of pyramidal neurons and through other mechanisms. Not shown here: activation of mGlu3 and mGlu5 receptors may restrain neuroinflammation by driving microglia toward an antinflammatory phenotype, and activation of mGlu3 receptors stimulates the production of neurotrophic factors in astrocytes (see chapters 3 and 4).

mGlu2−/<sup>−</sup> mice (20). These findings suggest that the functional partnership between mGlu3 and mGlu5 receptors may be instrumental for the biochemical and functional specialization of cortical GABAergic interneurons. An attractive hypothesis is that the two receptors may also serve to regulate the numeric balance between pyramidal neurons and GABAergic interneurons in the developing cerebral cortex. An elegant article by Prof. Oscar Marin and his Associates (9) demonstrates that the balanced network between excitatory and inhibitory cortical neurons becomes established through mechanisms of activity-dependent cell death and survival, with the decrease in the total number of pyramidal neurons preceding the decrease in the number of interneurons in the first few days of postnatal life. Surviving interneurons are those that are innervated by pyramidal cells because synaptic excitation enhances the activity of the phosphatidylinositol-3-kinase (PI3K) pathway by suppressing the expression of the PI3K negative regulator, phosphatase and tensin homolog PTEN (9). Both mGlu3 and mGlu5 receptors may signal through the activation of the PI3K pathway (3), and we are currently examining whether genetic and pharmacological manipulation of the two receptors alters the expression of total and phosphorylated PTEN in the developing cerebral cortex. Another interesting area of investigation is whether mGlu5 and mGlu3 receptors regulate KCC2 expression and GABA responses in pyramidal cells during postnatal development. Changes in cortical and cerebellar KCC2 levels were found in mice lacking mGlu3 and mGlu5 receptors, respectively (20, 21).

How all this can be relevant to the pathophysiology and experimental treatment of schizophrenia? Cell-specific deletion of mGlu5 receptors in PV<sup>+</sup> neurons in mice results into a behavioral phenotype modeling schizophrenia, characterized by spontaneous hyperactivity, and defects in pre-pulse inhibition and long-term memory (22). A psychotic-like phenotype was also observed in mice lacking mGlu3 receptors (23, 24). Wang et al. (2) found that mGlu5 receptor signaling was largely reduced in the post-mortem DLPFC of individuals affected by schizophrenia, as shown by tyrosine hyperphosphorylation (leading to receptor desensitization), reductions in receptor coupling to Gq/11, Homer, and PI3K, and reduced receptor association with adaptor and scaffolding proteins, such as RGS4, norbin, Preso 1 and tamalin. Remarkably, the reduced signaling of mGlu5 receptors disrupted the physiological interplay between mGlu5 and NMDA receptors in the DLPFC of individuals affected by schizophrenia (2). Abnormalities in mGlu3 receptor dimerization have also been reported in the frontal cortex of individuals affected by schizophrenia (25). Polymorphic variants of GRM3 (the gene encoding for the mGlu3 receptor) have been consistently associated with schizophrenia (26–37), and genetic alterations of GRM5 have also been linked to schizophrenia (38).

We hypothesize that, in schizophrenia, a reduced expression and/or activity of mGlu3 and mGlu5 receptors may have a strong impact on the maturation of cortical GABAergic interneurons leading to a permanent alteration of network activity in the frontal cortex. If turned to be correct, this hypothesis raises the interesting possibility that a pharmacological intervention with mGlu3 or mGlu5 receptor PAMs early after birth may restore the correct developmental trajectory rescuing the pathological phenotype. This is a testable hypothesis in animal models, but translation to humans is an extremely difficult task for ethical and practical reasons. Regardless of any potential therapeutical application, defining the precise role of mGlu3 and mGlu5 receptors in neuronal development will give an answer to an intriguing question raised since the discovery of mGlu receptors: why mGlu receptor-mediated PI hydrolysis is so large in the early postnatal life (much larger than PI hydrolysis mediated by any other receptor coupled to Gq/11) and progressively declines afterwards?

Another potential link between mGlu5 receptor signaling and schizophrenia is suggested by the evidence that the endogenous D-amino acid, D-aspartate, has a great efficacy in stimulating PI hydrolysis in cortical or hippocampal slices prepared from 8-9 day-old rats, and its action is largely mediated by mGlu5 receptors (39). D-Aspartate levels are reduced in the dorsolateral prefrontal cortex of patients affected by schizophrenia as a result of an enhanced activity of the catabolic enzyme, D-aspartate oxidase (40, 41). In addition, the antipsychotic drug, olanzapine, enhances glutamate release in the prefrontal cortex by inhibiting D-aspartate catabolism (42). Thus, a defective D-aspartate/mGlu5 receptor axis might contribute to the pathophysiology of schizophrenia, and some antipsychotic drugs might reinforce mGlu5 receptor signaling by enhancing endogenous D-aspartate levels. Whether D-aspartate directly activates mGlu5 receptors, or receptor activation is secondary to the enhanced glutamate release induced by D-aspartate remains to be determined.

### mGlu5 RECEPTOR PAMS AS NOVEL ANTIPSYCHOTIC AGENTS

The leading hypothesis of NMDA receptor hypofunction in schizophrenia stems from the strong psychotomimetic effects of slow NMDA receptor channel blockers, such as phenyclidine (PCP) and ketamine, the cognitive dysfunction and psychosis caused by NMDA receptor autoantibodies in patients affected by autoimmune encephalitis and the large body of evidence obtained in mice lacking NMDA receptors or treated with NMDA receptor antagonists. NMDA receptors are found in most CNS neurons; however, they are highly expressed and functional in cortical interneurons (43). Several drugs that activate NMDA receptors have been developed for the treatment of schizophrenia, but with limited success in clinical trials. One notable example is bitopertin, a drug that inhibits the high affinity uptake of the NMDA receptor co-agonist, glycine. After promising preclinical and phase-2 clinical studies, bitopertin failed in phase-3 trials (44). Although clinical data with bitopertin were disappointing, active work on NMDA receptor agonists continues. A functional cross-talk exists between mGlu5 and NMDA receptors. Activation of mGlu5 receptors facilitates NMDA receptor function (45–50), whereas activation of NMDA receptors amplifies mGlu5 receptor activity by restraining receptor desensitization (51). This and other observations laid the groundwork for the development of mGlu5 reptor PAMs for the treatment of schizophrenia. Jeff Conn, Carrie Jones, Craig Lidsley and their Associates (Vanderbilt University) are actively involved in the design and development of highly selective and brain permeant mGlu5 receptor PAMs, and they are so kind to make their molecules available to the scientific community. PAMs differ from orthosteric agonists because they have no intrinsic efficacy but amplify the action of endogenous glutamate. Thus, PAMs are particularly advantageous from a therapeutical standpoint because their action is activitydependent. Systemic treatment with mGlu5 PAMs corrects the pathological phenotype in pharmacological and genetic models that have predictive validity for antipsychotic treatment, including models of NMDA receptor hypofunction (52–58) (**Figure 1**). Excitotoxicity mediated by the enhanced NMDA receptor activity is a potential limtation to the use of mGlu5 receptor PAMs in the treatment of schizophrenia. Accordingly, high doses of mGlu5 receptor PAMs may cause seizures and neurotoxicity in rodents (4, 59, 60). An elegant way to overcome these limitations is the use of biased mGlu5 receptor PAMs that amplify receptor function without recruiting NMDA receptors (4, 5). One of these molecules, compound VU0409551, showed antipsychotic-like activity without activating NMDA receptors (5).

Biased mGlu5 PAMs might be particularly helpful in the treatment of patients affected by schizophrenia who are resistant to conventional antipsychotic medication. Our expectation is that these drugs may improve both positive and negative symptoms and exert pro-cognitive effects by restoring the balance between excitation and inhibition in the prefrontal cortex of patients affected by schizophrenia (58). While an early treatment with mGlu5 PAMs might correct the developmental abnormalities associated with schizophrenia (see above), the impact of a late treatment on disease progression cannot be predicted. A large body of evidence suggests that neuroinflammation has a key role in the progressive degeneration of the gray and white matter associated with schizophrenia (61–63). Interestingly, mGlu5 receptors are present in microglial cells, and their activation drives microglia toward an anti-inflammatory phenotype (64–69).

A potential anti-inflammatory effect of mGlu5 PAMs may slow disease progression in the absence of a direct induction of excitotoxic neuronal death. Thus, biased PAMs that enhance mGlu5 receptor activation without potentiating NMDA receptor currents cater the potential to behave as disease modifyiers in schizophrenia. This interesting hypothesis warrants further investigation in animal models of psychosis associated with neuroinflammation.

## TARGETING mGlu1 RECEPTORS IN THE TREATMENT OF PSYCHOSIS

An elegant manuscript describes a novel form of receptorreceptor interaction involving M4 muscarinic acetylcholine receptors and mGlu1 receptors (70). This interaction might pave the way to novel therapeutic strategies in schizophrenia. M4 receptor activation inhibits striatal dopamine release by enhancing the production of endocannabinoids, which behave as retrograde messengers at dopaminergic nerve terminals. Interestingly, this mechanism requires the co-activation of mGlu1 receptors, which are coupled to Gq/11, and, therefore, have the fisique du role to enhance the production of endocannabinoids. M4 receptors, which are coupled to Gi/o, are also able to inhibit D1 receptor signaling, but this function is independent of mGlu1 receptors (70) (**Figure 1**). Inhibition of dopamine release mediated by mGlu1 receptors explains the antipsychotic-like activity of selective mGlu1 PAMs in rodents (70) and holds promise for the treatment of positive symptoms of schizophrenia. Mutations of GRM1 (the gene encoding for the mGlu1 receptor) that reduce mGlu1 receptor signaling have been associated with schizophrenia (71), and mice lacking mGlu1 receptors display a psychotic-like phenotype (72). This suggests that a defective mGlu1 receptor expression and/or activity might contribute to the striatal dopaminergic hyperactivity associated with schizophrenia, and that mGlu1 receptor PAMs may correct this defect. Of note, co-activation of M4 and mGlu1 receptors selectively inhibits dopamine release in the striatum without affecting dopaminergic transmission in other brain regions. This suggests that mGlu1 receptor PAMs may reduce dopamine release only where is needed, i.e., in the hyperactive meso-striatal system. This is advantageous with respect to most antisychotic agents, which block D2 receptors with no regional discrimination.

The interplay between mGlu1 and M4 receptors is one of the many examples of functional interactions between Gq/11-coupled group-I mGlu receptors and other Gi/<sup>o</sup> coupled receptors. In the cerebellum, GABA<sup>B</sup> receptors cooperate with mGlu1 receptors at the synapses between parallel fibers and Purkinje cells (73). Andrzej Pilc and his Associates (Krakow University, Polland) have described a synergistic effect between mGlu5 and GABA<sup>B</sup> receptor PAMs in behavioral tests with pharmacological validity for the treatment of positive, negative and cognitive symptoms of schizophrenia (74).

## mGlu2 RECEPTORS AND SCHIZOPHRENIA: TO BE OR NO TO BE?

The study of mGlu2 receptors as druggable targets in schizophrenia has been a leading theme of research in the whole mGlu receptor field in the last two decades. The evidence that compound LY354740 (a potent, selective, and brain permeant mGlu2/3 agonist) was able to reverse behavioral and neurochemical effects of PCP in rats (75) gave the impetus to the development of orthosteric mGlu2/3 receptor agonists or mGlu2 receptor PAMs as novel "non-monoaminergic" antipsychotic agents. mGlu2 receptors are presynaptic receptors that inhibit neurotransmitter release (3). It is believed that mGlu2 receptor activation produces antipsychotic effects by restraining the activity of 5-HT2A receptors in the frontal cortex. Gerard Marek and his Associates were first to demonstrate that mGlu2 receptors inhibit electrophysiological responses mediated by 5-HT2A receptors (76, 77). In the following years, Javier Gonzalez-Maeso and his Associates showed that mGlu2 receptors form a multimeric complex with 5-HT2A receptors and negatively modulate 5-HT2A receptor signaling in response to serotonin-like hallucinogenic drugs (78–81). Expression of mGlu2 receptors in the mouse frontal cortex is under the control of 5-HT2A receptors, and repeated administrations of second-generation antipsychotics, which block 5-HT2A receptors, down-regulates the expression of mGlu2 receptors through an epigenetic mechanism mediated by reduced histone acetylation at the GRM2 gene promoter (82, 83). Interestingly, mGlu2 receptors are down-regulated in the frontal cortex of mice subjected to prenatal stress (84, 85), which models the epigenetic and behavioral modifications associated with schizophrenia (84, 86). Not surprisingly, both mGlu2/3 agonists and mGlu2 PAMs display robust antipsychotic-like activity in a variety of behavioral tests (87–90).

The clinical development of the mGlu2/3 agonist, LY404039 (under the form of the oral prodrug, pomaglumetad methionil) is a remarkable example of how translational research in schizophrenia is conditioned by a number of unpredictable variables. Pomaglumetad methionil was as efficacious as olanzapine in improving positive and negative symptoms of schizophrenia in the first phase 2 clinical trial (91) (**Figure 1**). Treatment with pomaglumetad methionil did not cause the typical adverse effects of antipsychotic drugs, such as extrapyramidal symptoms, increase in body weight, and hyperprolactinemia (91). These findings generated great enthusiasm, demonstrating for the first time the efficacy of a non-monoaminergic agent in the treatment of schizophrenia. Unfortunately, the antipsychotic activity of pomaglumetad methionil was not confirmed in subsequent phase 2 or 3 clinical trials (92–94), and this led to discontinuation of the development program of LY404039 in schizophrenia with negative consequences for the whole mGlu receptor field. However, a metanalysis of all clinical studies showed that pomeglumetad methionil displayed therapeutic efficacy in subgroups of patients who were early-in-disease and had never been treated with 5-HT2A blocking agents (95). Research on the effect of LY404039 in humans is still ongoing, and both pomaglumetad methionil and LY2979165 (the oral prodrug of the selective mGlu2 receptor agonist, 2812223) were found to reduce ketamine-evoked blood oxygenation level dependent (BOLD) MRI signal in healthy subjects (96). It will be interesting to design clinical trials in which mGlu2/3 receptor agonists (or mGlu2 receptor PAMs) are tested in selected populations of patients affected by schizophrenia recruited on the basis of their genetic background, disease course, and history of previous medication. We are in favor of the development of mixed mGlu2/3 agonists or PAMs (rather than selective mGlu2 PAMs) if we consider the potential impact of all these drugs on disease progression. We have evidence that selective activation of mGlu2 receptors may produce neurotoxic effect (97–99) by enhancing the proinflammatory activity of microglia (100–102) or through other mechanisms. In contrast, activation of mGlu3 receptors is consistently neuroprotective through a mechanism of astrocyte-neuronal interaction mediated by the production of neurotrophic factors, such as transforming-growth factor-β1 or glial cell-derived neurotrophic factor (98, 103–105).

### mGlu4 RECEPTORS AS NOVEL TARGETS FOR THE TREATMENT OF SCHIZOPHRENIA

The study of individual group-III mGlu receptor subtypes in schizophrenia is now facilitated by the availability of brain permeant subtype-selective agents that are suitable for in vivo studies. One of these drugs is compound LSP4-2022, which behaves as a preferential orthosteric agonist of mGlu4 receptors (106). Systemic treatment with LSP4-2022 has been shown to induce antipsychotic-like activity in a number of behavioral tests and to attenuate neurotransmitter release induced by the NMDA receptor antagonist, MK-801 (107, 108). Interestingly, mGlu4 receptors co-operate with other neurotransmitter receptors coupled to Gi/<sup>o</sup> in improving psychotic-like symptoms in mice. Wozniak et al. (108) showed that the antipsychotic-like activity of LSP4-2022 in mice was prevented by pharmacological blockade of 5-HT1A receptors, whereas sub-threshold doses of LSP4-2022 and the 5-HT1A agonist, 8-hydroxy-dipropylaminotetraline (8- OH-DPAT), acted synergistically in producing antipsychotic- like effects. A similar synergism was shown between LSP4-2022 and drugs that activate GABA<sup>B</sup> receptors, although in this case the synergism could only be demonstrated in behavioral tests that model positive symptoms of schizophrenia (107).

A more recent study exended the analysis to the interaction between mGlu4 and M4 muscarinic receptors in an attempt to develop novel pharmacological strategies with good efficacy in improving positive, negative and cognitive symptoms, and good profile of safety and tolerability. A robust effect was seen by combining subactive doses of LSP4-2022 and the selective M4 muscarinic receptor PAM, VU152100, in behavioral tests that model positive, negative, and cognitive symptoms. This combination also reduced 5-HT2A-mediated spontaneous excitatory postsynaptic currents in frontal cortical slices, indicating that mGlu4 and M4 receptors cooperate in reducing glutamate release (109). Remarkably, the association of subactive doses of drugs that activate mGlu4 and M4 receptors did not cause motor impairment in mice, suggesting that the functional cross-talk between mGlu4 and M4 receptors can be targeted by new safer antipsychotic agents (**Figure 1**).

Finally, the study of mGlu4 receptors provides a new potential link beween the kynurenine pathway and schizophrenia. The kynurenine pathway of tryptophan metabolism generates a series of neuroactive componds, including the mGlu4 receptor agonist, cinnabarinic acid (110). We have evidence that cinnabarinic acid inhibits behavioral and biochemical responses to MK-801 at very low doses (111). Cinnabarinic acid is synthesized in very low amounts in the normal brain, but its production increases considerably under conditions of neuroinflammation. A protective activity of endogenous cinnabarinic acid against neuroinflammation (112) and psychotic symptoms (113) might be lost in schizophrenia because of a lower activity of kynurenine monoxygenase (114–116), the enzyme that converts kynurenine into 3-hydroxykynurenine giving rise to all metabolites that lie downstream of 3-hydroxykynurenine. This is certainly a field of great interest that warrants further investigation.

# CONCLUSIONS

Subtype-selective mGlu receptor ligands offer the opportunity of a precision medicine based pharmacological approach to target multiple domains of the psychopathological spectrum of schizophrenia. For example, patients with mutations of GRM1 or GRM3 genes might benefit from a treatment with mGlu1 and mGlu3 receptor PAMs, respectively, whereas patients who are early in disease or had never been treated with second-generation antipsychotics might respond to mGlu2/3 receptor agonists. The increasing knowledge of the role played by mGlu3 and mGlu5 receptors in the developmental trajectory of GABAergic interneurons might pave the way to an early use of mGlu3 or

## REFERENCES


mGlu5 receptor PAMs as disease modifying agents, being aware that treatment of schizophrenia in the preclinical phase awaits the discovery of predictive biomarkers endowed with high specificity and selectivity. Finally, some mGlu receptor ligands might have a positive impact on neuroinflammation by influencing microglial function or through other mechanisms. There are no reasons to be skeptical and continue to develop new mGlu receptor ligands for the treatment of schizophrenia with good hope of success.

# AUTHOR CONTRIBUTIONS

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


positive allosteric modulators–development challenges for a promising novel antipsychotic target. Neuropharmacology (2014) 82:161–73. doi: 10.1016/j.neuropharm.2012.12.003


reveals new possibilities for developing subtype selective ligands with therapeutic potential. FASEB J. (2012) 26:1682–93. doi: 10.1096/fj.11-1 95941


**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 Nicoletti, Orlando, Di Menna, Cannella, Notartomaso, Mascio, Iacovelli, Matrisciano, Fazio, Caraci, Copani, Battaglia and Bruno. 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.

# Autism Associated With Anti-NMDAR Encephalitis: Glutamate-Related Therapy

*Ruu-Fen Tzang1,2, Chuan-Hsin Chang3, Yue-Cune Chang4 and Hsien-Yuan Lane5,6,7\**

*1 Department of Psychiatry, Mackay Memorial Hospital, Taipei, Taiwan, 2 Mackay Medicine, Nursing and Management College, Taipei, Taiwan, 3 Agricultural Biotechnology Research Center, Academia Sinica, Taipei, Taiwan, 4 Department of Mathematics, Tamkang University, Taipei, Taiwan, 5 Department of Psychiatry, China Medical University Hospital, Taichung, Taiwan, 6 Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan, 7 Department of Psychology, College of Medical and Health Sciences, Asia University, Taichung, Taiwan*

The purpose of this review is to correlate autism with autoimmune dysfunction in the absence of an explanation for the etiology of autism spectrum disorder. The anti-N-methyl-Daspartate receptor (anti-NMDAR) autoantibody is a typical synaptic protein that can bind to synaptic NMDA glutamate receptors, leading to dysfunctional glutamate neurotransmission in the brain that manifests as psychiatric symptoms (psychosis, hallucinations, and personality changes). Detection of autoantibodies, cytokines, decreased lymphocytes, serum immunoglobulin level imbalance, T-cell mediated immune profile, maternal infection history, and children's infection history can all be vital biological markers of autoimmune autism. Diagnosing autoimmune encephalitis sooner can increase the effectiveness of curative treatments—such as immune therapy or immune modulatory therapy—that may prevent the long-term consequence of being misdiagnosed with autism spectrum disorder. Glutamate therapy primarily normalizes glutamate neurotransmission and can be a new add-on intervention alongside antipsychotics for treating autoimmune autism.

Keywords: autoimmune autism, glutamate-related therapy, anti-N-methyl-D-aspartate receptor encephalitis, immune therapy, dysfunctional glutamate neurotransmission

# INTRODUCTION

Autism spectrum disorder (ASD) is characterized by symptoms of impaired communication, stereotyped behavior, difficulty with social interaction, and certain repetitive or unusual behaviors. The current prevalence of ASD is estimated to be 1.5% or higher in developed countries (1); however, the etiology of autism remains largely unexplained. At the beginning of the 1980s, researchers began to apply the "immune hypothesis of schizophrenia" to explain immune-dysfunction– induced neuroinflammation as a cause for the symptoms of schizophrenia (2). Since autism has some overlapping features with schizophrenia, for example both have disturbed cognitive and social function, and there are neurobiological (brain volumes) and genetic (e.g., involvement of the same genes or chromosomal locations) domains both in autism and schizophrenia (3). Therefore, researchers also apply "neuroinflammation hypothesis of ASD" to regard autism as a disorder of autoimmune dysfunction (4, 5) or as autoimmune autism.

At least 69% of individuals with a diagnosis of ASD have been known to have neuroinflammation or encephalitis (6). Specifically, the so-called "anti-brain autoantibody" may damage fetal or children's brain cells, eventually leading to children falling into an autistic or regressive state.

*Edited by:*

*Ming D. Li, Zhejiang University, China*

### *Reviewed by:*

*Alessandro Usiello, Second University of Naples, Italy Ming D. Li, Zhejiang University, China Zhongli Yang, Zhejiang University, China*

> *\*Correspondence: Hsien-Yuan Lane hylane@gmail.com*

### *Specialty section:*

*This article was submitted to Molecular Psychiatry, a section of the journal Frontiers in Psychiatry*

*Received: 07 November 2018 Accepted: 03 June 2019 Published: 21 June 2019*

### *Citation:*

*Tzang R-F, Chang C-H, Chang Y-C and Lane H-Y (2019) Autism Associated With Anti-NMDAR Encephalitis: Glutamate-Related Therapy. Front. Psychiatry 10:440. doi: 10.3389/fpsyt.2019.00440*

**108**

Such brain-reactive antibodies causing autistic symptoms may elucidate the exploration of autism's etiology and suggest practical anti-inflammatory management protocols for children with ASD (7). Thus, this review introduces the potential pathogenicity of autism, which could explain why autoimmune dysfunction leads to autistic symptoms (indicated herein as "autoimmune autism").

Furthermore, we emphasize that psychiatrists should more quickly recognize autoimmune encephalitis—anti-N-methyl-Daspartate receptor (anti-NMDAR) encephalitis in children and adolescent patients and treat them as soon as possible, because timely immune therapy or glutamate therapy is beneficial for children with autism. In addition, through a literature review, we review four hypothetical pathways that correlate autism with dysfunctional autoimmunity.

## DYSFUNCTIONAL AUTOIMMUNITY-ORIGIN FROM MOTHER

On the basis of evidence that some of children with autism have mothers with rheumatoid arthritis, celiac disease, or a family history of type 1 diabetes (8), the symptoms of autism can be explained using the following exposure model. The fetus is exposed to prenatal antibodies from the dysfunctional maternal familial autoimmune system. The maternal pathogenic IgG antibodies or "induced maternal autoantibody-related ASD antibodies" cross the placenta and interact with the receptors in the fetal brain, eventually resulting in disruption of the brain cells (9–11).

An infection in the mother or child can trigger more brain damage in the fetus. If a mother with possible maternal familial autoimmune dysfunction develops an infection during pregnancy, this infection can trigger the autoimmune dysfunction to become severe, and maternal prenatal antibodies will then start attacking the brain of the fetus, leading to symptoms of autism (8); this is called "maternal-infection–triggered immune dysfunction." Children may develop the social, functional, and behavioral deficits observed in ASD owing to impaired neurodevelopment triggered by infection (5). Therefore, history of maternal immune dysfunctional disease or maternal infection is vital for diagnosing autoimmune autism.

### DYSFUNCTIONAL AUTOIMMUNITY-ORIGIN FROM CHILD

Laing et al. (1989) demonstrated that when rabbits were inoculated with influenza A virus, the induced antibodies cross-reacted with a brain-specific protein found in the hippocampus, cortex, and cerebellum of the rabbits and caused brain cell damage. Various viral infections or immunizations during a child's growth period such as encephalitis associated with potassium channel complex antibody, NMDAR encephalitis, and Hashimoto's encephalitis or vaccination induced anti-NMDAR encephalitis (12, 13)—all may cause several types of autoimmune encephalitis (5).

The anti-NMDAR autoantibody is a typical synaptic protein found in the nerve cells of teratomas. This autoantibody can bind with synaptic NMDA glutamate receptors in the brain, thereby causing dysfunctional glutamate neurotransmission and leading to psychiatric symptoms, such as psychosis, hallucinations, and personality changes (14). Recently, anti-NMDAR encephalitis has been frequently noticed after viral infection rather than originating from teratomas. The virally induced anti-NMDAR autoantibody may also attack the NMDA receptors of brain cells, causing hypofunction of glutamate neurotransmission in children with autism.

Such anti-NMDAR encephalitis is a type of reversible neuron dysfunction caused by autoimmune dysfunctional autoantibodies against NMDAR (15). Several cases of anti-NMDAR encephalitis have indeed been reported in patients with autism (16–18). Children with anti-NMDAR encephalitis symptoms that mimicked autistic regression recovered after timely immune therapy (19–21).

Approximately 30%–70% of autistic patients have circulating anti-brain autoantibodies (4). The autoantibodies in autistic patients—including anticardiolipin, β2-Glycoprotein 1, and antiphosphoserine antibodies (21); anti-double-stranded DNA antibodies (22); and antinucleosome-specific antibodies (23) all act as anti-brain autoantibodies and play roles in disrupting neurotransmission by attacking specific neurotransmitter receptors. More specific autoantibody detection tools should be developed clinically to help autistic patients receive an accurate diagnosis instead of being misdiagnosed with autism.

# AUTOIMMUNE AUTISM AFTER INTERFACE THEORIES

Not all children develop autoimmune dysfunction after a viral infection. Through the "Immune-Mediated Two-Hit Model for Psychosis" theory, it can be understood that only the brains of a specific group of children will be attacked. These children possibly have mothers with activated monocytes or microglia caused by maternal stress and inflammation paradigms during a prior pregnancy. When inflammation or some viral infection hits the memorized immune system or activating microglia of children, these children develop severe autoimmune abnormalities, and the second hit blows their neuronal circuitry (24). Here, prior maternal stress and inflammation induce a memorized immune system or activate microglia in children's brains, leading to the development of autistic symptoms in children during infections that occur in the early neurodevelopmental period.

## AUTOIMMUNE AUTISM—CYTOKINE LEVEL

Cytokines [including chemokines, interferons, interleukins (ILs), lymphokines, and tumor necrosis factor (TNF)] are small proteins used in cell signaling. Cytokine deregulation occurs after overactivation of the immune system because of infection, injury, or inflammation (4). Elevated serum levels of IL-6 and IL-17A (25), increased serum IL-6 level (26), and increased expression of the inflammatory molecules IL-1β, IL-6, IL-17, and TNF (27) have been observed in children with autism.

Proinflammatory cytokines promote inflammation and have also been found to be highly increased in the peripheral blood of patients with an ASD diagnosis in comparison with controls (28, 29). Furthermore, the proinflammatory cytokines TNFalpha, IL-6, granulocyte-macrophage colony-stimulating factor, Th1 cytokine (interferon-gamma), and chemokine (IL-8) are increased in the brains of individuals with ASD (30).

Most crucially, an increased level of one type of cytokine (IL-6) may alter the tryptophan/kynurenine pathway, which is closely related to glutamatergic neurotransmission (31). Assessing the aforementioned cytokine markers would allow researchers to diagnose autoimmune encephalitis accurately instead of misdiagnosing patients with autism. For children with autoimmune ASD symptom, the role of cytokines in ASD symptoms is that an autistic child may have autoimmune encephalitis induced glutamate hypofunction in his brain. Therefore, glutamate therapy could be used as the effective therapeutic strategy in restoring glutamate receptor's function (32) and cytokine alterations. However, no study is available on using cytokine as the biomarkers of immune function to assess the effect of glutamate therapy in either ASD or schizophrenia. That will be an interesting study in future.

### AUTOIMMUNE AUTISM—T-CELL-MEDIATED AUTOIMMUNITY

In addition to an imbalance of cytokine levels in the serum or cerebrospinal fluid, signs of T-cell activation are reliable indicators of autoimmune dysfunction in autism. T-cell-mediated autoimmunity participates in the functions of physiological defense, maintenance, and brain repair. Autoimmune T-cell deficiency or T-cell imbalance alters the activation profile of T-cells (33)—the innate proinflammatory response—with increased T-cell activation or a skewed display (34). Moreover, it significantly increases the T helper cell 1 and 2 (Th1/Th2) ratio (30). All these actions are associated with autoimmune autism.

In summary, in *ex vivo* studies, immune cells including abnormal or skewed T helper cells (Th1 and Th2), cytokine profiles, decreased lymphocytes, decreased T-cell mitogen response, and an imbalance of serum immunoglobulin levels (35) have all been linked with autoimmune encephalitis and autistic symptoms.

### AUTOIMMUNE AUTISM—NUCLEAR FACTOR KAPPA-LIGHT-CHAIN-ENHANCER OF ACTIVATED B-CELLS

Nuclear factor kappa-light-chain-enhancer of activated B-cells (NF-κB) is a protein found in almost all cell types. This protein mediates the regulation of cellular immune responses by promoting the expression of inflammatory cytokines and chemokines as well as by establishing a feedback mechanism that can produce chronic or excessive inflammation. Approximately 45% of a subgroup of children with autism have low natural killer (NK) cytotoxic cell activity (36). NF-κB is more aberrantly expressed in the orbitofrontal cortex of autistic patients than in controls. Specifically, the NF-κB of resident immune cells in brain regions are part of a molecular cascade indicating a more severe inflammation, which is associated with the behavioral and clinical symptoms of those with an ASD diagnosis (37).

## AUTOIMMUNE AUTISM—IMMUNE-RELATED GENETIC POLYMORPHISM

Autism has been associated with autoimmune dysfunction and with immune-based genes including human leukocyte antigen (HLA)-DRB1 and complement C4 alleles. Such genes show aberrant immune activity during vulnerable and critical periods of neurodevelopment, participating in the generation of the neurological dysfunction characteristic of ASD (35). Higher expression of T-cell activation markers (HLA-DR, CD26) was noticed during a study of immune phenotyping of peripheral blood mononuclear cells in young autistic children but not in controls (38).

Additionally, patients with autism were discovered to have a significantly higher frequency of *HLA-DRB1\*11* allele than controls (39). This joint analysis of genotype and DNA methylation broadly demonstrates the potential of both brain and bloodbased DNA methylation for insights into ASD and psychiatric phenotypes (40). The 16p11.2 mutations altered kynurenine pathway metabolism leading to abnormal glutamatergic activity in autism and may be the pathogenesis of ASD (31). Ghaleiha et al. suggested to use Memantine as an adjunctive treatment to restore NMDAR-dependent functionality before (41). Moreover, Memantine had a function targeting glutamate neurotransmission and already found to be the potential new and safe adjunctive treatment in children with ASD (42, 43).

### AUTOIMMUNE AUTISM—DIAGNOSIS

Social cognitive impairment in children with autism originates from dysfunction in dopaminergic, serotonergic, noradrenergic, and glutamatergic neurotransmission in the brain after dysfunctional autoimmunity. Consequently, patients who develop autoimmune autism early in life may be misdiagnosed if their anti-NMDAR encephalitis or potential autoimmune-related disease remains unrecognized (44). Autoimmune dysfunctional autism requires immune therapy; therefore, earlier detection is essential to prevent a misdiagnosis of autism. Detection of autoantibody, cytokines, decreased lymphocytes, imbalance of serum immunoglobulin levels, and T-cell-mediated immune profile in addition to maternal infection history or children's infection history can all be employed as biological markers of autoimmune autism.

# AUTOIMMUNE AUTISM—TREATMENT

Autoimmune dysfunctional autism requires immune therapy, which involves first-line immune therapy with pulse therapy in addition to intravenous immunoglobulin and plasmapheresis. Secondline immune therapy comprises rituximab or cyclophosphamide. In several cases, favorable treatment effects were reported after immune therapy if the child was noticed to have NMDAR-Ab in the serum and cerebrospinal fluid (16, 19–20, 21, 34, 45, 46). Recently, ASD drug development has focused on correcting synaptic dysfunctions; abnormalities in central oxytocin, vasopressin, and serotonin neurotransmissions, as well as neuroinflammation targets for new strategies to treat the core symptoms of ASD (47).

### TREATMENT: SECOND-LINE IMMUNE THERAPY

Up to half of all patients treated for anti-NMDAR encephalitis reported poor treatment response and the failure of first-line immunotherapy (46). Among these patients with inadequate treatment response, approximately 65% showed substantial improvement after well-tolerated second-line immunotherapy (46, 48). Second-line therapy, most commonly rituximab and/or cyclophosphamide, is often required in patients without tumors and those with a delayed diagnosis (49). Rituximab directed against the CD20 antigen on the surface of B-lymphocytes leads to decreased maturation of B-cells into antibody-secreting cells and is a favorable immunotherapy option in anti-NMDAR encephalitis (50). Second-line immunotherapy using rituximab has been reported to have improved the long-term outcome in a 16-yearold female patient with autoimmune NMDAR encephalitis and autistic traits diagnosed earlier in childhood (20, 21). Therefore, this study suggests using second-line immunotherapy as soon as possible to help patients with autoimmune autism who have had a poor response to treatment after first-line immune therapy of pulse therapy, plasmapheresis, or immunoglobulins.

### AUTOIMMUNE AUTISM—GLUTAMATE-RELATED THERAPY

For patients with autoimmune autism who have responded poorly to immune therapy, have difficulty starting immune therapy, or are presently using antipsychotic treatment, immune-modulating therapy is the treatment of choice. Brain cells need glutamate (glutamic acid) to regulate the central nervous system. Accumulated neuroimaging, family, genetic and animal studies have demonstrated that glutamate (glutamic acid) can improve mood disturbance and executive function. In anti-NMDAR encephalitis, dysfunctional postsynaptic glutamatergic transmission in synapses leads to enhanced release of γ-Aminobutyric acid and hypofunction of glutamate secretion, thus the NMDAR agonist, e.g., D-cycloserine, sarcosine, and GLYX-13 (rapastinel) can modulate glutamatergic transmission (51). D-Cycloserine (DCS), a partial glycine B agonist at the NMDA receptor site, has been shown to improve sociability in patient with autism earlier (52) by normalizing glutamate neurotransmission. In

### REFERENCES


recent years, D-cycloserine was also reported to be effective in improving stereotyped symptoms (53) and social reciprocity (54) in older adolescents and young adults with ASD.

Sodium benzoate, a D-amino acid oxidase inhibitor, was reported to be effective in children with communication difficulty with ASDs (55). Considering the challenge of earlier recognition of autoimmune autism, and the exorbitant cost of immune therapy in countries with poor medical knowledge of autoimmune autism, immune-modulating therapy or glutamate therapy can be a new add-on intervention alongside the current use of antipsychotics in treating autoimmune autism.

# CONCLUSION

Earlier diagnosis of autoimmune encephalitis increases the potential of curative treatments by enabling provision of timely immune therapy or immune modulatory therapy, which can prevent long-term consequences, such as being misdiagnosed with autism (16). An immunophenotyped patient with symptoms of autism may need to obtain a diagnosis of autoimmune encephalitis to avoid being misdiagnosed with ASD. Additionally, in children with the underlying syndrome presentation of childhood disintegrative disorder, early-onset schizophrenia, or late-onset autism (56); in children with first episode of psychosis (20); and especially when autistic symptoms follow a febrile illness (44), autoimmune autism including diagnosis of anti-NMDAR encephalitis or other autoimmune dysfunctional encephalitis should be considered as a possible organic cause.

### AUTHOR CONTRIBUTIONS

R-FT, C-HC, and H-YL produced the idea to this review. R-FT, C-HC, Y-CC, and H-YL did the literature review. R-FT and C-HC made the draft of this paper. H-YL critically revised this manuscript. All authors read and approved the final manuscript.

### ACKNOWLEDGEMENTS

Ministry of Science and Technology, Taiwan (MOST 107-2314- B-039-039); China Medical University, Taiwan (CMU107-BC-4); and China Medical University Hospital, Taiwan (DMR-108-093).


with anti-NMDA receptor encephalitis: an observational cohort study. *Lancet Neurol* (2013) 12(2):157–65. doi: 10.1016/S1474-4422(12)70310-1


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

digital media

of impactful research

article's readership