Bioenergetic abnormalities in schizophrenia

A dissertation submitted to the

Graduate School

of the University of Cincinnati

in partial fulfillment of the

requirements for the degree of

Doctor of Philosophy

in the Graduate Program in Neuroscience

of the College of Medicine

by

Courtney René Sullivan

B.S. University of Pittsburgh, 2013

Dissertation Committee:

Mark Baccei, Ph.D. (chair)

Robert McCullumsmith, M.D., Ph.D. (advisor)

Michael Lieberman, Ph.D.

Temugin Berta, Ph.D.

Robert McNamara, Ph.D.

ABSTRACT

Schizophrenia is a devastating illness that affects over 2 million people in the U.S. and displays a wide range of psychotic symptoms, as well as cognitive deficits and profound negative symptoms that are often treatment resistant. Cognition is intimately related to synaptic function, which relies on the ability of cells to obtain adequate amounts of energy.

Studies have shown that disrupting bioenergetic pathways affects working memory and other cognitive behaviors. Thus, investigating bioenergetic function in schizophrenia could provide important insights into treatments or prevention of cognitive disorders. There is accumulating evidence of bioenergetic dysfunction in chronic schizophrenia, including deficits in energy storage and usage in the brain. However, it is unknown if glycolytic pathways are disrupted in this illness. This dissertation employs a novel reverse translational approach to explore glycolytic pathways in schizophrenia, effectively combining human postmortem studies with bioinformatic analyses to identify possible treatment strategies, which we then examine in an animal model.

To begin, we characterized a major pathway supplying energy to neurons (the lactate shuttle) in the dorsolateral prefrontal cortex (DLPFC) in chronic schizophrenia. We found a significant decrease in the activity of two key glycolytic enzymes in schizophrenia

(hexokinase, HXK and phosphofructokinase, PFK), suggesting a decrease in the capacity to generate bioenergetic intermediates through glycolysis in this illness. Notably, we did not detect protein changes in enzymes or transporters in this pathway in the DLPFC. This suggests the bioenergetic interplay of astrocytes and neurons in schizophrenia is highly complex and may not be fully appreciated at the region-level. Thus, we utilized a cell-level approach (laser capture microdissection) to generate populations of cells enriched for astrocytes or neurons cut from layers 3 and 5 of the DLPFC of schizophrenia and control

ii subjects. We found significant mRNA changes in glycolytic enzymes (HXK1, PFKM, PFKL, and glucose-6-phosphate isomerase, GPI), lactate transporters (monocarboxylate transporter 1, MCT1), and glucose transporters (glucose transporter 1, GLUT1 and GLUT3) in pyramidal neurons in schizophrenia. Interestingly, we did not find any changes in astrocytes. This suggests a neuron-specific deficit in glycolytic pathways in the DLPFC in schizophrenia, which could contribute to pathophysiology of this illness.

To build on these findings, we performed bioinformatic analyses to examine the implications of an altered bioenergetic profile in schizophrenia. We first sought to replicate our findings in additional cohorts of schizophrenia and control subjects. We probed 2 independent transcriptomic datasets (Stanley Medical Research Institute Online Genomics

Database and Mount Sinai Microarray Dataset) for our metabolic targets. Supporting our hypothesis, we found several glycolytic enzymes and transporters to also be dysregulated in schizophrenia in these databases. Next, we utilized the Library of Integrated Network-

Based Cellular Signatures (LINCS) database to generate transcriptional signatures containing differentially expressed genes associated with bioenergetic abnormalities in schizophrenia. Using these signatures, we performed enrichment analyses with Enrichr to probe for connected pathways and biological significance and found hits for cell metabolism, proliferation, immunity, and inflammation pathways. Furthermore, we compared our disease signatures to a library of “drug activity transcriptional signatures” to identify possible perturbagens with the ability to “reverse” the disease signature and inform future preclinical experiments. Top perturbagens included peroxisome proliferator-activated receptor (PPAR) agonists, capable of bolstering metabolic pathways and possibly reversing cognitive deficits.

iii

To further elucidate the role of bioenergetics in cognitive dysfunction, we examined metabolic pathways in the GluN1 knockdown (KD) model of schizophrenia. This genetic model has 10% of normal functioning n-methyl-D-aspartate receptor (NMDA) receptors levels and exhibits several endophenotypes of schizophrenia including impaired social interaction, increased stereotypic behaviors, and decreased performance in spatial and working memory tasks. Using mass spectrometry and pathway analyses, we found abnormal metabolic pathways in GluN1 KD mice, as well as decreases in lactate and glucose transporter transcripts. With the goal of reversing these deficits, we selected a top perturbagen from our drug discovery bioinformatic analysis (PPAR agonists) with the hypothesis that this drug intervention may help restore schizophrenia endophenotypes in this model. Pioglitazone (pio) is a synthetic ligand for PPARγ, which can alter the transcription and expression of glucose transporters, leading to changes in glucose uptake.

We investigated the effects of (pio) treatment in the GluN1 KD model and found that pioglitazone treatment helped to restore explicit memory. This suggests bolstering metabolic pathways via pioglitazone may improve specific subtypes of cognition. This work has important implications for the treatment of cognitive illnesses with bioenergetic deficits such as schizophrenia.

iv

v

ACKNOWLEDGEMENTS

I want to extend great thanks to the many people who helped me reach success in this journey. You have all made me a better thinker, teammate, and person. First, I would like to thank my mentor, Dr. Robert McCullumsmith. Over the years, I realized just how lucky I was to work with you, and I am confident you were exactly what I needed. You pushed me to be the best scientist I could be, always supporting but never coddling me. I immediately noticed your great passion for research, as evidenced (you hate that phrase) by your Saturday phone calls to talk about awesome ideas that came to you about my project. You made me excited about science. I have appreciated the way you cared about my well-being, both personally and professionally. I always valued your enthusiasm and belief in me, and I feel honored to have been taken under your wing.

I want to thank all the current and former members of the McCullumsmith laboratory.

We have always prided ourselves on having an excellent lab milieu, and my career working with all of you is a testament to that. This work could never have been completed without you, and it has been a pleasure to work with you all. A special thanks to Sinead O’Donovan and Adam Funk, who have been there the duration of my time in the lab and never called any of my questions stupid or got annoyed with me for my numerous lab supply orders. You truly are some of the most gifted scientists I have met, and a joy to be around. I wouldn’t want to go to conferences or Christmas parties with any other postdocs. Also to Jennifer

McGuire, Erica Carey, Emily Devine, Rachael Koene, Jarek Meller, Eduard Bentea, and

Rebekka Meeks for their great help on these projects. The hard work you have done has not gone unnoticed.

I would also like to extend a special thanks to my friends up north in Toronto,

Canada. Dr. Amy Ramsey was kind enough to host me in her lab and work with her animals to complete the last chapter of my thesis. Our collaboration together has been wonderful,

vi and you are an inspiring female role model. Living in another country for a month is not easy. Thus, I would also like to send a special thanks to Catharine Mielnik, who made

Canada feel like a home. I learned more under your training than I could have ever hoped.

You truly have a talent for what you do and teaching others, and my gratitude for our time together is great. Not only did I become a more well-rounded scientist, I gained a lifelong friend.

I would next like to thank the Neuroscience Graduate Program, who proved repeatedly why I came to Cincinnati. Never have I seen faculty care so much about each individual student, and for this I never felt lost or alone. I would especially like to thank my dissertation committee: Dr. Mark Baccei, Dr. Robert McNamara, Dr. Michael Lieberman, and Dr. Temugin Berta, who guided me throughout this journey. Their feedback and thoughtful comments helped shape my ideas into excellent projects. Coming to UC was one of my best decisions thanks to the outstanding science and outstanding people.

Of course, words cannot describe how amazing my family has been. From high school, to college, to graduate school, to say you were my biggest cheerleaders is an understatement. It is probably accurate to say you thought my work was cooler than I did, and your excitement for my future was never lost on me. You always made me feel special, and the pride you had in me motivated me every day. To my parents, no one has ever believed in me like you. To my mom, I am the luckiest daughter in the world to have someone who loves me enough to actually want to talk to me on the phone every few days.

You have coached me through every part of life, and made me into the woman I am today.

To my dad, who taught me that hard work pays off, you instilled in me the work ethic and drive that is essential to survive graduate school. You did everything for me along this journey, from rent money in college to moving me from 3rd floor apartments multiple times- I

vii hope to attain your level of selflessness one day. You retain the right to go car shopping with me when I get a big girl job, a day you’ve always dreamed about.

A huge thanks to my Grandma, I had no idea being away from you would be so hard. You made me feel close to you with every card and every bite of the (literally) thousands of homemade meals you made me to be eaten miles away. To my Grandpa, whom I have always shared a special bond with, thank you for being there for me whenever

I needed you. Discussing sports, science, and life with you over the years (usually accompanied with Mexican food) has been a highlight in my life. You are truly one of the most intelligent people I know, even though 10 years and several books later, you and I cannot decide what consciousness might be. To Nannie, thank you for always making time for me when I am home, visiting with me and making family time together so special. And to my brothers, aunts and uncles, and cousins, thank you for reminding me constantly what a blessing it is to have a wonderful family. Of course, I must also thank the amazing friends I have made along the way. I hold my friends dearly and am so proud of what each of you have accomplished- I only hope you realize your own importance in my success.

Finally, I would like to thank Josiah Combs, who has been there throughout my graduate school career. Having you in my corner has empowered me to achieve more than

I thought possible. Thank you for sticking by my side even when I was stressed or tired and no fun to be around. I have never felt more loved and supported, be it through hand delivered coffees or morning work drop offs so I don’t walk in the cold. The amount of laughter, memories, and adventures you have brought me is immeasurable, and I am eternally grateful for having you in my life.

viii

ix

TABLE OF CONTENTS

General abstract...... ii

Acknowledgements...... vi

List of figures and tables...... xii

CHAPTER 1: Introduction...... 1

Genetic risk for schizophrenia...... 2

Bioenergetic coupling and energy supply at the synapse...... 3

Glucose versus lactate as primary energy source...... 4

Glucose/lactate utilization in normal cognition...... 7

Evidence for abnormal bioenergetic function in schizophrenia from transcriptomic

and proteomic studies...... 7

In vivo evidence for metabolic dysfunction...... 9

Cell-specific metabolic changes in schizophrenia...... 9

Do certain cell-subtypes exhibit greater bioenergetic susceptibility in

schizophrenia...... 11

Bioenergetic profile of GABAergic interneurons...... 12

The metabolic role of astrocytes in glutamatergic function...... 13

Astrocyte dysfunction in schizophrenia ...... 14

Oxidative stress and chronic inflammation in schizophrenia...... 14

What drives metabolic abnormalities in schizophrenia?...... 15

Genomic variation of synaptic and metabolic systems confers risk for

schizophrenia...... 15

Does synaptic dysfunction during brain development yield an intermediate metabolic

phenotype?...... 16

x

Changes in metabolic systems in schizophrenia due to antipsychotic drug

treatment...... 16

Metabolic abnormalities in drug-naïve patients...... 17

Animal models as a valuable tool...... 19

Models informing antipsychotic treatment on metabolic systems...... 20

Evidence for an intermediate metabolic phenotype in genetic models of synaptic

dysfunction...... 21

Metabolic abnormalities in pharmacological models of NMDA receptor

dysfunction...... 22

Possible therapeutic strategies targeting metabolic systems...... 23

Dissertation objectives...... 23

Rationale for studying the DLPFC...... 24

CHAPTER 2: Decreased Chloride Channel Expression in the Dorsolateral Prefrontal

Cortex in Schizophrenia...... 33

Abstract...... 34

Introduction...... 35

Materials and methods...... 36

Results...... 41

Discussion...... 42

Summary...... 46

CHAPTER 3: Neuron-specific deficits of bioenergetic processes in the dorsolateral prefrontal cortex in schizophrenia...... 50

xi

Abstract...... 51

Introduction...... 51

Materials and methods...... 53

Results...... 62

Discussion...... 66

Summary...... 73

CHAPTER 4: Bioinformatic analysis of bioenergetic changes in schizophrenia...... 96

Abstract...... 97

Introduction...... 97

Materials and methods...... 99

Results...... 103

Discussion...... 105

Summary...... 118

CHAPER 5: Bioenergetic deficits and reversal of memory deficits with pioglitazone in the GluN1 model of schizophrenia...... 152

Abstract...... 153

Introduction...... 153

Materials and methods...... 155

Results...... 164

Discussion...... 165

Summary...... 170

CHAPTER 6: Dissertation summary...... 183

xii

Summary of dissertation research...... 184

Future directions...... 191

References...... 195

xiii

LIST OF FIGURES AND TABLES

CHAPTER 1

Table 1.1. Characteristics of selected animal models of schizophrenia with “synaptic

dysfunction”...... 27

Table 1.2. Summary of metabolic abnormalities in schizophrenia and models of

schizophrenia...... 28

Figure 1.1. Bioenergetic coupling in normal brain...... 29

Figure 1.2. Synaptic and bioenergetic function in schizophrenia...... 31

Figure 1.3. Description of bioenergetic terminology...... 32

CHAPTER 2

Table 2.1. Subject’s characteristics...... 47

Figure 2.1. KCC2 Antibody Specificity in Rat...... 48

Figure 2.2. NKCC1 and KCC2 Expression in DLPFC...... 49

CHAPTER 3

Table 3.1. Summary subject demographics...... 75

Table 3.2. Extended subjects demographics...... 76

Table 3.3. Table of qPCR primers...... 78

Table 3.4. Summary of in silico analyses of schizophrenia patients on/off

antipsychotics...... 79

Table 3.5. In silico analyses of suicide as COD (Y/N), EtOH (Y/N), drug abuse (Y/N),

and sex...... 80

Figure 3.0. Summary of targets in glycolytic pathway...... 81

xiv

Figure 3.1. Enzyme activity curves...... 83

Figure 3.2. Region-level mRNA expression...... 84

Figure 3.3. Region-level protein expression...... 85

Figure 3.4. Enzyme activity...... 86

Figure 3.5. Specific activity...... 87

Figure 3.6. Major depressive disorder enzyme activity...... 88

Figure 3.7. Substrate concentrations...... 89

Figure 3.8. Pyramidal neuron mRNA expression...... 90

Figure 3.9. Astrocyte mRNA expression...... 91

Figure 3.10. Secondary analysis of antipsychotic medication...... 92

Figure 3.11. Secondary analysis of EtOH use...... 93

Figure 3.12. Secondary analysis of smoking...... 94

Figure 3.13. Secondary analysis of laterality...... 95

CHAPTER 4

Table 4.1. Seed profile for bioinformatic analyses...... 120

Table 4.2. Summary of in silico analyses...... 121

Table 4.3. Consistently upregulated and consistently downregulated genes across

clustered seed gene knockdown signatures...... 123

Table 4.4. Summary of Enrichr analyses...... 125

Table 4.5. Top 20 discordant chemical perturbagens per seed gene...... 125

Table 4.6. Top 20 discordant chemical perturbagens across all seed gene

knockdown signatures...... 127

Table 4.7. Top 12 unique chemical perturbagens...... 128

Figure 4.1. Summary figure of workflow overview...... 129

xv

Figure 4.2. Detailed summary figure of workflow...... 130

Figure 4.3. Heatmap of seed gene knockdown signatures...... 132

Figure 4.4. Clustered heatmap of seed gene knockdown signatures...... 134

Figure 4.5. Selection of upregulated genes...... 136

Figure 4.6. Selection of downregulated genes...... 137

Figure 4.7. KEGG cell signaling pathways for consistently upregulated genes...... 138

Figure 4.8. Cellular components for consistently upregulated genes...... 139

Figure 4.9. Molecular function for consistently upregulated genes...... 140

Figure 4.10. Protein-protein interaction hubs for consistently upregulated genes..141

Figure 4.11. Kinase analysis for consistently upregulated genes...... 142

Figure 4.12. Transcription factor analysis for consistently upregulated genes...... 143

Figure 4.13. dbGaP for consistently upregulated genes...... 144

Figure 4.14. KEGG cell signaling pathways for consistently downregulated

genes...... 145

Figure 4.15. Cellular components for consistently downregulated genes...... 146

Figure 4.16. Molecular function for consistently downregulated genes...... 147

Figure 4.17. Protein-protein interaction hubs for consistently downregulated

genes...... 148

Figure 4.18. Kinase analysis for consistently downregulated genes...... 149

Figure 4.19. Transcription factor analysis for consistently downregulated genes..150

Figure 4.20. dbGaP for consistently downregulated genes...... 151

CHAPTER 5

Table 5.1. Top pathways from LCMS analysis of affinity purified PSD95 protein

interactome...... 172

xvi

Table 5.2. Full demographics for animal cohort...... 173

Table 5.3. Behavioral assay timeline...... 175

Figure 5.1. Metabolic transcripts in GluN1 KD mice...... 176

Figure 5.2. Animal weight and food consumption...... 177

Figure 5.3. Puzzle box assay...... 178

Figure 5.4. Locomotor activity and stereotypy...... 179

Figure 5.5. Elevated plus maze...... 180

Figure 5.6. Social paradigm...... 181

Figure 5.7. Prepulse inhibition and acoustic startle response...... 182

xvii

CHAPTER 1

INTRODUCTION

1

Schizophrenia is a devastating illness that affects over 2 million people in the U.S. and displays a wide range of psychotic symptoms, as well as cognitive deficits and profound negative symptoms that are often treatment resistant (1-6). This illness is highly heritable, suggesting a major role for genetic variants in its complex pathophysiology. Additionally, environmental factors largely influence the development of schizophrenia, suggesting there is genetic risk combined with adverse life events that do not fully manifest until adulthood (7,

8).

Genetic risk for schizophrenia

The genetic contribution to schizophrenia susceptibility is strong. Schizophrenia affects 1% of the general population, while the risk of schizophrenia patients’ relatives is around 10% (9). Large genome-wide association studies have reported over 100 genetic loci containing common alleles conveying minor schizophrenia associations (10), while rare de novo copy number variations (CNVs), which often span multiple genes, confer higher effects on risk (11, 12). There is accumulating evidence of bioenergetic dysfunction in chronic schizophrenia, including deficits in energy storage and usage in the brain. While it is possible that genetic variation in metabolic genes contributes to these energetic deficits, genetic risk for schizophrenia is conferred by a large number of alleles, with risk variants each typically conferring a small portion of overall risk (13). Interestingly, these studies demonstrate a convergence of de novo mutations and altered gene expression on sets of functionally related proteins, pointing to the regulation of plasticity at excitatory synapses as a pathogenic mechanism in schizophrenia (12). Taken together, the bioenergetic deficits and genetic risk for synaptic dysfunction in schizophrenia lead to the following question: how do defects in bioenergetic function develop and contribute to the pathophysiology of this illness?

2

Bioenergetic coupling and energy supply at the synapse

Bioenergetic coupling in the brain requires the coordination of multiple systems and cell types to deliver energetic substrates in a spatio-temporal manner. There are multiple mechanisms in the brain to meet neuronal energy demands, including glycolysis, oxidative phosphorylation, and lactate uptake. Additionally, glutamate released at the synapse signals increased energetic demand to astrocytes and enhances production of bioenergetic substrates via increased glucose uptake, glycolytic rate, and lactate generation (14-16). In order to shape plasticity, glutamate levels in the synapse are normally tightly controlled by astrocytes, which remove extracellular glutamate via excitatory amino acid transporters

(EAATs)(17). These transporters rely on the electrochemical gradient maintained by the adenosine triphosphate (ATP) dependent Na+/K+ ATP pump. Thus, the clearance of synaptic glutamate is bioenergetically costly as well.

While the role of glutamate clearance in bioenergetic homeostasis is generally well understood, the principal mechanism fulfilling the energy requirements of neurons has been debated. Two bioenergetic ideologies offer viable energy production pathways under normal and pathological conditions. An early hypothesis stated that the main mechanism of energy production for neurotransmission was systemically derived glucose taken up by neurons and metabolized by oxidative phosphorylation (18). Conversely, a more recent and well- supported hypothesis suggests that astrocytes produce lactate in aerobic conditions (called the Warburg effect), with lactate shuttling from astrocytes to meet the bioenergetic needs of neurons. Pellerin and Magistretti have termed the net flow of lactate from astrocytes to neurons the “astrocyte-neuron lactate shuttle” (19) (Figure 1), which may help fuel neuronal oxidative phosphorylation (20, 21). This hypothesis posits that neuronal activation increases the concentration of glutamate in the synapse, activates glycolysis in glycogen rich glial

3 cells even in the presence of normal oxygen levels, and generates lactate which is transported out of astrocytes and into neurons via monocarboxylate transporters

(MCTs)(18, 22, 23). For example, lactate generated by glycolysis in glial cells constitutively supports synaptic transmission even under conditions in which a sufficient supply of glucose and intracellular ATP are present (20). Lactate production in astrocytes and the lactate shuttle are now thought to be the main mechanisms supporting bioenergetic coupling between astrocytes and neurons (24-26).

Glucose versus lactate as primary energy source

Initially, several lines of evidence suggested a role for glucose uptake into neurons as the main bioenergetic substrate. Glucose metabolism in neurons is an important modulator of memory, demonstrated in numerous studies and animal models (27-30)(reviewed in

(31)). Glucose is normally readily available to neurons, which have abundant levels of glucose transporter 3 (GLUT3) and glycolytic enzymes such as hexokinase (HXK) and phosphofructokinase (PFK)(32). Around 90% of the brain’s energy is generated by oxidative phosphorylation during resting conditions, which ATP consuming neurons are well suited for

(33). In line with these observations, a series of experiments demonstrated that lactate produced by astrocytes is not required for the short-term increase in oxidative phosphorylation that occurs after an increase in neuronal activity (34). This was demonstrated by using oxamate to block lactate dehydrogenase (LDH), a key enzyme in utilizing astrocyte derived lactate in neurons, which had no effect on the initial decrease in oxygen consumption (34). However, a more elegant set of experiments examining electrical stimulation in neurons found that inhibiting LDH in astrocytes via oxamate caused neighboring neurons to become hyperpolarized, which could be reversed by the addition of lactate to the medium and uptake into neighboring cells (35). These results show that

4 pyramidal cells are electrically regulated by changes in lactate release and the astrocyte- neuron lactate shuttle.

Pellerin and Magistretti proposed that during increased neuronal activation, glycolysis and lactate production in astrocytes predominates that of neurons (“astrocyte-neuron lactate shuttle”)(14, 19, 36). Astrocytes are positioned to readily take up glucose from the bloodstream and distribute it to neighboring neurons (23). They also contain glycogen, the brain’s only energy store, which is necessary for sustained neuronal firing as glucose and oxygen are depleted (37, 38). Other support for this hypothesis includes activity dependent activation of glycolysis in astrocytes and the close association of glycolytic enzymes to energy dependent ion pumps (14, 36, 39). A sophisticated set of experiments has also demonstrated that when energetic demands rise, neurons are unable to upregulate glycolysis to meet bioenergetic needs in favor of preserving normal levels of glucose shunting into the pentose phosphate pathway, thus preventing oxidative stress and cell death (40-42). This is due to the constitutive degradation of 6‑phosphofructo‑2- kinase/fructose‑2, 6‑bisphosphatase‑3 (PFKFB3) in neurons, an enzyme responsible for generating a potent activator of PFK activity (fructose‑2,6-bisphosphate)(42). This results in a much lower rate of glycolysis in neurons versus astrocytes. PFKFB3 is present in astrocytes and can respond to the dynamic energetic needs of synapses. In summary, neurons utilize glucose to maintain their antioxidant status during stress, meeting their bioenergetic requirements through closely coupled astrocytes (and the lactate shuttle)(40).

While glucose taken into neurons and low glycolytic flux may be important for transient synaptic transmission, another series of experiments demonstrated that the lactate shuttle is necessary for long-term synaptic activity. Long-term potentiation was impaired following

5 perturbation of lactate generation via glycogenolysis in astrocytes. This effect was rescued by the addition of exogenous lactate in normal conditions, but not when the expression of the primary neuronal lactate transporter monocarboxylate transporter 2 (MCT2) was disrupted (43). Inhibition of LDH with malonate or oxamate (inhibitors of succinate dehydrogenase and LDH) also decreased neuronal function when hippocampal slices were maintained in glucose or lactate, while neuronal function could be preserved with the addition of pyruvate, supporting the idea of a mitochondrial lactate oxidative complex

(Figure 1.3, Box 1) and a lactate shuttle (44, 45). However, effects of malonate and oxamate on neuronal function were only seen after a minimum of 30–45 minutes or after excitotoxic challenge. Using fMRI and NMR imaging, a different study demonstrated that the energy demands for acute, transient increases in neuronal activity in vivo are small and that heavy increases in neuronal activity cannot be accompanied by large increases in oxygen consumption alone (12-17%)(46). The lactate shuttle hypothesis states that as neuronal activation continues, oxidative metabolism is expected to increase due to the uptake of lactate produced by astrocytes and transport into the tricarboxylic acid cycle for neuronal fuel. This coincides with their findings that prolonged neuronal stimulation induces increasing levels of the cerebral metabolic rate of oxygen consumption (46).

Due to these variable results, it is not surprising that glucose versus lactate as the primary energy substrate of the brain has been debated. It is conceivable that both of these mechanisms play vital roles in neurotransmission, with proportional contributions determined by factors such as brain region, developmental stage, cell-type, synapse environment, and bioenergetic needs. These factors may also contribute to variability in which metabolic systems are vulnerable or compromised in different disease states.

6

However, the current thinking in the field is that the lactate shuttle is a primary source of energy for neurons (reviewed in (24-26, 47, 48)).

Glucose/lactate utilization in normal cognition

The importance of glucose/lactate utilization in cognitive function is more resolved.

The coupling mechanism between neuronal activity and astrocyte lactate production is essential for working memory performance and long-term memory formation in rodents, which is impaired following disruption of the MCTs and bioenergetic coupling (43, 49).

“Breaking” the lactate shuttle disrupts synaptic transmission, resulting in cognitive impairment (47, 50). Patients with schizophrenia experience a wide range of psychotic symptoms, as well as profound negative symptoms and cognitive deficits (1-6). Since bioenergetic coupling and neurotransmission are tightly coupled to cognitive function, these pathways could be important pathophysiological substrates in schizophrenia.

Evidence for abnormal bioenergetic function in schizophrenia from transcriptomic and proteomic studies

Schizophrenia pathology features a number of abnormalities associated with glucose metabolism, the lactate shuttle, and bioenergetic coupling, suggesting energy storage and usage deficits in the brain in this illness (Table 1)(51-64). Studies employing microarrays found significant decreases in the expression of genes encoding proteins involving the malate shuttle, tricarboxylic acid (TCA) cycle, ornithine–polyamine, aspartate–alanine, and ubiquitin metabolism in the dorsolateral prefrontal cortex (DLPFC) in schizophrenia. These changes were not attributable to antipsychotic treatment, which may have a restorative effect (61). Alterations in these genes might have significant implications for oxidative phosphorylation, which is a key mechanism of ATP production for neurotransmission. Other studies implicate mitochondrial dysfunction in the pathophysiology of schizophrenia (63,

7

64). Further, a genetic study demonstrated evidence in schizophrenia for linkage between enzymes that control glycolysis, such as 6-phosphofructo-2-kinase/fructose-2,6- bisphosphatase 2 (PFKFB2), hexokinase (HXK) 3, and pyruvate kinase (PK) 3, suggesting that genetic risk for this illness includes bioenergetic substrates (65).

Proteomic analyses also highlight the abnormal expression of bioenergetic targets in schizophrenia (57). Two-dimensional gel electrophoresis and mass spectrometry identified

11 downregulated and 14 upregulated proteins in the posterior superior temporal gyrus in schizophrenia. About half of these hits are enzymes involved the regulation of energy metabolism, such as aldolase C and glyceraldehyde-3-phosphate dehydrogenase, likely to impact bioenergetic systems. The same study also reported differentially expressed ATP synthase subunits, which may result in altered ATP metabolism and ultimately contribute to bioenergetic uncoupling in schizophrenia. Importantly, all schizophrenia subjects in this study were treated with antipsychotic medication. Some proteomic studies suggest that these metabolic abnormalities could be an effect of medication, while others implicate bioenergetic dysfunction as a central element of the disease (60, 65).

Several other postmortem studies have found abnormalities in the activity of metabolic enzymes in schizophrenia. For instance, there is a decrease in first half and an increase in second half TCA cycle enzyme activity in the DLPFC of schizophrenia (51).

Most of these subjects (9/13) were off of antipsychotic medications for at least 6 months prior to death, suggesting that alterations in TCA cycle enzyme activity are a core feature of the illness (65). Other studies have also shown a decrease in specific activity of mitochondrial respiratory chain enzymes in the frontal cortex (66, 67). These data suggest that functionality of metabolic proteins, and not expression levels alone, could be important in the pathophysiology of chronic schizophrenia.

8

In vivo evidence for metabolic dysfunction

Postmortem studies have several limitations, such as postmortem interval, mRNA/protein integrity, lifetime effects of medication on brain neurochemistry, and samples that reflect the later or more “mature” stages of the illness. These factors may variably impact dependent measures, including mRNA, protein, and receptor binding site expression

(68). One way to circumvent many of these challenges is to perform studies in living patients. In vivo proton magnetic resonance spectroscopy (MRS) studies offer a noninvasive approach to directly study brain bioenergetics in schizophrenia. Interestingly, a study employing phosphorous spectroscopy (31P-MRS) found a decrease (22%) in creatine kinase activity in schizophrenia, an enzyme critical for maintaining stable ATP levels during altered neuronal activity (53). While 22/26 patients in this study were taking antipsychotic medication, other MRS studies in medication naïve patients also implicate abnormal bioenergetic pathways in schizophrenia, suggesting that decreases in the availability of high-energy phosphates may be a common feature of the illness (69). For example, using high field MRS, one study demonstrated elevated in vivo brain lactate levels in patients with schizophrenia, possibly indicating metabolic dysfunction with a shift towards anaerobic glycolysis (70). However, MRS studies also face limitations. This includes large voxel sizes unable to differentiate contributions from white or gray matter, low signal to noise ratio, long acquisition times, and (often) small sample sizes. Variability in patient demographics and methodological differences could contribute to the inconsistent reports from imaging studies on bioenergetics in schizophrenia. However, MRS studies overall have provided meaningful indexes of brain activity and disruption of metabolic energy pathways in schizophrenia at a macroscopic level.

Cell-specific metabolic changes in schizophrenia

9

Two groups previously found cell-specific changes in gene expression of bioenergetic factors in schizophrenia (71-73). Using dentate granule neuron samples from the hippocampus, one group observed decreases in mRNA expression for clusters of genes that facilitate mitochondrial oxidative energy metabolism, ubiquitin-proteasome systems, and synaptic plasticity (71). This included transcripts for lactate dehydrogenase A, reduced nicotinamide adenine dinucleotide dehydrogenases, and ATP synthases. Most schizophrenia subjects in this study were on antipsychotic medication at the time of death, although unmedicated and medicated subjects contributed almost equally to these findings.

Changes were not observed in bipolar disorder (BPD) or major depression disorder (MDD) subjects, some of which were also on psychotropic medications (71). In two separate studies, another group found marked decreases in mitochondrial and ubiquitin related genes in layer 3 and 5 pyramidal neurons in the DLPFC of schizophrenia subjects (n=36 and n=19)(72, 73). These results also did not extend to a cohort of BPD or MDD subjects

(n=19), suggesting that cell-subtype specific alterations of metabolic gene expression may be unique to schizophrenia. Authors posit these changes in schizophrenia are not due to medication effects, since this cohort of BPD and MDD subjects also included patients on antipsychotic medication, and changes in these transcripts in pyramidal cells in the DLPFC of antipsychotic-treated monkeys were not detected (72). These findings support a molecular link between signatures of mitochondrial dysfunction and spine pathology in schizophrenia, which is well-documented in this brain region (74-77). Supporting this hypothesis, ubiquitin-proteasome systems are strongly linked to metabolics and the control of synaptic protein connectivity, signaling, and turnover (78-82). For instance, degradation of the main positive regulator of glycolysis (PFKFB3) in neurons through the ubiquitin– proteasome pathway results in an inability to upregulate glycolytic flux during increased synaptic activity (40). Taken together, these findings suggest that bioenergetic pathways

10 function differently across different brain regions and are cell-subtype specific. The differential involvement of brain regions, circuits, and cell types is in keeping with the diversity of cognitive symptoms in schizophrenia, since persons afflicted with this illness have heterogeneity in the onset, prognosis, and phenotype of cognitive impairment.

Do certain cell-subtypes exhibit greater bioenergetic susceptibility in schizophrenia?

Apart from pyramidal cells, another class of neurons raises interest due to its particularly high-energy usage and susceptibility to oxidative stress: GABAergic interneurons. Parvalbumin positive (PV+) interneurons are highly sensitive to redox states and reactive oxygen species signaling, and oxidative stress is linked to long lasting PV+ interneuron defects and cognitive deficits in adulthood (83). Particularly, there is evidence that oxidative stress during the critical window in development leads to loss of PV+ interneurons, and may contribute to abnormal brain development and schizophrenia pathology (83, 84). Early in development, the accumulation of Cl− by Na-K-Cl cotransporter

(NKCC1) results in GABAA receptors exhibiting excitatory properties, stimulating synaptic growth and requiring large amounts of energy. Later in life, GABAA receptors become inhibitory due to the delayed expression of the chloride exporter K-Cl cotransporter

(KCC2)(85). One study found increased mRNA expression of two chloride channel regulatory kinases (OXSR1 and WNK3) in the DLPFC in schizophrenia, suggesting a further dysregulation of chloride transport and energy consumption (86). These findings suggest an abnormal GABAergic metabolic profile in schizophrenia, which could be due to oxidative stress.

Glia are another cell type with vital role in bioenergetic homeostasis that may be abnormal in schizophrenia. For instance, a recent study shows that childhood-onset patient

11 derived induced pluripotent stem cells (iPSCs) show delayed differentiation into astrocytes with glial pathology (87). However, limited work has been done examining bioenergetic processes of glial cells in schizophrenia, and there is little direct evidence for cell-subtype metabolic dysfunction in astrocytes in this illness (see below for astrocyte dysfunction in schizophrenia). However, astrocyte and neuron metabolics are tightly coupled via the glutamate/glutamine cycle, and extensive work has been done examining these substrates in schizophrenia.

A brain with dysfunction in multiple brain regions and cell-subtypes such as in schizophrenia may not have the reserve capacity to compensate for this deficit. While there is strong evidence for disruption of limbic circuits in schizophrenia (including frontal cortex, hippocampus, striatum, and thalamus), nearly every brain region has been implicated to an extent (including cerebellum) in schizophrenia pathology. This supports the idea that chronic schizophrenia, often viewed as a developmental illness with synaptic abnormalities, could be accompanied by widespread metabolic dysfunction, attributable to high metabolic demands placed on neurons by the processes involved in neurotransmission.

Bioenergetic profile of GABAergic interneurons

The bioenergetic profile of GABAergic interneurons in adulthood, particularly fast- spiking interneurons, has also been thoroughly reviewed (88). One study combining high- resolution 2-deoxyglucose and immunohistochemistry suggested that glucose metabolism might be significantly higher in GABAergic neurons than in glutamatergic neurons (89).

Additionally, there is evidence that glucose metabolism increases during long-term recurrent inhibition of hippocampal pyramidal cells (90). It is possible that these high-energy processes in GABAergic neurons make them highly vulnerable to bioenergetic deficits, and

12 that decreases in GABAergic inhibitory tone in schizophrenia might reflect a decrease in glucose utilization.

The metabolic role of astrocytes in glutamatergic function

It is well established that glutamatergic systems are disrupted in schizophrenia. In a normal brain, neurons have lower capacity than astrocytes for glutamate reuptake.

Astrocytes are responsible for the majority of glutamate uptake (about 75%) via EAATs and recycle glutamate to the precursor glutamine, which neurons can readily transport (91-93).

This is referred to as the glutamate/glutamine cycle, and is bioenergetically costly. However, glutamate entering astrocytes can meet several metabolic fates, including entering the TCA cycle or lactate and/or ATP production (94-96). Studies have confirmed that a significant amount of glutamate is oxidatively metabolized in astrocytes to lactate (by glutamate dehydrogenase and the TCA cycle) when energetic demand is high, and that the amount converted to glutamine is proportionately decreased (92, 95, 96). Interestingly, EAATs are co-expressed with Na+/K+ ATPases, mitochondria, and glycolytic enzymes to signal rapid glycolysis and lactate generation when neuronal activity is high (97, 98). Since the glutamate/glutamine cycle in astrocytes is tightly linked to both metabolics and neurotransmission, alterations in this cycle may indicate disrupted bioenergetics coupling between neurons and astrocytes.

We and other groups have found changes in cellular and subcellular localization of glutamate transporters in schizophrenia (97, 99-102). Abnormal EAAT expression on astrocytes may lead to pathological glutamate spillover, as well as a decrease in the generation of bioenergetic substrates for neuronal consumption (101). Localization of

EAATs impacts synaptic plasticity (99), and changes in localization suggest uncoupling of glutamate transporter protein complexes from mitochondria (100). Supporting this

13 hypothesis, one study demonstrated decreased labeling of astrocytes adjacent to blood vessels in schizophrenia, suggesting decreased access to the vascular space, which is the primary source of glucose. Such changes could contribute to diminished metabolic capacity (103). Since astrocytes are integral to the bioenergetic homeostasis and fidelity of synaptic function, targeted studies examining changes in these cells is a promising avenue for understanding the pathophysiology of schizophrenia (104).

Astrocyte dysfunction in schizophrenia

Converging evidence suggests that astrocyte dysfunction contributes to the pathophysiology of schizophrenia, including changes in molecular, structural, and functional properties, as well as defects in astrocyte-mediated glutamate reuptake (97, 105, 106). For instance, one of the largest GWAS to date found astrocyte and oligodendrocyte gene sets associated with increased risk for schizophrenia, particularly those related to astrocyte signaling at the synapse (107). Another microarray study found that most abnormally expressed genes in the prefrontal cortex in schizophrenia were oligodendrocyte and astrocyte related (108). These substrates and bioenergetic systems in astrocytes warrant further study.

Oxidative stress and chronic inflammation in schizophrenia

Another possible contributor to abnormal brain metabolism in schizophrenia is oxidative stress and chronic inflammation. There is strong evidence for immunological and oxidative stress abnormalities in schizophrenia, both of which have been firmly established as major contributing factors in the development of mitochondrial dysfunction and impaired bioenergetics (reviewed in (109)). Additionally, peripheral markers of oxidative stress, nitrosative stress, and inflammation have been linked to mitochondrial complex I

14 dysfunction and correlate with specific aspects of cognitive function in subjects with schizophrenia (110, 111).

What drives metabolic abnormalities in schizophrenia?

Pharmacologic, genetic and theoretical considerations suggest schizophrenia as a developmental disorder with synaptic dysfunction (112-116). The accumulating evidence discussed above suggests metabolic disturbances are also a key feature of this illness. As the brain develops, bioenergetic organization and the formation of synapses occur simultaneously, creating a fundamentally interdependent system. Reflecting the heterogeneous nature of schizophrenia, some “indirect” cases may develop an intermediate metabolic phenotype secondary to inherited genetic risk for synaptic dysfunction, while some “direct” cases may have genetic risk for impaired bioenergetic systems, leading to the inability of cells to meet the energy demands of synaptic machinery. Thus, metabolic dysfunction may be a primary cause of schizophrenia and/or an intermediate phenotype secondary to synaptic dysfunction.

Genomic variation of synaptic and metabolic systems confers risk for schizophrenia

The genetic risk for schizophrenia is complex and includes numerous synaptic risk factors that appear to contribute to its pathophysiology. For example, genetic susceptibility factors include genes that play roles in NMDAR function, synapse development/plasticity, and postsynaptic pathways (10-12, 113, 117-126). These prominent abnormalities are part of a complex genetic profile that includes genomic variation in other functionally related groups such as metabolic proteins (61-65, 71-73). Thus, the metabolic phenotype observed in schizophrenia could be driven by inherited synaptic and metabolic risk factors, including single-nucleotide polymorphisms or rare CNVs. The combination of synaptic and metabolic

15 genetic insults could coalesce over development, resulting in a brain with synaptic disturbances and diminished bioenergetic capacity. This could contribute to abnormalities described in this illness such as decreased spine density (74, 127), loss of neuropil (128), decreased expression of glutamate transporters (102, 112-116, 129-131), altered expression of glutamate receptors, and other molecular changes (129, 132). This hypothesis is also supported by genetic linkage studies and numerous findings of abnormally regulated transcripts related to mitochondrial function, glucose utilization, and other high-energy pathways (61-65, 71-73).

Does synaptic dysfunction during brain development yield an intermediate metabolic phenotype?

Although alterations in metabolic gene expression may confer some risk for schizophrenia, it is also possible that genetic risk culminating in synaptic dysfunction could be driving perturbations of metabolic systems. Metabolism and normal synaptic function are highly interrelated, thus abnormal synapses are likely to have altered bioenergetic capacity

(and vice versa). Developing a brain with synaptic dysfunction could result in an intermediate metabolic phenotype in schizophrenia, which may contribute to cognitive symptoms (Figure 2). This coincides with the developmental nature of schizophrenia, where the age of onset is typically later in life (18-25 years)(133). Taken together, these data raise the question of whether or not metabolic dysfunction in schizophrenia is genetic, an intermediate phenotype acquired secondary to antipsychotic treatment, or some combination of both.

Changes in metabolic systems in schizophrenia due to antipsychotic drug treatment

16

The effects of medication in studies that include patient populations taking multiple classes of antipsychotic medications are often inconsistent and likely contribute to divergent bioenergetic findings in schizophrenia. For example, the typical antipsychotic drug haloperidol decreases glucose uptake and lactate production in cultured oligodendrocytes, suggesting reduced glycolysis (134). Using positron emissions tomography (PET), another study demonstrated haloperidol decreases glucose metabolism in the dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) in schizophrenia (135). Conversely, the atypical drug clozapine increases glucose uptake, the production of lactate, and enhances the efficacy of oxidative phosphorylation in cultured oligodendrocytes (134). Increases in glucose utilization may contribute to the therapeutic efficacy of some antipsychotic drugs, supported by the ability of glucose treatment to improve cognitive function when administered to schizophrenia patients (136, 137)(reviewed in (138)).

One study examined the effects of neuroleptics (haloperidol, chlorpromazine, risperidone, zotepine, and clozapine) on mitochondrial respiratory chain activity in human cortex. They reported complex I activity of the mitochondrial respiratory chain was progressively inhibited by all neuroleptics (139). This suggests, in at least the cortex, antipsychotic treatment can suppress mitochondrial function. TCA cycle enzymes are also dysregulated in the prefrontal cortex, ACC, and Wernicke’s area in schizophrenia (58, 60,

140, 141). These changes could be influenced by antipsychotic treatment, as TCA cycle enzymes are decreased in rat frontal cortex following haloperidol or olanzapine treatment

(142).

Metabolic abnormalities in drug-naïve patients

17

It is important to consider glucose abnormalities in drug-naïve patients, as antipsychotics could exert diverse effects on these systems. Previous findings consistently implicate impaired glucose tolerance and metabolic syndrome in first episode drug-naïve schizophrenia patients, suggesting metabolic abnormalities are a core feature of the illness

(143-145). Glycolytic pathways also appear to be disrupted in schizophrenia. In peripheral blood mononuclear cells from antipsychotic-naive patients and controls, 18 differentially expressed proteins were reported, of which 8 were related to glycolytic pathways and none were altered in chronically ill antipsychotic-treated patients (146). It is also important to consider metabolic abnormalities across brain regions, which could result in non-uniform antipsychotic modulation on these systems. For instance, local cerebral metabolic rates were determined for multiple brain regions in schizophrenia patients on and off antipsychotic treatment and in healthy controls using PET (147). Before treatment, schizophrenia subjects had lower metabolic activity than did normal controls in frontal and temporal regions, while activity was higher in basal ganglia areas. Treatment with antipsychotic medication nearly normalized metabolic rates in all regions except for frontal cortices, where hypofrontality persisted (147). Another PET study using fluorodeoxyglucose demonstrated that drug free patients with schizophrenia (n=12) had significantly lower regional cerebral metabolic rates of glucose in the hippocampus and the anterior cingulate cortex than did normal controls, but not in neocortical areas or in the extrapyramidal system

(148). A similar decrease in glucose metabolism was found in other cohorts of drug free schizophrenia patients (n=20)(149). Taken together, these findings suggest metabolic pathology in schizophrenia varies by brain region, is not entirely due to treatment with antipsychotics, and is likely modulated by antipsychotic medication diversely across the brain.

18

Animal models as a valuable tool

Given the complex and heterogeneous nature of schizophrenia, it is likely that metabolic genetic predisposition, an intermediate metabolic phenotype secondary to genetic risk for synaptic dysfunction, and some modulation from treatment with antipsychotic drugs may all contribute to the bioenergetic deficits observed in this illness. Multifaceted contributions can make interpreting human results challenging, highlighting the utility of animal models to better investigate these important questions. Animal models of synaptic dysfunction may provide a tool to address the hypothesis that metabolic disturbances may be an intermediate phenotype secondary to genetic risk for synaptic dysfunction. The evidence suggesting a central role for the glutamatergic system in this illness has led to development of several animal models of synaptic dysfunction. Specifically, defects in

NMDARs in schizophrenia (150-158) have led to generation of NMDA receptor hypofunction models (159-165). For example, the NMDA receptor GluN1 subunit knockdown model demonstrates impaired social interaction, increased stereotypic behaviors, decreased performance in spatial and working memory tasks, and increased auditory and visual event related potentials (suggesting decreased inhibitory tone)(161, 163, 165-167). Other targeted mutations in genes encoding glutamate receptors result in similar schizophrenia-like phenotypes, several of which have been implicated in schizophrenia (reviewed in (168)). In addition to genetic models, pharmacological models such as administration of phencyclidine, MK-801, and other NMDA-receptor antagonists can induce positive and negative symptoms associated with schizophrenia (169, 170). NMDA receptor hypofunction models are widely used in schizophrenia research, but are not the only models of dysfunctional synapses. Several other genetic models also exhibit behavioral abnormalities considered endophenotypes for schizophrenia (including astrocyte pathology models), such

19 as GluR-A knockout mice, disrupted in schizophrenia 1 (DISC1) transgenic mice, serine racemase knockout mice, nicotinic receptor knockout mice, and SynGAP heterozygotes

(Table 2)(165, 171-175).

Models informing antipsychotic treatment on metabolic systems

Antipsychotics also modulate the expression of glucose transporters. In PC12 cells, fluphenazine, chlorpromazine, clozapine, and haloperidol all produced significant inhibition of glucose uptake into cells after a brief 30 minute incubation. However, after a 24 hour incubation with these drugs, there was a significant increase in the induction of GLUT3 expression, and to a lesser extent GLUT1 expression (176). In another study, following 48 hour incubation with either haloperidol, olanzapine, or mirtazapine, human leukemic U937 cells had increased expression of GLUT4 and GLUT5 mRNA (177). It is possible that the inhibition of glucose accumulation in neuronal cells leads those cells to perceive a state of glucose deprivation (178). As a result, metabolic systems may have homeostatic responses such as upregulation of GLUTs or increased glycogenolysis. GLUT inhibition may also contribute to the hyperglycemia seen in schizophrenia patients, as demonstrated by the strong correlation between the ability of a drug to inhibit glucose transport in vitro and its ability to induce hyperglycemia in vivo (179).

Animal models are also useful to investigate the effects of antipsychotic medication on metabolic systems. Several published experiments mentioned in this introduction contain antipsychotic treated animal cohorts in an attempt to control for possible medication effects on dependent measures. While these metabolic changes appear to be inherent to the illness, several experiments have not controlled for the possibility of drug interactions, highlighting the need to consistently control for antipsychotic treatment due to the majority

20 of schizophrenia subjects receiving treatment at the time of death. For instance, oral administration of clozapine (10 mg/kg) and haloperidol (1 mg/kg) for 31 days had no effect on gene expression of two glucose metabolism related genes, HXK and aldolase, in the cortex of mice (180). However, in another study, clozapine, but not haloperidol, altered numerous transcripts related to glucose metabolism in mouse forebrain, with downregulation of pathways including the TCA cycle and pyruvate metabolism (181). It is also possible that antipsychotic drugs produce divergent effects on metabolic systems in genetic models with synaptic pathology compared to wild type animals. Several animal models of synaptic dysfunction are currently utilized in schizophrenia research, many of which have metabolic disturbances similar to those seen in schizophrenia (Table 2)(165,

171-175, 182-191). Animal models will continue to be valuable tools in investigating metabolic changes in this illness; however, it is apparent that the interplay between synaptic pathology, antipsychotic medication, and bioenergetic systems is particularly complex and warrants closer scrutiny.

Evidence for an intermediate metabolic phenotype in genetic models of synaptic dysfunction

There is evidence suggesting developing a brain with synaptic dysfunction yields an intermediate metabolic phenotype. This may be due to failure of astrocytes to develop or maintain metabolic coupling with neurons. There is strong evidence linking glycolysis/lactate shuttle defects to developmental NMDA receptor dysfunction in GluN1 subunit KD mice, including abnormal mitochondrial PK protein expression (54, 192). It is unclear which isoform (PKM1 or PKM2) is abnormally expressed, an important consideration as isoform expression varies in cells with different glycolytic profiles. We have also shown marked decreases in two metabolic transporters important in this pathway in the frontal cortex of

21

GluN1 KD mice (MCT4, 63% and GLUT3, 60%)(193). MCT4 plays a vital role in transporting lactate generated from glycolysis in astrocytes into the synaptic cleft for neurons to take up, while GLUT3 transports glucose directly into neurons to be metabolized

(194). Neurons are unable to sustain a high glycolytic flux during prolonged synaptic activity and instead meet their bioenergetic requirements from other sources such as the lactate shuttle, suggesting that defects in lactate transporter MCT4 could lead to bioenergetic uncoupling between cell types (40). Both synaptic function and meeting of energetic demands are essential for cognition, and failure of either could contribute to the cognitive symptoms seen in schizophrenia. It is possible that defects in excitatory synapses in the

GluN1 KD model could lead to a metabolic intermediate phenotype and contribute to poor cognitive performance in these animals. Similar metabolic phenotypes are seen in pharmacological models of NMDA receptor dysfunction.

Metabolic abnormalities in pharmacological models of NMDA receptor dysfunction

Converging evidence from studies using N-methyl-D-aspartate subtype glutamate receptor (NMDAR) antagonists suggests that NMDA receptor blockade is associated with metabolic abnormalities. An NMR study showed a transient increase in lactate in rats treated with a single dose of the NMDA receptor antagonist MK-801. This was not coupled with immediate increases in oxidative metabolism, which is consistent with the hypothesis that rapid increases in energy demand stimulates glycolysis and later recruits oxidative metabolism (195). Studies also report an increase in glucose utilization in both phencyclidine (PCP) and MK-801 treated rats in frontal areas, possibly indicating a shift to glycolysis in astrocytes and lactate production (196, 197). Conversely, after daily injections of MK-801 for 6 days, rats showed decreases in lactate in striato-thalamo-cortical circuits, as well as decreases in mitochondrial function and glycolysis (most severe in the

22 cortex)(198). Tricarboxylic acid (TCA) cycle substrates were also dysregulated in the cortex and hippocampus of subchronic MK-801 treated rats (55). These data suggest that both short and long-term pharmacological blockade of NMDA receptors have the ability to modulate bioenergetic systems.

Possible therapeutic strategies targeting metabolic systems

Augmenting affected systems such as glucose utilization pathways could offer a novel approach to restoring cognitive function in schizophrenia. This could include targeting pro- metabolic substrates pharmacologically. Pioglitazone (Pio), a synthetic ligand for peroxisome proliferator-activated receptor gamma (PPARγ), can alter the transcription and expression of glucose transporters, including GLUT1, leading to changes in glucose uptake through PPARγ and other mechanisms (199, 200). An increase in glucose uptake stimulates glycolytic pathways and may restore some cognitive deficits (201). Interestingly, the effects of 6 months treatment with pioglitazone was assessed in 42 Alzheimer’s patients with comorbid type II diabetes mellitus. Patients receiving pioglitazone treatment had increased regional cerebral blood flow in the parietal lobe and cognitive improvement, as well as enhanced insulin sensitivity (202). Pioglitazone has also been used as an adjunct to antipsychotics to reduce negative symptoms in schizophrenia (203, 204). Other studies administering similar drugs, such as the antibiotic ceftriaxone, which increases glucose metabolism via increased glutamate transporter 1 expression and glutamate uptake, have shown modest decreases in psychotic symptoms in schizophrenia patients (205-207).

DISSERTATION OBJECTIVES

This recent work in both human and animal models highlights the importance of bioenergetic dysfunction in schizophrenia pathophysiology. Pathogenic mechanisms

23 underlying metabolic defects in schizophrenia are complex and not readily explained by genetic variance or neurochemical changes alone. Adding to this complexity, antipsychotic medications appear to interact with metabolic systems in diverse ways. In vivo and postmortem studies reveal the consequences of abnormal metabolic gene expression and developing a brain with dysfunctional synapses. In addition, reverse translational approaches have become increasingly valuable, as it is likely that a combination of diverse variables contribute to bioenergetic deficits in schizophrenia. Teasing apart primary versus secondary causes also poses a challenge not easily overcome by human studies alone, particularly between intertwined entities such as metabolic and synaptic systems. Some bioenergetic disturbances in schizophrenia may be cell, cell-subtype, or micro-domain specific, leading to recent studies examining specific cortical lamina, cell-subtypes, and microdomains. Therefore, the primary goal of this dissertation is to examine synaptic and metabolic interplay in the pathophysiology of schizophrenia, modulate bioenergetic systems through pro-metabolic drug administration in the GluN1 knockdown model of schizophrenia, and provide the framework for future studies.

Rationale for studying the DLPFC

We chose to focus our studies on the DLPFC for several reasons. The DLPFC is involved in top-down regulation of goal-directed behavior (208). In schizophrenia, there is altered activation of the DLPFC in working memory tasks (209) and working memory may be impaired for years before diagnosis (210). Cognitive impairment is a hallmark feature of the illness, and antipsychotics generally lack the ability to treat cognitive decline (211, 212).

Development of adjunct therapies that target cognitive symptoms would greatly improve the quality of patients’ lives. There is also reduced grey matter volume as a consequence of reduced neuropil in the DLPFC in schizophrenia (213). A variety of abnormalities (many of

24 which are metabolic in nature) have been reported in the DLPFC of schizophrenia patients, including gene/protein expression changes, cell density changes, changes in the number of various receptors, and blood flow (68, 97, 214). Alterations in metabolic systems including the ones studied here could result in inefficient functioning of the DLPFC mediated neural circuitry involved in higher order processes/behaviors (215). An inability to respond to dynamic energetic demands or regulate glycolytic pathways connecting astrocytes and neurons could contribute to the phenotypes seen in schizophrenia.

Hypothesis

The central hypothesis of this dissertation is that bioenergetic pathways involving multiple cell-subtypes are abnormal in schizophrenia, and these deficits contribute to cognitive dysfunction in this illness.

Aim 1

This aim investigates excitatory and inhibitory balance in schizophrenia. We test the hypothesis that chloride channels are abnormally expressed in schizophrenia, contributing to impaired GABAergic transmission in schizophrenia.

Aim 2

This aim investigates glycolytic pathways in schizophrenia at the region and cell- level. We test the hypothesis that glycolytic enzymes are disrupted in the dorsolateral prefrontal cortex, and that there are neuron-specific deficits in key components of glycolysis and the lactate shuttle.

25

Aim 3

This aim investigates biological pathways connected to our postmortem metabolic findings using bioinformatic analyses and builds a disease signature that can then be used to identify therapeutic drug candidates.

Aim 4

This aim investigates the efficacy of a pro-metabolic drug implicated in our bioinformatic analyses as capable of reversing metabolic deficits/associated pathway pathologies in improving cognition in an animal model of schizophrenia. We test the hypothesis that increasing glucose uptake via pioglitazone will help restore behavioral deficits in the GluN1 knockdown model.

26

CHAPTER 1 TABLES:

Table 1.1. Characteristics of selected animal models of schizophrenia with “synaptic dysfunction.”

27

Table 1.2. Summary of metabolic abnormalities in schizophrenia and models of schizophrenia.

28

CHAPTER 1 FIGURES:

Figure 1.1. Bioenergetic coupling in normal brain. Glycolysis and oxidative metabolism via tricarboxylic acid (TCA) cycle are key pathways in maintaining synaptic function. Both neurons and astrocytes undergo glycolysis even during aerobic conditions. Glucose, which feeds the glycolytic pathway, can enter cells through glucose transporters (GLUTs). Meeting the energy demand of neurons is highly reliant on the metabolic coupling of neurons to glycolysis and lactate production in astrocytes. There are several key enzymes in glycolysis, including hexokinase (HXK) and lactate dehydrogenase (LDH). This metabolic coupling also requires monocarboxylate transporters (MCTs), which rapidly transport lactate generated by

29 astrocytes into the extracellular space and into neurons. Lactate can also be derived from the breakdown of glycogen in astrocytes. Once in neurons, lactate is converted back to pyruvate by LDH. Pyruvate may then enter the TCA cycle and oxidative phosphorylation to generate 30-36 molecules of adenosine triphosphate (ATP). This net flow of energetic substrates from astrocytes to neurons to support neuronal activity is termed the “astrocyte- neuron lactate shuttle.”

30

Figure 1.2. Synaptic and bioenergetic function in schizophrenia. Genetic and environmental risk factors for schizophrenia include genomic variants and stressful events that impact the NMDA receptor signaling complex. There is a close interrelationship between the development of glutamatergic synapses and the meeting of bioenergetic demands, which if disrupted could in return affect synaptic function, generating a pathological cycle and possibly an intermediate metabolic phenotype. This phenotype could include a metabolic uncoupling of astrocytes and neurons, affecting pathways such as the astrocyte neuron lactate shuttle, and result in an inability to support increases in neuronal activity. There is evidence that the astrocyte-neuron lactate shuttle is necessary for cognitive functions such as long-term memory, suggesting bioenergetic uncoupling could contribute to cognitive deficits in adulthood in this illness. Abbreviations. Dorsolateral prefrontal cortex (DLPFC); hippocampus (HC).

31

Figure 1.3. Description of bioenergetic terminology.

32

CHAPTER 2

Decreased Chloride Channel Expression in the Dorsolateral Prefrontal Cortex in

Schizophrenia

Courtney R Sullivan1,2, Adam J Funk2, Dan Shan3, Vahram Haroutunian4,5, Robert E. McCullumsmith2*

1 University of Cincinnati College of Medicine, Neuroscience Graduate Program, Cincinnati,

United States

2 Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati,

Cincinnati, United States

3 Department of Nephrology, University of Alabama Birmingham, Birmingham, United

States

4 Department of Psychiatry, Mount Sinai School of Medicine, New York, United States

5 James J Peters Veterans Affairs Medical Center, New York, United States

33

ABSTRACT

Alterations in GABAergic neurotransmission are implicated in several psychiatric illnesses, including schizophrenia. The Na-K-Cl (NKCC1) and K-Cl (KCC2) cotransporters regulate intracellular chloride levels. Abnormalities in cotransporter expression levels could shift the chloride electrochemical gradient and impair GABAergic transmission. In this study, we performed Western blot analysis to investigate whether NKCC1 or KCC2 proteins are abnormally expressed in the dorsal lateral prefrontal cortex and the anterior cingulate cortex in patients with schizophrenia versus a control group. We found decreased KCC2 cotransporter protein expression in the dorsal lateral prefrontal cortex, but not the anterior cingulate cortex, in subjects with schizophrenia, supporting the hypothesis of region-level abnormal GABAergic function in the pathophysiology of schizophrenia. Subjects with schizophrenia off antipsychotic medication at the time of death had decreased KCC2 cotransporter protein expression compared to both normal controls and subjects with schizophrenia on antipsychotics. Our results provide evidence for KCC2 protein abnormalities in schizophrenia and suggest that antipsychotic medications might reverse deficits of this protein in the illness.

34

INTRODUCTION

Schizophrenia is a complex and disabling illness characterized by impairments in attention, cognition, planning, and social function (216). The schizophrenia phenotype may include positive, negative, and cognitive symptoms. Examples of these symptoms may include auditory hallucinations, social withdrawal, and working memory deficits, respectively

(2, 217-220). Many of the cognitive and negative symptoms emerge from dysregulation of the prefrontal cortex (PFC) (221-223). Recent evidence suggests these deficits may be secondary to alterations in the regulation of dorsolateral prefrontal cortex (DLPFC) pyramidal neurons by gamma-aminobutyric acid (GABA) interneurons.

Injection of GABA antagonists into DLPFC yields deficits in working memory similar to those found in schizophrenia (224, 225). There is evidence that during the critical window in development, oxidative stress leads to loss of PV+ GABAergic interneurons, and may contribute to abnormal brain development and schizophrenia pathology (83, 84). Moreover, studies have shown increased (>100%) GABAA alpha2 subunit expression on the axon initial segment of pyramidal neurons, without an increase in the pyramidal neurons themselves (224, 226). This up-regulation could be a compensatory mechanism due to reduced inhibitory input from presynaptic GABAergic terminals (224). Taken together, these data indicate a disturbance of inhibitory GABAergic neurons in this illness (227).

GABAergic interneuron function is bioenergetically costly and is mediated, in part, by chloride channels. The efficiency of GABAergic neurotransmission relies on the balance of intracellular chloride concentrations in the postsynaptic cell. Two chloride cotransporters,

Na-K-Cl cotransporter (NKCC1) and K-Cl cotransporter (KCC2), are responsible for uptake and release of chloride ions, respectively (228-230). Early in development, the accumulation of Cl− by NKCC1 results in GABAA receptors exhibiting excitatory properties, stimulating

35 synaptic growth and requiring large amounts of energy. Later in life, GABAA receptors become inhibitory due to the delayed expression of the chloride exporter KCC2 (85). Thus, we investigated chloride channel protein expression levels to assess the roles of these molecules in severe mental illness. Specifically, we hypothesize that abnormalities in the expression of these proteins may contribute to the pathophysiology of schizophrenia.

MATERIALS AND METHODS

Subjects and Tissue Preparation

Anterior cingulate cortex (ACC) and DLPFC postmortem brain samples were provided by the Mount Sinai Medical Center and Bronx Veterans Administration Medical

Center Brain Bank and consisted of thirty-four subjects with schizophrenia and twenty-nine nonpsychiatrically ill comparison subjects. Subjects were diagnosed with schizophrenia based on Diagnostic and Statistical Manual of Mental Disorders III Revision (DSM-III-R) criteria. The medical records of the subjects were examined using a formal blinded medical chart review instrument, as well as in person interviews with the subjects and/or their caregivers. The subjects were evaluated for National Institute of Neurological Disorders and

Stroke and Association Internationale pour la Recherché et l'Enseignement en

Neurosciences (NINDS-AIREN) criteria for a diagnosis of vascular dementia; National

Institute of Neurological and Communicative Disorders and Stroke (NINCDS), DSM-IV and

Consortium to Establish a Registry for Alzheimer's Disease (CERAD) diagnosis of dementia; Consensus criteria for a clinical diagnosis of Probable or Possible diffuse Lewy body disease; unified Parkinson's disease rating scale (UPDRS) for Parkinson’s disease; clinical criteria for diagnosis of Frontotemporal dementia; medical history of psychiatric disease; history of drug or alcohol abuse; and other tests of cognitive function including the

36 mini–mental state examination (MMSE) and clinical dementia rating (CDR). In addition, each brain tissue specimen was examined neuropathologically using systematized macro- and microscopic evaluation using CERAD guidelines. Since the patients in our cohort were elderly at the time of death, many of the subjects have the cognitive impairment associated with aged subjects with schizophrenia (131, 231-233). All samples were derived from the left side of the brain. Subjects with schizophrenia were diagnosed with this illness for at least 30 years.

Subjects were excluded for a history of alcoholism, death by suicide, or coma for more than 6 hours before death. Next of kin consent was obtained for each subject (100).

Schizophrenia and comparison groups were matched for sex, age, pH, and PMI (Table 2.1).

Information regarding the rationale for starting or stopping medications prior to death was not available. The 6 week cut off prior to death for on or off antipsychotic medications is based on reports of the effects of haloperidol on the brain lasting about 4-6 weeks in rodents after antipsychotics are stopped (234, 235). The range of the time off antipsychotic medication for at least 6 weeks was 6-246 weeks, with the mean of 73 weeks, 6 days. We were unable to assess the effects of other psychotropic medications, such as benzodiazepines (BZD) or opiates, as we did not have enough subjects on these compounds to perform valid statistical tests (Table 2.1). Patients did not have access to alcohol or drugs of abuse. They did have access to tobacco, but data regarding tobacco use are not available for this cohort.

For western blot studies, the ACC was dissected from Brodmann areas 24 and 32.

The DLPFC was dissected from Brodmann areas 9 and 46. Samples were then stored at

−80°C. Frozen tissue was first pulverized then homogenized in 300 µl of homogenization buffer containing 5mM Tris-HCl pH 7.4, 320 mM sucrose, and protease inhibitor tablet

37

(Roche Diagnostics, Mannheim, Germany). All subjects were analyzed individually (they were not pooled).

Antibodies

Commercial primary antibodies for human subjects were used as described below:

NKCC1 (#P55011, UniProt, 1:500; Millipore, Billerica, Massachusetts, USA) and KCC2

(#Q9H2X9, UniProt, 1:1,000; Millipore, Billerica, Massachusetts, USA). Valosin-containing protein (VCP) (1:10,000; Novus Biological, Littleton, Colorado, USA) was used as a loading control.

Commercial primary antibodies for rodent studies were used as described below:

KCC2 (1:2,000; Millipore, Billerica, Massachusetts, USA) and KCC2 (1:2000; Abcam,

Cambridge, Massachusetts, USA). Valosin-containing protein (VCP) (1:10,000; Novus

Biological, Littleton, Colorado, USA) was used as a loading control.

Two different antibodies gave similar bands for KCC2 at the predicted molecular weight (123-126 kDa) and no other bands were on the gel (Figure 2.1, Panel A).

Human Western Blot Analysis

Postmortem brain tissue was homogenized with a large handheld homogenizer in

300 µl of homogenization buffer and aliquoted mixture into 50 µl tubes. Total protein concentration was determined with a bicinchoninic acid protein assay kit (Pierce

Biotechnology, Inc., Rockford, Illinois, USA), and absorbance was measured on a

SpectraCount absorbance microplate reader (Packard/Perkin Elmer, Wellesley,

Massachusetts, USA) at 562 nm. Homogenates were stored in 50 µl aliquots at − 80 °C until assayed.

38

Samples for Western blot analyses were prepared with milli-Q water and sample buffer (6 X solution: 4.5% sodium dodecyl sulfate (SDS), 15% b-mercaptoethanol,

0.018% bromophenol blue, and 36% glycerol in 170mM Tris-HCl, pH 6.8) and heated at

70°C for 10 min. Samples (10 µl/well) were then run on 17 well 4–12% gradient gels. The same amount of protein (20 µg) was loaded for each sample. 14 wells were loaded per gel, balanced for control and schizophrenia in duplicate within gels. Investigators were not blinded to group identity. Gels were then transferred to PVDF membranes by BioRad semi- dry transblotters (Bio-Rad, Hercules, CA, USA). The membranes were blocked at 4°C in cold room overnight with LiCor Odyssey blocking buffer (LiCor, Lincoln, NE, USA) for all antibodies. After three 10 min washes in 1 X PBS with 0.1% Tween, the membranes were then incubated with the appropriate second antibody with IR-Dye 680 or 800cw labeled in

LiCor blocking buffer with 0.2% Tween and SDS for 1 h at room temperature. Washes were repeated after secondary antibody incubation with 1 X PBS block. Membranes were scanned using a LiCor Odyssey scanner, and the intensity value for each protein band was measured using the Odyssey 3.0 software. We tested each antibody using varying concentrations of total protein from homogenized human cortical tissue to confirm we were in the linear range of the assay.

We initially measured NKCC1 and KCC2 in the DLPFC. We found a significant change in KCC2 in this region, and then measured KCC2 in the ACC to determine if this change was region specific.

Rat Western Blot Analysis

Rodent studies were performed in accordance with the IACUC guidelines at the

University of Alabama at Birmingham. Adult male Sprague-Dawley rats (250 g) were

39 housed in pairs and maintained on a 12 hour light/dark cycle. Rats received 28.5 mg/kg haloperidol-decanoate or vehicle (sesame oil) by intramuscular injection every 3 weeks for 9 months. This dose was chosen based on previous reports (235-239). Brain tissue was dissected and stored at -80°C.

Rat brain tissue sections were scraped from slides using a razor blade and homogenization buffer. Total protein concentration was determined with a bicinchoninic acid protein assay kit (Pierce Biotechnology, Inc., Rockford, Illinois, USA), and absorbance was measured on a SpectraCount absorbance microplate reader (Packard/Perkin Elmer,

Wellesley, Massachusetts, USA) at 562 nm. Homogenates were stored in 50 µl aliquots at

− 80 °C until assayed.

Samples for Western blot analyses were prepared with milli-Q water and sample buffer (6 X solution: 4.5% sodium dodecyl sulfate (SDS), 15% b-mercaptoethanol,

0.018% bromophenol blue, and 36% glycerol in 170mM Tris-HCl, pH 6.8) and heated at

70°C for 10 min. Samples (10 µl/well) were then run on 4–12% gradient gels. The same amount of protein (10 µg) was loaded for each sample. Seventeen samples were loaded per gel, alternating treated and untreated in duplicate within gels. Investigators were not blinded to group identity. Gels were then transferred to PVDF membranes by BioRad semi- dry transblotters (Bio-Rad, Hercules, CA, USA). The membranes were blocked for 1 hour at room temperature with LiCor Odyssey blocking buffer (LiCor, Lincoln, NE, USA) and incubated at 4°C in cold room overnight with for all antibodies. After three 10 min washes in

1 X PBS with 0.1% Tween, the membranes were then incubated with the appropriate second antibody with IR-Dye 680 or 800cw labeled in LiCor blocking buffer with 0.2%

Tween and SDS for 1 hour at room temperature. Washes were repeated after secondary antibody incubation with 1 X PBS block. Membranes were scanned using a LiCor Odyssey

40 scanner, and the intensity value for each protein band was measured using the Odyssey

3.0 software. We tested each antibody using varying concentrations of total protein from homogenized rat cortical tissue to confirm we were in the linear range of the assay.

Data analysis

Using Odyssey Version 3.0 analytical software, near-infrared emissions detected by the LiCor Odyssey scanner were expressed as integrated intensity with top-bottom median intra-lane background subtraction. In each subject, duplicate lanes of NKCC1 and KCC2 protein expression were normalized to VCP as an in-lane loading control and then averaged for each subject. VCP was chosen as a loading control because VCP was previously determined to be unchanged in schizophrenia compared to control subjects (240).

Data were analyzed using Statistica (Statsoft, Tulsa, Oklahoma, USA). Correlation analyses were performed to probe for associations between the expression of NKCC1 or

KCC2 and tissue pH, age, and postmortem interval. One-way analysis of variance was used for all human analyses. Analysis of variance was performed to assess the effects of antipsychotic medication exposure in subjects with schizophrenia and in rats treated with antipsychotic medications.

RESULTS

No significant associations were found between pH, post-mortem interval (PMI) or age and any of our dependent measures. In subjects with schizophrenia, we found a 22% decrease in KCC2 protein expression in the DLPFC [F(1,58)=2.140, p<0.05]. We did not detect changes in NKCC1 protein levels in this region (Figure 2.2 Panel A and 2.2 Panel B).

To determine if changes in KCC2 expression were region-specific, we measured KCC2

41 protein levels in the ACC. We did not any detect changes in KCC2 expression in the ACC

(Figure 2.2, Panel C).

In both the DLPFC and the ACC, subjects with schizophrenia off antipsychotics for 6 weeks or more at the time of death had a 38% and 35% decrease in KCC2 expression compared to subjects with schizophrenia on antipsychotics, respectively [F(1,30)=2.159, p<0.05 and F(1,31)=2.439, p<0.05] (Figure 2E and 2F). There was no significant change in

NKCC1 between subjects with schizophrenia on and off medication in the DLPFC (Figure

2D). We also found a significant decrease in KCC2 expression in subjects with schizophrenia off medication compared to controls in the DLPFC [F(1,57)=4.415, p<0.05].

We did not find any changes in KCC2 expression in the cortex of rats treated with

28.5 mg/kg haloperidol-decanoate or vehicle every 3 weeks for 9 months [F(1,18)=0.05611, p=0.9559] (Figure 2.1, Panel B).

DISCUSSION

While multiple markers of GABAergic neurotransmission are abnormal in schizophrenia, the contribution of chloride channels to these irregularities is less well understood (86, 224, 241, 242). Alterations in expression levels of chloride transporter regulatory kinases could shift the chloride electrochemical gradient and ultimately impair

GABA transmission in schizophrenia (86). Previous studies found that mRNA expression levels of oxidative stress response kinase (OXSR1) and with no lysine kinase 3 (WNK3) were increased in schizophrenia in the DLPFC (86). Maintenance of the electrochemical gradient is dependent on the phosphorylation or dephosphorylation of cotransporter proteins by these regulatory kinases (230). Changes in NKCC1 or KCC2 activity alone may result in unbalanced intracellular chloride levels, leading to abnormal GABAergic function

42

(86, 243). NKCC1 is phosphorylated and activated by OXSR1, resulting in an increase in intracellular chloride (86). The other regulatory kinase, WNK3, both activates NKCC1 and inhibits KCC2, also resulting in increased intracellular chloride (244). The authors of this study proposed that increases in these kinases could disrupt the balance of chloride transport by increasing intracellular chloride levels in GABAergic targets (224). No change in cotransporter mRNA was found in this study, suggesting that changes in expression of these regulatory kinases do not directly impact cotransporter gene expression (86).

However, lower KCC2 activity could diminish the extrusion of chloride from GABAergic target neurons, resulting in GABA channels on the postsynaptic membrane allowing passive outflow of Cl- and an excitatory effect upon GABA binding (86, 244, 245). This could create diminished GABAergic inhibitory tone in subjects with schizophrenia following GABA binding (245).

In the present study, we found decreased levels of KCC2 protein expression in the

DLPFC in schizophrenia. Interestingly, KCC2 is exclusively expressed in neurons, while

NKCC1 is localized to glia (246-249). Lower KCC2 protein levels would further reduce the extrusion of chloride from GABAergic targeted neurons, leading to an imbalance of intracellular chloride levels that could also lead to poor efficiency of GABA channels (86,

250). Taken together, these findings are consistent with a pathophysiologic process suggesting increased intracellular chloride levels in schizophrenia. These data support the hypothesis that a disruption of intraneuronal chloride homeostasis may contribute to altered

GABAergic function in the DLPFC in schizophrenia.

We found decreased KCC2 protein expression in the DLPFC but not in the ACC.

This suggests a region specific down-regulation of KCC2 in schizophrenia. Numerous studies have found region-specific gene expression in schizophrenia, possibly due to

43 differences in neurocircuitry between the ACC and DLPFC (251-254). Studies have shown a loss of interneurons in the ACC in subjects with schizophrenia (255, 256) as well as a loss of serotonin 5HT2 receptor binding sites on GABAergic interneurons. These changes may lead to altered GABAergic inhibitory regulation of ACC pyramidal cells, and possibly contribute to the pathophysiology of schizophrenia (257). We hypothesize our findings contribute to the GABA-related changes in the prefrontal cortex, specifically in the DLPFC.

We also found that subjects with schizophrenia who were off antipsychotic medications for 6 weeks or more at the time of death had decreased KCC2 expression in the ACC and DLPFC, compared to those who were on antipsychotic medications. This decrease in schizophrenia subjects off medication suggests that expression of chloride channels might be modulated via dopaminergic systems and D2 antagonists. To further assess the effects of antipsychotic medication, we treated rats for 9 months with haloperidol-decanoate (28.5 mg/kg every 3 weeks) or vehicle (sesame oil) to simulate a lifetime of antipsychotic treatment. We chose the antipsychotic haloperidol because nearly all of our subjects with schizophrenia were taking typical antipsychotics. We did not find any changes in KCC2 expression in the haloperidol treated rats, suggesting that the decrease in

KCC2 protein expression is not due to a medication effect, and that the absence of a decrease in KCC2 protein levels in medicated subjects with schizophrenia is a disease by treatment interaction (Figure 2.1, Panel B).

While very few subjects were taking BZDs or opioids at the time of death, it is worthwhile noting the potential effects these drugs might have on related neural networks.

Anxiety associated with schizophrenia may be attenuated with the use of benzodiazepines; however, their efficacy as an anxiolytic varies widely and they are used on a limited basis

(258, 259). BZDs work by increasing the inhibitory effects GABAergic neurons have on their

44 targets (260, 261). After GABA receptor binding, they modulate the GABAA receptor complex allosterically, increasing total chloride ion flux across the postsynaptic membrane.

This could lead to stronger GABAergic tone and hyperpolarization of the GABAergic target, reversing the possible effects of the changes in chloride transporter expression we observed in subjects with schizophrenia. In contrast, opioid use could potentially antagonize the inhibitory effects of GABA (262). Studies have shown that opioid binding to the mu- opioid receptor results in a decrease in presynaptic vesicular GABA release and inhibition of

GABAergic neurotransmission (262-264). Subjects taking opioids at the time of death could have an exacerbation of diminished inhibitory tone exerted by GABA. Finally, we were unable to account for the putative effects of nicotine as smoking data were not available for the study cohort.

Our present findings of decreased KCC2 protein in the DLPFC varies from one previous report that did not find changes in KCC2 mRNA (86). These divergent findings could be secondary to factors such as age, PMI, and/or antipsychotic medication treatment that vary between the two studies, which utilized tissues from different brain collections. For example, our subjects were on average 20 years older, with PMIs about half of the previous work (86). In addition, while the prior study measured mRNA, we measured protein levels, which do not always have the same valence of change as one another (254, 265).

Postmortem studies commonly present challenges such as these, necessitating development of more innovative techniques, and larger, better characterized cohorts of subjects (68, 266).

Taken together with previous findings of altered GABA signaling in this illness, our data support the hypothesis that decreased KCC2 expression in the DLPFC contributes to the pathophysiology of schizophrenia. Our data also support the conclusion that decreases

45 in KCC2 protein expression are not a medication effect, and raise the possibility that antipsychotic medications might restore expression of this protein in subjects with schizophrenia. The glutamate hypothesis of schizophrenia posits that lessened NMDA- subtype glutamate receptor activity in schizophrenia results in decreased stimulation of

GABAergic inhibitory neurons. The loss of GABAergic activity thereby disinhibits glutamatergic neurons, resulting in altered glutamatergic tone (169, 267). Our data suggest an abnormal GABAergic profile, and warrant further studies in glutamatergic cell types.

SUMMARY

It is possible that decreases in KCC2 reflect higher cortical energetic demands similar to that of early in life when KCC2 expression is low and GABAergic neurons produce excitatory effects (268). This energetic strain could have implications for the bioenergetic profile of GABAergic interneurons in schizophrenia, and consequently affect inhibitory regulation of other cell types such as pyramidal neurons. This warrants further investigation bioenergetic pathways in pyramidal neurons and supporting glial cells.

ACKNOWLEDGEMENTS

This study was funded by the Veterans Affairs Mental Illness Research, Education, and Clinical Center. I would especially like to thank Dan Shan for running the Western blot experiments and Adam for help in data analysis.

46

CHAPTER 2 TABLES:

Table 2.1. Subjects’ Characteristics

DLPFC NKCC1 DLPFC KCC2 ACC KCC2

Category Comparison Schizophrenia Comparison Schizophrenia Comparison Schizophrenia

n 28 34 28 32 29 33

Sex 11 m / 17 f 23 m / 11 f 11 m / 17 f 21 m / 11 f 12 m / 17 f 22 m / 11 f

AOD (years) 78 ± 15 74 ± 12 79 ± 14 73 ± 11 78 ± 14 74 ± 11

Tissue pH 6.4 ± .2 6.4 ± .3 6.4 ± .3 6.3 ± .3 6.4 ± .2 6.3 ± .3

PMI (hours) 7.8 ± 6.9 12.3 ± 6.5 8.3 ± 7.1 11.9 ± 6.6 7.5 ± 6.5 12.4 ± 6.9

APD

on/off/unknown 0/28/0 23/11/0 0/28/0 23/9/0 0/29/0 23/10/0

BZD

on/off/unknown 0/28/0 6/28/0 0/28/0 6/26/0 1/28/0 7/26/0

OP

on/off/unknown 8/20/0 0/34/0 8/20/0 0/32/0 8/21/0 0/33/0

Table 2.1. Subject’s characteristics. Abbreviations: female (f); male (m); age of death

(AOD); post mortem interval (PMI); antipsychotic drugs (APD), benzodiazepine (BZD),

opiates (OP), “off” refers the subject being off the medication for at least 6 weeks prior to

death., “on” refers to being on the medication within 6 weeks of death. Data expressed

mean +/- standard deviation.

47

CHAPTER 2 FIGURES:

Figure 2.1. KCC2 Antibody Specificity in Rat. Expression of K-Cl cotransporter (KCC2) in rat prefrontal cortex with varying protein concentration (5 µg, 10 µg, 20 µg). Two different antibodies gave similar bands for KCC2 at the predefined molecular weight (123-126 kDa)

(A). Expression of K-Cl cotransporter (KCC2) in prefrontal cortex in rats treated with haloperidol or vehicle for 9 months (B). Data are shown as mean ± SEM.

48

Figure 2.2. NKCC1 and KCC2 Expression in DLPFC. Expression of Na-K-Cl cotransporter (NKCC1) (A, D) and K-Cl cotransporter (KCC2) (B, E) protein in the DLPFC in control subjects (CTL) and subjects with schizophrenia (SCZ) (A, B) and in subjects with schizophrenia on and off medication for at least 6 weeks (D, E). Expression of KCC2 protein in the ACC in control subjects (CTL) and subjects with schizophrenia (SCZ) (C) and in subjects with schizophrenia on and off medication for at least 6 weeks (F). *P<0.05

**P<0.05 compared to KCC2 expression in the DLPFC of control subjects. Data are shown as mean ± SEM.

49

CHAPTER 3

Neuron-specific deficits of bioenergetic processes in the dorsolateral prefrontal

cortex in schizophrenia

Courtney R. Sullivan*, BS 1, Rachael H. Koene, BA 1, Kathryn Hasselfeld, BS 1, Sinead M. O’Donovan, PhD 1, Amy Ramsey, PhD 2, Robert E. McCullumsmith, MD,PhD 1

1Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati

2 Department of Pharmacology and Toxicology, University of Toronto, ON, Canada

50

ABSTRACT

Schizophrenia is a devastating illness that affects over 2 million people in the U.S. and costs society billions of dollars annually. New insights into the pathophysiology of schizophrenia are needed to provide the conceptual framework to facilitate development of new treatment strategies. We examined bioenergetic pathways in the dorsolateral prefrontal cortex (DLPFC) of subjects with schizophrenia and control subjects using western blot analysis, quantitative real-time polymerase chain reaction, and enzyme/substrate assays.

Laser-capture microdissection-qPCR was used to examine these pathways at the cellular level. We found decreases in hexokinase (HXK) and phosphofructokinase (PFK) activity in the DLPFC, as well as decreased PFKM mRNA expression. In pyramidal neurons, we found an increase in monocarboxylate transporter 1 (MCT1) mRNA expression, and decreases in HXK1, PFKM, PFKL, glucose-6-phosphate isomerase (GPI), glucose transporter 1 (GLUT1), and GLUT3 mRNA expression. These results suggest abnormal bioenergetic function, as well as a neuron-specific defect in glucose utilization, in the

DLPFC in schizophrenia.

INTRODUCTION

A growing body of evidence suggests abnormal bioenergetic function in chronic schizophrenia, including deficits in energy storage and usage processes in the brain.

Microarray studies found significant decreases in expression of genes encoding proteins involving the malate shuttle, tricarboxylic acid (TCA) cycle, as well as the ornithine– polyamine, aspartate–alanine, and ubiquitin metabolism groups in the dorsolateral prefrontal cortex (DLPFC)(51-62, 269). These changes were not attributable to antipsychotic treatment, which may have a restorative effect (61). Further, a genetic study demonstrated evidence for linkage between enzymes that control glycolysis, such as 6-

51 phosphofructo-2-kinase/fructose-2,6-bisphosphatase 2 (PFKFB2), hexokinase 3 (HXK3), and pyruvate kinase 3 (PK3), suggesting that genetic risk for this illness includes bioenergetic substrates (65).

Several postmortem studies have also found abnormalities in metabolic enzyme activity. In schizophrenia, there was abnormal activity of several TCA cycle enzymes in unpooled DLPFC samples (n=13), as well as decreased specific activity of mitochondrial respiratory chain enzymes in the frontal cortex (51, 66, 67). In vivo studies also implicate alteration of bioenergetic pathways in schizophrenia. Direct evidence for bioenergetic dysfunction as a core feature of schizophrenia was reported using magnetic resonance spectroscopy (MRS). A decrease (22%) in creatine kinase activity was found in schizophrenia (53), an enzyme critical for maintaining stable adenosine triphosphate (ATP) levels during altered neuronal activity (53, 270, 271). Other MRS studies in medication naïve patients suggest that decreases in the availability of high-energy phosphates may be a core feature of the illness (69, 272, 273).

While compelling, prior work has not explored critical elements of bioenergetic pathways for glucose utilization in schizophrenia. In normal brain, ∼80% of the energy consumed supports neurotransmission and neurotransmitter cycling (274). Abnormalities in bioenergetic function are likely to have implications for neuronal and circuit activity. Glucose is converted to glucose-6-phosphate (G6P) via HXK in the first step of glycolysis, where it may be further converted to other bioenergetic intermediates by highly regulated enzymes, such as phosphofructokinase (PFK) and glucose-6-phosphate isomerase (GPI), ultimately yielding pyruvate and ATP (275). Pyruvate may be converted to lactate by lactate dehydrogenase (LDH) and shuttled via monocarboxylate transporters (MCTs) into neurons, providing energy for receptor trafficking, spine formation and other neurotransmission

52 events (18, 19). In schizophrenia, cognitive impairment correlated with decreases in striatal cytochrome oxidase and cortical glucose utilization (276, 277). Motivated by these findings, we investigated key metabolic proteins, transcripts, and enzyme activities to assess glycolysis and glucose uptake in severe mental illness. Specifically, we hypothesize that cell-subtype specific abnormalities in glucose utilization pathways may contribute to a common pathophysiology found in chronic schizophrenia (Figure 3.0).

MATERIALS AND METHODS

Tissue acquisition and preparation

Dorsolateral prefrontal cortex (DLPFC, Brodmann area 9) postmortem brain samples originated from the Maryland Brain Collection and were distributed by both the Maryland

Brain Collection and the Alabama Brain Collection. The cohort consisted of subjects with schizophrenia (n=16) and nonpsychiatrically ill comparison subjects (n=16) (Table 3.1,

Table 3.2). Thirteen control subjects and twelve subjects with schizophrenia were common to all experiments (Table 3.1). Subjects were diagnosed with schizophrenia based on DSM-

IV criteria. The medical records of the subjects were examined using a formal blinded medical chart review instrument, as well as in person interviews with the subjects and/or their caregivers, as previously described (278). Schizophrenia, depression, and comparison groups were matched for sex, age, pH, and PMI (Table 3.1).

Overall study design

In this study, there were two overarching experiments: 1) A region-level analysis of mRNA and protein levels, as well as enzyme activity and substrate levels. This experiment utilized DLPFC brain tissue sections scraped from glass slides and homogenized. We used real time quantitative polymerase chain reaction (RT-qPCR), Western blot analyses, and

53 enzyme assays to generate this data. 2) A cell-subtype specific analysis of mRNA expression in DLPFC in enriched populations of pyramidal neurons and astrocytes. This experiment utilized cells cut using laser capture from fresh frozen tissue sections. We used laser capture microdissection (LCM) coupled with qPCR to generate this data.1000 cells were cut from each subject, and all of the cells were used for the qPCR analyses. The tissue sections used for each of these experiments (14 µm) were cut from tissue blocks, and a number of the subjects from experiments 1 and 2 overlap, as detailed in Table 3.1.

There was no pooling of subjects in any of our experiments.

Laser capture microdissection

Subjects for cell-level studies are described in Table 3.1. LCM was performed as previously described (97, 99, 279). Briefly, 14 µm frozen tissue sections from superficial and deep layers of DLPFC were removed from -80°C storage and allowed to air dry. Tissue sections were rehydrated with distilled H20 and then underwent rapid Nissl staining with

RNase-treated (N8080119, Applied Biosystems; Foster City, CA) cresyl-violet (1% cresyl violet, 1% glacial acetic acid, pH 4.0). Slides were dehydrated in increasing ethanol concentrations (30 seconds 70% EtOH, 30 seconds 95% EtOH, 30 seconds 100% EtOH,

30 seconds 100% EtOH) and cleared in Histoclear II (64110-01, Electron Microscopy

Services, USA) for 10 minutes. After drying in a hood for 15 minutes, enriched populations of pyramidal neurons or astrocytes (1000 of each cell type per subject) were identified via morphology under the 20X objective lens and cut using the Applied Biosystems

ArcturusXT™ LCM instrument and CapSure Macro LCM caps (Life Technologies, formerly

Arcturus, Mountain View, CA, USA)(97, 99, 279). Laser settings were adjusted prior to each session to produce optimal cutting and capturing with laser settings ranging from 70-100 mW in power, and 2,000-3,000 µsec in duration. Separate caps were used for each subject

54 and each cell population. Cells were captured in 4 hour sessions as we have previously demonstrated this time frame has minimal effects on messenger RNA integrity (97).

Following cell capture, each cap was incubated with 50 µl of PicoPure RNA extraction buffer (Molecular Devices, Sunnyvale, CA, USA) for 30 min at 42º C. Samples were then centrifuged for 2 min at 800 × g and stored at -80ºC until further processing. Before PCR experiments, samples were preamplified as previously described (97, 99, 279).

RT-qPCR

For region-level studies, RNA was extracted from DLPFC sections using the RNeasy

Minikit (Qiagen, NL) according to manufacturer’s instructions. RNA was treated with RNase- free DNase (79254; Qiagen, NL) during processing. RNA concentration was determined using UV spectrophotometry at 260nm (NanoDrop, ThermoScientific, USA). Equalized RNA was reverse-transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems,

Foster City, CA, USA).

TaqMan PCR assays for each target gene (Table 3.3) were performed in duplicate on cDNA samples in MicroAmp Fast Optical 96-well optical plates (Applied Biosystems;

Foster City, CA) on an Applied Biosystems detection system (ABI SteponePlus, Applied

Biosystems, Life Technologies, USA). For each reaction, 3 µl of cDNA was placed in a 20 µl reaction containing 10 µl of mastermix and 1x dilution of each primer (Applied Biosystems,

Life Technologies, USA). Reactions were performed with an initial ramp time of 10 minutes at 95°C, and 40 subsequent cycles of 15 seconds at 95°C and 1 minute at 60°C. For negative controls for the qPCR reactions, cDNA was omitted (non-template control) or cDNA was generated with reverse transcriptase excluded from the reaction (no RT control).

Relative concentrations of the transcripts of interest were calculated with comparison to a

55 standard curve made with dilutions of cDNA from a pooled sampling of all the subjects.

Values for the transcripts of interest were normalized to the geometric mean of B2M, β-actin

(ACTB), GAPDH and cyclophilin A (PPIA), housekeeping genes whose expression is unchanged in control and schizophrenia groups (Student’s t-test, p>0.05), for the same samples. All TaqMan PCR data were captured using StepOnePlus Software (StepOnePlus

Real-Time PCR System; Thermoscientific, Waltham, MA)

Western blot analyses

Western blot analyses were performed as previously described (100, 154, 268).

Samples were processed in duplicate and analyzed using Odyssey 3.0 analytical software

(LI-COR Biosciences, Lincoln, NE). We were unable to demonstrate working antibodies in postmortem brain for PFK1, HXK2, MCT4, and GLUT1 (GPI was not assessed due to resources/timing). Prior to examining protein expression, we tested our LDH, LDHA, LDHB,

MCT1, HXK1, GLUT3, and VCP western blot assays using varying concentrations of total protein of human cortical tissue homogenate. These control studies demonstrated that our assays were linear for the protein concentrations used in our studies (data not shown).

Primary antibodies for western blot studies were diluted in LI-COR Odyssey blocking buffer and 0.2% Tween. Secondary antibodies were diluted in LI-COR Odyssey blocking buffer,

0.2% Tween, and 0.01% sodium dodecyl sulfate. Antibodies are described in the

“Antibodies” section below.

LDH and HXK activity assays

Two 14 µm sections per subject were scraped from glass slides and samples resuspended in 100 µl cold 1x PBS buffer with HALT protease and phosphatase inhibitor and then centrifuged at 10,000 g (LDH) or 13,000 g (HXK) in a Sorvall Legend Micro 21R

56 centrifuge for 15 (LDH) or 10 (HXK) minutes at 4 ͦ C. The supernatant was collected, assayed for protein content (using the bicinchoninic acid method, as described first by

(280)) and evaluated using the manufacturer’s protocol (Lactate Dehydrogenase Activity

Assay MAK066-1KT and Hexokinase Colorimetric Assay MAK091-1KT, Sigma Aldrich, St.

Louis, Missouri) with the following changes. A final amount of 1 µg (LDH assay) or 3.75 µg

(HXK assay) protein was added to each assay well. Each sample was assayed with and without a specific inhibitor (LDH: GSK 2837808A, 100uM, 10 MG 5189/10, R&D Systems,

Inc., Minneapolis, Minnesota; HXK: Lonidamine 10 mM, L4900-5MG, Sigma Aldrich, St.

Louis, Missouri). Samples were run in duplicate then read at the recommended wavelength

(450 nm) every 1 (LDH) or 2 (HXK) minutes for 3 hours using a BioTek ELx800. Optical density readings were converted to nmoles of activity/µg/min, and data expressed as enzyme velocity without inhibitor minus enzyme velocity with inhibitor. Assays had a coefficient of variability of 1-3%.

Phosphofructokinase activity assay

Two 14 µm sections per subject were scraped from glass slides and samples resuspended in 100 µl cold 1x PBS buffer with HALT protease and phosphatase and then centrifuged at 13,000 g in a Sorvall Legend Micro 21R centrifuge for 10 minutes at 4 ͦ C. The supernatant was collected, assayed for protein content, and evaluated using the manufacturer’s protocol (Phosphofructokinase (PFK) Activity Colorimetric Assay Kit,

MAK093, Sigma Aldrich, St. Louis, Missouri) with the following changes. A final amount of

3.75 µg protein was added to each assay well. Each sample was assayed with and without an inhibitor (Aurintricarboxylic acid (ATA), 20 µM, A1895, Sigma Aldrich, St. Louis,

Missouri). Samples were run in duplicate then read at the recommended wavelength (450 nm) every 5 minutes for 3 hours using a BioTek ELx800. Optical density readings were

57 converted to nmoles of activity/µg/min, and data expressed as enzyme velocity without inhibitor minus enzyme velocity with inhibitor.

Inhibitors

Inhibitors for LDH (GSK 2837808A, 100uM), HXK (Lonidamine 10 mM), and PFK

(ATA 20 µM) were selected based on literature and optimized for postmortem brain. GSK potently inhibits both LDHA and LDHB (IC50 values are 1.9 and 14 nM for LDHA and LDHB respectively), however higher doses were required to achieve near complete inhibition in postmortem brain (281, 282). The inhibition of HXK by lonidamine is well established in cell lines, where maximum inhibition occurs near 0.5 mM (283-287). However, we observed maximal inhibition in postmortem brain near 10 mM. ATA was selected for its potent inhibition of PFK activity in previous studies, achieving 50% inhibition in purified PFK enzyme at 0.2 µM (288, 289). We found more optimal inhibition in postmortem brain at 20

µM. The activity assays used to measure LDH and HXK activity have been previously used to assess these enzymes (290-296).

Lactate and Glucose-6-phosphate assays

Two 14 µm sections per subject were scraped from glass slides and samples resuspended in 30-50 µl cold 1x PBS buffer with HALT protease and phosphatase inhibitor and then centrifuged at max speed in a Sorvall Legend Micro 21R centrifuge for 4 (lactate) or 10 (G6P) minutes at 4 ͦ C. The supernatant was collected, assayed for protein content, and evaluated using the manufacturer’s protocol (L-Lactate Assay kit colorimetric, ab65331 and Glucose 6 Phosphate Assay kit colorimetric ab83426, Abcam, Cambridge,

Massachusetts, USA) with the following changes. Samples (0.5 µg total protein per well for lactate assay and 30 µg total protein per well for G6P assay) were heated at 95 ͦ C for 5

58 minutes prior to the assay to inactivate any endogenous enzymes. Each sample was assayed with and without enzyme to control for any remaining endogenous enzyme activity.

Samples were incubated with reaction mix for 30 minutes at room temperature and then read at the 450 nm using a BioTek ELx800. Data are presented as sample lactate or G6P concentration with enzyme minus concentration without enzyme. Assays had a coefficient of variability of 1-3%.

Primers

All primers were obtained from Thermo Fisher Scientific, Waltham, MA, USA.

Human (Hs); rat (Rn). Targets are as follows (Table 3.3): MCT1 (Hs00161826_m1), MCT1

(Rn00562332_m1), MCT4 (Hs00358829_m1), HXK1 (Hs00175976_m1), HXK1

(Rn00562436_m1), HXK2 (Hs00606086_m1), LDHA (Hs01378790_g1), LDHB

(Hs00929956_m1), PFKM (Hs01075411_m1), PFKM (Rn00581848_m1), PFKL

(Hs00160027_m1), GPI (Hs00976715_m1), GLUT1 (Hs00892681_m1), GLUT1

(Rn01417099_m1), GLUT3 (Hs00359840_m1), GLUT3 (Rn00567331_m1), B2M

(Hs99999907_m1), B2M (Rn00560865_m1), ACTB (Hs99999903_m1), GAPDH

(Rn01775763_g1), PPIA (Hs99999904_m1), PPIA (Rn00690933_m1).

Antibodies

Commercial primary antibodies were used as described below: LDHA and LDHB

(1:5000, Santa Cruz Biotechnology, Santa Cruz, California, USA), LDH (1:500, Santa Cruz

Biotechnology, Santa Cruz, California, USA), MCT1 (1:500, Thermo Fisher Scientific,

Waltham, MA, USA), HXK1 (1:1000, Santa Cruz Biotechnology, Santa Cruz, California,

USA), GLUT3 (1:100, abcam, Cambridge, Massachusetts, USA), VCP (1:5000, Abcam,

Cambridge, Massachusetts, USA). Secondary antibodies were used at a 1:2000 dilution as

59 described below: Donkey anti-mouse (IRDye 680RD, 925-68072; LI-COR Biosciences,

Lincoln, NE), goat anti-rabbit (IRDye 800CW, 926-32211; LI-COR Biosciences, Lincoln,

NE), goat anti-mouse (IRDye 680RD, 925-68070; LI-COR Biosciences, Lincoln, NE).

Rodent studies

Rodent studies were performed in accordance with the IACUC guidelines at the

University of Alabama-Birmingham. Sample size was chosen to minimize the number of animals necessary to be sufficiently powered based on a priori power calculations from previous studies (97). Animal groups were alternated in terms sacrifice. Investigators were blinded until after data collection. Adult male Sprague–Dawley rats (250 g) were housed in pairs and maintained on a 12-h light/dark cycle. Dissection of rodent brains included removal of olfactory bulb and sectioning tissue from the most rostral point of the prefrontal cortex, the frontal pole.

For antipsychotic studies, rats received 28.5 mg kg−1 haloperidol-decanoate or vehicle (sesame oil) by intramuscular injection every 3 weeks for 9 months (97). We used haloperidol, a typical antipsychotic, for these studies because the majority of our schizophrenia subjects (11/16) were taking typical antipsychotics at time of death. Brain tissue enzyme activity and metabolic concentrations were assayed in the frontal cortex

(n=10 per group) following the 9 month treatment period. Antipsychotic treated rodent LCM- qPCR experiments were performed similar to human studies (97, 99).

In a different study, postmortem intervals (PMI) were simulated by keeping rat brain tissue at room temperature for time points of 0, 4, 8, 12, 24, and 48 hours (n=3 per time point) prior to dissection. Enzyme activity was assayed in the frontal cortex of the rat brains simulating varying PMIs.

60

Statistical analysis

Our general statistical procedure is as follows: Actual values (not percent) were used in all statistical tests. All outliers were removed using the ROUT method (Q=5%). All dependent measures were tested for normalcy and homogeneity of variance using

D'Agostino & Pearson omnibus normality test. If samples were normal, Student’s t-tests

(enzyme, protein studies-immunoblot/western blot, and substrate assays) or ANOVA with

Tukey multiple comparison corrections was performed (qPCR experiments with >10 observations). If data were not normal, Y values were log2 transformed and retested for normalcy. Non-parametric tests were applied data were still not normal: Mann-Whitney

(enzyme, protein studies/western blot, and substrate assays) or Kruskal-Wallis test (qPCR) was used. All LCM-qPCR pyramidal neuron findings remained significant after applying the false discovery rate (FDR) correction.

We probed for associations between our dependent measures and pH, PMI, and age (or RIN for PCR studies) using correlation analyses. If significant, we further investigated using analysis of covariance (ANCOVA). We did not detect any significant correlations between age, pH, and PMI with any of our dependent measures, including mRNA (region and cell-level), protein levels, enzyme activity, or substrate levels.

In addition to postmortem interval, pH, and age, we performed secondary analyses

(using t-tests) to examine EtOH use, smoking, laterality, and on/off antipsychotic medication as grouping variables in our schizophrenia subjects and found no significant effects. No subjects had any dependence for cocaine, sedatives, cannabis, amphetamines, hallucinogens, inhalants, or opioid dependence (only subject 828 took morphine for pain).

Subjects with prolonged agonal states were not included in this cohort.

61

We performed reverse power calculations on 5 of our significant cell-level findings

(power=1-β, α=0.05) and found that power for all experiments was ≥ 0.80 (MCT1=0.89,

HXK1=0.80, PFK=0.99, GLUT1=0.86, and GLUT3=0.92). We also performed reverse power calculations for lactate dehydrogenase A in neurons and found that power=0.74. As this falls below the conventional threshold for power (0.80), it may be that we did not have sufficient power to detect an effect for this particular transcript. For our LCM-qPCR pyramidal neuron studies, we applied FDR analysis post hoc (Q = 5%) to account for making 11 primary analyses for this experiment. These analyses were not necessary for astrocyte results since no targets were significant.

RESULTS

No significant associations were found between pH, post-mortem interval (PMI),

RNA integrity number (RIN), or age and any of our dependent measures (mRNA, protein, enzyme activity, or substrate levels). We adapted all enzyme assays to postmortem brain and used inhibitors to demonstrate specificity of our assays. Enzyme inhibitors used were able to achieve near complete inhibition, similar to 95 ͦ C heat inactivation (Figure 3.1).

Region-level Studies

We detected a significant decrease in phosphofructokinase muscle subtype (PFKM) mRNA expression (24%) in DLPFC in schizophrenia (t=2.16, p<0.05). We did not detect any changes in transcripts for monocarboxylate transporter 1 (MCT1), MCT4, lactate dehydrogenase A (LDHA), LDHB, hexokinase 1 (HXK1), HXK2, glucose transporter 1

(GLUT1), or GLUT3 in schizophrenia versus control at the region-level (Figure 3.2). PFKL and GPI were not assessed at the region-level due to resources and timing. Since transcript expression is not always indicative of protein levels, we also measured protein expression

62 for these genes (68). We did not detect any changes in LDH, LDHA, LDHB, HXK1, MCT1, or GLUT3 protein levels in the DLPFC in schizophrenia (Figure 3.3).

Next, we assayed enzyme activity for key pathways in glucose utilization in the brain.

In subjects with schizophrenia, we found decreases in HXK (26%) and PFK (16%) activity, but not LDH, in the DLPFC (Figure 3.4). We also found a decrease in HXK (27%) activity normalized to protein expression in the DLPFC (t=1.68, p<0.05) (Figure 3.5). We were unable to assess PFK activity normalized to protein expression due to lack of working PFK antibody for human brain. We did not find any changes in LDH, HXK, or PFK enzyme activity in rats treated with 28.5 mg/kg haloperidol-decanoate or vehicle every 3 weeks for 9 months (Figure 3.4). In rats with varying PMIs, no changes were detected in LDH, HXK, or

PFK activity in rat brain samples at 4, 8, or 12 hour time points compared to 0 hours (Figure

3.4). At 24 and 48 hour time points, we detected increases in LDH and PFK activity. To address if the changes in enzyme activity were disease specific, we also examined HXK and PFK activity in a cohort of subjects with major depressive disorder (MDD) (n=20) and did not detect any differences (Figure 3.6). Finally, we detected an increase in lactate, but not G6P, in the DLPFC in subjects with schizophrenia (Figure 3.7). We did not find any changes in lactate or G6P levels in rats treated with haloperidol-decanoate (Figure 3.7).

Postmortem enzyme activity and substrate levels were not associated with PMI or pH

(Figures 3.4).

Cell-level Studies

To address the possibility that changes in expression may be cell-subtype specific, we used LCM-qPCR to assess mRNA expression of our metabolic targets in enriched populations of astrocytes and pyramidal neurons. We have previously shown our LCM

63 samples are enriched for specific cell-subtypes using neurochemical markers (97, 99, 279).

In a sample enriched for pyramidal neurons, we found an increase in MCT1 mRNA expression (22%, p=0.038), as well as decreases in HXK1 (p=0.023, t=3.18, 19%), PFKM

(p=0.003, t=3.20, 22%), PFKL (p=0.010, t=2.74, 27%), GPI (p=0.015, t=2.58, 26%), GLUT1

(p=0.008, t=2.78, 20%), and GLUT3 (p=0.023, t=2.66, 20%) mRNA expression (Figure 3.8, p<0.05). We did not detect any significant changes in mRNA expression in samples enriched for astrocytes (Figure 3.9). Since 11/16 of our schizophrenia subjects were on typical antipsychotic medications, we assessed MCT1, HXK1, PFKM, GLUT1, and GLUT3 transcripts in enriched populations of pyramidal neurons from rats treated with 28.5 mg/kg haloperidol-decanoate or vehicle every 3 weeks for 9 months to see if our findings in pyramidal neurons in schizophrenia were secondary to a medication effect. PFKL and GPI were not assessed due to limited sample and timing. We found increases in MCT1 (17%, t=2.52, p<0.05) and GLUT3 (20%, t=2.41, p<0.05), but no changes in HXK1, PFKM, or

GLUT1, mRNA expression in enriched pyramidal neuron samples of antipsychotic treated rats (Figure 3.8). All LCM-qPCR pyramidal neuron findings remained significant after applying the FDR correction.

Secondary analyses

To assess the use of antipsychotic medication in our patient population, we reanalyzed our significant cell-level targets with and without antipsychotic naïve subjects. All findings remained significant without these subjects (MCT1 p=0.014, HXK1 p=0.001, PFKM p=0.011, PFKL p=0.037, GPI p=0.034, GLUT1 p=0.030, and GLUT3 p=0.036) (Figure

3.10).

64

To assess the effects of EtOH use on our dependent measures in schizophrenia subjects (11 on, 9 off), we performed secondary analyses of schizophrenia subjects on and off EtOH and did not detect an effect (Figure 3.11).

To assess the effects of smoking on our significant findings, we performed secondary analyses of schizophrenia subjects who were smokers/nonsmokers and did not detect any effects (Figure 3.12).

To assess the effects of laterality on our significant findings, we performed secondary analyses of samples from the left or right brain in schizophrenia subjects and did not detect any effects (Figure 3.13).

Confirmation studies for potential variables from online databases

Using the Stanley Medical Research Institute (SMRI) Online Genomics Database

(supported by Dr. Michael Elashoff) (297), we examined the effects of antipsychotic drug treatment on the expression of our significant cell-level gene targets in 50 schizophrenia subjects from 12 independent studies. We performed two analyses using different p-value and fold change (FC) cut offs: 1) p-value cut off of p<0.05 and 1.1 FC (similar to our studies) and 2) p-value cut-off of p<0.001 and 1.2 FC based on previous work (298). In silico analyses at the region-level indicated antipsychotic (AP) treatment in patients increased the expression of GLUT1 and GLUT3 (p<0.05 and 1.1 FC) (Table 3.4). Our cell- level rat study also demonstrated that AP treatment increases GLUT3 expression in pyramidal neurons. However, changes due to AP treatment in the database and our studies are in the opposite direction of our findings, suggesting our results (other than MCT1) are not due to the effects of antipsychotic medication.

65

We also performed in silico analyses in the SMRI Online Genomics Database and examined the effects of lifetime EtOH use, drug abuse, sex, and suicide as cause of death in schizophrenia subjects. Both sex and suicide as cause of death have region-level effects on metabolic targets in this database (Table 3.5).

Non-significant findings

We did not find any change in LDH activity between schizophrenia and control subjects at the region-level (p=0.180). We also did not detect changes in G6P levels

(p=0.907). We did not detect any region-level changes in mRNA expression of LDHA

(p=0.359), LDHB (p=0.703), HXK1 (p=0.397), HXK2 (p=0.323), MCT1 (p=0.165), MCT4

(p=0.752), GLUT1 (p=0.776), or GLUT3 (p=0.907) in the DLPFC. We did not detect changes in LDHA (p=0.978), LDHB (p=0.855), LDH (p=0.634), HXK1 (p=0.326), MCT1

(p=0.176), or GLUT3 (p=0.413) protein expression in the DLPFC.

At the cell-level, we did not detect changes in LDHA (p=0.159), LDHB (p=0.860),

HXK2 (p=0.524), or MCT4 (p=0.23) in samples enriched for pyramidal neurons in schizophrenia versus control subjects. We did not detect any changes in astrocytes cut from schizophrenia and control subjects, including LDHA (p=0.306), LDHB (p=0.921), HXK1

(p=0.080), HXK2 (p=0.346), PFKM (p=0.057), PFKL (p=0.052), GPI (p=0.725), MCT1

(p=0.436), MCT4 (p=0.607), GLUT1 (p=0.767), or GLUT3 (p=0.470).

In major depressive disorder subjects, we did not detect changes in HXK (p=0.65) or

PFK (p=0.828) activity in the DLPFC.

DISCUSSION

To our knowledge, this study is the first to suggest cell-subtype specific changes in glucose utilization in postmortem brain in severe mental illness. In laser-captured cell

66 populations enriched for pyramidal neurons from superficial (2-3) and deep (5-6) layer

DLPFC, we found significant decreases in mRNA expression of four glycolytic enzymes

(HXK1, PFKM, PFKL, and GPI), two glucose transporters (GLUT1 and GLUT3), and an increase in lactate/pyruvate transporter MCT1 mRNA expression in schizophrenia. We did not detect any changes in a cell population enriched for astrocytes, suggesting our findings are cell-subtype specific. In normal brain, glucose enters cells through GLUT1/GLUT3 and is processed by HXK1, GPI, and PFKM/PFKL via glycolysis to produce pyruvate. Pyruvate can then be converted to lactate and transported between cells or intracellularly by MCTs to be oxidized in the TCA cycle when neuronal energy demand is high (299, 300). Our data suggests a decrease in the capacity of pyramidal neurons to generate bioenergetic substrates from glucose via glycolytic pathways. Additionally, if neurons were unable to take up adequate amounts of glucose for glycolysis, the intracellular pool of available pyruvate/lactate for transport into mitochondria may be diminished, ultimately impacting energy supply. Under aerobic conditions, pyruvate generated from glycolysis is oxidatively decarboxylated to form acetyl CoA, which serves as the main input into the TCA cycle. Our data suggests this mechanism is impaired in neurons, which could result in TCA cycle abnormalities and impaired oxidative phosphorylation.

Consistent with this hypothesis, other studies demonstrated decreases in genes related to oxidative phosphorylation in LCM captured pyramidal neurons in schizophrenia

(71-73). Decreases in clusters of genes that encode for mitochondrial oxidative energy metabolism were found in dentate granule pyramidal neurons from the hippocampus. This included transcripts for lactate dehydrogenase A, NADH dehydrogenases, and ATP synthases. These changes were not found in MDD or bipolar affective disorder, suggesting possible specificity for schizophrenia (71). Other studies also observed decreases in

67 mitochondrial related genes in pyramidal neurons in the DLPFC of schizophrenia subjects

(72, 73). Metabolic systems are strongly linked to the control of synaptic protein connectivity, signaling, and turnover (78-82). Thus, decreases in mitochondrial function coupled with abnormal glucose utilization in neurons could reduce the capacity of pyramidal cells to sustain a normal complement of dendritic spines, contributing to the lower DLPFC spine density reported in schizophrenia (71, 74, 75).

Neurons are unable to synthesize glucose and thus are fully dependent upon glucose transporters for glucose uptake/supply (301). Although not usually considered the rate-limiting step in glucose utilization in normal brain, in pathological states decreased

GLUT1 and GLUT3 expression may diminish glucose transport capacity to a threshold resulting in impaired glucose metabolism (302). Interestingly, decreases in GLUT1 and

GLUT3 and impaired brain glucose utilization have been reported in other cognitive disorders. For instance, a reduction in GLUT1/GLUT3 expression has been implicated as a possible cause, rather than a consequence, of neurodegeneration in Alzheimer’s disease

(302-304). Our finding of decreased GLUT1 and GLUT3 transporter expression in neurons suggests a similar impairment in glucose uptake and metabolism as a key feature of chronic schizophrenia, possibly contributing to cognitive impairment. However, it is important to consider mRNA changes in neurons may not have an effect on protein levels or transport activity. Interestingly, McDermott and deSilva hypothesized that genetic deficits in

GLUT1/GLUT3 and poor uptake of glucose in the brain would result in a backlog effect and mild systemic hyperglycemia (305), which has been reported in schizophrenia (143, 306).

Reduced glucose availability in neurons may disrupt bioenergetic coupling systems such as the glutamine/glutamate cycle, which is also perturbed in this illness (307, 308). For example, elevated glutamine to glutamate ratio in the CSF of schizophrenia first episode

68 drug naïve patients suggests endogenous substrates which communicate between neurons and glia are altered (307).

Since most subjects (11/16) in our cohort were taking typical antipsychotics, we examined our dependent measures in haloperidol treated rats. Increases in GLUT3 transcripts in pyramidal neurons following haloperidol treatment suggest our finding of decreased GLUT3 mRNA in schizophrenia is not due to a medication effect. In contrast, our findings of increased MCT1 transcripts in pyramidal neurons in schizophrenia could be secondary to the administration of typical antipsychotic medications. However, it is possible that there is a disease x drug interaction that may not be appreciated in our rodent studies.

To address this limitation, we performed additional in silico analyses using a publically available online database (Table 3.4). In brain samples from subjects with schizophrenia on versus off medications, we found no changes in MCT1, HXK1, PFKM, PFKL, or GPI and increased levels of the glucose transporters GLUT1 and GLUT3. These changes are in the opposite direction of our findings in schizophrenia, and for GLUT3 mirror our findings in antipsychotic treated rats. Interestingly, these increases in glucose and lactate transporters following antipsychotic drug administration offer a novel mechanism for haloperidol’s antipsychotic effect. Increased glucose transporter expression could restore intracellular glucose levels in neurons, while increases in monocarboxylate transporters could circumvent bioenergetic deficits by scavenging extracellular lactate generated by astrocytes.

We detected decreased HXK (26%) enzyme activity in the DLPFC of schizophrenia, possibly indicating a functional defect in glucose utilization in this brain region, impacting

ATP production, oxidative phosphorylation, and synaptic events. HXK1 is normally localized to the outer membrane of mitochondria through specific binding to voltage dependent anion

69 channel (VDAC), where it couples cytosolic glycolysis to mitochondrial ATP production and interacts with the Na+/K+ ATPase (275, 309). This confers HXK1 direct access to ATP generated by mitochondria and facilitates increased activity/high glycolytic rates when needed. Previous studies reported altered subcellular localization of HXK1 in schizophrenia, with a shift in HXK1 partitioning from the mitochondrial fraction to the cytosolic fraction in the DLPFC and parietal cortex (100, 310, 311). Such a shift may diminish ATP production and increase vulnerability to oxidative damage. This functional uncoupling in schizophrenia may contribute to our finding of decreased HXK activity. Such a decrease in HXK activity could further diminish the capacity of cells to coordinate oxidative phosphorylation and glycolysis, which is necessary to respond to bioenergetic demands during neuroplastic events (312).

Our finding of decreased PFK activity (16%) in schizophrenia also supports the hypothesis of impaired glycolytic function in neurons. Increased rates of glycolysis and lactate production (below toxic levels) are necessary for long-term memory formation (43,

49, 313). A decrease in PFK activity could slow the rate of glycolysis to avoid lactate accumulation under resting conditions, but may influence the ability of cells to meet energy demands during neuronal activation. These findings are consistent with previous reports of abnormal enzyme activity in metabolic pathways in schizophrenia, such as decreased creatine kinase activity, as well as a decrease in cytochrome-c oxidase activity in the caudate nucleus (63%) and frontal cortex (43%) (53, 66, 67). Interestingly, similar changes in enzyme activity were not found in a cohort of MDD subjects. Many complex yet subtle abnormalities underlie severe psychiatric illnesses, and while many of these changes may be shared between MDD and schizophrenia, our findings do not appear to extend to unipolar depression. However, in MDD there are also metabolic abnormalities such as

70 alterations in high-energy phosphate metabolism and regulation of oxidative phosphorylation (314). It may be important to examine our dependent measures in other illnesses, such as bipolar disorder, that share high levels of genetic and environmental risk with schizophrenia.

Postmortem interval might impact enzyme activity; however, our HXK and PFK enzyme activity findings are not due to a PMI effect (Figure 3.4). The average PMI in our brain samples was 12-13 hours. Although we found increases in LDH and PFK activity at 24 and 48 hour time points in our PMI rodent studies, these changes are in the opposite direction of our findings in schizophrenia. We also did not detect any changes in G6P levels in the DLPFC in schizophrenia (Figure 3.7), or an effect of PMI on these factors. It may not be possible to detect localized or cell-specific changes in substrate/product levels (such as

G6P) in whole tissue homogenates, and region-level changes (such as increased lactate levels) may be difficult to accurately interpret. Techniques that may provide such specificity have not yet been adapted to postmortem substrate.

A common limitation to region-level studies is the inability to determine the cell type or types in which changes are occurring. Our study addresses this concern using a cell- level approach to assess metabolic pathways in schizophrenia, as well as extensive rodent studies to probe for possible medication and PMI effects. The present study is not without limitations. First, the LCM technique does not produce entirely homogeneous samples. We have previously used neurochemical markers to demonstrate enrichment of populations of pyramidal neurons and astrocytes (97, 99, 279). Due to the labor-intensive nature of the

LCM studies presented here, we have not yet examined our dependent measures at the cellular level in multiple brain regions, other cell-subtypes, a non-schizophrenia disease cohort, or rats with varying PMIs. Additionally, our findings need to be replicated in an

71 independent sample set and a cohort with additional female subjects. The present study would also be strengthened by examining our dependent measures in an animal model of schizophrenia treated with antipsychotics. Finally, schizophrenia is characterized by hypofrontality and thus the impaired glucose metabolism in pyramidal neurons reported here could be causative or a consequence of this abnormality. Further studies are needed to determine if glycolytic disturbances are a primary or secondary effect.

The cell-subtype specific nature of the bioenergetic defects observed here suggest other cell types may have unique bioenergetic profiles in schizophrenia. GABAergic interneurons, particularly fast-spiking interneurons, may be particularly susceptible due to their high-energy processes. Studies suggest that in normal brain, GABAergic neurons consume a substantial fraction of glucose, and glucose metabolism might be higher in

GABAergic neurons than in glutamatergic neurons, making them more vulnerable to bioenergetic insults (89, 315). Additionally, there is evidence that glucose metabolism increases during long-term recurrent inhibition of hippocampal pyramidal cells, and decreases in GABAergic inhibitory tone in schizophrenia might reflect a decrease in glucose utilization (90). Abnormal bioenergetic function in these cells could further disrupt excitatory/inhibitory balance in schizophrenia (discussed in Chapter 2). Further studies examining glycolytic pathways in interneurons could provide insight into the circuity involved.

Taken together with previous studies, the findings reported here suggest metabolic systems are an important target in delineating the pathophysiology of schizophrenia.

Augmenting affected systems such as glucose utilization pathways could offer a novel approach to restoring cognitive function in schizophrenia. This could include targeting pro- metabolic substrates pharmacologically. Pioglitazone (Pio), a synthetic ligand for

72 peroxisome proliferator-activated receptor gamma (PPARγ), can alter the transcription and expression of GLUT1, leading to changes in glucose uptake through PPARγ and other mechanisms (199, 200). An increase in glucose uptake stimulates glycolytic pathways and may restore some cognitive deficits. Previously, pioglitazone was assessed in 42

Alzheimer’s patients accompanied with type II diabetes mellitus for 6 months. Interestingly, patients receiving pioglitazone treatment had increased regional cerebral blood flow in the parietal lobe and cognitive improvement, as well as enhanced insulin sensitivity (202).

Pioglitazone has also been used as an adjunct to antipsychotics, resulting in the reduction of negative symptoms in schizophrenia (203, 204). Other studies administering similar drugs, such as the antibiotic ceftriaxone, which increases glucose metabolism via increased

GLUT1 expression and glutamate uptake, have shown modest decreases in psychotic symptoms in schizophrenia subjects (205-207).

In summary, our novel data implicate functional deficits of glucose metabolism and a cell-subtype specific defect of glycolytic processes in the DLPFC in schizophrenia, possibly impacting the ability of neurons to respond to the high energy demands associated with neuroplastic events (51, 100, 316). Since bioenergetics are tightly coupled to cognitive function, abnormal metabolism in the prefrontal cortex may directly impact cognitive tasks such as working memory in schizophrenia (43, 49, 313). There remain significant challenges in developing high efficacy therapeutics for schizophrenia, but substrates modulating bioenergetic systems such as glycolysis and oxidative phosphorylation could offer plausible avenues for development of novel pharmacological interventions.

SUMMARY

Previously we found decreases in chloride cotransporter KCC2 in the DLPFC, which could affect the efficacy of GABAergic neurotransmission and disturb excitatory/inhibitory

73 balance in schizophrenia. We have now demonstrated metabolic deficits in excitatory pyramidal neurons from the DLPFC, including decreases in mRNA expression of glycolytic enzymes (HXK1, GPI, PFKM, PFKL) and glucose transporters (GLUT1, GLUT3) in schizophrenia. Decreased glycolysis in neurons could increase the strain on metabolic systems, such as the lactate shuttle, to meet energetic demands. Taken together, this suggests a cell-subtype specific dysfunction of metabolic processes in schizophrenia, and warrants further investigation.

ACKNOWLEDGEMENTS

I would like to thank all those that have supported this research in the

McCullumsmith laboratory, as well as the L.I.F.E. Foundation, Lindsay Brinkmeyer

Schizophrenia Research Fund, Alabama Brain Collection, and Maryland Brain Collection.

74

CHAPTER 3 TABLES:

Table 3.1. Subjects Table. mRNA Enzyme/Protein

CTL SCZ CTL SCZ MDD N 16 16 16 16 20 Sex 14m,2f 14m,2f 12m,4f 13m,3f 17m, 3f pH 6.6±0.2 6.6±0.3 6.6±0.2 6.6±0.3 6.6±0.6 PMI 13±4 15±5 12±5 13±6 20±6 Age 44±9 45±11 43±9 45±11 40±9 Rx 0/16 3/11/2 0/16 2/11/3 0/20

Table 3.1. Summary subject demographics. Control subjects (CTL); schizophrenia

(SCZ); postmortem interval (PMI); male (m); female (f); off or unknown / on typical / atypical antipsychotics (Rx).

75

76

Table 3.2. Extended subjects demographics. RNA integrity number (RIN), postmortem interval (PMI), cause of death (COD), alcohol (EtOH), hypertensive arteriosclerotic cardiovascular disease (HASCVD), arteriosclerotic cardiovascular (ASCVD), deep vein thrombosis (DVT), pulmonary embolism (PE). * indicates subjects used for mRNA studies only. ** indicates subjects used for enzyme assays and protein studies only

77

Table 3.3. Table of qPCR primers. Monocarboxylate transporter (MCT); hexokinase

(HXK); lactate dehydrogenase (LDH); phosphofructokinase muscle type (PFKM); phosphofructokinase liver type (PFKL); glucose-6-phosphate isomerase (GPI); glucose transporter (GLUT); beta-2-microglobulin (B2M); beta actin (ACTB); cyclophilin a (PPIA); and glyceraldehyde 3-phosphate dehydrogenase (GAPDH).

78

Table 3.4. Summary of in silico analyses of schizophrenia patients on/off antipsychotics. Database Database Rat Pyramidal Pyramidal SCZ on vs SCZ SCZ on vs SCZ Human DLPFC Neurons Target Neurons off APD, off APD, (SCZ vs CTL) (Haloperidol (SCZ vs CTL) p<0.001, p<0.05, vs CTL) FC>1.2 FC>1.1 1.24 FC, 1.22 FC, 1.17 FC, 1.03 FC, 1.03 FC, MCT1 p=0.165 p=0.039 p=0.021 p=0.345 p=0.345 1.10 FC, -1.24 FC, 1.13 FC, 1.12 FC, 1.12 FC, HXK1 p=0.397 p=0.003 p=0.054 p=0.065 p=0.065 -1.32 FC, -1.28 FC, 1.13 FC, 1.05 FC, 1.05 FC, PFKM p=0.039 p=0.003 p=0.061 p=0.225 p=0.225 -1.27 FC, 1.06 FC, 1.06 FC, PFKL n/a n/a p=0.011 p=0.087 p=0.087 -1.26 FC, 1.09 FC, 1.09 FC, GPI n/a p=0.015 n/a p=0.006 p=0.006

-1.04 FC, -1.19 FC, 1.02 FC, 1.07 FC, 1.07 FC, GLUT1 p=0.776 p=0.009 p=0.775 p=0.029 p=0.029 1.01 FC, -1.19 FC, 1.19 FC, 1.12 FC, 1.12 FC, GLUT3 p=0.907 p=0.012 p=0.027 p=0.001 p=0.001

Table 3.4. Summary of in silico analyses of schizophrenia patients on/off antipsychotics. Findings in schizophrenia versus control patients at the region –level and in pyramidal neurons (first two columns). Last three columns examine targets in haloperidol treated rats and schizophrenia subjects on versus off antipsychotic medication in two databases. Schizophrenia (SCZ); control (CTL); antipsychotic drug (APD); monocarboxylate transporter 1 (MCT1); hexokinase 1 (HXK1); phosphofructokinase muscle type (PFKM); phosphofructokinase liver type (PFKL); glucose-6-phosphate isomerase (GPI); glucose transporter 1 (GLUT1); and glucose transporter 3 (GLUT3). Red text indicates finding was significant according to p-value and fold change cut offs indicated in column headings.

79

Table 3.5. In silico analyses of suicide as COD (Y/N), EtOH (Y/N), drug abuse (Y/N), and sex. Human Pyramidal Database Database Database DLPFC Neurons SCZ +/- SCZ lifetime Database SCZ +/- drug (SCZ vs (SCZ vs suicide as EtOH, M/F, p<0.05 abuse, p<0.05 CTL) CTL) COD, p<0.05 p<0.05 1.24 FC, 1.22 FC, -1.12 FC, -1.04 FC, -1.00 FC, -1.05 FC, MCT1 p=0.165 p=0.039 p=2.05E-06 p=0.059 p=0.870 p=0.012 1.10 FC, -1.24 FC, -1.01 FC, -1.05 FC, -1.00 FC, 1.20 FC, HXK1 p=0.397 p=0.003 p=0.749 p=0.246 p=0.960 p=0.015 -1.32 FC, -1.28 FC, 1.14 FC, 1.02 FC, 1.00 FC, 1.10 FC, PFKM p=0.039 p=0.003 p=0.024 p=0.672 p=0.756 p=0.011 -1.27 FC, -1.06 FC, 1.02 FC, -1.01 FC, 1.10 FC, PFKL n/a p=0.011 p=0.026 p=0.419 p=0.790 p=0.002 -1.26 FC, 1.03 FC, -1.03 FC, 1.05 FC, 1.12 FC, GPI n/a p=0.015 p=0.340 p=0.284 p=0.0261 p=0.0007

-1.04 FC, -1.19 FC, -1.08 FC, -1.07 FC, -1.04 FC, 1.08 FC, GLUT1 p=0.776 p=0.009 p=0.024 p=0.185 p=0.348 p=0.018 1.01 FC, -1.19 FC, 1.01 FC, -1.00 FC, -1.02 FC, 1.11 FC, GLUT3 p=0.907 p=0.012 p=0.689 p=0.947 p=0.369 p=0.0001

Table 3.5. In silico analyses of suicide as COD (Y/N), EtOH (Y/N), drug abuse (Y/N), and sex. Red text indicates finding was significant according to p-value cut off of 0.05.

Cause of death (COD); alcohol (EtOH), schizophrenia (SCZ); control (CTL); monocarboxylate transporter 1 (MCT1); hexokinase 1 (HXK1); phosphofructokinase muscle type (PFKM); phosphofructokinase liver type (PFKL); glucose-6-phosphate isomerase (GPI); glucose transporter 1 (GLUT1); and glucose transporter 3 (GLUT3).

80

CHAPTER 3 FIGURES:

Figure 3.0. Summary of targets in glycolytic pathway. Summary figure of glycolytic pathway with descriptions of targets studied in each experiment. Dorsolateral prefrontal cortex (DLPFC); messenger ribonucleic acid (mRNA); lactate dehydrogenase (LDH); monocarboxylate transporter 1 (MCT1); hexokinase 1 (HXK1); phosphofructokinase muscle

81 type (PFKM); phosphofructokinase liver type (PFKL); glucose-6-phosphate isomerase (GPI); glucose transporter 1 (GLUT1); and glucose transporter 3 (GLUT3).

82

Figure 3.1. Enzyme activity curves. Activity curves in postmortem dorsolateral prefrontal cortex for lactate dehydrogenase (LDH) (A), hexokinase (HXK) (B), and phosphofructokinase (PFK) (C) expressed as nmoles NADH over time and evaluated with and without specific inhibitors and heat inactivation. Inhibitors are as follows: LDH (GSK

2837808A, 100uM), HXK (Lonidamine 10 mM), and PFK (ATA 20 µM).

83

Figure 3.2. Region-level mRNA expression. Lactate dehydrogenase A (LDHA), lactate dehydrogenase B (LDHB), hexokinase 1 (HXK1), HXK2, phosphofructokinase muscle

(PFKM), monocarboxylate transporter 1 (MCT1), MCT4, glucose transporter 1 (GLUT1), and GLUT3 mRNA expression in the dorsolateral prefrontal cortex (DLPFC) of schizophrenia patients (SCZ) and control subjects (CTL) (n=16 per group). Data are expressed as mean ± SEM. *P<0.05.

84

Figure 3.3. Region-level protein expression. Expression of lactate dehydrogenase (LDH), lactate dehydrogenase A (LDHA), lactate dehydrogenase B (LDHB), LDHA:LDHB, hexokinase 1

(HXK1), monocarboxylate transporter 1 (MCT1), and glucose transporter 3 (GLUT3) in the dorsolateral prefrontal cortex (DLPFC) of schizophrenia patients (SCZ) normalized to valosin- containing protein (VCP) and expressed as percent control (CTL) (A). Western blot representation of LDH, LDHA, LDHB, HXK1, MCT1 and VCP expression at predicted band weights in SCZ and CTL subjects (B) (n=16 per group). Data are expressed as mean ± SEM.

85

Figure 3.4. Enzyme activity. Lactate dehydrogenase (LDH), hexokinase (HXK), and phosphofructokinase (PFK) activity in the dorsolateral prefrontal cortex (DLPFC) in control subjects (CTL) and subjects with schizophrenia (SCZ) measured in nmoles nicotinamide adenine dinucleotide hydrate (NADH) over time (A, F, K). LDH, HXK, and PFK activity in rat prefrontal cortex (n=10 per group) treated with haloperidol-decanoate (28.5 mg kg−1) (HAL) or vehicle (CTL) for 9 months (B, G, L), and LDH, HXK, and PFK activity in rats simulating varying postmortem intervals (PMIs) expressed as percent of 0 hour PMI (E, J, O). Correlation of LDH,

HXK, and PFK activity and pH (C, H, M) or PMI (D, I, N) in CTL subjects and subjects with SCZ.

Data are expressed as mean ± SEM (n=16 per group). *P<0.05.

86

Figure 3.5. Specific activity. Specific activity of lactate dehydrogenase (LDH) (A) and hexokinase (HXK) (B) in schizophrenia (SCZ) and control (CTL) subjects. Data are expressed as nmoles of activity/µg/min normalized to protein. Data are expressed as mean

± SEM. N=16/group *P<0.05.

87

Figure 3.6. Major depressive disorder enzyme activity. Hexokinase (HXK) and phosphofructokinase (PFK) activity in the dorsolateral prefrontal cortex (DLPFC) in control subjects (CTL) and subjects with major depressive disorder (MDD) measured in nmoles nicotinamide adenine dinucleotide hydrate (NADH) over time (A, D). Correlation of HXK and

PFK activity and PMI (C, F) and pH (B, E) in CTL subjects and subjects with MDD. Data are expressed as mean ± SEM (n=16 per group).

88

*

Figure 3.7. Substrate concentrations. Lactate (A) and G6P (E) concentration measured in the dorsolateral prefrontal cortex (DLPFC) of control subjects (CTL) and subjects with schizophrenia (SCZ) expressed as nmoles/µg protein. Lactate (B) and G6P (F) concentration in rat prefrontal cortex treated with haloperidol (HAL) or vehicle (CTL) for 9 months. Correlation of lactate (C) and G6P (G) concentration and pH in CTL subjects and subjects with SCZ.

Correlation of lactate (D) and G6P (H) concentration and PMI in CTL subjects and subjects with

SCZ. Data (A-F) are expressed as mean ± SEM (n=10 per group).

89

Figure 3.8. Pyramidal neuron mRNA expression. Relative expression levels of lactate dehydrogenase A (LDHA), LDHB, hexokinase 1 (HXK1), HXK2, glucose-6-phosphate isomerase (GPI), phosphofructokinase muscle (PFKM), phosphofructokinase liver (PFKL), monocarboxylate transporter 1 (MCT1), MCT4, glucose transporter 1 (GLUT1), and GLUT3 transcripts in enriched pyramidal neuron populations from schizophrenia (SCZ) and control

(CTL) subjects (A). Relative expression levels of HXK1, PFKM, MCT1, GLUT1, and GLUT3 transcripts in enriched pyramidal neuron populations from haloperidol and control treated rats

(B). Data from pyramidal neuron enriched samples were normalized to the geometric mean of three housekeeping genes. Data are expressed as percent control ± SEM. *P<0.05.

90

Figure 3.9. Astrocyte mRNA expression. Relative expression levels of lactate dehydrogenase A (LDHA), LDHB, hexokinase 1 (HXK1), HXK2, glucose-6-phosphate isomerase (GPI), phosphofructokinase muscle (PFKM), PFKL, monocarboxylate transporter 1

(MCT1), MCT4, glucose transporter 1 (GLUT1), and GLUT3 transcripts in enriched astrocyte populations from schizophrenia (SCZ) and control (CTL) subjects. Data from astrocyte enriched samples were normalized to the geometric mean of three housekeeping genes. Data are expressed as percent control ± SEM. *P<0.05.

91

Figure 3.10. Secondary analysis of antipsychotic medication. Antipsychotic naïve patients were excluded from cell-level analysis and targets remained significant. Data were normalized to the geometric mean of three housekeeping genes and expressed ± SEM.

*P<0.05.

92

Figure 3.11. Secondary analysis of EtOH use. We performed secondary analyses of schizophrenia subjects on and off EtOH and did not detect an effect.

93

Figure 3.12. Secondary analysis of smoking. To assess the effects of smoking on our significant findings, we performed secondary analyses of schizophrenia subjects who were smokers/nonsmokers and did not detect any effects.

94

Figure 3.13. Secondary analysis of laterality. To assess the effects of laterality on our significant findings, we performed secondary analyses of samples from the left or right brain in schizophrenia subjects and did not detect any effects.

95

CHAPTER 4

Bioinformatic analysis of bioenergetic changes in schizophrenia

Courtney R. Sullivan, BS 1, Adam Funk, PhD 1, Sinead O’Donovan, PhD 1, Eduard Bentea, PhD 1, Erica Carey, BA 2, Jarek Meller, PhD 2, Robert E. McCullumsmith, MD, PhD 1

1Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati

2 Division of Bioinformatics, Cincinnati Children’s Hospital Medical Center

96

ABSTRACT

Bioinformatic analyses can address important biological questions preserving valuable resources, including offering insights on the connectivity of biological networks in human disease. First, to replicate our bioenergetic findings, we performed in silico analyses to probe 2 independent schizophrenia datasets for metabolic targets and found transcripts of glucose/lactate transporters and glycolytic enzymes to be dysregulated. Next, we used bioinformatic tools to build a model of disease based on findings from our postmortem schizophrenia studies and current literature. Using a schizophrenia-specific bioenergetic signature, we investigated connected cellular pathways, kinases, and transcription factors using Enrichr analyses. These results implicated MAPK/ERK, Wnt/β-Catenin, and p53 signaling pathways, as well as other targets involved in cell metabolism, inflammation and immunity, cell cycle regulation, development, and chromatin remodeling. Finally, with the goal of identifying drugs capable of “reversing” our disease signature and the connected pathways, we performed discovery based bioinformatics using the Library of Integrated

Network-based Cellular Signatures and identified 12 unique chemical perturbagens. Of these perturbagens, PPAR agonists presented as promising therapeutic targets for future preclinical experiments. Taken together, these analyses build upon our previous reports of glycolytic defects in schizophrenia, and suggest possible mechanisms to restore these deficits and functionally related pathways include PPAR agonist intervention.

INTRODUCTION

Bioinformatic analyses are powerful tools for generating hypotheses, providing mechanistic insights, and confirming experimental results. These analyses often lead to relevant biological discoveries by integrating the functions of software tools with large

97 datasets generated by genomics, proteomics, metabolomics, and transcriptomics. We previously found bioenergetic defects in schizophrenia, suggesting glycolytic pathways are abnormally regulated. To replicate these postmortem findings at a large scale with minimal resource investment, we probed 2 independent datasets for our metabolic targets: 1) the

Stanley Medical Research Institute (SMRI) Genomics Database, a publicly available database compiling microarray and consortium data from 12 independent studies of schizophrenia and control subjects. 2) Microarray expression data in the DLPFC in cases with schizophrenia (n=16) and controls (n=15) from the Brain Bank of the Department of

Psychiatry of the Mount Sinai School of Medicine (New York, New York)/James J. Peters

Veterans Affairs Medical Center (Bronx, New York). However, changes in brain bioenergetics may have wide spread effects on non-bioenergetic systems, meaning drug interventions only targeting one bioenergetic pathway may not fully restore the comprehensive pathology. Next we aim to use a systems biology approach to model our postmortem schizophrenia findings, analyze connected transcriptomic changes via pathway analyses, and identify therapeutic compounds with the potential to “reverse” glycolytic deficits as well as the predicted associated changes.

We developed a workflow (Figure 4.1 and Figure 4.2.) using the Library of Integrated

Network-based Cellular Signatures web portal (iLINCS) (http://ilincs.org) to extrapolate our postmortem schizophrenia findings by building a model of glycolytic pathology. LINCS is an online reference library of cell-based perturbation-response “signatures” employing a wide range of assay technologies cataloging diverse cellular responses (317). This reference library includes data on the expression of 978 landmark genes (L1000 genes) following the genetic perturbation of the target genes in various cell lines (the so-called “knockdown signature”). The rationale is that the selected landmark genes would capture most of the

98 information contained within the entire transcriptome (318). Over 1000 cell types have been used in LINCS experiments, and model systems include proliferating immortal cell lines, primary cells, and induced pluripotent stem cells (iPSCs) (317). In our previous studies, we found decreases in functionally related glycolytic genes (genes of interest) in pyramidal neurons in schizophrenia. iLINCS may be queried for each gene of interest to determine if there is a “knockdown signature” in the database. By clustering the available knockdown

(KD) signatures in iLINCS, we are able to analyze the connectivity of the KD signatures

(and thus our schizophrenia profile), generating lists of genes that change together.

Furthermore, we can perform Enrichr analyses on these genes to gain perspective on bioenergetic pathology, describe potential multisystem pathological mechanisms, and inform future studies.

Finally, in order to select a potential drug intervention for preclinical testing, we also developed a discovery based workflow to generate a list of perturbagens with L1000 transcript changes in the opposite direction of our schizophrenia signature (i.e. generating an “inverse signature”) (Figure 4.2). iLINCS provides over 40,000 transcriptomic profiles of cell lines following treatment with chemical perturbagens such as FDA approved drugs, chemical probes, and screening library compounds including those with clinical utility and known mechanisms of action (317). FDA approved drugs generated from this “knockdown signature concordance” analysis will serve as candidates for therapeutic treatment in a preclinical model of schizophrenia.

MATERIALS AND METHODS

Database in silico analyses

99

To replicate our bioenergetic findings (Chapter 3), we probed a publicly available microarray database (Stanley Medical Research Institute Online Genomics Database) and microarray data from the DLPFC of control and schizophrenia subjects (Mount Sinai dataset). The Stanley Medical Research Institute Online Genomics Database (supported by

Dr. Michael Elashoff) was used to assess transcriptomic changes of metabolic targets between 50 schizophrenia and 50 control subjects in multiple brain regions (Brodmann area

(BA) 6, BA8/9, BA10, BA46, and the cerebellum) (297). The Mount Sinai dataset includes microarray data from the DLPFC of schizophrenia (n=16) and control subjects (n=15). We reported fold change and p values for our metabolic targets in Table 4.2.

Disease-based seed profile

We began by selecting genes of interest from schizophrenia pathophysiology to build a disease “seed profile.” We selected targets from the glycolytic pathway that were selectively downregulated in our postmortem studies (Chapter 3, GPI, HXK1, PFKM, PFKL), as well as dysregulated in the literature (LDHA, PFKFB2) (Table 4.1). These “seed genes” were used in subsequent iLINCS analyses.

Generation of knockdown signatures for seed genes

We used iLINCS (http://ilincs.org) to retrieve knockdown (KD) KD signatures for each seed gene individually. Seed genes often have multiple knockdown signatures from experiments done in different cell lines. We selected the KD signature from a vertebral- cancer of the prostate (VCAP) cell line for each seed gene. This was the only cell line with

KD signatures available for all seed genes. Each seed gene KD signature (GPI KD signature, HXK1 KD signature, PFKM KD signature, PFKL KD signature, LDHA KD signature, PFKFB2 KD signature) is comprised of the transcriptional changes of 978

100 landmark genes (L1000 genes) when that seed gene is knocked down. We downloaded the knockdown signatures for each seed genes. Visual representations of the transcriptional changes in L1000 genes for each KD signature are presented as heatmaps generated in R

Software (R version 3.4.2, 2017 The R Foundation for Statistical Computing) (Figure 4.3).

The L1000 genes are arranged in alphabetical order on the y-axis and is consistent across figures. A full list of L1000 genes is available on http://ilincs.org.

Knockdown signature concordance analyses

With the goal of identifying small drug-like molecules with inverse signatures, we probed iLINCS for chemical perturbagens that result in L1000 transcriptomic signatures that are highly discordant (anti-correlated) with each seed gene knockdown signature

(potentially reversing the seed gene knockdown signature). This was done individually for each seed gene knockdown signature and the top 20 discordant chemical perturbagen signatures were recorded (Table 4.5). PFKFB2 had no discordant signatures, while GPI and

PFKM had less than 20. Perturbagens appearing multiple times for the same seed gene are either in different cell lines or were administered to cells at different concentrations. We also recorded the top 20 overall discordant chemical perturbagen signatures across all seed gene KD signatures (Table 4.6). The top 20 overall discordant chemical perturbagen signatures were in the VCAP cell line. When duplicates were removed, 12 unique discordant chemical perturbagen signatures remained (Table 4.7).

Clustering of seed gene knockdown signatures

With the goal of identifying shared transcriptomic changes in our seed profile, we used iLINCS to cluster our 6 seed gene knockdown signatures (Figure 4.4). To generate this heatmap, we used the interactive “gene clusters” function in iLINCS to cluster

101 expression data from all of the top 50 differentially expressed genes of each seed gene knockdown. This “union of the top 50 differentially expressed genes” resulted in less than the maximum of 6*50=300 genes clustered due to gene overlap (Figure 4.4). Using the y- axis dendrogram function, we then selected genes that displayed similar changes in expression across the signatures. We refer to genes that are upregulated across all signatures as “consistently upregulated” and genes that are downregulated across all signatures as “consistently downregulated.” Consistently upregulated and consistently downregulated gene groups were selected separately and the data was exported to Excel

(Table 4.3, Figures 4.5 and Figure 4.6).

Pathway analyses of consistently upregulated and downregulated genes

To assess the biological underpinning of consistent transcriptional changes in our seed gene knockdown signatures from iLINCS, we performed traditional pathway analyses on consistently upregulated and downregulated genes using Enrichr

(http://amp.pharm.mssm.edu/Enrichr/) (Table 4.4, Figures 4.7-4.20) (319, 320). Analyses for consistently upregulated genes (67 genes) and consistently downregulated genes (69 genes) were performed separately. We report results for KEGG cell signaling pathway, gene ontology (GO) molecular function, GO cellular components, protein-protein interacting partners, and dbGaP disease implications. Using the ChEA 2016 database in Enrichr, we also report transcription factors that have occupancy sites for a significant number of our consistently changed genes. Finally, using LINCS L1000 database in Enrichr, we report kinases that when knocked down in specific cell lines, increase or decrease genes that are in our consistently increased or consistently decreased gene data sets, respectively (Table

4.4, Figures 4.7-4.20).

102

RESULTS

In silico replication studies for metabolic targets

Using the SMRI Online Genomics Database, we performed in silico analyses probing for changes in metabolic gene targets in schizophrenia subjects across multiple brain regions (Table 4.2). We found two targets to be dysregulated in these datasets

(LDHA, -1.11 FC, p=0.022 and GLUT3, 1.12 FC, p=0.001). In the Mount Sinai dataset, we found decreases in MCT1 (-1.15 FC), MCT4 (-1.16 FC), PFKP (-1.23 FC, p=0.047), and

GLUT3 (-1.34 FC) mRNA expression.

Bioinformatic analyses of altered glycolytic genes in schizophrenia

Consistently upregulated gene analysis

Top cell signaling pathways included forkhead box O (FOXO), mitogen-activated protein kinase (MAPK), and Wnt signaling pathways (Figure 4.7). These pathways often interact and all have important roles in cell metabolism, inflammation, cellular proliferation, oxidative stress prevention, and apoptosis (321-324). Analysis of GO cellular components for consistently upregulated genes included several microtubule and cytoskeletal components (Figure 4.8), while the top hits for GO molecular function included protein kinase C (PKC) and MAPK binding (Figure 4.9). Our protein-protein interactions analysis

(Figure 4.10) yielded multiple hits for protein kinases such as RPS6KA3, MAPK14, GSK3B, and CDK1. We also found hits for proteins involved in protein degradation and protein ubiquitination (CUL1 and SKP1), chromatin remodeling (PCNA and HDAC1), and an estrogen receptor (ESR1).

We next probed for kinases that when knocked down in various cell lines, have upregulated genes similar to the genes consistently upregulated in our signatures. Top hits included several serine/threonine-protein kinases (AKT1, MAP2K2, PTK2, ATR) as well as

G-protein coupled receptors (GPCRs) involved in energy homeostasis (NMUR2, FFAR1)

103

(Figure 4.11). We next probed for transcription factors that had occupancy sites for our consistently upregulated genes. We found several transcription factors that play important roles in developmental processes, such as WT1, TRIM28, NFIB, NUCKS1, and TCF3

(Figure 4.12). Finally, our database of genotypes and phenotypes (dbGaP) analysis implicates diabetes as the most relevant phenotype (Figure 4.13). A summary of these findings can be found in Table 4.4.

Consistently downregulated gene analysis

Top cell signaling pathways include pathways relevant to immune/inflammatory responses and cell cycle regulation (Figure 4.14). The top 5 hits in our GO cellular components analysis for consistently downregulated genes implicated mitochondrial function (Figure 4.15), while the top hits for GO molecular function included several RNA polymerase promotors/activators (Figure 4.16). Our protein-protein interactions analysis

(Figure 4.17) yielded multiple hits for chromatin remodeling proteins (PCNA, HDAC1,

HDAC2, CREBBP). We also found hits for proteins involved in immunity/cell cycle regulation (NFKB2, CSNK2A1, MYC), and ubiquitination (BRCA1).

We next probed for kinases that when knocked down in various cell lines, have downregulated genes similar to the genes consistently downregulated in our signatures.

Top hits included several kinases involved in fatty acid metabolism and energy homeostasis

(RXRA, PPARG, PXK), cellular growth (RYK, WEE1, ABL2), GPCRs (GPR37, LPHN1), and a glucocorticoid receptor (NR3C1) (Figure 4.18).

We next probed for transcription factors that had occupancy sites for our consistently downregulated genes. We found several transcription factors that play important roles in cell proliferation (MYCN, MYC, E2F1, E2F4, FOXM1), development (NUCKS1, POU5F1), and erythroid function (EKLF) (Figure 4.19). Finally, our dbGaP analysis implicates carcinoma, memory, and Grave’s Disease as the most relevant phenotypes (Figure 4.20). A summary

104 of these findings can be found in Table 4.4.

Knockdown signature concordance analyses

Using the connected signatures function in iLINCS, we identified chemical perturbagens that produce transcriptomic signatures that are highly discordant (anti- correlated) with the transcriptomic profile of our seed gene knockdown signatures. We reported the top 20 discordant perturbagens for each seed gene knockdown signature in

Table 4.5 (neither GPI nor PFKM had 20 discordant chemical perturbagens, PFKFB2 had none). Both HXK1 and PFKL generated top 20 perturbagens with high discordance

(average of -0.4), LDHA, GPI, and PFKM had perturbagen signatures with an average discordance of -0.2 to -0.3, while PFKFB2 had no discordant perturbagen signatures.

Next, we report the top 20 overall discordant perturbagen signatures across all seed gene knockdown signatures by ranking all discordant perturbagens for all seed gene KD signatures together (Table 4.6). After removing duplicate hits, 12 unique perturbagens remained. The top hit was Valproic acid, a voltage-gated sodium channel blocker and

HDAC inhibitor. Other hits included typical antipsychotic drugs (trifluoperazine, thioridazine, and fluphenazine), two PI3K inhibitors (Wortmannin and LY-294002), and PPARγ agonists

(troglitazone and genistein).

DISCUSSION

Bioinformatics is an interdisciplinary field that integrates a wide variety of data sources to generate meaningful comparisons and can offer key insights into the mechanistics of human disease. The goal of this study was 3 parts: first, to replicate our metabolic findings in additional databases; second, to analyze the connectivity of several transcriptomic signatures relevant to bioenergetic pathology in schizophrenia through generating a common set of consistently up- and down-regulated genes and using pathway

105 analyses; and third, to generate a list of promising drug interventions with the goal of future preclinical testing. First, we discuss our in silico analyses of bioenergetic targets in independent transcriptomic disease datasets and in the context of our previous postmortem findings (Chapter 3). Next, we discuss in detail the implicated pathways from Enrichr pathway analyses of the consistently changed genes from our clustered KD signatures.

Finally, we discuss the top discordant chemical perturbagens and the implications of glucose uptake modulators as schizophrenia intervention strategies.

In silico analyses

Our postmortem findings demonstrate abnormalities in glycolytic enzymes, as well as glucose/lactate transporters, in the DLPFC in schizophrenia. We sought to strengthen these findings by replicating our results in additional cohorts through in silico analyses. In both the SMRI and the Mount Sinai database, we find several metabolic genes to be abnormal in schizophrenia, including changes in glucose transporters (GLUT3), lactate transporters (MCT1, MCT4), and two glycolytic enzymes (PFKP, LDHA). These changes were found at the region-level, indicating strong metabolic defects in multiple brain regions across datasets and that these abnormalities are a common feature of schizophrenia. Next, we generated a disease signature consisting of glycolytic deficits for further bioinformatic analyses.

Pathway analyses of consistently upregulated genes

We began by using Enrichr to identify pathways associated with our consistently upregulated genes from our clustering of seed gene knockdown signatures (Figure 4.5,

Figures 4.7-4.13). Using KEGG, we generated top cell signaling pathways which returned hits such as FOXO, Wnt, and MAPK signaling. FOXO1 is localized to the nucleus where it binds to the insulin response sequence located in the promoter for glucose 6-phosphatase

106 and increases its rate of transcription (325). The FOXO transcription factor family is regulated by an array of posttranslational modifications, including phosphorylation (322).

FOXO has also been implicated in diabetes complications, and transgenic mice that overexpress FOXO1 have impaired glucose tolerance and a suppression of genes involved in glucose utilization by glycolysis, the pentose phosphate shunt, and lipogenesis (326).

Wnt signaling pathways play a role in inflammation and metabolic regulation, including shifts in fuel utilization between glycolysis and fatty acid oxidation (324, 327).

Abnormal Wnt signaling is linked to altered expression of key metabolic genes, transcription factors, and proteins associated with mitochondrial dysfunction and carbohydrate metabolism (327-330). Wnt pathways can modulate T-cell factor (TCF) transcriptional activity, and in turn target genes such as MYC and PPARδ, which have been linked to mitochondrial glutaminolysis and oxidative capacity (331, 332).

MAPKs have the ability to respond to and regulate key metabolic targets, and are activated in response to insulin. However, inappropriate MAPK signaling can result in metabolic syndrome (323). When phosphorylated, MAPKs such as ERK1/2 and p38 have the ability to enhance the transcriptional activity of peroxisome proliferator-activated receptors (PPARs), and in turn alter glucose uptake (333). Alternatively, prolonged inhibition of the p38 alters the expression levels of glucose transporters GLUT1/GLUT4 and decreases glucose uptake (334). The top hits in our GO molecular function analysis for consistently upregulated genes included the PKC family of enzymes, which typically catalyze phosphorylation reactions and transduce signals that promote lipid hydrolysis, regulated cell growth, and mediate the inflammatory response (335). PKCs also have the ability to activate MAPKs, altering the gene expression, the cellular phenotype, and induce diabetic complications (336). Taken together, top signaling pathways for our consistently upregulated genes implicate glucose utilization, energy homeostasis, mitochondrial

107 function, and cell growth.

The top hit of our protein-protein interaction analysis for consistently upregulated genes was RPS6KA3, a gene that encodes a member of the ribosomal S6 kinase (RSK) family of growth factor-regulated serine/threonine kinases. These kinases phosphorylate many targets, and are key in in cell cycle progression, differentiation, and cell survival (337).

The second hit in our analysis was MAPK14. RSK proteins appear to have important roles in cell cycle progression, differentiation, and cell survival, and are directly phosphorylated and activated by MAPK proteins (such as ERK1) (337). In primary neurons, n-methyl-D- aspartate receptor (NMDAR) activation leads to ERK and RSK2 activation. (338). Recent evidence suggests that RSK2 regulates AMPA receptor transmission, and could be important for neuroplastic events such as learning and memory (338). Other kinase hits include glycogen synthase kinase 3 beta (GSK3B) and cyclin dependent kinase 1 (CDK1), which are heavily involved in energy metabolism, mitochondrial respiration, and cell-cycle progression (339, 340). We also found hits for two core subunits of the SKP1-CUL1-F-box protein E3 ubiquitin ligase complex, CUL1 and SKP1, that regulate ubiquitination of cell proliferation proteins (341). Taken together, our top protein-protein interaction hits suggest a role for pathways involved in energy metabolism and cell proliferation.

Next, we probed for kinases that when knocked down in specific cell lines, increased the expression of genes that were consistently upregulated in our cluster of KD signatures.

We found several serine/threonine-protein kinases including AKT1, PTK2, MAP2K2, and

ATR. AKT1 is one of 3 closely related serine/threonine-protein kinases (AKT1, AKT2 and

AKT3) thought to regulate metabolism and promote cellular proliferation (342). AKT also phosphorylates GSK-3 (stimulating glycogen synthesis) and modulates glucose uptake via increased GLUT1 transcription and targeting of GLUT4 to the plasma membrane (343).

Interestingly, genetic linkage studies have identified AKT as a candidate susceptibility gene

108 for schizophrenia, and AKT1 protein expression and kinase activity are decreased in schizophrenia (344-346). Protein-tyrosine kinase 2 (PTK2) plays an important role in cellular adhesion, migration, and cytoskeletal functions (347). Mitogen-activated protein kinase kinase 2 (MAP2K2) codes for the protein MEK2, which is part of the RAS/MAPK signaling, and regulates cell proliferation, differentiation, migration, and apoptosis (348). Similarly,

ATR regulates the cell cycle by activating checkpoint damage signaling (349).

We also had significant hits for G-protein coupled receptor proteins NMUR2 and

FFAR1. Neuromedin U (NMUR2) encodes a G-protein coupled receptor protein widely expressed in the central nervous system (CNS) that binds the neuropeptide neuromedin U in order to regulate energy homeostasis (350, 351). Free fatty acid receptor 1 (FFAR1) is another membrane protein in the G-protein coupled receptor family that binds medium to long chain free fatty acids, is widely expressed in the brain, and helps in the regulation of energy homeostasis (352). FFAR1 is transcriptionally upregulated by glucose, which is blocked via PI3K inhibitors such as Wortmannin, suggesting a PI3K dependent mechanism

(353). Interestingly, FFAR1 agonists are considered potential drug interventions to enhance insulin secretion in type 2 diabetes (353). Taken together, these results suggest that kinases involved in glucose metabolism and cell cycle regulation are relevant to our consistently upregulated genes.

Finally, we performed pathway analyses to identify transcription factors with occupancy sites for our consistently upregulated genes, several of which played key roles in neurodevelopment. For example, WT1 is a tumor suppressor gene that plays an important role in cellular development and is essential in the development of the urogenital system

(354), while NFIB is essential in tissue differentiation in embryonic development (355).

Similarly, TCF3 regulates many developmental processes such as CNS development (356).

For instance, TCF3 represses Wnt–β-Catenin signaling during neocortical development

109

(357). Lastly, TRIM28 is a ubiquitously expressed protein involved in transcriptional regulation, cellular differentiation and proliferation, and apoptosis (358). Interestingly, deletion of this protein in adult mice results in increased anxiety-like behaviors and stress- induced alterations in stress-induced alterations in spatial learning and memory (359). Our transcription factor analysis also generated hits for estrogen receptors (ESR1, ESR2), tumor suppressors/activators (ELK3), and transcription factors involved in immune responses and macrophage function (CEBPB)(360, 361).

Pathway analyses of consistently downregulated genes

We began by using Enrichr to identify pathways associated with our consistently downregulated genes from our clustering of seed gene knockdown signatures (Figure 4.6,

Figures 4.14-4.20). Using KEGG, we generated top cell signaling pathways which returned hits such as HTLV infection, hepatitis B, p53 signaling, and cell cycle pathways. This implicates inflammation, impaired immunity, and cell proliferation. For instance, p53 is a highly studied tumor suppressor that regulates the expression of >2,500 target genes impacting cellular processes such as cell longevity, while hepatitis B causes acute and chronic necroinflammatory liver diseases involving the adaptive immune response (362,

363).

The top 5 hits in our GO cellular components analysis for consistently downregulated genes implicate the mitochondria and oxidative phosphorylation (Figure 4.15). Implicated cellular components included include 2 cardiac mitochondria (subsarcolemmal and interfibrillar mitochondria), two spermatid mitochondria (Nebenkern and mitochondrial derivative), and general mitochondrion, possibly indicating widespread bioenergetic abnormalities.

The top hit of our protein-protein interaction analysis (Figure 4.17) for consistently

110 downregulated genes was histone deacetylase 2 (HDAC2), a ubiquitously expressed protein that removes acetyl groups from lysine residues on core histones (364). When histones, the primary components of chromatin, are modified by deacetylation (or other posttranslational modifications), chromatin properties and thus DNA packaging and many central biological processes can be affected (365). HDAC1, another histone deacetylase, and its binding partners proliferating cell nuclear antigen (PCNA) and BRCA1 are involved in epigenetics and DNA replication, and DNA repair (366). Epigenetic repression via histone deacetylation could play key roles in development, transcriptional regulation, and cell cycle progression (365). For instance, the deacetylation of p53, one of our significantly implicated cell pathways, is achieved by an HDAC1 complex, which in turn can modulate p53- mediated cell growth arrest and apoptosis (367). Interestingly, HDAC1 is increased in the prefrontal cortex and hippocampus in schizophrenia and overexpression of HDAC1 leads to impairments in working memory (368-370). Similarly, neuron-specific overexpression of

HDAC2 in mice lead to suppression of spine formation, reduced synapse number, and impaired synaptic plasticity and memory formation, characteristics commonly observed in schizophrenia (371). These morphological changes and learning impairments were ameliorated by treatment with the HDAC inhibitor suberoylanilide hydroxamic acid (SAHA; vorinostat), suggesting a possible therapeutic role for HDAC inhibitors in schizophrenia treatment (371). There is mounting instances of schizophrenia patients with mutations in genes encoding chromatin regulators (such as histone modifying enzymes and transcription factors) (372). On a broader scale, meta-analyses from genome-wide association studies

(GWAS) and postmortem transcriptomic data implicate chromatin and nucleosome assembly machinery may contribute to the genetic risk of schizophrenia (373, 374).

Another one of our top hits was CREB-binding protein (CREBBP), a ubiquitously expressed protein with intrinsic histone acetyltransferase activity (375). It is involved in the

111 transcriptional coactivation of several transcription factors, playing crucial roles in development, promoting cell growth, and energy homeostasis (376). For example, CREBBP acetylates p53, E2F, and TCF4 transcription factors affecting their transcriptional activation and DNA binding activities (377-379). Additionally, normal levels of CREBBP are essential for repressing MYC, another top protein-protein interactor, in the G1 phase of the cell cycle, thereby preventing inappropriate entry of cells into S phase (380). MYC is normally activated by Wnt and EGF (via the MAPK/ERK pathway), and universally upregulates gene expression to promote cell cycle progression (381-383). Taken together, these results indicate a disruption in the balance of activation and suppression of cellular proliferation processes.

Another hit in our protein-protein interaction analysis was casein kinase 2 alpha 1

(CSNK2A1), a catalytic subunit of casein kinase 2 (CK2), the constitutively active serine/threonine-protein kinase. CK2 promotes cell survival in part by phosphorylating AKT and stimulating the Wnt signaling pathway (384, 385), systems that were implicated in our

KEGG and kinase analysis of consistently upregulated genes. CK2 is constituently active and also has a regulatory role in p53/MAPK signaling, further implicating cellular processes such as cell cycle progression, proliferation, and suppression of apoptosis (386-388).

Lastly, we had a hit for NFKB2, a transcription factors with essential roles in innate immunity and inflammation (389, 390). Mutations in the NFKB2 gene or protein can result in immunodeficiency syndromes (391, 392). The NFKB family is known for directly binding several transcription factors including p53 (393). Stimulation of NFKB promotes resistance to p53 mediated apoptosis, as p53 and NFKB inhibit each other's ability to stimulate gene expression (394). There is also evidence that NFKB interacts with numerous upstream kinases, chromatin-modifiers such as HDACs, and CREBBP (395-397).

Next, we probed for kinases that when knocked down in specific cell lines,

112 decreased the expression of genes that were consistently downregulated in our cluster of

KD signatures (Figure 4.18). We found multiple kinases with important roles in glucose homeostasis and energy metabolism. The nuclear receptor PPARγ forms a heterodimer with the retinoid X receptor (RXR), two of our top hits, increasing the transcription of various genes that stimulate glucose uptake and carbohydrate metabolism (398). PPARγ ligands such as thiazolidinediones (TZDs), increase glucose utilization, treat hyperglycemia, and represent therapeutic possibilities in disease signatures with deficient glycolytic systems

(399). PPARγ also regulates gut homeostasis by potently inhibiting inflammatory mediator- induced NFKB transcriptional activity (400). For other PPAR family members such as

PPARα, RXRA heterodimerization is required for transcriptional activation on fatty oxidation genes (398). PXK, another hit, also is highly involved in energy metabolism via modulation of Na⁺/K⁺-ATPase enzymatic and ion pump activities (401). Recent work also demonstrated our hit ABL2, a cytoplasmic tyrosine kinase, is activated in the cellular response to oxidative stress (402).

We also found kinases that are involved in axonal growth during CNS development

(RYK), cell growth and cycle progression (WEE1, ABL2), and cytoskeletal remodeling

(ABL2). RYK is a Wnt receptor important in neurogenesis and axon guidance that also interacts with significant transcription factors from our analyses such as FOXO (403, 404).

ABL2, although involved in oxidative stress, also has significant cytoskeletal remodeling functions (405). Similarly, top hit LPHN1 functions in cell adhesion a growth (406). WEE1 is a nuclear kinase important in the regulation of cell size, which it accomplishes via inhibition of CDK1, effectively governing the time point of mitosis entry (407). We also report NR3C1 and GPR37 as significant kinases. NR3C1 is a glucocorticoid receptor regulating genes involved in development, metabolism, and immune responses (408). NR3C1 has the ability

113 to bind to and transrepress other transcription factors such as NFKB, and thus exert anti- inflammatory actions (409). Taken together, these results suggest that kinases involved in carbohydrate/fatty acid metabolism, inflammation, and cell cycle regulation are relevant to our consistently downregulated genes.

Finally, we performed pathway analyses to identify transcription factors with occupancy sites for our consistently downregulated genes, several of which played key roles in cellular proliferation/development and were similar to hits from previous analyses

(MYC, MYCN, NUCKS1, forkhead box M1) (Figure 4.19). We also report additional transcription factors that play crucial roles in cell cycle control such as tumor suppressor proteins E2F1 and E2F4 (410). These transcription factors suggest a connection between our disease signature and regulatory elements of the cell cycle throughout the course of development.

In summary, we extrapolated on our postmortem data that suggested a dysregulation of glycolytic pathways using Enrichr analyses on the consistent gene changes in our clustered seed gene KD signatures. The analyses yield additional pathways and regulators involved in cellular processes such as cell cycle regulation and inflammatory responses. Inappropriate cell cycle regulation could contribute to metabolic deficits, while the immune system has previously been implicated in schizophrenia (411-413). Several of our metabolic and immunity findings have also been recently replicated in other large scale bioinformatic analyses of psychiatric disorders (414). These results highlight the connectivity of glycolytic pathology in schizophrenia to several systems that could be important targets in future studies (Table 4.4).

Knockdown signature concordance analyses

We have divided the top 12 unique perturbagens from our signature concordance analysis into 5 classifications: PI3K inhibitors, antipsychotic drugs, HDAC inhibitors, PPAR

114 agonists, and other. PI3K inhibitors constituted 2 of the drugs that may “reverse” our disease signature. These antitumor drugs inhibit PI3K enzymes, which are part of the

PI3K/AKT/mTOR signaling pathway and play critical roles in many cellular functions such as cellular proliferation, metabolism, and immune cell activation (415-417). Recent work has linked schizophrenia pathology to the PI3K-AKT-mTOR signaling cascade (418, 419).

Additionally, PI3K signaling can modulate synaptic formation and plasticity, and has been implicated in both schizophrenia and autism disorder (reviewed in (420, 421)). Previous work demonstrated pharmacological inhibition of the catalytic subunit of PI3K blocks amphetamine induced psychosis in mice, highlighting PI3K inhibitors as therapeutic target for the treatment of psychiatric disorders (422). However, PI3K inhibition may be more suitable for “positive symptoms,” as PI3K inhibition results in diminished insulin-stimulated glucose uptake by inhibiting translocation of GLUT4 glucose transporters to the plasma membrane (423). Three of our drug discovery analysis hits were typical antipsychotics, which also treat positive symptoms of schizophrenia via dopamine receptor D2 antagonism

(424). It is possible that the glycolytic knockdown signatures used to generate these perturbagens could cause transcriptional changes in L1000 genes involved in the regulation of dopaminergic synapses, suggesting dopamine and metabolic systems could be connected in schizophrenia. This is not entirely surprising as antipsychotics have well documented metabolic effects (425).

The top unique perturbagen in our signature concordance analysis (Figure 4.7) was valproic acid, an HDAC inhibitor that modulates sodium channels and enhances gamma- aminobutyric acid (GABA)-mediated neurotransmission (traditionally used to treat epilepsy and bipolar disorder) (426). Interestingly, sodium valproate (valproic acid) is commonly used as an adjunctive therapy for the treatment of schizophrenia, and a recent 4 week randomized clinical trial demonstrated faster improvement in psychopathology with a

115 combination therapy of valproate and risperidone or olanzapine compared to antipsychotics alone (n=249) (427). There is also some evidence for positive effects on aggression and tardive dyskinesia in schizophrenia, although sample sizes are small (428). Valproic acid also affects ERK and Wnt pathways, which were implicated in our pathway analyses and regulate cell survival and cytoskeletal modifications (429). Stimulation of ERK1/2 has the ability to enhance the transcriptional activity of PPARs and increase glucose metabolism, indicating PPAR agonists might also be potential therapeutic targets (333). Interestingly, 2 of our unique perturbagen hits included PPAR agonists (troglitazone and Genistein) (430).

Troglitazone, part of the TZD family, is an anti-inflammatory and antidiabetic drug developed to treat type 2 diabetes. Troglitazone is a ligand to both PPARα and (more strongly) PPARγ, promotes glucose uptake by increasing two transporters we found decreased in pyramidal neurons in schizophrenia (GLUT1 and GLUT3), and inhibits the pro- inflammatory factor NFKB (implicated in our Enrichr analyses) (200, 431, 432). The TZD family of PPAR agonists present an interesting therapeutic intervention for studies in preclinical models of schizophrenia due to their ability to modulate glucose systems we find abnormal in schizophrenia (433, 434). Pioglitazone, a current FDA approved member of the

TZA family, also stimulates glycolytic systems and can attenuate mitochondrial dysfunction, reverse memory impairment, and decrease the incidence of dementia (199, 202, 204, 434-

438). Thus, pioglitazone is a strong, easily accessible drug candidate to “reverse” pathology and possibly restore cognitive defects related to schizophrenia.

Caveats

Through these approaches, bioinformatics has the potential to offer key insights into our understanding of how specific human diseases or healthy states manifest themselves.

However, the techniques presented here have several limitations. Our pathway analyses are inherently biased as Enrichr is limited by previously published experiments, meaning

116 some systems are more extensively studied and may skew results. Additionally, when selecting “consistently changed” genes as inputs for Enrichr using dendrograms and clustering (Figure 4.5 and Figure 4.6), it is not uncommon for 5/6 of the knockdown signatures to change in the same direction for a gene, while 1/6 KD signatures have no change or a change in the opposite direction. However, since the goal of the study is to analyze an aggregate signature, we selected clusters of genes that largely change together in the same direction. Enrichr merges human, mouse and rat genes, which has advantages and disadvantages (320).

For our chemical perturbagen analyses, two seed genes (HXK1 and PFKL) had the strongest discordant perturbagen signatures (Table 4.5), leading to the overrepresentation of these two seed genes in the overall top 20 chemical perturbagens table (Table 4.6).

Furthermore, PFKFB2 did not have any discordant signatures to contribute. However, although not in the top 20 overall discordant perturbagen signatures, the top hit for PFKM was also valproic acid (albeit a weak discordance), suggesting our overall top hits may still reverse this signature.

iLINCS, although powerful, also has important caveats in itself. Knockdown signatures are generated in a variety of cell lines (mostly cancer) and can have differing transcriptomic changes to the same perturbagen. While the selection of a relevant cell line is important, not every target gene has been knocked down in every cell line, and comparing signatures from different cell lines can yield different results. In this study, none of our seed genes had knockdown signatures in neuronal cell lines. To minimize variability, we compared signatures for seed genes using one consistent cell line (VCAP). While these cells are not neuronal cell lines, downstream regulation of transcriptional profiles is often comparable across cell lines, and still provides a useful substrate for comparison of mRNA signatures. Further, the LINCS database includes 978 so-called “landmark” genes that are

117 deemed as important biological substrates. There may be instances where this panel does not fully capture the changes found in disease states.

Our in silico analyses of metabolic targets in schizophrenia also present challenges.

The SMRI database analysis function performs a meta-analysis across 12 independent studies, combines molecularly and functionally distinct brain regions. It is possible that for some genes, regions such as the cerebellum drive results, while more subtle differences in cortical regions are not reported. Additionally, both the SMRI and the Mount Sinai databases were generated for whole brain regions, and are not cell-subtype specific. It is not surprising that several of our cell-subtype specific findings were not appreciated at the region-level in our studies. It may be up in some cells and down in others, with no net changes. The SMRI database particularly raises concern due to the number of individual studies used for comparison, of which have varying subject demographics. Limitations aside, it will be interesting to probe these datasets (and more publicly accessible databases) for targets in pathways that were implicated in our Enrichr analyses (Table 4.4).

SUMMARY

Important biological questions can be addressed by bioinformatics including understanding biological networks in human disease. Here we have utilized bioinformatic analyses to gain insight into the connectivity of bioenergetic pathology in schizophrenia

(Chapter 3) and signaling systems such as inflammatory and cell growth pathways. Our discovery approaches allowed us to generate candidate drugs with the potential to reverse the connected pathological signature as opposed to a single target. PPAR agonists emerged as promising therapeutic targets for preclinical drug trials in models of schizophrenia.

ACKNOWLEDGEMENTS

118

I would like to thank Dr. Meller and Erica Carey, for their training and guidance in developing the bioinformatic workflow. I would also like to thank Adam, Sinead, Eduard, and

Rob on engaging and learning these techniques with me.

119

CHAPTER 4 TABLES:

Table 4.1. Seed profile for bioinformatic analyses. Target Main finding

GPI decrease PN mRNA

decrease PN mRNA HXK1 decrease HXK DLPFC enzyme activity decrease DLPFC mRNA (PFKM) PFKM and PFKL decrease PN mRNA (PFKM/PFKL) decrease PFK DLPFC enzyme activity increase in lactate in ACC# LDHA increase in lactate in DLPFC*

PFKFB2 genetic linkage and function##

Table 4.1. Seed profile for bioinformatic analyses. Summary of metabolic findings from human postmortem studies (region and cell-level) and literature. # Rowland 2016. ## Stone

2004. * unpublished observation.

120

Table 4.2 Summary of in silico analyses (disease versus control). Human Pyramidal SMRI Target DLPFC Neurons Genomics Mount Sinai mRNA mRNA mRNA 1.24 FC, 1.22 FC, 1.03 FC, MCT1 -1.15 FC# p=0.165 p=0.039 p=0.345 1.03 FC, 1.04 FC, MCT2 NM -1.16 FC# p=0.846 p=0.055 1.07 FC, -1.19 FC, 1.01 FC, -1.31 FC, MCT4 p=0.752 p=0.230 p=0.160 p=0.137 -1.15 FC, -1.11 FC, -1.11 FC, -1.06 FC, LDHA p=0.359 p=0.285 p=0.022 p=0.320 -1.02 FC, 1.01 FC, -1.07 FC, LDHB 1.08 FC# p=0.703 p=0.860 p=0.041 1.10 FC, -1.24 FC, 1.12 FC, -1.01 FC, HXK1 p=0.397 p=0.003 p=0.065 p=0.831 -1.23 FC, -1.02 FC, -1.02 FC, HXK2 ND p=0.323 p=0.589 p=0.267 -1.32 FC, -1.43 FC, 1.05 FC, 1.03 FC, PFKM p=0.039 p=0.0001 p=0.225 p=0.694 -1.27 FC, -1.00 FC, PFKL NM -1.03 FC# p=0.011 p=0.920 -1.09 FC, -1.02 FC, -1.23 FC, PFKP NM p=0.249 p=0.679 p=0.047 -1.04 FC, -1.19 FC, 1.07 FC, -1.03 FC, GLUT1 p=0.776 p=0.009 p=0.029 p=0.790 1.01 FC, -1.19 FC, 1.12 FC, GLUT3 -1.34 FC# p=0.907 p=0.012 p=0.001 -1.26 FC, -1.01 FC, -1.01 FC, GPI NM p=0.015 p=0.659 p=0.878

Table 4.2. Summary of in silico analyses. Summary of metabolic findings from human postmortem studies (region and cell-level) and online databases. Mount Sinai database compares microarray data from the dorsolateral prefrontal cortex in schizophrenia and control subjects. SMRI Genomics database compares microarray and consortium data from schizophrenia and control subjects from 5 postmortem brain regions (BA6, BA8/9, BA10,

BA46, cerebellum). Stanley Medical Research Institute (SMRI) Online Genomics Database; dorsolateral prefrontal cortex (DLPFC); not measured (NM); fold change (FC); not detected

(ND); lactate dehydrogenase A (LDHA), LDHB, hexokinase 1 (HXK1), HXK2, glucose-6-

121 phosphate isomerase (GPI), phosphofructokinase muscle (PFKM), phosphofructokinase liver

(PFKL), monocarboxylate transporter 1 (MCT1), MCT4, glucose transporter 1 (GLUT1), and

GLUT3. All results are presented as diseased state versus control. Red text indicates finding was significant according to p-value cut off of 0.05 and fold change cut off of 1.1. # denotes multiple for probes for the same gene (geomean was used to calculate fold change).

*P<0.05 and FC>1.10.

122

123

Table 4.3. Consistently upregulated and consistently downregulated genes across clustered seed gene knockdown signatures.

124

Table 4.4. Summary of Enrichr analyses. CONSISTENTLY UPREGULATED GENES ANALYSIS Analysis Select Targets Function cell metabolism, inflammation, KEGG Signaling Pathways FOXO, MAPK, Wnt cell growth, mitochondrial function RPS6KA3, MAPK14, GSK3B, metabolism, cell proliferation, Protein-Protein Interactions CDK1, CUL1, SKP1, PCNA, degradation/ubiquitination, HDAC1, ESR1 chromatin remodeling AKT1, MAP2K2, PTK2, ATR, glucose metabolism, cell cycle Kinases NMUR2, FFAR1 regulation, energy homeostasis developmental processes, WT1, TRIM28, NFIB, NUCKS1, Transcription Factors immune response, estrogen TCF3, ESR1, ESR2, ELK3, CEBPB receptors CONSISTENTLY DOWNREGULATED GENES ANALYSIS p53 signaling, HTLV infection, inflammation, impaired KEGG Signaling Pathways hepatitis B immunity, and cell proliferation Chromatin remodeling/HDAC HDAC2, HDAC1, PCNA, BRCA1, Protein-Protein Interactions inhibitors, immunity/cell cycle CREBBP, CSNK2A1, NFKB2, MYC regulation, ubiquitination CNS development, cell cycle PPARγ, RXR, PXK, ABL2, RYK, progression, cytoskeletal Kinases WEE1, LPHN1, NR3C1, GPR37 remodeling carbohydrate/fatty acid metabolism, inflammation MYC, MYCN, NUCKS1, FOXM1, cellular proliferation and cell Transcription Factors E2F1, E2F4, POU5F1, EKLF cycle control, development

Table 4.4. Summary of Enrichr analyses. Hits and biological function for Enrichr analyses of consistently upregulated and consistently downregulated gene groups.

125

Table 4.5. Top 20 discordant chemical perturbagens per seed gene.

126

Table 4.6. Top 20 discordant chemical perturbagens across all seed gene knockdown signatures.

127

Table 4.7. Top 12 unique chemical perturbagens. The top unique chemical perturbagens with the ability to “reverse” the disease-based signature.

128

CHAPTER 4 FIGURES:

Figure 4.1. Summary figure of workflow overview.

129

EMPIRICAL SEED PROFILE -SCZ , MDD, CTL Postmortem Brain -5 Brain Regions -Detailed Demographics LDH HXK DLPFC activity -Transcriptomic ACC lactate HXK1 mRNA in PN Interrogate In PFK Silico Databases DLPFC activity -Di erentiated Cortical Neurons (iPSC) PFKm mRNA in dlpfc -CTL and DISC1 Mutants PFKM mRNA in PN PFKLmRNA in PN -Transcriptomic and Proteomic GPI PFKFB2 GPI mRNA in PN Genetic linkage

-SCZ and CTL Postmortem Brain -4 Brain Regions -Transcriptomic

Generate KD Signatures For Each Seed Gene KD Signatures

PFKL PFKM GPI PFKFB2 LDHA HXK1 GPI LDHA PFKM PFKFB2 HXK1 PFKL

CLUSTER KD SIGNATURES

Top 50 Di erentially Genes Expressed Genes

L1000 Genes Per Signature

Select Concordant Gene Changes Find Consensus Top Discordant Chemical Perturbagen Signatures Downregulated Upregulated

Pioglitazone

Reverse Disease Signature in Animal Model?

PATHWAY ANALYSIS 130 Figure 4.2. Detailed summary figure of workflow. Phosphofructokinase liver type (PFKL);

PFK muscle type (PFKM); G lucose-6-phosphate isomerase (GPI); 6-phosphofructo-2- kinase (PFKFB2); lactate dehydrogenase A (LDHA); hexokinase 1 (HXK1); schizophrenia

(SCZ); major depressive disorder (MDD); control (CTL); Brodmann area (BA); dorsolateral prefrontal cortex (DLPFC).

131

Figure 4.3. Heatmap of seed gene knockdown signatures. Representative heatmaps of transcriptional changes in L1000 genes in seed gene knockdown signatures.

132

Phosphofructokinase liver type (PFKL); PFK muscle type (PFKM); G lucose-6-phosphate isomerase (GPI); 6-phosphofructo-2-kinase (PFKFB2); lactate dehydrogenase A (LDHA); hexokinase 1 (HXK1).

133

Figure 4.4. Clustered heatmap of seed gene knockdown signatures. The clustering of the top 50 differentially expressed genes in each knockdown signature.

Phosphofructokinase liver type (PFKL); PFK muscle type (PFKM); G lucose-6-phosphate

134 isomerase (GPI); 6-phosphofructo-2-kinase (PFKFB2); lactate dehydrogenase A (LDHA); hexokinase 1 (HXK1).

135

Figure 4.5. Selection of upregulated genes. Upregulated genes were selected using the dendrogram clustering function (fold change and p-values were exported to Excel).

Phosphofructokinase liver type (PFKL); PFK muscle type (PFKM); G lucose-6-phosphate isomerase (GPI); 6-phosphofructo-2-kinase (PFKFB2); lactate dehydrogenase A (LDHA); hexokinase 1 (HXK1).

136

Figure 4.6. Selection of downregulated genes. Downregulated genes were selected using the dendrogram clustering function (fold change and p-values were exported to

Excel). Phosphofructokinase liver type (PFKL); PFK muscle type (PFKM); G lucose-6- phosphate isomerase (GPI); 6-phosphofructo-2-kinase (PFKFB2); lactate dehydrogenase A

(LDHA); hexokinase 1 (HXK1).

137

Figure 4.7. KEGG cell signaling pathways for consistently upregulated genes.

138

Figure 4.8. Cellular components for consistently upregulated genes.

139

Figure 4.9. Molecular function for consistently upregulated genes.

140

Figure 4.10. Protein-protein interaction hubs for consistently upregulated genes.

141

Figure 4.11. Kinase analysis for consistently upregulated genes.

142

Figure 4.12. Transcription factor analysis for consistently upregulated genes.

143

Figure 4.13. dbGaP for consistently upregulated genes.

144

Figure 4.14. KEGG cell signaling pathways for consistently downregulated genes.

145

Figure 4.15. Cellular components for consistently downregulated genes.

146

Figure 4.16. Molecular function for consistently downregulated genes.

147

Figure 4.17. Protein-protein interaction hubs for consistently downregulated genes.

148

Figure 4.18. Kinase analysis for consistently downregulated genes.

149

Figure 4.19. Transcription factor analysis for consistently downregulated genes.

150

Figure 4.20. dbGaP for consistently downregulated genes.

151

CHAPTER 5

Bioenergetic deficits and reversal of memory deficits with pioglitazone in the GluN1

model of schizophrenia

Courtney R. Sullivan, BS 1, Catharine Mielnik, PhD 2, Sinead O’Donovan, PhD 1, Adam Funk, PhD 1, Rachael Koene 2, BA, Guillaume Labilloy 3, Jarek Meller, PhD 3, Amy Ramsey, PhD 2, Robert E. McCullumsmith, MD, PhD 1

1Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati

2 Department of Pharmacology and Toxicology, University of Toronto, ON, Canada

3 Division of Biomedical Informatics, Cincinnati Children’s Hospital Medical Center

152

ABSTRACT

The N-methyl-D-aspartate subtype glutamate receptor (NMDAR)-subunit GluN1 knockdown mouse model has well-characterized behavioral endophenotypes for schizophrenia, including social deficits, increased stereotypy, decreased habituation to novel objects, and decreased performance in spatial and working memory tasks (165). Previously we found that metabolic pathways are abnormally regulated in schizophrenia (Chapter 3). Pioglitazone (pio) is a synthetic ligand for peroxisome proliferator-activated receptor gamma (PPARγ), and can alter the transcription and expression of GLUT1 leading to increases in glucose uptake (199). Due to the pro-metabolic nature of pio, we hypothesize that increasing glucose uptake via glucose transporter 1 (GLUT1) will diminish cognitive deficits associated with schizophrenia in GluN1 knockdown mice. To examine the relationship between impaired bioenergetics and endophenotypes in schizophrenia, we performed behavioral tasks examining locomotor activity, anxiety and mania, cognition and executive function, social affiliation, and sensorimotor gating in GluN1 knockdown mice and WT littermate controls on and off of pio. We found that pioglitazone treatment may restore explicit memory in an executive function task.

Further studies are needed to elucidate potential therapeutic effects including alternative dosing regimens of pioglitazone and biochemical studies of bioenergetic pathways in this model.

INTRODUCTION

One of the leading hypotheses of the etiology of schizophrenia is glutamatergic hypofunction (164, 169, 439). N-methyl-D-aspartate subtype glutamate (NMDA) receptor dysfunction is strongly implicated in the pathophysiology of schizophrenia, particularly in human pharmacological studies (155-158). For instance, phencyclidine (PCP) and ketamine are capable of producing psychosis, social withdrawal and working-memory deficits via NMDA receptor blockade (157, 169). Thus, the glutamate synapse has emerged as a prominent target

153 for potential therapeutic intervention in schizophrenia (440). Converging evidence suggests that

NMDA receptor hypofunction may actually reflect a dysregulation at the receptor level, such as mutations in NMDA receptor subunits (441). Several studies have found associations between the gene that encodes the essential GluN1 subunit of the NMDA receptor (Grin1) and schizophrenia (442-450). The GluN1 knockdown (KD) mouse model has a ~90% global reduction of normal functioning NMDA receptors, achieved by targeted insertion of ~2 kb of foreign DNA into intron 19 of the Grin1 gene (166). Additionally, mice with mutations in the

Grin1 gene have well-characterized behavioral endophenotypes for schizophrenia; this includes impairments in executive function and working memory, increased stereotypic behavior, decreased anxiety like behavior, decreased sensorimotor gating, and abnormal social interaction (165). Several of these behaviors are attenuated with typical and atypical antipsychotics at doses similar to those used in pharmacological models of schizophrenia (MK-

801 or PCP models), including hyperactivity, stereotypy, and sociability deficits (166, 451).

Therefore, the GluN1 knockdown mouse model is a viable and useful tool for studying the interface of NMDA receptor hypofunction, bioenergetics, and schizophrenia symptomology.

There is evidence of abnormal bioenergetic pathways in GluN1 KD animals and other models of impaired synapses. For example, there is strong evidence linking lactate shuttle defects to N-methyl-D-aspartate (NMDA) receptor dysfunction in MK-801 treated rats and in

GluN1 subunit knockdown mice (54, 55, 192). Dysregulation of the energetic flow of monocarboxylates inhibits long term potentiation (43, 313), and may contribute to cognitive deficits in GluN1 KD mice. Additionally, when lactate flow from astrocytes to neurons is stimulated, or when lactate is introduced to extracellular space, it enhances working memory and memory formation in rats (43, 49, 313). Thus, selective manipulation of glucose uptake into astrocytes coupled with behavioral studies on GluN1 KD mice would allow further examination of bioenergetic abnormalities and cognitive dysfunction in schizophrenia.

154

Pioglitazone (Pio) is a member of the thiolazinedione (TZD) drug family that is approved to treat type 2 diabetes and hyperglycemia (199). Pio is a synthetic ligand for peroxisome proliferator-activated receptor gamma (PPARγ), a nuclear receptor that is responsible for the regulation of several bioenergetic functions such as lipid homeostasis, adipocyte differentiation, and insulin sensitivity (199, 452). Interestingly, PPARγ agonists have the ability to reduce oxidative stress (via mediating nitric oxide oxygen production) as well as modifying mitochondrial metabolism by interacting with proteins associated with mitochondrial function

(mitoNEET proteins) (452-456). Activation of PPARγ via pio can also alter the transcription and expression of glucose transporter 1 (GLUT1), leading to changes in glucose uptake through

PPARγ and other mechanisms (199, 200). Thus, we hypothesize that treatment with pioglitazone will help restore cognitive endophenotypes associated with schizophrenia in the

GluN1 KD model via stimulation of metabolic pathways.

To inform our hypothesis, we treated animals with pioglitazone and measure various behavioral outputs. First, we assessed locomotor activity, stereotypic behaviors, and anxiety like behavior. Next, we assessed executive function and cognitive flexibility using the puzzle-box test. This test requires the mouse to display problem-solving behavior as well as short- and long-term explicit memory to reach a dark box (GluN1 KD mice have significant delays in time to complete these tasks). To assess pio’s effect on sociability, we used an age- and sex-matched mouse as a social stimulus as previously published (457). Finally, we assessed sensorimotor gating using the pre-pulse inhibition of acoustic startle paradigm (PPI)(458).

MATERIALS AND METHODS

Experimental mouse lines and genotyping

Animal housing and experimentation were carried out in accordance with the

Canadian Council in Animal Care (CCAC) guidelines for the care and use of animals. Mice were group housed with littermates on a 12-h light-dark cycle (0700 to 1900h) and were

155 given access to food (2018 Teklad Global 18% Protein Rodent Diet, Envigo, Madison

Wisconsin USA, www.envigo.com) ad libitum, unless otherwise specified. Mice were tail clipped and had their toes tattooed at P13 (± 3 days) for genotyping and weaned at P21.

Toe tattooing was used to identify all experiment mice.

Grin1flneo/flneo mice were generated in house, based on the previously described

GluN1 KD mouse (166). All experimental animals were of F1 progeny of Grin1flneo/flneo heterozygotes; C57Bl/6 background and 129/SvlmJ background. The floxed insertion mutation (neo) was identified using the following primers: wildtype forward 5’- TGA GGG

GAA GCT CTT CCT GT-3’, mutant forward 5’- GCT TCC TCG TGC TTT ACG GTA T-3’, common reverse 5’-AAG CGA TTA GAC AAC TAA GGG T-3’.

Grin1flneo/flneo heterozygotes (C57Bl/6) were bred to heterozygous Grin1flneo/flneo mice

(129/SvlmJ) resulting in F1 progeny at the expected Mendelian frequency for all proposed studies; 12.5% WT, 12.5%WTCre, 12.5% GluN1 and 12.5% GluN1Cre. Primers mentioned above were used to identify the Grin1 insertion mutation. For preliminary quantitative reverse transcription PCR (RT-qPCR) studies, we had a n of 5 animals per group (10 total).

For our preliminary liquid chromatography–mass spectrometry (LC-MS) study, we had an n of 3 per group (6 total). For our pioglitazone studies, we had a total of 27 animals, 14 WT and 13 GluN1 KD mice (Table 5.2).

RT-qPCR

TaqMan PCR assays for each target gene were performed in duplicate on cDNA samples in 96-well optical plates on a Stratagene MX3000P (Stratagene, La Jolla,

California). All TaqMan PCR data were captured using Sequence Detector Software (SDS version 1.6; PE Applied Biosystems). All primers were obtained from Thermo Fisher

Scientific, Waltham, MA, USA. Targets are as follows: MCT1 (Mm01306379_m1), MCT2

156

(Mm00441442_m1), MCT4 (Mm00446102_m1), GLUT1 (Mm00441480_m1), and GLUT3

(Mm00441483_m1).

PSD-95 affinity purification, mass spectrometry, and pathway analysis

PSD-95 affinity purification

We used a mouse anti-PSD-95 antibody (Millipore, catalogue # MAB1596) to capture PSD-95 protein complexes from samples (3 male WT pooled, 3 male GluN1 KD).

We verified the specificity of this antibody using multiple reaction monitoring mass spectrometry analysis of PSD-95 peptides captured by affinity purification. 5ug of PSD-95 antibody was coupled per 1mg of Dynabeads (Life Technologies) according to the antibody coupling kit protocol (#14311D). For each sample, 1000ul of 10mg/ml antibody coupled beads were washed 2x 1ml with ice cold 1x PBST (#9809S, Cell Signaling) then incubated with mouse brain lysate brought to a final volume of 1000ul with ice cold 1x PBST for 1 hour at room temperature. The supernatant was removed and the beads were washed 4 x 10 minutes at room temperature in 1ml ice cold 1x PBST. Captured protein complexes were eluted with 30ul of 1N Ammonium Hydroxide (#320145, Sigma), 5mM EDTA, pH 12 for 10 minutes at room temperature. 6ul of 6x protein denaturing buffer (4.5% SDS, 15% β- mercaptoethanol, 0.018% bromophenol blue, and 36% glycerol in 170 mM Tris-HCl, pH 6.8) was added to each sample elution. The eluted samples were heated at 70°C for 10 minutes then processed for mass spectrometry.

Mass Spectrometry sample preparation

All samples (3 male WT pooled, 3 male GluN1 KD pooled) were loaded on a 1.5 mm, 4-12% Bis-Tris Invitrogen NuPage gel (NP0335BOX) and electrophoresed in 1x MES buffer (NP0002) for 10 minutes at 180v. The gel was fixed in 50% ethanol/10% acetic acid overnight at RT, then washed in 30% ethanol for 10 min followed by two 10 min washes in

157

MilliQ water (MilliQ Gradient system). The lanes were harvested, cut into small (~2mm) squares, and subjected to in-gel tryptic digestion and peptide recovery. Samples were resuspended in 0.1% formic acid.

Nano liquid chromatography coupled electrospray tandem mass spectrometry (nLC-

ESI-MS/MS).

nLC-ESI-MS/MS analyses were performed on a 5600+ QTOF mass spectrometer

(Sciex, Toronto, On, Canada) interfaced to an Eksigent (Dublin, CA) nanoLC.ultra nanoflow system. Peptides were loaded (via an Eksigent nanoLC.as-2 autosampler) onto an

IntegraFrit Trap Column (outer diameter of 360 µm, inner diameter of 100, and 25 µm packed bed) from New Objective, Inc. (Woburn, MA) at 2 µl/min in formic acid/H2O 0.1/99.9

(v/v) for 15 min to desalt and concentrate the samples. For the chromatographic separation of peptides, the trap-column was switched to align with the analytical column, Acclaim

PepMap100 (inner diameter of 75 µm, length of 15 cm, C18 particle sizes of 3 µm and pore sizes of 100 Å) from Dionex-Thermo Fisher Scientific (Sunnyvale, CA). The peptides were eluted using a variable mobile phase (MP) gradient from 95% phase A (Formic acid/H2O

0.1/99.9, v/v) to 40% phase B (Formic Acid/Acetonitrile 0.1/99.9, v/v) for 70 min, from 40% phase B to 85% phase B for 5 min and then keeping the same mobile phase composition for 5 additional min at 300 nL/min. The nLC effluent was ionized and sprayed into the mass spectrometer using NANOSpray® III Source (Sciex). Ion source gas 1 (GS1), ion source gas 2 (GS2) and curtain gas (CUR) were respectively kept at 8, 0 and 35 vendor specified arbitrary units. The mass spectrometer method was operated in positive ion mode and the interface heater temperature and ion spray voltage were kept at 150ºC, and at 2.6 kV, respectively. The data was recorded using Analyst-TF (version 1.7) software.

Data independent acquisition (DIA)

158

The DIA method was set to go through 1757 cycles for 99 minutes, where each cycle performed one TOF-MS scan type (0.25 sec accumulation time, in a 550.0 to 830.0 m/z window) followed by 56 sequential overlapping windows of 6 Daltons each. Note that the Analyst software automatically adds 1 Dalton to each window to provide overlap, thus an input of 5 Da in the method set up window results in an overlapping 6 Da collection window width (e.g. 550-556, then 555-561, 560-566, etc). Within each window, a charge state of +2, high sensitivity mode, and rolling collision energy with a collision energy spread

(CES) of 15 V was selected.

DIA data analysis parameters

Protalizer DIA software (Vulcan Analytical, Birmingham, AL) was used to analyze every DIA file (Sciex 5600 QTOFs). The Swiss-Prot Homo Sapien database, downloaded

March 17th 2015, was used as the reference database for all MS/MS searches. A precursor and fragment-ion tolerance for QTOF instrumentation was used for the Protalizer Caterpillar spectral-library free identification algorithm. Potential modifications included in the searches were phosphorylation at S, T, and Y residues, N-terminal acetylation, N-terminal loss of ammonia at C residues, and pyroglutamic acid at N-terminal E and Q residues.

Carbamidomethylation of C residues was searched as a fixed modification. The maximum valid protein and peptide expectation score from the X! Tandem search engine used for peptide and protein identification on reconstructed spectra was set to 0.005.

For DIA quantification by Protalizer the maximum number of b and y series fragment-ion transitions were set to nine excluding those with m/z values below 300 and not containing at least 10% of the relative intensity of the strongest fragment-ion assigned to a peptide. A minimum of five fragment-ions were required for a peptide to be quantified

(except where indicated otherwise). In datasets where a minimum of seven consistent fragment-ions were not detected for the same peptide ion in each of the three files

159 compared in a triplicate analysis, the algorithm identified the file with the largest sum fragment-ion AUC and extracted up to seven of these in the other files using normalized retention time coordinates based on peptides detected by the Caterpillar algorithm in all the files in a dataset.

Ingenuity pathway analysis (IPA)

We performed 2 analyses using IPA pathway analysis. To generate the top pathways for WT mice we input the top 20 increased proteins in WT mice relative to GluN1

KD mice from the mass spectrometry experiments into IPA. Next, to generate the top pathways for GluN1 KD mice we input the top 20 increased proteins in GluN1 KD mice relative to WT mice into IPA.

Drug administration

Pioglitazone is an FDA approved drug for the treatment of type II diabetes. It increases the transport of glucose from the bloodstream into tissues by increasing the expression of the glucose transporter GLUT1. We divided mice into 4 groups, WTveh (n=6), WTpio (n=8),

GluN1veh (n=4) and GluN1pio (n=9). On day 1 of the experimental paradigm, mice were given free access to either a 2018 Teklad Global 18% Protein Rodent Diet (Envigo,

Madison Wisconsin USA, www.envigo.com) as control chow or chow infused with pioglitazone at 100 ppm (2018 Teklad with 0.01% pio, BOCSCI, Shirley, NY, USA) for 7 days. Dose was selected based on previous studies (200). On day 8, behavioral tests began

(animals remained on respective diets until sacrificed on day 17). Throughout the experiments, we assessed caloric intake and body weight to ensure pio treatment did not have adverse effects (Figure 5.2).

Behavioral testing

F1 male and female Grin1flneo/Cre mice were used for all behavioral testing, along with

WT littermates as controls. Animals were aged 10-12 weeks. All behavioral tests were

160 completed between 09:00 and 15:00h. Animals were weighed at the beginning of the experiment and on Days 7 and 14. Food was also weighed throughout experiments (Figure

5.2). On Day 7, all mice had their locomotor activity and stereotypy behavior measured in the Open Field Test. On day 9, mice were tested on the Elevated Plus Maze (EPM). Days

10 through 12, mice were tested in the puzzle box assay. Day 15 mice were subjected to the social affiliative paradigm, and finally on day 16, pre-pulse inhibition was tested. A timetable is presented in Table 5.3.

Open Field Test

Locomotor activity and stereotypy were measured using digital activity monitors

(Omnitech Electronics, Columbus, OH, USA) on the first day of behavioral testing (Day 8 of overall experimental paradigm) as previously described (459). Naïve mice were placed in novel Plexiglas arenas (20 x 20 x 45 cm) and their locomotor and stereotypic activity were recorded over a 120-min period in dim light (15-16lux). Activity was tracked via infrared light beam sensors; total distance traveled and stereotypic movements were collected in 5-min bins.

Elevated Plus Maze

Anxiety behavior was assessed on day 9 of testing via the elevated plus maze (460).

The elevated plus maze was composed of 4 opaque-white arms (2 opposite arms closed, 2 opposite arms open), arranged in a plus shape, with an open center. The dimensions were as follows; maze elevation (38.7cm), open arm (L:30.5cm, W:5cm, H:0cm), closed arm

(L:30.5cm, W:5cm, H:15.2cm) and center (5cm x 5cm). The experiment mouse was placed in the center of the maze, and allowed to freely explore the maze for 8 min. in dim light (15-

16lux), while being tracked with an overhead camera. Open and closed arm times were recorded and collected by Biobserve Viewer3 software. The percentage of time spent in the

161 open arms, as compared to closed arms and center time, was calculated and expressed as a percent of time spent in the open arms of the maze.

Puzzle Box Assay

Mice were run on the Puzzle Box Assay on days 10-12 of testing to assess cognition and executive function as previously described (459). Consisting of two compartments, the puzzle box contains a start area (58 x 28 x 27.5 cm) in bright light (250lux), and a goal zone

(14 x 28 x 27.5 cm) in dim light (5lux). The two areas are separated by a black Plexiglas divider, but connected via an underpass large enough for mice to pass through easily. Mice were placed in the start box facing away from the divider, and the time to move to the goal zone (through the underpass, with both hind legs in the goal zone) was manually scored and recorded. Mice were tested over three days, 3 trials/day, with each day consisting of increasingly difficult obstacles present in the underpass connecting the start area and the goal zone. 2 min were given between each trial on a given day, with a max 300 sec allowed for the completion of each trial. The trials (and obstacles) for the puzzle box were as follows

(Figure 5.3, Panel D):

Day 10: T1 (training) open door and unblocked underpass, T2 and T3 (challenge, then learning) doorway closed and underpass open

Day 11: T4 (explicit memory) identical to T3, T5 and T6 (challenge, then learning) underpass filled with bedding (similar to that found in home cage)

Day 12: T7 (explicit memory) identical to T6, T8 and T9 (challenge, then learning) underpass blocked by a removable cardboard plug

Social Affiliative Paradigm

Social affiliative behavior was assessed on day 15 of testing, as previously described (459, 461). Sociability was measured via video recording the motion and

162 exploration of the experimental mouse, tracked via Biobserve Viewer (version 2) software

(center body – reference point). Experimental mice were allowed to explore the open area

(opaque white walls, 62 x 42 x 22 cm) for 10-min. in dim lighting (15-16lux). The area contained two inverted wire cups, one containing a stimulus mouse (‘social’) and the other empty (‘non-social’). Time spent in each zone (3 cm zone around the cup) was recorded via the Biobserve software. Mice used as a social stimulus were novel, wildtype, inbred

C57Bl/6 mice that were age- and sex-matched to the test mouse.

Pre-Pulse Inhibition/Acoustic Startle Reflex

Pre-pulse inhibition of the acoustic startle response was measured on day 16, via

SR-LAB equipment and software from San Diego Instruments (San Diego Instruments, San

Diego, CA, USA). Accelerometers were calibrated to 700±5 mV and output voltages were amplified and analyzed for voltage changes using SR Analysis, and exported as an excel file. Background white noise was maintained at 65dB. PPI was measured in a 30-min test with 80 randomized trials of: (1) 10 trials pulse alone (2) 10 trials pre-pulse alone (for each pre-pulse), (3) 10 trials pre-pulse plus pulse (for each pre-pulse), and (4) 10 trials no pulse.

5 pulse alone trials were performed before and after the 80 trials, totaling 90 trials per run.

The pre-pulse (4dB, 8dB, or 16dB) was presented 100ms prior to the startle pulse (165dB).

The inter-stimulus interval (ISI) was randomized between 5 and 20s. Experimental mice were placed in a cylindrical tube on a platform in a soundproof chamber. Mice were allowed to acclimatize in the chamber and to the background noise for 300s, followed by 5 consecutive pulse alone trials, then by 80 randomized trials (as described above) and then

5 consecutive pulse alone trials. Pre-pulse inhibition was measured as a decrease in the amplitude of startle response to a 100dB acoustic startle pulse, following each pre-pulse

(4dB, 8dB and 16dB).

Harvesting whole brain tissue

163

On day 17, following all behavioral testing, mice were sacrificed via live cervical dislocation. Brains were removed and frozen in ice cold isopentane (sitting on dry ice).

Brains were then put into 5ml Eppendorf tubes and stored at -80°C until further use

(biochemical studies, future directions).

Statistics

All dependent measures were tested for normalcy and homogeneity of variance using D'Agostino & Pearson omnibus normality test. For locomotor activity, stereotypy, vertical activity, EPM, social and social novelty, and PPI/acoustic startle response, we performed 2-way ANOVA with Bonferonni multiple comparison corrections. Each dB in PPI was treated as an individual ANOVA (4 dB, 8 dB, 16 dB).

RESULTS

Our first pilot experiment examined the synaptic composition of GluN1 KD mice using our PSD95 nLC-ESI-MS/MS protocol (n=3/group, pooled). We performed pathway analysis with the Enrichr suite of bioinformatic tools and compared WT to GluN1 KD PSD95 interactomes using the top 20 differentially expressed proteins. WT mice showed pathways relevant for synaptic plasticity, while GluN1 KD analyses yielded proteins related to glucose metabolism and utilization (Table 5.1). We then further examined glucose metabolism in GluN1 KD mice via region-level RT-qPCR (n=5/group). We detected decreases (about 50%) in transcripts for glucose and lactate transporters in the frontal cortex of GluN1 KD mice versus WT mice (Figure

5.1). Thus, we proceeded to modulate glycolytic pathways in GluN1 KD mice with pioglitazone in an attempt to modulate pathological behaviors.

Throughout the pioglitazone experiments, we assessed caloric intake and body weight to ensure pio treatment did not have adverse effects. Pio did not affect the intake of chow or the body weight of the animals (p=0.5096) (Figure 5.2).

164

In the puzzle box assay, we found that pio increased time to perform the task in WT mice in Trial 1 (training) and Trial 6 (learning) (Figure 5.3, Panel B). However, we found that pio treatment decreased (improved) time to perform task in GluN1 mice in Trial 4 (explicit memory)

(Figure 5.3 Panels B and C). This suggests a significant genotype by drug interaction (F (24,

184) = 6.291, p< 0.0001).

In the open field test, we detected a genotype effect on both locomotor activity and stereotypy (p< 0.0001, F (1, 23) = 363.4) (Figure 5.4, Panels B and C). GluN1 mice, when compared to WT littermates, displayed an increase in hyperlocomotion, as quantified by an increase in total distance traveled measured in cm, which was accompanied by a lack of habituation over the two-hour time course and an increase in overall stereotypic movements.

Following pio treatment, GluN1 KD mice did not show an improvement in hyperactivity

(p=0.4752, F (1, 23) = 0.5269), nor lack of habituation, or exaggerated stereotypic movements

(p=0.7737, F (1, 23) = 0.08467) (Figure 5.4, Panels B and C).

In the elevated plus maze, we detected an effect of genotype (p< 0.0001). GluN1 KD animals spent significantly more time in the open arms (s) than did WT mice, regardless of pio/vehicle treatment (Figure 5.5). We did not detect any significant differences in genotype or drug treatment in our social paradigm (Figure 5.6).

Finally, in our PPI experiment, we detected an effect of genotype and genotype by drug interaction at 4 dB. GluN1pio animals had decreased PPI compared to all other groups (Figure

5.7, Panel A). At 8 dB we see an effect of genotype. GluN1 pio mice have significantly lower PPI than either WT group (Figure 5.7, Panel A). At 16 dB we also see an effect of genotype.

GluN1pio mice have significantly lower PPI than do WTpio mice (Figure 5.7, Panel A). We did not detect any differences in the acoustic startle response between any groups, although GluN1

KD animals tended to have higher startle responses (Figure 5.7, Panel B).

DISCUSSION

165

Initially, we affinity purified postsynaptic density 95 (PSD95) protein complexes from

GluN1KD and WT brains. PSD95 is a scaffolding protein for excitatory synapses, and about

95% of PSD95 expression localizes to the postsynaptic density of excitatory synapses making these protein complexes an excellent proxy for synaptic composition. We performed pathway analysis to compare WT to GluN1 KD PSD95 interactomes and found that WT mice showed pathways relevant for synaptic plasticity (as expected), while GluN1 KD analyses yielded proteins related to glucose metabolism and utilization (Table 5.1).These data suggest a profound shift in the composition of the cortical excitatory synaptic proteome in GluN1 KD mice, with apparent increases in neuroenergetic substrates in neurons. Next, we detected decreases

(about 50%) in transcripts for GLUT1, GLUT3, and lactate transporter MCT4 in the frontal cortex of GluN1 KD mice versus WT mice (Figure 5.1). These data suggest diminished capacity to take up glucose as well as lactate, a neuroenergetic substrate that is shuttled from astrocytes to neurons for ATP synthesis. Pioglitazone, a PPARγ agonist, stimulates uptake of glucose through glucose transporters, and could help restore this deficit.

To test this hypothesis, we examined the effects of pioglitazone on behavioral endpoints in the GluN1 KD model. GluN1 KD animals display a wide variety of endophenotypes associated with schizophrenia, including deficits in cognition (165). Our previous studies suggest glycolytic pathways in schizophrenia are abnormal and could contribute to cognitive dysfunction (Chapter 3, (462)). Our pilot data also indicate abnormal metabolic transporter expression and bioenergetic changes at the synapse in GluN1 KD mice (Table 5.1 and Figure

5.1). Thus, pharmacological manipulation of glycolytic pathways via pio could stimulate glucose uptake and metabolism, possibly restoring some of these deficits. We are also able to control for medication treatments, which can be difficult to interpret in human studies. It is important to distinguish significant metabolic changes as underlying features of the illness and not epiphenomena related to treatment factors.

166

We hypothesized that pio treatment would improve executive function in the GluN1 KD model. The puzzle box assay examines executive function and is progressively more difficult, evident in the similarly poor performance by all groups in later trials. However, in all earlier trials

(where GluN1 deficit is more robust), pio tended to increase WT mean latencies while decreasing GluN1 KD mean latencies. Further, our statistical analyses detected a genotype by drug interaction. In Trials 1 and 6, pioglitazone significantly increased the latency to perform the task in WT mice, while in Trial 4, pio significantly decreased the latency to perform the task in

GluN1 KD mice (Figure 5.3, Panels B and C). This suggests that in normal physiology, pioglitazone could have a negative effect on learning, while in a pathological brain with impaired synapses, pio may have restorative effects on explicit memory.

We also examined the effects of pio on non-cognitive behaviors. In line with current literature (166), our data suggest that GluN1 KD animals have increased locomotor activity and stereotypy compared to WT littermate controls, as well as increased mania and anti-anxiety like behavior (increased % time spent in open arms in EPM compared to WT mice) (Figures 5.4 and

5.5). Pio did not have an effect on these behaviors regardless of genotype, suggesting that either glucose metabolism does not play a role in this phenotype of GluN1 KD mice or that this is a neurodevelopmental phenotype that requires earlier intervention.

We did not detect any changes in sociability or social novelty between any groups. All groups choose to spend more time in the social zone (Figure 5.6). Previous studies demonstrate inherent social abnormalities in GluN1 KD mice, suggesting our experiment may be underpowered to detect an effect of sociability (166). We also did not detect any changes in social novelty- however our social novelty data has high variability, which could contribute to lack of significant findings.

Finally, we assessed sensorimotor gating using prepulse inhibition and acoustic startle response. While there were no changes in acoustic startle response across all groups, we detected an effect of genotype on PPI at 4 dB (F (1, 23) = 20.06, p=0.0002), 8 dB (F (1, 23) =

167

15.63, p=0.0006), and 16 dB (F (1, 23) = 9.177, p=0.0060). GluN1 KD mice had generally lower PPI than did WT controls at all dBs. Additionally, at 4 dB, GluN1pio mice had a lower PPI than any other group, suggesting pioglitazone interacts in a negative way with the KD pathophysiology (drug by genotype interaction, F (1, 23) = 6.579, p=0.0173). The 4dB pre-pulse is the lowest level of pre-pulse and thus the “most sensitive” measure of the deficit (vs. 8dB and

16dB, where the deficit is sometimes not seen in a disease state). This could explain why we only detect this negative drug interaction at the 4 dB pre-pulse. Interestingly (although not significant), WTpio mice tended to have a larger PPI versus WTveh mice in all dB tests, while

GluN1 pio mice tended to have a lower PPI versus GluN1veh mice in all dB tests, reinforcing the idea of different genotype by drug interactions. Additional test animals are needed to explore this possibility.

Our findings suggest that pio may selectively exert its effects on specific endophenotypes of schizophrenia, suggesting circuit-specific modulation of neuroplastic substrates. One possibility is that pioglitazone improves the function of brain regions and circuits that are hypometabolic, while having no effect/negative consequences on brain regions that are normal or hypermetabolic. Many behavioral endpoints that are markers of “positive symptoms” of schizophrenia could correspond to “hypermetabolic” circuity, and thus not be amenable to rescue via pioglitazone. We found pio had no effect on behavioral deficits such as hyperactivity, increased stereotypy, and social impairments, which are normalized by typical and atypical antipsychotics in GluN1 KD animals (166, 167, 451, 463-465). However, pio did have a restorative effect on explicit memory (puzzle box assay), while typical antipsychotics do not

(211, 466). Studies examining the effects of antipsychotic treatment on cognition in GluN1 KD animals has not yet been done and is warranted.

Cognition includes several subprocesses, not all of which are uniquely sustained by the frontal cortex, but also by distributed cortical networks including frontal regions which may not be associated with the frontal lobes (467, 468). These processes can include task management,

168 planning, flexibility of behavior and thought, working memory, attention, or others. Thus, poor performance in any of these parameters could reflect a number of possible cognitive mechanisms

(467, 469). For instance, two meta-analyses of cannabis use on cognition in schizophrenia demonstrated cannabis improved certain subtasks for elements of cognition (visual memory, planning, working memory) while having no effect or worsening other tasks (attention, verbal memory, processing speed)(470, 471). It is also possible that pioglitazone may only improve a metabolic subtype of schizophrenia. Cognition and other complex functions disrupted in schizophrenia are often multicomponential, and modulation of these outputs could be treated with personalized combination therapeutics.

Another study examined the effects of pioglitazone as an adjunct treatment to antipsychotic drugs in schizophrenia subjects (n=40) (203). Subjects received risperidone plus either pioglitazone (30 mg/day) or placebo for 8 weeks. Patients in the pioglitazone group showed significantly more improvement in Positive and Negative Syndrome Scale (PANSS) negative subscale scores (p < 0.001) as well as PANSS total scores (p = 0.01) compared with the placebo group, suggesting pio and risperidone work through different circuitry and could be used in conjunction to treat both cognitive/negative and positive symptoms. There is also evidence suggesting pio treatment could help combat glucose–lipid metabolic abnormalities and diabetes that are often exacerbated by antipsychotics (204).These findings suggest the possibility of pioglitazone as an augmentation therapy in reducing difficult to treat symptoms in schizophrenia.

Our findings also justify future pioglitazone studies examining different types of memory.

For example, pioglitazone may independently influence associative learning, spatial or working memory, cognitive flexibility, long-term memory, or recognition memory. Pio is also an FDA approved drug, and further human trials examining pio, antipsychotic treatment, and a larger number of specific cognitive aspects would be useful. This could be achieved through measures such as the Stroop word-color task (selective attention), n back test and backward digit span tests

169

(working memory), Brown-Peterson test (memory capacity), or Wisconsin card sorting (planning and set-shifting) (467, 472). Other indexes of cognition such as the Verbal Comprehension Index

(VCI), the Perceptual Organizational Index (POI), the Working Memory Index (WMI), the Processing

Speed Index (PSI), the Immediate Memory Index (IMI) and the General Memory Index are useful in human cognitive battery (467).

In summary, these data suggest that pio may impact specific elements of cognition in models of schizophrenia. The present work is not without limitations. First, this study has an average of n=6 mice per group, which may not be large enough to fully elucidate subtle behavioral differences. We performed reverse power calculations (α=.05, 1-ß=0.20) and found we need an average of 18 animals per group to detect an effect. As discussed above, cognition is multifaceted and the current study only utilizes one approach examining executive function

(puzzle box assay). Several behavioral tasks such as fear conditioned learning (context versus cue), water mazes (contextual memory), radial arm maze (working memory), and novel object task would be useful in determining more nuanced effects of pioglitazone. Additionally, pioglitazone doses used here may not fully produce therapeutic effects, and future work with varying dosing regimens is warranted. Finally, biochemical experiments in GluN1 KD mice could help elucidate other potential metabolic pathways to target. This would provide the frame work for future intervention studies using promising new pro-metabolic drugs in GluN1 KD mice or of pioglitazone in more models of schizophrenia (i.e. pharmacological, other mutants such as

DISC-1, SHANK3, NRG-1, etc.). This reverse translational approach could also generate targeted questions for future postmortem work.

SUMMARY

We have previously shown that there are functional changes in glycolytic enzymes and expression changes in metabolic pathways in schizophrenia and GluN1 KD animals.

Bioinformatic analyses indicate PPAR agonists may help reverse the glycolytic signature in

170 schizophrenia. In the present work, we investigated whether administration of a pioglitazone could provide insight into the mechanism underlying cognitive deficits in the GluN1 KD model of schizophrenia. We demonstrated that pioglitazone improved explicit memory in

GluN1 KD mice, which provides important insights into the cognitive symptoms in schizophrenia. Much more work is needed to fully elucidate the effects of pro-metabolic drugs on multiple cognitive abilities, such as memory, executive functions, attention, or visuospatial ability, taking into account unique cognitive circuity in the normal and disease state.

ACKNOWLEDGEMENTS

I would first like to thank Rachael Koene and Dr. Adam Funk for generating the qPCR and LCMS data used to guide this study, and Dr. Meller for his bioinformatic analyses. I would like to thank Dr. Robert McCullumsmith for encouraging me to follow my data and allowing me the opportunity to take part in an international collaboration. Also thank you to Dr. Amy Ramsey for hosting me in her laboratory and her generous donation of GluN1 KD and WT mice. Finally, I would like to thank Dr. Catharine Mielnik for her excellent behavioral training and continued support for this project.

171

CHAPTER 5 TABLES:

Table 5.1. Top pathways from nLC-ESI-MS/MS analysis of affinity purified PSD95 protein interactome. For each group, n=3 brains were pooled. The top 20 interacting proteins for postsynaptic density 95 (PSD95) for each group were included in the Enrichr analysis. Liquid chromatography-tandem mass spectrometry (LCMS); Nano liquid chromatography coupled electrospray tandem mass spectrometry (nLC-ESI-MS/MS); n- methyl-D-aspartate receptor (NMDA).

172

Mouse ID GluN1 Cre Cage Age (wks) Sex Rx Dose

NLX598.2 -/- Tg+ RAEF 11.3 M PIO 100ppm

NLX599.5 WT nTg RAEF 11.1 M PIO 100ppm

NLX598.3 -/- nTg RAEG 11.3 F PIO 100ppm

NLX599.2 -/- Tg+ RAEG 11.1 F PIO 100ppm

NLX599.1 WT Tg+ RAEH 11.1 M vehicle 0

NLX599.8 WT nTg RAEH 11.1 M vehicle 0

NLX600.5 -/- Tg+ RAEI 10.4 M vehicle 0

NLX600.7 WT Tg+ RAEI 10.4 M vehicle 0

NLX642.2 -/- Tg+ RAGH 12.3 F PIO 100ppm

NLX642.9 -/- nTg RAGH 12.3 F PIO 100ppm

NLX643.1 -/- nTg RAGH 12.3 F PIO 100ppm

NLX643.6 WT Tg+ RAGH 12.3 F PIO 100ppm

NLX644.1 -/- nTg RAGN 10.7 F vehicle 0

NLX645.2 -/- nTg RAGN 10.7 F vehicle 0

NLX645.6 WT Tg+ RAGN 10.7 F vehicle 0

NLX645.7 WT Tg+ RAGN 10.7 F vehicle 0

NLX644.5 WT nTg RAGM 10.7 M PIO 100ppm

NLX644.9 -/- nTg RAGM 10.7 M PIO 100ppm

NLX645.4 WT nTg RAGM 10.7 M PIO 100ppm

NLX644.2 -/- nTg RAGL.1 10.7 F vehicle 0

NLX644.4 -/- nTg RAGL.1 10.7 F vehicle 0

NLX644.7 WT nTg RAGL.2 10.7 F PIO 100ppm

NLX644.8 -/- nTg RAGL.2 10.7 F PIO 100ppm

NLX647.6 WT Tg+ RAGP.1 10.1 M vehicle 0

NLX647.7 WT nTg RAGP.1 10.1 M vehicle 0

NLX647.9 WT nTg RAGP.2 10.1 M PIO 100ppm

NLX646.2 WT nTg RAGP.2 10.0 M PIO 100ppm

173

Table 5.2. Full demographics for animal cohort. Weeks (wks); wildtype (WT); treatment

(Rx); transgenic (Tg+); nontransgenic (nTg); pioglitazone (PIO); female (F); male (M); parts per million (ppm). WTveh (n=6), WTpio (n=8), GluN1veh (n=4), and GluN1pio (n=9). Total n=27.

174

Table 5.3. Experimental timeline for behavioral assays. Behavioral Assay Endophenotype Day of experiment Hyperactivity and Open Field Test 7 Stereotypy Elevated Plus Maze Anxiety 9

Puzzle Box Assay Executive Function 10-12

Social Paradigm Social Behaviors 15

Prepulse Inhibition Sensorimotor Gating 16

Table 5.3. Behavioral assay timeline. On days 1-6, mice began a diet of either vehicle or pioglitazone chow. On Day 7, all mice had their locomotor activity and stereotypy behavior measured in the Open Field Test. On day 9, mice were tested on the Elevated Plus Maze

(EPM). Days 10 through 12, mice were tested in the puzzle box assay. Day 15 mice were subjected to the social affiliative paradigm, and finally on day 16, pre-pulse inhibition was tested.

175

CHAPTER 5 FIGURES:

Figure 5.1. Metabolic transcripts in GluN1 KD mice. Metabolic transporter transcripts in frontal cortex of GluN1 KD mice (n=5). Monocarboxylate transporter (MCT); Glucose transporter (GLUT); Wildtype (WT). Values were calculated as percent control. Data is % mean ± SEM. *P<0.05.

176

Figure 5.2. Animal weight and food consumption. Average weight (in grams) of groups over 14 days (A). Total amount of food consumed (in grams) by each group over 14 days

(B). Data is % mean ± SEM.

177

Figure 5.3. Puzzle box assay. Latency (s) to complete trials in puzzle box expressed as a time course (significance not denoted) (A). Latency (s) to goal zone in puzzle box in WTveh,

WTpio, GluN1veh, and GluN1pio mice (B). Trial 4 of the puzzle box assay demonstrating a decrease in latency to goal zone in GluN1pio mice when compared to GluN1veh (C). There were 3 trials per day; descriptions of each trial obstacle are shown in panel (D). Open (O); underpass (U); bedding (B); plug (P). Data shown as mean ± SEM. P<0.05.

178

Figure 5.4. Locomotor activity and stereotypy. Total distance (cm) over time (mins) in the open field test (significance not denoted) (A). Total distance (cm) traveled during open field test (B). Stereotypy number for each group in open field test (C). # denotes significantly different from both WTveh and WTpio groups. Data shown as mean ± SEM. P<0.05.

179

Figure 5.5. Elevated plus maze. Time spent in open arm expressed as % time in open arm versus closed arm. # denotes significantly different from both WTveh and WTpio groups.

Data shown as mean ± SEM. P<0.05.

180

Figure 5.6. Social paradigm. Time (s) spent in social or nonsocial zones distance in the social paradigm for WTveh, WTpio, GluN1veh, and GluN1pio groups (A). Social novelty

(time spent in social zone (s) - time spent in nonsocial zone (s)) takes into account that the

“cup” that the social mouse is under is also a novel object to the experimental mouse, and takes out the “object novelty” from the data (B). Data shown as mean ± SEM. P<0.05.

181

Figure 5.7. Prepulse inhibition and acoustic startle response. Sensorimotor gating (pre- pulse inhibition) represented as percent inhibition of acoustic startle response at three pre- pulse tones; 4db, 8db, and 16dB (A). Acoustic startle response to 165dB startle pulse in pre-pulse inhibition test (B). * denotes significantly different from WTveh; ** denotes significantly different from WTpio; # denotes significantly different from WTveh and WTpio;

## denotes significantly different from WTveh, WTpio, and GluN1veh. Data shown as mean

± SEM. P<0.05.

182

CHAPTER 6

Dissertation summary

183

SUMMARY OF DISSERTAION RESEARCH

In the present studies, we investigated bioenergetic pathways underlying cognitive function in the pathophysiology of schizophrenia. Mounting evidence suggests energy metabolism is disrupted in schizophrenia, and many biochemical and morphological abnormalities are present in excitatory-inhibitory circuitry of the dorsolateral prefrontal cortex (DLPFC) which could contribute to cognitive deficits. For example, processes such as excitatory-inhibitory balance (E/I balance) rely on neurons appropriately adapting energy expenditure necessary for optimal function as the environmental conditions evolve during development. During early postnatal development, metabolic demand is high as neurons form these networks via enhanced neurogenesis, migration, synaptogenesis, and dendrite arborization (473). In schizophrenia, there is a decrease in the expression of glutamic acid decarboxylase 67 (GAD67), the enzyme responsible for most gamma-aminobutyric acid

(GABA) synthesis, in layer 3 parvalbumin positive (PV+) interneurons (474, 475).

Additionally, there are ~20% fewer dendritic spines in layer 3 in schizophrenia, suggesting a loss of excitatory inputs (75, 127). This suggests a loss of the finely tuned balance of E/I interplay in these circuits. To further investigate this, we characterized the expression of two chloride channels in the DLPFC and anterior cingulate cortex (ACC) responsible for maintaining the electrochemical gradient gating GABAergic neurotransmission.

In individuals with schizophrenia, there is diminished expression of the calcium binding protein PV in GABAergic neurons in the DLPFC (224, 476, 477). However, the overall density of PV+ neurons is not changed, suggesting a functional impairment in

GABAergic neurons in schizophrenia (224). Gene expression of PV appears to change in an activity-dependent manner, and changes in these chloride channel cotransporters could contribute to this impairment via an imbalance of intracellular chloride levels (478). Here we demonstrate that K-Cl (KCC2) cotransporter protein expression is decreased in the DLPFC,

184 but not the ACC, in schizophrenia. KCC2 is exclusively expressed on neurons, suggesting a neuron and region-specific abnormality in the GABAergic system (246-249). Lower KCC2 protein levels could diminish the extrusion of chloride from GABAergic target neurons, resulting in GABA channels on the postsynaptic membrane allowing passive outflow of Cl- and an excitatory effect upon GABA binding (86, 244, 245, 250). Sustained periods of high energy demand during development can also become periods of enhanced vulnerability if lesions occur in metabolic pathways, contributing to pathologically imbalanced neurotransmission. There is evidence that decreases in GABAergic inhibitory tone in schizophrenia might reflect a decrease in glucose utilization (90). Furthermore, fast-spiking interneurons require a substantial fraction of glucose making them vulnerable to bioenergetic insults (89, 315). These findings could have implications for the bioenergetic profile of both GABAergic interneurons and excitatory pyramidal neurons.

The lactate shuttle is a key bioenergetic pathway that constitutively supports synaptic transmission at excitatory synapses. Neuronal activation increases the concentration of glutamate in the synapse, activates glycolysis in glial cells, and generates lactate which is transported out of astrocytes and into neurons via monocarboxylate transporters (MCTs)(18, 22, 23). This coupling mechanism between neuronal activity and astrocyte lactate production is essential for working memory performance and long-term memory formation, which are impaired if the lactate shuttle is “broken” (43, 47, 49, 50).

Thus, we next investigated the lactate shuttle and glycolytic pathways in pyramidal neurons and supporting glial cells in the DLPFC. Our novel data suggest glycolytic pathways and energetic substrates are abnormally regulated in DLPFC pyramidal neurons in severe mental illness. We report decreased mRNA expression of four glycolytic enzymes

(hexokinase 1 (HXK1), phosphokinase muscle (PFKM), phosphokinase liver (PFKL), and glucose-6-phosphate isomerase (GPI)), two glucose transporters (glucose transporter 1

185

(GLUT1) and GLUT3), and an increase in lactate/pyruvate transporter MCT1 mRNA expression in schizophrenia. In addition to the possibility of inefficient GABAergic transmission due to high levels of intracellular chloride, pyramidal neurons expressing low levels of glycolytic enzymes and glucose transporters may not contain adequate amounts of intracellular bioenergetic substrates for transport into mitochondria, ultimately impacting mitochondrial function and energy supply. We were unable to assess interneurons for changes in glycolytic targets due to the technical limitations of our approach.

Supporting this notion, several studies have reported mitochondrial dysfunction in the pathophysiology of schizophrenia (63, 64). This includes decreases in specific activity of mitochondrial respiratory chain enzymes in the frontal cortex (66, 67). Additionally, peripheral markers have been linked to mitochondrial complex I dysfunction and correlate with specific aspects of cognitive function in subjects with schizophrenia (110, 111). Two groups also demonstrate cell-subtype specific decreases in mitochondrial related genes in dentate granule neurons and layer 3 and 5 pyramidal neurons in the DLPFC (71-73). These data support the hypothesis that neurons are unable to generate a sufficient amount of bioenergetic substrates in schizophrenia.

Psychiatric disorders such as schizophrenia, bipolar depression (BPD), and major depressive disorder (MDD) often display significant overlap in their pathophysiology and genetic risk factors, including metabolic defects (479). For instance, positron emission tomography (PET) measurements from schizophrenia and MDD patients show reduced glucose metabolism in the caudate nucleus, ACC, and prefrontal cortex in resting and stimulated states (480, 481). MDD also displays decreased mitochondrial enzyme activity in glutamatergic neurons and impaired energy metabolism compared to controls (482, 483).

Similarly, medication naïve BPD patients have elevated levels of lactate in the prefrontal cortex, as well as mutations in mitochondrial DNA (484, 485). Similarity in energy

186 metabolism dysfunction in psychiatric disorders has been previously reviewed (486).

However, here we report decreases in the enzyme activity of HXK and PFK in the DLPFC of schizophrenia, but not MDD, suggesting disease specific deficits in glycolysis at the region- level in this illness. In keeping with this idea, PET studies have converged on

“hypofrontality” and impaired glucose metabolism in the frontal cortex in schizophrenia (487,

488). Taken together, our work showing dysregulation of glycolytic pathways in schizophrenia is a novel finding in this field, and provides the framework for examining the relationship between bioenergetic abnormalities in schizophrenia and cognition.

As discussed above, the heterogeneity of schizophrenia and overlap with other neuropsychiatric disorders necessitate the replication of these findings in additional cohorts.

We probed 2 independent databases of schizophrenia and control subjects for glycolytic targets and found several metabolic enzymes and glucose transporters were dysregulated.

In silico analyses such as these are powerful and efficient means of confirming pathological pathways or consistent themes in specific diseases in large datasets, and often permit the researcher to control for variables such as medication. It would be valuable to generate cell- subtype specific databases and probe for our targets in additional disease states.

Bioinformatic tools also integrate many resources for an unprecedented potential for new discoveries in systems biology (318). For example, we used the Library of Integrated

Network-based Cellular Signatures web portal (iLINCS) to generate a “disease signature” based on our findings of abnormally expressed glycolytic genes in schizophrenia. iLINCS generated a transcriptomic profile containing the downstream expression data of 978 landmark genes when glycolytic enzymes are knocked down in a cell line. By analyzing the changes in the 978 landmark genes across the entire disease meta-signature, we are able to investigate the biological underpinning of gene networks that are most affected by the pathology. This represents an important step in understanding the relationship between

187 cognitive dysfunction and biological systems connected to metabolic pathology in schizophrenia. We used Enrichr to perform enrichment analyses on gene groups that were consistently affected across the disease meta-signature. Enrichr analyses are useful in associating a collection of genes with a biological function, analyzing known protein-protein interaction networks and signaling pathways, and computationally predicting targets of kinases and transcription factors (320). Our Enrichr results identified targets and pathways reflecting cell cycle regulation, cellular metabolism, inflammation, and immunity as associated with our disease signature. These enrichment analyses can help build a connectivity map for glycolytic abnormalities in schizophrenia, which can in turn provide the framework for whole system based treatment strategies as opposed to targeted approaches. For example, a long standing hypothesis supports a role for inflammation and the immune system in the pathogenesis of schizophrenia (412, 413). There is evidence that mild systemic inflammation alters glucose metabolism in the medial temporal lobe and produces impairments in spatial memory in humans (489). This suggests inflammatory processes and cell metabolism could both be contributors to cognitive dysfunction in schizophrenia, and drugs targeting these pathways could be therapeutic (490).

There remain significant challenges in developing high efficacy therapeutics for cognitive domains in schizophrenia. Severity of deficits in attention, executive function, and memory often vary widely among individuals with schizophrenia. Antipsychotics work globally in the brain by targeting dopamine receptors (and sometimes serotonin receptors), and generally lack the ability to treat cognitive decline (211, 212). Modeling pathological changes from multiple studies (a disease meta-signature) via bioinformatic analyses could help identify novel pharmacological interventions with the ability to restore multiple facets of the illness. Therefore, we built a disease-based meta-signature using our postmortem glycolytic findings in conjunction with published studies describing abnormal metabolic

188 targets in schizophrenia and probed iLINCS for chemical perturbagens that produce an inverse signature. These perturbagens are predicted to have the ability to “reverse” not only the glycolytic genes used to generate the signature, but also many of the associated transcript abnormalities. Our bioinformatics analyses identified PPAR agonists such as pioglitazone as possible therapeutic interventions. Activation of PPARs induce anti- inflammatory and insulin sensitizing responses, which were Enrichr pathways associated with our glycolytic disease signature (491, 492). PPAR agonists also stimulate glycolytic systems and can attenuate mitochondrial dysfunction (200, 453). There has been limited work done examining PPAR agonists and cognition. However, there is evidence that pioglitazone and other members of the thiazolidinedione (TZD) family of PPAR agonists decrease the incidence of dementia, reverse cognitive impairments in animal models of

Alzheimer’s disease, and improve negative symptoms in schizophrenia (202-204, 434-437).

These studies demonstrate the ability of bioinformatic analyses to identify promising novel pharmacological interventions.

More work is needed examining pioglitazone’s (and other PPAR agonists) effect on cognition in schizophrenia. An important goal of this study is to characterize bioenergetic pathways that underlie cognition in schizophrenia and restore abnormalities in these systems. We next utilized an animal model of schizophrenia to modulate “broken” metabolic pathways and the disease meta-signature using pioglitazone, with the goal of ultimately restoring cognitive function. The GluN1 knockdown (KD) model has various endophenotypes of schizophrenia (including cognitive dysfunction) as well as metabolic abnormalities (Chapter 5). Interestingly, pioglitazone selectively restored explicit memory in the puzzle box assay in GluN1 KD mice, while it did not have an effect on sociability or anxiety-like behaviors. Notably, pioglitazone worsened learning and memory in wildtype

(WT) mice in puzzle box trials. This suggests that PPAR agonist administration may affect

189 intact neurocircuitry negatively, while in a pathological state with impaired metabolism pioglitazone “normalizes” broken circuits and restores explicit memory. Prefrontal- hippocampal interactions underlie explicit memory, and functional neuroimaging studies have shown failure to recruit the hippocampus during episodic memory retrieval in schizophrenia (493, 494). Similarly, there is a failure to selectively raise cerebral blood flow in areas of the frontal and temporal lobes during certain neuropsychological tasks (495). It is possible that GluN1 KD animals have a comparable defect in specific cortico-hippocampal circuits, and treatment with pioglitazone augments these regions in KD mice while causing an overactive system in WT mice. Experiments on several different cognitive domains in

GluN1 KD mice would also be informative. If the pathology in GluN1 KD mice affects cortex- based cognitive processes differentially, then it could provide evidence for abnormalities in selective cortico-hippocampal circuits, rather than an overall impairment of hippocampal function. In exploring the effects of pioglitazone in schizophrenia, the heterogeneity of the patient population highlights the importance of carefully evaluating individual patients, noting specific cognitive deficits and related circuity in each case, as pioglitazone may affect patient subpopulations differently. For example, pioglitazone differentially affected PANSS depression scores in U.S. and Chinese schizophrenia populations (204).

Collectively, our results suggest that GABAergic neurotransmission and glycolytic pathways in pyramidal neurons in the DLPFC are abnormal in schizophrenia, which was confirmed by replication studies in multiple databases. While many groups have examined metabolic pathways in schizophrenia, this is the first evidence for cell-subtype specific dysregulation of glycolytic enzymes/transporters in schizophrenia. Our studies also demonstrate the power of bioinformatic analyses through disease modeling and the identification of candidate drugs to reverse pathological systems. We validated this technique by giving a preclinical model of schizophrenia a bioinformatic candidate drug.

190

Following administration of pioglitazone, we found circuit-specific modulation of neuroplastic substrates and an improvement in a specific type of memory (explicit memory). It is important to consider if metabolic deficits proceed or are subsequent to synaptic dysfunction in schizophrenia. For instance, in the GluN1 KD model of genetically “broken” glutamate synapses metabolic abnormalities are secondary. Teasing apart primary versus secondary causes poses a challenge not easily overcome, particularly between intertwined entities such as metabolic and synaptic systems; however, our animal studies demonstrate the promise for metabolic treatment strategies regardless of developmental timing. Further work is needed to address how pathological brain circuits and metabolic disturbances interface with cognition in schizophrenia.

FUTURE DIRECTIONS

Our work provides important insights into bioenergetic abnormalities in schizophrenia by employing cell-specific techniques in postmortem brain, discovery based bioinformatics, and behavioral analyses following intervention in an animal model of schizophrenia. In order to fully understand abnormal brain metabolism in schizophrenia, and how these defects affect cognition, additional work is necessary. Our work characterizing chloride channels gating GABAergic neurotransmission combined with our novel laser capture microdissection studies demonstrate bioenergetic abnormalities in the DLPFC in schizophrenia. An initial step in building upon these findings is extending our studies to other brain regions and other cell types such as interneurons, hippocampal neurons, oligodendrocytes, and microglia. For instance, laser capture microdissection (LCM) of interneurons in the DLPFC could further elucidate the relationship between abnormal chloride channel expression and GABAergic function. Additionally, recent imaging studies show microglial activation and altered antioxidant status in the brain of schizophrenia patients (496). Evidence suggests that microglia increase aerobic glycolysis when activated

191 by various stimuli, and abnormal mitochondrial function and glucose availability in disease states could influence microglia function (497). It could be useful to first perform single cell

RNA sequencing for several cell types and compare schizophrenia and control transcriptomes to inform more targeted experiments. Currently, we do not have protein data at the cell-level, and combining LCM with protein quantification methods that require minimal sample input would strengthen our gene expression results.

The next step is to extend both our in silico and iLINCS bioinformatic analyses. It is important to generate new cell-level datasets that can be publically accessed. In addition to more cell-level experiments, we can probe existing schizophrenia databases for enriched expression of genes specific to a cell type (i.e. astrocytes, microglia, and interneurons). We can also apply Enrichr pathway analyses to these datasets to determine biological functions relevant to transcriptomic/proteomic changes in schizophrenia. Similarly, we can go back and probe in silico databases for target hits from the Enrichr analyses of our disease meta- signature. For instance, we could examine the expression of kinases or transcription factors connected to our disease based signature in postmortem schizophrenia datasets. This could help elucidate the mechanistic relationship between the disease based signature and other biological systems. Our in silico analyses could also be extended to examine the effects of sex, drug use, medication, and suicide on the expression of our targets. These remain difficult factors to control for in individual studies and contribute to data variability, thus large consortium datasets offer a way to circumvent this issue.

There are several ways to expand our iLINCS analyses. Similar to building a schizophrenia based signature, we could measure additional glycolytic transcripts in GluN1

KD animals and generate a GluN1 KD signature. This signature could be used to query the

LINCS database according to our bioinformatic workflow (Figure 4.2). Furthermore, there are additional user features in iLINCS allowing users to upload an L1000 transcriptomic

192 signature. We could perform RNAseq on WT and GluN1 KD mice and upload the transcriptomic changes when GluN1 is knocked down. Coupled with Enrichr analyses, this would allow us to identify connected pathways (and possible transcriptional mechanisms) to

GluN1 KD pathology, as well as candidate drugs that may reverse phenotypic deficits.

Finally, pioglitazone (and other PPAR agonists) is available in the LINCS chemical perturbagen database, and it would be interesting to query this chemical perturbagen signature against a GluN1 KD signature as well as our schizophrenia disease based signature (a reverse workflow strategy).

Additionally, in order to fully understand the restorative effects of pioglitazone in specific types of memory in GluN1 KD animals, more work is necessary. First, we could measure the expression of glucose transporter 1 (GLUT1) in pioglitazone treated animals, which is the proposed mechanism of action for increasing glucose uptake. This will help determine if pioglitazone increases GLUT1 expression differentially in WT and GluN1 KD animals. Next, we could utilize anatomical and modeling techniques to detect markers of synaptic plasticity in cortico-hippocampal circuits. This would help elucidate the relationship between the modulation of glycolytic systems with pioglitazone and cognitive behaviors mapped to synaptic plasticity. Furthermore, we believe a larger battery of behavioral tests would be useful in determining the nuances of pioglitazone’s effects on cognitive function.

Specifically, further investigation into cognitive deficits would be informative since there exist a number of phenotypic cognitive subdomains that can be examined in animal models. For instance, spatial and working memory function in rodents can be measured using the Y maze spontaneous alternation task, as well as variants of the Morris water maze and radial arm maze task. Paradigms such as the novel object recognition could be used to measure recognition memory. Probing multiple types of learning and memory that map to specific neurocircuitry could lead to targeted biochemical experiments after PPAR

193 agonist administration in a pathological state. Other drugs, such as those identified in our concordance analyses (i.e. HDAC inhibitors, PI3K inhibitors), could also be evaluated in animal models.

Finally, the pioglitazone dose in the present study was chosen based on prior cell culture studies (200). It is possible that the selected dose (100 ppm) and treatment regimen

(1 week prior to behavioral battery, continuing throughout experiments) were not substantial enough to produce measurable effects in certain behavioral measures. It is also possible that pioglitazone intervention is not efficacious in the adult brain, and an earlier intervention is necessary to produce more robust results. Future experiments using dose-response testing with pioglitazone (including method of delivery), as well as varying developmental intervention points, will help determine the role of glucose metabolism in GluN1 KD cognitive dysfunction. Additional animals (male and female) should also be added to increase power, and possible sex effects should be considered.

In summary, we have taken postmortem observations, confirmed our hypotheses in silico, applied discovery based bioinformatic analyses, and performed reverse translational drug trails in an animal model of schizophrenia. This workflow can be widely applied and demonstrates the application for iLINCS as a cross-translational tool in studying human disease.

194

REFERENCES

1. Association AP. Diagnostic and Statistical Manual of Mental Disorders. Fourth, Text Revision ed. Washington, D.C.: American Psychiatric Association; 2000. 2. Buchanan RW, Carpenter WT. Schizophrenia: Introduction and overview. In: Sadock BJ, Sadock VA, editors. Comprehensive Textbook of Psychiatry. 1. Philadelphia: Lippincott, Williams, and Wilkins; 2000. p. 1096-110. 3. Fleischhacker W. Negative symptoms in patients with schizophrenia with special reference to the primary versus secondary distinction. L'Encephale. 2000;26 Spec No 1:12-4. 4. Zanello A, Curtis L, Badan Ba M, Merlo MC. Working memory impairments in first-episode psychosis and chronic schizophrenia. Psychiatry research. 2009;165(1-2):10-8. 5. Potkin SG, Turner JA, Brown GG, McCarthy G, Greve DN, Glover GH, et al. Working memory and DLPFC inefficiency in schizophrenia: the FBIRN study. Schizophrenia bulletin. 2009;35(1):19-31. 6. Wobrock T, Schneider M, Kadovic D, Schneider-Axmann T, Ecker UK, Retz W, et al. Reduced cortical inhibition in first-episode schizophrenia. Schizophrenia research. 2008;105(1-3):252-61. 7. van Os J, Kenis G, Rutten BPF. The environment and schizophrenia. Nature. 2010;468:203. 8. Tsuang M. Schizophrenia: genes and environment. Biological psychiatry. 2000;47(3):210-20. 9. McGuffin P, Owen M, Farmer A. Genetic basis of schizophrenia. The Lancet. 1995;346(8976):678-82. 10. Schizophrenia Working Group of the Psychiatric Genomics C. Biological insights from 108 schizophrenia-associated genetic loci. Nature. 2014;511(7510):421-7. 11. Fromer M, Pocklington AJ, Kavanagh DH, Williams HJ, Dwyer S, Gormley P, et al. De novo mutations in schizophrenia implicate synaptic networks. Nature. 2014;506(7487):179-84. 12. Kirov G, Pocklington AJ, Holmans P, Ivanov D, Ikeda M, Ruderfer D, et al. De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia. Mol Psychiatry. 2012;17(2):142-53. 13. Owen MJ, Craddock N, O'Donovan MC. Suggestion of roles for both common and rare risk variants in genome-wide studies of schizophrenia. Archives of general psychiatry. 2010;67(7):667-73. 14. Pellerin L, Magistretti PJ. Glutamate uptake into astrocytes stimulates aerobic glycolysis: a mechanism coupling neuronal activity to glucose utilization. Proceedings of the National Academy of Sciences of the United States of America. 1994;91(22):10625-9. 15. Chatton JY, Marquet P, Magistretti PJ. A quantitative analysis of L-glutamate-regulated Na+ dynamics in mouse cortical astrocytes: implications for cellular bioenergetics. The European journal of neuroscience. 2000;12(11):3843-53. 16. Loaiza A, Porras OH, Barros LF. Glutamate triggers rapid glucose transport stimulation in astrocytes as evidenced by real-time confocal microscopy. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2003;23(19):7337-42. 17. Balcar VJ, Johnston GAR. THE STRUCTURAL SPECIFICITY OF THE HIGH AFFINITY UPTAKE OF l- GLUTAMATE AND l-ASPARTATE BY RAT BRAIN SLICES. Journal of neurochemistry. 1972;19(11):2657-66. 18. Chih CP, Roberts Jr EL. Energy substrates for neurons during neural activity: a critical review of the astrocyte-neuron lactate shuttle hypothesis. Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism. 2003;23(11):1263-81. 19. Pellerin L, Pellegri G, Bittar PG, Charnay Y, Bouras C, Martin JL, et al. Evidence supporting the existence of an activity-dependent astrocyte-neuron lactate shuttle. Developmental neuroscience. 1998;20(4-5):291-9.

195

20. Nagase M, Takahashi Y, Watabe AM, Kubo Y, Kato F. On-site energy supply at synapses through monocarboxylate transporters maintains excitatory synaptic transmission. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2014;34(7):2605-17. 21. Pellerin L, Bouzier-Sore AK, Aubert A, Serres S, Merle M, Costalat R, et al. Activity-dependent regulation of energy metabolism by astrocytes: an update. Glia. 2007;55(12):1251-62. 22. Magistretti PJ, Chatton J-Y. Relationship between L-glutamate-regulated intracellular Na+ dynamics and ATP hydrolysis in astrocytes. Journal of Neural Transmission. 2005;112(1):77-85. 23. Rouach N, Koulakoff A, Abudara V, Willecke K, Giaume C. Astroglial metabolic networks sustain hippocampal synaptic transmission. Science (New York, NY). 2008;322(5907):1551-5. 24. Weber B, Barros LF. The Astrocyte: Powerhouse and Recycling Center. Cold Spring Harbor Perspectives in Biology. 2015;7(12). 25. Schurr A. Lactate: The Ultimate Cerebral Oxidative Energy Substrate? Journal of Cerebral Blood Flow & Metabolism. 2005;26(1):142-52. 26. Magistretti PJ, Allaman I. A cellular perspective on brain energy metabolism and functional imaging. Neuron. 2015;86(4):883-901. 27. Gold PE. The many faces of amnesia. Learning & memory (Cold Spring Harbor, NY). 2006;13(5):506-14. 28. Benton D, Owens DS, Parker PY. Blood glucose influences memory and attention in young adults. Neuropsychologia. 1994;32(5):595-607. 29. Benton D, Owens DS. Blood glucose and human memory. Psychopharmacology. 1993;113(1):83- 8. 30. Gold PE. Glucose and age-related changes in memory. Neurobiology of aging. 2005;26 Suppl 1:60-4. 31. Messier C. Glucose improvement of memory: a review. European journal of pharmacology. 2004;490(1–3):33-57. 32. Pfeuffer J, Tkac I, Gruetter R. Extracellular-intracellular distribution of glucose and lactate in the rat brain assessed noninvasively by diffusion-weighted 1H nuclear magnetic resonance spectroscopy in vivo. Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism. 2000;20(4):736-46. 33. Sokoloff L. The metabolism of the central nervous system in vivo. In: J. Field HWM, V.E. Hall (Eds.), editor. Handbook of Physiology. 3. American Physiological Society, Washington D.C. 1960. p. 1843–64. 34. Hall CN, Klein-Flugge MC, Howarth C, Attwell D. Oxidative phosphorylation, not glycolysis, powers presynaptic and postsynaptic mechanisms underlying brain information processing. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2012;32(26):8940-51. 35. Sada N, Lee S, Katsu T, Otsuki T, Inoue T. Targeting LDH enzymes with a stiripentol analog to treat epilepsy. Science (New York, NY). 2015;347(6228):1362-7. 36. Magistretti PJ, Pellerin L, Rothman DL, Shulman RG. Energy on demand. Science (New York, NY). 1999;283(5401):496-7. 37. Allen NJ, Karadottir R, Attwell D. A preferential role for glycolysis in preventing the anoxic depolarization of rat hippocampal area CA1 pyramidal cells. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2005;25(4):848-59. 38. Gruetter R. Glycogen: the forgotten cerebral energy store. Journal of neuroscience research. 2003;74(2):179-83. 39. Knull HR. Association of glycolytic enzymes with particulate fractions from nerve endings. Biochimica et Biophysica Acta (BBA) - Enzymology. 1978;522(1):1-9.

196

40. Herrero-Mendez A, Almeida A, Fernandez E, Maestre C, Moncada S, Bolanos JP. The bioenergetic and antioxidant status of neurons is controlled by continuous degradation of a key glycolytic enzyme by APC/C-Cdh1. Nature cell biology. 2009;11(6):747-52. 41. Almeida A, Almeida J, Bolaños JP, Moncada S. Different responses of astrocytes and neurons to nitric oxide: the role of glycolytically generated ATP in astrocyte protection. Proc Natl Acad Sci U S A. 2001;98. 42. Almeida A, Moncada S, Bolanos JP. Nitric oxide switches on glycolysis through the AMP protein kinase and 6-phosphofructo-2-kinase pathway. Nature cell biology. 2004;6(1):45-51. 43. Suzuki A, Stern SA, Bozdagi O, Huntley GW, Walker RH, Magistretti PJ, et al. Astrocyte-neuron lactate transport is required for long-term memory formation. Cell. 2011;144(5):810-23. 44. Schurr A, Payne RS. Lactate, not pyruvate, is neuronal aerobic glycolysis end product: an in vitro electrophysiological study. Neuroscience. 2007;147(3):613-9. 45. Schurr A, Gozal E. Aerobic production and utilization of lactate satisfy increased energy demands upon neuronal activation in hippocampal slices and provide neuroprotection against oxidative stress. Frontiers in pharmacology. 2011;2:96. 46. Lin AL, Fox PT, Hardies J, Duong TQ, Gao JH. Nonlinear coupling between cerebral blood flow, oxygen consumption, and ATP production in human visual cortex. Proceedings of the National Academy of Sciences of the United States of America. 2010;107(18):8446-51. 47. Jolivet R, Allaman I, Pellerin L, Magistretti PJ, Weber B. Comment on recent modeling studies of astrocyte-neuron metabolic interactions. Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism. 2010;30(12):1982-6. 48. Belanger M, Allaman I, Magistretti PJ. Brain energy metabolism: focus on astrocyte-neuron metabolic cooperation. Cell metabolism. 2011;14(6):724-38. 49. Newman LA, Korol DL, Gold PE. Lactate produced by glycogenolysis in astrocytes regulates memory processing. PLoS One. 2011;6(12):e28427. 50. Mangia S, DiNuzzo M, Giove F, Carruthers A, Simpson IA, Vannucci SJ. Response to 'comment on recent modeling studies of astrocyte-neuron metabolic interactions': much ado about nothing. Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism. 2011;31(6):1346-53. 51. Bubber P, Hartounian V, Gibson GE, Blass JP. Abnormalities in the tricarboxylic acid (TCA) cycle in the brains of schizophrenia patients. Eur Neuropsychopharmacol. 2011;21(3):254-60. 52. Kung L, Roberts RC. Mitochondrial pathology in human schizophrenic striatum: a postmortem ultrastructural study. Synapse (New York, NY). 1999;31(1):67-75. 53. Du F, Cooper AJ, Thida T, Sehovic S, Lukas SE, Cohen BM, et al. In vivo evidence for cerebral bioenergetic abnormalities in schizophrenia measured using 31P magnetization transfer spectroscopy. JAMA psychiatry. 2014;71(1):19-27. 54. Zhou K, Yang Y, Gao L, He G, Li W, Tang K, et al. NMDA receptor hypofunction induces dysfunctions of energy metabolism and semaphorin signaling in rats: a synaptic proteome study. Schizophrenia bulletin. 2012;38(3):579-91. 55. Sun L, Li J, Zhou K, Zhang M, Yang J, Li Y, et al. Metabolomic analysis reveals metabolic disturbance in the cortex and hippocampus of subchronic MK-801 treated rats. PloS one. 2013;8(4):e60598. 56. Regenold WT, Phatak P, Marano CM, Sassan A, Conley RR, Kling MA. Elevated cerebrospinal fluid lactate concentrations in patients with bipolar disorder and schizophrenia: implications for the mitochondrial dysfunction hypothesis. Biological psychiatry. 2009;65(6):489-94. 57. Martins-de-Souza D, Gattaz WF, Schmitt A, Novello JC, Marangoni S, Turck CW, et al. Proteome analysis of schizophrenia patients Wernicke's area reveals an energy metabolism dysregulation. BMC Psychiatry. 2009;9:17.

197

58. Prabakaran S, Swatton JE, Ryan MM, Huffaker SJ, Huang JT, Griffin JL, et al. Mitochondrial dysfunction in schizophrenia: evidence for compromised brain metabolism and oxidative stress. Mol Psychiatry. 2004;9(7):684-97, 43. 59. Pennington K, Beasley CL, Dicker P, Fagan A, English J, Pariante CM, et al. Prominent synaptic and metabolic abnormalities revealed by proteomic analysis of the dorsolateral prefrontal cortex in schizophrenia and bipolar disorder. Mol Psychiatry. 2008;13(12):1102-17. 60. Beasley CL, Pennington K, Behan A, Wait R, Dunn MJ, Cotter D. Proteomic analysis of the anterior cingulate cortex in the major psychiatric disorders: Evidence for disease-associated changes. Proteomics. 2006;6(11):3414-25. 61. Middleton FA, Mirnics K, Pierri JN, Lewis DA, Levitt P. Gene expression profiling reveals alterations of specific metabolic pathways in schizophrenia. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2002;22(7):2718-29. 62. Beasley CL, Dwork AJ, Rosoklija G, Mann JJ, Mancevski B, Jakovski Z, et al. Metabolic abnormalities in fronto-striatal-thalamic white matter tracts in schizophrenia. Schizophrenia research. 2009;109(1-3):159-66. 63. Ben-Shachar D. Mitochondrial dysfunction in schizophrenia: a possible linkage to dopamine. Journal of neurochemistry. 2002;83(6):1241-51. 64. Ben-Shachar D, Laifenfeld D. Mitochondria, synaptic plasticity, and schizophrenia. International review of neurobiology. 2004;59:273-96. 65. Stone WS, Faraone SV, Su J, Tarbox SI, Van Eerdewegh P, Tsuang MT. Evidence for linkage between regulatory enzymes in glycolysis and schizophrenia in a multiplex sample. American journal of medical genetics Part B, Neuropsychiatric genetics : the official publication of the International Society of Psychiatric Genetics. 2004;127B(1):5-10. 66. Maurer I, Zierz S, Möller H-J. Evidence for a mitochondrial oxidative phosphorylation defect in brains from patients with schizophrenia. Schizophrenia research. 2001;48(1):125-36. 67. Cavelier L, Jazin EE, Eriksson I, Prince J, Bave U, Oreland L, et al. Decreased cytochrome-c oxidase activity and lack of age-related accumulation of mitochondrial DNA deletions in the brains of schizophrenics. Genomics. 1995;29(1):217-24. 68. McCullumsmith RE, Hammond JH, Shan D, Meador-Woodruff JH. Postmortem brain: an underutilized substrate for studying severe mental illness. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2014;39(1):65-87. 69. Pettegrew JW, Keshavan MS, Panchalingam K, et al. Alterations in brain high-energy phosphate and membrane phospholipid metabolism in first-episode, drug-naive schizophrenics: A pilot study of the dorsal prefrontal cortex by in vivo phosphorus 31 nuclear magnetic resonance spectroscopy. Archives of general psychiatry. 1991;48(6):563-8. 70. Rowland LM, Pradhan S, Korenic S, Wijtenburg SA, Hong LE, Edden RA, et al. Elevated brain lactate in schizophrenia: a 7 T magnetic resonance spectroscopy study. Translational psychiatry. 2016;6(11):e967. 71. Altar CA, Jurata LW, Charles V, Lemire A, Liu P, Bukhman Y, et al. Deficient hippocampal neuron expression of proteasome, ubiquitin, and mitochondrial genes in multiple schizophrenia cohorts. Biological psychiatry. 2005;58(2):85-96. 72. Arion D, Corradi JP, Tang S, Datta D, Boothe F, He A, et al. Distinctive transcriptome alterations of prefrontal pyramidal neurons in schizophrenia and schizoaffective disorder. Mol Psychiatry. 2015;20(11):1397-405. 73. Arion D, Huo Z, Enwright JF, Corradi JP, Tseng G, Lewis DA. Transcriptome alterations in prefrontal pyramidal cells distinguish schizophrenia from bipolar and major depressive disorders. Biological psychiatry.

198

74. Garey LJ, Ong WY, Patel TS, Kanani M, Davis A, Mortimer AM, et al. Reduced dendritic spine density on cerebral cortical pyramidal neurons in schizophrenia. J Neurol Neurosurg Psychiatry. 1998;65(4):446-53. 75. Glantz LA, Lewis DA. Decreased dendritic spine density on prefrontal cortical pyramidal neurons in schizophrenia. Archives of general psychiatry. 2000;57(1):65-73. 76. Glausier JR, Lewis DA. Dendritic spine pathology in schizophrenia. Neuroscience. 2013;251:90- 107. 77. Kolluri N, Sun Z, Sampson AR, Lewis DA. Lamina-specific reductions in dendritic spine density in the prefrontal cortex of subjects with schizophrenia. The American journal of psychiatry. 2005;162(6):1200-2. 78. Hegde AN, DiAntonio A. Ubiquitin and the synapse. Nature reviews Neuroscience. 2002;3(11):854-61. 79. Murphey RK, Godenschwege TA. New roles for ubiquitin in the assembly and function of neuronal circuits. Neuron. 2002;36(1):5-8. 80. Pak DTS, Sheng M. Targeted Protein Degradation and Synapse Remodeling by an Inducible Protein Kinase. Science (New York, NY). 2003;302(5649):1368-73. 81. Ehlers M. Activity level controls postsynaptic composition and signaling via the ubiquitin- proteasome system. Nature neuroscience. 2003;6(3):231-42. 82. Speese SD, Trotta N, Rodesch CK, Aravamudan B, Broadie K. The ubiquitin proteasome system acutely regulates presynaptic protein turnover and synaptic efficacy. Current Biology. 2003;13(11):899- 910. 83. Hardingham GE, Do KQ. Linking early-life NMDAR hypofunction and oxidative stress in schizophrenia pathogenesis. Nature reviews Neuroscience. 2016;17(2):125-34. 84. Powell SB, Sejnowski TJ, Behrens MM. Behavioral and neurochemical consequences of cortical oxidative stress on parvalbumin-interneuron maturation in rodent models of schizophrenia. Neuropharmacology. 2012;62(3):1322-31. 85. Ben-Ari Y. Excitatory actions of gaba during development: the nature of the nurture. Nature reviews Neuroscience. 2002;3(9):728-39. 86. Arion D, Lewis DA. Altered expression of regulators of the cortical chloride transporters NKCC1 and KCC2 in schizophrenia. Archives of general psychiatry. 2011;68(1):21-31. 87. Windrem MS, Osipovitch M, Liu Z, Bates J, Chandler-Militello D, Zou L, et al. Human iPSC Glial Mouse Chimeras Reveal Glial Contributions to Schizophrenia. Cell Stem Cell.21(2):195-208.e6. 88. Kann O, Papageorgiou IE, Draguhn A. Highly energized inhibitory interneurons are a central element for information processing in cortical networks. Journal of Cerebral Blood Flow & Metabolism. 2014;34(8):1270-82. 89. McCasland JS, Hibbard LS. GABAergic neurons in barrel cortex show strong, whisker-dependent metabolic activation during normal behavior. The Journal of neuroscience : the official journal of the Society for Neuroscience. 1997;17(14):5509-27. 90. Ackermann RF, Finch DM, Babb TL, Engel J, Jr. Increased glucose metabolism during long- duration recurrent inhibition of hippocampal pyramidal cells. The Journal of neuroscience : the official journal of the Society for Neuroscience. 1984;4(1):251-64. 91. Arne Schousboe LKB, Karsten Kirkegaard Madsen, Helle S Waagepetersen. Amino acid neurotransmitter synthesis and removal. Neuroglia. 3 ed. Oxford: Oxford University Press; 2013. p. 443- 56. 92. Schousboe A, Scafidi S, Bak LK, Waagepetersen HS, McKenna MC. Glutamate Metabolism in the Brain Focusing on Astrocytes. Advances in neurobiology. 2014;11:13-30. 93. Petr GT, Sun Y, Frederick NM, Zhou Y, Dhamne SC, Hameed MQ, et al. Conditional Deletion of the Glutamate Transporter GLT-1 Reveals That Astrocytic GLT-1 Protects against Fatal Epilepsy While

199

Neuronal GLT-1 Contributes Significantly to Glutamate Uptake into Synaptosomes. The Journal of Neuroscience. 2015;35(13):5187-201. 94. McKenna MC, Sonnewald U, Huang X, Stevenson J, Zielke HR. Exogenous glutamate concentration regulates the metabolic fate of glutamate in astrocytes. Journal of neurochemistry. 1996;66(1):386-93. 95. McKenna MC. Glutamate pays its own way in astrocytes. Frontiers in endocrinology. 2013;4:191. 96. Sonnewald U, Westergaard N, Petersen SB, Unsgard G, Schousboe A. Metabolism of [U- 13C]glutamate in astrocytes studied by 13C NMR spectroscopy: incorporation of more label into lactate than into glutamine demonstrates the importance of the tricarboxylic acid cycle. Journal of neurochemistry. 1993;61(3):1179-82. 97. McCullumsmith RE, O'Donovan SM, Drummond JB, Benesh FS, Simmons M, Roberts R, et al. Cell- specific abnormalities of glutamate transporters in schizophrenia: sick astrocytes and compensating relay neurons? Mol Psychiatry. 2016;6:823-30. 98. Genda EN, Jackson JG, Sheldon AL, Locke SF, Greco TM, O'Donnell JC, et al. Co- compartmentalization of the astroglial glutamate transporter, GLT-1, with glycolytic enzymes and mitochondria. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2011;31(50):18275-88. 99. O'Donovan SM, Hasselfeld K, Bauer D, Simmons M, Roussos P, Haroutunian V, et al. Glutamate transporter splice variant expression in an enriched pyramidal cell population in schizophrenia. Translational psychiatry. 2015;5:e579. 100. Shan D, Mount D, Moore S, Haroutunian V, Meador-Woodruff JH, McCullumsmith RE. Abnormal partitioning of hexokinase 1 suggests disruption of a glutamate transport protein complex in schizophrenia. Schizophrenia research. 2014;154(1-3):1-13. 101. Sinead O'Donovan CS, Robert McCullumsmith. The role of glutamate transporters in the pathophysiology of neuropsychiatric disorders. NPJ Schizophrenia 2017(In press). 102. Shan D, Lucas EK, Drummond JB, Haroutunian V, Meador-Woodruff JH, McCullumsmith RE. Abnormal expression of glutamate transporters in temporal lobe areas in elderly patients with schizophrenia. Schizophrenia research. 2013. 103. Webster MJ, Knable MB, Johnston-Wilson N, Nagata K, Inagaki M, Yolken RH. Immunohistochemical localization of phosphorylated glial fibrillary acidic protein in the prefrontal cortex and hippocampus from patients with schizophrenia, bipolar disorder, and depression. Brain, behavior, and immunity. 2001;15(4):388-400. 104. Magistretti PJ. Neuron–glia metabolic coupling and plasticity. Journal of Experimental Biology. 2006;209(12):2304-11. 105. Benes FM, Davidson J, Bird ED. Quantitative cytoarchitectural studies of the cerebral cortex of schizophrenics. Archives of general psychiatry. 1986;43(1):31-5. 106. Cotter D, Mackay D, Chana G, Beasley C, Landau S, Everall IP. Reduced neuronal size and glial cell density in area 9 of the dorsolateral prefrontal cortex in subjects with major depressive disorder. Cerebral cortex (New York, NY : 1991). 2002;12(4):386-94. 107. Goudriaan A, de Leeuw C, Ripke S, Hultman CM, Sklar P, Sullivan PF, et al. Specific Glial Functions Contribute to Schizophrenia Susceptibility. Schizophrenia bulletin. 2014;40(4):925-35. 108. Sugai T, Kawamura M, Iritani S, Araki K, Makifuchi T, Imai C, et al. Prefrontal abnormality of schizophrenia revealed by DNA microarray: impact on glial and neurotrophic gene expression. Annals of the New York Academy of Sciences. 2004;1025:84-91. 109. Morris G, Berk M. The many roads to mitochondrial dysfunction in neuroimmune and neuropsychiatric disorders. BMC Medicine. 2015;13(1):68.

200

110. Martínez-Cengotitabengoa M, Mac-Dowell KS, Leza JC, Micó JA, Fernandez M, Echevarría E. Cognitive impairment is related to oxidative stress and chemokine levels in first psychotic episodes. Schizophr Res. 2012;137. 111. Gubert C, Stertz L, Pfaffenseller B, Panizzutti BS, Rezin GT, Massuda R. Mitochondrial activity and oxidative stress markers in peripheral bloodmononuclear cells of patients with bipolar disorder, schizophrenia, and healthy subjects. J Psychiatr Res. 2013;47. 112. Pongrac J, Middleton FA, Lewis DA, Levitt P, Mirnics K. Gene expression profiling with DNA microarrays: advancing our understanding of psychiatric disorders. Neurochemical research. 2002;27(10):1049-63. 113. Mirnics K, Middleton FA, Lewis DA, Levitt P. Analysis of complex brain disorders with gene expression microarrays: schizophrenia as a disease of the synapse. Trends in neurosciences. 2001;24(8):479-86. 114. Frankle WG, Lerma J, Laruelle M. The synaptic hypothesis of schizophrenia. Neuron. 2003;39(2):205-16. 115. Stephan KE, Friston KJ, Frith CD. Dysconnection in schizophrenia: from abnormal synaptic plasticity to failures of self-monitoring. Schizophrenia bulletin. 2009;35(3):509-27. 116. Khvotchev M. Schizophrenia and synapse: emerging role of presynaptic fusion machinery. Biological psychiatry. 2010;67(3):197-8. 117. Hayashi-Takagi A, Sawa A. Disturbed synaptic connectivity in schizophrenia: convergence of genetic risk factors during neurodevelopment. Brain research bulletin. 2010;83(3-4):140-6. 118. Harrison PJ, Law AJ. Neuregulin 1 and schizophrenia: genetics, gene expression, and neurobiology. Biological psychiatry. 2006;60(2):132-40. 119. Millar JK, Wilson-Annan JC, Anderson S, Christie S, Taylor MS, Semple CA, et al. Disruption of two novel genes by a translocation co-segregating with schizophrenia. Human molecular genetics. 2000;9(9):1415-23. 120. Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, Sigmundsson T, Ghosh S, et al. Neuregulin 1 and susceptibility to schizophrenia. American journal of human genetics. 2002;71(4):877- 92. 121. St Clair D, Blackwood D, Muir W, Carothers A, Walker M, Spowart G, et al. Association within a family of a balanced autosomal translocation with major mental illness. Lancet. 1990;336(8706):13-6. 122. Gulsuner S, Walsh T, Watts AC, Lee MK, Thornton AM, Casadei S, et al. Spatial and Temporal Mapping of De novo Mutations in Schizophrenia To a Fetal Prefrontal Cortical Network. Cell. 2013;154(3):518-29. 123. Mirnics K, Middleton F, Marquez A, Lewis D, Levitt P. Molecular characterization of schizophrenia viewed by microarray analysis of gene expression in prefrontal cortex. Neuron. 2000;28(1):53-67. 124. Purcell SM, Moran JL, Fromer M, Ruderfer D, Solovieff N, Roussos P, et al. A polygenic burden of rare disruptive mutations in schizophrenia. Nature. 2014;506(7487):185-90. 125. Kirov G, Rujescu D, Ingason A, Collier DA, O'Donovan MC, Owen MJ. Neurexin 1 (NRXN1) deletions in schizophrenia. Schizophrenia bulletin. 2009;35(5):851-4. 126. Kirov G, Gumus D, Chen W, Norton N, Georgieva L, Sari M, et al. Comparative genome hybridization suggests a role for NRXN1 and APBA2 in schizophrenia. Human molecular genetics. 2008;17(3):458-65. 127. Glantz LA, Lewis DA. Dendritic spine density in schizophrenia and depression. Archives of general psychiatry. 2001;58(2):203. 128. Selemon LD, Goldman-Rakic PS. The reduced neuropil hypothesis: a circuit based model of schizophrenia. Biological psychiatry. 1999;45(1):17-25.

201

129. Meador-Woodruff J, Clinton S, Beneyto M, McCullumsmith R. Molecular abnormalities of the glutamate synapse in the thalamus in schizophrenia. Annals of the New York Academy of Sciences. 2003;1003:75-93. 130. Spangaro M, Bosia M, Zanoletti A, Bechi M, Cocchi F, Pirovano A, et al. Cognitive dysfunction and glutamate reuptake: effect of EAAT2 polymorphism in schizophrenia. Neuroscience letters. 2012;522(2):151-5. 131. Oni-Orisan A, Kristiansen L, Haroutunian V, Meador-Woodruff J, McCullumsmith R. Altered vesicular glutamate transporter expression in the anterior cingulate cortex in schizophrenia. Biological psychiatry. 2008;63(8):766-75. 132. Meador-Woodruff J, Healy D. Glutamate receptor expression in schizophrenic brain. Brain research Brain research reviews. 2000;31(2-3):288-94. 133. Alda M, Ahrens B, Lit W, Dvorakova M, Labelle A, Zvolsky P, et al. Age of onset in familial and sporadic schizophrenia. Acta Psychiatr Scand. 1996;93(6):447-50. 134. Steiner J, Martins-de-Souza D, Schiltz K, Sarnyai Z, Westphal S, Isermann B, et al. Clozapine promotes glycolysis and myelin lipid synthesis in cultured oligodendrocytes. Frontiers in cellular neuroscience. 2014;8:384. 135. Holcomb HH, Cascella NG, Thaker GK, Medoff DR, Dannals RF, Tamminga CA. Functional sites of neuroleptic drug action in the human brain: PET/FDG studies with and without haloperidol. The American journal of psychiatry. 1996;153(1):41-9. 136. Stone WS, Seidman LJ, Wojcik JD, Green AI. Glucose effects on cognition in schizophrenia. Schizophrenia research. 2003;62(1-2):93-103. 137. Fucetola R, Newcomer JW, Craft S, Melson AK. Age- and dose-dependent glucose-induced increases in memory and attention in schizophrenia. Psychiatry research. 1999;88(1):1-13. 138. Dwyer DS, Lu XH, Iii AM. Neuronal glucose metabolism and schizophrenia: therapeutic prospects? Expert review of neurotherapeutics. 2003;3(1):29-40. 139. Maurer I, Moller HJ. Inhibition of complex I by neuroleptics in normal human brain cortex parallels the extrapyramidal toxicity of neuroleptics. Mol Cell Biochem. 1997;174(1-2):255-9. 140. Martins-de-Souza D, Gattaz WF, Schmitt A, Maccarrone G, Hunyadi-Gulyas E, Eberlin MN, et al. Proteomic analysis of dorsolateral prefrontal cortex indicates the involvement of cytoskeleton, oligodendrocyte, energy metabolism and new potential markers in schizophrenia. Journal of psychiatric research. 2009;43(11):978-86. 141. Martins-de-Souza D, Gattaz WF, Schmitt A, Novello JC, Marangoni S, Turck CW. Proteome analysis of schizophrenia patients Wer-nicke’s area reveals an energy metabolism dysregulation. BMC Psychiatry. 2009;30. 142. Ma D, Chan MK, Lockstone HE, Pietsch SR, Jones DN, Cilia J, et al. Antipsychotic treatment alters protein expression associated with presynaptic function and nervous system development in rat frontal cortex. J Proteome Res. 2009;8(7):3284-97. 143. Ryan MC, Collins P, Thakore JH. Impaired fasting glucose tolerance in first-episode, drug-naive patients with schizophrenia. The American journal of psychiatry. 2003;160(2):284-9. 144. Spelman LM, Walsh PI, Sharifi N, Collins P, Thakore JH. Impaired glucose tolerance in first- episode drug-naive patients with schizophrenia. Diabetic medicine : a journal of the British Diabetic Association. 2007;24(5):481-5. 145. Newcomer JW. Metabolic syndrome and mental illness. The American journal of managed care. 2007;13(7 Suppl):S170-7. 146. Herberth M, Koethe D, Cheng TMK, Krzyszton ND, Schoeffmann S, Guest PC, et al. Impaired glycolytic response in peripheral blood mononuclear cells of first-onset antipsychotic-naive schizophrenia patients. Mol Psychiatry. 2011;16(8):848-59.

202

147. Wolkin A, Jaeger J, Brodie JD, Wolf AP, Fowler J, Rotrosen J, et al. Persistence of cerebral metabolic abnormalities in chronic schizophrenia as determined by positron emission tomography. The American journal of psychiatry. 1985;142(5):564-71. 148. Tamminga CA, Thaker GK, Buchanan R, Kirkpatrick B, Alphs LD, Chase TN, et al. Limbic system abnormalities identified in schizophrenia using positron emission tomography with fluorodeoxyglucose and neocortical alterations with deficit syndrome. Archives of general psychiatry. 1992;49(7):522-30. 149. Wiesel FA, Wik G, Sjögren I, Blomqvist G, Greitz T, Stone-Elander S. Regional brain glucose metabolism in drug free schizophrenic patients and clinical correlates. Acta Psychiatrica Scandinavica. 1987;76(6):628-41. 150. Akbarian S, Sucher NJ, Bradley D, Tafazzoli A, Trinh D, Hetrick WP, et al. Selective alterations in gene expression for NMDA receptor subunits in prefrontal cortex of schizophrenics. The Journal of neuroscience : the official journal of the Society for Neuroscience. 1996;16(1):19-30. 151. Balu DT, Coyle JT. Neuroplasticity signaling pathways linked to the pathophysiology of schizophrenia. Neuroscience & Biobehavioral Reviews. 2011;35(3):848-70. 152. Coyle J, Tsai G, Goff D. Converging evidence of NMDA receptor hypofunction in the pathophysiology of schizophrenia. Annals of the New York Academy of Sciences. 2003;1003:318-27. 153. Coyle JT. The glutamatergic dysfunction hypothesis for schizophrenia. Harvard review of psychiatry. 1996;3(5):241-53. 154. Funk A, Rumbaugh G, Harotunian V, McCullumsmith R, Meador-Woodruff J. Decreased expression of NMDA receptor-associated proteins in frontal cortex of elderly patients with schizophrenia. Neuroreport. 2009;20(11):1019-22. 155. Coyle JT, Tsai G, Goff DC. Ionotropic glutamate receptors as therapeutic targets in schizophrenia. Current drug targets CNS and neurological disorders. 2002;1(2):183-9. 156. Luby ED, Cohen BD, Rosenbaum G, Gottlieb JS, Kelley R. Study of a new schizophrenomimetic drug; sernyl. AMA archives of neurology and psychiatry. 1959;81(3):363-9. 157. Krystal JH, Karper LP, Seibyl JP, Freeman GK, Delaney R, Bremner JD, et al. Subanesthetic effects of the noncompetitive NMDA antagonist, ketamine, in humans. Psychotomimetic, perceptual, cognitive, and neuroendocrine responses. Archives of general psychiatry. 1994;51(3):199-214. 158. Lahti AC, Weiler MA, Tamara Michaelidis BA, Parwani A, Tamminga CA. Effects of ketamine in normal and schizophrenic volunteers. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2001;25(4):455-67. 159. Mouri A, Noda Y, Enomoto T, Nabeshima T. Phencyclidine animal models of schizophrenia: approaches from abnormality of glutamatergic neurotransmission and neurodevelopment. Neurochemistry international. 2007;51(2-4):173-84. 160. Olney JW, Farber NB. Glutamate receptor dysfunction and schizophrenia. Archives of general psychiatry. 1995;52(12):998-1007. 161. Duncan G, Miyamoto S, Gu H, Lieberman J, Koller B, Snouwaert J. Alterations in regional brain metabolism in genetic and pharmacological models of reduced NMDA receptor function. Brain research. 2002;951(2):166-76. 162. Dzirasa K, Ramsey AJ, Takahashi DY, Stapleton J, Potes JM, Williams JK, et al. Hyperdopaminergia and NMDA receptor hypofunction disrupt neural phase signaling. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2009;29(25):8215-24. 163. Halene TB, Ehrlichman RS, Liang Y, Christian EP, Jonak GJ, Gur TL, et al. Assessment of NMDA receptor NR1 subunit hypofunction in mice as a model for schizophrenia. Genes, Brain and Behavior. 2009;8(7):661-75. 164. Jentsch JD, Roth RH. The neuropsychopharmacology of phencyclidine: from NMDA receptor hypofunction to the dopamine hypothesis of schizophrenia. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 1999;20(3):201-25.

203

165. Ramsey AJ. NR1 knockdown mice as a representative model of the glutamate hypothesis of schizophrenia. Progress in brain research. 2009;179:51-8. 166. Mohn AR, Gainetdinov RR, Caron MG, Koller BH. Mice with reduced NMDA receptor expression display behaviors related to schizophrenia. Cell. 1999;98(4):427-36. 167. Duncan GE, Moy SS, Lieberman JA, Koller BH. Effects of haloperidol, clozapine, and quetiapine on sensorimotor gating in a genetic model of reduced NMDA receptor function. Psychopharmacology. 2006;184(2):190-200. 168. Inta D, Monyer H, Sprengel R, Meyer-Lindenberg A, Gass P. Mice with genetically altered glutamate receptors as models of schizophrenia: a comprehensive review. Neuroscience and biobehavioral reviews. 2010;34(3):285-94. 169. Javitt DC, Zukin SR. Recent advances in the phencyclidine model of schizophrenia. The American journal of psychiatry. 1991;148(10):1301-8. 170. Rung JP, Carlsson A, Ryden Markinhuhta K, Carlsson ML. (+)-MK-801 induced social withdrawal in rats; a model for negative symptoms of schizophrenia. Progress in neuro-psychopharmacology & biological psychiatry. 2005;29(5):827-32. 171. Guo X, Hamilton P, Reish N, Sweatt J, Miller C, Rumbaugh G. Reduced Expression of the NMDA Receptor-Interacting Protein SynGAP Causes Behavioral Abnormalities that Model Symptoms of Schizophrenia. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2009. 172. Barkus C, Feyder M, Graybeal C, Wright T, Wiedholz L, Izquierdo A, et al. Do GluA1 knockout mice exhibit behavioral abnormalities relevant to the negative or cognitive symptoms of schizophrenia and schizoaffective disorder? Neuropharmacology. 2012;62(3):1263-72. 173. Bannerman DM, Deacon RM, Brady S, Bruce A, Sprengel R, Seeburg PH, et al. A comparison of GluR-A-deficient and wild-type mice on a test battery assessing sensorimotor, affective, and cognitive behaviors. Behav Neurosci. 2004;118(3):643-7. 174. Hikida T, Jaaro-Peled H, Seshadri S, Oishi K, Hookway C, Kong S, et al. Dominant-negative DISC1 transgenic mice display schizophrenia-associated phenotypes detected by measures translatable to humans. Proceedings of the National Academy of Sciences of the United States of America. 2007;104(36):14501-6. 175. Pletnikov MV, Ayhan Y, Nikolskaia O, Xu Y, Ovanesov MV, Huang H, et al. Inducible expression of mutant human DISC1 in mice is associated with brain and behavioral abnormalities reminiscent of schizophrenia. Mol Psychiatry. 2008;13(2):173-86, 15. 176. Dwyer DS, Pinkofsky HB, Liu Y, Bradley RJ. Antipsychotic drugs affect glucose uptake and the expression of glucose transporters in PC12 cells. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 1999;23(1):69-80. 177. Heiser P, Singh S, Krieg JC, Vedder H. Effects of different antipsychotics and the antidepressant mirtazapine on glucose transporter mRNA levels in human blood cells. Journal of psychiatric research. 2006;40(4):374-9. 178. Dwyer DS, Donohoe D, Lu XH, Aamodt EJ. Mechanistic Connections between Glucose/Lipid Disturbances and Weight Gain induced by Antipsychotic Drugs. International review of neurobiology. 2005;65:211-47. 179. Dwyer DS, Donohoe D. Induction of hyperglycemia in mice with atypical antipsychotic drugs that inhibit glucose uptake. Pharmacology, biochemistry, and behavior. 2003;75(2):255-60. 180. Mehler-Wex C, Grünblatt E, Zeiske S, Gille G, Rausch D, Warnke A, et al. Microarray analysis reveals distinct gene expression patterns in the mouse cortex following chronic neuroleptic and stimulant treatment: implications for body weight changes. Journal of Neural Transmission. 2006;113(10):1383-93.

204

181. Rizig MA, McQuillin A, Ng A, Robinson M, Harrison A, Zvelebil M, et al. A gene expression and systems pathway analysis of the effects of clozapine compared to haloperidol in the mouse brain implicates susceptibility genes for schizophrenia. Journal of psychopharmacology (Oxford, England). 2012;26(9):1218-30. 182. Wiedholz L, Owens W, Horton R, Feyder M, Karlsson R, Hefner K, et al. Mice lacking the AMPA GluR1 receptor exhibit striatal hyperdopaminergia and 'schizophrenia-related' behaviors. Mol Psychiatry. 2008;13(6):631-40. 183. Lee PR, Brady DL, Shapiro RA, Dorsa DM, Koenig JI. Social interaction deficits caused by chronic phencyclidine administration are reversed by oxytocin. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2005;30(10):1883-94. 184. Paylor R, Nguyen M, Crawley JN, Patrick J, Beaudet A, Orr-Urtreger A. Alpha7 nicotinic receptor subunits are not necessary for hippocampal-dependent learning or sensorimotor gating: a behavioral characterization of Acra7-deficient mice. Learning & memory (Cold Spring Harbor, NY). 1998;5(4-5):302- 16. 185. Fernandes C, Hoyle E, Dempster E, Schalkwyk LC, Collier DA. Performance deficit of alpha7 nicotinic receptor knockout mice in a delayed matching-to-place task suggests a mild impairment of working/episodic-like memory. Genes, brain, and behavior. 2006;5(6):433-40. 186. Salas R, Orr-Urtreger A, Broide RS, Beaudet A, Paylor R, De Biasi M. The nicotinic acetylcholine receptor subunit alpha 5 mediates short-term effects of nicotine in vivo. Molecular pharmacology. 2003;63(5):1059-66. 187. Labrie V, Fukumura R, Rastogi A, Fick LJ, Wang W, Boutros PC, et al. Serine racemase is associated with schizophrenia susceptibility in humans and in a mouse model. Human molecular genetics. 2009;18(17):3227-43. 188. Pascual O, Casper KB, Kubera C, Zhang J, Revilla-Sanchez R, Sul JY, et al. Astrocytic purinergic signaling coordinates synaptic networks. Science (New York, NY). 2005;310(5745):113-6. 189. Weckmann K, Labermaier C, Asara JM, Müller MB, Turck CW. Time-dependent metabolomic profiling of Ketamine drug action reveals hippocampal pathway alterations and biomarker candidates. Translational psychiatry. 2014;4(11):e481. 190. Wesseling H, Chan MK, Tsang TM, Ernst A, Peters F, Guest PC, et al. A combined metabonomic and proteomic approach identifies frontal cortex changes in a chronic phencyclidine rat model in relation to human schizophrenia brain pathology. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2013;38(12):2532-44. 191. Yang J, Chen T, Sun L, Zhao Z, Qi X, Zhou K, et al. Potential metabolite markers of schizophrenia. Mol Psychiatry. 2013;18(1):67-78. 192. Wesseling H, Guest PC, Lee CM, Wong EH, Rahmoune H, Bahn S. Integrative proteomic analysis of the NMDA NR1 knockdown mouse model reveals effects on central and peripheral pathways associated with schizophrenia and autism spectrum disorders. Molecular autism. 2014;5:38. 193. Courtney Sullivan KC, Rachael Koene, Amy Ramsey, Robert McCullumsmith. Decreased lactate dehydrogenase activity and abnormal expression of lactate shuttle transporters in schizophrenia. Abstract, Society for Neuroscience Meeting. 2016. 194. Bergersen LH. Lactate transport and signaling in the brain: potential therapeutic targets and roles in body–brain interaction. Journal of Cerebral Blood Flow & Metabolism. 2015;35(2):176-85. 195. Loubinoux I, Meric P, Borredon J, Correze J-L, Gillet B, Beloeil J-C, et al. Cerebral metabolic changes induced by MK-801: a 1D (phosphorus and proton) and 2D (proton) in vivo NMR spectroscopy study. Brain research. 1994;643(1–2):115-24. 196. Clow DW, Lee SJ, Hammer RP, Jr. Competitive (AP7) and non-competitive (MK-801) NMDA receptor antagonists differentially alter glucose utilization in rat cortex. Synapse (New York, NY). 1991;7(4):260-8.

205

197. Nehls DG, Park CK, MacCormack AG, McCulloch J. The effects of N-methyl-D-aspartate receptor blockade with MK-801 upon the relationship between cerebral blood flow and glucose utilisation. Brain research. 1990;511(2):271-9. 198. Eyjolfsson EM, Nilsen LH, Kondziella D, Brenner E, Håberg A, Sonnewald U. Altered 13C Glucose Metabolism in the Cortico—Striato—Thalamo—Cortical Loop in the MK-801 Rat Model of Schizophrenia. Journal of Cerebral Blood Flow & Metabolism. 2010;31(3):976-85. 199. Smith U. Pioglitazone: mechanism of action. International journal of clinical practice Supplement. 2001(121):13-8. 200. Dello Russo C, Gavrilyuk V, Weinberg G, Almeida A, Bolanos JP, Palmer J, et al. Peroxisome proliferator-activated receptor gamma thiazolidinedione agonists increase glucose metabolism in astrocytes. The Journal of biological chemistry. 2003;278(8):5828-36. 201. Owen L, Sunram-Lea SI. Metabolic Agents that Enhance ATP can Improve Cognitive Functioning: A Review of the Evidence for Glucose, Oxygen, Pyruvate, Creatine, and L-Carnitine. Nutrients. 2011;3(8):735-55. 202. Sato T, Hanyu H, Hirao K, Kanetaka H, Sakurai H, Iwamoto T. Efficacy of PPAR-gamma agonist pioglitazone in mild Alzheimer disease. Neurobiology of aging. 2011;32(9):1626-33. 203. Iranpour N, Zandifar A, Farokhnia M, Goguol A, Yekehtaz H, Khodaie-Ardakani MR, et al. The effects of pioglitazone adjuvant therapy on negative symptoms of patients with chronic schizophrenia: a double-blind and placebo-controlled trial. Human psychopharmacology. 2016;31(2):103-12. 204. Smith RC, Jin H, Li C, Bark N, Shekhar A, Dwivedi S, et al. Effects of pioglitazone on metabolic abnormalities, psychopathology, and cognitive function in schizophrenic patients treated with antipsychotic medication: a randomized double-blind study. Schizophrenia research. 2013;143(1):18-24. 205. Research Foundation for Mental Hygiene I. A Placebo-controlled Efficacy Study of IV Ceftriaxone for Refractory Psychosis. In: Depression NAfRoSa, editor. 2009. 206. Stoessl AJ. Glucose utilization: still in the synapse. Nature neuroscience. 2017;20(3):382-4. 207. Zimmer ER, Parent MJ, Souza DG, Leuzy A, Lecrux C, Kim H-I, et al. [18F]FDG PET signal is driven by astroglial glutamate transport. Nature neuroscience. 2017;20(3):393-5. 208. Furman DJ, Hamilton JP, Gotlib IH. Frontostriatal functional connectivity in major depressive disorder. Biology of mood & anxiety disorders. 2011;1(1):11. 209. Van Snellenberg JX, Torres IJ, Thornton AE. Functional neuroimaging of working memory in schizophrenia: task performance as a moderating variable. Neuropsychology. 2006;20(5):497-510. 210. Lesh TA, Niendam TA, Minzenberg MJ, Carter CS. Cognitive control deficits in schizophrenia: mechanisms and meaning. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2011;36(1):316-38. 211. Meltzer HY, McGurk SR. The effects of clozapine, risperidone, and olanzapine on cognitive function in schizophrenia. Schizophrenia bulletin. 1999;25(2):233-55. 212. Meltzer HY, Park S, Kessler R. Cognition, schizophrenia, and the atypical antipsychotic drugs. Proceedings of the National Academy of Sciences. 1999;96(24):13591. 213. Glantz LA, Lewis DA. Reduction of synaptophysin immunoreactivity in the prefrontal cortex of subjects with schizophrenia. Regional and diagnostic specificity. Archives of general psychiatry. 1997;54(7):660-9. 214. McCullumsmith R, Clinton S, Meador-Woodruff J. Schizophrenia as a disorder of neuroplasticity. International review of neurobiology. 2004;59:19-45. 215. Sullivan CR, O’Donovan S, McCullumsmith RE, Ramsey A. Defects in bioenergetic coupling in schizophrenia. Biological psychiatry. 2017. 216. Trivedi JK. Cognitive deficits in psychiatric disorders: Current status. 2006;48(1):10-20. 217. Badcock JC. The cognitive neuropsychology of auditory hallucinations: a parallel auditory pathways framework. Schizophrenia bulletin. 2010;36(3):576-84.

206

218. Kay SR. Significance of the positive-negative distinction in schizophrenia. Schizophrenia bulletin. 1990;16(4):635-52. 219. Szoke A, Meary A, Trandafir A, Bellivier F, Roy I, Schurhoff F, et al. Executive deficits in psychotic and bipolar disorders - implications for our understanding of schizoaffective disorder. Eur Psychiatry. 2008;23(1):20-5. 220. Rajji TK, Mulsant BH. Nature and course of cognitive function in late-life schizophrenia: a systematic review. Schizophrenia research. 2008;102(1-3):122-40. 221. Weinberger DR, Berman KF, Zec RF. Physiologic dysfunction of dorsolateral prefrontal cortex in schizophrenia. I. Regional cerebral blood flow evidence. Archives of general psychiatry. 1986;43(2):114- 24. 222. Carter CS, Perlstein W, Ganguli R, Brar J, Mintun M, Cohen JD. Functional hypofrontality and working memory dysfunction in schizophrenia. The American journal of psychiatry. 1998;155(9):1285-7. 223. Cohen JD, Servan-Schreiber D. Context, cortex, and dopamine: a connectionist approach to behavior and biology in schizophrenia. Psychological review. 1992;99(1):45-77. 224. Lewis DA, Hashimoto T, Volk DW. Cortical inhibitory neurons and schizophrenia. Nature reviews Neuroscience. 2005;6(4):312-24. 225. Sawaguchi T, Matsumura M, Kubota K. Delayed response deficits produced by local injection of bicuculline into the dorsolateral prefrontal cortex in Japanese macaque monkeys. Experimental brain research. 1989;75(3):457-69. 226. Benes F, Todtenkopf M, Vincent S. Meta-analysis of nonpyramidal neuron (NP) loss in layer II in anterior cingulate cortex (ACCx-II) from three studies of postmortem schizophrenic (SZ) brain. Soc Neurosci Abstract. 1998; 24(1275). 227. Benes FM, Berretta S. GABAergic interneurons: implications for understanding schizophrenia and bipolar disorder. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2001;25(1):1-27. 228. Payne JA, Rivera C, Voipio J, Kaila K. Cation-chloride co-transporters in neuronal communication, development and trauma. Trends in neurosciences. 2003;26(4):199-206. 229. Lauf PK, Bauer J, Adragna NC, Fujise H, Zade-Oppen AM, Ryu KH, et al. Erythrocyte K-Cl cotransport: properties and regulation. The American journal of physiology. 1992;263(5 Pt 1):C917-32. 230. Flatman PW. Regulation of Na-K-2Cl cotransport by phosphorylation and protein-protein interactions. Biochimica et biophysica acta. 2002;1566(1-2):140-51. 231. Powchik P, Davidson M, Haroutunian V, Gabriel SM, Purohit DP, Perl DP, et al. Postmortem studies in schizophrenia. Schizophrenia bulletin. 1998;24(3):325-41. 232. Beeri MS, Rapp M, Silverman JM, Schmeidler J, Grossman HT, Fallon JT, et al. Coronary artery disease is associated with Alzheimer disease neuropathology in APOE4 carriers. Neurology. 2006;66(9):1399-404. 233. Purohit DP, Davidson M, Perl DP, Powchik P, Haroutunian VH, Bierer LM, et al. Severe cognitive impairment in elderly schizophrenic patients: a clinicopathological study. Biological psychiatry. 1993;33(4):255-60. 234. Roberts RC, Gaither LA, Gao XM, Kashyap SM, Tamminga CA. Ultrastructural correlates of haloperidol-induced oral dyskinesias in rat striatum. Synapse (New York, NY). 1995;20(3):234-43. 235. Nyberg S, Farde L, Halldin C. Delayed normalization of central D2 dopamine receptor availability after discontinuation of haloperidol decanoate. Preliminary findings. Archives of general psychiatry. 1997;54(10):953-8. 236. Drummond JB, Tucholski J, Haroutunian V, Meador-Woodruff JH. Transmembrane AMPA receptor regulatory protein (TARP) dysregulation in anterior cingulate cortex in schizophrenia. Schizophrenia research. 2013;147(1):32-8.

207

237. Mithani S, Atmadja S, Baimbridge KG, Fibiger HC. Neuroleptic-induced oral dyskinesias: effects of progabide and lack of correlation with regional changes in glutamic acid decarboxylase and choline acetyltransferase activities. Psychopharmacology. 1987;93(1):94-100. 238. Kashihara K, Sato M, Fujiwara Y, Harada T, Ogawa T, Otsuki S. Effects of intermittent and continuous haloperidol administration on the dopaminergic system in the rat brain. Biological psychiatry. 1986;21(7):650-6. 239. Harte MK, Bachus SB, Reynolds GP. Increased N-acetylaspartate in rat striatum following long- term administration of haloperidol. Schizophrenia research. 2005;75(2-3):303-8. 240. Bauer D, Haroutunian V, McCullumsmith R, Meador-Woodruff J. Expression of four housekeeping proteins in elderly patients with schizophrenia. J Neural Transm. 2009;116(4):487-91. 241. Wassef A, Baker J, Kochan LD. GABA and schizophrenia: a review of basic science and clinical studies. Journal of clinical psychopharmacology. 2003;23(6):601-40. 242. Hyde TM, Lipska BK, Ali T, Mathew SV, Law AJ, Metitiri OE, et al. Expression of GABA signaling molecules KCC2, NKCC1, and GAD1 in cortical development and schizophrenia. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2011;31(30):11088-95. 243. Nezu A, Parvin MN, Turner RJ. A Conserved Hydrophobic Tetrad near the C Terminus of the Secretory Na+-K+-2Cl- Cotransporter (NKCC1) Is Required for Its Correct Intracellular Processing. J Biol Chem. 2009;284(11):6869-76. 244. Rinehart J, Vázquez N, Kahle KT, Hodson CA, Ring AM, Gulcicek EE, et al. WNK2 Kinase Is a Novel Regulator of Essential Neuronal Cation-Chloride Cotransporters. J Biol Chem. 2011;286(34):30171-80. 245. Kahle KT, Staley K. Altered neuronal chloride homeostasis and excitatory GABAergic signaling in human temporal lobe epilepsy. Epilepsy currents / American Epilepsy Society. 2008;8(2):51-3. 246. Payne JA, Stevenson TJ, Donaldson LF. Molecular characterization of a putative K-Cl cotransporter in rat brain. A neuronal-specific isoform. The Journal of biological chemistry. 1996;271(27):16245-52. 247. Plotkin MD, Snyder EY, Hebert SC, Delpire E. Expression of the Na-K-2Cl cotransporter is developmentally regulated in postnatal rat brains: a possible mechanism underlying GABA's excitatory role in immature brain. Journal of neurobiology. 1997;33(6):781-95. 248. Plotkin MD, Kaplan MR, Peterson LN, Gullans SR, Hebert SC, Delpire E. Expression of the Na(+)- K(+)-2Cl- cotransporter BSC2 in the nervous system. The American journal of physiology. 1997;272(1 Pt 1):C173-83. 249. Kanaka C, Ohno K, Okabe A, Kuriyama K, Itoh T, Fukuda A, et al. The differential expression patterns of messenger RNAs encoding K-Cl cotransporters (KCC1,2) and Na-K-2Cl cotransporter (NKCC1) in the rat nervous system. Neuroscience. 2001;104(4):933-46. 250. Markkanen M, Karhunen T, Llano O, Ludwig A, Rivera C, Uvarov P, et al. Distribution of neuronal KCC2a and KCC2b isoforms in mouse CNS. The Journal of comparative neurology. 2014;522(8):1897-914. 251. Funk AJ, McCullumsmith RE, Haroutunian V, Meador-Woodruff JH. Abnormal activity of the MAPK- and cAMP-associated signaling pathways in frontal cortical areas in postmortem brain in schizophrenia. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2012;37(4):896-905. 252. Katsel P, Davis KL, Gorman JM, Haroutunian V. Variations in differential gene expression patterns across multiple brain regions in schizophrenia. Schizophrenia research. 2005;77(2-3):241-52. 253. Kristiansen L, Patel S, Haroutunian V, Meador-Woodruff J. Expression of the NR2B-NMDA receptor subunit and its Tbr-1/CINAP regulatory proteins in postmortem brain suggest altered receptor processing in schizophrenia. Synapse (New York, NY). 2010;64(7):495-502. 254. Oni-Orisan A, Kristiansen LV, Haroutunian V, Meador-Woodruff JH, McCullumsmith RE. Altered vesicular glutamate transporter expression in the anterior cingulate cortex in schizophrenia. Biological psychiatry. 2008;63(8):766-75.

208

255. Benes FM, McSparren J, Bird ED, SanGiovanni JP, Vincent SL. Deficits in small interneurons in prefrontal and cingulate cortices of schizophrenic and schizoaffective patients. Archives of general psychiatry. 1991;48(11):996-1001. 256. Benes FM. Emerging principles of altered neural circuitry in schizophrenia. Brain research Brain research reviews. 2000;31(2-3):251-69. 257. Zavitsanou K, Huang XF. Decreased [3H]spiperone binding in the anterior cingulate cortex of schizophrenia patients: an autoradiographic study. Neuroscience. 2002;109(4):709-16. 258. Wolkowitz OM, Pickar D. Benzodiazepines in the treatment of schizophrenia: a review and reappraisal. The American journal of psychiatry. 1991;148(6):714-26. 259. Gaillard R, Ouanas A, Spadone C, Llorca PM, Loo H, Bayle FJ. [Benzodiazepines and schizophrenia, a review of the literature]. L'Encephale. 2006;32(6 Pt 1):1003-10. 260. Costa E, Guidotti A, Mao CC, Suria A. New concepts on the mechanism of action of benzodiazepines. Life sciences. 1975;17(2):167-85. 261. Rudolph U, Mohler H. GABA-based therapeutic approaches: GABAA receptor subtype functions. Current opinion in pharmacology. 2006;6(1):18-23. 262. Vaughan CW, Ingram SL, Connor MA, Christie MJ. How opioids inhibit GABA-mediated neurotransmission. Nature. 1997;390(6660):611-4. 263. Capogna M, Gahwiler BH, Thompson SM. Mechanism of mu-opioid receptor-mediated presynaptic inhibition in the rat hippocampus in vitro. The Journal of physiology. 1993;470:539-58. 264. Vaughan CW, Christie MJ. Presynaptic inhibitory action of opioids on synaptic transmission in the rat periaqueductal grey in vitro. The Journal of physiology. 1997;498 ( Pt 2):463-72. 265. Eastwood SL, Harrison PJ. Decreased expression of vesicular glutamate transporter 1 and complexin II mRNAs in schizophrenia: further evidence for a synaptic pathology affecting glutamate neurons. Schizophrenia research. 2005;73(2-3):159-72. 266. McCullumsmith RE, Meador-Woodruff JH. Novel approaches to the study of postmortem brain in psychiatric illness: old limitations and new challenges. Biological psychiatry. 2011;69(2):127-33. 267. Lisman JE, Coyle JT, Green RW, Javitt DC, Benes FM, Heckers S, et al. Circuit-based framework for understanding neurotransmitter and risk gene interactions in schizophrenia. Trends in neurosciences. 2008;31(5):234-42. 268. Sullivan CR, Funk AJ, Shan D, Haroutunian V, McCullumsmith RE. Decreased chloride channel expression in the dorsolateral prefrontal cortex in schizophrenia. PloS one. 2015;10(3):e0123158. 269. Vawter MP, Barrett T, Cheadle C, Sokolov BP, Wood WH, 3rd, Donovan DM, et al. Application of cDNA microarrays to examine gene expression differences in schizophrenia. Brain research bulletin. 2001;55(5):641-50. 270. Saks VA, Ventura-Clapier R, Aliev MK. Metabolic control and metabolic capacity: two aspects of creatine kinase functioning in the cells. Biochimica et biophysica acta. 1996;1274(3):81-8. 271. Kemp GJ. Non-invasive methods for studying brain energy metabolism: what they show and what it means. Developmental neuroscience. 2000;22(5-6):418-28. 272. Bertolino A, Callicott JH, Elman I, Mattay VS, Tedeschi G, Frank JA, et al. Regionally Specific Neuronal Pathology in Untreated Patients with Schizophrenia: A Proton Magnetic Resonance Spectroscopic Imaging Study. Biological psychiatry. 1998;43(9):641-8. 273. Cecil KM, Lenkinski RE, Gur RE, Gur RC. Proton Magnetic Resonance Spectroscopy in the Frontal and Temporal Lobes of Neuroleptic Naive Patients with Schizophrenia. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 1999;20(2):131-40. 274. Shulman RG, Rothman DL, Behar KL, Hyder F. Energetic basis of brain activity: implications for neuroimaging. Trends in neurosciences. 2004;27(8):489-95. 275. Wilson JE. Isozymes of mammalian hexokinase: structure, subcellular localization and metabolic function. J Exp Biol. 2003;206(Pt 12):2049-57.

209

276. Buchsbaum MS, Shihabuddin L, Hazlett EA, Schroder J, Haznedar MM, Powchik P, et al. Kraepelinian and non-Kraepelinian schizophrenia subgroup differences in cerebral metabolic rate. Schizophrenia research. 2002;55(1-2):25-40. 277. Prince JA, Harro J, Blennow K, Gottfries CG, Oreland L. Putamen mitochondrial energy metabolism is highly correlated to emotional and intellectual impairment in schizophrenics. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. 2000;22(3):284-92. 278. Roberts RC, Roche JK, Conley RR, Lahti AC. Dopaminergic synapses in the caudate of subjects with schizophrenia: relationship to treatment response. Synapse (New York, NY). 2009;63(6):520-30. 279. Sodhi MS, Simmons M, McCullumsmith R, Haroutunian V, Meador-Woodruff JH. Glutamatergic gene expression is specifically reduced in thalamocortical projecting relay neurons in schizophrenia. Biological psychiatry. 2011;70(7):646-54. 280. Smith PK, Krohn RI, Hermanson GT, Mallia AK, Gartner FH, Provenzano MD, et al. Measurement of protein using bicinchoninic acid. Analytical biochemistry. 1985;150(1):76-85. 281. Billiard J, Dennison JB, Briand J, Annan RS, Chai D, Colón M, et al. Quinoline 3-sulfonamides inhibit lactate dehydrogenase A and reverse aerobic glycolysis in cancer cells. Cancer & Metabolism. 2013;1:19-. 282. Xie H, Hanai J, Ren JG, Kats L, Burgess K, Bhargava P, et al. Targeting lactate dehydrogenase--a inhibits tumorigenesis and tumor progression in mouse models of lung cancer and impacts tumor- initiating cells. Cell metabolism. 2014;19(5):795-809. 283. Floridi A, Paggi MG, D'Atri S, De Martino C, Marcante ML, Silvestrini B, et al. Effect of lonidamine on the energy metabolism of Ehrlich ascites tumor cells. Cancer research. 1981;41(11 Pt 1):4661-6. 284. Floridi A, Lehninger AL. Action of the antitumor and antispermatogenic agent lonidamine on electron transport in ehrlich ascites tumor mitochondria. Archives of Biochemistry and Biophysics. 1983;226(1):73-83. 285. Brawer MK. Lonidamine: Basic Science and Rationale for Treatment of Prostatic Proliferative Disorders. Reviews in Urology. 2005;7(Suppl 7):S21-S6. 286. Floridi A, Bruno T, Miccadei S, Fanciulli M, Federico A, Paggi MG. Enhancement of doxorubicin content by the antitumor drug lonidamine in resistant ehrlich ascites tumor cells through modulation of energy metabolism. Biochemical Pharmacology. 1998;56(7):841-9. 287. Floridi A, Paggi MG, Marcante ML, Silvestrini B, Caputo A, De Martino C. Lonidamine, a selective inhibitor of aerobic glycolysis of murine tumor cells. Journal of the National Cancer Institute. 1981;66(3):497-9. 288. Zeevalk GD, Schoepp D, Nicklas WJ. Excitotoxicity at Both NMDA and Non-NMDA Glutamate Receptors Is Antagonized by Aurintricarboxylic Acid: Evidence for Differing Mechanisms of Action. Journal of neurochemistry. 1995;64(4):1749-58. 289. McCune SA, Foe LG, Kemp RG, Jurin RR. Aurintricarboxylic acid is a potent inhibitor of phosphofructokinase. Biochemical Journal. 1989;259(3):925-7. 290. Sinthujaroen P, Wanachottrakul N, Pinkaew D, Petersen JR, Phongdara A, Sheffield-Moore M, et al. Elevation of serum fortilin levels is specific for apoptosis and signifies cell death in vivo. BBA clinical. 2014;2:103-11. 291. Holmgren G, Synnergren J, Bogestal Y, Ameen C, Akesson K, Holmgren S, et al. Identification of novel biomarkers for doxorubicin-induced toxicity in human cardiomyocytes derived from pluripotent stem cells. Toxicology. 2015;328:102-11. 292. Kadri Z, Lefevre C, Goupille O, Penglong T, Granger-Locatelli M, Fucharoen S, et al. Erythropoietin and IGF-1 signaling synchronize cell proliferation and maturation during erythropoiesis. Genes & development. 2015;29(24):2603-16.

210

293. Wang W, Qin Z, Zhu D, Wei Y, Li S, Duan L. Synthesis, Bioactivity Evaluation, and Toxicity Assessment of Novel Salicylanilide Ester Derivatives as Cercaricides against Schistosoma japonicum and Molluscicides against Oncomelania hupensis. Antimicrobial agents and chemotherapy. 2015;60(1):323- 31. 294. Diehl-Jones W, Archibald A, Gordon JW, Mughal W, Hossain Z, Friel JK. Human Milk Fortification Increases Bnip3 Expression Associated With Intestinal Cell Death In Vitro. Journal of pediatric gastroenterology and nutrition. 2015;61(5):583-90. 295. Schriner SE, Coskun V, Hogan SP, Nguyen CT, Lopez TE, Jafari M. Extension of Drosophila Lifespan by Rhodiola rosea Depends on Dietary Carbohydrate and Caloric Content in a Simplified Diet. Journal of medicinal food. 2016;19(3):318-23. 296. Ghosh S, Canugovi C, Yoon JS, Wilson DM, 3rd, Croteau DL, Mattson MP, et al. Partial loss of the DNA repair scaffolding protein, Xrcc1, results in increased brain damage and reduced recovery from ischemic stroke in mice. Neurobiology of aging. 2015;36(7):2319-30. 297. Higgs BW, Elashoff M, Richman S, Barci B. An online database for brain disease research. BMC Genomics. 2006;7:70-. 298. Mistry M, Gillis J, Pavlidis P. Genome-wide expression profiling of schizophrenia using a large combined cohort. Molecular psychiatry. 2013;18(2):215-25. 299. Hashimoto T, Hussien R, Cho H-S, Kaufer D, Brooks GA. Evidence for the Mitochondrial Lactate Oxidation Complex in Rat Neurons: Demonstration of an Essential Component of Brain Lactate Shuttles. PloS one. 2008;3(8):e2915. 300. Brooks GA, Dubouchaud H, Brown M, Sicurello JP, Butz CE. Role of mitochondrial lactate dehydrogenase and lactate oxidation in the intracellular lactate shuttle. Proceedings of the National Academy of Sciences. 1999;96(3):1129-34. 301. McEwen BS, Reagan LP. Glucose transporter expression in the central nervous system: relationship to synaptic function. European journal of pharmacology. 2004;490(1-3):13-24. 302. Liu Y, Liu F, Iqbal K, Grundke-Iqbal I, Gong CX. Decreased glucose transporters correlate to abnormal hyperphosphorylation of tau in Alzheimer disease. FEBS letters. 2008;582(2):359-64. 303. Hoyer S. Causes and consequences of disturbances of cerebral glucose metabolism in sporadic Alzheimer disease: therapeutic implications. Advances in experimental medicine and biology. 2004;541:135-52. 304. Cunnane S, Nugent S, Roy M, Courchesne-Loyer A, Croteau E, Tremblay S, et al. Brain fuel metabolism, aging, and Alzheimer's disease. Nutrition (Burbank, Los Angeles County, Calif). 2011;27(1):3-20. 305. McDermott E, de Silva P. Impaired neuronal glucose uptake in pathogenesis of schizophrenia - can GLUT 1 and GLUT 3 deficits explain imaging, post-mortem and pharmacological findings? Medical hypotheses. 2005;65(6):1076-81. 306. Henneman DH, Altschule MD, Goncz R. Carbohydrate metabolism in brain disease: Ii. glucose metabolism in schizophrenic, manic-depressive, and involutional psychoses. AMA Archives of Internal Medicine. 1954;94(3):402-16. 307. Hashimoto K, Engberg G, Shimizu E, Nordin C, Lindström LH, Iyo M. Elevated glutamine/glutamate ratio in cerebrospinal fluid of first episode and drug naive schizophrenic patients. BMC Psychiatry. 2005;5(1):6. 308. Kenji H, Eiji S, Masaomi I. Dysfunction of Glia-Neuron Communication in Pathophysiology of Schizophrenia. Current Psychiatry Reviews. 2005;1(2):151-63. 309. Abu-Hamad S, Zaid H, Israelson A, Nahon E, Shoshan-Barmatz V. Hexokinase-I protection against apoptotic cell death is mediated via interaction with the voltage-dependent anion channel-1: mapping the site of binding. The Journal of biological chemistry. 2008;283(19):13482-90.

211

310. Regenold WT, Pratt M, Nekkalapu S, Shapiro PS, Kristian T, Fiskum G. Mitochondrial detachment of hexokinase 1 in mood and psychotic disorders: implications for brain energy metabolism and neurotrophic signaling. Journal of psychiatric research. 2012;46(1):95-104. 311. Lindberg D, Shan D, Ayers-Ringler J, Oliveros A, Benitez J, Prieto M, et al. Purinergic signaling and energy homeostasis in psychiatric disorders. Current molecular medicine. 2015;15(3):275-95. 312. BeltrandelRio H, Wilson JE. Coordinated regulation of cerebral glycolytic and oxidative metabolism, mediated by mitochondrially bound hexokinase dependent on intramitochondrially generated ATP. Archives of Biochemistry and Biophysics. 1992;296(2):667-77. 313. Steinman MQ, Gao V, Alberini CM. The Role of Lactate-Mediated Metabolic Coupling between Astrocytes and Neurons in Long-Term Memory Formation. Frontiers in Integrative Neuroscience. 2016;10:10. 314. Harper DG, Jensen JE, Ravichandran C, Perlis RH, Fava M, Renshaw PF, et al. Tissue Type-Specific Bioenergetic Abnormalities in Adults with Major Depression. 2017;42(4):876-85. 315. Duarte JMN, Gruetter R. Glutamatergic and GABAergic energy metabolism measured in the rat brain by 13C NMR spectroscopy at 14.1 T. Journal of neurochemistry. 2013;126(5):579-90. 316. Bubber P, Tang J, Haroutunian V, Xu H, Davis KL, Blass JP, et al. Mitochondrial enzymes in schizophrenia. Journal of molecular neuroscience : MN. 2004;24(2):315-21. 317. Koleti A, Terryn R, Stathias V, Chung C, Cooper DJ, Turner JP, et al. Data Portal for the Library of Integrated Network-based Cellular Signatures (LINCS) program: integrated access to diverse large-scale cellular perturbation response data. Nucleic acids research. 2018;46(D1):D558-d66. 318. Duan Q, Flynn C, Niepel M, Hafner M, Muhlich JL, Fernandez NF, et al. LINCS Canvas Browser: interactive web app to query, browse and interrogate LINCS L1000 gene expression signatures. Nucleic acids research. 2014;42(Web Server issue):W449-W60. 319. Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC bioinformatics. 2013;14:128. 320. Kuleshov MV, Jones MR, Rouillard AD, Fernandez NF, Duan Q, Wang Z, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic acids research. 2016;44(W1):W90-7. 321. Gross DN, Wan M, Birnbaum MJ. The role of FOXO in the regulation of metabolism. Current diabetes reports. 2009;9(3):208-14. 322. Wang Y, Zhou Y, Graves DT. FOXO Transcription Factors: Their Clinical Significance and Regulation. BioMed research international. 2014;2014:925350. 323. Gehart H, Kumpf S, Ittner A, Ricci R. MAPK signalling in cellular metabolism: stress or wellness? EMBO Reports. 2010;11(11):834-40. 324. Sethi JK, Vidal-Puig A. Wnt signalling and the control of cellular metabolism. The Biochemical journal. 2010;427(1):1-17. 325. Nakae J, Kitamura T, Silver DL, Accili D. The forkhead transcription factor Foxo1 (Fkhr) confers insulin sensitivity onto glucose-6-phosphatase expression. Journal of Clinical Investigation. 2001;108(9):1359-67. 326. Zhang W, Patil S, Chauhan B, Guo S, Powell DR, Le J, et al. FoxO1 regulates multiple metabolic pathways in the liver: effects on gluconeogenic, glycolytic, and lipogenic gene expression. The Journal of biological chemistry. 2006;281(15):10105-17. 327. Chafey P, Finzi L, Boisgard R, Cauzac M, Clary G, Broussard C, et al. Proteomic analysis of beta- catenin activation in mouse liver by DIGE analysis identifies glucose metabolism as a new target of the Wnt pathway. Proteomics. 2009;9(15):3889-900. 328. Cadoret A, Ovejero C, Terris B, Souil E, Levy L, Lamers WH, et al. New targets of beta-catenin signaling in the liver are involved in the glutamine metabolism. Oncogene. 2002;21(54):8293-301.

212

329. Rui Y, Xu Z, Lin S, Li Q, Rui H, Luo W, et al. Axin stimulates p53 functions by activation of HIPK2 kinase through multimeric complex formation. The EMBO journal. 2004;23(23):4583-94. 330. Kim NG, Xu C, Gumbiner BM. Identification of targets of the Wnt pathway destruction complex in addition to beta-catenin. Proceedings of the National Academy of Sciences of the United States of America. 2009;106(13):5165-70. 331. Jeong AY, Lee MY, Lee SH, Park JH, Han HJ. PPARdelta agonist-mediated ROS stimulates mouse embryonic stem cell proliferation through cooperation of p38 MAPK and Wnt/beta-catenin. Cell cycle (Georgetown, Tex). 2009;8(4):611-9. 332. Wise DR, DeBerardinis RJ, Mancuso A, Sayed N, Zhang X-Y, Pfeiffer HK, et al. Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proceedings of the National Academy of Sciences of the United States of America. 2008;105(48):18782- 7. 333. Juge-Aubry CE, Hammar E, Siegrist-Kaiser C, Pernin A, Takeshita A, Chin WW, et al. Regulation of the transcriptional activity of the peroxisome proliferator-activated receptor alpha by phosphorylation of a ligand-independent trans-activating domain. The Journal of biological chemistry. 1999;274(15):10505-10. 334. Sozen B, Ozturk S, Yaba A, Demir N. The p38 MAPK signalling pathway is required for glucose metabolism, lineage specification and embryo survival during mouse preimplantation development. Mechanisms of Development. 2015;138:375-98. 335. Newton AC. Protein Kinase C: Structure, Function, and Regulation. Journal of Biological Chemistry. 1995;270(48):28495-8. 336. Tomlinson DR. Mitogen-activated protein kinases as glucose transducers for diabetic complications. Diabetologia. 1999;42(11):1271-81. 337. Marques Pereira P, Schneider A, Pannetier S, Heron D, Hanauer A. Coffin–Lowry syndrome. European Journal Of Human Genetics. 2009;18:627. 338. Thomas GM, Rumbaugh GR, Harrar DB, Huganir RL. Ribosomal S6 kinase 2 interacts with and phosphorylates PDZ domain-containing proteins and regulates AMPA receptor transmission. Proceedings of the National Academy of Sciences of the United States of America. 2005;102(42):15006- 11. 339. Wang Z, Fan M, Candas D, Zhang TQ, Qin L, Eldridge A, et al. Cyclin B1/Cdk1 coordinates mitochondrial respiration for cell-cycle G2/M progression. Developmental cell. 2014;29(2):217-32. 340. Plyte SE, Hughes K, Nikolakaki E, Pulverer BJ, Woodgett JR. Glycogen synthase kinase-3: functions in oncogenesis and development. Biochimica et Biophysica Acta (BBA) - Reviews on Cancer. 1992;1114(2):147-62. 341. Nakayama KI, Nakayama K. Regulation of the cell cycle by SCF-type ubiquitin ligases. Seminars in cell & developmental biology. 2005;16(3):323-33. 342. Nicholson KM, Anderson NG. The protein kinase B/Akt signalling pathway in human malignancy. Cellular signalling. 2002;14(5):381-95. 343. Hajduch E, Litherland GJ, Hundal HS. Protein kinase B (PKB/Akt)--a key regulator of glucose transport? FEBS letters. 2001;492(3):199-203. 344. Emamian ES, Hall D, Birnbaum MJ, Karayiorgou M, Gogos JA. Convergent evidence for impaired AKT1-GSK3beta signaling in schizophrenia. Nature genetics. 2004;36(2):131-7. 345. Mathur A, Law MH, Megson IL, Shaw DJ, Wei J. Genetic association of the AKT1 gene with schizophrenia in a British population. Psychiatric genetics. 2010;20(3):118-22. 346. Karege F, Perroud N, Schurhoff F, Meary A, Marillier G, Burkhardt S, et al. Association of AKT1 gene variants and protein expression in both schizophrenia and bipolar disorder. Genes, brain, and behavior. 2010;9(5):503-11.

213

347. Zachary I, Rozengurt E. Focal adhesion kinase (p125FAK): a point of convergence in the action of neuropeptides, integrins, and oncogenes. Cell. 1992;71(6):891-4. 348. Shapiro P. Ras-MAP kinase signaling pathways and control of cell proliferation: relevance to cancer therapy. Critical reviews in clinical laboratory sciences. 2002;39(4-5):285-330. 349. Maréchal A, Zou L. DNA Damage Sensing by the ATM and ATR Kinases. Cold Spring Harbor Perspectives in Biology. 2013;5(9):a012716. 350. Guan X-M, Yu H, Jiang Q, Van der Ploeg LHT, Liu Q. Distribution of neuromedin U receptor subtype 2 mRNA in the rat brain☆☆Published on the World Wide Web on 27 November 2000. Gene Expression Patterns. 2001;1(1):1-4. 351. Nakazato M, Hanada R, Murakami N, Date Y, Mondal MS, Kojima M, et al. Central Effects of Neuromedin U in the Regulation of Energy Homeostasis. Biochemical and Biophysical Research Communications. 2000;277(1):191-4. 352. Nakamoto K, Nishinaka T, Matsumoto K, Kasuya F, Mankura M, Koyama Y, et al. Involvement of the long-chain fatty acid receptor GPR40 as a novel pain regulatory system. Brain research. 2012;1432:74-83. 353. Kebede M, Ferdaoussi M, Mancini A, Alquier T, Kulkarni RN, Walker MD, et al. Glucose activates free fatty acid receptor 1 gene transcription via phosphatidylinositol-3-kinase-dependent O- GlcNAcylation of pancreas-duodenum homeobox-1. Proceedings of the National Academy of Sciences of the United States of America. 2012;109(7):2376-81. 354. Hamilton TB, Barilla KC, Romaniuk PJ. High affinity binding sites for the Wilms' tumour suppressor protein WT1. Nucleic acids research. 1995;23(2):277-84. 355. Steele-Perkins G, Plachez C, Butz KG, Yang G, Bachurski CJ, Kinsman SL, et al. The Transcription Factor Gene Nfib Is Essential for both Lung Maturation and Brain Development. Molecular and cellular biology. 2005;25(2):685-98. 356. Ikawa T, Kawamoto H, Goldrath AW, Murre C. E proteins and Notch signaling cooperate to promote T cell lineage specification and commitment. The Journal of Experimental Medicine. 2006;203(5):1329-42. 357. Kuwahara A, Sakai H, Xu Y, Itoh Y, Hirabayashi Y, Gotoh Y. Tcf3 Represses Wnt–β-Catenin Signaling and Maintains Neural Stem Cell Population during Neocortical Development. PloS one. 2014;9(5):e94408. 358. Iyengar S, Farnham PJ. KAP1 Protein: An Enigmatic Master Regulator of the Genome. Journal of Biological Chemistry. 2011;286(30):26267-76. 359. Jakobsson J, Cordero MI, Bisaz R, Groner AC, Busskamp V, Bensadoun J-C, et al. KAP1-Mediated Epigenetic Repression in the Forebrain Modulates Behavioral Vulnerability to Stress. Neuron. 2008;60(5):818-31. 360. Pal R, Janz M, Galson DL, Gries M, Li S, Johrens K, et al. C/EBPbeta regulates transcription factors critical for proliferation and survival of multiple myeloma cells. Blood. 2009;114(18):3890-8. 361. Ruffell D, Mourkioti F, Gambardella A, Kirstetter P, Lopez RG, Rosenthal N, et al. A CREB-C/EBPβ cascade induces M2 macrophage-specific gene expression and promotes muscle injury repair. Proceedings of the National Academy of Sciences of the United States of America. 2009;106(41):17475- 80. 362. Stegh AH. Targeting the p53 signaling pathway in cancer therapy - The promises, challenges, and perils. Expert Opinion on Therapeutic Targets. 2012;16(1):67-83. 363. Iannacone M, Sitia G, Ruggeri ZM, Guidotti LG. HBV PATHOGENESIS IN ANIMAL MODELS: RECENT ADVANCES ON THE ROLE OF PLATELETS. Journal of hepatology. 2007;46(4):719-26. 364. Kelly RD, Cowley SM. The physiological roles of histone deacetylase (HDAC) 1 and 2: complex co- stars with multiple leading parts. Biochemical Society transactions. 2013;41(3):741-9. 365. Kouzarides T. Chromatin modifications and their function. Cell. 2007;128(4):693-705.

214

366. Moldovan G-L, Pfander B, Jentsch S. PCNA, the Maestro of the Replication Fork. Cell.129(4):665- 79. 367. Luo J, Su F, Chen D, Shiloh A, Gu W. Deacetylation of p53 modulates its effect on cell growth and apoptosis. Nature. 2000;408:377. 368. Jakovcevski M, Bharadwaj R, Straubhaar J, Gao G, Gavin DP, Jakovcevski I, et al. Prefrontal cortical dysfunction after overexpression of histone deacetylase 1. Biological psychiatry. 2013;74(9):696- 705. 369. Narayan S, Tang B, Head SR, Gilmartin TJ, Sutcliffe JG, Dean B, et al. Molecular Profiles of Schizophrenia in the CNS at Different Stages of Illness. Brain research. 2008;1239:235-48. 370. Sharma RP, Grayson DR, Gavin DP. Histone deactylase 1 expression is increased in the prefrontal cortex of schizophrenia subjects: analysis of the National Brain Databank microarray collection. Schizophrenia research. 2008;98(1-3):111-7. 371. Guan J-S, Haggarty SJ, Giacometti E, Dannenberg J-H, Joseph N, Gao J, et al. HDAC2 negatively regulates memory formation and synaptic plasticity. Nature. 2009;459(7243):55-60. 372. Akbarian S. Epigenetic mechanisms in schizophrenia. Dialogues in Clinical Neuroscience. 2014;16(3):405-17. 373. Luo X, Huang L, Jia P, Li M, Su B, Zhao Z, et al. Protein-Protein Interaction and Pathway Analyses of Top Schizophrenia Genes Reveal Schizophrenia Susceptibility Genes Converge on Common Molecular Networks and Enrichment of Nucleosome (Chromatin) Assembly Genes in Schizophrenia Susceptibility Loci. Schizophrenia bulletin. 2014;40(1):39-49. 374. Gilman SR, Chang J, Xu B, Bawa TS, Gogos JA, Karayiorgou M, et al. Diverse types of genetic variation converge on functional gene networks involved in schizophrenia. Nature neuroscience. 2012;15(12):1723-8. 375. Ogryzko VV, Schiltz RL, Russanova V, Howard BH, Nakatani Y. The transcriptional coactivators p300 and CBP are histone acetyltransferases. Cell. 1996;87(5):953-9. 376. Goto NK, Zor T, Martinez-Yamout M, Dyson HJ, Wright PE. Cooperativity in transcription factor binding to the coactivator CREB-binding protein (CBP). The mixed lineage leukemia protein (MLL) activation domain binds to an allosteric site on the KIX domain. The Journal of biological chemistry. 2002;277(45):43168-74. 377. Waltzer L, Bienz M. Drosophila CBP represses the transcription factor TCF to antagonize Wingless signalling. Nature. 1998;395:521. 378. Martínez‐Balbás MA, Bauer UM, Nielsen SJ, Brehm A, Kouzarides T. Regulation of E2F1 activity by acetylation. The EMBO journal. 2000;19(4):662. 379. Avantaggiati ML, Ogryzko V, Gardner K, Giordano A, Levine AS, Kelly K. Recruitment of p300/CBP in p53-Dependent Signal Pathways. Cell. 1997;89(7):1175-84. 380. Rajabi HN, Baluchamy S, Kolli S, Nag A, Srinivas R, Raychaudhuri P, et al. Effects of Depletion of CREB-binding Protein on c-Myc Regulation and Cell Cycle G1-S Transition. Journal of Biological Chemistry. 2005;280(1):361-74. 381. Campisi J, Gray HE, Pardee AB, Dean M, Sonenshein GE. Cell-cycle control of c-myc but not c-ras expression is lost following chemical transformation. Cell. 1984;36(2):241-7. 382. Denis N, Kitzis A, Kruh J, Dautry F, Corcos D. Stimulation of methotrexate resistance and dihydrofolate reductase gene amplification by c-myc. Oncogene. 1991;6(8):1453-7. 383. de Alboran IM, O'Hagan RC, Gartner F, Malynn B, Davidson L, Rickert R, et al. Analysis of C-MYC function in normal cells via conditional gene-targeted mutation. Immunity. 2001;14(1):45-55. 384. Di Maira G, Salvi M, Arrigoni G, Marin O, Sarno S, Brustolon F, et al. Protein kinase CK2 phosphorylates and upregulates Akt/PKB. Cell Death And Differentiation. 2005;12:668.

215

385. Song DH, Dominguez I, Mizuno J, Kaut M, Mohr SC, Seldin DC. CK2 Phosphorylation of the Armadillo Repeat Region of β-Catenin Potentiates Wnt Signaling. Journal of Biological Chemistry. 2003;278(26):24018-25. 386. Litchfield DW. Protein kinase CK2: structure, regulation and role in cellular decisions of life and death. Biochemical Journal. 2003;369(Pt 1):1-15. 387. McKendrick L, Milne D, Meek D. Protein kinase CK2-dependent regulation of p53 function: Evidence that the phosphorylation status of the serine 386 (CK2) site of p53 is constitutive and stable. In: Ahmed K, Issinger OG, Chambaz E, editors. A Molecular and Cellular View of Protein Kinase CK2. Boston, MA: Springer US; 1999. p. 187-99. 388. Chakraborty A, Werner JK, Koldobskiy MA, Mustafa AK, Juluri KR, Pietropaoli J, et al. Casein kinase-2 mediates cell survival through phosphorylation and degradation of inositol hexakisphosphate kinase-2. Proceedings of the National Academy of Sciences. 2011;108(6):2205. 389. May MJ, Ghosh S. Signal transduction through NF-kappa B. Immunology today. 1998;19(2):80-8. 390. Duncan JA, Reeves JR, Cooke TG. BRCA1 and BRCA2 proteins: roles in health and disease. Molecular pathology : MP. 1998;51(5):237-47. 391. Brue T, Quentien M-H, Khetchoumian K, Bensa M, Capo-Chichi J-M, Delemer B, et al. Mutations in NFKB2 and potential genetic heterogeneity in patients with DAVID syndrome, having variable endocrine and immune deficiencies. BMC Medical Genetics. 2014;15:139. 392. Chen K, Coonrod Emily M, Kumánovics A, Franks ZF, Durtschi Jacob D, Margraf Rebecca L, et al. Germline Mutations in NFKB2 Implicate the Noncanonical NF-κB Pathway in the Pathogenesis of Common Variable Immunodeficiency. The American Journal of Human Genetics.93(5):812-24. 393. Hoesel B, Schmid JA. The complexity of NF-κB signaling in inflammation and cancer. Molecular Cancer. 2013;12(1):86. 394. Webster GA, Perkins ND. Transcriptional cross talk between NF-kappaB and p53. Molecular and cellular biology. 1999;19(5):3485-95. 395. Ashburner BP, Westerheide SD, Baldwin AS. The p65 (RelA) Subunit of NF-κB Interacts with the Histone Deacetylase (HDAC) Corepressors HDAC1 and HDAC2 To Negatively Regulate Gene Expression. Molecular and cellular biology. 2001;21(20):7065-77. 396. CHAPMAN NR, WEBSTER GA, GILLESPIE PJ, WILSON BJ, CROUCH DH, PERKINS ND. A novel form of the RelA nuclear factor κB subunit is induced by and forms a complex with the proto-oncogene c-Myc. Biochemical Journal. 2002;366(2):459-69. 397. Gerritsen ME, Williams AJ, Neish AS, Moore S, Shi Y, Collins T. CREB-binding protein/p300 are transcriptional coactivators of p65. Proceedings of the National Academy of Sciences of the United States of America. 1997;94(7):2927-32. 398. Chandra V, Huang P, Hamuro Y, Raghuram S, Wang Y, Burris TP, et al. Structure of the intact PPAR-γ–RXR-α nuclear receptor complex on DNA. Nature. 2008;456(7220):350-6. 399. Lehmann JM, Moore LB, Smith-Oliver TA, Wilkison WO, Willson TM, Kliewer SA. An Antidiabetic Thiazolidinedione Is a High Affinity Ligand for Peroxisome Proliferator-activated Receptor γ (PPARγ). Journal of Biological Chemistry. 1995;270(22):12953-6. 400. Remels AH, Langen RC, Gosker HR, Russell AP, Spaapen F, Voncken JW, et al. PPARgamma inhibits NF-kappaB-dependent transcriptional activation in skeletal muscle. Am J Physiol Endocrinol Metab. 2009;297(1):E174-83. 401. Mao H, Ferguson TS, Cibulsky SM, Holmqvist M, Ding C, Fei H, et al. MONaKA, a Novel Modulator of the Plasma Membrane Na,K-ATPase. The Journal of Neuroscience. 2005;25(35):7934. 402. Cao C, Leng Y, Li C, Kufe D. Functional Interaction between the c-Abl and Arg Protein-tyrosine Kinases in the Oxidative Stress Response. Journal of Biological Chemistry. 2003;278(15):12961-7.

216

403. Tourette C, Farina F, Vazquez-Manrique RP, Orfila A-M, Voisin J, Hernandez S, et al. The Wnt Receptor Ryk Reduces Neuronal and Cell Survival Capacity by Repressing FOXO Activity During the Early Phases of Mutant Huntingtin Pathogenicity. PLOS Biology. 2014;12(6):e1001895. 404. Hollis ER, Ishiko N, Yu T, Lu C-C, Haimovich A, Tolentino K, et al. Ryk controls remapping of motor cortex during functional recovery after spinal cord injury. Nature neuroscience. 2016;19(5):697-705. 405. Colicelli J. ABL Tyrosine Kinases: Evolution of Function, Regulation, and Specificity. Science signaling. 2010;3(139):re6-re. 406. Silva J-P, Ushkaryov YA. The latrophilins, “split-personality” receptors. Advances in experimental medicine and biology. 2010;706:59-75. 407. McGowan CH, Russell P. Cell cycle regulation of human WEE1. The EMBO journal. 1995;14(10):2166-75. 408. Rhen T, Cidlowski JA. Antiinflammatory action of glucocorticoids--new mechanisms for old drugs. The New England journal of medicine. 2005;353(16):1711-23. 409. Ray A, Prefontaine KE. Physical association and functional antagonism between the p65 subunit of transcription factor NF-kappa B and the glucocorticoid receptor. Proceedings of the National Academy of Sciences of the United States of America. 1994;91(2):752-6. 410. DeGregori J, Leone G, Miron A, Jakoi L, Nevins JR. Distinct roles for E2F proteins in cell growth control and apoptosis. Proceedings of the National Academy of Sciences of the United States of America. 1997;94(14):7245-50. 411. Muller N. Inflammation and the glutamate system in schizophrenia: implications for therapeutic targets and drug development. Expert Opin Ther Targets. 2008;12(12):1497-507. 412. Muller N, Myint AM, Schwarz MJ. Inflammation in schizophrenia. Adv Protein Chem Struct Biol. 2012;88:49-68. 413. Müller N, Schwarz MJ. Immune System and Schizophrenia. Current immunology reviews. 2010;6(3):213-20. 414. Gandal MJ, Haney JR, Parikshak NN, Leppa V, Ramaswami G, Hartl C, et al. Shared molecular neuropathology across major psychiatric disorders parallels polygenic overlap. Science (New York, NY). 2018;359(6376):693. 415. Kurtz J-E, Ray-Coquard I. PI3 Kinase Inhibitors in the Clinic: An Update. Anticancer Research. 2012;32(7):2463-70. 416. Crabbe T. Exploring the potential of PI3K inhibitors for inflammation and cancer. Biochemical Society transactions. 2007;35(Pt 2):253-6. 417. Ito K, Caramori G, Adcock IM. Therapeutic potential of phosphatidylinositol 3-kinase inhibitors in inflammatory respiratory disease. The Journal of pharmacology and experimental therapeutics. 2007;321(1):1-8. 418. Gururajan A, van den Buuse M. Is the mTOR-signalling cascade disrupted in Schizophrenia? Journal of neurochemistry. 2014;129(3):377-87. 419. Kalkman HO. The role of the phosphatidylinositide 3-kinase-protein kinase B pathway in schizophrenia. Pharmacology & therapeutics. 2006;110(1):117-34. 420. Enriquez-Barreto L, Morales M. The PI3K signaling pathway as a pharmacological target in Autism related disorders and Schizophrenia. Molecular and Cellular Therapies. 2016;4:2. 421. Kalkman HO. The role of the phosphatidylinositide 3-kinase–protein kinase B pathway in schizophrenia. Pharmacology & therapeutics. 2006;110(1):117-34. 422. Law AJ, Wang Y, Sei Y, O'Donnell P, Piantadosi P, Papaleo F, et al. Neuregulin 1-ErbB4-PI3K signaling in schizophrenia and phosphoinositide 3-kinase-p110delta inhibition as a potential therapeutic strategy. Proceedings of the National Academy of Sciences of the United States of America. 2012;109(30):12165-70.

217

423. Cheatham B, Vlahos CJ, Cheatham L, Wang L, Blenis J, Kahn CR. Phosphatidylinositol 3-kinase activation is required for insulin stimulation of pp70 S6 kinase, DNA synthesis, and glucose transporter translocation. Molecular and cellular biology. 1994;14(7):4902-11. 424. Dixon LB, Lehman AF, Levine J. Conventional antipsychotic medications for schizophrenia. Schizophrenia bulletin. 1995;21(4):567-77. 425. Tsygankov BD, Agasarain EG, Zykova AS. [Antipsychotic drugs and their influence on the carbohydrate metabolism in patients with schizophrenia-spectrum disorders]. Zhurnal nevrologii i psikhiatrii imeni SS Korsakova. 2014;114(5):86-91. 426. Davies JA. Mechanisms of action of antiepileptic drugs. Seizure. 1995;4(4):267-71. 427. Citrome L. Schizophrenia and valproate. Psychopharmacology bulletin. 2003;37 Suppl 2:74-88. 428. Schwarz C, Volz A, Li C, Leucht S. Valproate for schizophrenia. The Cochrane database of systematic reviews. 2008(3):Cd004028. 429. Rosenberg G. The mechanisms of action of valproate in neuropsychiatric disorders: can we see the forest for the trees? Cellular and Molecular Life Sciences. 2007;64(16):2090-103. 430. Wang L, Waltenberger B, Pferschy-Wenzig E-M, Blunder M, Liu X, Malainer C, et al. Natural product agonists of peroxisome proliferator-activated receptor gamma (PPARγ): a review. Biochemical Pharmacology. 2014;92(1):73-89. 431. Aljada A, Garg R, Ghanim H, Mohanty P, Hamouda W, Assian E, et al. Nuclear factor-kappaB suppressive and inhibitor-kappaB stimulatory effects of troglitazone in obese patients with type 2 diabetes: evidence of an antiinflammatory action? The Journal of clinical endocrinology and metabolism. 2001;86(7):3250-6. 432. Henry RR. Effects of troglitazone on insulin sensitivity. Diabetic medicine : a journal of the British Diabetic Association. 1996;13(9 Suppl 6):S148-50. 433. #xe9, rez M, #xed, Jos a, #xe9, Quintanilla RA. Therapeutic Actions of the Thiazolidinediones in Alzheimer’s Disease. PPAR Research. 2015;2015:8. 434. Jiang L-Y, Tang S-S, Wang X-Y, Liu L-P, Long Y, Hu M, et al. PPARγ Agonist Pioglitazone Reverses Memory Impairment and Biochemical Changes in a Mouse Model of Type 2 Diabetes Mellitus. CNS Neuroscience & Therapeutics. 2012;18(8):659-66. 435. Heneka MT, Fink A, Doblhammer G. Effect of pioglitazone medication on the incidence of dementia. Annals of Neurology. 2015;78(2):284-94. 436. Masciopinto F, Di Pietro N, Corona C, Bomba M, Pipino C, Curcio M, et al. Effects of long-term treatment with pioglitazone on cognition and glucose metabolism of PS1-KI, 3xTg-AD, and wild-type mice. Cell Death Dis. 2012;3:e448. 437. Papadopoulos P, Rosa-Neto P, Rochford J, Hamel E. Pioglitazone Improves Reversal Learning and Exerts Mixed Cerebrovascular Effects in a Mouse Model of Alzheimer’s Disease with Combined Amyloid- β and Cerebrovascular Pathology. PloS one. 2013;8(7):e68612. 438. Sauerbeck A, Gao J, Readnower R, Liu M, Pauly JR, Bing G, et al. Pioglitazone attenuates mitochondrial dysfunction, cognitive impairment, cortical tissue loss, and inflammation following traumatic brain injury. Experimental Neurology. 2011;227(1):128-35. 439. Allen R, Young S. Phencyclidine-induced psychosis. The American journal of psychiatry. 1978;135(9):1081-4. 440. Javitt DC. Glutamate as a therapeutic target in psychiatric disorders. Molecular Psychiatry. 2004;9:984. 441. Kantrowitz JT, Javitt DC. N-methyl-d-aspartate (NMDA) receptor dysfunction or dysregulation: the final common pathway on the road to schizophrenia? Brain research bulletin. 2010;83(3-4):108-21. 442. Begni S, Moraschi S, Bignotti S, Fumagalli F, Rillosi L, Perez J, et al. Association between the G1001C polymorphism in the GRIN1 gene promoter region and schizophrenia. Biological psychiatry. 2003;53(7):617-9.

218

443. Georgi A, Jamra RA, Klein K, Villela AW, Schumacher J, Becker T, et al. Possible association between genetic variants at the GRIN1 gene and schizophrenia with lifetime history of depressive symptoms in a German sample. Psychiatric genetics. 2007;17(5):308-10. 444. Galehdari H, Pooryasin A, Foroughmand A, Daneshmand S, Saadat M. Association between the G1001C polymorphism in the GRIN1 gene promoter and schizophrenia in the Iranian population. Journal of molecular neuroscience : MN. 2009;38(2):178-81. 445. Itokawa M, Yamada K, Yoshitsugu K, Toyota T, Suga T, Ohba H, et al. A microsatellite repeat in the promoter of the N-methyl-D-aspartate receptor 2A subunit (GRIN2A) gene suppresses transcriptional activity and correlates with chronic outcome in schizophrenia. Pharmacogenetics. 2003;13(5):271-8. 446. Iwayama-Shigeno Y, Yamada K, Itokawa M, Toyota T, Meerabux JM, Minabe Y, et al. Extended analyses support the association of a functional (GT)n polymorphism in the GRIN2A promoter with Japanese schizophrenia. Neuroscience letters. 2005;378(2):102-5. 447. Tang J, Chen X, Xu X, Wu R, Zhao J, Hu Z, et al. Significant linkage and association between a functional (GT)n polymorphism in promoter of the N-methyl-D-aspartate receptor subunit gene (GRIN2A) and schizophrenia. Neuroscience letters. 2006;409(1):80-2. 448. Qin S, Zhao X, Pan Y, Liu J, Feng G, Fu J, et al. An association study of the N-methyl-D-aspartate receptor NR1 subunit gene (GRIN1) and NR2B subunit gene (GRIN2B) in schizophrenia with universal DNA microarray. European journal of human genetics : EJHG. 2005;13(7):807-14. 449. Martucci L, Wong AH, De Luca V, Likhodi O, Wong GW, King N, et al. N-methyl-D-aspartate receptor NR2B subunit gene GRIN2B in schizophrenia and bipolar disorder: Polymorphisms and mRNA levels. Schizophrenia research. 2006;84(2-3):214-21. 450. Demontis D, Nyegaard M, Buttenschon HN, Hedemand A, Pedersen CB, Grove J, et al. Association of GRIN1 and GRIN2A-D with schizophrenia and genetic interaction with maternal herpes simplex virus-2 infection affecting disease risk. American journal of medical genetics Part B, Neuropsychiatric genetics : the official publication of the International Society of Psychiatric Genetics. 2011;156b(8):913-22. 451. Gandal MJ, Sisti J, Klook K, Ortinski PI, Leitman V, Liang Y, et al. GABAB-mediated rescue of altered excitatory-inhibitory balance, gamma synchrony and behavioral deficits following constitutive NMDAR-hypofunction. Translational psychiatry. 2012;2:e142. 452. Seto SW, Yan Lam T, Leung G, L.S. Au A, Ngai SM, Wan Chan S, et al. Comparison of vascular relaxation, lipolysis and glucose uptake by peroxisome proliferator-activated receptor-γ activation in + db/+ m and + db/+ db mice2007. 40-8 p. 453. Colca JR, McDonald WG, Waldon DJ, Leone JW, Lull JM, Bannow CA, et al. Identification of a novel mitochondrial protein ("mitoNEET") cross-linked specifically by a thiazolidinedione photoprobe. Am J Physiol Endocrinol Metab. 2004;286(2):E252-60. 454. Paddock ML, Wiley SE, Axelrod HL, Cohen AE, Roy M, Abresch EC, et al. MitoNEET is a uniquely folded 2Fe 2S outer mitochondrial membrane protein stabilized by pioglitazone. Proceedings of the National Academy of Sciences of the United States of America. 2007;104(36):14342-7. 455. Hwang J, Kleinhenz DJ, Rupnow HL, Campbell AG, Thule PM, Sutliff RL, et al. The PPARgamma ligand, rosiglitazone, reduces vascular oxidative stress and NADPH oxidase expression in diabetic mice. Vascular pharmacology. 2007;46(6):456-62. 456. Wright MB, Bortolini M, Tadayyon M, Bopst M. Minireview: Challenges and Opportunities in Development of PPAR Agonists. Molecular Endocrinology. 2014;28(11):1756-68. 457. Ramsey AJ, Milenkovic M, Oliveira AF, Escobedo-Lozoya Y, Seshadri S, Salahpour A, et al. Impaired NMDA receptor transmission alters striatal synapses and DISC1 protein in an age-dependent manner. Proc Natl Acad Sci U S A. 2011;108(14):5795-800.

219

458. Geyer MA, McIlwain KL, Paylor R. Mouse genetic models for prepulse inhibition: an early review. Mol Psychiatry. 2002;7(10):1039-53. 459. Milenkovic M, Mielnik CA, Ramsey AJ. NMDA receptor-deficient mice display sexual dimorphism in the onset and severity of behavioural abnormalities. Genes, brain, and behavior. 2014;13(8):850-62. 460. Moy SS, Nadler JJ, Young NB, Perez A, Holloway LP, Barbaro RP, et al. Mouse behavioral tasks relevant to autism: phenotypes of 10 inbred strains. Behav Brain Res. 2007;176(1):4-20. 461. Mielnik CA, Horsfall W, Ramsey AJ. Diazepam improves aspects of social behaviour and neuron activation in NMDA receptor-deficient mice. Genes, brain, and behavior. 2014;13(7):592-602. 462. Courtney Sullivan RK, Kathryn Hasselfeld, Sinead O'Donovan, Amy Ramsey, Robert McCullumsmith. Neuron-specific deficits of neuroenergetic processes in the dorsolateral prefrontal cortex in schizophrenia. Molecular Psychiatry. 2018;in press. 463. Duncan GE, Moy SS, Lieberman JA, Koller BH. Typical and atypical antipsychotic drug effects on locomotor hyperactivity and deficits in sensorimotor gating in a genetic model of NMDA receptor hypofunction. Pharmacol Biochem Behav. 2006;85(3):481-91. 464. Fradley RL, O'Meara GF, Newman RJ, Andrieux A, Job D, Reynolds DS. STOP knockout and NMDA NR1 hypomorphic mice exhibit deficits in sensorimotor gating. Behav Brain Res. 2005;163(2):257-64. 465. Halene TB, Ehrlichman RS, Liang Y, Christian EP, Jonak GJ, Gur TL, et al. Assessment of NMDA receptor NR1 subunit hypofunction in mice as a model for schizophrenia. Genes Brain Behav. 2009;8(7):661-75. 466. Lee MA, Jayathilake K, Meltzer HY. A comparison of the effect of clozapine with typical neuroleptics on cognitive function in neuroleptic-responsive schizophrenia. Schizophrenia research. 1999;37(1):1-11. 467. Bhattacharya K. Cognitive Function in Schizophrenia: A Review. J Psychiatry Neurosci. 2015;18(187). 468. Baddeley A, Della Sala S, Papagno C, Spinnler H. Dual-task performance in dysexecutive and nondysexecutive patients with a frontal lesion. Neuropsychology. 1997;11(2):187-94. 469. Miyake A, Friedman NP, Emerson MJ, Witzki AH, Howerter A, Wager TD. The unity and diversity of executive functions and their contributions to complex "Frontal Lobe" tasks: a latent variable analysis. Cognitive psychology. 2000;41(1):49-100. 470. Rabin RA, Zakzanis KK, George TP. The effects of cannabis use on neurocognition in schizophrenia: a meta-analysis. Schizophrenia research. 2011;128(1-3):111-6. 471. Yücel M, Bora E, Lubman DI, Solowij N, Brewer WJ, Cotton SM, et al. The Impact of Cannabis Use on Cognitive Functioning in Patients With Schizophrenia: A Meta-analysis of Existing Findings and New Data in a First-Episode Sample. Schizophrenia bulletin. 2012;38(2):316-30. 472. Goldberg TE, Green MF. Neurocognitive functioning in patients with schizophrenia: an overview; in Neuropsychopharmacology - fifth generation of progress (eds) KL Davis, D Charney, JT Coyle2002. 657-69 p. 473. Gao R, Penzes P. Common Mechanisms of Excitatory and Inhibitory Imbalance in Schizophrenia and Autism Spectrum Disorders. Current molecular medicine. 2015;15(2):146-67. 474. Beasley CL, Reynolds GP. Parvalbumin-immunoreactive neurons are reduced in the prefrontal cortex of schizophrenics. Schizophrenia research. 1997;24(3):349-55. 475. Fung SJ, Webster MJ, Sivagnanasundaram S, Duncan C, Elashoff M, Weickert CS. Expression of interneuron markers in the dorsolateral prefrontal cortex of the developing human and in schizophrenia. The American journal of psychiatry. 2010;167(12):1479-88. 476. Mody I, Pearce RA. Diversity of inhibitory neurotransmission through GABA(A) receptors. Trends in neurosciences. 2004;27(9):569-75.

220

477. Hashimoto T, Volk DW, Eggan SM, Mirnics K, Pierri JN, Sun Z, et al. Gene expression deficits in a subclass of GABA neurons in the prefrontal cortex of subjects with schizophrenia. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2003;23(15):6315-26. 478. Carder RK, Leclerc SS, Hendry SHC. Regulation of Calcium-Binding Protein Immunoreactivity in GABA Neurons of Macaque Primary Visual Cortex. Cerebral Cortex. 1996;6(2):271-87. 479. Doherty JL, Owen MJ. Genomic insights into the overlap between psychiatric disorders: implications for research and clinical practice. Genome Medicine. 2014;6(4):29-. 480. Buchsbaum MS, Hazlett EA. Positron emission tomography studies of abnormal glucose metabolism in schizophrenia. Schizophrenia bulletin. 1998;24(3):343-64. 481. Videbech P. PET measurements of brain glucose metabolism and blood flow in major depressive disorder: a critical review. Acta Psychiatr Scand. 2000;101. 482. Abdallah CG, Jiang L, De Feyter HM, Fasula M, Krystal JH, Rothman DL, et al. Glutamate metabolism in major depressive disorder. The American journal of psychiatry. 2014;171(12):1320-7. 483. Gardner A, Johansson A, Wibom R, Nennesmo I, Dobeln U, Hagenfeldt L. Alterations of mitochondrial function and correlations with personality traits in selected major depressive disorder patients. J Affect Disord. 2003;76. 484. Dager SR, Friedman SD, Parow A, Demopulos C, Stoll AL, Lyoo IK, et al. Brain metabolic alterations in medication-free patients with bipolar disorder. Archives of general psychiatry. 2004;61(5):450-8. 485. Shao L, Martin MV, Watson SJ, Schatzberg A, Akil H, Myers RM, et al. Mitochondrial involvement in psychiatric disorders. Ann Med. 2008;40(4):281-95. 486. Zuccoli GS, Saia-Cereda VM, Nascimento JM, Martins-de-Souza D. The Energy Metabolism Dysfunction in Psychiatric Disorders Postmortem Brains: Focus on Proteomic Evidence. Frontiers in Neuroscience. 2017;11:493. 487. Gur RE, Resnick SM, Gur RC, Alavi A, Caroff S, Kushner M, et al. Regional brain function in schizophrenia. II. Repeated evaluation with positron emission tomography. Archives of general psychiatry. 1987;44(2):126-9. 488. Siegel BV, Jr., Buchsbaum MS, Bunney WE, Jr., Gottschalk LA, Haier RJ, Lohr JB, et al. Cortical- striatal-thalamic circuits and brain glucose metabolic activity in 70 unmedicated male schizophrenic patients. The American journal of psychiatry. 1993;150(9):1325-36. 489. Harrison NA, Doeller CF, Voon V, Burgess N, Critchley HD. Peripheral inflammation acutely impairs human spatial memory via actions on medial temporal lobe glucose metabolism. Biological psychiatry. 2014;76(7):585-93. 490. Khandaker GM, Cousins L, Deakin J, Lennox BR, Yolken R, Jones PB. Inflammation and immunity in schizophrenia: implications for pathophysiology and treatment. The lancet Psychiatry. 2015;2(3):258- 70. 491. Kapadia R, Yi J-H, Vemuganti R. Mechanisms of anti-inflammatory and neuroprotective actions of PPAR-gamma agonists. Frontiers in bioscience : a journal and virtual library. 2008;13:1813-26. 492. Youssef J, Badr M. Role of Peroxisome Proliferator-Activated Receptors in Inflammation Control. Journal of Biomedicine and Biotechnology. 2004;2004(3):156-66. 493. Jin J, Maren S. Prefrontal-Hippocampal Interactions in Memory and Emotion. Frontiers in Systems Neuroscience. 2015;9:170. 494. Heckers S. Neuroimaging studies of the hippocampus in schizophrenia. Hippocampus. 2001;11(5):520-8. 495. A Bachneff S. Regional Cerebral Blood Flow in Schizophrenia and the Local Circuit Neurons Hypothesis1996. 163-82 p. 496. Kato TA, Hyodo F, Yamato M, Utsumi H, Kanba S. [Redox and microglia in the pathophysiology of schizophrenia]. Yakugaku zasshi : Journal of the Pharmaceutical Society of Japan. 2015;135(5):739-43.

221

497. Ghosh S, Castillo E, Frias ES, Swanson RA. Bioenergetic regulation of microglia. 2017.

222