Feature-Expression Heat Maps 51

University of Groningen

The Dysregulated Brain Haarman, Bartholomeus Cornelius Maria

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2017

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Haarman, B. C. M. (2017). The Dysregulated Brain: A psychoimmunological approach to bipolar disorder. University of Groningen.

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license. More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne- amendment.

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 04-10-2021 The Dysregulated Brain A psychoimmunological approach to bipolar disorder B.C.M. Haarman The Dysregulated Brain A psychoimmunological approach to bipolar disorder

Illustration cover Alan Flemming Layout Marieke van der Vliet / dit-s.nl Printed by Pumbo.nl

The studies described in this thesis were performed at the Department of Psychiatry, University Medical Center Groningen, Groningen, The Netherlands; Department of Immunology, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands; and Radiology Morphological Solutions, Berkel en Rodenrijs, The Netherlands.

The studies were financially supported by the European Commission EU-FP7-HEALTH-222963 ‘MOODINFLAME’ and EU-FP7-PEOPLE- 286334 ‘PSYCHAID’.

ISBN 978-90-824638-1-1 (printed version) 978-90-824638-2-8 (digital version)

© 2017 B.C.M. Haarman. All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without written permission from the author. The Dysregulated Brain

A psychoimmunological approach to bipolar disorder

PhD thesis

to obtain the degree of PhD at the University of Groningen on the authority of the Rector Magnificus Prof. E. Sterken and in accordance with the decision by the College of Deans.

This thesis will be defended in public on

1 March 2017 at 16:15 hours

Bartholomeus Cornelius Maria Haarman

born on 7 August 1978 in Raalte Supervisor(s) Prof. W.A. Nolen Prof. H.A. Drexhage

Co-supervisor(s) Dr. R.F. Riemersma – Van der Lek Dr. H. Burger

Assessment committee Prof. A. Aleman Prof. V. Arolt Prof. J.D. Laman Table of Contents

CHAPTER 1 General background 7

PART 1 Peripheral immune system 25

CHAPTER 2 Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder 27

CHAPTER 3 Feature-expression heat maps 51

CHAPTER 4 Inflammatory monocyte gene expression 67

CHAPTER 5 Does CRP predict outcome in bipolar disorder in regular outpatient care? 81

PART 2 Neuroimmune system 99

CHAPTER 6 PET and SPECT in bipolar disorders 101

CHAPTER 7 Neuroinflammation in bipolar disorder 121

CHAPTER 8 Volume, metabolites and neuroinflammation of the hippocampus in bipolar disorder 141

CHAPTER 9 Diffusion tensor imaging in euthymic bipolar disorder 171

CHAPTER 10 Summary and general discussion 195

Dutch summary 223 List of publications 243 Acknowledgements 247 Curriculum vitae 253

CHAPTER 1 General background

Bartholomeus C.M. Haarman

Adapted from Haarman BCM, Riemersma - Van der Lek RF, Ruhé HG, Groot JC, Nolen WA, Doorduin J. Bipolar Disorders. In: Dierckx RAJO, Otte A, de Vries EFJ, Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Heidelberg: Springer Berlin Heidelberg; 2014. p. 223–51. Introduction

Bipolar disorder: diagnostic description Bipolar disorder (BD)1 is a mood disorder characterized by episodic pathologic disturbances in mood: (hypo)manic episodes and depressive episodes which alternate with euthymic periods, i.e. with normal mood. BD has to be distinguished from (unipolar) major depressive disorder (MDD), which is characterized by only depressive episodes. According to DSM-IV, the core criterion of a (hypo)manic episode is the occurrence of a pathologically elated (euphoria), expansive or irritable mood. DSM-5 added increased energy or activity to this list. In addition to these core criteria, there are other symptoms such as inflated self-esteem or grandiosity, decreased need for sleep, being more talkative than usual, flight of ideas, distractibility, increase in goal-directed activity or psychomotor agitation and excessive involvement in plea- surable activities that have a high potential for painful consequences. A depressive episode consists of at least one of the core symptoms: depressed mood and loss of interest or pleasure, completed with symptoms such as sleep problems, psychomotor changes, fatigue or loss of energy, feelings of worthlessness or excessive feelings of guilt, difficulty concentrating or making decisions and recurrent thoughts of death1. Two types of BD are recognized: bipolar I disorder (BD-I) and bipolar II disorder (BD-II), characterized by the occurrence of manic episode(s) or by only hypomanic episode(s), respectively. The difference between manic and hypomanic episodes (and thus between BD-I and BD-II) is that manic episodes are associated with marked impairment in occupational, relational or social functioning, which can lead to hospitalization, while hypomanic episodes do not have this marked impairment and do not lead to hospitalization. When manic and depressive symptoms co-occur (or alternate very quickly) in the same episode, in DSM-IV it is labeled as a mixed episode and in DSM-5 as a bipolar disorder, manic or depressive episode with mixed features. Manic, depressive and mixed episodes can also be complicated by the presence of concurrent psychotic symptoms. Besides the mood symptoms, many patients with BD also show cognitive dysfunctions which may persist during euthymic periods, and which involve disturbances in various domains such as attention, verbal memory and executive functioning2,3.

8 Epidemiology and burden of disease The lifetime prevalence of BD is about 2% across different countries, women being affected as frequently as men4,5. Across the world, the disorder is sixth among all 1 health conditions in terms of causing disability6 with poor clinical and functional 7 8 9 outcome , increased risk for suicidality and significant societal costs . It has been background General calculated that in the United States the average cost per case ranged from $11,720 to $624,785, based on the severity of the illness9. In the European countries societal costs for managing BD are considered to be high as well10–12.

Risk factors Established risk factors for BD include history of BD, and interestingly also other psychiatric disorders, in parents or siblings; severe childhood adversities; and excessive alcohol use, although the last two are viewed as triggers rather than causes in the presence of biological vulnerability13,14. A recent systematic review concluded that it is still unclear whether perinatal infection has a role in the etiology of BD15. This is particularly relevant in view of the present thesis.

Diagnostic and treatment complications Although the clinical picture seems clear at first glance, making the diagnosis is more complicated in practice. On average, there is a time lag of about 6 years between the first episode and the making of the right diagnosis, and another six years before the start of adequate treatment. This is in most cases partly impeded by the precedence of depressive episodes without obvious (hypo)manic symptoms at the beginning of the disease16. Because antidepressants appear less effective for the treatment of bipolar depressive episodes than for unipolar MDD17, delayed diagnosis often leads to prolonged illness and dysfunction.

Current pathophysiological models

It is generally accepted that the cause of BD is multifactorial, with multiple genes making someone vulnerable, and with psychological and social factors causing the genes to be expressed. Moreover, somatic factors are assumed to play a role. To unravel the complex interplay between genotype and phenotype researchers have tried to find intermediary processes that are related to both the underlying genotype and the ultimate phenotype. Over the last 50 years several pathophysiological theories have been proposed for BD. Of these, we will shortly address the monoamine theory, the neuroinflammation theory, the white matter tract integrity disruption theory and the mitochondrial dysfunction theory.

9 Monoamine theory Since the 1960s, after the discovery of the first antipsychotic and antidepressant drugs, the monoamine theory has been the leading pathophysiological theory for various psychiatric disorders, including MDD and BD18–20. Based on the working mechanism of these drugs, disruption of serotonergic and noradrenergic neurotransmission in mood disorders and dopaminergic neurotransmission in schizophrenia underlies these disorders. In this regard noradrenaline is related to alertness and energy as well as to anxiety and attention; lack of serotonin to anxiety, obsessions, and compulsions; and dopamine to attention, motivation, pleasure, and reward19. Although the neurotransmitter theory originated in the discovery of psychotropic medication and has also given rise to the development of new pharmacological treatments, this model has also been criticized. The model is often regarded to be (too) simple, not explaining all patients’ symptoms, and having an effect that likely depends more on indirect effects, e.g. receptor changes21.

Corticolimbic dysregulation theory Based on extensive molecular imaging results (see chapter 6), complemented with functional MRI (fMRI) research, overall hyperactivation of limbic brain regions in BD patients relative to controls were displayed in a meta-analysis, along with an overall hypoactivation of frontal regions22 (see figure 1). This corresponds to findings in other mood disorders, especially MDD, which is known as the corticolimbic theory of depression23. Hypo- and hyperactivity in frontal and limbic regions, respectively, was most pronounced in manic patients, although also present in depressed and euthymic ones. Depressed patients exhibit more pronounced hypoactivation of frontal regions than euthymic patients, whereas euthymic patients display, surprisingly, more hyperactivity in limbic regions than their depressed counterparts. The corticolimbic theory has some overlap with several neurological networks that have been described and are thought to lie at the basis of physiological emotional processing. These networks can be divided into circuits that lie within the cerebral cortex and those that extend to other parts of the brain24. The limbic-cortical-striatal-pallidal-thalamic (LCSPT) circuit connects the PFC to the limbic and subcortical areas of the brain25. This LCSPT circuit is thought to be particularly important in the mediation of emotional expression, because of its relation to visceral control structures26. The mood related cortico-cortical networks interact with and extend to the LCSPT27 via top-down inhibitory control28. The orbital prefrontal network consists of the central and caudal part of the orbital cortex and the ventrolateral PFC; it includes sensory association areas such as the visual associated areas in the inferior temporal cortex and somatic-sensory associated areas in the insula and frontal operculum, as well as the olfactory and taste areas. In addition to sensory integration, this system

10 FIGURE 1 Neuroanatomical regions important in mood disorders 1 anterior cingulate cortex

prefrontal cortex background General

subgenual cingulate

orbitofrontal cortex ventral striatum

hypothalamus pituitary raphe nuclei amygdala

hippocampus locus coeruleus

nervus vagus

Neuroanatomical regions important in mood disorders. (Adapted from Patrick J. Lynch, medical illustrator, and C. Carl Jaffe, MD, cardiologist76, under the Creative Commons Attribution 2.5 Generic license (CC BY 2.5)).

codes for affective characteristics of stimuli such as reward, aversion and relative value (salience)26. The medial prefrontal network of cortical areas includes the ventromedial PFC, the dorsolateral PFC, the anterior and posterior cingulate cortex, anterior temporal cortex and the enthorhinal and posterior parahippocampal corteces. This system does not have substantial sensory connections, but is a visceromotor system that is particularly involved in introspective functions such as mood and emotion, and in visceral reactions to emotional stimuli24. It is widely known as the “default system”, because in fMRI imaging it appeared activated as a network of areas that become inactive in most tasks that involve external attention29. It has been proposed that the “ventral” orbital prefrontal network and the “dorsal” medial prefrontal network are reciprocally connected and that the orbital PFC may mediate connections between higher-order dorsolateral prefrontal regions and subcortical limbic regions such as the amygdala during emotion regulation30.

11 Neuroinflammation theory The “macrophage theory of depression” postulates an aberrant pro-inflammatory state of monocytes/macrophages in patients with mood disorder, and considers this aberrant state of the cells as a driving force behind the illness31. The theory is founded on a higher frequency of autoimmune diseases in mood disorders, aberrant pro-inflammatory cytokines and elevated pro-inflammatory gene expression in monocytes. Autoimmune thyroiditis is considered to be an endophenotype of BD32. Patients with BD and MDD have a raised prevalence of autoimmune thyroiditis33–35. Not only BD patients, but also their offspring (affected as well as non-affected) and their monozygotic (affected and non-affected) and dizygotic (affected, but not as much unaffected) co-twins have a raised prevalence of autoimmune thyroiditis32,36. It was hypothesized that an activated inflammatory response system in monocytes consti- tutes the shared genetic susceptibility factor for both BD and thyroid autoimmunity; this has led to the extensive investigations of neopterine, IL-1β, IL-6 and TNF-α in mood disorders, and in particular in MDD. With regard to the serum concentration of these compounds, increased levels were also described in BD when compared to controls, although not in all studies37,38. To investigate the pro-inflammatory state of monocytes in a more precise and robust manner, a Q-PCR analysis of CD14+ purified monocytes was performed in which 22 mRNAs for inflammatory, chemokinesis/motility, cell survival/apoptosis and MAP kinases pathway molecules were found to have an increased expression in BD patients compared to controls39. Interactions between the immune system and the HPA-axis, as well as interactions between the immune system and the neuronal system via indoleamine 2,3 dioxygenase (IDO) pathways have been suggested to result in mood disorder symptomatology. The HPA-axis is a complex set of direct influences and feedback interactions among the hypothalamus, the pituitary gland, and the adrenal glands that controls reactions to stress and regulates many body processes. The adrenal glands produce cortisol, which is a major stress hormone and has effects on many tissues in the body, including the brain, where it binds to glucocorticoid receptors in the PFC, the amygdala and the hippocampus40. Moreover, glucocorticoid insensitivity has been associated with a higher risk of developing a depressive episode40. In various in vivo and ex vivo studies a strong association between the activation of the inflammatory response system and glucocorticoid insensitivity has been demonstrated, linking at least in part the overproduction of pro-inflammatory cytokines to the HPA-axis disturbances in major mood disorders41–43. Tryptophan, the precursor amino acid of serotonin, can also be metabolized to downstream metabolites, known as kynurenines, via an alternative pathway. IDO is an oxygenase that catabolizes the first and rate-limiting step in this oxidative degradation. The IDO activity in monocytes/macrophages is enhanced by proin-

12 flammatory cytokines, e.g. during infections and when there is physical or mental stress44. Under such circumstances tryptophan breakdown is increased, making it less available for serotonin synthesis. When tryptophan is degraded, the next in vivo 1 product is kynurenine, which is the first metabolite of tryptophan45. This kynurenine

is again broken down into two pathways: (1) a neuroprotective, kynurenic acid, background General NMDA receptor antagonist pathway and (2) a neurotoxic 3-hydroxy kynurenine and quinolinic acid, NMDA receptor agonist pathway46. In the brain, this latter part of tryptophan catabolism, the kynurenine pathway, occurs in the astrocytes and microglia where astrocytes produce mainly neuroprotective kynurenic acid while macrophages produce mainly neurotoxic metabolites like quinolinic acid. Normally, formation of quinolinic acid is faster, while kynurenic acid has a counteractive protective role against quinolinic acid47. Based on the above, a hypothesis was proposed, that an imbalance between the neurodegenerative and neuroprotective pathways leads to neurodegeneration and brings a person to a chronically depressive episode. This imbalance might be due either to a highly increased neurodegenerative pathway activity or to a lack of sufficient neuroprotective factor activity48. Tryptophan levels and the neuroprotective kynurenic acid were significantly decreased in MDD patients when compared to controls49. Also, in IFN-α treatment of hepatitis C patients, associated with depression and fatigue, IFN-α was found to up-regulate the expression of IDO50. Furthermore, the decrease of plasma tryptophan and the increase of kynurenine and neopterine during IFN-α treatment were found to correlate with the development of depression51,52. Molecular imaging can be of added importance in investigating the neuroinflam mation theory. Microglia are the central cells involved in immune regulation in the brain. When activated, these cells present the translocator protein (TSPO) on their mitochondrial membrane53. Using the positron emission tomography (PET) ligand 11C-PK11195, areas of microglia activation in the brain can be visualized. Besides in various neurological disorders, microglia activation has been found in schizophrenia, where a clear focus of neuroinflammation was found in the hippocampus54,55.

White matter microstructure integrity disturbances Interest in the white matter tracts in BD started with the observation of diffuse cortical and callosal white matter pathology in structural MRI studies in BD patients56,57. With the development of diffusion tensor imaging (DTI), a MRI technique allowing for the investigation of the preferred direction and rate of water diffusion, the integrity of the white matter microstructure can be investigated in more detail, because in the physiological situation water diffusion is restricted by the axonal structures58. The main parameters derived from DTI are the fractional anisotropy (FA) and mean diffusivity (MD). MD measures the magnitude of water molecule diffusion and FA is an index of the degree of directionality of water diffusivity. FA is reduced in diseased

13 states known to be associated with axonal loss and destruction of myelin sheaths in several diseases, e.g. multiple sclerosis, leukoencephalopathies and Alzheimer’s disease59. In BD most studies reported reduced FA and/or elevated MD, compared to controls, involving the prefrontal lobe, corpus callosum, internal capsule, uncinate fasciculus and superior and inferior longitudinal fasciculi, and suggesting a role for white matter microstructure disturbances in BD pathophysiology60. The studies focusing on the specific mood states of BD patients revealed FA to be altered in the different mood states61. In the euthymic state FA was usually found to be increased in the genu of corpus callosum, internal capsule, anterior thalamic radiation and uncinate fasciculus compared to controls, whereas during depressive episodes a lower FA has been shown in the genu of the corpus callosum and in corona radiate compared controls. In mixed samples, higher and lower FA values were found in different brain regions62. The place of white matter microstructure disturbances in the pathophysiology, with regard to other disease mechanisms, is still controversial. It has been suggested that FA changes could be related to inflammation related processes in BD, analogous to multiple sclerosis61.

Mitochondrial dysfunction Using various techniques, scientific evidence for a cellular energy metabolism disturbance has been presented. When observed in cell biological research, abnormal mitochondrial morphology is often linked to altered energy metabolism. In BD patients, compared to controls, mitochondria in neurons and fibroblasts have been reported to be smaller and concentrated proportionately more within the perinuclear region than in distal processes of the cells63. Conversely, patients with mitochondrial diseases have a higher lifetime prevalence of MDD (54%) or BD (17%) than the average population64. Magnetic resonance spectroscopy (MRS) is a neuroimaging technique that allows the investigation of the metabolism on a cellular level. This MRI technique provides additional biochemical information of a selected voxel compared to a regular T1 or T2 image. The cellular metabolites are presumed to represent different cell functions: N-acetyl-aspertate (NAA) relates to cell viability and choline to cell membrane phospholipid integrity, and creatine is a measure of cellular metabolism65. Creatine plays an important role as a cell energy buffer, especially in high energy consum- ing cells such as muscular and brain cells. Using the creatine energy buffer reaction (figure 2), cells with an abundance of ATP can store energy by converting creatine to phosphocreatine. When in energy demanding circumstances the ATP stock becomes depleted, ATP can temporarily be supplied by reconverting phosphocreatine to creatine until the phosphocreatine stock is also depleted or energy is resupplied

14 FIGURE 2 Creatine energy buffer reaction 1 General background General

via other routes such as the oxidative phosphorylation. With 31P-MRS, creatine and phosphocreatine concentrations can be measured separately, as can the total concentration of both metabolites. The total concentration can also be measured with 1H-MRS, but the separate concentrations to a lesser degree when advanced quantification tools are being used. In BD patients a decreased phosphocreatine66 and reduced total creatine67,68 were described, when compared to controls, supporting the mitochondrial dysfunction theory. Findings in other MRS metabolites such as a reduced pH and increased lactate, exponents of cell metabolism exhaustion, add indirectly to this theory66,69. A study of the nature of the metabolic dysfunction revealed a paradoxical down regulation of mitochondria-related genes to glucose deprivation in fresh lymphocytes derived from BD patients, whereas control subjects showed an upregulation. This finding would suggest that patients with BD might have impairment in molecular adaptation to energy stress70. However, there is still debate as to whether this dys regulation is based on mitochondrial DNA disturbances, mitochondria-related nuclear DNA disturbances, or the effects of other mechanisms71.

15 Aim and outline of this thesis

Although multiple pathophysiological theories on BD exist, a comprehensive pathophysiological framework, integrating the various illness models for BD, is still lacking. Since immune cells play an important role, not only protecting neurons from pathogens but as microglia cells also helping in maintaining homeostasis in the brain72,73, they may play a key role in such a framework. This thesis aims to clarify the role of the immune system in the pathophysiology of BD via several different approaches, using bio-assay and neuroimaging techniques. The immune system consists of two parts: the peripheral immune system and the neuroimmune system. Structurally distinct from the peripheral immune system, the neuroimmune system is a system of structures and processes, involving the biochemical and electrophysiological interactions between the nervous system and immune system, which protect neurons from pathogens. Unlike the peripheral system, the neuroimmune system is composed primarily of glial cells, in particular microglia72–74.

Part 1: Peripheral immune system The first part of the thesis describes studies of the function of the peripheral immune system in BD, focusing on monocyte pro-inflammatory gene-expression and CRP, using bio assay techniques.

Association between monocyte gene-expression and clinical features Chapter 2 discusses the associations between monocyte pro-inflammatory gene expression and illness characteristics. We a priori hypothesized that pro-inflammatory gene expression would be found more frequently in BD patients with a lifetime history of psychotic symptoms. We go on to explore the associations with the individual manic and depressive symptoms of BD. Then we elucidate the course of the pro-inflammatory gene expression by investigating in more detail the relation between the pro-inflammatory gene expression and age at onset, and duration of illness. Finally, in chapter 2, we examine the association between the current use of psychotropic medication and monocyte pro-inflammatory gene expression. The feature-expression heat map, an in-house developed method used in chapter 2 to visualize the complex associations between symptoms and gene-expression, is explained in more detail in chapter 3.

Monocyte gene-expression: state or trait? In chapter 4 we present the results of the study among the bipolar cohort of the MOODINFLAME project75, in which we investigated whether monocyte pro-inflammatory gene-expression is more a state or more a trait marker. We compare the monocyte pro-inflammatory gene-expression in euthymic BD patients

16 with HC. Moreover, we present the results of a small additional study in which BD patients are compared during a mood episode and when euthymic. 1 Does CRP predict outcome in clinical practice?

Chapter 5 describes a historic cohort study in which we investigate whether, in background General a clinical setting, higher CRP levels at baseline may predict a worse BD outcome, defined as a shorter time to relapse or a longer time to recover, depending on the mood state at baseline.

Part 2: Neuroimmune system The second part of the thesis investigates the function of the neuroimmune system, focusing on microglia activation, and using PET, MRI, MRS, DTI neuroimaging techniques.

Previous PET/SPECT studies Chapter 6 describes the findings of previous PET / single-photon emission computed tomography (SPECT) research efforts in BD, based on the literature. First, we discuss the cerebral blood flow and cerebral metabolism findings, followed by the neurotransmitter studies. Finally, we summarize the most important conclusions, and follow with remarks about the observed molecular imaging study designs specific for BD.

Microglial activation in the hippocampus In chapter 7 we present the results of the first neuroinflammation PET study in BD. We aimed to demonstrate an increased [11C]-(R)-PK11195 binding to activated microglia in BD-I in comparison to a healthy control group. We a priori hypothesized the hippocampus to be the main focus of neuroinflammation in BD. In a second model we explore the presence of neuroinflammation in other brain regions.

Associations between volume, metabolites and microglial activation In chapter 8 we investigate the relations between volume, metabolites and microglial activation of the hippocampus in a contemporaneously executed PET/MRI study. We compare hippocampal volume and metabolites in BD-I patients with HC, using MRI and MRS. We a priori hypothesized hippocampal volume and the N-acetylaspartate (NAA) metabolite to be decreased in BD patients, compared to HC. Subsequently, within the BD-I and HC groups, we post-hoc investigate whether hippocampal volume and metabolites were associated with microglial activation. Furthermore, we explore if potential illness modifying factors such as duration of illness, medication use, body mass index (BMI), exercise, smoking, number of caffeine consumptions and alcohol use did affect these hippocampal measurements within the BD-I group, and whether

17 these hippocampal measurements were associated with experienced mood and functioning.

White matter microstructure disturbances and lithium usage Using DTI, in chapter 9 we investigate the white matter microstructure in BD and HC, and differences among patients in relation to lithium usage. We first compare estimates of white matter microstructure (fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD)) between euthymic BD-I patients and HC. We a priori hypothesized that we would find a widespread decrease in FA in several white matter tracts of BD-I patients compared to HC, associated with reciprocal alterations of other white matter microstructural parameters. Subsequently, we divide the patient group into lithium-users and non-lithium-users and analyze the estimates of white matter microstructure across these three groups (non-lithium-users, lithium-users, healthy controls). In this analysis, supposing a restoring effect of lithium on myelination, we a priori hypothesized FA to be increased - and consistently a decrease of other white matter microstructural parameters - in lithium-using patients compared to non-lithium-using patients, possibly even attaining healthy control values.

18 References

1. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Arlington, VA: American Psychiatric Publishing, Inc; 2013. 1 2. Arts B, Jabben N, Krabbendam L, van Os J. Meta-analyses of cognitive functioning in euthymic

bipolar patients and their first-degree relatives. Psychol Med. 2008 Jun;38(6):771–85. background General 3. Bora E, Yucel M, Pantelis C. Cognitive endophenotypes of bipolar disorder: a meta-analysis of neuropsychological deficits in euthymic patients and their first-degree relatives. J Affect Disord. 2009 Feb;113(1-2):1–20. 4. Pini S, de Queiroz V, Pagnin D, Pezawas L, Angst J, Cassano GB, et al. Prevalence and burden of bipolar disorders in European countries. Eur Neuropsychopharmacol. 2005 Aug;15(4):425–34. 5. Merikangas KR, Akiskal HS, Angst J, Greenberg PE, Hirschfeld RMA, Petukhova M, et al. Lifetime and 12-month prevalence of bipolar spectrum disorder in the National Comorbidity Survey replication. Arch Gen Psychiatry. 2007 May;64(5):543–52. 6. World Health Organization. The world health report 2001: mental health: new understanding, new hope. World Health. Geneva, Switzerland; 2001. 7. Goodwin FK. Manic-Depressive Illness: Bipolar Disorders and Recurrent Depression. 2nd ed. New York: Oxford University Press; 2007. 8. Baldessarini RJ, Tondo L. Suicide risk and treatments for patients with bipolar disorder. JAMA. 2003 Sep 17;290(11):1517–9. 9. Begley CE, Annegers JF, Swann AC, Lewis C, Coan S, Schnapp WB, et al. The lifetime cost of bipolar disorder in the US: an estimate for new cases in 1998. Pharmacoeconomics. 2001 Jan;19(5 Pt 1):483–95. 10. Hakkaart-Van Roijen L, Hoeijenbos MB, Regeer EJ, Ten Have M, Nolen WA, Veraart CPWM, et al. The societal costs and quality of life of patients suffering from bipolar disorder in the Netherlands. Acta Psychiatr Scand. 2004;110(5):383–92. 11. Young AH, Rigney U, Shaw S, Emmas C, Thompson JM. Annual cost of managing bipolar disorder to the UK healthcare system. J Affect Disord. Elsevier B.V.; 2011;133(3):450–6. 12. Ekman M, Granström O, Omérov S, Jacob J, Landén M. The societal cost of bipolar disorder in Sweden. Soc Psychiatry Psychiatr Epidemiol. 2013;48(10):1601–10. 13. Rush AJ. Toward an understanding of bipolar disorder and its origin. J Clin Psychiatry. 2003;64 Suppl 6:4–8; discussion 28. 14. Post RM, Leverich GS, Xing G, Weiss RB. Developmental vulnerabilities to the onset and course of bipolar disorder. Dev Psychopathol. 2001;13(3):581–98. 15. Barichello T, Badawy M, Pitcher MR, Saigal P, Generoso JS, Goularte JA, et al. Exposure to Perinatal Infections and Bipolar Disorder: A Systematic Review. Curr Mol Med. 2016;16(2):106–18. 16. Suppes T, Leverich GS, Keck PE, Nolen W a, Denicoff KD, Altshuler LL, et al. The Stanley Foundation Bipolar Treatment Outcome Network. II. Demographics and illness characteristics of the first 261 patients. J Affect Disord. 2001 Dec;67(1-3):45–59. 17. Sachs GS, Nierenberg AA, Calabrese JR, Marangell LB, Wisniewski SR, Gyulai L, et al. Effectiveness of adjunctive antidepressant treatment for bipolar depression. N Engl J Med. 2007 Apr 26;356(17):1711–22. 18. Schildkraut JJ. The catecholamine hypothesis of affective disorders: a review of supporting evidence. Am J Psychiatry. 1965 Nov;122(5):509–22. 19. Nutt DJ. Relationship of neurotransmitters to the symptoms of major depressive disorder. J Clin Psychiatry. 2008;69 Suppl E:4–7. 20. Malhi GS, Tanious M, Das P, Coulston CM, Berk M. Potential mechanisms of action of lithium in bipolar disorder: Current understanding. CNS Drugs. 2013. p. 135–53. 21. Hirschfeld RM. History and evolution of the monoamine hypothesis of depression. J Clin Psychiatry. 2000 Jan;61 Suppl 6:4–6.

19 22. Kupferschmidt D a, Zakzanis KK. Toward a functional neuroanatomical signature of bipolar disorder: quantitative evidence from the neuroimaging literature. Psychiatry Res. Elsevier Ireland Ltd; 2011 Aug 30;193(2):71–9. 23. Mayberg HS. Limbic-cortical dysregulation: a proposed model of depression. J Neuropsychiatry Clin Neurosci. 1997 Jan;9(3):471–81. 24. Price JL, Drevets WC. Neurocircuitry of mood disorders. Neuropsychopharmacology. Nature Publishing Group; 2010 Jan;35(1):192–216. 25. al-Mousawi a. H, Evans N, Ebmeier KP, Roeda D, Chaloner F, Ashcroft GW. Limbic dysfunction in schizophrenia and mania. A study using 18F- labelled fluorodeoxyglucose and positron emission tomography. Br J Psychiatry. 1996 Oct 1;169(4):509–16. 26. Drevets WC, Price JL, Furey ML. Brain structural and functional abnormalities in mood disorders: implications for neurocircuitry models of depression. Brain Struct Funct. 2008 Sep;213(1-2):93–118. 27. Ongür D, Ferry AT, Price JL. Architectonic subdivision of the human orbital and medial prefrontal cortex. J Comp Neurol. 2003 Jun 2;460(3):425–49. 28. Savitz J, Drevets WC. Bipolar and major depressive disorder: neuroimaging the developmental- degenerative divide. Neurosci Biobehav Rev. 2009 May;33(5):699–771. 29. Gusnard DA, Akbudak E, Shulman GL, Raichle ME. Medial prefrontal cortex and self-referential mental activity: relation to a default mode of brain function. Proc Natl Acad Sci U S A. 2001 Mar 27;98(7):4259–64. 30. Phillips ML, Ladouceur CD, Drevets WC. A neural model of voluntary and automatic emotion regulation: implications for understanding the pathophysiology and neurodevelopment of bipolar disorder. Mol Psychiatry. 2008 Sep;13(9):829, 833–57. 31. Smith RS. The macrophage theory of depression. Med Hypotheses. 1991 Aug;35(4):298–306. 32. Vonk R, van der Schot AC, Kahn RS, Nolen W a, Drexhage H a. Is autoimmune thyroiditis part of the genetic vulnerability (or an endophenotype) for bipolar disorder? Biol Psychiatry. 2007 Jul 15;62(2):135–40. 33. Bunevicius R, Peceliuniene J, Mickuviene N, Bunevicius A, Pop VJ, Girdler SS. Mood and thyroid immunity assessed by ultrasonographic imaging in a primary health care. J Affect Disord. 2007 Jan;97(1-3):85–90. 34. Carta MG, Loviselli A, Hardoy MC, Massa S, Cadeddu M, Sardu C, et al. The link between thyroid autoimmunity (antithyroid peroxidase autoantibodies) with anxiety and mood disorders in the community: a field of interest for public health in the future. BMC Psychiatry. 2004 Jan;4:25. 35. Benros ME, Waltoft BL, Nordentoft M, Ostergaard SD, Eaton WW, Krogh J, et al. Autoimmune diseases and severe infections as risk factors for mood disorders: A nationwide study. JAMA Psychiatry. 2013;70(8):812–20. 36. Hillegers MH, Reichart CG, Wals M, Verhulst FC, Ormel J, Nolen W a. Five-year prospective outcome of psychopathology in the adolescent offspring of bipolar parents. Bipolar Disord. 2005 Aug;7(4):344–50. 37. O’Brien SM, Scully P, Scott L V, Dinan TG. Cytokine profiles in bipolar affective disorder: focus on acutely ill patients. J Affect Disord. 2006 Feb;90(2-3):263–7. 38. Hoekstra R, Fekkes D, Pepplinkhuizen L, Loonen AJM, Tuinier S, Verhoeven WMA. Nitric oxide and neopterin in bipolar affective disorder. Neuropsychobiology. 2006 Jan;54(1):75–81. 39. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 40. Spijker a T, van Rossum EFC. Glucocorticoid sensitivity in mood disorders. Neuroendocrinology. 2012 Jan;95(3):179–86. 41. Pariante CM, Pearce BD, Pisell TL, Sanchez CI, Po C, Su C, et al. The proinflammatory cytokine, interleukin-1alpha, reduces glucocorticoid receptor translocation and function. Endocrinology. 1999 Sep;140(9):4359–66.

20 42. Almawi WY, Lipman ML, Stevens AC, Zanker B, Hadro ET, Strom TB. Abrogation of glucocorticoid- mediated inhibition of T cell proliferation by the synergistic action of IL-1, IL-6, and IFN-gamma. J Immunol. 1991 May 15;146(10):3523–7. 43. Ito K, Chung KF, Adcock IM. Update on glucocorticoid action and resistance. J Allergy Clin 1 Immunol. 2006 Mar;117(3):522–43. 44. Babcock TA, Carlin JM. Transcriptional activation of indoleamine dioxygenase by interleukin 1 and

tumor necrosis factor alpha in interferon-treated epithelial cells. Cytokine. background General 2000 Jun;12(6):588–94. 45. Bender DA, McCreanor GM. Kynurenine hydroxylase: a potential rate-limiting enzyme in tryptophan metabolism. Biochem Soc Trans. 1985 Apr;13(2):441–3. 46. Chiarugi A, Calvani M, Meli E, Traggiai E, Moroni F. Synthesis and release of neurotoxic kynurenine metabolites by human monocyte-derived macrophages. J Neuroimmunol. 2001 Nov 1;120(1- 2):190–8. 47. Perkins MN, Stone TW. An iontophoretic investigation of the actions of convulsant kynurenines and their interaction with the endogenous excitant quinolinic acid. Brain Res. 1982 Sep 9;247(1):184–7. 48. Myint AM, Kim YK. Cytokine-serotonin interaction through IDO: a neurodegeneration hypothesis of depression. Med Hypotheses. 2003;61(5-6):519–25. 49. Myint A-M, Kim YK, Verkerk R, Scharpé S, Steinbusch H, Leonard B. Kynurenine pathway in major depression: evidence of impaired neuroprotection. J Affect Disord. 2007 Feb;98(1-2):143–51. 50. Curreli S, Romerio F, Mirandola P, Barion P, Bemis K, Zella D. Human primary CD4 + T cells activated in the presence of IFN-alpha 2b express functional indoleamine 2,3-dioxygenase. J Interferon Cytokine Res. 2001 Jun;21(6):431–7. 51. Capuron L, Ravaud A, Neveu PJ, Miller AH, Maes M, Dantzer R. Association between decreased serum tryptophan concentrations and depressive symptoms in cancer patients undergoing cytokine therapy. Mol Psychiatry. 2002 Jan;7(5):468–73. 52. Wichers MC, Koek GH, Robaeys G, Verkerk R, Scharpé S, Maes M. IDO and interferon-alpha- induced depressive symptoms: a shift in hypothesis from tryptophan depletion to neurotoxicity. Mol Psychiatry. 2005 Jun;10(6):538–44. 53. Doorduin J, de Vries EFJ, Dierckx RA, Klein HC. PET imaging of the peripheral benzodiazepine receptor: monitoring disease progression and therapy response in neurodegenerative disorders. Curr Pharm Des. 2008 Jan;14(31):3297–315. 54. van Berckel BN, Bossong MG, Boellaard R, Kloet R, Schuitemaker A, Caspers E, et al. Microglia activation in recent-onset schizophrenia: a quantitative (R)-[11C]PK11195 positron emission tomography study. Biol Psychiatry. 2008 Nov 1;64(9):820–2. 55. Doorduin J, de Vries EFJ, Willemsen ATM, de Groot JC, Dierckx R a, Klein HC. Neuroinflammation in schizophrenia-related psychosis: a PET study. J Nucl Med. 2009 Nov;50(11):1801–7. 56. Kempton MJ, Geddes JR, Ettinger U, Williams SCR, Grasby PM. Meta-analysis, database, and meta-regression of 98 structural imaging studies in bipolar disorder. Arch Gen Psychiatry. 2008 Sep;65(9):1017–32. 57. Vita A, De Peri L, Sacchetti E. Gray matter, white matter, brain, and intracranial volumes in first-episode bipolar disorder: a meta-analysis of magnetic resonance imaging studies. Bipolar Disord. 2009 Dec;11(8):807–14. 58. Le Bihan D. Molecular diffusion, tissue microdynamics and microstructure. NMR Biomed. 1996;8(7-8):375–86. 59. Le Bihan D. Looking into the functional architecture of the brain with diffusion MRI. Nat Rev Neurosci. 2003 Jun;4(6):469–80. 60. Heng S, Song AW, Sim K. White matter abnormalities in bipolar disorder: insights from diffusion tensor imaging studies. J Neural Transm. 2010 May;117(5):639–54.

21 61. Zanetti M V, Jackowski MP, Versace A, Almeida JRC, Hassel S, Duran FLS, et al. State-dependent microstructural white matter changes in bipolar I depression. Eur Arch Psychiatry Clin Neurosci. 2009 Sep;259(6):316–28. 62. Bellani M, Brambilla P. Diffusion imaging studies of white matter integrity in bipolar disorder. Epidemiol Psychiatr Sci. 2011 Jun;20(2):137–40. 63. Cataldo AM, McPhie DL, Lange NT, Punzell S, Elmiligy S, Ye NZ, et al. Abnormalities in mitochondrial structure in cells from patients with bipolar disorder. Am J Pathol. American Society for Investigative Pathology; 2010 Aug;177(2):575–85. 64. Fattal O, Link J, Quinn K, Cohen BH, Franco K. Psychiatric comorbidity in 36 adults with mitochondrial cytopathies. CNS Spectr. 2007 Jun;12(6):429–38. 65. Gillard JH, Waldman AD, Barker PB, editors. Clinical MR Neuroimaging. Spectroscopy. Cambridge: Cambridge University Press; 2004. 66. Kato T, Takahashi S, Shioiri T, Inubushi T. Alterations in brain phosphorous metabolism in bipolar disorder detected by in vivo 31P and 7Li magnetic resonance spectroscopy. J Affect Disord. 1993 Jan;27(1):53–9. 67. Port JD, Unal SS, Mrazek D a, Marcus SM. Metabolic alterations in medication-free patients with bipolar disorder: a 3T CSF-corrected magnetic resonance spectroscopic imaging study. Psychiatry Res. 2008 Feb 28;162(2):113–21. 68. Frey, Stanley J a, Nery FG, Monkul ES, Nicoletti M a, Chen H-H, et al. Abnormal cellular energy and phospholipid metabolism in the left dorsolateral prefrontal cortex of medication-free individuals with bipolar disorder: an in vivo 1H MRS study. Bipolar Disord. 2007 Jun;9 Suppl 1(1):119–27. 69. Dager SR, Friedman SD, Parow A, Demopulos C, Stoll AL, Lyoo IK, et al. Brain metabolic alterations in medication-free patients with bipolar disorder. Arch Gen Psychiatry. 2004 May;61(5):450–8. 70. Naydenov A V, MacDonald ML, Ongur D, Konradi C. Differences in lymphocyte electron transport gene expression levels between subjects with bipolar disorder and normal controls in response to glucose deprivation stress. Arch Gen Psychiatry. 2007 May;64(5):555–64. 71. Kato T. Molecular neurobiology of bipolar disorder: a disease of “mood-stabilizing neurons”? Trends Neurosci. 2008 Oct;31(10):495–503. 72. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 73. Stertz L, Magalhães PVS, Kapczinski F. Is bipolar disorder an inflammatory condition? The relevance of microglial activation. Curr Opin Psychiatry. 2013 Jan;26(1):19–26. 74. Eyre H, Baune BT. Neuroplastic changes in depression: A role for the immune system. Psychoneuroendocrinology. Elsevier Ltd; 2012;37(9):1397–416. 75. MOODINFLAME website. http://moodinflame.eu. 2014. 76. Lynch PJ, Jaffe CC. Brain human sagittal section [Internet]. Wikimedia Commons. 2006. Available from: https://commons.wikimedia.org/wiki/File:Brain_human_sagittal_section.svg?uselang=en

22 1 General background General

PART 1 Peripheral immune system

CHAPTER 2 Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder Towards a better understanding of psychoimmunological interactions

Bartholomeus C.M. Haarman, Rixt F. Riemersma-Van der Lek, Huibert Burger, Mina Netkova, Roosmarijn C. Drexhage, Florian Bootsman, Esther Mesman, Manon H.J. Hillegers, Anne T. Spijker, Erik Hoencamp, Hemmo A. Drexhage, Willem A. Nolen

Bipolar Disorders. 2014 Mar 29;16(2):137–50. Abstract

Objectives Existing and previously published datasets were examined for associations between illness and treatment characteristics and monocyte pro-inflammatory gene expression in bipolar disorder (BD) patients. We a priori hypothesized that increased monocyte pro-inflammatory gene expression would be found more frequently in patients with a lifetime history of psychotic symptoms.

Methods Monocyte QPCR and symptom data of 64 BD patients were collected from three Dutch studies. Regression analyses were performed to analyze the various associations of which feature-expression heat maps were drawn.

Results Symptoms No associations were found with lifetime psychotic symptoms, while a positive association was identified between sub-cluster 2 genes and manic symptoms.

Age at onset / duration of illness For several sub-cluster 1a genes a negative association was found with age at onset. For most sub-cluster 2 genes a positive association was found with the duration of illness.

Medication Current use of antidepressants and of anti-epileptics were associated with sub-cluster 2 gene expression, and current use of lithium and antipsychotics with sub-cluster 1a gene expression.

Conclusions Our hypothesis that lifetime psychotic features would be associated with the pro-inflammatory monocyte gene expression was not confirmed. In an explorative analysis we found: (1) a possible relation between pro-inflammatory gene expression and manic symptomatology, (2) a differential immune activation related to age at onset and duration of illness, and (3) support for the concept of an immune suppressive action of some of the mood regulating medications.

28 Introduction

The pathophysiology of bipolar disorder (BD) is complex. While there is no doubt that both genetic predisposition and environmental factors play a role, it remains important to further investigate their neurobiological underpinnings and their interaction1. As both the stress system and the immune system interact with the brain and are influenced by the environment, they can be regarded as a linking pin. The “mono- 2 cyte-T-cell theory of mood disorders”2 considers an activated inflammatory response system (IRS) in mood disorders as a driving force behind the illness. IRS activation disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship can be regarded as a disbalance in immune regulatory processes. Pro-inflammatory cytokines are capable of destabilizing brain function3, which makes the brain vulnerable to stress and possibly other yet unknown factors with mood disturbances as the consequence. Padmos et al.4 described a sensitive quantitative polymerase chain reaction (Q-PCR) assay system to detect the pro-inflammatory state of circulating monocytes of naturalistically treated patients with BD and detected in the monocytes a coherent, mutually correlating set of 19 aberrantly expressed inflammatory genes (‘a pro-inflammatory signature’), supporting the concept of an activated IRS in mood disorders2. We expect the IRS to be an intermediate process between genotype and phenotype (figure 1). The pro-inflammatory signature occurred in 55% of BD patients versus 18% in healthy controls.

FIGURE 1 Schematic presentation of the monocyte-T-cell theory of mood disorders in BD

Genetic Stress vulnerability

Activated Medication - inflammatory response system

Intermediary mechanisms

Bipolar disorder Other (auto-immune) phenotype pathology

29 In a subsequent study, Drexhage et al.5 found nine inflammation related schizophrenia (SZ) genes using the same method, indicating that monocytes of SZ and BD patients partly overlap, but also differ in inflammatory gene expression. Combined with six specific auto-immune genes, related to BD, this resulted in a set of 34 genes (table 1). Moreover, via cluster analysis they identified the expression of sub-clusters 1a en 1b within the signature genes, relating to core inflammation and transcription, and sub-cluster 2, relating to adhesion, motility and chemotaxis (figure 2). Several of these 34 genes were also investigated in other studies in BD. Savitz et al. recently demonstrated an increased expression of TNF and eleven other genes in monocytes of a combined sample of depressed BD patients and major depressive disorder patients compared to healthy controls. On the genomic level CCL2 was found to be weakly overexpressed in BD patients in a Nordic genome wide association study. PDE4B6,7, IL68, IL19, TNF8,10,11, EGR312,13, CCL214–16 and FABP17 were investigated in gene polymorphism studies.

TABLE 1 Description of genes

Gene symbol Name of corresponding protein DUSP2 Dual specificity protein phosphatase 2 ATF3 Cyclic AMP-dependent transcription factor 3 PDE4B cAMP-specific 3',5'-cyclic phosphodiesterase 4B IL6 Interleukin 6

IL1 Interleukin 1 TNF Tumor necrosis factor TNFAIP3 Tumor necrosis factor, alpha-induced protein 3 BCL2A1 B-cell lymphoma-2-related protein A1 PTX3 Pentraxin-related protein 3 Inflammation PTGS Prostaglandin G/H synthase (cyclooxygenase) CCL20 C-C chemokine ligand 20 CXCL2 C-X-C chemokine ligand 2 EREG Epiregulin CXCL3 C-X-C chemokine ligand 3 MXD MAD protein

EGR3 Early growth response protein 3 F3 Tissue factor 3 MAFF Musculoaponeurotic fibrosarcoma oncogene homolog F THBS Thrombospondin 1

Transcription SERPINB2 Plasminogen activator inhibitor-2 RGC32 Response gene to complement 32 protein PTPN Protein tyrosine phosphatase, non-receptor type 7 NAB2 Nerve growth factor-induced protein A binding protein 2 MAPK6 Mitogen-activated protein kinase 6 EMP1 Epithelial membrane protein 1 STX1 Syntaxin-1A DHRS3 Short-chain dehydrogenase/reductase 3

chemotaxis CCL2 C-C chemokine ligand 2 CCL7 C-C chemokine ligand 7 Adhesion / motility / / motility / Adhesion CDC42 Cell division control protein 42 homolog

FABP Fatty acid-binding protein 5 CD9 Cluster of differentiation 9 antigen HSPA1 Heat shock 70 kDa protein 1 Other CCR2 C-C chemokine receptor type 2

30 FIGURE 2 The hypothetical relationship of the molecules of the pro-inflammatory gene expression in monocytes of BD patients 2 Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship

These molecules have regulatory functions in the monocyte of which some are predominantly active in inflammation (sub-cluster 1a, squared, e.g. TNF, IL1, IL6), others in regulating transcription (sub-cluster 1b, diamonds, e.g. EGR3, MXD), whereas others have their main function in cell movement functions such as adhesion, motility or chemotaxis (sub-cluster 2, encircled, e.g. CCL2, CCL7). Additional abbreviations: Gα and Gq, G protein α and q; cAMP, Cyclic adenosine monophosphate; PKA and PKC, protein kinase A and C; raf, rapidly accelerated fibrosarcom; TNF-R, tumor necrosis factor receptor; p38, protein 38; jnks, jun N-terminal kinase; elk-1, ETS domain- containing protein; egr-1, early growth response protein 1; GRE, glucocorticoid response element; RXR, retinoid X receptor; RAR, retinoid acid receptor; PPAR, peroxisome proliferator-activated receptor. (Adapted from Drexhage55, reprinted with permission)

Although in the studies by Padmos and Drexhage the BD patients as a group were characterized by a positive pro-inflammatory signature, a proportion of BD patients did not show a positive signature. This led to the subsequent question whether specific clinical characteristics in BD patients would be associated with the pro-inflammatory gene expression. In view of the aforementioned overlap of pro-inflammatory gene expression between BD and SZ we a priori hypothesized that the pro-inflammatory gene expression would be found more frequently in BD patients with a lifetime history of psychotic symptoms. Secondly, we explored the associations with the individual manic and depressive symptoms of BD, extending on previous findings relating the mood state to specific

31 pro-inflammatory gene expression changes4. Thirdly, we elucidated the course of the pro-inflammatory gene expression by investigating the relation between the pro-inflammatory gene expression and age at onset and duration of illness in more detail. We presumed that patients with an earlier age at onset would have a biologically more severe form of the illness, therefore demonstrating an increased pro-inflammatory gene expression compared to patients with a later age at onset. Furthermore, we expected the pathogenic processes to cause progressing derailment of the illness, demonstrating a positive association between pro-inflammatory gene expression and duration of illness irrespective of age at onset.

Finally, we examined the association between current use of psychotropic medication and monocyte pro-inflammatory gene expression, as (part of) these medications have been reported to be immune suppressive in nature5,18–21.

Patients and Methods

Participants Study participants comprised of patients and healthy controls from three different studies4,5,22–25. All patients (n=64) had a DSM-IV BD I or II disorder, confirmed by the Structured Clinical Interview for DSM-IV axis I disorders (SCID-I)26 (table 2 and figure 3). For this extension analysis we selected all participants (64 outpatients, age range 16-61 years, 24 (38%) male) who were included in the original studies and of whom Q-PCR and questionnaire data were available. Regarding the twin study, only data from the index twin were used. There was no family relation between any of the participants. Patients needed to be free of any additional severe psychiatric disorders and of any relevant medical illness that might affect inflammation status at least 2 weeks before blood withdrawal. Healthy controls (n=63) were also recruited within the three aforementioned studies and comprised of laboratory or medical staff, medical students and high school students. All controls were free of any lifetime psychiatric disorder and of a history of these disorders in first-degree family members (self report). Controls did not use any psychotropic or other medication and were also free of any relevant medical illness that might affect inflammation status at least 2 weeks before blood withdrawal.

Ethical considerations The Medical Ethical Review Committee of the University Medical Center Utrecht approved the original studies, which were performed in accordance with the Helsinki Declaration of 1975. Written informed consent was obtained from all participants.

32 TABLE 2 Characteristics of the participants

Bipolar disorder Healthy controls Group size 64 63 DSM IV classification Bipolar I disorder 52 (81%) Bipolar II disorder 12 (19%) Age (yr) 40 (16-61) 40 (16-58) Gender 2 Male 24 (38%) 22 (34%) Female 40 (62%) 39 (66%)

Duration of illness (mean (range), yr) 15 (3-38) disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship Age at onset (mean (range), yr) 25 (6-50) Medication (n=58) Antidepressants 19/58 (33%) Lihium 43/58 (74%) Antipsychotics 9/58 (16%) Valproate 12/58 (21%) Carbamazepine 6/58 (10%) Lamotrigine 2/58 (3%) Benzodiazepines 18/58 (31%)

FIGURE 3 Patient flow diagram

58 BD patients 41 BD twin pairs 54 children of BD parents Stanley Foundation Bipolar Dutch Twin study Dutch BD Offspring study Network, Dutch site

41 index BD patients with QPCR 19 BD patients with QPCR and 4 BD patients with QPCR and and questionnaire data available questionnaire data available questionnaire data available

64 patients

62 duration of 56 item level 58 item level 58 medication 57 item level illness IDS YMRS lists SCID-I measurements

We used the data of forty-one patients who took part in an ongoing Dutch twin study on BD22, nineteen patients participated in the Dutch site of the former Stanley Foundation Bipolar Network (SFBN) 56, and four patients participated in a ongoing prospective study among the adolescent offspring of BD patients in the Netherlands24,25.

Assessments Clinical features used in the analyses were extracted from the interviews held, which included the SCID-I, the Young Mania Rating Scale (YMRS), the Inventory of Depressive Symptoms (IDS) and a demographic questionnaire. The lifetime history of psychotic symptoms was derived from the SCID-I. The YMRS is an eleven-item,

33 multiple-choice questionnaire to assess manic symptoms27. The IDS is a thirty-item, multiple-choice questionnaire to assess depressive symptoms of all symptom domains of depression28. Individual YMRS and IDS item scores were used to qualify the presence of specific mood symptoms. The scores of the individual questionnaire items were transformed to a 0-1 scale to facilitate mutual comparison. Age at onset was defined as the age when the first mood episode occurred. Current medication use was dichotomously described as being present or absent.

Laboratory methods Blood (drawn in the morning and within three days of psychiatric assessment in all cases) was collected in clotting tubes for serum preparation and stored at -80°C, and in sodium-heparin tubes for immune cell preparation. From the heparinized blood, peripheral blood mono-nuclear cell (PBMC) suspensions were prepared in the after- noon by low-density gradient centrifugation within 8 hours to avoid activation of the monocytes29. To obtain cDNA for Q-PCR, 1 mg RNA was reversed-transcribed using the cDNA high-capacity cDNA Reverse Transcription kit (Applied Biosystems, USA). Q-PCR was performed on thirty-four genes selected by whole genome profiling, previously described by R. Drexhage et al.5, using a method described by Staal et al.30.

Preprocessing The quantitative value obtained from Q-PCR is a cycle threshold (Ct) that was used to calculate household-gene (Abl) normalized Ct values (ΔCt = Ct gene – Ct housekeeping gene) via the ΔΔCt method31. By subtracting the ΔCt with the mean ΔCt for the healthy control group, the relative gene expression (ΔΔCt) was determined. The healthy controls were solely used for this purpose and were not needed in the other analyses. Z-score transformation of the relative gene expression was applied to facilitate mutual comparison. The relative gene expression can consequently be expressed as fold change after transformation via the the ΔΔCt method (fold change = 2-ΔΔCt)31.

Statistical analyses Regression methods were used to determine size and statistical significance of the association between gene expression and clinical features. Analyses were performed using individual immunologic details and individual psychiatric details. These regarded associations between specific mRNA gene expressions e.g. PDE4B, ATF3 and specific psychiatric symptoms, e.g. auditory hallucinations, depressed mood. The associations between pro-inflammatory gene expression and lifetime history of psychotic symptoms were analyzed using ordinal regression, taking lifetime history of psychotic symptoms as dependent variable, in accordance with our pathophysiological model (figure 1).

34 Associations of pro-inflammatory gene expression with current manic symptoms and with current depressive symptoms were analyzed using ordinal regression, taking symptoms as dependent variable. Associations between age at onset and pro-inflammatory gene expression and between duration of illness and pro-inflammatory gene expression were analyzed using linear regression. This was done separately in univariable and in combined multivariable models without and with an interaction term for effect of age at onset 2 on the duration of illness. Theses analyses were repeated while correcting for gender. Pro-inflammatory gene expression was the dependent variable. disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship Associations between pro-inflammatory gene expression and medications were analyzed with linear regression using the pro-inflammatory gene expression as dependent variable. Continuous outcome measures (relative gene expression, duration of illness) were checked for normal distribution by graphical inspection and using the skewness and kurtosis normality test. Univariable and multivariable linear regression were used for continuous, normally distributed outcome variables (disease duration, gene expression in the medication analyses). Because of the multiple individual statistical tests in the analyses, there is an increased risk of a wrongful rejection of the null hypothesis (type I error). To control for this problem, we applied correction for the false discovery rate (FDR), as described by Benjamini-Hochberg32, in most analyses and only considered clustering groups of associations to be of importance. Statistical analyses were performed using Stata Statistical Software, release 11 (StataCorp. 2009, College Station, TX).

Feature-expression heat maps To display the associations we have drawn adapted heat maps, visualizing the independent gene expression variable on the vertical y-axis and the dependent clinical symptom variables on the horizontal x-axis. In these feature-expression heat maps the associations between independent and dependent variables are represented by circles in each respective compartment. The regression coefficients indicating effect size are represented by the type (red=positive, blue=negative) and intensity of the color, whereas the statistical significance of the analyses is represented by the radius of the circles. To facilitate the visual identification of meaningful clusters of association both axes were ordered into functional categories and sequences. Monocyte gene expression was ordered on the sub-clusters 1a (inflammation), 1b (transcription) and 2 (adhesion, motility and chemotaxis) categories, based on a cluster analyses performed by R.C. Drexhage5. Clinical symptoms were ordered into symptom categories. For psychotic symptoms a division into delusions, hallucinations and psychomotor symptoms

35 is used. Depressive symptoms were divided into general symptoms, melancholic symptoms and atypical symptoms. Manic symptoms were not subdivided into categories. Furthermore, symptoms were ordered into a phenomenological sequence for mood symptoms (core mood symptoms, thought symptoms, psychosomatic symptoms, motor symptoms, food intake symptoms, sleep symptoms, higher functional symptoms) and psychotic symptoms. The heat maps were drawn with the corrplot package, by T. Wei33, on R, release 2.14.1 (R Development Core Team 2011, Vienna, Austria). Centered dots were added to the compartments that complied with a FDR below 0.2, thus allowing 1/5 to be false positive. See chapter 3 for a more detailed description of the feature-expression heat map method.

Results

Associations between lifetime psychotic symptoms and monocyte pro-inflammatory gene expression The associations between the psychotic symptoms and the expression of genes belonging to each subcluster were depicted in the psychosis feature-expression heat map (figure 4). This heat map demonstrated a clustering group of associations between the sub-cluster 2 genes and the psychomotor symptoms. Clustering groups of associations with delusional or hallucinatory symptoms could not be demon strated. None of the individual analyses were statistically significant after FDR correction. Although not fitting into a clustering group of associations, the negative association between thought withdrawal and sub-cluster 1 gene expression, and the positive association between olfactory hallucinations and sub-cluster 1 gene expression are conspicuous.

Associations between current mood symptoms and monocyte pro-inflammatory gene expression The manic feature-expression heat map (figure 5) was created of the associations between manic symptoms and the expression of genes belonging to each subcluster. This feature-expression heat map showed a clustering group of associations between the sub-cluster 2 genes and manic symptoms. Sixteen of these associations were significant after FDR correction, especially in the associations with the symptoms increased speech and increased motor activity. The associations between depressive symptoms and the expression of genes belonging to each subcluster were drawn up in the depression feature-expression heat map (figure 6). With regard to this feature-expression heat map, clustering groups of associations could not be identified and none of the individual analyses were significant

36 FIGURE 4 Association between lifetime psychotic symptoms and monocyte pro-inflammatory gene expression

s t 2

elusions of being controlled DelusionsPersecutory of referenceGrandiose delusionsSomatic delusionsReligious delusionsDelusions delusionsD of guilThoughtThought insertionThought withdrawaAuditory broadcastingl Visual hallucinations hallucinationTactile GustatoryhallucinationsOlfactory hallucinationsExcessive hallucinationsEcholalia motorGrossly oractivity echopraxiaDisorganized disorganized speechbehavior disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship DUSP2 ATF3 PDE4B IL6 p IL1B 0.001 TNF 0.005 TNFAIP3 0.01 BCL2A1 0.05 PTX 0.1 PTGS inflammation 0.2 CCL20 CXCL2 EREG effect size CXCL3 3 MXD

EGR3 2 F3 MAFF 1 THBS

SERP transcription RGC32 0 PTPN NAB2 -1 MAPK EMP1 STX1 -2

SDR motility / adhesion / chemotaxis CCL2 -3 CCL7 CDC42 FABP CD9 HSPA1 CCR2 delusions hallucinations psychomotor psychotic symptoms

Heat map depicting the regression coefficient of the association between lifetime psychotic symptoms and gene expression. Psychotic symptoms were ordinally measured. gene expression was expressed as z-transformed –∆∆Ct. Statistical analysis was performed using ordered logistic regression. Contrast was set to 3. Blank compartments either represent a small regression coefficient and statistical probability or an analysis with insufficient observations. None of the analyses were significant below the 0.2 false discovery rate (FDR) threshold for multiple testing.

37 FIGURE 5 Association between manic symptoms and monocyte pro-inflammatory gene expression

ontent ElevatedIrritability MoodC LanguageIncreased / ThoughtIncreased Motor DisorderDecreased ActivitySpeechDisruptive orSlee EnergSexualp ory AggressiveAppearance InterestInsight Behavio DUSP2 ATF3 p PDE4B 0.001 IL6 0.005 IL1B 0.01 TNF 0.05 TNFAIP3 0.1 BCL2A1 0.2 PTX inflammation PTGS FDR<0.2 CCL20 CXCL2 EREG CXCL3 MXD

EGR3 effect F3 size 3 MAFF THBS 2 SERP transcription RGC32 PTPN 1 NAB2 MAPK 0 EMP1 STX1 -1

SDR motility / adhesion / chemotaxis CCL2 CCL7 -2 CDC42 FABP CD9 -3 HSPA1 CCR2 manic symptoms

Heat map depicting the regression coefficient of the association between actual manic symptoms (within three days of blood sampling) and gene expression. Depressive symptoms were ordinally measured. gene expression was expressed as z-transformed –∆∆Ct. Statistical analysis was performed using ordered logistic regression. Contrast was set to 3. Blank compartments either represent a small regression coefficient and statistical probability or an analysis with insufficient observations. Dotted circles represent significance below the 0.2 false discovery rate (FDR) threshold for multiple testing.

38 FIGURE 6 Association between depressive symptoms and monocyte pro-inflammatory gene expression

omni t 2 n d

Mood s utlook disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship O

Sad MoodIrritableAnxiou MoodSomaticSympathetic ComplaintsPanic /Arousal PhobicGastrointestinalSleep Symptoms OnsetMid-Nocturnal Symptom InsomniaSexualsConcentration Interes Ins Selft Future / Decision SuicidalOutlook. MakingInvolvemen IdeatioMoodt VariationQualityPsychomotor of MooPsychomotor SlowingDecreased AgitationDecreased AppetiteEarly Weigh MorningPleasure ReactivityInsomnia / EnjoymentLeaden of MoodIncreased ParalysisIncreased /Appetite PhysicalEnergy Weight Energy Hypersomni/ FatiguabilityInterpersonala Sensitivity DUSP2

ATF3 p PDE4B 0.001 IL6 0.005 IL1B 0.01 TNF 0.05 TNFAIP3 0.1 BCL2A1 0.2 PTX inflammation PTGS CCL20 CXCL2 EREG CXCL3 MXD EGR3 effect size F3 3 MAFF

THBS 2

SERP transcription RGC32 PTPN 1 NAB2

MAPK 0 EMP1 STX1 -1 SDR motility / adhesion / CCL2 chemotaxis CCL7 -2 CDC42 FABP -3 CD9 HSPA1 CCR2 general melancholical atypical depressive symptoms depressive symptoms depressive symptoms

Heat map depicting the regression coefficient of the association between actual depressive symptoms (within three days of blood sampling) and gene expression. Depressive symptoms were ordinally measured. Gene expression was expressed as z-transformed –∆∆Ct. Statistical analysis was performed using ordered logistic regression. Contrast was set to 3. Blank compartments either represent a small regression coefficient and statistical probability or an analysis with insufficient observations. None of the analyses were significant below the 0.2 false discovery rate (FDR) threshold for multiple testing.

39 after FDR correction. Although not fitting into clustering groups of associations, some associations e.g. with a decreased appetite, with sympathetic arousal and with sad and irritable mood are notable. No noticeable difference exists in the number of associations between the atypical, melancholic and general depressive symptoms as separate groups.

Association between age at onset, duration of illness and monocyte pro-inflammatory gene expression The associations between age at onset, duration of illness and the expression of genes belonging to each subcluster are described in table 3. Using univariable analyses many sub-cluster 1a genes associated significantly with the age at onset in a negative manner. Some sub-cluster 1a genes associated significantly with duration of illness in a positive manner. When entered into multivariable models, PDE4B, IL6, TNFAIP3 and PTX3 gene expression associated significantly with the age at onset in a negative manner, where no concurrent associations were demonstrated with duration of illness (table 3). The genes ATF3 and IL1 associated significantly with duration of illness in a positive manner in these analyses, where concurrent associations with age at onset could not be demonstrated. Associations between any of the sub-cluster 1b genes and age at onset or duration of illness could not be found. Many sub-cluster 2 genes associated significantly with the duration of illness in a positive manner and some sub-cluster 2 genes associated significantly with age at onset in a negative manner (table 3). When entered into multivariable models, MAPK6, EMP1, STX1, DHRS3 and CCL2 gene expression associated significantly with duration of illness in a positive manner, where age at onset was found not to be associated in any of these models for sub-cluster 2 genes (table 3). Addition of an interaction term or gender to the multivariable analyses did not alter the results markedly. Significant associations could not be demonstrated between individual gene expression and age, except for MAPK6.

40 TABLE 3 Associations between age at onset, duration of illness and individual monocyte pro-inflammatory gene expression

Univariable Multivariable mRNA Age at onset - Duration of illness - Age at onset – Duration of illness genes RC (ci) RC (ci) RC (ci) RC (ci) DUSP2 -0.0716 (-0.128 - -0.0152) ** 0.0478 (-0.0124 - 0.108) -0.0639 (-0.128 - 0.000227) * 0.0172 (-0.0490 - 0.0835) ATF3 -0.0535 (-0.0955 - -0.0115) ** 0.0678 (0.0257 - 0.110) *** -0.0294 (-0.0752 - 0.0163) 0.0537 (0.00641 - 0.101) ** PDE4B -0.0785 (-0.126 - -0.0308) *** 0.0669 (0.0161 - 0.118) ** -0.0618 (-0.115 - -0.00831) ** 0.0373 (-0.0180 - 0.0926) IL6 -0.233 (-0.403 - -0.0626) *** 0.106 (-0.126 - 0.338) -0.232 (-0.417 - -0.0478) ** 0.00281 (-0.230 - 0.236) IL1 -0.0798 (-0.163 - 0.00370) * 0.122 (0.0340 - 0.211) *** -0.038 (-0.128 - 0.0522) 0.104 (0.00546 - 0.203) ** 2 TNF -0.0653 (-0.126 - -0.00512) ** 0.0383 (-0.0316 - 0.108) -0.0623 (-0.130 - 0.00518) * 0.00781 (-0.0681 - 0.0838) TNFAIP3 -0.0948 (-0.161 - -0.0283) *** 0.0616 (-0.0283 - 0.151) -0.0886 (-0.160 - -0.0169) ** 0.022 (-0.0686 - 0.113) BCL2A1 -0.0514 (-0.107 - 0.00432) * 0.0685 (0.0121 - 0.125) ** -0.0263 (-0.0881 - 0.0354) 0.0559 (-0.00795 - 0.120) *

PTX3 -0.1000 (-0.177 - -0.0231) ** 0.0794 (-0.0227 - 0.181) -0.0889 (-0.172 - -0.00628) ** 0.0398 (-0.0647 - 0.144) disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship PTGS -0.0714 (-0.145 - 0.00273) * 0.0355 (-0.0618 - 0.133) -0.0702 (-0.150 - 0.0100) * 0.00426 (-0.0972 - 0.106) CCL20 -0.204 (-0.371 - -0.0368) ** 0.203 (-0.0139 - 0.420) * -0.168 (-0.346 - 0.00970) * 0.128 (-0.0965 - 0.353) CXCL2 -0.112 (-0.204 - -0.0206) ** 0.131 (0.0377 - 0.224) *** -0.068 (-0.169 - 0.0330) 0.0983 (-0.00613 - 0.203) * EREG -0.0931 (-0.185 - -0.00100) ** 0.0892 (-0.0118 - 0.190) * -0.0681 (-0.183 - 0.0471) 0.0458 (-0.0787 - 0.170) CXCL3 -0.086 (-0.229 - 0.0568) 0.149 (0.000770 - 0.297) ** -0.00711 (-0.180 - 0.166) 0.144 (-0.0426 - 0.331) MXD -0.0187 (-0.0571 - 0.0196) 0.0309 (-0.00966 - 0.0714) -0.00288 (-0.0502 - 0.0444) 0.029 (-0.0221 - 0.0802) EGR3 0.0152 (-0.0614 - 0.0917) 0.0177 (-0.0650 - 0.100) 0.0382 (-0.0573 - 0.134) 0.0422 (-0.0611 - 0.145) F3 -0.0885 (-0.203 - 0.0265) 0.115 (-0.00756 - 0.237) * -0.0398 (-0.182 - 0.102) 0.0891 (-0.0642 - 0.242) MAFF -0.0498 (-0.123 - 0.0234) 0.0759 (-0.000765 - 0.153) * -0.0128 (-0.102 - 0.0765) 0.0677 (-0.0289 - 0.164) THBS -0.0604 (-0.143 - 0.0221) 0.0544 (-0.0357 - 0.145) -0.0471 (-0.151 - 0.0566) 0.0243 (-0.0878 - 0.136) SERPINB2 -0.0536 (-0.173 - 0.0664) 0.0956 (-0.0309 - 0.222) -0.00196 (-0.150 - 0.146) 0.0944 (-0.0653 - 0.254) RGC32 -0.0448 (-0.140 - 0.0501) 0.107 (0.0106 - 0.203) ** 0.0209 (-0.0913 - 0.133) 0.120 (-0.00101 - 0.241) * PTPN -0.0529 (-0.104 - -0.00184) ** 0.0691 (0.0205 - 0.118) *** -0.0175 (-0.0792 - 0.0443) 0.0590 (-0.00181 - 0.120) * NAB2 -0.0246 (-0.0875 - 0.0383) 0.0284 (-0.0365 - 0.0934) -0.0151 (-0.0865 - 0.0564) 0.0212 (-0.0526 - 0.0950) MAPK6 -0.00521 (-0.0705 - 0.0601) 0.0619 (-0.00358 - 0.127) * 0.0287 (-0.0430 - 0.100) 0.0756 (0.00148 - 0.150) ** EMP1 -0.0655 (-0.124 - -0.00698) ** 0.0894 (0.0309 - 0.148) *** -0.0324 (-0.0963 - 0.0314) 0.0739 (0.00787 - 0.140) ** STX1 -0.101 (-0.213 - 0.0107) * 0.157 (0.0525 - 0.262) *** -0.0166 (-0.146 - 0.113) 0.148 (0.0164 - 0.279) ** DHRS3 -0.062 (-0.142 - 0.0183) 0.117 (0.0437 - 0.190) *** 0.00491 (-0.0842 - 0.0940) 0.120 (0.0289 - 0.211) ** CCL2 -0.0475 (-0.123 - 0.0281) 0.0969 (0.0205 - 0.173) ** -0.0059 (-0.0887 - 0.0769) 0.0940 (0.00692 - 0.181) ** CCL7 -0.130 (-0.260 - 0.000696) * 0.182 (0.0485 - 0.316) *** -0.0632 (-0.206 - 0.0800) 0.151 (-0.000107 - 0.303) * CDC42 -0.0586 (-0.100 - -0.0170) *** 0.0637 (0.0210 - 0.106) *** -0.0383 (-0.0842 - 0.00763) 0.0454 (-0.00208 - 0.0928) * FABP 0.0262 (-0.0240 - 0.0763) -0.00999 (-0.0616 - 0.0417) 0.0309 (-0.0318 - 0.0936) 0.00828 (-0.0553 - 0.0719) CD9 -0.0204 (-0.0723 - 0.0314) 0.0524 (0.00279 - 0.102) ** 0.0146 (-0.0464 - 0.0755) 0.0610 (-0.000852 - 0.123) * HSPA1 0.0530 (0.000640 - 0.105) ** -0.0357 (-0.0909 - 0.0195) 0.0492 (-0.0164 - 0.115) -0.00663 (-0.0732 - 0.0599) CCR2 0.00109 (-0.0527 - 0.0549) -0.00043 (-0.0463 - 0.0455) 0.00132 (-0.0704 - 0.0730) 0.000299 (-0.0609 - 0.0615)

Z-score transformation of the relative gene expression was applied to facilitate mutual comparison. Statistical analysis was performed using linear regression. Multivariable regression analyses consisted of age at onset and duration of illness as dependent variables. Correction for gender or the addition of an interaction term in the multivariable analyses did not alter results significantly. Results are reported as regression coefficients (RC) with 95% confidence intervals (ci). Significance is reported at the 10% (p=0.1)*, 5% (p=0.05)** and 1% (p=0.01)*** level.

41 Association between monocyte pro-inflammatory gene expression and current medication use The associations between the expression of genes belonging to each subcluster and current medication use were drawn up in the medication feature-expression heat map (figure 7). This feature-expression heat map showed a clustering group of negative associations between current use of an antidepressant and sub-cluster 2 genes. Furthermore, a clustering group of weaker positive associations was found between current use of anti-epileptics, especially carbamazepine, and sub-cluster 2 genes. Also noticeable were relative weak, statistical non-significant associations between current use of lithium and some sub-cluster 1a genes, and between current use of antipsychotics and sub-cluster 1a genes.

FIGURE 7 Association between current medication use and monocyte pro-inflammatory gene expression

3 S F 9

DUSP2 ATF3 PDE4b Il6 Il1 TN TNFaip3 BCL2a1 PTX PTG CCL20 CXCL2 EREG CXCL3 MXD EGR3 F3 MAF THBS SERP RGC32 PTPN NAB2 MAPK EMP1 STX1 DHRS CCL2 CCL7 CDC42 FABP5 CD HSPa1a CCR2 Lithium

Valproate

Carbamazepine

Lamotrigine

Antidepressants

Antipsychotics

Benzodiazepines

p Effect size 0. 1

0. 2 -3 -2 -1 0 1 23 0.0 1 0.0 5 0.001 0.005 FDR<0. 2

Heat map depicting the regression coefficient of the association between medication and gene expression. Medication was qualitatively measured. gene expression was expressed as z-transformed –∆∆Ct. Statistical analysis was performed using linear regression. Contrast was set to 1.5. Blank compartments represent a small regression coefficient and statistical probability. The dotted circle represents significance below the 0.2 false discovery rate (FDR) threshold for multiple testing.

42 Discussion

To our knowledge this is the first study on the associations between an extensive set of clinical features and monocyte gene expression in BD and it can be regarded as a next step in the converging approach between immunology and psychopathology in unraveling the complex pathophysiological mechanisms of BD. Our a priori for mulated hypothesis that lifetime psychotic features would be associated with the 2 pro-inflammatory monocyte gene expression could not be confirmed. However, our method visualized the following interesting findings in BD patients: (1) a possible disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship relation between pro-inflammatory gene expression and manic symptomatology, (2) a differential immune activation related to an earlier age at onset, (3) an increased immune system dysregulation during the course of the disorder and (4) support for the concept of an immune suppressive action of some of the mood regulating medications. The association between immune activation and manic / psychomotor symptoms is evident in both the analyses with the lifetime psychotic symptoms and in the analyses with the manic symptoms, in other words irrespective of the questionnaire used. This association between immune activation and manic symptoms is supported by a previous finding by Dickerson et al., who described the YMRS and individual manic symptoms (speech, appearance, irritability, language–thought disorder, thought content and increased motor activity/energy) to be related to an increased serum C-reactive protein(CRP)34. Findings by Brietzke et al. who described pro-inflammatory cytokines to be increased in manic BD patients35 and Cunha et al. who Cnha et al. ards this, there arepatients. found CRP to be increased in manic BD patients, but not in euthymic and depressed ones, are also compatible with this association. Where CRP levels are known to be induced by cytokines36, it must be noted that a dis crepancy between cytokine gene expression and protein synthesis is known to exist4. In general, the association between the pro-inflammatory gene expression and depressive symptoms is less outspoken than it is with manic symptoms. However, some associations seem to exist with individual symptoms, e.g. decreased appetite, sympathetic arousal and sad and irritable mood. These associations seem to have some analogy with what is known as sickness behavior during the course of an infec- tion37–39. We could not find indications for a specific relation between atypical or melancholic symptoms and monocyte activation. To our knowledge this is the first study investigating an association between the subtypes of depression and biological parameters in BD. In unipolar major depressive disorder a differential role of HPA-axis function and inflammation, including CRP and cytokine levels, has been reported in melancholic versus atypical depression40.

43 The association between age at onset and monocyte activation supports the notion that patients with an earlier age at onset have a biologically more severe form of the illness. Several authors have demonstrated increased morbidity in BD patients with an earlier age at onset41–43, which is in agreement with this observation. On the biological side, genetic differences between BD patients with earlier and later age at onset have been described44,45. Manenschijn, Spijker et al. demonstrated an increase in cortisol in patients with a later age at onset, suggesting BD beginning on an earlier age to be less associated with HPA-axis disturbances46. Additionally, the present study supports an association with the duration of illness and the theory that the pathogenic mechanisms cause further impairment in immune system regulatory mechanisms with the progression of the illness. This association is consistent with the findings from Drexhage et al.5, partly based on the same dataset, and from Soreca et al., who described an increased medical burden, e.g. more cardio- vascular, endocrine and metabolic disease, in BD patients with al longer duration of illness47. The observation of a differential immune activation related to an earlier age of onset and an increased immune dysregulation during the course of the disorder can also be interpreted in view of the staging hypothesis of BD. A staging model founded on neurobiological correlates of distinct stages of BD could potentially predict clinical care needs and assist in refining treatment options48. Kauer-Sant’Anna et al. described an increased TNF and decreased IL6 cytokine levels in late-stage versus earlier-stage BD patients, all levels being elevated when compared to healthy controls49. In the present study TNF and IL6 gene expression was not found to be associated with longer duration of illness, but increased gene expression of these molecules was related to an earlier age at onset. In an attempt to explain the known discrepancy between cytokine gene expression and protein synthesis4 it could be argued that earlier-stage patients in the Kauer-Sant’Anna study potentially had a biologically more severe form of the illness. That would be consistent with the earlier age at onset of this group and may explain the differentiated cytokine expression between the early-stage and late-stage groups. Based on the present study additional research is warranted on the ATF3, IL1, MAPK6, STX1, DHRS3, CCL2 gene expression as potential neurobiological staging markers of the progression of the illness, whereas PDE4b, IL6, TNFAIP3, PTX3 gene expression could perhaps predict a more immune mediated profile in the earlier stages of the disorder. In addition to previous findings by Padmos and Drexhage4,5, who described relations between current use of lithium and antipsychotics and pro-inflammatory gene expression, we found associations between current use of antidepressants (negative), and of anti-epileptics (positive) and pro-inflammatory gene expression. Our study results support the theory that the effects of lithium and antipsychotics mainly concern down regulation of some genes and this could also be the case for antide-

44 pressants. Based on the relation between monocyte activation and manic symptoms, it can be argued that the addition of anti-inflammatory medication to standard anti-mania treatment could be a beneficial addition treatment strategy for manic episodes in addition to depressive episodes50. Although treatment studies have thus far not been performed in this regard, low dose (30-80mg/day) acetylsalicylic acid was found to produce a statistically significant duration-independent reduction in the relative risk of clinical deterioration in subjects on lithium, in a large pharma- 2 co-epidemiological study51 and further research into this treatment possibility is warranted. disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship The present study has several limitations. Firstly, the present study focuses on the pro-inflammatory gene expression of monocytes, which is a select part of the complex immune system, and generalized statements should be considered in that regard. Secondly, all patients were naturalistically treated and none of them was ‘medication naive’. The positive association between monocyte activation and current use of anti-epileptic leaves room for argumentation that medication is a causal factor for the increased monocyte pro-inflammatory gene expression in BD. Thirdly, an effect of age or gender cannot completely be ruled out. A fourth limitation is contained within the original selection design of the analyzed genes, where only highly over- and under-expressed genes, which were clearly involved in inflammatory processes, where selected, possibly ruling out important genes. The cross-sectional design of the study being suboptimal for analyses with regard to disease progression is a specific fifth limitation concerning the age at onset and duration of illness analyses. The final and in our opinion most important limitation concerns the multiple testing in the analyses leading to an increased risk of type I errors despite the application of FDR correction, combined with relatively small sample size. Nevertheless, we consider it as important for hypothesis forming and testing in further studies. Because of the relatively small sample size of BD-II patients no separate sub-group analyzes were performed comparing patients with BD-I or BD-II disorder. The nature of the associations between psychiatric symptoms and pro-inflammatory monocyte activation is as yet unknown. With the current state of knowledge about these interactions we are unable to clarify why and how activation of sub-clusters of genes, or even specific gene activation, is related to a specific (group) of symptoms. However, we regard them as intermediary phenomenon in the neuroinflammation theory of BD, positioned between the activated inflammatory response system and the phenotypical expression (symptoms), as depicted in figure 1. A theoretical model of action originates in activation of the mononuclear phagocyte system at the level of the brain (microglia), the circulation (monocytes), and the tissues (macrophages) as a key element in the pathogenesis of major psychiatric disorders. Whether there exists a direct migration of activated monocytes to the brain in psychiatric disease needs further exploration. Also, the role of inflammatory cytokine

45 exchange among the various compartments (circulation, brain, and peripheral tissues, such as adipose tissue and lymphoid tissue) needs clarification. Several routes have been proposed for cytokines to enter and act on the brain, e.g. by altering the blood- brain barrier and by affecting neuronal afferents such as the vagus nerve52. Indeed, the vagus nerve is known to be essential for balancing anti- and pro-inflammation during sepsis53. Supported by multiple observations of mania in vagus nerve stimulation, a treatment for refractory epilepsy thought to have its effect on the limbic system and frontal cortex54, it is tempting to question the importance of the role of the vagus nerve in the association between monocyte activation and the psychiatric symptoms in BD. In our opinion, especially the association between manic symptoms and immune activation deserves verification in further studies. The psychoimmunological model in BD does not stand on its own, but concerns other psychiatric disorders as well. This makes it interesting to study our findings also in other disorders, like SZ where motor function symptoms also play an important role.

Acknowledgements

We thank Harm de Wit and Annemarie Wijkhuijs for their excellent technical assistance and Sanne Kemner for her assistance in retrieving archived data.

46 References

1. Schulze TG. Genetic research into bipolar disorder: the need for a research framework that integrates sophisticated molecular biology and clinically informed phenotype characterization. Psychiatr Clin North Am. 2010 Mar;33(1):67–82. 2. Maes M, Smith R, Scharpe S. The monocyte-T-lymphocyte hypothesis of major depression. Psychoneuroendocrinology. 1995 Jan;20(2):111–6. 3. Calabrese JR, Skwerer RG, Barna B, Gulledge AD, Valenzuela R, Butkus A, et al. Depression, 2 immunocompetence, and prostaglandins of the E series. Psychiatry Res. 1986 Jan;17(1):41–7. 4. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 5. Drexhage RC, van der Heul-Nieuwenhuijsen L, Padmos RC, van Beveren N, Cohen D, Versnel M a, et al. Inflammatory gene expression in monocytes of patients with schizophrenia: overlap and difference with bipolar disorder. A study in naturalistically treated patients. Int J Neuropsychopharmacol. 2010 Nov;13(10):1369–81. 6. Pedrosa E, Locker J, Lachman HM. Survey of schizophrenia and bipolar disorder candidate genes using chromatin immunoprecipitation and tiled microarrays (ChIP-chip). J Neurogenet. 2009 Jan;23(3):341–52. 7. Kähler AK, Otnaess MK, Wirgenes K V, Hansen T, Jönsson EG, Agartz I, et al. Association study of PDE4B gene variants in Scandinavian schizophrenia and bipolar disorder multicenter case-control samples. Am J Med Genet B Neuropsychiatr Genet. 2010 Jan 5;153B(1):86–96. 8. Clerici M, Arosio B, Mundo E, Cattaneo E, Pozzoli S, Dell’osso B, et al. Cytokine polymorphisms in the pathophysiology of mood disorders. CNS Spectr. 2009 Aug;14(8):419–25. 9. Papiol S, Molina V, Desco M, Rosa A, Reig S, Sanz J, et al. Gray matter deficits in bipolar disorder are associated with genetic variability at interleukin-1 beta gene (2q13). Genes Brain Behav. 2008 Oct;7(7):796–801. 10. Pae C-U, Lee K-U, Han H, Serretti a, Jun T-Y. Tumor necrosis factor alpha gene-G308A polymorphism associated with bipolar I disorder in the Korean population. Psychiatry Res. 2004 Jan 30;125(1):65–8. 11. Czerski PM, Rybakowski F, Kapelski P, Rybakowski JK, Dmitrzak-Weglarz M, Leszczyńska- Rodziewicz A, et al. Association of tumor necrosis factor -308G/A promoter polymorphism with schizophrenia and bipolar affective disorder in a Polish population. Neuropsychobiology. 2008 Jan;57(1-2):88–94. 12. Mansour HA, Talkowski ME, Wood J, Chowdari K V, McClain L, Prasad K, et al. Association study of 21 circadian genes with bipolar I disorder, schizoaffective disorder, and schizophrenia. Bipolar Disord. 2009 Dec;11(7):701–10. 13. Gallitano AL, Tillman R, Dinu V, Geller B. Family-based association study of early growth response gene 3 with child bipolar I disorder. J Affect Disord. 2012 May;138(3):387–96. 14. Pae C-U, Kim J-J, Yu H-S, Lee C-U, Lee S-J, Jun T-Y, et al. Monocyte chemoattractant protein-1 promoter -2518 polymorphism may have an influence on clinical heterogeneity of bipolar I disorder in the Korean population. Neuropsychobiology. 2004 Jan;49(3):111–4. 15. Roh M-S, Lee KY, Joo E-J, Lee N, Kim YS. No association of the MCP-1 promoter A-2518G polymorphism with bipolar disorder in the Korean population. Neurosci Lett. 2007 Oct 29;427(1):1–5. 16. Altamura a C, Mundo E, Cattaneo E, Pozzoli S, Dell’osso B, Gennarelli M, et al. The MCP-1 gene (SCYA2) and mood disorders: preliminary results of a case-control association study. Neuroimmunomodulation. 2010 Jan;17(2):126–31. 17. Iwayama Y, Hattori E, Maekawa M, Yamada K, Toyota T, Ohnishi T, et al. Association analyses between brain-expressed fatty-acid binding protein (FABP) genes and schizophrenia and bipolar

47 disorder. Am J Med Genet B Neuropsychiatr Genet. 2010 Mar 5;153B(2):484–93. 18. Drzyzga L, Obuchowicz E, Marcinowska A, Herman ZS. Cytokines in schizophrenia and the effects of antipsychotic drugs. Brain Behav Immun. 2006 Nov;20(6):532–45. 19. Pollmächer T, Haack M, Schuld A. Effects of antipsychotic drugs on cytokine networks. J Psychiatr …. 2000;34(6):369–82. 20. Rybakowski JK. Antiviral and immunomodulatory effect of lithium. Pharmacopsychiatry. 2000 Sep;33(5):159–64. 21. Nahman S, Belmaker RH, Azab AN. Effects of lithium on lipopolysaccharide-induced inflammation in rat primary glia cells. Innate Immun. 2012 Jun;18(3):447–58. 22. Vonk R, van der Schot AC, Kahn RS, Nolen W a, Drexhage H a. Is autoimmune thyroiditis part of the genetic vulnerability (or an endophenotype) for bipolar disorder? Biol Psychiatry. 2007 Jul 15;62(2):135–40. 23. Padmos RC, Van Baal GCM, Vonk R, Wijkhuijs AJM, Kahn RS, Nolen W a, et al. Genetic and environmental influences on pro-inflammatory monocytes in bipolar disorder: a twin study. Arch Gen Psychiatry. 2009 Sep;66(9):957–65. 24. Reichart CG, Wals M, Hillegers MHJ, Ormel J, Nolen W a., Verhulst FC. Psychopathology in the adolescent offspring of bipolar parents. J Affect Disord. 2004 Jan;78(1):67–71. 25. Hillegers MH, Reichart CG, Wals M, Verhulst FC, Ormel J, Nolen W a. Five-year prospective outcome of psychopathology in the adolescent offspring of bipolar parents. Bipolar Disord. 2005 Aug;7(4):344–50. 26. APA. Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR). Arlington, VA: American Psychiatric Association; 2000. 27. Young RC, Biggs JT, Ziegler VE, Meyer D a. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978 Nov 1;133(5):429–35. 28. Rush AJ, Weissenburger J, Giles E, Schlesser MA, Fulton CL. The Inventory Preliminary for Depressive Findings Symptomatology ( IDS ): Depression. 1986;65–87. 29. Knijff EM, Breunis MN, van Geest MC, Kupka RW, Ruwhof C, de Wit HJ, et al. A relative resistance of T cells to dexamethasone in bipolar disorder. Bipolar Disord. 2006 Dec;8(6):740–50. 30. Staal FJT, Weerkamp F, Baert MRM, van den Burg CMM, van Noort M, de Haas EFE, et al. Wnt target genes identified by DNA microarrays in immature CD34+ thymocytes regulate proliferation and cell adhesion. J Immunol. 2004 Jan 15;172(2):1099–108. 31. User Bulletin # 2 ABI P RISM 7700 Sequence Detection System SUBJECT: Relative Quantitation of Gene Expression. System. 2001;1997. 32. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med. 1990 Jul;9(7):811–8. 33. Wei T. corrplot: Visualization of a correlation matrix [Internet]. R package version 0.73. 2013. Available from: http://cran.r-project.org/package=corrplot 34. Dickerson F, Stallings C, Origoni A, Boronow J, Yolken R. Elevated serum levels of C-reactive protein are associated with mania symptoms in outpatients with bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry. 2007 May 9;31(4):952–5. 35. Brietzke E, Stertz L, Fernandes BS, Kauer-Sant’anna M, Mascarenhas M, Escosteguy Vargas A, et al. Comparison of cytokine levels in depressed, manic and euthymic patients with bipolar disorder. J Affect Disord. Elsevier B.V.; 2009 Aug;116(3):214–7. 36. Ramadori G, Van Damme J, Rieder H, Meyer zum Büschenfelde KH. Interleukin 6, the third mediator of acute-phase reaction, modulates hepatic protein synthesis in human and mouse. Comparison with interleukin 1 beta and tumor necrosis factor-alpha. Eur J Immunol. 1988 Aug;18(8):1259–64. 37. Hart BL. Biological basis of the behavior of sick animals. Neurosci Biobehav Rev. 1988 Jan;12(2):123–37. 38. Charlton BG. The malaise theory of depression: major depressive disorder is sickness behavior and antidepressants are analgesic. Med Hypotheses. 2000 Jan;54(1):126–30.

48 39. Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci. 2008 Jan;9(1):46–56. 40. Lamers F, Vogelzangs N, Merikangas KR, de Jonge P, Beekman a TF, Penninx BWJH. Evidence for a differential role of HPA-axis function, inflammation and metabolic syndrome in melancholic versus atypical depression. Mol Psychiatry. Nature Publishing Group; 2012 Oct 23;(February):1–8. 41. Bellivier F, Golmard JL, Henry C, Leboyer M, Schürhoff F. Admixture analysis of age at onset in bipolar I affective disorder. Arch Gen Psychiatry. 2001 May 1;58(5):510–2. 42. Leboyer M, Henry C, Paillere-Martinot M-L, Bellivier F. Age at onset in bipolar affective disorders: a 2 review. Bipolar Disord. 2005 Apr;7(2):111–8. 43. Coryell W, Fiedorowicz J, Leon AC, Endicott J, Keller MB. Age of onset and the prospectively Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder bipolar in expression gene monocyte related inflammation and features clinical between Relationship observed course of illness in bipolar disorder. J Affect Disord. Elsevier; 2012 Oct 10;1–5. 44. Lin P-I, McInnis MG, Potash JB, Willour VL, Mackinnon DF, Miao K, et al. Assessment of the effect of age at onset on linkage to bipolar disorder: evidence on chromosomes 18p and 21q. Am J Hum Genet. 2005 Oct;77(4):545–55. 45. Etain B, Mathieu F, Rietschel M, Maier W, Albus M, McKeon P, et al. Genome-wide scan for genes involved in bipolar affective disorder in 70 European families ascertained through a bipolar type I early-onset proband: supportive evidence for linkage at 3p14. Mol Psychiatry. 2006 Jul;11(7):685–94. 46. Manenschijn L, Spijker AT, Koper JW, Jetten AM, Giltay EJ, Haffmans J, et al. Long-term cortisol in bipolar disorder: Associations with age of onset and psychiatric co-morbidity. Psychoneuroendocrinology. Elsevier Ltd; 2012 Dec;37(12):1960–8. 47. Soreca I, Fagiolini A, Frank E, Houck PR, Thompson WK, Kupfer DJ. Relationship of general medical burden, duration of illness and age in patients with bipolar I disorder. J Psychiatr Res. 2008 Sep;42(11):956–61. 48. Berk M, Hallam KT, McGorry PD. The potential utility of a staging model as a course specifier: a bipolar disorder perspective. J Affect Disord. 2007 Jun;100(1-3):279–81. 49. Kauer-Sant’Anna M, Kapczinski F, Andreazza AC, Bond DJ, Lam RW, Young LT, et al. Brain-derived neurotrophic factor and inflammatory markers in patients with early- vs. late-stage bipolar disorder. Int J Neuropsychopharmacol. 2009 May;12(4):447–58. 50. Müller N, Schwarz MJ, Dehning S, Douhe a, Cerovecki a, Goldstein-Müller B, et al. The cyclooxygenase-2 inhibitor celecoxib has therapeutic effects in major depression: results of a double-blind, randomized, placebo controlled, add-on pilot study to reboxetine. Mol Psychiatry. 2006 Jul;11(7):680–4. 51. Stolk P, Souverein PC, Wilting I, Leufkens HGM, Klein DF, Rapoport SI, et al. Is aspirin useful in patients on lithium? A pharmacoepidemiological study related to bipolar disorder. Prostaglandins Leukot Essent Fatty Acids. Elsevier; 2010 Jan;82(1):9–14. 52. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 53. Kessler W, Diedrich S, Menges P, Ebker T, Nielson M, Partecke LI, et al. The role of the vagus nerve: modulation of the inflammatory reaction in murine polymicrobial sepsis. Mediators Inflamm. 2012 Jan;2012(2012):467620. 54. Gerson R, Murray E, Price B, Frankel M, Douglass LM, Cunningham M. Mania following vagus nerve stimulation: a case report and review of the literature. Epilepsy Behav. Elsevier Inc.; 2011 Jan;20(1):138–40. 55. Drexhage RC. Immuno-neuro-endocrine networks: A study on the inflammatory state of circulating monocytes and CD4+ T cells in psychiatric and endocrine autoimmune disease. University Medical Center Rotterdam, The Netherlands; 2011. 56. Suppes T, Leverich GS, Keck PE, Nolen W a, Denicoff KD, Altshuler LL, et al. The Stanley Foundation Bipolar Treatment Outcome Network. II. Demographics and illness characteristics of the first 261 patients. J Affect Disord. 2001 Dec;67(1-3):45–59.

CHAPTER 3 Feature-expression heat maps A new visual method to explore complex associations between two variable sets

Bartholomeus C.M. Haarman, Rixt F. Riemersma-Van der Lek, Willem A. Nolen, R. Mendes, Hemmo A. Drexhage, Huibert Burger

Journal of Biomedical Informatics. 2015 Oct 14;53:156–61. Abstract

Introduction Existing methods such as correlation plots and cluster heat maps are insufficient in the visual exploration of multiple associations between genetics and phenotype, which is of importance to achieve a better understanding of the pathophysiology of psychiatric and other illnesses. The implementation of a combined presentation of effect size and statistical significance in a graphical method, added to the ordering of the variables based on the effect-ordered data display principle was deemed useful by the authors to facilitate in the process of recognizing meaningful patterns in these associations.

Materials and Methods The requirements, analyses and graphical presentation of the feature-expression heat map are described. The graphs display associations of two sets of ordered variables where a one-way direction is assumed. The associations are depicted as circles representing a combination of effect size (color) and statistical significance (radius).

Results An example dataset is presented and relation to other methods, limitations, areas of application and possible future enhancements are discussed.

Conclusion The feature-expression heat map is a useful graphical instrument to explore associations in complex biological systems where one-way direction is assumed, such as genotype-phenotype pathophysiological models.

52 Introduction

Tukey emphasized that exploratory data analysis relies more on graphical display, whereas confirmatory data analysis is easier to computerize1,2. Existing graphical methods to explore associations in a set of multiple variables are cluster heat maps and correlation plots. Heat maps originated from two-dimensional displays of a two-by-two data matrix. Larger values were represented by darker squares and smaller values by lighter squares3. E.g., in gene expression studies, these values correspond to the amount of a particular RNA or protein expressed. The further development of the cluster heat map, which includes ordering of the columns and 3 rows to reveal structure, has been a multi-step process. Facilitating the process of detecting meaningful patterns in the visual presentation, Sneath4 displayed the Feature-expression heat maps results of a cluster analysis by permuting the rows and the columns of a matrix to place similar values adjacent to each other according to the clustering, which is based on the effect-ordered data display principle5. This principle says that in any data table or graph, unordered variables should be ordered according to what we aim to show. The ideas of similarity and grouping are derived from Gestalt psychology, but have shown to be equally useful in biology6. Ling ultimately formed the idea for joining cluster trees to the rows and columns of the heat map7. Technical advances in printing let the presentation of the graphs develop from overstruck printer characters to the use of computer programs to produce cluster heat maps with high-resolution color graphics8, as can be seen in figure 1. Correlation plots are used to visualize association matrices. These plots can be regarded as heat map style displays of multiple correlation statistics. These statistics may be drawn in several forms: as numbers, circles, ellipses, squares, bars or “pac-man” symbols. In each symbol both the sign and magnitude of the correlation coefficient is represented. This is done so by using two colors printed with varying intensity. The color indicates the sign of the coefficient and the intensity of the color increases proportionally with the magnitude of the correlation coefficient5. We entertained the idea that these visual methods could be of help in the exploration of associations between genetic data and phenotypical presentation in the investigation of the pathophysiology of psychiatric disorders. The pathophysiology of these disorders is still largely unknown. In an effort to unravel the genetic basis of mood disorders, many genome-wide association studies have been performed. However, these studies found evidence for only a few susceptibility genes, which in turn accounted for a very minor part of disease liability. This fuelled the idea that to grasp the mechanism of these complex illness, it is important to have a framework integrating biology and clinical phenotype9. In this model the intermediary processes that occur between the genetic information and the specific phenotypical expression of these illnesses are regarded as a black box10.

53 FIGURE 1

Cluster heat map27,28. The columns of the heat map represent genes and the rows represent samples. Each cell is colorized based on the level of expression of that gene in that sample.

To achieve a better understanding of these intermediary underlying pathophysio logical processes, we wanted to investigate patterns in the associations between specific symptoms and specific gene expression11. We hypothesized that patterns in these associations would come to light most effectively at the intersection of related genes and related symptoms, embroidering on the above mentioned principle of effect-ordered data display. Because we were exploring the physiology of these intermediary black box processes, we preferred to use an effect size measure instead of the correlation coefficient. Contrary to the correlation coefficient effect size measures describe the magnitude of an association in measurement units. This is generally of more interest in the biomedical sciences than just the degree of linearity of an association, which is measured by the correlation coefficient. This is of special importance in explorative biology based research, which can be compared to a field biologist visiting a new

54 habitat who will begin describing the most striking features, i.e. analogous to the largest effects sizes. In addition to a measure of the magnitude of the associations of interest we wanted to implement inferential statistics to aid in drawing conclusions incorporating their certainty. Statistical significance for the given sample size was used in this regard. Summarizing, the scope of the method we had in mind was to visualize a large set of associations of variables in two sets in which one-way association was assumed, i.e. from gene expression to phenotype. This approach required a contiguously and ordered arrangement of the variables, incorporating the direction of the associations, an effect size measure and the statistical significance of individual associations. 3 Elaborating on this reasoning we developed feature-expression heat maps. In this article we will first describe the method of creating a feature-expression heat Feature-expression heat maps map. Secondly we will present an example. Finally, we will comment on this method and theorize on other areas of application.

Materials and Methods

Preprocessing The dataset for a feature-expression heat map analysis needs to consist of two sets of variables that differ in their nature. The variables in these sets are assumed to represent phenomena that occur in a certain time order according to an underlying theoretical model and have a one-way relationship. The variables of each set are then transformed in a way that facilitates comparability within the set, such as z-transformation. Because of reasons of in-between comparability, similarity in data type (binary, ordinal, continuous) of the variables within each set is a stringent requirement. In order to be able to recognize meaningful patterns in the final feature-expression heat map the variables need to be arranged contiguously and ordered a priori in a way that similar variables are placed adjacent to each other in both variable sets, consistent with the effect ordered data display principle. This may be achieved by performing a cluster analysis on a correlation matrix of the variables in each variable set. Alternatively, the variables can be arranged on phenomenological similarity.

Analysis Regression methods may be used to determine the effect size and statistical significance of the associations of the individual variables of the preceding with those of the subsequent variable sets. In case of linear regression the regression coefficient β and its p-value are used as measures of effect size and statistical significance, respectively. In case of (ordered) logistic regression the β is used which indicates the change in

55 the logit (or log-odds of the outcome) and may be preferred over the odds ratio, the latter being asymmetric and ranging from zero to infinity. In order to control for the increased risk of wrongful rejection of the null hypothesis (type I error, false positive results) correction for the false discovery rate (FDR) can be applied, as described by Benjamini and Hochberg12. Deriving from this method, approximations of the power and sample size can be calculated13,14. The separate effect size, statistical significance and FDR results of each association are put into individual data matrices. They are ordered in such a way that for each association the statistical property is placed on the intersection between the preceding variables (columns) and the subsequent variables (rows), where the variables of the preceding and subsequent variable sets are ordered according to the above-described procedure.

Graphical presentation To display the associations of two sets of variables adapted heat maps are drawn, visualizing the preceding variable set in the columns and the subsequent variable set in the rows, each ordered facilitating the visual identification of meaningful clusters of association later on. An underlying cluster analysis tree may be added to one or both of the axes. In the feature-expression heat maps the associations between preceding and subsequent variables are represented by circles (figure 2). The effect size measure is represented by the type and intensity of the color, whereas the statistical significance of the analyses is represented by the radius of the circles. Shades of red are used for positive effect sizes, whereas shades of blue are used for negative effect sizes.

FIGURE 2 Overview of a feature-expression heat map

Preceding Variable set A (ordered by cluster analysis or phenomenologically)

Variable 1 Variable 2 Variable 3 ... Variable n color intensity = effect size Variable 1 circle radius = statistical significance Variable 2 blue = negative effect

Variable 3 red = positive effect

Variable 4 centered dot = association surviving false discovery rate correction

... or phenomenologically) converging associations Subsequent Variable set B Subsequent Variable (ordered by cluster analysis cluster analysis by (ordered

Variable n

56 TABLE 1 R code for corrplot Table 1 R code for corrplot

Import tables > plot_size <- as.matrix(read.table("plot_size.txt", sep = "\t", header = TRUE)) > plot_significance <- as.matrix(read.table("plot_significance.txt", sep = "\t", as Stepmatrices Codeheader = TRUE)) > fdr_template <- as.matrix(read.table("fdr_template.txt", sep = "\t", header = TRUE)) > fdr_significance <- as.matrix(read.table("fdr_significance.txt", sep = "\t", header = TRUE)) Define colors > col <- colorRampPalette(c("blue", "white", "red")) Create > corrplot(plot_significance, method =c("circle"), col=("black"), tl.cex = 0.6, tl.col = ("black"), cl.pos="n") significance plot > corrplot(plot_size, is.corr = FALSE, method =c("color"), addgrid.col = "grey", 3 Create size col = col(200), tl.cex = 0.6, tl.col = "black", cl.pos="n") measure plot > corrplot(fdr_template, method =c("circle"), col=("black"), tl.cex = 0.6, tl.col Create FDR Feature-expression heat maps = ("black"), p.mat = fdr_significance, insig = "blank", sig.level = 0.009, plot cl.pos="n") Create legends > corrplot(legend_significance, method =c("circle"), col=("grey"), tl.cex = 0.6, tl.col = ("black"), cl.pos="n") > corrplot(plot_size, is.corr = FALSE, method =c("color"), addgrid.col = "grey", col = col(200), tl.cex = 0.6, tl.col = "black", cl.pos="r", cl.lim = c(-3,3), cl.ratio=0.4, cl.length=7) R code for corrplot version 0.7315 R code for corrplot version 0.7315

Centered dots may be added to the compartments that comply with a FDR below a certain threshold, thus allowing for a selected portion to be expected false positive. Optionally, when needed in the process of drawing statistical decisions it can be preferred to only visualize the circles of associations of which the statistical significance is below a pre-defined threshold. To create these heat maps, separate plots are drawn for the effect size, statistical significance and FDR data matrices. These plots can be drawn with the corrplot package, by T. Wei15, on R (R Development Core Team 2013, Vienna, Austria)16. R pro- gramming code examples are given in table 1. In order to magnify the more significant associations, i.e. those in which p is approaching 0, applying a transformation 1-3√p to the statistical significance parameter was found to be useful empirically. Finally, the separate plots can be merged with the transparency function of a vector graphics editor.

57 Results and discussion

Example We will use some results of our study on monocyte gene expression and psychiatric symptoms of patients with bipolar disorder presented in chapter 2 to demonstrate the use of the feature-expression heat map11. In this dataset we analyzed the relation between gene expression and manic symptoms containing information from 64 patients. According to the underlying pathophysiological model manic symptoms are associated with inflammation related monocyte gene expression, albeit indirectly. The manner of how these processes interact is as yet unknown. Resulting from the pathophysiological model gene expression variables were placed in the preceding variable set, whereas manic symptoms were put in the subsequent variable symptom set. The gene variables were ordered based on a hierarchical cluster analysis of a Pearson correlation matrix, previously executed and published17. They were subdivided into three subclusters and a rest group based on the molecular function. Symptom variables were ordered into a phenomenological sequence ranging from core mood symptoms via thought symptoms, psychosomatic symptoms, motor symptoms, food intake symptoms, sleep symptoms, to higher functional symptoms. The gene expression variables contained continuous, normally distributed data and Z-transformation was applied. The symptom variables contained ordered categorical data, which were all transformed on a 0-1 scale for standardization. In this example analysis associations between individual gene expression and individual manic symptoms were analyzed using ordered logistic regression. The ordered logistic regression model is a direct generalization of the commonly used two-outcome logistic model. In ordered logistic regression, an ordinal dependent variable is estimated as a linear function of independent variables and a set of cut-points18,19. For each association used to create the feature-expression heat map in this example the effect size was defined as the magnitude of the regression coefficient β indicating the change in the log-odds of the outcome variable per unit increase in the independent variable. Subsequently, the z-statistic was calculated for each association as the ratio of the coefficient β to its standard error. The statistical significance then was defined as the statistical probability (p) that this statistic, assumed to follow a normal distribution, is as extreme as, or more so, than what would have been observed under the null hypothesis, defined by p>|z|. As stated previously, the transformation 1-3√p was applied to the statistical significance and maximization of the effect size to 3 was applied. Effect size and significance plots were exported as vector graphs, displayed in figure 3a and figure 3b.

58 . Manic . 11 11 Individual effect size plot statistical(a), significance plot and (b) FDR plot (c) constitutethat the feature- expression heat map (d) depicting the associations between manic symptoms monocyte and inflammatory expression gene symptoms were ordinally measured. Gene expression and distributed normally was z-transformed. Statistical analysis was performed using ordered logistic regression. Circles with a center dot represent significance below the 0.2 false discovery rate (FDR) threshold for multiple testing. The legends, black have annotations and lines been added manually after the creation of the heat map. Figure 3d has been reprinted permissionwith 3 2 1 0 -1 -2 -3 FDR<0.2 0.001 0.005 0.01 0.05 0.1 0.2 3 p size effect

Feature-expression heat maps chemotaxis

motility / motility

inflammation transcription / adhesion

Insight

Appearance

Sexual Interest Sexual

Disruptive or Aggressive Behavior Aggressive or Disruptive

Decreased Sleep Decreased

Increased Speech Increased

Increased Motor Activity or Energy or Activity Motor Increased Language / Thought Disorder Thought / Language

manic symptoms Content

Irritability Elevated Mood Elevated d F3 IL6 PTX TNF SDR IL1B CD9 MXD STX1 ATF3 SERP FABP CCL7 PTGS CCL2 PTPN EMP1 EREG EGR3 THBS CCR2 MAFF NAB2 MAPK HSPA1 CCL20 CXCL2 CXCL3 PDE4B DUSP2 RGC32 CDC42 BCL2A1 TNFAIP3 FDR plot c Significance plot b Effect size plot Effect a FIGURE 3

59 We set the FDR at 0.2, thus allowing 1/5 to be false positive. This resulted in a statistical significance threshold (q-value) of 0.009. A FDR plot was drawn with corrplot marking the corresponding circles (figure 3c). Finally, the effect size plot, statistical significance plot and FDR plot were merged with Adobe Illustrator (Adobe Systems Inc., San José, California; figure 3d). This mania feature-expression heat map allowed for the identification of a converging group of associations between the genes PTPN – CDC42, so called sub-cluster 2 genes, and manic symptoms. Sixteen of these associations were significant after FDR correction, especially in the associations with the symptoms increased speech and increased motor activity11.

Relation to other methods Marked properties of the feature-expression heat maps are the combined display of an effect size measure and the statistical significance and use of effect-ordered data display on two sets of variables. This combination aids in the recognition of association patterns in complex systems, e.g. pathophysiological models. These feature-expression heat maps are based on the original cluster heat maps and use some features that can be found in correlation plots. Both the original cluster heat maps and feature-expression heat maps facilitate the visual analysis of extensive data sets for patterns. Where original cluster heat maps allow displaying all kinds of data matrices, the feature-expression heat map limits its applicability to one-way associations between two variable sets. While limiting the area of usability, it facilitates the use of regression methods. These are almost essential to analyze the strength of the phenomena involved and are an asset in explorative research focused on deconstructing pathophysiological models. Of the various ways of displaying correlation plots the circle correlation plot, where the radius as well as the type and intensity of the color were derived from the correlation coefficient, drew our attention. In doing so the correlation plot utilizes two ways of visual display to show one test outcome. The visual combination of two measures of association in the feature-expression heat maps, i.e. effect size and statistical significance in one graphical display, increases the usefulness of the method. While visualizing the effect size allows observation of the strongest associations, adding the significance of the association adds the ability to observe the signal-to-noise ratio of the observations, thus aiding in the process of inference of the explorative process. In visually integrating and presenting this distinctive information this method allows for a balanced interpretation of the associations. Especially, the method allows for salient complex (patterns of) associations to become apparent. Bipartite network graphs can be regarded as an alternative to the feature-expression heat map method in visualizing two dimension (bipartite) variables20,21. For example, a bipartite network could represent genes and symptoms as nodes, and the edge

60 weight connecting the nodes could represent the significance of the respective association. The color of the edges could represent the effect size, and its style (e.g. solid versus dotted line) could represent whether or not the significance met the FDR correction. The possibility of adding more than two variable dimensions in one graph (multipartite graph) is a benefit of this method. However, in medium size datasets consisting of two variable dimensions the authors consider the surveyability of the feature-expression heat map to be favorable due to its convenient arrangement.

Limitations The scalability of the feature-expression heat map is principally limited by the 3 perception and interpretation capabilities of the interpreter. To facilitate the inter- pretability of more complex feature-expression heat maps separate panels can be Feature-expression heat maps created containing subgroups of the variables. For example, the presentation of the separate symptom dimensions (manic, depressive, psychotic symptoms) has been performed using separate panels in our recent study11. Like all methods exploring data sets with large variable lists, feature-expression heat maps may bring about an increased risk of type I errors. Although an extensive discussion about controlling for this problem is beyond the scope of this article, we have endeavored to restrict the extend of this limitation by deploying the Benjamini-Hochberg method. It is known that the Benjamini-Hochberg method offers a more powerful alternative to the traditional Bonferroni method22,23. Although a wrongful rejection of the null hypotheses cannot be fully eliminated with this method, considering only clusters of adjoining associations to be meaningful can further diminish this risk, which is a strength of the heat map method.

Future perspectives Originating in a study on the relation between monocyte gene expression and psychiatric symptoms we expect the feature-expression heat map method to be useful for studying many other illnesses by benefiting from a combined effect size and statistical significance plot. High throughput screening studies24,25 involving complex relations between genetics and biological features are obvious candidates. Even more so, the method is not limited to genotype – phenotype relations, but can easily be applied in any explorative analysis exploring multiple variables, within a group of subjects, of which a two-variable-set one-way dependency is assumed. As this method can be regarded as an evolution of the original cluster heat maps, it is tempting to reflect on possible future enhancements. At present the method relies on separate ordering of the row and column variables, based on individual cluster analyses or phenomenological similarity. Biclustering is a cluster method that allows simultaneous clustering of both rows and columns26. Biclustering of the data matrix to obtain the ordering of the variables would increase the extent of the interaction

61 effects in the feature-expression heat map. Furthermore, instead of relying on visual identification of meaningful clusters of associations, future development incorporating more advanced, automated pattern recognition may aid in the discrimination between more meaningful and less meaningful clusters. By automation assigned cluster associations could be visually marked by a distinguishable background color of the compartments involved or with a bold line surrounding these compartments.

Conclusion

The feature-expression heat map is a useful graphical instrument to explore associations in complex biological systems where one-way direction is assumed, such as genotype-phenotype pathophysiological models. It utilizes the combined display of an effect size measure and the statistical significance as well as the use of effect-ordered data display of two sets of variables, both aiding in the recognition of meaningful association patterns.

62 References

1. Tukey JW. Exploratory Data Analysis. Addison-Wesley; 1977. 2. Tukey JW. We Need Both Exploratory and Confirmatory. Am Stat. 1980 Feb;34(1):23–5. 3. Loua T. Atlas statistique de la population de Paris. J. Dejey & cie; 1873. 4. Sneath PH. The application of computers to taxonomy. J Gen Microbiol. 1957 Aug;17(1):201–26. 5. Friendly M. Corrgrams: Exploratory displays for correlation matrices. Am Stat. 2002;56(4):316–24. 6. Friendly M, Kwan E. Effect ordering for data displays. Comput Stat Data Anal. 2003;43:509–39. 7. Ling RL. A computer generated aid for cluster analysis. Commun ACM. ACM; 1973 Jun 1;16(6):355–61. 8. Wilkinson L, Friendly M. The history of the cluster heat map. Am Stat. American Statistical Association; 2009;63(2):179–84. 9. Schulze TG. Genetic research into bipolar disorder: the need for a research framework that 3 integrates sophisticated molecular biology and clinically informed phenotype characterization. Psychiatr Clin North Am. 2010 Mar;33(1):67–82. Feature-expression heat maps 10. Cauer W. Theorie der linearen Wechselstromschaltungen, Vol.I. Leipzig: Akad. Verlags- Gesellschaft Becker und Erler; 1941. 11. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50. 12. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med. 1990 Jul;9(7):811–8. 13. Ferreira JA, Zwinderman AH. Approximate power and sample size calculations with the Benjamini-Hochberg method. Int J Biostat. 2006;2(1). 14. Efron B. Size, power and false discovery rates. Ann Stat. 2007 Aug;35(4):1351–77. 15. Wei T. corrplot: Visualization of a correlation matrix [Internet]. R package version 0.73. 2013. Available from: http://cran.r-project.org/package=corrplot 16. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2014. 17. Drexhage RC, van der Heul-Nieuwenhuijsen L, Padmos RC, van Beveren N, Cohen D, Versnel M a, et al. Inflammatory gene expression in monocytes of patients with schizophrenia: overlap and difference with bipolar disorder. A study in naturalistically treated patients. Int J Neuropsychopharmacol. 2010 Nov;13(10):1369–81. 18. Brant R. Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics. 1990 Dec;46(4):1171–8. 19. StataCorp. Ordered logistic regression. In: Stata 13 Base Reference Manual. College Station, TX: Stata Press; 2013. 20. Dulmage AL, Mendelsohn NS. Coverings of bipartite graphs. Can J Math. 1958 Jan 1;10:517–34. 21. Asratian AS, Denley TMJ, Häggkvist R. Bipartite Graphs and Their Applications. Cambridge University Press; 1998. 22. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B …. 1995; 23. Noble WS. How does multiple testing correction work? Nat Biotechnol. Nature Publishing Group; 2009 Dec;27(12):1135–7. 24. Giuliano KA, Haskins JR, Taylor DL. Advances in high content screening for drug discovery. Assay Drug Dev Technol. 2003 Aug;1(4):565–77. 25. Abraham VC, Taylor DL, Haskins JR. High content screening applied to large-scale cell biology. Trends Biotechnol. 2004 Jan;22(1):15–22.

63 26. Eren K, Deveci M, Küçüktunç O, Çatalyürek Ü V. A comparative analysis of biclustering algorithms for gene expression data. Brief Bioinform. 2013 May;14(3):279–92. 27. Andrade M. Heatmap [Internet]. Wikipedia. 2006. Available from: http://en.wikipedia.org/wiki/File:Heatmap.png 28. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A. 1998 Dec 8;95(25):14863–8.

64 3 Feature-expression heat maps

CHAPTER 4 Inflammatory monocyte gene expression Trait or state marker in bipolar disorder?

Karlijn Becking / Bartholomeus C.M. Haarman, Rixt F. Riemersma-van der Lek, Laura Grosse, Willem A. Nolen, Stephan Claes, Hemmo A. Drexhage, Robert A. Schoevers

International Journal of Bipolar Disorders. 2015 Dec 17;3(1):20. Abstract

Background This study aimed to examine whether inflammatory gene expression was a trait or a state marker in patients with bipolar disorder (BD).

Methods 69 healthy controls (HC), 82 euthymic BD patients and 8 BD patients with a mood episode (7 depressed, 1 manic) were included from the MOODINFLAME study. Six of the 8 patients who had a mood episode were also investigated when they were euthymic (6 of the 82 euthymic patients). Of these participants the expression of 35 inflammatory genes were determined in monocytes using quantitative- polymerase chain reaction of which a total gene expression score was calculated as well as a gene expression score per sub-cluster.

Results There were no significant differences in inflammatory monocyte gene expression between healthy controls, euthymic patients. Patients experiencing a mood episode however had a significantly higher total gene expression score (10.63±2.58) compared to healthy controls (p=.004) and euthymic patients (p=.009), as well as when compared to their own scores when they were euthymic (p=.02). This applied in particular for the sub-cluster 1 gene expression score, but not for the sub-cluster 2 gene expression score.

Conclusions Our study indicates that in BD inflammatory monocyte gene expression is especially elevated while in a mood episode compared to being euthymic.

68 Background

Disturbances in the immune system have frequently been reported in bipolar disorder (BD) 1. Several meta-analyses found peripheral cytokines to be raised in patients compared to healthy controls (HC)2,3. However, results are heterogeneous, with also studies reporting on normal4 or even lower cytokine levels5 in BD compared to HC. This may be due to the fact that peripheral cytokines are strongly influenced by lifestyle and disease factors6. Focusing on the main cellular producers of these cytokines, such as circulating monocytes and macrophages, may be a better approach to find stable markers for BD. Indeed, studies from our group focusing on gene expression of circulating monocytes, found a discriminating pro-inflammatory gene expression in BD patients compared to HC7,8. It remains unclear whether these immunological disturbances are related to the mood state, or are a trait phenomenon. Most studies compared BD patients to HC, 4 without differentiating between patients in different mood states. The few available studies that examined immune disturbances across mood states found significantly Inflammatory gene monocyte expression higher levels of peripheral inflammatory markers during a mood episode compared to euthymia9–13. Regarding inflammatory gene expression, our original hypothesis prior to the study described below, was that monocyte activity might be a diagnostic biomarker for BD and thus a trait factor. However, in further analysis of our previous study we already found the expression of specific inflammatory genes to be higher in a small subsample of depressed versus euthymic patients and to a lesser extent in manic compared to euthymic patients7. Furthermore, we reported a possible relation between a sub-cluster of genes and manic symptomatology14 in BD. In this report we present the results of the MOODINFLAME study, in which we compared euthymic BD patients with HC. Moreover, we present the results in of a small additional study in which BD patients were compared both in a mood episode and when euthymic. Thus, our study aimed to examine whether inflammatory gene expression in monocytes is a trait or a state marker in BD.

69 Methods

Participants Data were derived from the EU funded MOODINFLAME study15 carried out to investigate possible inflammatory biomarkers in order to advance early diagnosis, treatment and prevention of mood disorders. In the MOODINFLAME study adult male and female subjects were included who were free of inflammation related symptoms including fever and current or recent infectious or inflammatory disease, uncontrolled systemic disease, uncontrolled metabolic disease or other significant uncontrolled somatic disorders known to affect mood. They did not use somatic medication, specifically any medication known to affect mood or the immune system, such as corticosteroids, non-steroid anti-inflammatory drugs and statins. Female candidates who were pregnant or recently gave birth were excluded. The present study has been set up as a cross-sectional case-control study extended with a within-patient longitudinal design. Blood was analyzed of a sample of 159 adult participants recruited from two university psychiatry clinics in Groningen (the Netherlands) and Leuven (Belgium). The sample consisted of 69 HC, 82 euthymic BD patients (BD-Eu) and 8 BD patients with a mood episode (BD-Ep) (seven depressed, one manic). Six of the 8 BD-Ep patients were also investigated when they were euthymic (6 of the 82 BD-Eu patients). Of these patients, 1 euthymic patient was resampled after he had an episode, whereas the other five were first measured when they were euthymic and later when they were in an episode. The study was approved by the ethical committees of the participating universities, and written informed consent was obtained from all participants.

Assessments DSM-IV BD diagnoses were established using the Mini-International Neuropsychiat- ric Interview (MINI)16,17. The severity of depression was measured by the Inventory of 18 Depressive Symptoms (IDS-C30) for BD patients in a face-to-face interview, for HC with a self-report questionnaire (IDS-SR30). To determine the presence or intensity of manic symptomatology in patients the Young Mania Rating Scale (YMRS)19 was used. Mood states were defined as euthymic, manic or depressed, based on the MINI. BD-Eu patients were neither in a depressed nor (hypo-)manic episode at the time of measurement as indicated by an IDS-C30 score <22 and an YMRS score <12, respectively. Remaining clinical characteristics were obtained with the Patient Questionnaire from the former Stanley Foundation Bipolar Network, including separate clinician and patient chapters covering a spectrum of clinical features20. In the event of a mismatch of results from the MINI in relation to the Patient Questionnaire, diagnoses were checked with the treating physician. Age of onset was defined as the age when the first mood episode occurred, and information on psychiatric medication was dichotomized.

70 Laboratory methods To detect the expression of inflammatory genes of monocytes, similar methods were used as described in the original study by Padmos et al.7. In short, RNA was isolated from purified monocytes and to obtain c-DNA for quantitative-polymerase chain reaction (q-PCR), 1 μg of RNA was reverse-transcribed using the cDNA high-capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA, USA). Then, relative to the housekeeping gene ABL1, the expression of ADM, ATF3, BCL2A1, BTG3, CCL2, CCL20, CCL7, CD9, CDC42, CXCL2, DHRS3, DUSP2, EMP1, EREG, FABP5, HSPA1A/HSPA1B, IL-1α, IL-1β, IL1R1, IL-6, IRAK2, MAFF, MAPK6, MXD1, NAB2, PDE4B, PTGS2, PTPN7, PTX3, RGCC32, SERPINB2, STX1A, THBD, TNF and TNFAIP3 was determined, using the comparative threshold cycle (CT) method21. See Table 1, for the list of genes and corresponding proteins. Data were expressed as ΔCT values (values corrected to ABL1) and to control for site (Groningen and Leuven), fold change transformation was applied. By dividing the ΔCT-scores of patients from Groningen 4 by the mean of healthy controls from Groningen and subsequently the scores of

patients from Leuven by the mean of healthy controls from Leuven, the relative gene Inflammatory gene monocyte expression expression was expressed as a fold change (FC) value21.

Gene score calculation In order to obtain a simple measure for overall monocyte activation, we calculated a gene score from the expression levels as described by Grosse et al.22. For each of the 35 genes we determined a range in HC fold change gene expression, defined by the HC mean ± 1 standard deviation (SD). Then, we used this range as a standard to compare the gene expression across the different groups. A gene was considered up-regulated if the FC-value was higher than the HC mean + 1 SD, and down- regulated if the FC-value was lower than the HC mean – 1 SD. Then, we calculated a total gene expression score, by adding all up-regulated (+1), all down-regulated (-1) and all normally expressed (0) genes for each patient. This method proved to be valid, since the total gene scores showed highly significant correlations with the majority of the genes22. Additionally, we calculated two separate sub-cluster gene scores, based on previous cluster analyses performed by Drexhage et al.8. The first sub-cluster consisted primarily of pro-inflammatory genes (see Table 1) and the second sub-cluster consisted of chemotaxis, adhesion, differentiation and motility genes (see Table 1).

Statistical analyses All data were analyzed with SPSS version 20.0 (SPSS, Chicago, IL, USA). Sample characteristics were compared using Pearson’s chi-square and Fisher’s exact tests for dichotomous and categorical variables, and for continuous variables ANOVA and t-tests were used. To compare inflammatory gene expression scores across HC,

71 TABLE 1 List of genes with corresponding proteins

Gene symbol Name of corresponding protein ATF3 Cyclic AMP-dependent transcription factor 3 BCL2A1 B-cell lymphoma-2-related protein A1 CCL20 C-C chemokine ligand 20

CXCL2 C-X-C chemokine ligand 2 DUSP2 Dual specificity protein phosphatase 2 EREG Epiregulin IL-1β Interleukin 1β IL-6 Interleukin 6 PDE4B cAMP-specific 3',5'-cyclic phosphodiesterase 4B Inflammation PTGS2 Prostaglandin G/H synthase (cyclooxygenase) PTX3 Pentraxin-related protein 3 TNF Tumor necrosis factor TNFAIP3 Tumor necrosis factor, alpha-induced protein 3 CCL2 C-C chemokine ligand 2 CCL7 C-C chemokine ligand 7

CDC42 Cell division control protein 42 homolog DHRS3 Short-chain dehydrogenase/reductase 3 EMP1 Epithelial membrane protein 1 MAPK6 Mitogen-activated protein kinase 6 motility adhesion / / adhesion

Chemotaxis/ NAB2 Nerve growth factor-induced protein A binding protein 2 differentiation/ differentiation/ PTPN7 Protein tyrosine phosphatase, non-receptor type 7 STX1A Syntaxin-1A ADM Adrenomedullin BTG3 BTG family, member 3 CD9 Cluster of differentiation 9 antigen FABP5 Fatty acid-binding protein 5 HSPA1/HSPA1B Heat shock 70 kDa protein 1 IL-1α Interleukin 1α IL1R1 Interleukin 1 receptor, type 1 Other IRAK2 Interleukin 1 receptor associated kinase-like 2 MAFF Musculoaponeurotic fibrosarcoma oncogene homolog F MXD1 MAD protein RGCC32 Regulator of cell cylcle SERPINB2 Plasminogen activator inhibitor-2 THBD Thrombomodulin

euthymic patients, mood episode patients, and within-patient analyses, ANOVA was used. Results were reported as mean ± standard error. Because an overall inflammatory gene expression score was used, correction for multiple testing was not applied. As a set of sensitivity analyses, we repeated all analyses using ANCOVA controlling for sex, age and body mass index (BMI).

72 Results

Table 2 shows the sample characteristics of HC, BD-Eu patients and BD-Ep patients. Figure 1 shows he total inflammatory gene expression scores in these groups. We found no significant differences between HC (2.58±0.88) and BP-Eu (3.48±0.84), or BD-Ep patients when they were euthymic (1.17±0.94) (all p>.44). However, BD-Ep patients had a significantly higher total gene expression score (10.63±2.58) compared to HC (p=.004) and BD-Eu patients (p=.009) and compared to their own scores when they were euthymic (p=.020). For the means of sub-cluster 1 score, again no significant differences were found between HC (1.13±0.41), BD-Eu patients (1.59±0.39) and BD-Ep patients when they were euthymic (0.50±1.38) (all p>.40). BP-Ep patients (5.13±1.20) again had a significantly higher sub-cluster 1 score compared to healthy controls, BD-Eu patients and compared to their own scores when they were euthymic (p=.002, p=.006 and p=.01, respectively). The mean sub-cluster 2 scores of HC 4 (0.74±0.28), BD-Eu patients (0.59±0.26), BD-Ep patients (2.0±0.81), BD-Ep patients when they were euthymic (0.17±0.94) did not differ significantly between any of the Inflammatory gene monocyte expression groups (all p>.10). When repeating the analyses adjusted for sex, age and BMI, this resulted in essentially the same results.

TABLE 2 Characteristics of patients and healthy controls (N=159)

Healthy Controls Euthymic Mood episode (n=69) (n=82) (n=8) pa Female, n (%) 39 (56.5) 41 (50.0) 5 (62.5) .63 Age, mean (sd) 44.7 (16.1) 43.1 (12.1) 41.8 (12.7) .72 BMI, mean (sd) 23.9 (3.2) 25.7 (4.2) 26.89 (4.2) .005 IDS score, mean (sd) 4.8 (3.4) 8.7 (8.0) 42.1 (14.1) <.001 Clinical characteristics YMRS score, mean (sd) - 1.3 (1.2) 4.7 (6.4) .001 Bipolar I disorder, n (%) - 53 (64.4) 5 (62.5) .90 Bipolar II disorder, n (%) - 29 (35.4) 3 (37.5) Age of onset, mean (sd) - 23.3 (9.6) 21.8 (9.1) .68 Lifetime psychotic features, n (%) 26 (31.7) 2 (25.0) .70 Psychotropic medication - Melatonin - 1 (1.2) 2 (25.0) .02 SSRI - 4 (4.9) 1 (12.5) .38 Antipsychotics - 14 (17.1) 1 (12.5) .74 Lithium - 62 (75.6) 2 (25.0) .003 Benzodiazepines - 11 (13.4) 2 (25.0) .37 Anti-epileptics - 22 (26.8) 2 (25.0) .91

Abbreviations: sd, standard deviation; BMI, body mass index; IDS, inventory of depressive symptoms; YMRS, Young mania rating scale; SSRI, selective serotonin reuptake inhibitor. a Based on χ2-tests and Fisher’s exact tests for dichotomous and categorical variables, ANOVA tests when comparing age and BMI, and t-tests when comparing continuous variables between euthymic and mood episode BD patients.

73 FIGURE 1 Total monocyte gene expression score of healthy controls, euthymic and mood episode BD patients (N=159)

Black lines represent mean and standard error of the mean per group. Lines connecting values from the BD-Eu and BD-Ep group represent the euthymic patients who were measured again when they had a mood episode. Abbreviations: HC, healthy controls; BD, bipolar disorder.

Discussion To our knowledge, the present study is the first to show an elevated inflammatory monocyte gene expression in BD patients when experiencing a mood episode, compared to both HC and euthymic BD patients. Furthermore, BD-Ep patients had an increased inflammatory gene expression than when they were euthymic. This indicates that inflammatory gene expression in BD is related to the mood state, rather than being a trait marker. Our findings are supported by several other studies examining peripheral cytokines, where the highest levels of cytokines are found in BD patients with a mood episode, although findings in these studies were not equivocal9–13. Serum levels of cytokines are known to follow a different pattern than monocyte gene expression23. Belonging to the same developmental lineage as brain microglia, monocyte activation may be more directly related to psychopathology than circulating cytokines24,25. Previous studies from our group in different samples examining inflammatory monocyte gene expression in relation to BD found specifically the sub-cluster 2 genes to be related to a mood episode7,8 or to severity of manic symptoms14. Although in our study the scores were also higher in BD-Ep patients, we did not find a significant difference

74 in sub-cluster 2 gene score compared to HC or BD-Eu patients. This can probably be explained by the fact that we included only one manic patient, whereas in our previous studies more manic patients were included and by the fact that we used a total gene score calculation, whereas the previous studies examined the separate genes. Since sub-cluster 2 genes are associated with adhesion, cell differentiation and cell shape changes and sub-cluster 1 consists of the classic pro-inflammatory genes, it seems that in our sample having a mood episode is specifically associated with activation of the inflammatory response system. Although our finding that an increased inflammatory gene expression is more likely to be a state than a trait phenomenon, the causality and timely sequence of these associations is still difficult to interpret. Based on the present data we cannot be sure whether an increase in inflammatory gene expression preceded the mood episode, or vice versa. In an earlier study we showed that increased immune activation represented by peripheral markers preceded the onset of manic symptoms in MDD patients26. 4 The only way to examine a clear causal role for immune activation in the development of a mood episode is to measure euthymic patients multiple times prior, during and Inflammatory gene monocyte expression after a mood episode. Our study has several limitations. First and most important, although the total sample consisted of 159 persons, we had only few BP-Ep patients: seven patients with a depression and only one manic patient. And since we examined only one manic patient, it is difficult to draw conclusions about inflammatory gene expression in a manic state. However, because we found already significant differences in this small group, we would encourage future studies to include more patients with a mood episode and also to assess patients repeatedly during both an episode and when euthymic. Second, all our patients were treated naturalistically, which resulted in a variety of medications that are also known to influence inflammatory gene expression, including lithium, anti-epileptics, antipsychotics and several antidepres- sants7,14,27,28. Since these effects are typically suppressive in nature, medication may have obscured a real difference in inflammatory gene expression between BD-Eu patients and HC. However, the BP-Ep patients demonstrating significantly increased gene expression compared to both HC and BD-Eu patients used approximately the same medications, suggesting a pathophysiological cause. Third, our selection of genes was based on the study of Padmos et al.7 which found these specific signature genes, possibly ruling out other important genes. Finally, our study only focused on inflammatory gene expression of monocytes, which make up around 2-8% of the total white blood cell population and is still a peripheral measurement. It would also be of interest to examine other parts of the peripheral immune system (e.g. leuko- cyte subsets), or more proximal factors such as microglial activation in the brain or cytokine concentrations in cerebrospinal fluid.

75 Conclusions

In conclusion, our study showed that in BD patients the presence of a mood episode was associated with elevated inflammatory monocyte gene expression. This may imply that immune activation found in BD may rather be a state marker than a trait marker and can be detected in monocytes. Studies in peripheral cytokines corrob- orate our findings, however our results in gene expression need to be replicated in larger samples before a firm conclusion can be drawn.

Acknowledgements

We thank Harm de Wit and Annemarie Wijkhuijs for their excellent technical assistance and Juliëtte Kalkman for accompanying the patients.

76 References

1. Leboyer M, Soreca I, Scott J, Frye M, Henry C, Tamouza R, et al. Can bipolar disorder be viewed as a multi-system inflammatory disease? J Affect Disord. Elsevier B.V.; 2012 Dec;141(1):1–10. 2. Modabbernia A, Taslimi S, Brietzke E, Ashrafi M. Cytokine alterations in bipolar disorder: a meta-analysis of 30 studies. Biol Psychiatry. Elsevier; 2013 Jul 1;74(1):15–25. 3. Munkholm K, Braüner JV, Kessing LV, Vinberg M. Cytokines in bipolar disorder vs. healthy control subjects: a systematic review and meta-analysis. J Psychiatr Res. Elsevier Ltd; 2013 Sep;47(9):1119–33. 4. Guloksuz S, Cetin EA, Cetin T, Deniz G, Oral ET, Nutt DJ. Cytokine levels in euthymic bipolar patients. J Affect Disord. Elsevier B.V.; 2010 Nov;126(3):458–62. 5. Boufidou F, Nikolaou C, Alevizos B, Liappas I a, Christodoulou GN. Cytokine production in bipolar affective disorder patients under lithium treatment. J Affect Disord. 2004 Oct;82(2):309–13. 6. O’Connor M-F, Bower JE, Cho HJ, Creswell JD, Dimitrov S, Hamby ME, et al. To assess, to control, to exclude: effects of biobehavioral factors on circulating inflammatory markers. Brain Behav Immun. Elsevier Inc.; 2009 Oct;23(7):887–97. 7. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger 4 RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 8. Drexhage RC, van der Heul-Nieuwenhuijsen L, Padmos RC, van Beveren N, Cohen D, Versnel Inflammatory gene monocyte expression M a, et al. Inflammatory gene expression in monocytes of patients with schizophrenia: overlap and difference with bipolar disorder. A study in naturalistically treated patients. Int J Neuropsychopharmacol. 2010 Nov;13(10):1369–81. 9. Brietzke E, Stertz L, Fernandes BS, Kauer-Sant’Anna M, Mascarenhas M, Escosteguy Vargas A, et al. Comparison of cytokine levels in depressed, manic and euthymic patients with bipolar disorder. J Affect Disord. Elsevier B.V.; 2009 Aug;116(3):214–7. 10. Cunha AB, Andreazza AC, Gomes F a, Frey BN, da Silveira LE, Gonçalves C a, et al. Investigation of serum high-sensitive C-reactive protein levels across all mood states in bipolar disorder. Eur Arch Psychiatry Clin Neurosci. 2008 Aug;258(5):300–4. 11. Barbosa IG, Morato IB, de Miranda AS, Bauer ME, Soares JC, Teixeira AL. A preliminary report of increased plasma levels of IL-33 in bipolar disorder: further evidence of pro-inflammatory status. J Affect Disord. Elsevier; 2014 Mar;157:41–4. 12. Tsai S-Y, Chung K-H, Wu J-Y, Kuo C-J, Lee H-C, Huang S-H. Inflammatory markers and their relationships with leptin and insulin from acute mania to full remission in bipolar disorder. J Affect Disord. 2012;136(1-2):110–6. 13. Ortiz-Domínguez A, Hernandez M, Berlanga C, Gutierez-Mora D, Moreno J, Heinze G, et al. Immune variations in bipolar disorder: phasic differences. Bipolar Disord. 2007;(11):596–602. 14. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50. 15. MOODINFLAME website. http://moodinflame.eu. 2014. 16. Sheenan D V, Lecrubier Y, Sheehan H, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): The Development and Validation of a Structured Diagnostic Psychiatric Interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59(Suppl 20):22–33. 17. Lecrubier Y, Sheehan D, Weiller E, Amorim P, Bonora I, Harnett Sheehan K, et al. The Mini International Neuropsychiatric Interview (MINI). A short diagnostic structured interview: reliability and validity according to the CIDI. Eur Psychiatry. Éditions scientifiques et médicales Elsevier, Paris; 1997;12(5):224–31.

77 18. Rush AJ, Gullion CM, Basco MR, Jarrett RB, Trivedi MH. The Inventory of Depressive Symptomatology (IDS): psychometric properties. Psychol Med. 1996;26:477–86. 19. Young RC, Biggs JT, Ziegler VE, Meyer D a. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978 Nov 1;133(5):429–35. 20. Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck PE, Denicoff KD, et al. The Stanley Foundation Bipolar Treatment Outcome Network. I. Longitudinal methodology. J Affect Disord. 2001 Dec;67(1-3):33–44. 21. Biosystems A. User Bulletin #2, Applied Biosystems PRISM 7700 Sequence Detection System: Relative Quantitation of Gene Expression. In: Applied Biosystems. 2001. 22. Grosse L, Carvalho L a, Wijkhuijs AJM, Bellingrath S, Ruland T, Ambrée O, et al. Clinical characteristics of inflammation-associated depression: Monocyte gene expression is age-related in major depressive disorder. Brain Behav Immun. Elsevier Inc.; 2014 Aug 20;2. 23. Mesman E, Hillegers MH, Ambree O, Arolt V, Nolen W a., Drexhage H a. Monocyte activation, brain-derived neurotrophic factor (BDNF), and S100B in bipolar offspring: A follow-up study from adolescence into adulthood. Bipolar Disord. 2014;39–49. 24. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 25. Haarman BCM (Benno), Riemersma-Van der Lek RF, de Groot JC, Ruhé HG (Eric), Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder – A [11C]-(R)-PK11195 positron emission tomography study. Brain Behav Immun. Elsevier Inc.; 2014 Aug 3;40:219–25. 26. Becking K, Boschloo L, Vogelzangs N, Haarman BCM, Riemersma-van der Lek R, Penninx BWJH, et al. The association between immune activation and manic symptoms in patients with a depressive disorder. Transl Psychiatry. Nature Publishing Group; 2013 Jan;3(10):e314. 27. Rybakowski JK. Antiviral and immunomodulatory effect of lithium. Pharmacopsychiatry. 2000 Sep;33(5):159–64. 28. Tourjman V. In vivo immunomodulatory effects of antipsychotics on inflammatory mediators: A review. Adv Biosci Biotechnol. 2012;03(04):551–65.

78 4 Inflammatory gene monocyte expression

CHAPTER 5 Does CRP predict outcome in bipolar disorder in regular outpatient care?

Sonya M. Balukova, Bartholomeus C.M. Haarman, Rixt.F. Riemersma – van der Lek, Robert A. Schoevers

International Journal of Bipolar Disorders. 2016 Dec;4(1):14. Abstract

Background The association between inflammation and the course of mood disorders is receiving increased attention. This study aims to investigate whether a sub-group of patients with BD can be identified for which a higher CRP (C-reactive protein) level at baseline is associated with an unfavorable prognosis.

Methods Historic cohort study using CRP at baseline, with 15 months follow-up of mood status and medication. Cross-sectional analyses include boxplots, one-way ANOVA, Receiver Operating Characteristics (ROC) curve and chi-square test, and the longitudinal analysis using multivariate Cox-regression.

Results 84 bipolar disorder patients were included in the analyses. Cross-sectionally, no statistically significant difference was found in CRP distribution across mood states (p=0.372) or rapid cycling state (p=0.656). Also, no CRP cut-off level was distinguished between euthymic and non-euthymic patients according to the ROC curve (p=0.449, AUC=0.452, 95%CI: 0.327, 0.576), and a literature-derived cut-off value (3 mg/L) again demonstrated no difference (p=0.530). Longitudinally, no association was found between CRP and prognosis of disease neither in euthymic (-2 Log Likelihood= 120.460; CRP: p=0.866, B= -0.011, OR= 0.989 (95% CI: 0.874 – 1.120) or non-euthymic patients (-2 Log Likelihood= 275.028; CRP: p=0.802, B= 0.010, OR=1.010 (95% CI: 0.937 – 1.088)). Medication use did not affect these associations.

Conclusions We found no statistically significant association between CRP and a more unfavorable BD prognosis, suggesting that the application of CRP as a practical biomarker to predict outcome in a naturalistic outpatient care setting is not as straightforward as it may seem.

82 Background

Bipolar disorder (BD) is associated with a significant decrease in quality of life and social functioning of patients1. Despite the availability of pharmacological treatment, its efficacy is far from optimal2,3. A promising approach for optimizing the treatment is to tailor it to the specific characteristics of a patient as part of a ‘personalized medicine’ approach which encompasses not only psychological but also biological markers4. There is increasing evidence to suggest that immunological processes may contribute to the emergence as well as the prognosis and severity of BD5. Apart from pro-inflammatory cytokines such as IL-2 and IL-66, C-reactive protein (CRP) is an acute phase protein that is produced in response to infection and inflammation. It is also considered to be another candidate biomarker for detecting immune dysregulation in BD. A number of studies suggest an association between CRP and mood disorders, especially during a manic episode7–9. Based on the idea that immunological processes play a role in the pathophysiology of BD, it can be hypothesized that an increased activity of these processes, measured with CRP, would lead to more instability in BD symptomatology and course of disease. 5 Increased inflammatory activity has been shown to be related to therapy resistance Does CRP predict outcome in bipolar disorder in regular outpatient care? outpatient regular in disorder bipolar in outcome predict CRP Does and chronicity in unipolar depression10–12. Recently, Becking et al.13 demonstrated an increased CRP to predict future development of manic symptoms in a sample of MDD (major depressive disorder) patients, also suggesting that this is a subtype with an untoward prognosis. However, to date, no studies have examined this issue in bipolar patients. The current study investigated whether, in a clinical setting, higher CRP levels at baseline may predict a worse BD outcome, defined as a shorter time to relapse (if euthymic) or a longer time to recover. As some medication may influence inflammatory processes, this was taken into account.

83 Methods

Participants and ethical considerations For the present historic cohort study we used medical files from 84 BD patients from the BD outpatient department of the psychiatry department of the University Medical Center Groningen (UMCG), the Netherlands. Patients who provided a written informed consent to participate, were included if the following criteria were fulfilled: DSM-IV-TR diagnosis of BD, age between 18 and 65, recorded CRP value, not pregnant or less than 6 months postpartum, no current serious somatic illnesses (current infections or liver disease, serious un- or undertreated heart, lung or neurological disorders). All patient data were collected as part of regular outpatient care and were anony- mously used for research according to the Data Protection Act (WBP) and Medical Treatment Agreement (WGBO), as formulated in the Code of Conduct for the Use of Data in Health Research, also known as the Research Code of Conduct (see also IRB in Supporting Information.

Assessments of study parameters The study parameters included CRP, measurement date and value, BD type, presence of a rapid cycling course, and mood episodes and medication one month before and 15 months following the CRP measurement. Assessment of the psychiatric condition of patients was determined by the first author based on information from two sources: the electronic medical records of the treating psychiatrists and the LifeChart Methodology (LCM) records– (a systematic collection of data on the course of illness and treatment presented in a graphical form)14. The condition was noted as one of five categories: 0. Euthymia; 1. Depressive episode; 2. Hypomania or mania; 3. Mixed episode; 4. Unstable mood. Longer-lasting clinically significant mood instability that did not fulfill the criteria for any mood episode was assessed as an “unstable” mood, while subthreshold symptoms were assessed as no change in episode15. Serum CRP (high sensitive CRP) is routinely measured at hospital admission in this Psychiatry department (and additional measurements are performed when there is a psychiatric or somatic event for a patient). CRP was assayed using a wide range turbi- dimetric CRP assay (CRPL3 assay) on a Roche Modular platform (Roche, Mannheim, Germany). Starting from the CRP measurement date at baseline (T0), the time to episode change (T1) and the corresponding psychiatric condition were gathered from the medical file. Cases with unclear information and/or diagnosis of current mental status were discussed with the psychiatrists of the BD outpatient department. Remaining decisions were taken in discussion with all authors of this study.

84 Statistical analysis Both cross-sectional and longitudinal analyses were performed. Cross-sectionally, data were visualized using boxplots and tested using histograms, P-P plots, Kolmogorov-Smirnov test, Kruskal-Wallis, a receiver operating characteristics (ROC) curve and Fisher’s exact test. The longitudinal analysis comprised a multivariate Cox-regression based on the time passed until a change of episode has occurred where the main covariate was the scale of CRP values. This analysis was done separately for euthymic patients and those who were in a mood episode at baseline (depressed, manic, mixed, unstable). This was done because the variable signifying elapsed time has a different meaning for patients in an euthymic state than for a non-euthymic one, and should thus be interpreted differently to describe the progression of disease. Longer time before episode change signifies a better BD prognosis if a subject is euthymic (means longer time in remission), but it means a worse course if a subject is non-euthymic (means longer time in an episode). Medication was added to the Cox-regression to examine whether it affected CRP or changed the effect of CRP on BD progress. Separate analyses were done after excluding those non-psychopharmaceutical drugs that are known to have 5 anti-inflammatory and/or CRP-affecting properties16–18. Does CRP predict outcome in bipolar disorder in regular outpatient care? outpatient regular in disorder bipolar in outcome predict CRP Does As part of sensitivity analyses, all tests were repeated in each of the following subgroups for each timepoint and then for the whole study: excluding subjects taking anti-inflammatory medication; excluding all medication that affects CRP; excluding outliers, defined as CRP values above 10 mg/L.

85 Results

The sample of this study consisted of 84 patients and Table 1 shows their characteristics. Testing the data for normality showed a non-normal distribution of the CRP data, with a positive skewness and a significant difference from a normal distribution seen by Kolmogorov-Smirnov test (D(84)=0.258, p<.001). Because of this, median values are provided for each mood state in table 1.

TABLE 1 Characteristics of patient population

Euthymic Depressed (Hypo)manic Mixed Unstable All Mean age (SD) 44.9 (12.4) 42.7 (12.4) 43.2 (10.3) 46 (7.1) 40.5 (10.2) 43.6 (11.7) Number of Subjects (%) 37 (44.0) 27 (32.1) 12 (14.3) 2 (2.4) 6 (7.1) 84 (100) Male Gender (%) 11 (37.9) 10 (34.5) 4 (13.8) 2 (6.9) 2 (6.9) 29 (34.5) Female Gender (%) 26 (47.3) 17 (30.9) 8 (14.5) 0 4 (7.3) 55 (65.5) BD type I (%)* 26 (41.9) 21 (33.9) 11 (17.7) 2 (3.2) 2 (3.2) 62 (100) BD type II (%)* 7 (41.2) 5 (29.4) 1 (5.9) 0 4 (23.5) 17 (100) Rapid Cycling (%) 4 (30.8) 3 (23.1) 2 (15.4) 0 4 (30.8) 13 (100) Metabolic Syndrome (%)** 9 (13.8) 6 (9.2) 2 (3.1) 1 (1.5) 2 (3.1) 20 (30.8)*** Median C-reactive protein (mg/L) 1.10 2.10 0.90 3.08 1.39 1.37

Abbreviations: SD =Standard Deviation; BD=bipolar disorder * Bipolar disorder type information missing for 5 patients ** Metabolic syndrome information missing for 19 patients (Criteria: fasting glucose >5.6 mmol/L and two or more of following: BMI >30 kg/m2, Hypertriglyceridemia >1.7 mmol/L, HDL-C <0.9 mmol/L in men and <1.0 mmol/L in women, Hypertension > 140/90 mmHg) *** The number in brackets in this cell represents the percentage of all subjects who have metabolic syndrome, while the rest of the percentages in this row represent the distribution of subjects with metabolic syndrome across the mood states

Cross-sectional analysis The differences between the distribution of CRP values among the mood states were first examined using Kruskal-Wallis. No statistical significance was found (p=0.372) (see table 1). Figure 1 shows the distribution of mood episodes at baseline. Using Kolmogorov-Smirnov no statistical significant difference was found between the patients that were rapid cycling and patients that were not (p=0.656, see figure 2). Repeating the cross-sectional analyses excluding all medications that are known to affect CRP, as well as excluding CRP outliers, generally yielded the same, non-significant results.

86 FIGURE 1 Boxplot of CRP distribution at baseline across mood state

The Y-axis depicts CRP concentration in mg/L starting from 0 mg/L and incrementing care? outpatient regular in disorder bipolar in outcome predict CRP Does with 5 mg/L. On the X-axis bars signify the interquartile range of CRP values in each mood group, and the thickened line inside them shows the median value of CRP within this mood group. The stars and circle show the boxplot outliers, while the numbers next to them signify the specific number of the patient with this outlier value. The circle is a patient with CRP value closest to the threshold value but above it, so it is also not included.

After excluding 6 CRP outliers, there were 37 (44%) who were in euthymia and 47 (56%) were non-euthymic: 27 (32.1%) were in a depressive episode, 12 (14.3%) in a manic episode, 2 (2.4%) in a mixed episode and 6 (7.1%) were unstable. Using a ROC curve, the data was tested to identify a cut-off value of CRP which could suggest whether subjects at baseline would be euthymic or in an episode. Based on the curve (Figure S1 in Supporting Information), no such cut-off value was found (AUC=0.452,p=0.449, 95%CI: 0.327, 0.576). Consequently, the cross-sectional analyses were performed using a literature-based CRP cut-off value of 3.0 mg/L and a chi-square test for the not-normal distribution of CRP value. The results showed that, cross-sectionally, higher CRPs were almost equally distributed among euthymic and non-euthymic patients (42.1% euthymic and 57.9% non-euthymic by CRP>3mg/L; 44.6% euthymic and 55.4% non-euthymic by CRP≤3mg/L). Fisher’s exact test confirmed that these results have no statistical significance (p=0.530).

87 FIGURE 2 Boxplot of CRP distribution between patients with and without rapid cyclingline

The Y-axis depicts CRP concentration in mg/L starting from 0 mg/L and incrementing with 5 mg/L. On the X-axis bars signify the interquartile range of CRP values in each patient group, and the thickened line inside them shows the median value of CRP within this mood group. The stars and circle show the boxplot outliers, while the numbers next to them signify the specific number of the patient with this outlier value.

TABLE 2 Longitudinal results from Cox regression

CRP after adjusting for: Patient Group -2 Log Sig. B Odds 95% CI of Odds Ratio Likelihood coef. Ratio Lower Upper Euthymic Patients at 120.460 .866 -.011 .989 .874 1.120 Baseline Non-Euthymic Patients at 275.028 .802 .010 1.010 .937 1.088 Baseline

Longitudinal analysis For euthymic subjects at baseline From the 37 subjects who were euthymic at Baseline, 20 had a change of episode during the trial period, while 17 patients stayed euthymic. The results are shown in table 2 and the hazard function of the covariate CRP is on figure 3. There is no correlation between CRP value and the event of episode change (relapsing). The odds

88 FIGURE 3 Hazard function at mean of the covariate CRP for subjects euthymic at baseline

5 This figure illustrates what the hazard ratio is for recovering of a subject with a given CRP

value compared to a subject with a CRP value of one unit lower in the course of the studied care? outpatient regular in disorder bipolar in outcome predict CRP Does period. The Y-axis represents the rate of recovering of all subjects sick at baseline. The elapsed time period in days until an euthym state has occurred is depicted on the X-axis.

ratio (=0.989) approaches equality for both groups (results are very close to the neutral line). Moreover, these findings are not statistically significant and so the null hypothesis could not be rejected.

For non-euthymic subjects at baseline There were 47 subjects who were in an episode at baseline of which 30 became euthymic, while 17 patients remained in a mood episode during the study period. The results are shown in the table 2 and in figure 4. The odds ratio is approaching equality (results are very close to the neutral line) and these findings are not statistically significant. As seen in the euthymic group, these results also do not show an association between CRP and relapsing. After adjusting for covariates and testing for interactions of CRP with medication at the two timepoints, there were no results with statistical significance in each of the two mood-state groups – euthymic and non-euthymic. Adjusting for separate medication types used in any moment during the study again yielded no associations with statistical significance.

89 FIGURE 4 Hazard function at mean of the covariate CRP for subjects sick at baseline

This figure illustrates what the hazard ratio is for recovering of a subject with a given CRP value compared to a subject with a CRP value of one unit lower in the course of the studied period. The Y-axis represents the rate of recovering of all subjects sick at baseline. The elapsed time period in days until an euthym state has occurred is depicted on the X-axis.

TABLE 3 Number of subjects per medication type

Parameter T0a T1b Totalc Lithium (%) 31 (36.9) 47 (56) 48 (57.1) Valproic acid (%) 14 (16.7) 9 (10.7) 16 (19.0) Anti-Inflammatory drugs (%)* 12 (14.3) 12 (14.3) 16 (19.0) Tricyclic Antidepressants (TCA) (%)* 5 (6.0) 5 (6.0) 7 (8.3) Antidepressants (non-TCA) (%) 31 (36.9) 27 (32.1) 32 (38.1) Antipsychotics (%) 35 (41.7) 34 (40.5) 39 (46.4) Benzodiazepines (%) 31 (36.9) 31 (36.9) 39 (46.4)

a Medication used up to and including the baseline (T0); b Medication used up to and including the time of episode change is such occurred (T1); c Medication used for the total period of the study (1 month before and up to 15 months after CRP measurement, unless episode change occurred earlier); * Values are the same for both T0 and T1 timepoints;

90 Table 3 shows a summary of the used medication at the T0 and T1 timepoints as well as in total for the whole study period.

Discussion

Principal findings To our knowledge, this is the first study examining longitudinal associations between CRP level and clinical outcome in BD patients in a naturalistic treated and real-life measurement outpatient setting. In a first cross-sectional analysis, we could not distinguish a sub-group of BD patients with an elevated baseline CRP level based on affective state or rapid cycling state. In the longitudinal analysis, no statistically significant association was found between higher CRP values and relapsing in either euthymic or non-euthymic patients, as well as when comparing them.

Comparison to previous studies The results of the cross-sectional analysis can be compared to several previous studies on the association of CRP with BD. Two of the studies with a total of 202 5 patients reported that for their BD subjects higher CRP was significantly associated Does CRP predict outcome in bipolar disorder in regular outpatient care? outpatient regular in disorder bipolar in outcome predict CRP Does with manic state compared to the other mood episodes8,9. Three other studies with a total of 248 patients however did not demonstrate significant differences in CRP across the mood states, which is corresponding to the current study results19–21. The findings of the influence of medication on the reported associations between CRP values and mood states in the above studies are also conflicting. Hornig et al 19 found a significant negative interaction between lithium and CRP within BD patients, while Dickerson et al.8 found that no medication affected the CRP association with BD, as we found in the current study. In a recent meta-analysis Dargél et al. demonstrated an overall cross-sectional association between BD and CRP22. In this meta-analysis, CRP levels were elevated in manic and euthymic patients compared to HC, but not in depressed BD patients compared to HC. Recent prospective longitudinal studies demonstrated CRP alterations across mood states23, as well as before and after various treatments24. In addition an increased CRP was found to be associated with an increased risk for developing late-onset BD in a large Danish demographic sample25. Adding to this body of knowledge, the present study shows that using CRP as a practical biomarker to predict outcome in a naturalistic outpatient care setting is not as easily applicable as it may seem.

91 Limitations The present study has several limitations. A limitation pertaining the use of histori- cal data is that CRP measurements were more or less routinely measured, and more in those patients who were psychiatrically or somatically ill. Since it was not purely measured as a routine, a selection bias cannot be ruled out. Furthermore, body fat percentage, smoking, blood pressure, physical activity and SNP polymorphisms are suggested to be associated with different baseline levels of CRP26–30 and in the present study we were not able to correct for these variations. Finally, the study would have benefitted from a larger sample size.

Proposed mechanisms There are two factors to consider which could play a role in shaping the results from this study: use of medication and the pathophysiological model of BD. Due to the naturalistic design of this study, a great variety of medications was used, some of which have been suggested to have anti-inflammatory properties. Lithium, part of the standard treatment of dipolar disorder in the Netherlands, was used by more than half of the patients in this study (57.1%). Although its mechanism of action in bipolar depression is not fully understood yet16, several studies have found evidence that lithium causes a decrease in the inflammatory marker CRP11,19 and a similar effect is also described for SSRIs (selective serotonin reuptake inhibitors) used by 38% of the subjects here18. However, there are reports that lithium and antidepressants may also exhibit pro-inflammatory properties16,31,32. These properties were postulated after observing the stimulating effect of lithium on pro-inflammatory cytokines such as Tumor Necrosis Factor-α (TNF-α), Interleukin-4 (IL-4) and Interleukin-6 (IL-6). It is not yet clear what causes these reverse effects. There could be various mechanisms via which medication is influencing CRP. However, at the same time no such effect was demonstrated in this study. Therefore other factors should be considered. Our findings therefore suggest that the inflammatory model of BD is probably much more complex than what can be shown by straightforward inflammatory alterations demonstrated by elevated markers such as CRP33 at one point in time. Compared to the inflammatory mechanism in somatic disorders where CRP, being an acute phase protein, directly correlates to acute worsening of a disease, it may be that immunological processes in BD are affected by more factors and additional mechanisms, which need to be added to the model to gain a better understanding. One such hypothesis is that inflammatory changes in psychiatric disorders are mediated through a shift towards the tryptophan/kynurenine pathway, which leads to the formation of further factors that may affect the functioning of the brain, such as neurotoxic accumulation of the metabolite 3-Hydroxykynurenine34. Another hypothesis is the co-occurrence of autoimmunity alongside immune dysregulation in BD. It is thus likely that BD is related to a state of immune dysregulation, rather than

92 more pure immune activation35,36. It has been suggested that the immune mediating pathogenesis of BD occurs at a much younger age, and what follows is a dysregulation of the bodily systems. The long-lasting results are then what we see at an older age: unhealthy and unexpected interactions and responses of the whole body, and of inflammatory markers as well37. Another interesting possibility is that what we see is not necessarily an inflamed but a damaged organism, given that CRP reacts not only to inflammatory stimuli, but also to damaged cells38. To increase our understanding of the position of immune-system-mediated pathophysiological processes in BD it is necessary to measure CRP and other relevant immune bio-assays prospectively in a larger sample size. In that regard it is important to investigate what happens on an intra-individual level: additional trials are necessary measuring CRP on different timepoints in a prospective longitudinal manner so that individual changes could be followed and analyzed. Perhaps, these studies will elucidate models, adjusting for variables known to influence CRP, that eventually enable CRP to be used as a practical biomarker to predict outcome in naturalistic treatment settings. 5

Acknowledgements care? outpatient regular in disorder bipolar in outcome predict CRP Does We thank dr. J.E. Kootstra-Ros for her valuable assistance in retrieving archived data.

93 References

1. ten Have M, Vollebergh W, Bijl R, Nolen W a. Bipolar disorder in the general population in The Netherlands (prevalence, consequences and care utilisation): results from The Netherlands Mental Health Survey and Incidence Study (NEMESIS). J Affect Disord [Internet]. 2002 Apr;68(2–3):203–13. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12063148 2. Frecska E, Balla P, Falussy L, Ferencz A, Varga Z. The message of the survival curves: I. Composite analysis of long-term treatment studies in bipolar disorder [Internet]. Vol. 14, Neuropsychopharmacologia Hungarica . 2012. p. 155–64. Available from: http://www.mppt.hu/images/magazin/pdf/xiv-evfolyam-3-szam/frecska.pdf 3. Burcusa SL, Iacono WG. Risk for recurrence in depression. Vol. 27, Clinical Psychology Review. 2007. p. 959–85. 4. Hamdani N, Doukhan R, Kurtlucan O, Tamouza R, Leboyer M. Immunity, inflammation, and bipolar disorder: diagnostic and therapeutic implications. Curr Psychiatry Rep [Internet]. 2013;15(9):387. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23955004 5. Liu HC, Yang YY, Chou YM, Chen KP, Shen WW, Leu SJ. Immunologic variables in acute mania of bipolar disorder. J Neuroimmunol. 2004;150(1–2):116–22. 6. Brietzke E, Stertz L, Fernandes BS, Kauer-Sant’Anna M, Mascarenhas M, Escosteguy Vargas A, et al. Comparison of cytokine levels in depressed, manic and euthymic patients with bipolar disorder. J Affect Disord [Internet]. 2009 Aug [cited 2010 Aug 28];116(3):214–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/19251324 7. Maes M, Smith R, Scharpe S. The monocyte-T-lymphocyte hypothesis of major depression. Psychoneuroendocrinology [Internet]. 1995 Jan [cited 2011 Mar 6];20(2):111–6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/7899532 8. Dickerson F, Stallings C, Origoni A, Boronow J, Yolken R. Elevated serum levels of C-reactive protein are associated with mania symptoms in outpatients with bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry [Internet]. 2007 May 9 [cited 2011 Jun 1];31(4):952–5. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17391822 9. Cunha AB, Andreazza AC, Gomes F a, Frey BN, da Silveira LE, Gonçalves C a, et al. Investigation of serum high-sensitive C-reactive protein levels across all mood states in bipolar disorder. Eur Arch Psychiatry Clin Neurosci [Internet]. 2008 Aug [cited 2011 Jun 1];258(5):300–4. Available from: http://www.ncbi.nlm.nih.gov/pubmed/18297417 10. Raison CL, Rutherford RE, Woolwine BJ, Shuo C, Schettler P, Drake DF, et al. A randomized controlled trial of the tumor necrosis factor antagonist infliximab for treatment- resistant depression: the role of baseline inflammatory biomarkers. JAMA psychiatry [Internet]. 2013;70(1):31–41. Available from: http://archpsyc.jamanetwork.com/article. aspx?articleid=1356541 11. Sluzewska A, Sobieska M, Rybakowski JK. Changes in acute-phase proteins during lithium potentiation of antidepressants in refractory depression. Neuropsychobiology. 1997;35(3):123–7. 12. Miller AH, Maletic V, Raison CL. Inflammation and Its Discontents: The Role of Cytokines in the Pathophysiology of Major Depression. Biol Psychiatry [Internet]. 2009;65(9):732–41. Available from: http://dx.doi.org/10.1016/j.biopsych.2008.11.029 13. Becking K, Boschloo L, Vogelzangs N, Haarman BCM, Riemersma-van der Lek R, Penninx BWJH, et al. The association between immune activation and manic symptoms in patients with a depressive disorder. Transl Psychiatry [Internet]. 2013 Jan [cited 2013 Dec 2];3(10):e314. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=3818012&tool=pmcentrez&rendertype=abstract

94 14. Denicoff KD, Leverich GS, Nolen W a, Rush AJ, McElroy SL, Keck PE, et al. Validation of the prospective NIMH-Life-Chart Method (NIMH-LCM-p) for longitudinal assessment of bipolar illness. Psychol Med [Internet]. 2000 Nov [cited 2013 Sep 13];30(6):1391–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11097079 15. Perlis RH, Ostacher MJ, Marangell LB, Hongwei Z, Wisniewski SR, Ketter, Terrence a., et al. Predictors of Recurrence in Bipolar Disorder: Primary Program for Bipolar Disorder ( STEP-BD ). Am J Psychiatry. 2006;163(4):217–24. 16. Goldstein BI, Kemp DE, Soczynska JK, McIntyre RS. Inflammation and the phenomenology, pathophysiology, comorbidity, and treatment of bipolar disorder: a systematic review of the literature. J Clin Psychiatry [Internet]. 2009 Aug;70(8):1078–90. Available from: http://www.ncbi.nlm.nih.gov/pubmed/19497250 17. Ximenes JCM, de Oliveira Gonçalves D, Siqueira RMP, Neves KRT, Santos Cerqueira G, Correia AO, et al. Valproic acid: an anticonvulsant drug with potent antinociceptive and anti-inflammatory properties. Naunyn Schmiedebergs Arch Pharmacol [Internet]. 2013;386(7):575–87. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23584602 18. O’Brien SM, Scott L V, Dinan TG. Antidepressant therapy and C-reactive protein levels. Br J Psychiatry [Internet]. 2006;188:449–52. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16648531 19. Hornig M, Goodman D, Kamoun M, Amsterdam J. Positive and negative acute phase proteins in affective subtypes. J Affect Disord. 1998;49(1):9–18. 20. Hope S, Dieset I, Agartz I, Steen NE, Ueland T, Melle I, et al. Affective symptoms are associated with markers of inflammation and immune activation in bipolar disorders but not in schizophrenia. J Psychiatr Res [Internet]. 2011;45(12):1608–16. 5 Available from: http://dx.doi.org/10.1016/j.jpsychires.2011.08.003

21. Tsai S-Y, Chung K-H, Wu J-Y, Kuo C-J, Lee H-C, Huang S-H. Inflammatory markers and their care? outpatient regular in disorder bipolar in outcome predict CRP Does relationships with leptin and insulin from acute mania to full remission in bipolar disorder. J Affect Disord [Internet]. 2012;136(1–2):110–6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/21962564 22. Dargél AA, Godin O, Kapczinski F, Kupfer DJ, Leboyer M. C-Reactive Protein Alterations in Bipolar Disorder. J Clin Psychiatry [Internet]. 2015 Feb 25;76(2):142–50. Available from: http://www.psychiatrist.com/jcp/article/pages/2015/v76n02/v76n0203.aspx 23. Jacoby AS, Munkholm K, Vinberg M, Pedersen BK, Kessing LV. Cytokines, brain-derived neurotrophic factor and C-reactive protein in bipolar i disorder - Results from a prospective study. J Affect Disord [Internet]. 2016;197:167–74. Available from: http://dx.doi.org/10.1016/j. jad.2016.03.040 24. Uyanik V, Tuglu C, Gorgulu Y, Kunduracilar H, Uyanik MS. Assessment of cytokine levels and hs-CRP in bipolar I disorder before and after treatment. Psychiatry Res [Internet]. 2015;228(3):386–92. Available from: http://dx.doi.org/10.1016/j.psychres.2015.05.078 25. Wium-Andersen MK, Ørsted DD, Nordestgaard BG. Elevated C-reactive protein and late-onset bipolar disorder in 78 809 individuals from the general population. Br J Psychiatry. 2016;208(2):138–45. 26. Marnell L, Mold C, Du Clos TW. C-reactive protein: Ligands, receptors and role in inflammation. Clin Immunol. 2005;117(2):104–11. 27. Palosuo T, Husman T, Koistinen J, Aho K. C-reactive protein in population samples. Acta Med Scand [Internet]. 1986;220(2):175–9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/3776692 28. Visser M, Bouter LM, McQuillan GM, Wener MH, Harris TB. Elevated C-reactive protein levels in overweight and obese adults. JAMA [Internet]. 1999 Dec 8;282(22):2131–5. Available from: http://www.ncbi.nlm.nih.gov/pubmed/10591334

95 29. Davey Smith G, Lawlor DA, Harbord R, Timpson N, Rumley A, Lowe GDO, et al. Association of C-reactive protein with blood pressure and hypertension: life course confounding and mendelian randomization tests of causality. Arterioscler Thromb Vasc Biol [Internet]. 2005 May;25(5):1051– 6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/15731495 30. Ford ES. Does exercise reduce inflammation? Physical activity and C-reactive protein among U.S. adults. Epidemiology [Internet]. 2002 Sep;13(5):561–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12192226 31. Hamer M, Batty GD, Marmot MG, Singh-Manoux A, Kivimäki M. Anti-depressant medication use and C-reactive protein: Results from two population-based studies. Brain Behav Immun [Internet]. 2011;25(1):168–73. Available from: http://dx.doi.org/10.1016/j.bbi.2010.09.013 32. Nassar A, Azab A. Effects of Lithium on Inflammation. ACS Chem Neurosci [Internet]. 2014;5:451–8. Available from: http://pubs.acs.org/doi/abs/10.1021/cn500038f 33. Altamura a. C, Buoli M, Pozzoli S. Role of immunological factors in the pathophysiology and diagnosis of bipolar disorder: Comparison with schizophrenia. Psychiatry Clin Neurosci. 2014;68(1):21–36. 34. Myint A-MMA, Kim Y-KKY. Network beyond IDO in psychiatric disorders: revisiting neurodegeneration hypothesis. … Neuro-Psychopharmacology Biol Psychiatry [Internet]. 2013 Jan 3 [cited 2014 Sep 17];48:304–13. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24184687 35. Rege S, Hodgkinson SJ. Immune dysregulation and autoimmunity in bipolar disorder: Synthesis of the evidence and its clinical application. Aust N Z J Psychiatry [Internet]. 2013;47(12):1136–51. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23908311 36. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord [Internet]. 2014 Mar 29 [cited 2014 Apr 23];16(2):137–50. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24286609 37. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol [Internet]. 2012 Aug 8 [cited 2012 Nov 1];92(September):1–17. Available from: http://www.ncbi.nlm.nih.gov/pubmed/22875882 38. Kushner I, Rzewnicki D, Samols D. What does minor elevation of C-reactive protein signify? Am J Med. 2006;119(2):17–28.

96 5 Does CRP predict outcome in bipolar disorder in regular outpatient care? outpatient regular in disorder bipolar in outcome predict CRP Does

PART 2 Neuroimmune system

CHAPTER 6 PET and SPECT in bipolar disorders

Bartholomeus C.M. Haarman, Rixt F. Riemersma-Van der Lek, Henricus G. Ruhé, Jan Cees de Groot, Willem A. Nolen and Janine Doorduin

Adapted from Bipolar Disorders. In: Dierckx RAJO, Otte A, de Vries EFJ, Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Heidelberg: Springer Berlin Heidelberg; 2014. p. 223–51. Abstract

This chapter describes the findings of the PET/SPECT research efforts in bipolar disorder (BD). First, the cerebral blood flow and cerebral metabolism findings in the prefrontal cortex, limbic system, subcortical structures and other brain regions are discussed. Metabolism and blood flow oriented studies aided to study various aspects of the metabolism based disease model in which prefrontal cortex hypoactivity is accompanied by limbic hyperactivity. Second, the neurotransmitter studies are discussed. The serotonin transporter alterations are described and the variation in study results is explained, followed by an overview of the results of the various dopamine receptor and transporter molecules studies, taking into account also the relation to psychosis. Molecular imaging demonstrated the importance of serotonin transporter alterations in parts of the limbic system in BD and underscored the role of dopaminergic and cholinergic neurotransmission. Most molecular imaging studies in BD have unique designs, extending our knowledge of the pathophysiological mechanisms, but also complicating comparisons between studies.

102 Introduction

It is generally accepted that the cause of bipolar disorder (BD) is multifactorial, with multiple genes making someone vulnerable, and with psychological and social factors bringing the genes to expression. Moreover, somatic factors are assumed to play a role. To unravel the complex interplay between genotype and phenotype researchers have tried to find intermediary processes, so called endophenotypes. These are more related to the underlying genotype than the ultimate phenotype. Endophenotypes should be consistently associated with the illness and represent persistent “trait” rather than episodic or “state” features. By definition, they also should be found in high-risk individuals such as non-affected first-degree family members at a higher rate than in the general population1. The last two decades, many molecular neuroimaging studies have been performed in BD. Alterations of function assessed by molecular neuroimaging may be regarded as important endophenotypes. Probably the best approach in neuroimaging of BD is to study patients during their depressive and manic episodes as well as during the euthymic phase with different (functional) neuroimaging techniques. However, these are very complicated patients, both technically as well as practically (e.g. one can never be sure that the same patient will develop both manic and depressive episodes within a certain time frame). Positron emission tomography (PET) and single-photon emission computed 6 tomography (SPECT) are imaging techniques that use radiolabeled, biological active

compounds (PET or SPECT tracers) to gain information on specific functions of the disorders bipolar in SPECT and PET brain, by measuring brain metabolism or blood flow, or functions of individual cells, such as transporter mechanisms or receptors. The tracers involved are administered in such small doses that pharmacological activity or chemical toxicity is practically absent and due to the usual short half-life of the radionuclides total radiation remains within generally accepted safety levels. Where PET uses positron-emitting radionuclides, that give rise to two opposite directed 511kV gamma rays after annihilation of positrons with electrons, the radionuclides in SPECT directly emit gamma rays. Because the gamma rays being specifically in the opposite direction, PET is able to achieve higher spatial resolutions (about 4mm) than SPECT (7-12mm). SPECT is more widely accessible due to the lower maintenance costs and generally easier to handle tracers.

Cerebral blood flow and cerebral metabolism Accumulating scientific evidence supports the theory of metabolic alterations in specific parts of the brain in patients with mood disorders: the prefrontal cortex, the limbic system and subcortical regions (figure 1). With molecular imaging techniques the metabolic activity in the brain (cerebral metabolic rate (CMR)) as well as the blood

103 FIGURE 1 Neuroanatomical regions important in mood disorders

anterior cingulate cortex

prefrontal cortex

subgenual cingulate

orbitofrontal cortex ventral striatum

hypothalamus pituitary raphe nuclei amygdala

hippocampus locus coeruleus

nervus vagus

Neuroanamtomic regions important in mood disorders. (Adapted from Patrick J. Lynch, medical illustrator, and C. Carl Jaffe, MD, cardiologist62, under the Creative Commons Attribution 2.5 Generic license (CC BY 2.5)).

flow in specific regions (cerebral blood flow (CBF)) can be measured. It is generally accepted that CMR and CBF are physiologically coupled and both are indeed closely correlated in healthy controls2. This appeared also to be the case in BD. Dunn et al.3 demonstrated that CMR and CBF were coupled globally and in most regions in BD, except the left pregenual anterior cingulate cortex. CMR can be investigated with an 18F-labeled fluorodeoxyglucose (FDG) PET scan. CBF is measured in PET by 15O-labeled water. The most common SPECT tracers to measure CBF are 133Xe, 123I-labeled iodoamphetamine (IMP) and 99mTc-labeled hexamethylpropylene amine oxime (HMPAO). CMR and CBF can be measured in resting state or during various tasks. Across the whole brain level it remains unclear whether there is an overall global CMR and CBF change in BD when compared to healthy controls. When investigated across mood states some studies found reduced global CMR4–6, while in other studies no alterations where found in CMR7,8. In depressed patients CMR was found to be reduced when compared to controls and manic patients in some studies4,5 but increased in another study9 (table 1). One

104 study investigating CBF found an increased perfusion in manic patients compared tot controls10, but others did not find a difference between the different mood states11,12.

Prefrontal cortex The prefrontal cortex (PFC) is the area of the frontal lobes of the cerebral cortex that is located before the motor and premotor areas. It plays an important role in executive functioning such as planning complex behavior, personality expression, decision making and moderating social behavior13. Regions of the brain are defined as Brodmann areas (BA) based on their cytoarchitectonic structure. In general, BD patients in a depressive or manic episode have a decreased prefrontal cortex CMR and CBF, compared to euthymic patients or healthy controls. Blumberg et al. found a reduced CBF in the right orbital PFC (BA 11) and medial frontal gyrus (BA 10) in manic patients when compared to euthymic patients14 (table 1). CMR activation related to a decision making task was also decreased in manic patients in this region15. Euthymic patients demonstrated orbitofrontral CBF decrease16. The healthy siblings of BD patients demonstrated a comparable CBF decrease in the orbitofrontal PFC during induced sadness17. In manic patients, a decrease in dorsolateral PFC (BA 8, 9, 46) CBF has been demon- strated18,19. Manic patients also showed a decrease of CMR during a decision making task in the ventrolateral PFC (BA 47) when compared to controls15. Furthermore, 6 euthymic older BD patients (50-65 years) had a lower CMR in this region than

controls of the same age820. disorders bipolar in SPECT and PET

Limbic system and subcortical structures The limbic system is a combination of in origin different brain structures that are involved in visceral behavioral patterns (related to survival: eating, drinking, sexual activity), emotions, and memory. Some structures, such as the hippocampus, amygdala, anterior thalamic nuclei, are phylogenetically rather old structures (hence the other name paleomammalian brain), while the septum, fornix and limbic cortex are more recently developed structures. The limbic cortex consists of the parahippocampal gyrus (BA 34-36), the cingulate gyrus (BA 23-26; 29-33) and the dentate gyrus, which are parts of the frontal, parietal and temporal cortical lobes on the medial surfaces of both hemispheres, surrounding the corpus callosum. The anterior part of the cingulate gyrus, the anterior cingulate cortex (ACC, BA 24, 25, 32, 33), plays a role in autonomic functions (regulating blood pressure, heart rate), rational cognitive functions (reward anticipa- tion, decision making, empathy), pain perception and emotion21. In BD patients with depressive or manic episodes, an increased CMR and CBF were demonstrated in various parts of the limbic system. In depressed BD patients, Drevets

105

ulate, and left head of caudate during manic manic during caudate of head left and ulate,

activation during word generation and decreased orbitofrontal orbitofrontal decreased and generation word during activation

depressives has been found after correction for grey matter volume. matter grey for correction after been found has depressives

in the middle frontal gyri bilaterally. Levothyroxin decreased relative activity in in activity relative decreased Levothyroxin bilaterally. gyri frontal middle the in

basal temporal, occipital; medial frontal; parietal regions and in the cingulate gyrus. cingulate the in and regions parietal frontal; medial occipital; temporal, basal -

s g n i

d n i f

n i a episodes. M amygdala the manic and in cortex BD left compared prefrontal patients to HC. dorsolateral Decreased left subgenual cingulate the right in activity higher significantly BD treatment, exhibited patients levothyroxin Before and cerebellar striatum, hippocampus), ventral amygdala, right right thalamus, medial temporal (right lobe left cortex, activity and hadvermis; relative lower a to episode mixed a or depression from going increased depression bipolar with CMR patients for brain whole The episode. manic or state euthymic controls. in than lower but MDD same, PFC for BD and the D were anterolateral CMRdorsal the in of results The and function between CBF neuropsychological were identified. correlations corrected Several cortex prefrontal and orbital rostral Decreased right with were mania. associated rest during activity cing anterior dorsal left in were activity an increased findings The principal and parietal right temporal, superior the left lower in perfusion significantly of interest regions Examining individual found. was group patient the in regions occipital bilateral was found. CMR groups two in the between performance in differences significant No statistically and unipolar (both ness affective with subjects in higher significantly be to wasfound metabolism cerebral Global controls. normal to compared depressed) bipolar than lower those of the of the euthymic blood flow values BD wereThe significantly cerebral mean patients regional medial the bilateral in controls both in callosum corpus the genu of the to ventral cortex prefrontal the in activity increased abnormally of An area unipolar familial and depressives bipolar familial was depressed increased in which levels, BD was plasma with stress Amygdala correlated cortisol patients. activity, BD. remitted in activity amygdala the normalize Mood stabilizers the right subgenual cingulate cortex, left thalamus, right amygdala, right hippocampus, right dorsal and ventral and hippocampus, dorsal ventral right amygdala, right right thalamus, left cortex, subgenual cingulate the right vermis. and cerebellar striatum, O

HMPAO

d HMPAO HMPAO PET o - - h O PET O PET t Tc Tc 15 15 e 2 2 FDGPET FDGPET FDGPET FDGPET FDG PET FDG H FDGand FDGPET CPT word generation, word generation, letter repetition, state resting state resting SPECT SPECT M 11 state resting 11 treatment with levothyroxine state resting state resting 99m state resting H H 99m state resting 11 11 99mTc SPECT state resting 11 PET state resting 11 state resting CPT electrical to stimulation the forearm 11 11

- n

o -

d i + + - - - + + + + + + + t e +/ a M c

2 Mi, 5 D) D

s I (15 M)

I (6 E, 5 M) E, (6 I I (6 E, 5 M) E, (6 I t

- I (9 E) (9 I - - II (1 E) c

- - e j SZ b u 10 MDD 5 HC 27 HC S BD 15 14 10 MDD 10 HC 9 BD 1 BD M, 5 BD (5 11 MDD HC 5 M) D, BD15 (10 D OCD w/o 10 OCD w/ 14 12 HC 8 3 E, D, 43 BD (12 7M) HM, 6 HC 11 BD 5 HC 11 BD D) 9 BD (9 11 MDD 21 HC D) 8 BD (8 D) BD16 (16 4 MDD 24 HC E) BD16 (16 10 HC 4 M) 8 E, D, BD21 (9 17 MDD 51 HC 9 E) D, BD15 (7 21 MDD 12 HC

9 1

6 5 6 2 4 6 2 1

8 5 9 2 0 2 9

3 0 9

9 0 8 9 )

5 1 4 6 2 2 0 7 1 r

7 1

0 6 9 i 5 9 0 6 2 9 a 5

8 1 2 0 m

8 8 0 e 9 9 w e 0 0 0 u g g y r 9 9 1 2 a 9

r r 0 0 r 2 a , 1 1 s

1 s s

r e e

2 a 2 b s

r r t t u

o e b b s b r y k e e e e a o h n a e h t t v v d o h m m t l c n M n u x x e e o u u u u - r r u a a a e l l o r u t l a S a B B B B B B B B B C D D (

TABLE 1 Overview of PET/SPECT studies on cerebral blood flow and cerebral metabolism in BD patients

106 and- sex

- he

posterior cortical -

anteroposterior gradient. anteroposterior anhedonia symptom cluster correlated with lower absolute metabolism in in metabolism absolute lower with correlated symptom cluster anhedonia - subjects. control healthy increased compared to HC to compared increased l cortex in the patients but an increase in this region in the siblings. the in region this in increase an but patients the in cortex l

6 PET and SPECT in bipolar disorders bipolar in SPECT and PET

s g n pregenual ACC in depressive patients versus healthy control subjects. Post hoc exploratory analysis analysis exploratory hoc Post subjects. control healthy versus patients depressive ACC in pregenual i

d n i related activation was increased in the manic patients compared with the control patients in the left dorsal ACC dorsal left the in patients control the compared with patients manic the in was increased activation related f

- n i bipolar depression a pattern of prefrontal hypometabolism was observed Additionally a cerebello a Additionally observed was hypometabolism prefrontal of pattern a depression bipolar

a but decreased in the right frontal polar region. region. polar frontal right the in decreased but M psychomotor MDD the both and BD, In CMR in normalized higher and with cortex, and temporal caudate/putamen, anteroventral claustrum, insula, right cingulate. anterior inferior in increases with perfusion, brain of redistribution an important with was associated withdrawal Lithium ACC. particularly areas, limbic in decreases and regions posterior not but manic and depressive both in was asymmetric lobes temporal the of part anterior the in CBFThe distribution t in asymmetry lobe temporal showed same patient the on sequentially taken Images state. euthymic in state. euthymic the in disappeared or diminished that mood states pathological were both systems observed in areas and limbic paralimbic cortices, CBF in the decreases prefrontal in Significant group. control healthy the with compared groups depression In subgroups. bipolar normalized hypermetabolism was all seen in Common and cingulate groups three anterior with induced sadness to were all the dorsal/rostral CBF in increases were groups the Distinguishing cortices. temporal inferior and orbitofrontal the in decreases and insula anterior fronta medial the in decreases cortex accumbens amygdala, orbitofrontal putamen,CMR and anteroventral left area, the bilateral was in increased right and anterior CBF in of reductions showed significant groups patient Both CBF. global in equivalent were groups three The and areas reduction of the normal cortical Task CBFDuring manic episode global was age with compared distribution regional and level flow blood cerebral normal showed a groups patient Both controls. normal matched subjects. control healthy and the BD CBF patients the emerged between in differences No significant additionally revealed increased metabolism in left parahippocampal, posterior cingulate, and right anterior insular insular anterior and right cingulate, parahippocampal, metabolism posterior increased left in revealed additionally versus patients depressive in cortices

d EMZ HMPAO o - - h O PET O PET IMP SPECT IMP t - Tc Tc 15 15 Xe SPECT Xe SPECT Xe SPECT Xe SPECT Xe I e 2 2 FDGPET FDGPET FDGPET CPT transient sadness induction probability SPECT SPECT 11 M 11 CPTauditory 99m lithium withdrawal 123 state resting 99m state resting H 11 state resting 133 state resting H decision based making task 133 133 state resting 133 resting state resting between groups and before / after medication state resting

- n i

o - -

d i

- - + + + + + + + t e +/ +/ a M c

HM, 2 M) 2 HM,

s II (22 D, 7 E) I (11 D, 4 E) II (13 D)

I (11 M) (11 I t -

- -

I (9 E) (9 I - c

e j b u 4 BD (14 E) 4 BD (14 S D)27 BD (27 31 MDD 1 BD13 (7 D) 6 BD (6 11 MDD 9 HC BD 14 29 BD 43 HC 9 BD 9 HS BD 13 18 HC 11 BD 11 HC M)6 BD (6 6 MDD 10 HC BD 12 16 HC 30 M) D, 40 BD (10 22 MDD 61 HC D) BD (7 7 10 MDD 9 HC 11 MDD

5 1 1

7 0 2

7 7

) 4 1 2

8 2

6 3 9 r

1 1 0 6 2 1 d 4 a 1 7 n 9 l 5 8

2 i 0 2 0 e 1 2 0 9 ö

e 7 9 9 i 0 0 3 y 0 8 t

9 n 0 k , 9 9 2 0 i 6 2 z

1 9 r

s 1 1

0 r

2 1 s

9 i r r 1

w o 1

2 y e s a n n e e 9

n d h l i i 9 h t g d u v 1 t h o n

u t b b s t 8 f ü u u a l o y u o e r u u u t i u 9 a t 1 S D G G I K K M R R R S T ( HS=healthy sibling; D=depressive episode, E=euthymic episode, M=manic episode, HM=hypomanic episode, Mi=mixed episode, episode, HM=hypomanic episode, Mi=mixed episode, M=manic episode, sibling; D=depressive E=euthymic episode, HS=healthy task discrimination ADT=auditory performance test, CPT=continuous

107 et al. found an increased CMR in the subgenual portion of the ACC (BA 25) when compared to controls, after correction for grey matter volume22 (table 1). This finding was repeated in both treated7 as well as in untreated depressed patients23. Dunn reported an association between this CMR increase and the presence of psychomotor and anhedonia symptoms. A similar increase in CMR was demonstrated in the pregenual and ventral area (BA 33, 24) of the ACC24. In manic patients, an increase in CBF in the subgenual portion of the ACC (BA 25), was described compared to controls22. This increase was also found in the left dorsal ACC (BA 32) when compared to euthymic patients25. In the manic patients, CMR during a decision making task was increased in the left dorsal ACC, when compared with controls 15. In untreated manic patients a SPECT study showed that increased cingulate cortex CBF is associated with poor executive functioning 26. Goodwin et al.27 examined 14 euthymic patients on lithium with SPECT before and after acute double-blind withdrawal of lithium. As often seen clinically, rapid withdrawal was associated with an increase of manic symptoms. The increase of manic symptoms correlated with a CBF decrease in the limbic areas, particularly the ACC. Euthymic patients also demonstrated ACC CBF aberrations16. The healthy siblings of BD patients demonstrated an comparable CBF increase in the ACC during induced sadness17. The amygdala, part of the limbic system, is one of the subcortical areas that is known to be involved in BD. Others are the nucleus accumbens, globus pallidus, striatum (including nucleus caudatus), all part of the basal ganglia of the brain that play a role in higher order motor control. Individually they are involved in different functions, the nucleus accumbens in the reward circuitry, nucleus caudatus in learning and memory, particularly regarding feedback processing and the globus pallidus in visceral regulation such as fever induction and emotion induced tachycardia28. Initially, studies of depressed BD patients versus controls described a reduced CMR in the amygdala19 as well as the striatum4,29,30. However, thereafter, various PET studies in depressed patients showed increased activity in the striatum, together with increased activity in limbic structures including the amygdala, hippocampus and parahippocampal regions6,7,20,24,32. Additionally, amygdala and ventral striatal CMR correlated positively with depression severity and with cortisol levels6,32. The difference between these initial and later studies is most probably explained by a higher signal quality and more careful patient selection in the later studies33. High CMR or CBF were also observed in the nucleus caudatus in manic patients25 and nucleus accumbens in depressed patients26.

Other cortical regions An asymmetric CBF was found in the anterior temporal cortex in manic and depressed patients but not when the patients were euthymic34 (table 1). In a more

108 recent study it was demonstrated that euthymic older BD patients (50-65 years) have a higher CMR in this region than controls of the same age20. Furthermore, CBF in the temporal cortex of BD patients was positively associated with executive functions but negatively with attention and memory26.

Neurotransmitter studies Departing from the monoamine theory of affective disorders35 PET/SPECT radioligand studies have focused on the serotonergic, dopaminergic and cholinergic systems.

Serotonin Serotonin (5-hydroxytryptamine) is a monoamine neurotransmitter that is formed out of the amino acid tryptophan. It is mainly found in the gastrointestinal tract, where its secreting cells regulate intestinal movement, in platelets, where it is released during aggregation and in the central nervous system. Serotonin has a regulatory effect with regard to mood, sleep, sexual activity and appetite. The neurons located in the raphe nuclei, a cluster of nuclei in the brain stem, are the main source of serotonin in the brain. The axons from the raphe nuclei neurons project to nearly every part of the central nervous system. After serotonin is released in the synaptic cleft, it can bind to one of the various receptors or it can be removed by the presynaptic neuron for reuse via the serotonin transporter. 6 As the primary site of serotonergic antidepressant activity, the serotonin transporter

(SERT) is the part of the serotonin neurotransmitter system that has received the disorders bipolar in SPECT and PET most attention in molecular imaging. Among the various ligands that are available, the PET ligands trans- 1,2,3,5,6,10- -hexahydro-6-[4-(methylthio) phenyl] pyrro- lo-[2,1-a] isoquinoline (11C(+)-McNeil 5652), 3-11C-amino-4-(2-dimethylaminometh- ylphenylsulfanyl)benzonitrile (11C-DASB) and the SPECT ligand 2-([2-([dimethylami- no]methyl)phenyl]thio)-5-123I-iodophenylamine (123I-ADAM) are used in BD research. An increase of SERT density was found in the thalamus using 11C(+)-McNeil 5652 in a combined group of euthymic or mildly depressed patients36 and a reduction in the midbrain, hippocampus, thalamus, putamen and ACC in a group of untreated depressed patients37 (table 2). With the use of 123I-ADAM SPECT a lower SERT density was found in de midbrain of euthymic BD-I patients when compared to euthymic BD-II patients and healthy controls38. Using the more stable en selective 11C-DASB ligand, an increased SERT density was found in the thalamus, dorsal cingulate cortex, medial prefrontal cortex and insula of depressed untreated BD patients, which was comparable to MDD39,40. Although the results are inconsistent, it can be concluded that serotonin transporter alterations occur in BD, especially in parts of the limbic system. Taking the regulatory function and the observed metabolic changes into account, the SERT density alter-

109

ter

- I patients when patients I psychotic BDpsychotic

HT2A receptor .

- cingulate cortex cortex cingulate BD patients whenBD patients

HTT both to BP relative of mood state or striatal striatal or moodof state

- comparison subjects.

MDD and BD groups showed significantly significantly showed MDD BD groups and Is associated with genetic variation within with variation genetic associated Is - DOPA uptake rate constants in the striatum striatum the in constants rate DOPA uptake - binding was found between non - found was binding F DOPA rate constants were significantly reduced in in reduced significantly were constants DOPA rate - - 18 2 depressives had reduced 5 reduced had depressives F - II patients and healthy controls. patients II psychotic BDpsychotic patients. nuclei when compared to controls. whento compared nuclei - 18

to MDDto patients

Conversely, ventral brainstem binding was nearly nearly was binding brainstem ventral Conversely, mal controls, whereas those for striatum were striatum not for whereas those mal controls, - r betweenr the groups.

and SZ compared and SZ with controls for and BD psychotic

induced decrease in striatal binding. induced striatal decrease in - had significantly lower DAT availability relative to controls in in controls to relative DAT availability lower significantly had

different. HTT BP in the thalamus (24%, 14%, respectively), insula (15%) and and (15%) insula respectively), 14%, (24%, thalamus the HTT BP in s s - g g n n i i

d d binding potential was not significantly different in manic patients than in in than patients manic in different significantly not was potential binding n n

i i 2 f f

n n i i a a M was the ACC in binding Receptor be decreased found to of compared MDD to and patients controls. BD in binding receptor Decreased CHRM2. the comparison subjects in the striatum. Treatmentthe the in comparison had striatum. no with subjects valproate patients. manic in potential D2 binding on the effect significant M significantly increased 5 increased significantly a to led challenge Amphetamine baseline. at binding D2 receptor However, subjects. healthy in than BD in patients response behavioral greater between difference wasthere the two the in nogroups significant ampheta mine Treatment with valproate had no significant effect on brain 5 on brain effect no significant had valproate with Treatment patients. manic in binding with patients in increased thalamus the was in significantly potential Binding in potential whereas binding subjects, compared as mood control to disorders not the diffe midbrain did had 16%BD patients amygdala, 26% to midbrain, the lower SERTin density ACC. and putamen thalamus, hippocampus, BD euthymic of midbrain de in found was density SERT lower A compared euthymic to BD BD,In the mean SERT BP was thalamus, in dorsal increased both the group healthy to Relative HC and MDD the pontine raphe of nuclei groups the vicinity in D in difference No statistical terms in differ not did subjects and healthy BD patients in differences No significant treatment with valproate, significantly were patients the for cortex frontal the for potentials The binding nor for those than lower The D posterior right the in compared controls to binding had greater BD patients binding lower modestly had BD patients region. caudate left the in putamen and the right lower binding in examined, regions and brain a significantly all in caudate compared region higher significantly had BD patients euthymic the controls, the to Compared DAT. striatal of availability BPD subjects caudate dorsal bilateral control VMAT2 of in than Binding BD thalamus in the was patients in higher SZ patients. and subjects group. control the in than higher were and BDSZ patients and between identical (DCC), medial prefrontal cortex and insula and decreased in the brainstem at the the at brainstem the in decreased and insula and cortex prefrontal medial (DCC), raphe pontine the of level striatum (12%). The bipolar (12%). striatum the patients andthe than were patients the the in lower patients in were found between the manic patients and the comparison subjects. Af were comparison and the subjects. between found manic the patients patients and controls. Post hoc tests showed higher binding for psychotic psychotic for binding showed higher tests hoc Post controls. and patients withpatients BD patients compared to non - to compared patients

1 1

- -

SPECT

TZTP PET TZTP PET

McNeil 5652 McNeil McNeil 5652 McNeil d d TRODAT TRODAT N o o - - - h h ADAM IZBM SPECT IZBM 3 PET raclopide DASB PET DASB PET SCH23390 CFT PET DTBZ PET - FP - FP DOPA PET setoperone PET setoperone t t ------Tc Tc - - - - I I e e F F F F C(+) - C(+) - C C C C C C C M 18 18 PET after baseline, treatmentvalproate M SPECT 18 11 PET 11 123 11 11 11 PETmethylspiperone 123 after baseline, amphetamine induction 18 after baseline, treatment valproate 11 11 99m SPECT 99m 11 11 valproate treatmentvalproate

t t

2 e e

g g

r r

2 2 2 2 1 2 a - a M M T D D D D T SERT SERT SERT SERT SERT DOPA uptake DAT DAT DAT VMAT 5

- - n n i i

o o

d d i i ------+ + + + t t e e a a M M c c

s s

II I (13 M) I (13 M) I (15 E) (15 I I t I (6 D; 5 E) D; (6 I

- - - - II (5 D) - c c

- e e j j b b u u S D) BD16 (16 17 MDD 23 HC D) BD16 (16 24 MDD 25 HC S M)7 BD (7 5 E) D, 6 BD (1 21 HC D) BD18 (18 41 HC BD 10 BD 14 28 HC D) BD18 (18 37 HC D) BD18 (18 18 MDD 34 HC 14 BD (3 D, 11 M) SZ 10 12 HC E) BD13 (13 13 HC BD 13 14 HC M) 1 6 E, D, BD10 (3 21 HC BD 13 14 HC 5 BD 10 MD 46 HC E) BD17 (17 17 HC 11 BD 13 HC BD 15 SZ 12 15 HC 7 MDD

41 51 39 37 60

49 36

58 61 ) 52 )

50 56 r 57 r 38 a a e e y y

, , r r

o o ra 1992ra y y h h d d ham 2005b t t t u u u u t t a a S 2006a Cannon 2011 Cannon ( S Ya 2002 Ichimiya Oquendo 2007 Chou 2010 2006b Cannon 2007 Cannon 1995 Pearlson 2000 Anand 2002b Yatham Suha 2002a Yatham 2007 Amsterdam 2010 Chang 2011 Anand 2001 Zubieta (

r r e e t t t t i i m m s s

n e n n i a a n

i r r n e t t o m n o o i t r r l a o u o u p r e e h o e N C N S D

TABLE 2 Overview of PET/SPECT studies on neurotransmitter systems in BD patients episode, HM=hypomanic episode, Mi=mixed episode, M=manic episode, sibling; D=depressive E=euthymic episode, HS=healthy task discrimination ADT=auditory performance test, CPT=continuous

110 ations may be interpreted as an exponent of a dysfunctional fronto-limbic network. It furthermore suggests that there might be (yet to be identified) modulators of gene expression or that other effects, such as serotonin transporter internalization, occur during different mood states. At the level of the post-synaptic receptors a study investigating the treatment effect 18 of valproate on the 5-HT2-receptor binding, using F-setoperone, demonstrated no difference before or after treatment in manic patients41.

Dopamine Dopamine is a catecholamine neurotransmitter that is formed out of L-DOPA, which in turn is made out of the amino acid tyrosine, while dopamine itself is the precursor of norepinephrine and epinephrine. A dopaminergic imbalance plays an important role in Parkinson’s disease and psychotic symptomatology (psychotic symptoms during mood episodes and SZ)42. Additionally it is thought to be of importance in mania because of the antimanic effect of dopamine receptor blockers (antipsychotics) and the mania producing effect of dopamine inducing substances, such as amphetamines43.

Five subtypes of dopamine receptors are known. The D1-like family consists of D1 and D5 receptors, which lead to the inhibition of intracellular adenylate cyclase upon activation, causing cAMP to rise. The D2-like family consists of D2, D3 and D4 receptors, which lead to the stimulation of intracellular adenylate cyclase upon activation, 6 causing cAMP, decrease. Overall, the D1-receptor and D2-receptor are the most abun-

dant dopamine receptor subtypes in the brain, with particularly high expression in disorders bipolar in SPECT and PET the striatum and nucleus accumbens and lower levels in the olfactory tubercle. The

D2-receptor is the prominent receptor in the substantia nigra, a region where the 44 D1-receptor is absent . After release into the synaptic cleft and having its neurotransmitting effect via the receptors, dopamine is pumped back into the cytosol of the presynaptic neuron by the dopamine transporter (DAT) from where it can be broken down by enzymes or be reused in synaptic vessels via the vesicular monoamine transporter 2 (VMAT2)45. Parts of the dopaminergic neurotransmission than can be examined with molecular imaging are the various dopamine receptors, dopamine release and the dopamine transporter. These in turn can be investigated during resting state or after an amphetamine challenge (stimulating dopamine release).

The D2-receptor is an obvious research target because of the known effectiveness of D2-receptor blocking antipsychotic medication on manic and psychotic symp- toms46. Radioligands targeting this receptor are benzamides, such as raclopride and iodobenzamide, and butyrophenons, such as methylspiperone. The binding potential of the benzamides is known to fluctuate in with changing endogenous dopamine concentrations, e.g. after amphetamine challenge. It is proposed that benzamides

111 and butyrophenones do not bind to the same configuration of theD2-receptor. Buty- rophenones may bind primarily to the monomer form, whereas benzamides may bind to both the monomer and dimer forms of the receptor47. In untreated non-psychotic manic patients compared to controls studies with the butyrophenone methylspiperone48,49 and the benzamides iodobenzamide and 50,51 raclopride did not find striatal D2-density difference (table 2). Pearlson et al.

however, did find a higher D2-recepter density in the caudate nucleus of BD patients with psychotic features during their depressive or manic episodes when compared to BD patients during episodes without psychotic features49. Within the group with psychotic features, the severity of the psychotic symptoms correlated with the receptor density, which was not the case with severity of mood symptoms. This suggests that the D2-receptor density is specifically related to psychosis but not to mood symptoms. This theory is further supported by the finding that the mood stabilizing

anti-epileptic valproate sodium did not alter the D2-receptor density in non-psychotic manic patients51. 52 Concerning the D1-recepter, Suhara et al. found the binding potential of SCH23390 to be decreased in the frontal cortex of BD patients with various mood states when compared to controls. In the striatum, results were comparable among patients and controls. Dopamine synthesis can be investigated by measuring the 18F-labeled 6-fluoro-L-DOPA, which is a precursor to dopamine, as described above. Dopamine synthesis was found to be comparable among untreated non-psychotic manic patients and controls. In view of the finding that valproate did not change D2-receptor density, it is interesting that valproate was able to reduce dopamine synthesis in effectively treated manic patients51,53. Perhaps the valproate-induced reduction of dopamine synthesis might be explained by an improved function of the PFC and fronto-limbic network resulting in an enhanced regulation of dopamine in the striatum. Endogenous dopamine release can be measured with an amphetamine challenge, in which dopamine release is stimulated by blocking sequestering via DAT and VMAT2 and inhibiting the breakdown enzyme monoamine oxidase(MOA). In BD amphetamine challenge elicited a greater behavioral response, as measured with the Brief Psychiatric Rating Scale (BPRS) and the Young Mania Rating Scale (YMRS) in BD patients compared to controls. However, a difference between D2-receptor binding potential of 123I-iodobenzamide between these groups was not found50. Because it is known that benzamide binding can fluctuate during amphetamine induced endogenous dopamine binding, it cannot be ruled out that BD patients may have a more sensitive dopamine system to challenges with stimulants and treatment with mood stabilizers33.

112 In recent years the DAT gained scientific attention because it is hypothesized that some of the efficacy of mood stabilizing medication may be due to their action on DAT54. In SPECT studies using 99mTc TRODAT-1 DAT density was increased in the right posterior putamen and in the left caudate in depressive BD-II patients55 and in the striatum of euthymic BD-I and BD-II patients 56. However, in untreated BD-I patients, a study using [O-methyl-11C]β-CFT (11C-CFT) PET, showed decreased DAT density in the bilateral dorsal caudate. These contradictive results may be explained by differences in patient groups (BD-I versus BD-II) and the difference in spatial resolution between SPECT and PET57. Using the (+)-α-11C-dihydrotetrabenazine (11C-DTBZ) ligand, a elevated VMAT2 density was found in the thalamus en ventral striatum in euthymic BD patients with a history of psychotic symptoms, which was comparable to SZ patients, but differed from controls58, This would suggest a relation with psychotic symptoms in BD, however, in the absence of research describing the VMAT2 density in BD patients without psychosis, a relation with affective symptoms cannot be ruled out. Overall, it can be assumed that altered dopamine neurotransmission plays a disease modifying role, especially in BD patients that experience psychotic symptoms in addition to affective symptomatology. However, dopamine neurotransmission as a pathophysiological mechanism in non-psychotic BD patients needs further research.

Choline 6 Acetylcholine is a neurotransmitter in both the peripheral nervous system and central

nervous system. In the central nervous system, it has a variety of effects as a neuro- disorders bipolar in SPECT and PET modulator upon plasticity (specifically in learning and memory), salience of sensory stimuli, arousal and reward. Interestingly, cholinesterase inhibitors were found to increase depressive symptoms in BD and MDD patients59 (table 2). Muscarinic type 2 receptor binding was decreased in the ACC of depressed BD patients when compared to MDD patients and controls, using 3-(3-(3-[18F]Flouropropyl) thio)-1,2, 5-thiadiazol-4-yl)-1,2,5,6-tetrahydro-1-methylpyridine (18F-FP-TZTP)60. This decrease in muscarinic type 2 receptor binding in BD patients was associated with a genetic variation in cholinergic muscarinic-2 receptor gene61. Furthermore, the depression and anxiety severity in BD patients were negatively correlated with the binding potentials, emphasizing a contribution of the cholinergic neurotransmitter system in BD pathophysiology.

113 Conclusion

Since the beginning of the earliest PET and SPECT studies in patients with BD-In the 1980s this field of research gave rise to many new insights in the pathophysiology of BD. The first, mainly metabolism and blood flow oriented studies aided to study various aspects of the metabolism based disease model in which PFC hypoactivity is accompanied by limbic hyperactivity. This model in its comprehensive form is however probably not precise enough to account for most of the specific mood and cognitive disease features and efforts are being made to draw into detail. The role of molecular imaging as the main imaging technique in metabolism studies has been taken over by fMRI, but they are still used to answer specific questions in which fMRI falls short. Molecular imaging demonstrated the importance of serotonin transporter alterations in parts of the limbic system in BD and underscored the role of dopamine and cholinergic neurotransmission. Most molecular imaging studies in BD have unique designs, extending the knowledge on the pathophysiological mechanisms, but also complicating comparisons between studies. The earlier studies with selection of heterogeneous patient groups, including both BD-I and BD-II patients and being in different mood states (manic, depressed and euthymic) led to results that were difficult to interpret. Moreover, use of medication can affect study outcomes, while studies with only medication-naïve patients, studies with washout periods and naturalistic studies all have their specific advantages but also disadvantages. Naturalistic study designs have the advantage that they are generally easier to perform and less burdensome for patients with this serious psychiatric disorder, but the effect of medication use can never be evaluated in a valid way. The obvious advantage of medication-naïve studies is the exclusion of these medication effects. The question arises however in how far the uniqueness of these patients in that they can function without medication, interferes with the investigated mechanism (i.e. the internal validity) and limits the generalizability (i.e. the external validity). In washout studies one could argue that the withdrawal interferes with the investigated mechanism. Another complicating factor is that the molecular imaging studies are limited in patient size because of careful ethical considerations due to the ionizing nature of the technique, which complicates comparisons between subgroups. Finally, some ligands are generally expected to measure the same biological property but are later found to differ in some specific aspects of the measurement complicating comparison between studies. Nevertheless, because of its unique selectivity emanating from a continuous extending range of possible ligands, molecular imaging remains an important tool in BD research.

114 References

1. Gottesman II, Gould TD. The endophenotype concept in psychiatry: etymology and strategic intentions. Am J Psychiatry [Internet]. 2003 Apr [cited 2012 Mar 1];160(4):636–45. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12668349 2. Drevets WC. Neuroimaging studies of mood disorders. Biol Psychiatry [Internet]. 2000 Oct 15;48(8):813–29. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11301246 3. Dunn RT, Willis MW, Benson BE, Repella JD, Kimbrell T a, Ketter T a, et al. Preliminary findings of uncoupling of flow and metabolism in unipolar compared with bipolar affective illness and normal controls. Psychiatry Res [Internet]. 2005 Nov 30 [cited 2012 Feb 27];140(2):181–98. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16257515 4. Baxter LR, Phelps ME, Mazziotta JC, Schwartz JM, Gerner RH, Selin CE, et al. Cerebral metabolic rates for glucose in mood disorders. Studies with positron emission tomography and fluorodeoxyglucose F 18. Arch Gen Psychiatry [Internet]. 1985 May [cited 2012 Feb 24];42(5): 441–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/3872649 5. Baxter LR, Schwartz JM, Phelps ME, Mazziotta JC, Guze BH, Selin CE, et al. Reduction of prefrontal cortex glucose metabolism common to three types of depression. Arch Gen Psychiatry [Internet]. 1989 Mar [cited 2011 Nov 20];46(3):243–50. Available from: http://www.ncbi.nlm.nih. gov/pubmed/2784046 6. Ketter TA, Kimbrell TA, George MS, Dunn RT, Speer AM, Benson BE, et al. Effects of mood and subtype on cerebral glucose metabolism in treatment-resistant bipolar disorder. Biol Psychiatry [Internet]. 2001;49(2):97–109. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11164756 7. Bauer M, London ED, Rasgon N, Berman SM, Frye M a, Altshuler LL, et al. Supraphysiological doses of levothyroxine alter regional cerebral metabolism and improve mood in bipolar depression. Mol Psychiatry [Internet]. 2005 May [cited 2012 Feb 27];10(5):456–69. Available from: http://www.ncbi.nlm.nih.gov/pubmed/15724143 8. Brooks JO, Wang PW, Strong C, Sachs N, Hoblyn JC, Fenn R, et al. Preliminary evidence of 6 differential relations between prefrontal cortex metabolism and sustained attention in depressed

adults with bipolar disorder and healthy controls. Bipolar Disord [Internet]. 2006 Jun;8(3):248–54. disorders bipolar in SPECT and PET Available from: http://www.ncbi.nlm.nih.gov/pubmed/16696826 9. Buchsbaum MS, Wu J, DeLisi LE, Holcomb H, Kessler R, Johnson J, et al. Frontal cortex and basal ganglia metabolic rates assessed by positron emission tomography with [18F]2-deoxyglucose in affective illness. J Affect Disord [Internet]. 1986;10(2):137–52. Available from: http://www.ncbi.nlm.nih.gov/pubmed/2941470 10. Rush AJ, Schlesser MA, Stokeley E, Bonte FR, Altshuller KZ. Cerebral blood flow in depression and mania. Psychopharmacollogy Bull. 1982;(1):6–7. 11. Silfverskiöld P, Risberg J. Regional cerebral blood flow in depression and mania. Arch Gen Psychiatry [Internet]. 1989 Mar;46(3):253–9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/7568564 12. Tutus a, Simsek a, Sofuoglu S, Nardali M, Kugu N, Karaaslan F, et al. Changes in regional cerebral blood flow demonstrated by single photon emission computed tomography in depressive disorders: comparison of unipolar vs. bipolar subtypes. Psychiatry Res [Internet]. 1998 Sep 28;83(3):169–77. Available from: http://www.ncbi.nlm.nih.gov/pubmed/9849726 13. Miller EK, Freedman DJ, Wallis JD. The prefrontal cortex: categories, concepts and cognition. Philos Trans R Soc Lond B Biol Sci [Internet]. 2002 Aug 29 [cited 2012 Mar 5];357(1424):1123–36. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=1693009&tool=pmcentrez&rendertype=abstract 14. Blumberg HP, Stern E, Ricketts S, Martinez D, de Asis J, White T, et al. Rostral and orbital prefrontal cortex dysfunction in the manic state of bipolar disorder. Am J Psychiatry [Internet]. 1999 Dec;156(12):1986–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/10588416

115 15. Rubinsztein JS, Fletcher PC, Rogers RD, Ho LW, Aigbirhio FI, Paykel ES, et al. Decision-making in mania: a PET study. Brain [Internet]. 2001 Dec;124(Pt 12):2550–63. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11701607 16. Culha AF, Osman O, Dogangün Y, Filiz K, Suna K, Kalkan ON, et al. Changes in regional cerebral blood flow demonstrated by 99mTc-HMPAO SPECT in euthymic bipolar patients. Eur Arch Psychiatry Clin Neurosci [Internet]. 2008 Apr [cited 2011 Oct 19];258(3):144–51. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17990047 17. Krüger S, Alda M, Young LT, Goldapple K, Parikh S, Mayberg HS. Risk and resilience markers in bipolar disorder: brain responses to emotional challenge in bipolar patients and their healthy siblings. Am J Psychiatry [Internet]. 2006 Feb;163(2):257–64. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16449479 18. Rubin E, Sackeim HA, Prohovnik I, Moeller JR, Schnur DB, Mukherjee S. Regional cerebral blood flow in mood disorders: IV. Comparison of mania and depression. Psychiatry Res [Internet]. 1995 May 31 [cited 2012 Feb 27];61(1):1–10. Available from: http://www.ncbi.nlm.nih.gov/pubmed/7568564 19. al-Mousawi a. H, Evans N, Ebmeier KP, Roeda D, Chaloner F, Ashcroft GW. Limbic dysfunction in schizophrenia and mania. A study using 18F- labelled fluorodeoxyglucose and positron emission tomography. Br J Psychiatry [Internet]. 1996 Oct 1 [cited 2012 Feb 27];169(4):509–16. Available from: http://bjp.rcpsych.org/cgi/doi/10.1192/bjp.169.4.509 20. Brooks JO, Hoblyn JC, Woodard S a, Rosen AC, Ketter T a. Corticolimbic metabolic dysregulation in euthymic older adults with bipolar disorder. J Psychiatr Res [Internet]. 2009 Feb [cited 2012 Feb 27];43(5):497–502. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=2693401&tool=pmcentrez&rendertype=abstract 21. Luu P, Posner MI. Anterior cingulate cortex regulation of sympathetic activity. Brain [Internet]. 2003 Oct [cited 2011 Jul 23];126(Pt 10):2119–20. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12959940 22. Drevets WC, Price JL, Simpson JR, Todd RD, Reich T, Vannier M, et al. Subgenual prefrontal cortex abnormalities in mood disorders. Nature [Internet]. 1997 Apr 24 [cited 2011 Jul 27];386(6627):824–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/9126739 23. Dunn RT, Kimbrell TA, Ketter TA, Frye MA, Willis MW, Luckenbaugh DA, et al. Principal components of the Beck Depression Inventory and regional cerebral metabolism in unipolar and bipolar depression. Biol Psychiatry [Internet]. 2002 Mar 1 [cited 2012 Apr 25];51(5):387–99. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0006322301012446 24. Mah L, Zarate C a, Singh J, Duan Y-F, Luckenbaugh D a, Manji HK, et al. Regional cerebral glucose metabolic abnormalities in bipolar II depression. Biol Psychiatry [Internet]. 2007 Mar 15 [cited 2011 Oct 19];61(6):765–75. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17027930 25. Blumberg HP, Stern E, Martinez D, Ricketts S, de Asis J, White T, et al. Increased anterior cingulate and caudate activity in bipolar mania. Biol Psychiatry [Internet]. 2000 Dec 1;48(11):1045–52. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11094137 26. Benabarre A, Vieta E, Martínez-Arán A, Garcia-Garcia M, Martín F, Lomeña F, et al. Neuropsychological disturbances and cerebral blood flow in bipolar disorder. Aust N Z J Psychiatry [Internet]. 2005 Apr [cited 2012 Mar 6];39(4):227–34. Available from: http://www.ncbi.nlm.nih.gov/pubmed/15777358 27. Goodwin GM, Cavanagh JT, Glabus MF, Kehoe RF, O’Carroll RE, Ebmeier KP. Uptake of 99mTc- exametazime shown by single photon emission computed tomography before and after lithium withdrawal in bipolar patients: associations with mania. Br J Psychiatry [Internet]. 1997 May 1 [cited 2012 Feb 27];170(5):426–30. Available from: http://bjp.rcpsych.org/cgi/doi/10.1192/ bjp.170.5.426 28. Packard MG, Knowlton BJ. Learning and memory functions of the Basal Ganglia. Annu Rev Neurosci [Internet]. 2002 Jan [cited 2011 Jun 13];25:563–93. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12052921

116 29. Bonne O, Krausz Y, Gorfine M, Karger H, Gelfin Y, Shapira B, et al. Cerebral hypoperfusion in medication resistant, depressed patients assessed by Tc99m HMPAO SPECT. J Affect Disord [Internet]. 1996 Dec 16 [cited 2012 Apr 25];41(3):163–71. Available from: http://www.sciencedirect.com/science/article/pii/s0165032796000584 30. Ito H, Kawashima R, Awata S, Ono S, Sato K, Goto R, et al. Hypoperfusion in the limbic system and prefrontal cortex in depression: SPECT with anatomic standardization technique. J Nucl Med [Internet]. 1996 Mar;37(3):410–4. Available from: http://www.ncbi.nlm.nih.gov/pubmed/8772633 31. Cohen RM, Semple WE, Gross M, Nordahl TE, King AC, Pickar D, et al. Evidence for common alterations in cerebral glucose metabolism in major affective disorders and schizophrenia. Neuropsychopharmacology [Internet]. 1989 Dec [cited 2012 Feb 27];2(4):241–54. Available from: http://www.ncbi.nlm.nih.gov/pubmed/2610821 32. Drevets WC, Price JL, Bardgett ME, Reich T, Todd RD, Raichle ME. Glucose metabolism in the amygdala in depression: relationship to diagnostic subtype and plasma cortisol levels. Pharmacol Biochem Behav [Internet]. 2002 Mar;71(3):431–47. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11830178 33. Gonul AS, Coburn K, Kula M. Cerebral blood flow, metabolic, receptor, and transporter changes in bipolar disorder: the role of PET and SPECT studies. Int Rev Psychiatry [Internet]. 2009 Jan [cited 2011 Dec 8];21(4):323–35. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20374147 34. Gyulai L, Alavi A, Broich K, Reilley J, Ball WB, Whybrow PC. I-123 iofetamine single-photon computed emission tomography in rapid cycling bipolar disorder: a clinical study. Biol Psychiatry [Internet]. 1997;41(2):152–61. Available from: http://www.ncbi.nlm.nih.gov/pubmed/9018385 35. Schildkraut JJ. The catecholamine hypothesis of affective disorders: a review of supporting evidence. Am J Psychiatry [Internet]. 1965 Nov [cited 2012 Mar 7];122(5):509–22. Available from: http://www.ncbi.nlm.nih.gov/pubmed/5319766 36. Ichimiya T, Suhara T, Sudo Y, Okubo Y, Nakayama K, Nankai M, et al. Serotonin transporter binding in patients with mood disorders: a PET study with [11C](+)McN5652. Biol Psychiatry [Internet]. 2002 May 1 [cited 2013 Nov 28];51(9):715–22. Available from: 6 http://www.ncbi.nlm.nih.gov/pubmed/11983185 37. Oquendo M a, Hastings RS, Huang Y-Y, Simpson N, Ogden RT, Hu X-Z, et al. Brain serotonin transporter binding in depressed patients with bipolar disorder using positron emission disorders bipolar in SPECT and PET tomography. Arch Gen Psychiatry [Internet]. 2007 Feb;64(2):201–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17283287 38. Chou Y-H, Wang S-J, Lin C-L, Mao W-C, Lee S-M, Liao M-H. Decreased brain serotonin transporter binding in the euthymic state of bipolar I but not bipolar II disorder: a SPECT study. Bipolar Disord [Internet]. 2010 May [cited 2012 Oct 11];12(3):312–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20565438 39. Cannon DM, Ichise M, Fromm SJ, Nugent AC, Rollis D, Gandhi SK, et al. Serotonin transporter binding in bipolar disorder assessed using [11C]DASB and positron emission tomography. Biol Psychiatry [Internet]. 2006 Aug 1 [cited 2011 Jul 27];60(3):207–17. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16875929 40. Cannon DM, Ichise M, Rollis D, Klaver JM, Gandhi SK, Charney DS, et al. Elevated serotonin transporter binding in major depressive disorder assessed using positron emission tomography and [11C]DASB; comparison with bipolar disorder. Biol Psychiatry [Internet]. 2007 Oct 15 [cited 2011 Jun 15];62(8):870–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17678634 41. Yatham LN, Liddle PF, Lam RW, Adam MJ, Solomons K, Chinnapalli M, et al. A positron emission tomography study of the effects of treatment with valproate on brain 5-HT2A receptors in acute mania. Bipolar Disord [Internet]. 2005 Jan;7 Suppl 5:53–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16225561 42. Beaulieu J, Gainetdinov RR. The physiology, signaling, and pharmacology of dopamine receptors. Pharmacol Rev [Internet]. 2011 Mar [cited 2012 Jul 27];63(1):182–217. Available from: http://www.ncbi.nlm.nih.gov/pubmed/21303898

117 43. Cousins D a, Butts K, Young AH. The role of dopamine in bipolar disorder. Bipolar Disord [Internet]. 2009 Dec;11(8):787–806. Available from: http://www.ncbi.nlm.nih.gov/pubmed/19922550 44. Hartman DS, Civelli O. Molecular attributes of dopamine receptors: new potential for antipsychotic drug development. Ann Med [Internet]. 1996 Jun;28(3):211–9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/8811164 45. Little KY, Krolewski DM, Zhang L, Cassin BJ. Loss of striatal vesicular monoamine transporter protein (VMAT2) in human cocaine users. Am J Psychiatry [Internet]. 2003 Jan [cited 2012 Mar 8];160(1):47–55. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12505801 46. Yildiz A, Vieta E, Leucht S, Baldessarini RJ. Efficacy of antimanic treatments: meta-analysis of randomized, controlled trials. Neuropsychopharmacology [Internet]. 2011 Jan [cited 2012 Mar 7];36(2):375–89. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=3055677&tool=pmcentrez&rendertype=abstract 47. Ginovart N. Imaging the dopamine system with in vivo [11C]raclopride displacement studies: understanding the true mechanism. Mol Imaging Biol [Internet]. 2005 [cited 2012 Mar 7];7(1):45– 52. Available from: http://www.ncbi.nlm.nih.gov/pubmed/15912275 48. Wong DF, Wagner HN, Pearlson G, Dannals RF, Links JM, Ravert HT, et al. Dopamine receptor binding of C-11-3-N-methylspiperone in the caudate in schizophrenia and bipolar disorder: a preliminary report. Psychopharmacol Bull [Internet]. 1985 Jan [cited 2012 Mar 7];21(3):595–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/4034877 49. Pearlson GD, Wong DF, Tune LE, Ross CA, Chase GA, Links JM, et al. In vivo D2 dopamine receptor density in psychotic and nonpsychotic patients with bipolar disorder. Arch Gen Psychiatry [Internet]. 1995 Jun [cited 2012 Mar 7];52(6):471–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/7771917 50. Anand a, Verhoeff P, Seneca N, Zoghbi SS, Seibyl JP, Charney DS, et al. Brain SPECT imaging of amphetamine-induced dopamine release in euthymic bipolar disorder patients. Am J Psychiatry [Internet]. 2000 Jul;157(7):1108–14. Available from: http://www.ncbi.nlm.nih.gov/pubmed/10873919 51. Yatham LN, Liddle PF, Lam RW, Shiah I-S, Lane C, Stoessl AJ, et al. PET study of the effects of valproate on dopamine D(2) receptors in neuroleptic- and mood-stabilizer-naive patients with nonpsychotic mania. Am J Psychiatry [Internet]. 2002 Oct [cited 2012 Mar 7];159(10):1718–23. Available from: http://www.ncbi.nlm.nih.gov/pubmed/12359678 52. Suhara T, Nakayama K, Inoue O, Fukuda H, Shimizu M, Mori A, et al. D1 dopamine receptor binding in mood disorders measured by positron emission tomography. Psychopharmacology (Berl) [Internet]. 1992 Jan [cited 2012 Apr 25];106(1):14–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/1531387 53. Yatham LN, Liddle PF, Shiah I-S, Lam RW, Ngan E, Scarrow G, et al. PET study of [(18)F]6-fluoro-L- dopa uptake in neuroleptic- and mood-stabilizer-naive first-episode nonpsychotic mania: effects of treatment with divalproex sodium. Am J Psychiatry [Internet]. 2002 May [cited 2012 Mar 7];159(5):768–74. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11986130 54. Yatham LN, Goldstein JM, Vieta E, Bowden CL, Grunze H, Post RM, et al. Atypical antipsychotics in bipolar depression: potential mechanisms of action. J Clin Psychiatry [Internet]. 2005 Jan [cited 2012 Mar 8];66 Suppl 5:40–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16038601 55. Amsterdam JD, Newberg AB. A preliminary study of dopamine transporter binding in bipolar and unipolar depressed patients and healthy controls. Neuropsychobiology [Internet]. 2007 Jan [cited 2012 Mar 4];55(3–4):167–70. Available from: http://www.ncbi.nlm.nih.gov/pubmed/17657170 56. Chang TT, Yeh TL, Chiu NT, Chen PS, Huang HY, Yang YK, et al. Higher striatal dopamine transporters in euthymic patients with bipolar disorder: a SPECT study with [Tc] TRODAT-1. Bipolar Disord [Internet]. 2010 Feb;12(1):102–6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20148872

118 57. Anand A, Barkay G, Dzemidzic M, Albrecht D, Karne H, Zheng Q-H, et al. Striatal dopamine transporter availability in unmedicated bipolar disorder. Bipolar Disord [Internet]. 2011 Jun [cited 2012 Feb 27];13(4):406–13. Available from: http://www.ncbi.nlm.nih.gov/pubmed/21843280 58. Zubieta JK, Taylor SF, Huguelet P, Koeppe R a, Kilbourn MR, Frey K a. Vesicular monoamine transporter concentrations in bipolar disorder type I, schizophrenia, and healthy subjects. Biol Psychiatry [Internet]. 2001 Jan 15;49(2):110–6. Available from: http://www.ncbi.nlm.nih.gov/pubmed/11164757 59. Dilsaver SC. Pathophysiology of “cholinoceptor supersensitivity” in affective disorders. Biol Psychiatry [Internet]. 1986 Jul [cited 2012 Mar 8];21(8–9):813–29. Available from: http://www.ncbi.nlm.nih.gov/pubmed/3015271 60. Cannon DM, Carson RE, Nugent AC, Eckelman WC, Kiesewetter DO, Williams J, et al. Reduced muscarinic type 2 receptor binding in subjects with bipolar disorder. Arch Gen Psychiatry [Internet]. 2006 Jul;63(7):741–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16818863 61. Cannon DM, Klaver JK, Gandhi SK, Solorio G, Peck S a, Erickson K, et al. Genetic variation in cholinergic muscarinic-2 receptor gene modulates M2 receptor binding in vivo and accounts for reduced binding in bipolar disorder. Mol Psychiatry [Internet]. 2011 Apr [cited 2012 Feb 27];16(4):407–18. Available from: http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=2895691&tool=pmcentrez&rendertype=abstract 62. Lynch PJ, Jaffe CC. Brain human sagittal section [Internet]. Wikimedia Commons. 2006. Available from: https://commons.wikimedia.org/wiki/File:Brain_human_sagittal_section.svg?uselang=en

6 PET and SPECT in bipolar disorders bipolar in SPECT and PET

119

CHAPTER 7 Neuroinflammation in bipolar disorder

A [11C]-(R)-PK11195 positron emission tomography study

Bartholomeus C.M. Haarman, Rixt F. Riemersma – Van der Lek, Jan Cees de Groot, Henricus G. Ruhé, Hans C. Klein, Tjitske E. Zandstra, Huibert Burger, Robert A. Schoevers, Erik F.J. de Vries, Hemmo A. Drexhage, Willem A. Nolen, Janine Doorduin

Brain, Behavior, and Immunity. 2014 Aug 3;40:219–25. Abstract

Background The “monocyte-T-cell theory of mood disorders” regards neuroinflammation, i.e. marked activation of microglia, as a driving force in bipolar disorder. Microglia activation can be visualized in vivo using [11C]-(R)-PK11195 PET. Indirect evidence suggests the hippocampus as a potential focus of neuroinflammation in bipolar disorder. We aim to demonstrate that there is increased [11C]-(R)-PK11195 binding to activated microglia in the hippocampus of patients with bipolar I disorder when compared to healthy controls.

Material and Methods Fourteen patients with bipolar I disorder and eleven healthy controls were included in the analyses. Dynamic 60-min PET scans were acquired after the injection of [ 11C]-(R)-PK11195. All subjects underwent psychiatric interviews as well as an MRI scan, which was used for anatomic co-registration in the data analysis. The data from the PET scans was analyzed with a two-tissue-compartment model to calculate the binding potential, using the metabolite-corrected plasma and blood curve as input.

Results A significantly increased [11C]-(R)-PK11195 binding potential, which is indicative of neuroinflammation, was found in the right hippocampus of the patients when compared to the healthy controls (1.66 (CI 1.45 – 1.91) versus 1.33 (CI 1.16 – 1.53); p=0.033, respectively). Although the same trend was observed in the left hippo campus, this difference was not statistically significant.

Conclusion This study is the first to demonstrate the presence of focal neuroinflammation in the right hippocampus in bipolar I disorder.

122 Introduction

The pathophysiology of bipolar disorder (BD) is complex and its neurobiology remains largely unknown1. Both the stress system and the immune system interact with the brain and are influenced by the environment. Their interactions can be regarded as possible linking pins. The “monocyte-T-cell theory of mood disorders” 2 considers an activated inflammatory response system (IRS) in mood disorders to be the driving force behind these illnesses. IRS activation can be regarded as a disbalance in immune regulatory processes. In BD this theory is supported by altered concentrations of immune related peripheral bio-assays, e.g. elevated serum or plasma levels of pro-inflammatory cytokines, aberrant expression of pro-inflammatory genes in circulating monocytes3, alterations in the kynurenine pathway4 and a modulating effect of several psychopharmaceuticals on the immune system3,5–8. Activation of the IRS is thought to correspond to neuroinflammation, which is reflected by an increase in activated microglia, the resident macrophages of the brain9. Indirect evidence gathered from post-mortem studies, corticosteroid treatment related hippocampal changes and animal models suggest the hippocampus as a potential focus of neuroinflammation in BD. Post-mortem studies in humans demonstrated an increased expression of inflammation related pro-apoptosis genes10 and oxidative damage11 to the RNA in the hippocampus of BD patients as well as a decrease in growth-associated protein12 that has been proposed to be related to neuroinflammation13. In addition, medical treatment with corticosteroids, known for their anti-inflammatory effects and associated with not only depressive symptoms but also manic symptoms, is related to both functional and morphological changes in the hippocampus14. Furthermore, multiple rodent studies demonstrate a relationship 7 between a disturbed microglia function in the hippocampus and other pathophysi- 9,15–19 ological changes which are thought to relate to neuropsychiatric disorders . In a Neuroinflammation in bipolar disorder recent rodent study stress-induction led to dynamic microglia changes in the hippocampus alone, which were associated with depressive-like behavior18. Another study demonstrated an increased cytokine response to lipopolysaccharide challenge in the hippocampus of SERT mutant rats19. Microglia activation can be visualized in vivo with the radiopharmaceutical [ 11C]-(R)-PK11195 by means of positron emission tomography (PET). This radiopharmaceutical binds to the translocator protein (TSPO), a receptor that is upregulated in the mitochondria of activated microglia cells20. [11C]-(R)-PK11195 has been utilized successfully in models of central inflammation, such as following an injection of the endotoxin lipopolysaccharide in animal models21–23 and following infections of the central nervous system in humans24–26. In various psychiatric and neurodegenerative disorders [11C]-(R)-PK11195 PET has proven to be a useful tool for imaging neuro inflammation27–30. Using this radiopharmaceutical Doorduin et al. demonstrated the

123 hippocampus to be the primary focus of neuroinflammation in schizophrenia-related psychosis30. New radiopharmaceuticals such as [11C]-PBR28 have been developed that are potentially more sensitive for imaging of neuroinflammation. However, these revealed substantial heterogeneity in binding affinity due to polymorphisms in the TSPO, resulting in so-called high-, median- and low-affinity binders, complicating data interpretation31. Since [11C]-(R)-PK11195 binding is not affected by polymorphisms we have selected this radiopharmaceutical for our study. In the current study we aim to demonstrate an increased [11C]-(R)-PK11195 binding to activated microglia in BD patients in comparison to a healthy control group. We a priori hypothesized the hippocampus to be the main focus of neuroinflammation in BD. In a second model we explored the presence of neuroinflammation in other brain regions. In addition, we examined whether clinical characteristics would be associated with neuroinflammation.

124 Materials and Methods

Participants For the present cross-sectional case-control study we included 15 preferably euthymic patients with bipolar I disorder (BD-I) and a group of 12 controls matching demographically in age and sex that participated in the MOODINFLAME study. The patients were recruited from an outpatient clinic for BD. The healthy controls were recruited via advertisements, recruitment posters and by contacting healthy controls from previous studies that gave their consent to be asked for future studies. For the MOODINFLAME study adult male and female subjects were included who were free of inflammation related symptoms, including fever and infectious or inflammatory disease. Furthermore, they were free of uncontrolled systemic disease, uncontrolled metabolic disease or other significant uncontrolled somatic disorder known to affect mood. They did not use somatic medication known to affect mood or the immune system, such as corticosteroids, non-steroid anti-inflammatory drugs and statins. Female candidates who were pregnant or recently gave birth were excluded. Patients and controls were free of benzodiazepines at least in the last week prior to the PET- scan. They were also free of anticoagulant use or presence of coagulation disease, did not suffer from palmar arc artery insufficiency, did not participate in a prior research study involving radiation less than a year ago, and did not have any contraindication for MRI scanning. Patients were allowed to continue their regular psychopharmaceutical treatment. They were neither in a depressed nor (hypo-)manic episode at the time of scanning as indicated by an Inventory of Depressive Symptoms - Clinician Version (IDS-C30) score <22 and Young Mania Rating Scale (YMRS) score <12, respectively. Patients 7 with any other current primary major psychiatric diagnosis were excluded including: schizophrenia, schizoaffective disorder, anxiety disorder and substance use disorders. Neuroinflammation in bipolar disorder Healthy controls did not have any current or lifetime psychiatric diagnosis. Nine patients were excluded due to protocol violations (claustrophobia (3), presence of ferromagnetic objects (2), palmar arc artery insufficiency (1), coagulation disorder (1), use of benzodiazepines (1) and pregnancy (1)). Four healthy controls were excluded due to protocol violations (claustrophobia (2) and presence of ferromagnetic objects (2)). After completing inclusion one participant admitted to having used a benzodiazepine on the evening prior to the PET-scan. In another participant, experienced technical difficulties with the automatic blood sampling system prohibited valid determination of the input function. These two subjects were removed from the subsequent analyses.

125 Ethical Considerations The Medical Ethical Review Committee of the University Medical Center Gronin- gen approved the protocol, which was performed in accordance with the Helsinki Declaration of 1975. Written informed consent was obtained from all participants.

Assessment All subjects underwent a Mini-International Neuropsychiatric Interview 5.0.0 (MINI) to confirm the bipolar I disorder diagnosis in the patient group and the absence of psychiatric disorders in the healthy control group32. Clinical features were extracted from the interviews held according to the MOODINFLAME protocol. This included the Patient Questionnaire of the former Stanley Foundation Bipolar Network, the YMRS, the IDS-C30 and a somatic illness questionnaire. The Patient Questionnaire includes separate clinician and patient chapters covering a spectrum of clinical features including vocational, educational and economic status, onset and course of illness, family history, past treatment, cycling and seasonal patterns, medical problems, medications, ability to function and symptomatic status, precipitants of illness (e.g. substance use), treatment adherence and insight into the illness33. In the event of a mismatch of results from the MINI in relation to the Patient Questionnaire, diagnoses were checked with the treating physician. The YMRS is an 34 eleven-item, multiple-choice questionnaire to assess manic symptoms . The IDS-C30 is a thirty-item, multiple-choice questionnaire to assess depressive symptoms of all 35 symptom domains of depression . The YMRS and IDS-C30 were assessed shortly before the scans and used in the relevant analyses. The somatic illness questionnaire is a MOODINFLAME specific checklist exploring all the organ systems for current and lifetime medical symptoms.

Radiochemistry [11C]-(R)-PK11195 was labeled as described previously 30. [11C]-(R)-PK11195 was obtained in 32±18% radiochemical yield (n=27). The quality control was performed by

HPLC, using a Novapak C18 column (150x3.9 mm) with acetonitrile/25 mM NaH2PO4 (pH 3.5) (60/40) as the eluent at a flow of 1 ml/min. The radiochemical purity was always >95% and the specific activity was 111±130 GBq/μmol. No differences were found between healthy volunteers and patients for the injected dose (390±18 vs. 355±63 MBq, p=0.088) and injected mass (0.75±0.58 vs. 0.67±0.40 mg/L, p=0.696).

Positron Emission Tomography For arterial blood sampling a catheter was inserted in the radial artery after testing for collateral circulation with the Allen test and injection of 1% lidocaine (Fresenius Kabi Nederland BV, ‘s Hertogenbosch, The Netherlands) for local anesthesia. In the other arm, a venous catheter was inserted in the antebrachial vein for injection of

126 [11C]-(R)-PK11195. Positron emission tomography imaging was performed with the ECAT EXACT HR+ camera (Siemens, Knoxville, Tenessee). Head movement was minimized with a head-restraining adhesive band and a neuroshield was used to minimize the interference of radiation from the subject’s body. A 60-minute emission scan in 3D-mode was performed, starting simultaneously with the intravenous injection of [11C]-(R)-PK11195. The tracer was injected at a speed of 0.5 ml/sec (total volume of 8.3 ml). After radiotracer injection, arterial blood radioactivity was continuously monitored with an automated blood sampling system (Veenstra Instruments, Joure, The Netherlands). Five extra blood samples were collected at 10, 20, 30, 45 and 60 minutes after [11C]-(R)-PK11195 injection to determine the amount of radioactivity in blood and plasma to calibrate the sampling system. The arterial blood samples that were collected at 20, 45 and 60 minutes after [11C]-(R)-PK11195 injection were also used for metabolite analysis. The metabolite analysis was performed as described previously30.

Magnetic Resonance Imaging Axial T1 (gradient echo T1 3D, slice thickness 1.2 mm isotropic) and T2 Flair (3 mm) weighed images were acquired using a 3T MRI scanner and an eight-channel head coil (3T Intera, Philips, Best, The Netherlands). The T1 images provided the input for the normalization of the PET scans to standard brain morphology of all subjects. Both the MRI scan and the PET scan were preferably made on the same day, and no more than 1 week apart. Image Analysis 7 Attenuation correction was performed with the separate ellipse algorithm. Images were reconstructed by filtered back projection in 21 successive frames of increas- Neuroinflammation in bipolar disorder ing duration (6x 10 s, 2x 30 s, 3x 1 min, 2x 2min, 2x 3 min, 3x 5min, 3x 10 min). MRI images were co-registered to the sum of all frames of the PET scan, resulting in the most optimal co-registration, using statistical parametric mapping (SPM8; Welcome Trust Center Neuroimaging, University College London, UK). Grey matter regions of interest (ROIs) were defined by the co-registered MRI images using a probability map that was based on automatic delineation of ROIs with the PVElab software36. The ROIs were transferred to the dynamic PET images and time-activity curves were calculated. In total 15 ROIs were included: left and right hippocampus, left and right frontal cortex, left and right dorsolateral prefrontal cortex (PFC), left and right temporal cortex, left and right parietal cortex, bilateral occipital cortex, bilateral anterior cingulate, bilateral posterior cingulate, bilateral cerebellum and basal ganglia. The time-activity curves of all ROIs were used for kinetic modeling with software developed in Matlab 7.1 (Mathworks, Natick, Massachusets). The individual delay was

127 corrected for the delay in radioactivity measurements in blood, as a result of the distance between the subject and the automated blood sampling system. A two-tissue compartment model was used to calculate the k1-k4 with the metabolite corrected plasma and blood curve as an input function, correcting for the individual delay and a free blood volume. The binding potential was defined as k3/k4 and was calculated for each ROI individually.

Statistical Analysis Statistical analyses were performed using Stata Statistical Software, release 11 (StataCorp. 2009, College Station, TX). The differences in demographic data between the groups were investigated with Student’s t-test (age), Pearson’s chi-squared test (gender) and Kruskal-Wallis

equality-of-populations rank test (IDS-C30 score). Student’s t-test was used to determine differences in whole-brain grey matter binding potential between the patient and healthy control group. Statistical analyses of the binding potentials in the examined brain regions were performed using two general linear models. The first model investigated the hypothesis that the hippocampi are the focus of neuroinflammation, incorporating the binding potentials of the left and right hippocampus as dependent variables. In the second model, exploring the other brain regions, binding potentials of all investigated brain regions were added as dependent variables. In both models the whole-brain grey matter binding potential was used as a covariate to normalize for individual global cerebral [11C]-(R)-PK11195 uptake variations. Beforehand inverse square root transformation was applied to the binding potentials to meet the normality assumption in the general linear models. Results of the general linear models are presented as back-transformed means of the individual ROIs with 95% confidence intervals (CI). The level of significance was defined as 0.05, two-sided, in the first (hypotheses driven) model. To correct for multiple comparisons in the second (explorative) model false discovery rate (FDR<0.1) correction was applied to the results, as described by Benjamini-Hochberg37. In order to rule out possible epilepsy associated inflammation effects38 post hoc analyses were performed excluding the patient with this comorbidity. Furthermore, to increase the homogeneity of the sample additional post hoc analyses were performed excluding the patient with mild depressive symptoms and the medication free patient.

Correlations between the binding potentials and the IDS-C30 score, YMRS score, age at onset, number of episodes and duration of illness were assessed with Spearman’s rho.

128 Results

Demographics Subject characteristics are displayed in Table 1. While all but one patient were euthymic (IDS-C30 score <12), a statistical difference in the IDS-C30 score between the patient and healthy control groups was observed (H=4.676; p=0.031). Differences between the groups in gender or age were not statistically significant.

[11C]-(R)-PK11195 PET There was no statistical difference in the whole-brain grey matter binding potential of [11C]-(R)-PK11195 in patients versus controls (1.28 (CI 1.07 – 1.50) versus 1.35 (CI 1.10 – 1.60); p=0.70). The hypothesis driven general linear model with the left and right hippocampi binding potentials as the dependent variables demonstrated a significant increased [ 11C]-(R)-PK11195 binding potential in the right hippocampus of the patients when compared to the healthy controls (1.66 (CI 1.45 – 1.91) versus 1.33 (CI 1.16 – 1.53); p=0.033; figure 1). The difference between the [11C]-(R)-PK11195 binding potential of the left hippocampus of BD-I patients compared to the healthy controls was not statistically significant (1.55 (CI 1.30 – 1.90) versus 1.20 (CI 1.00 – 1.46); p=0.071). The subsequent explorative general linear model adding all the investigated brain regions revealed a lower [11C]-(R)-PK11195 binding potential in the left dorsolateral PFC of the BD-I patients when compared tot healthy controls (1.18 (CI 1.09 – 1.27) versus 1.40 (CI 1.28 – 1.53); p=0.009). However, this difference did not survive correction for the false discovery rate (table 2). Post hoc analyses that were performed excluding the patient with epilepsy demon- 7 strated an increased [11C]-(R)-PK11195 binding potential in both the right hippocampus (1.69 (CI 1.47 – 1.96) versus 1.33 (CI 1.17 – 1.54); p=0.029) as well as the left Neuroinflammation in bipolar disorder hippocampus (1.63 (CI 1.35 – 2.00) versus 1.21 (CI 1.02 – 1.47); p=0.042) of the BD-I patients when compared to healthy controls. In the explorative model the results were comparable to the original analyses in effect size and statistical significance. In post hoc analyses excluding the patient with mild depressive symptoms and the medication free patient the results were comparable to the original analyses in effect size and statistical significance in both models.

129

– Inventory of 30

Medication

COCP(5)

roate, Thyroxine, Omeprazol (32) Omeprazol Thyroxine, roate,

(time since last medication switch (months)) switch medication last since (time

(21) ithium, Valp

Valproate, Thyroxine, Trazodone, Lamotrigine, Quetiapine (11) Quetiapine Lamotrigine, Trazodone, Thyroxine, Valproate, L (63) Lamotrigine Lithium, Thyroxine, Lithium, (43) Trazodone (14) Citalopram Carbamazepine, (25) Lamotrigine Lithium, (11) Quetiapine Valproate, (13) Trazodone Lithium, (30) Valproate (4) Quetiapine Lithium, - (10) Levetiracetam Valproate, (10) Lithium ------

Medical Medical ypothyroidism h

comorbidities ypothyroidism

Epilepsy - DH; - H ------Current medical comorbidities requiring medical IDS-C care. requiring medical comorbidities Current

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 score YMRS

2 2 2 2 2 0 5 0 5 0 0 0 0 0 0 6 3 3 3 4 4 11

12 14 30 C - core s IDS

episodes episodes

№

/1) depressed/(hypo)manic) (>5/>11) (>11/>5) (>5/3) (>11/>11) (>5/>5) (>20/>20) (>20/>20) (>5/3) (>11/2) (4/4) (>20/>20) (0 (1/2) (1/>11)

17 17 12 21 21 18 15 18 16 14 22 25 30 43 Age at onset at

21 21 61 41 37 22 52 25 58 55 50 55 50 26 56 56 53 66 36 36 24 40 45 49 64 Age

Sex F F M F F F M M M F M M M F F F F M F F F F M M M

1 1 7 7 2 2 5 8 5 8 6 9 6 9 3 3 4 4 11 11 № 12 10 10 13 14 Patients controls Healthy

TABLE 1 characteristics Subject Overview of the characteristics of the BD-I patients and healthy controls. Depressive Symptoms Clinician version; YMRS Mania – Young Rating Scale; DH – diaphragmatic hernia; COCP – combined oral contraceptive pill

130 FIGURE 1 Results of the hypothesis driven analysis

Mean adjusted [11C]-(R)-PK11195 binding potentials in the left and right hippocampus of healthy controls and BD-I patients. Statistical analysis on the binding potentials was performed using a general linear model, with the whole-brain grey matter binding potential as a covariate to correct for global [11C]-(R)-PK11195 uptake. Beforehand inverse square root transformation was applied to the binding potentials to meet the normality assumption in the general linear model. Results are presented as bars (back-transformed means) with error stripes (standard error). A significantly increased [11C]-(R)-PK11195 binding potential was observed in the right hippocampus of the patients when compared to the healthy controls (p=0.033). * statistically significant p<0.05. 7 Neuroinflammation in bipolar disorder

131 TABLE 2 Results of the explorative analysis

Adjusted [11C]-(R)-PK11195 Binding Potential p Region of Interest Healthy Controls BD-I Patients (mean (CI)) (mean (CI)) Left Hippocampus 1.20 (1.00 - 1.46) 1.55 (1.30 - 1.90) 0.071 Right Hippocampus 1.33 (1.16 - 1.53) 1.66 (1.45 - 1.91) 0.033 Left Frontal Cortex 1.24 (1.10 - 1.41) 1.24 (1.12 - 1.39) 0.184 Left Dorsolateral PFC 1.40 (1.28 - 1.53) 1.18 (1.09 - 1.27) 0.009 Left Temporal Cortex 1.19 (1.09 - 1.29) 1.26 (1.17 - 1.36) 0.315 Left Parietal Cortex 1.37 (1.26 - 1.50) 1.33 (1.23 - 1.44) 0.638 Right Frontal Cortex 1.20 (1.12 - 1.28) 1.27 (1.20 - 1.35) 0.184 Right Dorsolateral PFC 1.25 (1.12 - 1.42) 1.25 (1.13 - 1.39) 0.970 Right Temporal Cortex 1.21 (1.13 - 1.31) 1.32 (1.24 - 1.42) 0.111 Righ Parietal Cortex 1.40 (1.27 - 1.57) 1.40 (1.28 - 1.54) 0.956 Occipital Cortex 1.28 (1.19 - 1.37) 1.40 (1.31 - 1.49) 0.081 Anterior Cingulate 1.39 (1.24 - 1.56) 1.29 (1.17 - 1.43) 0.397 Posterior Cingulate 1.26 (1.10 - 1.46) 1.36 (1.20 - 1.55) 0.456 Cerebellum 1.01 (0.94 - 1.10) 1.13 (1.05 - 1.22) 0.055 Basal Ganglia 1.24 (1.16 - 1.31) 1.24 (1.18 - 1.31) 0.891

Mean adjusted [11C]-(R)-PK11195 binding potentials in the left hippocampus, right hippocampus, left frontal cortex, left dorsolateral prefrontal cortex (PFC), left parietal cortex, left temporal cortex, right frontal cortex, right dorsolateral PFC, right parietal cortex, right, temporal cortex, occipital cortex, anterior cingulate, posterior cingulate, basal ganglia and cerebellum of healthy controls and BD-I patients. Statistical analysis on the binding potentials was performed using a general linear model, with the whole-brain grey matter binding potential as a covariate to correct for global [11C]-(R)-PK11195 uptake. Beforehand inverse square root transformation was applied to the binding potentials to meet the normality assumption in the general linear model. Results are presented as back-transformed means with confidence interval (CI). A lower [11C]-(R)-PK11195 binding potential in the left dorsolateral PFC of the patients when compared tot healthy controls (1.18 (CI 1.09 – 1.27) versus 1.40 (CI 1.28 – 1.53); p=0.009) was revealed. However this difference did not survive correction for the false discovery rate (FDR<0.1).

Association with Clinical Features

Correlations between the hippocampi binding potentials and the total IDS-C30 score or individual items were not statistically significant in both the patient and healthy control group. All patients and healthy controls scored 0 on the YMRS. Therefore, correlations could not be calculated with manic symptoms. The correlations between the hippocampal binding potentials and the illness progression characteristics (number of depressive or manic episodes, total number of mood episodes) were also not statistically significant.

132 Discussion

To our knowledge this is the first study to reveal actual neuroinflammation in vivo in BD. We partly confirmed our a priori hypothesis, demonstrating a statistically significant increased binding potential of [11C]-(R)-PK11195 in the right hippocampus of BD-I patients as compared to healthy controls. The left hippocampus [ 11C]-(R)-PK11195 binding potential showed the same trend as the right hippocampus, with a comparable increase in binding potential, but it was not statistically significant. Although the effect size had the same magnitude as the right hippocampus, a slightly larger standard error was observed in the calculations and the study may have been underpowered to demonstrate a difference in the left hippocampus between patients and controls. This supposition is supported by a post hoc analysis excluding the patient with epilepsy that demonstrated a significant difference in the left hippocampus as well. The finding of neuroinflammation in BD corroborates previous studies which used less direct indicators of immune activation: an increase in peripheral TSPO receptors in platelets of BD patients, described by Marazziti39; peripheral blood monocyte gene expression found to be related to hemodynamic changes measured by functional MRI in the hippocampus of a combined sample of unipolar and bipolar depressed patients described by Savitz et al.40 and multiple studies investigating immune system related peripheral blood derived bio-assays, described above3,4,8,9. To increase the understanding of the role of immune activation in the pathophysiology of BD further research on the relationship between the peripheral blood derived bioassays and central nervous system neuroinflammation is necessary. The same holds true on the relationship between the various functional neuroimaging 7 observations and neuroinflammation. Animal model studies can be directive in this regard. These demonstrated that microglia have an active role in the development of Neuroinflammation in bipolar disorder mature synapses during embryogenesis41, pruning synapses postnatally42, regulating neurogenesis43 and inducing apoptosis9 in the hippocampus as well as other regions. It is tempting to speculate that these cellular processes (partially) explain the metabolic disturbances44,45 and the decreased neuronal viability observed in neuroimaging studies46. The decreased [11C]-(R)-PK11195 binding potential in the left dorsolateral PFC of BD-I patients compared to controls in the explorative analysis is possibly a false-positive finding as it was no longer statistically significant after FDR correction for multiple testing. However, a possible differential [11C]-(R)-PK11195 binding potential between the right hippocampus and the left dorsolateral PFC could also be regarded in the view of recent resting state fMRI connectivity studies that demonstrated aberrant connectivity between the PFC regions and limbic system regions in BD47.

133 It must be noted that our patients were almost all in the euthymic state, so they were not markedly depressed. It remains uncertain whether the inflammatory response would be greater during a depressive or manic episode. Previously neuroinflammation PET studies have been performed in schizophrenia and unipolar major depressive disorder (MDD). Our finding corresponds with the result in the study by Doorduin et al.30, albeit that the effect size in the right hippocampus in BD is smaller than that was found in schizophrenia-related psychotic patients, possibly related to the more extensive symptomatology of these patients. Another study on TSPO binding in patients with mild or moderate MDD using the radiopharmaceutical [11C]-PBR28 did not demonstrate a difference between patients and controls48. Although the hippocampus was not a specific region of interest in their study and subjects with a high sensitive C-reactive protein level of more than 5mg/l were excluded, it could be argued that perhaps neuroinflammation plays a more important role in BD than in MDD. The present study has several inevitable limitations. Increased [11C]-(R)-PK11195 binding to the TSPO receptor in the brain is traditionally related to microglia ac- tivation9. It is important to note that the TSPO receptor can also be expressed in astrocytes, potentially influencing the [11C]-(R)-PK11195 binding potential signal49. However, because both cells are known to contribute to neuroinflammation50, it can be argued that regardless of activated microglia cells or astrocytes being responsible for the increased TSPO expression, the increased [11C]-(R)-PK11195 binding most likely represents a neuroinflammatory process either way. The naturalistic design of the study does not take the possible confounding effect of concomitant medication use into account. It is known that most mood stabilizing medications, including lithium, anticonvulsants and antipsychotics, have an effect on the immune system3,5–8. However, their effects are generally immunosuppressive in nature. It can be argued that in the present study most medications would actually have diminished the effect of the observed neuroinflammation, so the amount of microglia activation in medication-free euthymic BD-patients could be even larger, compared to controls.

134 Conclusion In conclusion, this study demonstrates the presence of focal neuroinflammation in the right hippocampus of BD-I patients, being a point of departure for unraveling the role of in vivo immune activation in the pathophysiology of BD.

Acknowledgements We thank Juliëtte Kalkman for accompanying the patients; Johan Wiegers, Aafke Zeilstra, Remko Koning, Eelco Severs en Paul van Snick for their assistance with the acquisition of PET scans; Anita Sibeijn-Kuiper, Judith Streurman and Remco Renken for their assistance with the acquisition of MRI images; Sarah Hamel-Brown for giving suggestions regarding grammar, flow, aesthetics and language of the manuscript.

7 Neuroinflammation in bipolar disorder

135 References

1. Schulze TG. Genetic research into bipolar disorder: the need for a research framework that integrates sophisticated molecular biology and clinically informed phenotype characterization. Psychiatr Clin North Am. 2010 Mar;33(1):67–82. 2. Maes M, Smith R, Scharpe S. The monocyte-T-lymphocyte hypothesis of major depression. Psychoneuroendocrinology. 1995 Jan;20(2):111–6. 3. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 4. Myint AM, Kim Y-K, Verkerk R, Park SH, Scharpé S, Steinbusch HWM, et al. Tryptophan breakdown pathway in bipolar mania. J Affect Disord. 2007 Sep;102(1-3):65–72. 5. Drzyzga L, Obuchowicz E, Marcinowska A, Herman ZS. Cytokines in schizophrenia and the effects of antipsychotic drugs. Brain Behav Immun. 2006 Nov;20(6):532–45. 6. Pollmächer T, Haack M, Schuld A. Effects of antipsychotic drugs on cytokine networks. J Psychiatr …. 2000;34(6):369–82. 7. Rybakowski JK. Antiviral and immunomodulatory effect of lithium. Pharmacopsychiatry. 2000 Sep;33(5):159–64. 8. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50. 9. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 10. Benes FM, Matzilevich D, Burke RE, Walsh J. The expression of proapoptosis genes is increased in bipolar disorder, but not in schizophrenia. Mol Psychiatry. 2006 Mar;11(3):241–51. 11. Che Y, Wang J-F, Shao L, Young T. Oxidative damage to RNA but not DNA in the hippocampus of patients with major mental illness. J Psychiatry Neurosci. 2010 Sep;35(5):296–302. 12. Tian SY, Wang J-F, Bezchlibnyk YB, Young LT. Immunoreactivity of 43 kDa growth-associated protein is decreased in post mortem hippocampus of bipolar disorder and schizophrenia. Neurosci Lett. 2007 Jan 10;411(2):123–7. 13. Kato N, Nemoto K, Arino H, Fujikawa K. Influence of peripheral inflammation on growth- associated phosphoprotein (GAP-43) expression in dorsal root ganglia and on nerve recovery after crush injury. Neurosci Res. 2003 Mar;45(3):297–303. 14. Brown ES. Effects of glucocorticoids on mood, memory, and the hippocampus. Treatment and preventive therapy. Ann N Y Acad Sci. 2009 Oct;1179:41–55. 15. Roumier A, Béchade C, Poncer J-C, Smalla K-H, Tomasello E, Vivier E, et al. Impaired synaptic function in the microglial KARAP/DAP12-deficient mouse. J Neurosci. 2004 Dec 15;24(50):11421–8. 16. Roumier A, Pascual O, Béchade C, Wakselman S, Poncer J-C, Réal E, et al. Prenatal activation of microglia induces delayed impairment of glutamatergic synaptic function. PLoS One. 2008 Jan;3(7):e2595. 17. Costello D a, Lyons A, Denieffe S, Browne TC, Cox FF, Lynch M a. Long term potentiation is impaired in membrane glycoprotein CD200-deficient mice: a role for Toll-like receptor activation. J Biol Chem. 2011 Oct 7;286(40):34722–32. 18. Kreisel T, Frank MG, Licht T, Reshef R, Ben-Menachem-Zidon O, Baratta M V, et al. Dynamic microglial alterations underlie stress-induced depressive-like behavior and suppressed neurogenesis. Mol Psychiatry. Nature Publishing Group; 2013 Dec 17;(October):1–11.

136 19. Macchi F, Homberg JR, Calabrese F, Zecchillo C, Racagni G, Riva M a, et al. Altered inflammatory responsiveness in serotonin transporter mutant rats. J Neuroinflammation. Journal of Neuroinflammation; 2013 Jan;10(1):116. 20. Doorduin J, de Vries EFJ, Dierckx RA, Klein HC. PET imaging of the peripheral benzodiazepine receptor: monitoring disease progression and therapy response in neurodegenerative disorders. Curr Pharm Des. 2008 Jan;14(31):3297–315. 21. Choi J, Ifuku M, Noda M, Guilarte TR. Translocator protein (18 kDa)/peripheral benzodiazepine receptor specific ligands induce microglia functions consistent with an activated state. Glia. 2011 Feb;59(2):219–30. 22. Dobos N, de Vries EFJ, Kema IP, Patas K, Prins M, Nijholt IM, et al. The role of indoleamine 2,3-dioxygenase in a mouse model of neuroinflammation-induced depression. J Alzheimers Dis. 2012 Jan;28(4):905–15. 23. Dickens AM, Vainio S, Marjamäki P, Johansson J, Lehtiniemi P, Rokka J, et al. Detection of Microglial Activation in an Acute Model of Neuroinflammation Using PET and Radiotracers 11C-(R)-PK11195 and 18F-GE-180. J Nucl Med. 2014 Mar 10;55(3):466–72. 24. Wiley C a, Lopresti BJ, Becker JT, Boada F, Lopez OL, Mellors J, et al. Positron emission tomography imaging of peripheral benzodiazepine receptor binding in human immunodeficiency virus-infected subjects with and without cognitive impairment. J Neurovirol. 2006 Aug;12(4):262– 71. 25. Grover VPB, Pavese N, Koh S-B, Wylezinska M, Saxby BK, Gerhard a, et al. Cerebral microglial activation in patients with hepatitis C: in vivo evidence of neuroinflammation. J Viral Hepat. 2012 Feb;19(2):e89–96. 26. Garvey LJ, Pavese N, Politis M, Ramlackhansingh A, Brooks DJ, Taylor-Robinson SD, et al. Increased microglia activation in neurologically asymptomatic HIV-infected patients receiving effective ART. AIDS. 2014 Jan 2;28(1):67–72. 27. Banati RB. Visualising microglial activation in vivo. Glia. 2002 Nov;40(2):206–17. 28. van Berckel BN, Bossong MG, Boellaard R, Kloet R, Schuitemaker A, Caspers E, et al. Microglia activation in recent-onset schizophrenia: a quantitative (R)-[11C]PK11195 positron emission tomography study. Biol Psychiatry. 2008 Nov 1;64(9):820–2. 29. Folkersma H, Boellaard R, Yaqub M, Kloet RW, Windhorst AD, Lammertsma A a, et al. Widespread and prolonged increase in (R)-(11)C-PK11195 binding after traumatic brain injury. J Nucl Med. 2011 Aug;52(8):1235–9. 7 30. Doorduin J, de Vries EFJ, Willemsen ATM, de Groot JC, Dierckx R a, Klein HC. Neuroinflammation in schizophrenia-related psychosis: a PET study. J Nucl Med. 2009 Nov;50(11):1801–7. Neuroinflammation in bipolar disorder 31. Kreisl WC, Jenko KJ, Hines CS, Lyoo CH, Corona W, Morse CL, et al. A genetic polymorphism for translocator protein 18 kDa affects both in vitro and in vivo radioligand binding in human brain to this putative biomarker of neuroinflammation. J Cereb Blood Flow Metab. Nature Publishing Group; 2013 Jan;33(1):53–8. 32. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998 Jan;59 Suppl 2:22–33;quiz 34–57. 33. Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck PE, Denicoff KD, et al. The Stanley Foundation Bipolar Treatment Outcome Network. I. Longitudinal methodology. J Affect Disord. 2001 Dec;67(1-3):33–44. 34. Young RC, Biggs JT, Ziegler VE, Meyer D a. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978 Nov 1;133(5):429–35. 35. Rush AJ, Weissenburger J, Giles E, Schlesser MA, Fulton CL. The Inventory Preliminary for Depressive Findings Symptomatology ( IDS ): Depression. 1986;65–87.

137 36. Svarer C, Madsen K, Hasselbalch SG, Pinborg LH, Haugbøl S, Frøkjaer VG, et al. MR-based automatic delineation of volumes of interest in human brain PET images using probability maps. Neuroimage. 2005 Feb 15;24(4):969–79. 37. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med. 1990 Jul;9(7):811–8. 38. Aronica E, Crino PB. Inflammation in epilepsy: clinical observations. Epilepsia. 2011 May;52 Suppl 3:26–32. 39. Marazziti D, Dell’Osso B, Baroni S, Masala I, Nasso E Di, Giannaccini G, et al. Changes in peripheral benzodiazepine receptors in patients with bipolar disorder. Neuropsychiatr Dis Treat. 2005 Jun;1(2):151–4. 40. Savitz J, Frank MB, Victor T, Bebak M, Marino JH, Bellgowan PSF, et al. Inflammation and neurological disease-related genes are differentially expressed in depressed patients with mood disorders and correlate with morphometric and functional imaging abnormalities. Brain Behav Immun. 2013 Jul;31:161–71. 41. Paolicelli RC, Bolasco G, Pagani F, Maggi L, Scianni M, Panzanelli P, et al. Synaptic pruning by microglia is necessary for normal brain development. Science. 2011 Sep 9;333(6048):1456–8. 42. Schafer DP, Lehrman EK, Kautzman AG, Koyama R, Mardinly AR, Yamasaki R, et al. Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron. Elsevier Inc.; 2012 May 24;74(4):691–705. 43. Sierra A, Encinas JM, Deudero JJP, Chancey JH, Enikolopov G, Overstreet-Wadiche LS, et al. Microglia shape adult hippocampal neurogenesis through apoptosis-coupled phagocytosis. Cell Stem Cell. Elsevier Inc.; 2010 Oct 8;7(4):483–95. 44. Deckersbach T, Dougherty DD, Savage C, McMurrich S, Fischman AJ, Nierenberg A, et al. Impaired recruitment of the dorsolateral prefrontal cortex and hippocampus during encoding in bipolar disorder. Biol Psychiatry. 2006 Jan 15;59(2):138–46. 45. Gonul AS, Coburn K, Kula M. Cerebral blood flow, metabolic, receptor, and transporter changes in bipolar disorder: the role of PET and SPECT studies. Int Rev Psychiatry. 2009 Jan;21(4):323–35. 46. Deicken RF, Pegues MP, Anzalone S, Feiwell R, Soher B. Lower concentration of hippocampal N-acetylaspartate in familial bipolar I disorder. Am J Psychiatry. 2003 May;160(5):873–82. 47. Vargas C, López-Jaramillo C, Vieta E. A systematic literature review of resting state network-- functional MRI in bipolar disorder. J Affect Disord. Elsevier; 2013 Sep 25;150(3):727–35. 48. Hannestad J, DellaGioia N, Gallezot J-D, Lim K, Nabulsi N, Esterlis I, et al. The neuroinflammation marker translocator protein is not elevated in individuals with mild-to-moderate depression: a [11C]PBR28 PET study. Brain Behav Immun. Elsevier Inc.; 2013 Oct;33:131–8. 49. Lavisse S, Guillermier M, Hérard A-S, Petit F, Delahaye M, Van Camp N, et al. Reactive astrocytes overexpress TSPO and are detected by TSPO positron emission tomography imaging. J Neurosci. 2012 Aug 8;32(32):10809–18. 50. Hostenbach S, Cambron M, D’haeseleer M, Kooijman R, De Keyser J. Astrocyte loss and astrogliosis in neuroinflammatory disorders. Neurosci Lett. Elsevier Ireland Ltd; 2013 Oct 12;10–2.

138 7 Neuroinflammation in bipolar disorder

139

CHAPTER 8 Volume, metabolites and neuroinflammation of the hippocampus in bipolar disorder A combined magnetic resonance imaging and positron emission tomography study

Bartholomeus C.M. Haarman, Huibert Burger, Janine Doorduin, Remco J. Renken, Anita J. Sibeijn-Kuiper,Jan-Bernard C. Marsman, Erik F.J. de Vries, Jan Cees de Groot, Hemmo A. Drexhage, Richard Mendes, Willem A. Nolen, Rixt F. Riemersma – Van der Lek

Brain, Behavior, and Immunity. 2015 Sep 5;60(1):1–5. Abstract

Background The hippocampus is one of the brain regions that is involved in several pathophysiological theories about bipolar disorder (BD), such as the neuroinflammation theory and the corticolimbic metabolic dysregulation theory. We compared hippocampal volume and hippocampal metabolites in bipolar I disorder (BD-I) patients versus healthy controls (HC) with magnetic resonance imaging (MRI) and spectroscopy (MRS). We post-hoc investigated whether hippocampal volume and hippocampal metabolites were associated with microglial activation and explored if potential illness modifying factors affected these hippocampal measurements and whether these were associated with experienced mood and functioning.

Material and Methods Twenty-two BD-I patients and twenty-four HCs were included in the analyses. All subjects underwent psychiatric interviews as well as an MRI scan, including a T1 scan and PRESS magnetic resonance spectroscopy (MRS). Volumetric analysis was performed with Freesurfer. MRS quantification was performed with LCModel. A subgroup of 14 patients and 11 HCs also underwent a successful [11C]-(R)- PK11195 neuroinflammation positron emission tomography scan.

Results In contrast to our hypothesis, hippocampal volumes were not decreased in patients compared to HC after correcting for individual whole-brain volume variations. We demonstrated a decreased N-acetylaspartate (NAA) + N-acetyl-aspartyl-glutamate (NAAG) concentration in the left hippocampus. In the explorative analyses in the left hippocampus we identified positive associations between microglial activation and the NAA+NAAG concentration, between alcohol use and NAA+NAAG concentration, between microglial activation and the depression score and a negative relation between Cr+PCr concentration and experienced occupational disability. Duration of illness associated positively with volume bilaterally.

Conclusion Compared to HCs, the decreased NAA+NAAG concentration in the left hippocampus of BD-I patients suggests a decreased neuronal integrity in this region. In addition we found a positive relation between microglial activation and neuronal integrity in vivo, corresponding to a differentiated microglial function where some microglia induce apoptosis while others stimulate neurogenesis.

142 Introduction

Bipolar disorder (BD) is a severe mental illness that is characterized by episodic pathologic disturbances in mood: (hypo)manic episodes and depressive episodes, which alternate with euthymic periods, i.e. with normal mood1. Besides the mood symptoms, many patients with BD also show cognitive dysfunctions which may persist during euthymic periods2,3. The lifetime prevalence of BD is about 2%, women being affected as frequently as men4,5. Across the world, BD is sixth among all health conditions in terms of causing disability6 with poor clinical and functional outcome7, increased risk for suicidality 8 and significant societal costs. It has been calculated Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder that in the United States the average cost per case ranged from $11,720 to $624,785, based on the severity of the illness9. In the European countries societal costs for managing BD are considered to be high as well10–12. Although the pathophysiology of BD is complex and its neurobiology remains largely unknown13, several pathophysiological mechanisms have been proposed to explain at least part of the illness. Among these are microglial activation resulting in neuro-inflammation14, corticolimbic metabolic alterations15, and mitochondrial dys- function16. The neuroinflammation theory proposes an aberrant state of microglia in the brain (microglial activation) and of other immune competent cells to be the driving force behind the illness17–19. This theory is supported by findings of higher frequencies of comorbid autoimmune diseases, aberrant cytokine concentrations (e.g. tumor necrosis factor alpha, interleukin 4) and elevated inflammation-related gene expression in circulating monocytes and T cell activation20–23. The corticolimbic theory is based on many studies demonstrating an overall hyperactivation of limbic brain regions in BD patients relative to controls, along with an overall hypoactivation of prefrontal regions24. Supporting the hypothesis that mitochondrial dysfunction plays an important role in BD pathophysiology, several magnetic resonance spectroscopy (MRS) studies demonstrated cellular energy metabolism disturbances25–27 and structural anomalies in mitochondria were observed in cells of patients with BD28. The hippocampus, part of the limbic system, is one of the brain regions that is involved in all these pathophysiological theories29. The limbic system is a combination of in origin different brain structures that are involved in visceral behavioral patterns 8 (related to survival: eating, drinking, sexual activity), emotions, and memory. Some structures, such as the hippocampus and the amygdala, are phylogenetically rather old structures (hence the other name paleomammalian brain), while other parts such as the cingulate gyrus and the anterior cingulate cortex are more recently developed structures30. More specifically, the hippocampus has been implicated in the inhibition of stress responses31 and in the regulation of affective states and emotional behavior32, leading to contextually appropriate emotional responses29. Also, the hippocampus is known to play an important role in encoding new information33.

143 Multiple rodent studies have demonstrated a relationship between a disturbed microglia function in the hippocampus and other pathophysiological changes which are thought to relate to neuropsychiatric disorders17,34–36. One study demonstrated an increased cytokine response to lipopolysaccharide challenge in the hippocampus of SERT mutant rats37. Another rodent study showed stress-induction to induce dynamic microglia changes in the hippocampus alone, which was associated with depressive-like behavior38. In a recently published study using [11C]-(R)-PK11195 positron emission tomography, we demonstrated focal microglial activation in the right hippocampus and non-significant, trend level microglial activation in the left hippocampus of BD patients compared to healthy controls (HCs)39. This radiopharmaceutical binds to the translocator protein (TSPO), a receptor that is upregulated in the mitochondria of activated microglia cells40. [11C]-(R)-PK11195 has been used successfully in models of central inflammation, such as following an injection of the endotoxin lipopolysaccharide in animal models41–43 and investigating infections of the central nervous system in humans44–46. In various psychiatric and neurodegenerative disorders [11C]-(R)-PK11195 PET has proven to be a useful tool for imaging neuroin- flammation47–50. Using another approach, N-acetylaspartate (NAA), a metabolite of which the absolute concentration can be measured using 1H-MRS, has repeatedly (but not always) been found to be decreased in the hippocampus of bipolar patients51. NAA is the second most abundant amino acid in the central nervous system. It is formed in the mitochondria from acetyl co-enzyme A and aspartate and a decreased NAA is thought to represent loss of neuronal integrity. NAA is usually measured together with N-acetyl-aspartyl-glutamate (NAAG) by MRS52. Other metabolites that are reliably measured by MRS in the hippocampus are (phospho)creatine and choline compounds. Phosphocreatine (PCr) functions as an important energy buffer system with creatine (Cr) in the muscles and the brain, transporting energy generated in the mitochondria to the cytoplasm to maintain a constant ATP concentration. Gradual loss of creatine in conjunction with other major metabolites indicates tissue death or major cell death resulting from disease53. Interestingly, gene expression of creatine kinase, the enzyme that facilitates the Cr-PCr conversion, was found decreased in the hippocampus and prefrontal cortex of BD patients in a post mortem analysis54. The GPC+PCh peak of the 1H-MRS detects mainly phosphocholine (PCh) and glycero- phosphocholine (GPC), products of synthesis and breakdown of the cell membrane55. Regarding structural brain alterations a recent review summarized the scientific evidence for a small decrease in hippocampal volume that seemed to be counter acted by lithium treatment56,57. This decrease should be looked upon in the light that the whole-brain volume of BD patients is known to be smaller than that of HCs58. The relation between the various functional pathophysiological mechanisms and structural alterations of the hippocampus remains, however, unclear56.

144 In the current study we compared hippocampal volume and metabolites in bipolar I disorder (BD-I) patients with HCs using magnetic resonance imaging (MRI) and spectroscopy (MRS). We a priori hypothesized hippocampal volume and the NAA metabolite to be decreased in BD patients, compared to HCs. Subsequently, within the BD-I and HC group, we post-hoc investigated whether hippocampal volume and metabolites were associated with microglial activation. Furthermore, we explored if potential illness modifying factors such as duration of illness, medication use, body mass index (BMI), exercise, smoking, number of caffeine consumptions and alcohol use did affect these hippocampal measurements within the BD-I group and whether these hippocampal measurements were associated with Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder experienced mood and functioning.

145 Material and Methods

Participants For the present cross-sectional case-control study we included 22 BD-I patients and 24 HCs demographically similar in age and gender that participated in the MOODIN- FLAME study (http://www.moodinflame.eu). All subjects underwent a MRI-scan and a subgroup of the subjects also received a PET-scan (15 patients, 12 controls)59. In the MOODINFLAME study adult male and female subjects were included who were free of inflammation related symptoms including fever and current or recent infectious or inflammatory disease, uncontrolled systemic disease, uncontrolled metabolic disease or other significant uncontrolled somatic disorders known to affect mood. They did not use somatic medication known to affect mood or the immune system, such as corticosteroids, non-steroid anti-inflammatory drugs and statins. Female candidates who were pregnant or recently gave birth were excluded. Patients and controls did not have a contraindication for MRI scanning. Patients and controls who participated in the PET-scan were free of benzodiazepines at least in the last week prior to the scan. They were also free of anticoagulant use or presence of coagulation disease and did not participate in a prior research study involving radiation less than a year ago. Patients were allowed to continue their regular psychopharmacological treatment. They were neither in a depressed nor (hypo-)manic episode at the time of scanning as indicated by an Inventory of Depressive Symptoms - Clinician Version (IDS-C30) score <22 and a Young Mania Rating Scale (YMRS) score <12, respectively. Patients with any other current primary major psychiatric diagnosis were excluded including: schizophrenia, schizoaffective disorder, anxiety disorder and substance use disorders. HCs did not have any current or lifetime psychiatric diagnosis.

Ethical Considerations The Medical Ethical Review Committee of the University Medical Center Groningen approved the protocol, which was performed in accordance with the Helsinki Decla- ration of 201360. Written informed consent was obtained from all participants.

Assessments All subjects underwent a Mini-International Neuropsychiatric Interview (MINI) to confirm the diagnosis in the patient group and the absence of psychiatric disorders in the HCs61. Clinical features were extracted from the interviews held according to the general MOODINFLAME protocol. This protocol included the Patient Question- naire of the former Stanley Foundation Bipolar Network, the YMRS, the IDS-C30 and a somatic illness questionnaire. The Patient Questionnaire includes separate clinician and patient chapters covering a

146 spectrum of clinical features including vocational, educational and economic status, onset and course of illness, family history, past treatment, cycling and seasonal patterns, medical problems, medications, ability to function and symptomatic status, precipitants of illness (e.g. substance use), treatment adherence and insight into the illness62. In the event of a mismatch of results from the MINI in relation to the Patient Questionnaire, diagnoses were checked with the treating physician. The YMRS is an 63 eleven-item, multiple-choice questionnaire to assess manic symptoms . The IDS-C30 is a thirty-item, multiple-choice questionnaire to assess depressive symptoms of 64 all symptom domains of depression . The YMRS and IDS-C30 were assessed shortly before the scans and used in the relevant analyses. The somatic illness questionnaire Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder is a MOODINFLAME specific checklist exploring all the organ systems for current and lifetime medical symptoms.

Magnetic Resonance Imaging Using a 3T MRI scanner and an eight-channel head coil (3T Intera, Philips, Best, The Netherlands), we acquired anatomical axial T1 weighed MRI images, as well as two single volume point-resolved spectroscopy sequence (PRESS) MRS scans. The T1 images were used in the volumetric analysis and were used to determine the place- ment of the MRS volume of interest (VOI). These images were also used in the tissue type analysis of the MRS VOI and provided the anatomical reference for the normalization of the PET scans to standard brain morphology of all subjects. In the subgroup that underwent both a PET scan and an MRI, these were made no more than 1 week apart and even on the same day in 11 of the 27 subjects (41%).

Volumetric analysis The high resolution 3D T1-weighted images were acquired by 3D ultrafast spoiled gradient echo sequence (SPGR) with the following acquisition parameters: repetition time (TR) = 9.758 ms, echo time (TE) = 4.59 ms, flip angle = 8°, slice thickness = 1.2 mm, single-shot, field of view (FOV) = 220 · 174 mm, matrix = 256 · 256, voxel size=0.859 · 0.859 · 1.200 mm. A total of 130 axial images were collected for each subject, encompassing the whole-brain. PAR-REC files collected from scanning were transferred into NIFTI format data using MRIcroX software (version 1.2, 8 http://www.mccauslandcenter.sc.edu/CRNL/tools/mricro). Cortical reconstruction and volumetric segmentation was performed with the Free- surfer image analysis suite (version 5.3.0, http://surfer.nmr.mgh.harvard.edu) and was used to obtain the volume of the left and right hippocampus (Figure 1), as well as the whole-brain volume (not including the ventricles). The technical details of these procedures are described in prior publications65,66. Freesurfer is known to be a robust tool for automated segmentation of the hippocampus67. The Freesurfer volumetric analyses of the participants all passed a visual quality control check.

147 FIGURE 1 Freesurfer hippocampus volumes

Three-dimensional model of the segmented left and right hippocampus (yellow) of a female patient using Freesurfer and displayed in 3D Slicer (www.slicer.org).

Magnetic resonance spectroscopy All 1H MRS data were acquired using a point-resolved spectroscopy sequence (PRESS) with echo time (TE) = 144 ms, repetition time (TR) = 2000 ms, 128 acquisitions, sample frequency = 2 kHz, 1024 complex data points. Water unsuppressed spectra were also acquired for absolute quantification of metabolites in units of mmol/kg wet weight. Two single volume 15 · 15 · 15 mm (3375 mm3) VOIs were placed on the head of the left and right hippocampus (Figure 2). Due to technical problems 1H MRS scans failed in two healthy controls. The absolute quantification of the spectral metabolites NAA+NAAG, glutamate, glutamine, myo-inositol, Cr+PCr, GPC+PCh, taurine, alanine, aspartate, gamma-amino-butyric acid and glucose, as well as lipid and macromol- ecule resonances, was performed using the Linear Combination Model (LC Model) software68, an operator-independent fitting routine (Figure 3). Only the results of the more reliable metabolites (NAA+NAAG, Cr+PCr and GPC+PCh) were used. Quantifi- cation results in which these metabolites had a Cramer-Rao Lower Bounds of more than 20% were excluded for analysis69, providing 87 reliable MRS results (left hippocampus 22 patients, 21 controls; right hippocampus 22 patients, 22 controls). Tissue-type segmentation was performed on the T1-weighed MRI image using an in- house developed automation script. FMRIB’s Automated Segmentation Tool (FAST) part of FMRIB’s Software Library (FSL 5.0.7, http://www. fmrib.ox.ac.uk/fsl) was used for segmentation of the whole-brain into gray matter (GM), white matter (WM) and

148 FIGURE 2 Freesurfer hippocampus volumes Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder

T1-weighted sagittal, coronal and axial MRI locations for left hippocampal 1H-MRS single voxel acquisitions in a male patient.

FIGURE 3 Examples of MRS spectra

ipolar disorder

rr ealthy control 8

hemia hit ppm

Examples of MRS spectra located in the left hippocampus of a female patient with bipolar I disorder and a female healthy control. GPC+PCh = glycerylphosphorylcholine plus choline, Cr+PCr = creatine plus phosphcreatine, NAA+NAAG = N-acetyl-aspartate plus N-acetylaspartylglutamate.

149 cerebrospinal fluid (CSF) content. Subsequently the VOI position and angulation was selected based on the MRS header information using the Analysis of Functional NeuroImages package (AFNI, version jan 8 2014, http://http://afni.nimh.nih.gov). The GM, WM and CSF content of each MRS VOI was calculated. To correct for the tissue composition in each voxel, the GM fraction of each VOI was included as a co-variate in the statistical analyses.

Positron emission tomography PET procedures have been described extensively previously59,48. In short, PET imaging was performed with the ECAT EXACT HR+ camera (Siemens, Knoxville, Tennessee). Head movement was minimized with a head-restraining adhesive band and a neuroshield was used to minimize the interference of radiation from the subject’s body. A 60-minute emission scan in 3D-mode was performed, starting simultaneously with the intravenous injection of [11C]-(R)-PK11195. During the PET scan arterial blood samples were collected to generate metabolite-corrected plasma and blood input functions. MRI images were co-registered to the sum of all frames of the PET scan, resulting in the most optimal co-registration, using statistical parametric mapping (SPM8; Welcome Trust Center Neuroimaging, University College London, UK). One participant admitted having used a benzodiazepine on the evening prior to the scan after completing inclusion. In another inclusion, technical difficulties were experienced with the automatic blood sampling system prohibiting valid determination of the input function. These two subjects were removed from the subsequent PET-analyses, providing 25 reliable scans (14 patients, 11 healthy controls). Gray matter regions of interest (ROIs) were defined by the co-registered MRI images using a probability map that was based on automatic delineation of ROIs with the PVElab software70. The ROIs were transferred to the dynamic PET images and time-activity curves were calculated. In total 15 ROIs were calculated, including the left and right hippocampus59. The time-activity curves of all ROIs were used for kinetic modeling with software developed in Matlab 7.1 (Mathworks, Natick, Massachusetts). A two-tissue compartment model was used to calculate the k1-k4 with the metabolite corrected plasma and blood curve as an input function, correcting for the individual delay and a free blood volume. The binding potential was defined as k3/k4 and was calculated for each ROI individually.

150 Statistical Analysis Statistical analyses were performed using Stata Statistical Software, release 11 (StataCorp. 2009, College Station, Texas). The differences in demographic data between the groups were investigated with Student’s t-test (age), Pearson’s chi-squared test (gender) and Wilcoxon-Mann-Whit- ney rank-sum test (IDS-C30 score, YMRS score). Differences in volumes, metabolites and binding potential between the BD-I patient and HC groups were analyzed using linear regression models, using covariates to control for individual variation known to influence the dependent variables. In the volumetric analysis the difference in whole brain volume between BD-I patients Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder and HCs was investigated beforehand, correcting for age. Because of a known age- related brain volume reduction71 which is also present in individuals without brain disorder, age was used as a covariate when analyzing the whole-brain volume. In the subsequent volumetric hippocampus analyses, the left hippocampus volume and right hippocampus volume were used as dependent variables. In these analyses the whole-brain volume was used as a covariate thereby taking age-related variations into account. In the MRS analyses, the concentration of GPC+PCh, Cr+PCr and NAA+NAAG were used as dependent variables in separate models for the left and the right hippocampus. Because besides GM fraction age is also known to influence the metabolites in the hippocampus 72 both GM fraction and age were used as covariates in the MRS analyses. In the PET analysis the binding potential of the left hippocampus and the binding potential of the right hippocampus were used as dependent variables. The whole-brain GM binding potential was used as a covariate to normalize for individual global cerebral [11C]-(R)-PK11195 uptake variations, thereby taking known age-related variations into account73. Beforehand inverse square root transformation was applied to the PET binding potentials to meet the normality assumption. The results of the group analyses are presented as means with 95% confidence intervals (CI), except for the PET results, which are presented as back-transformed means with 95% CI. To correct for multiple comparisons in the group analyses false discovery rate (FDR<0.1) correction was applied to the results per scan modality (MRI, MRS, PET), as described by Benjamini-Hochberg74. Subsequently, we performed explorative post-hoc regression analyses within the BD-I 8 group in accordance with the pathophysiological model in figure 4, on both the left and right hippocampus. Lifestyle habits (body mass index (BMI), exercise, smoking, number of caffeine consumptions, alcohol use) and medication use were solely used as independent variables. Hippocampal volume, metabolites and neuroinflammation

(binding potential) were used both as dependent and independent variables. IDS-C30 depression score and experienced occupational disability were solely used as dependent variables. For comparison, the associations within the hippocampal measurements were also analyzed within the HC group and across groups for independent

151 variable · group interaction analysis, taking neuroinflammation (binding potential) as a dependent variable and volume and metabolites as dependent variables in the respective analyses. All post-hoc analyses were performed with linear regression, except for the analyses in which IDS-C30 depression score and experienced occupational disability were involved. Those were executed with ordered logistic regression. Analyses with a specific medication were not thought useful if less than five patients received the treatment. In each regression analysis the same covariates were entered that were used in the corresponding group analysis (whole-brain volume for hippocampus volume; age and GM fraction for metabolites, whole-brain GM binding potential for neuroinflammation), except for the analyses between volume, metabolites and neuroinflammation, where age was not included. The results of the explorative regression analyses are presented as regression coefficients β with 95% CI. The level of significance was defined as 0.05, two-sided. Non-significant trends are mentioned when p<0.1, two-sided.

FIGURE 4 Explorative analyses model

ippoampus pathophsioo

iroia atiation

epression sore eiation

etaoites

ieste haits xperiene oupationa isaiit oume

uration o iness

Post-hoc univariate regression analyses were performed within the BD-I group in accordance with this pathophysiological model, on both the left and right hippocampus. Lifestyle habits (body mass index (BMI), exercise, smoking, number of caffeine consumptions, alcohol use) and medication use were solely used as independent variables. Hippocampal volume, metabolites and microglia activation (binding potential) were used both as dependent and

independent variables. IDS-C30 depression score and experienced occupational disability were solely used as dependent variables. In each regression analysis the same covariates were entered that were used in the corresponding group analysis (whole-brain volume for hippocampus volume; age and gray matter (GM) fraction for metabolites, whole-brain GM binding potential for neuroinflammation), except for the analyses between volume, metabolites and neuroinflammation, where age was not included.

152 Results

Demographics Subject characteristics are displayed in Table 1. Differences between the groups in gender, age, handedness and YMRS-score were not significant. The difference in the mean IDS-C30 score between BD-I patients and HCs was statistically significant (z=-3.595; p=0.0003).

TABLE 1 Characteristics of the subjects Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder

Bipolar disorder Healthy controls Group size 22 24 Gender Male 10 (45%) 12 (50%) Female 12 (55%) 12 (50%) Age (mean (range), yr) 44.5 (24-61) 38.5 (19-67) Right handedness 11 (50%) 17 (70%)

IDS-C30 score (mean (range)) 5.1 (0-14) 1.5 (0-12) YMRS score (mean (range)) 0.2 (0-2) 0 (0-0) Duration of illness (mean (range), yr) 25.3 (1-39) Age at onset (mean (range), yr) 20.2 (12-43) Medication Citalopram 3 (13%) Trazodon 3 (13%) Lithium 13 (59%) Valproate 6 (27%) Carbamazepine 1 (5%) Levetiracetam 1 (5%) Lamotrigine 6 (27%) Quetiapine 6 (27%) L-Thyroxine 4 (18%) Benzodiazepines 2 (9%)

Group analyses Volumetric analyses 8 Whole-brain volume in BD-I patients was significantly decreased with 6.0%, when calculated without correcting for age (respectively 1.10 (CI 1.05 - 1.14) dm3 versus 1.17 (CI 1.12 - 1.23) dm3, t=2.13, p=0.038). However, after correcting for age this difference was not significant anymore (respectively 1.11 (CI 1.07 - 1.16) dm3 versus 1.16 (CI 1.12 - 1.20) dm3, F=2.39, p=0.129). Both left and right hippocampal volumes were significantly decreased in BD-I patients with respectively 7.2% and 7.5%, when compared to HCs without correcting for whole-brain volume (left: respectively 4.14 (CI 3.96 - 4.31) cm3 versus

153 4.46 (CI 4.25-4.66) cm3, t=-2.44, p=0.019; right: respectively 4.21 (CI 4.01-4.42) cm3 versus 4.55 (CI 4.35-4.74) cm3, t=-2.48, p=0.017), both being significant below the FDR threshold). However, for both volumes the difference between the patient and control group was not significant anymore after correction for whole-brain volume (left: respectively 4.23 (CI 4.09 - 4.38) cm3 versus 4.37 (CI 4.23 - 4.51) cm3, t=-1.28, p=0.207; right: respectively 4.31 (CI 4.17 - 4.46) cm3 versus 4.45 (CI 4.31 - 4.60) cm3, t=-1.32, p=0.193; figure 5).

FIGURE 5 Hippocampus volumes

eath ontros ipoar isorer

olume cm V

et ippoampus iht ippoampus

Mean volumes of the left and right hippocampus of healthy controls and bipolar I disorder patients, corrected for whole-brain volume. Statistical analysis on the binding potentials was performed using linear regression models, with the whole-brain volume (not including ventricles) as a covariate. Results are presented as bars (mean) with error stripes (standard error). The differences between the patient and control group were not statistically significant.

Magnetic resonance spectroscopy Analysis of the quantitative MRS results for each VOI (left and right hippocampus) taking the metabolite concentrations as dependent variables, indicated that NAA+NAAG – but not Cr+PCr and GPC+PCh – was significantly decreased below the FDR threshold in the left hippocampus of BD-I patients when compared to the healthy controls. No significant differences in any of the metabolites were observed between patients and HCs in the right hippocampus (Table 2).

154 TABLE 2 Concentration of brain metabolites in bipolar disorder and healthy control subjects

Volume of interest / metabolite Bipolar disorder Healthy controls p Left hippocampus (n=22) (n=21) GPC+PCh 4.35 (4.01-4.70) 4.64 (4.28-4.99) 0.286 Cr+PCr 11.13 (10.42-11.85) 12.30 (11.56-13.02) 0.037 NAA+NAAG 14.55 (13.68-15.41) 16.28 (15.40-17.16) 0.011* Right hippocampus (n=22) (n=22) GPC+PCh 4.44 (4.05-4.84) 4.31 (3.92-4.70) 0.648 Cr+PCr 11.04 (10.06-12.02) 11.24 (10.26-12.22) 0.784 NAA+NAAG 13.18 (12.02-14.50) 14.14 (12.97-15.30) 0.271 Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder

Mean metabolite concentrations (mmol/kg wet weight, CI) in the left and right hippocampus of healthy controls and BD-I patients. Statistical analyses on the metabolite concentrations was performed using separate linear regression models for each metabolite, with age and gray matter fraction as covariates. The results are presented as means with 95% confidence intervals (CI). A significantly decreased NAA+NAAG concentration was observed in the left hippocampus of the patients when compared to the healthy controls after false discovery rate correction (FDR). GPC+PCh = glycerylphosphorylcholine plus choline, Cr+PCr = creatine plus phosphcreatine, NAA+NAAG = N-acetyl-aspartate plus N-acetylaspartylglutamate. * statistically significant below the 0.1 FDR threshold for multiple comparisons.

Positron emission tomography We previously published the results of a general linear model with the left and right hippocampi binding potentials as the dependent variables that demonstrated a significantly increased [11C]-(R)-PK11195 binding potential in the right hippocampus of the patients when compared to the healthy controls (1.66 (CI 1.45 – 1.91) versus 1.33 (CI 1.16 – 1.53), p=0.033). This result was statistically significant below the FDR threshold for multiple comparisons. The difference between the [11C]-(R)-PK11195 binding potential of the left hippocampus of BD-I patients compared to the healthy controls demonstrated a non-significant trend (1.55 (CI 1.30 – 1.90) versus 1.20 (CI 1.00 – 1.46), p=0.071) 59.

Explorative analyses 8 Neuroinflammation, metabolites and volumes When looking at the associations of the measurements within the hippocampus, a significant association was demonstrated between the [11C]-(R)-PK11195 binding potential and the NAA+NAAG concentration (transformed β=-8.77 (CI -17.29 – -.25), p=0.045, figure 6) in the left hippocampus of BD-I patients. A non-significant trend level association was found between the [11C]-(R)-PK11195 binding potential and the GPC+PCh concentration on the same side

155 FIGURE 6 Association between [11C]-(R)-PK11195 binding potential and NAA+NAAG concentration in the left hippocampus of bipolar I disorder patients and healthy controls mmol/kg onentration ipoar isorer

eath ontros

ransorme -(R)- inin potentia

Association between [11C]-(R)-PK11195 binding potential and NAA+NAAG concentration in the left hippocampus of bipolar I disorder patients and healthy controls. Statistical analysis on the associations was performed using separate linear regression models for each subject group. Squares and dots represent the transformed [11C]-(R)-PK11195 binding potential and the NAA+NAAG concentration in an individual BD-I patient or healthy control, respectively. Within the BD-I patients group the solid fit line represents the statistical significant association between both hippocampal measurements. Within the HC group the dashed fit line represents the non-significant trend level association between both hippocampal measurements. Note that the x-axis is reversed to display the natural relation after inverse square root transformation of the [11C]-(R)-PK11195 binding potential. NAA+NAAG = N-acetyl-aspartate plus N-acetylaspartylglutamate.

(transformed β=-3.38 (CI -7.43 – .67), p=0.093). These associations were not sig nificant in the right hippocampus. Associations between [11C]-(R)-PK11195 binding potentials and volumes as well as associations between metabolites and volumes were not significant in BD-I patients. In the HC group that was analyzed for comparison, no significant associations were found between [11C]-(R)-PK11195 binding potentials and volumes and between [ 11C]-(R)-PK11195 binding potentials and metabolites on either side. A non-significant trend level association was demonstrated between the [11C]-(R)-PK11195 binding

156 potential and the NAA+NAAG concentration (transformed β=-4.82 (CI -9.80 – .14), p=0.055, figure 6) of the left hippocampus of HCs. The association between the left hippocampus [11C]-(R)-PK11195 binding potential and the NAA+NAAG concentration was significant across both groups, when adding group as a covariate (transformed β=-7.88 (CI -13.1 – 2.6), p=0.006), with group also being significant (p=0.007). After adding an independent variable · group interaction term, the association, group variable and interaction term were not significant any more. The association between the left hippocampus [11C]-(R)-PK11195 binding potential and the GPC+PCh concentration was also significant across both groups, when adding group as a covariate (transformed β=-3.18 (CI -5.92 – -.45), p=0.025), Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder with group not being significant (p=0.131). Again, after adding an independent variable · group interaction term, the association, group variable and interaction term were not significant any more. The other across groups associations were not significant.

Depression score and occupational disability The depression score associated significantly with the [11C]-(R)-PK11195 binding potential in the left hippocampus (transformed β=-8.39 (CI -16.53 – -.24), p=0.044). A non-significant negative trend was observed in the association between the depression score and the left GPC+PCh concentration (β=-.827 (CI -1.756 – -.24), p=0.081). The depression score was not associated significantly with hippocampal volume on either side. Experienced occupational disability showed a significant negative association with Cr+PCr concentration in the left hippocampus (β=-.513 (CI -1.011 - -.015), p=0.043). The relation between experienced occupational disability and both left and right hippocampal volumes demonstrated non-significant trends (left β=.00272 (CI -.00027 - .0057), p=0.075; right β=.00243 (CI -.00035 - .00521), p=0.087). Associa- tions between work disability and the right hippocampal metabolites, as well as with the [11C]-(R)-PK11195 binding potentials were not significant.

Duration of illness, lifestyle habits, medication Duration of illness was significantly associated with both left and right hippocampal 8 volumes (left β=13.0 (CI .1 – 25.9), p=0.049; right β=18.6 (CI 2.0 – 35.2), p=0.030). Associations between duration of illness and hippocampal metabolites, as well as with the [11C]-(R)-PK11195 binding potentials were not significant. None of the associations between BMI, exercise level, number of caffeine consumptions, smoking and any of the hippocampal measurements were found to be statistical significant. Alcohol use was significantly associated with the NAA+NAAG concentration in the left hippocampus (β=1.22 (CI .14 – 2.30), p=0.029), but no significant association was found with the other hippocampal measurements.

157 Lithium, valproate, lamotrigine, levo-thyroxine and quetiapine were the most frequently used medications (table 1) and the effects of these medications when taken by at least five patients on the hippocampal measurements were explored. A significant association was found between lithium use and the GPC+PCh and Cr+PCr concentrations in the right hippocampus (β=1.10 (CI .00 – 2.19), p=0.050 and β=2.52 (CI .05 – 4.98), p=0.046), while the association with the NAA+NAAG concentration demonstrated a non-significant trend (β=2.96 (CI -0.07 – 5.99), p=0.055). A non-significant trend level association was found between lithium use and right hippocampus microglia activation (transformed β=-.122 (CI -.26 – -.01), p=0.074). In the left hippocampus, non-significant trend level associations were also found between lithium use and the GPC+PCh and Cr+PCr concentrations (β=0.95 (CI -.01 – 1.92), p=0.053 and β=1.73 (CI -.15 – 3.61), p=0.069), as well as a negative non-significant trend level association between valproate use and the NAA+NAAG concentration (β=-1.98 (CI -4.03 – .06), p=0.057). The association between quetiapine use and left hippocampal volume also demonstrated a non-significant trend (β=235 (CI -45 - 556), p=0.095). A visual summary of the left hippocampus explorative analyses results is displayed in figure 7.

FIGURE 7 Summary of the left hippocampus explorative analyses results

ippoampus pathophsioo

iroia atiation

๨ epression sore oho use ๨ etaoites ๨ ๨ aproate ๰ ๰ h ๨ xperiene ithium rr ๨ ๰ oupationa isaiit ๨

uetiapine ๨ oume

๨

uration o iness ๨ ositie assoiation p ๰ eatie assoiation p

Summary of the left hippocampus post-hoc explorative univariate regression analyses within the BD-I group in accordance with the pathophysiological model, displayed in figure 4. Solid lines represent the statistical significant associations with p<0.05 and dashed lines represent trend level associations with p<0.1.

158 Discussion

In the group analyses, we demonstrated decreased concentrations of NAA+NAAG and Cr+PCr in the left hippocampus of BD-I patients as compared to HCs. We were not able to prove decreased hippocampal volumes between these groups after correcting for individual whole-brain volume variations. In the subsequent explorative analyses that were executed in accordance with an a priori analyses model, in the left hippocampus we identified positive associations between microglial activation and the NAA+NAAG concentration, between alcohol use and NAA+NAAG concentration, between microglial activation and the depression score and a negative relation Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder between Cr+PCr concentration and experience occupational disability. Duration of illness was also associated with hippocampal volume bilaterally. The fact that we did not find smaller hippocampal volumes in BD-I patients than in HCs is in agreement with 14 out of 18 studies that also failed to demonstrate a significant difference between BD patients and HCs56. The absence of reduction in hippocampus volume in BD-I patients compared to controls differs from findings in unipolar major depressive disorder as well as in schizophrenia75,76. It has been suggested that the absence of hippocampal volume reduction in BD patients is caused by the use of lithium by BD patients77 as an increase in hippocampus volume has been demonstrated to be associated with lithium use78,79. We were however not able to find an association in hippocampus volume and lithium use, but this may be explained by lack of statistical power. Moreover, at least some of the patients who were not taking lithium at the time of the study had done so in the past. Our observation that the whole-brain volume is smaller in BD-I patients compared to HCs, but that this difference disappears after correction for age is in line with the meta-analysis of brain volume in BD by De Peri58. It is conceivable that these brain changes have occurred already before the first mood episode arose80. Otten and Meeter previously suggested that this may also be the case for the hippocampus56. The present study demonstrated a positive association between both left and right hippocampus volume and duration of illness, which would suggest a tendency to evolve into hippocampal volumes more comparable to healthy individuals with the course of the illness. We demonstrated a decreased NAA+NAAG concentration in the left hippocampus 8 of BD-I patients when compared to HCs. This finding suggests decreased neuronal viability in the left hippocampus of our BD-I patients. It is conspicuous that we found the NAA+NAAG decrease solely in the left hippocampus. The demonstrated associations between lithium use and the hippocampal metabolites that were more obvious in the right hippocampus may point towards an equilibrium restoring effect of lithium. Indeed, lithium is know to increase the concentration of brain metabolites81,82. Furthermore, our findings do not stand-alone. Although several 1H-MRS studies have demonstrated a bilateral decreased NAA concentration in the hippocampus83,84,

159 others have also demonstrated a unilateral increase85, decrease86 or failed to demonstrate any difference bilaterally87,88. In the left hippocampus, microglial activation was associated with both NAA+NAAG and the depression score after correction for variations in individual [11C]-(R)-PK11195 uptake. At first glance, the positive association with NAA+NAAG is counterintui- tive to the neuroinflammation theory, which hypothesizes that microglial activation induces a local inflammatory milieu with inflammatory toxic metabolites and compounds, which have a negative effect on neuronal survival thereby producing mood symptoms17. It could be argued that perhaps the NAA+NAAG increase is caused by an effect of the microglial activation attracting and activating more immune competent cells thereby increasing the NAA+NAAG concentrations directly. Our observations urge for another explanation, which could lie within a differential activation of microglia. As with peripheral macrophages activation could be in an inflammatory sense (M1 macrophages), an anti-inflammatory sense (M2 macrophages), and a regenerating/tissue support sense (M2b macrophages) animal models demon strated that microglia are also involved in tissue regeneration and play an active role in neuronal support, i.e. the development of mature synapses during embryogenesis89, pruning synapses postnatally90, regulating neurogenesis91 and inducing apoptosis17 in the hippocampus as well as in other regions. Unfortunately, this differentiated activation cannot be visualized in vivo with PET imaging. It may well be the case that some microglial cells induce apoptosis, while others are actively facilitating neurogenesis, which would explain an overall decreased NAA+NAAG, but positive relationship with microglial activation, represented by the neuroinflammation related [11C]-(R)-PK11195 binding potential. This would be consistent with an adapted variation of the neuro inflammation theory by Sterz et al.19. They postulate that after the first acute mood episode neuronal injury causes the release of damage-associated molecules that activate the microglia. The activated microglia in turn release both proinflammatory cytokines and neurotrophic factors. These molecules induce modifications of the synaptic environment by synaptic pruning as an adaption attempt to cope with the insult caused by the acute episode. Then, after several episodes, the excessive production of proinflammatory cytokines, exceeding the normal down-regulatory capacity, maintains the microglia in a constantly activated state19. This left side microglial activation - NAA+NAAG association in BD-I patients follows roughly the same positive direction as the trend-level association within the HCs (figure 6). It is conspicuous that within the BD-I patients group the scatter cloud is positioned more in the upper left quadrant, whereas within the HCs the scatter cloud is positioned more in the lower right quadrant, possibly indicating a shift in the microglia-neuronal equilibrium. The fact that the interaction analysis across groups was non-significant suggests an underlying, confounding mechanism steering this possible equilibrium shift.

160 The positive relation between alcohol use and NAA+NAAG contradicts a recent study in which alcohol abuse was associated with a decrease in NAA/Cr in the visual cortex92. However, none of our subjects had an alcohol abuse disorder and it could be postulated that perhaps alcohol use could lead to a NAA+NAAG increase, whereas abuse would lead to a decrease. This is supported by an animal model study demonstrating NAA+NAAG to be increased in the frontal cortex of low dose ethanol exposed rats compared to controls, whereas high dose ethanol exposed rats demonstrated a decrease in NAA+NAAG 93. The left hippocampal creatine concentration associated negatively with experienced occupational disability in the patient group, meaning patients with a lower left hippo- Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder campal creatine are more likely to experience hinder in their occupational activities due to BD. This is in line the mitochondrial dysfunction theory16. The current study has several limitations. First, all patients were naturalistically treated and none of them was ‘medication naive’ possibly leading to medication effects in the observations revealing false phenomenon or concealing true ones. It is known that most mood stabilizing medications, including lithium, anticonvulsants and antipsychotics, as well as several antidepressants (SSRI’s, clomipramine, imipramine, MAO inhibitors) have an effect on the immune system18,21,94–97. In general their effects are thought to be immunosuppressive in nature. Although we demonstrated a positive trend level association with lithium use and right hippocampal microglia activation, it can be argued that in the present study most medications would actually have reduced the effect of the observed neuroinflammation. Also in this regard, an effect of previous benzodiazepine use on the TSPO-receptor cannot be ruled out. Second, the cross-sectional design of the study is suboptimal for the explorative analyses considering cause and effect. Third, although the MRI and PET scans were acquired not more than one week apart, ideally the scanning would take place simultaneously, to avoid noise in the measurements due to pathophysiological changes within the scanning timeframe. Fourth, increased [11C]-(R)-PK11195 binding to the TSPO receptor in the brain is traditionally related to microglia activation17. It is important to note that the TSPO receptor can also be expressed in astrocytes, potentially influencing the [11C]-(R)-PK11195 binding potential signal98. However, because both cells are known to contribute to neuroinflammation99, it can be argued that regard- 8 less of activated microglia cells or astrocytes being responsible for the increased TSPO expression, the increased [11C]-(R)-PK11195 binding most likely represents a neuroinflammatory process either way. Fifth, the study would have taken advantage of a larger sample size thereby increasing its statistical power. Analyses with little statistical power are indicated by wide confidence intervals. This limitation furthermore relates to the problem of the risk of increased type I-errors in the multiple explorative analyses. We chose not to apply alpha or any other correction for multiple testing in these analyses because it would obscure most if not all

161 observations, thereby increasing the risk of type-II-errors100, and we consider these results important direction indicators in the process of future hypotheses forming. In this regard it must be noted that our patients were almost all in the euthymic state, so they were not markedly depressed. It remains uncertain whether the pathophysiological processes would be more pronounced or altered during a depressive or manic episode.

162 Conclusions

In conclusion, the main findings of this study are a decreased NAA+NAAG concentration in the left hippocampus of BD-I disorder patients compared to HCs, suggesting decreased neuronal integrity in this region. In addition we found a positive relation between microglial activation and neuronal integrity in vivo, corresponding to a differentiated microglial function. In our opinion, especially this relation between microglial activation and neuronal viability deserves verification in further neuroimaging studies because it provides an entrance into integrating individual pathophysiological theories of BD. In this regard, Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder there is ample need for pathophysiological oriented studies focusing on integrating multiple imaging and biomarker assays to unravel the pathophysiological processes of BD patients. However, the importance of these pathophysiological theories clearly extends beyond BD and it therefore would be of interest to replicate these findings, not only in BD, but also in other disorders, such as major depressive disorder and schizophrenia.

Acknowledgements

We thank Juliëtte Kalkman for accompanying the patients; Johan Wiegers, Aafke Zeilstra, Remko Koning, Eelco Severs en Paul van Snick for their assistance with the acquisition of PET scans; Judith Streurman for her assistance with the acquisition of MRI images.

163 References

1. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Arlington, VA: American Psychiatric Publishing, Inc; 2013. 2. Arts B, Jabben N, Krabbendam L, van Os J. Meta-analyses of cognitive functioning in euthymic bipolar patients and their first-degree relatives. Psychol Med. 2008 Jun;38(6):771–85. 3. Bora E, Yucel M, Pantelis C. Cognitive endophenotypes of bipolar disorder: a meta-analysis of neuropsychological deficits in euthymic patients and their first-degree relatives. J Affect Disord. 2009 Feb;113(1-2):1–20. 4. Pini S, de Queiroz V, Pagnin D, Pezawas L, Angst J, Cassano GB, et al. Prevalence and burden of bipolar disorders in European countries. Eur Neuropsychopharmacol. 2005 Aug;15(4):425–34. 5. Merikangas KR, Akiskal HS, Angst J, Greenberg PE, Hirschfeld RMA, Petukhova M, et al. Lifetime and 12-month prevalence of bipolar spectrum disorder in the National Comorbidity Survey replication. Arch Gen Psychiatry. 2007 May;64(5):543–52. 6. World Health Organization. The world health report 2001: mental health : new understanding, new hope. World Health. Geneva, Switzerland; 2001. 7. Goodwin FK. Manic-Depressive Illness: Bipolar Disorders and Recurrent Depression. 2nd ed. New York: Oxford University Press; 2007. 8. Baldessarini RJ, Tondo L. Suicide risk and treatments for patients with bipolar disorder. JAMA. 2003 Sep 17;290(11):1517–9. 9. Begley CE, Annegers JF, Swann AC, Lewis C, Coan S, Schnapp WB, et al. The lifetime cost of bipolar disorder in the US: an estimate for new cases in 1998. Pharmacoeconomics. 2001 Jan;19(5 Pt 1):483–95. 10. Hakkaart-Van Roijen L, Hoeijenbos MB, Regeer EJ, Ten Have M, Nolen WA, Veraart CPWM, et al. The societal costs and quality of life of patients suffering from bipolar disorder in the Netherlands. Acta Psychiatr Scand. 2004;110(5):383–92. 11. Young AH, Rigney U, Shaw S, Emmas C, Thompson JM. Annual cost of managing bipolar disorder to the UK healthcare system. J Affect Disord. Elsevier B.V.; 2011;133(3):450–6. 12. Ekman M, Granström O, Omérov S, Jacob J, Landén M. The societal cost of bipolar disorder in Sweden. Soc Psychiatry Psychiatr Epidemiol. 2013;48(10):1601–10. 13. Schulze TG. Genetic research into bipolar disorder: the need for a research framework that integrates sophisticated molecular biology and clinically informed phenotype characterization. Psychiatr Clin North Am. 2010 Mar;33(1):67–82. 14. Smith RS. The macrophage theory of depression. Med Hypotheses. 1991 Aug;35(4):298–306. 15. Mayberg HS. Limbic-cortical dysregulation: a proposed model of depression. J Neuropsychiatry Clin Neurosci. 1997 Jan;9(3):471–81. 16. Kato T, Kato N. Mitochondrial dysfunction in bipolar disorder. Bipolar Disord. 2000 Sep;2(3 Pt 1):180–90. 17. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 18. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50. 19. Stertz L, Magalhães PVS, Kapczinski F. Is bipolar disorder an inflammatory condition? The relevance of microglial activation. Curr Opin Psychiatry. 2013 Jan;26(1):19–26. 20. Vonk R, van der Schot AC, Kahn RS, Nolen W a, Drexhage H a. Is autoimmune thyroiditis part of the genetic vulnerability (or an endophenotype) for bipolar disorder? Biol Psychiatry. 2007 Jul 15;62(2):135–40.

164 21. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 22. Drexhage RC, Hoogenboezem TH, Versnel MA, Berghout A, Nolen WA, Drexhage HA. The activation of monocyte and T cell networks in patients with bipolar disorder. Brain Behav Immun. 2011;25(6):1206–13. 23. Munkholm K, Braüner JV, Kessing LV, Vinberg M. Cytokines in bipolar disorder vs. healthy control subjects: a systematic review and meta-analysis. J Psychiatr Res. Elsevier Ltd; 2013 Sep;47(9):1119–33. 24. Kupferschmidt D a, Zakzanis KK. Toward a functional neuroanatomical signature of bipolar disorder: quantitative evidence from the neuroimaging literature. Psychiatry Res. Elsevier Ireland Ltd; 2011 Aug 30;193(2):71–9. 25. Kato T, Takahashi S, Shioiri T, Inubushi T. Alterations in brain phosphorous metabolism in bipolar Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder disorder detected by in vivo 31P and 7Li magnetic resonance spectroscopy. J Affect Disord. 1993 Jan;27(1):53–9. 26. Frey, Stanley J a, Nery FG, Monkul ES, Nicoletti M a, Chen H-H, et al. Abnormal cellular energy and phospholipid metabolism in the left dorsolateral prefrontal cortex of medication-free individuals with bipolar disorder: an in vivo 1H MRS study. Bipolar Disord. 2007 Jun;9 Suppl 1(1):119–27. 27. Port JD, Unal SS, Mrazek D a, Marcus SM. Metabolic alterations in medication-free patients with bipolar disorder: a 3T CSF-corrected magnetic resonance spectroscopic imaging study. Psychiatry Res. 2008 Feb 28;162(2):113–21. 28. Cataldo AM, McPhie DL, Lange NT, Punzell S, Elmiligy S, Ye NZ, et al. Abnormalities in mitochondrial structure in cells from patients with bipolar disorder. Am J Pathol. American Society for Investigative Pathology; 2010 Aug;177(2):575–85. 29. Frey, Andreazza AC, Nery FG, Martins MR, Quevedo J, Soares JC, et al. The role of hippocampus in the pathophysiology of bipolar disorder. Behav Pharmacol. 2007 Sep;18(5-6):419–30. 30. Haarman BCM, Riemersma - Van der Lek RF, Ruhé HG, Groot JC, Nolen WA, Doorduin J. Bipolar Disorders. In: Dierckx RAJO, Otte A, de Vries EFJ, Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Heidelberg: Springer Berlin Heidelberg; 2014. p. 223–51. 31. Phillips M. Neurobiology of emotion perception I: the neural basis of normal emotion perception. Biol Psychiatry. 2003 Sep 1;54(5):504–14. 32. McEwen BS, Magarinos AM. Stress and hippocampal plasticity: Implications for the pathophysiology of affective disorders. Hum Psychopharmacol. 2001;16(SUPPL. 1). 33. Eichenbaum H. A cortical-hippocampal system for declarative memory. Nat Rev Neurosci. 2000;1(1):41–50. 34. Costello D a, Lyons A, Denieffe S, Browne TC, Cox FF, Lynch M a. Long term potentiation is impaired in membrane glycoprotein CD200-deficient mice: a role for Toll-like receptor activation. J Biol Chem. 2011 Oct 7;286(40):34722–32. 35. Roumier A, Béchade C, Poncer J-C, Smalla K-H, Tomasello E, Vivier E, et al. Impaired synaptic function in the microglial KARAP/DAP12-deficient mouse. J Neurosci. 2004 Dec 15;24(50):11421– 8. 8 36. Roumier A, Pascual O, Béchade C, Wakselman S, Poncer J-C, Réal E, et al. Prenatal activation of microglia induces delayed impairment of glutamatergic synaptic function. PLoS One. 2008 Jan;3(7):e2595. 37. Macchi F, Homberg JR, Calabrese F, Zecchillo C, Racagni G, Riva M a, et al. Altered inflammatory responsiveness in serotonin transporter mutant rats. J Neuroinflammation. Journal of Neuroinflammation; 2013 Jan;10(1):116. 38. Kreisel T, Frank MG, Licht T, Reshef R, Ben-Menachem-Zidon O, Baratta M V, et al. Dynamic microglial alterations underlie stress-induced depressive-like behavior and suppressed neurogenesis. Mol Psychiatry. Nature Publishing Group; 2013 Dec 17;(October):1–11.

165 39. Haarman BCM, Riemersma-Van der Lek RF, de Groot JC, Ruhé HEG, Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder - A [(11)C]-(R)-PK11195 positron emission tomography study. Brain Behav Immun. 2014 Apr 3; 40. Doorduin J, de Vries EFJ, Dierckx RA, Klein HC. PET imaging of the peripheral benzodiazepine receptor: monitoring disease progression and therapy response in neurodegenerative disorders. Curr Pharm Des. 2008 Jan;14(31):3297–315. 41. Choi J, Ifuku M, Noda M, Guilarte TR. Translocator protein (18 kDa)/peripheral benzodiazepine receptor specific ligands induce microglia functions consistent with an activated state. Glia. 2011 Feb;59(2):219–30. 42. Dickens AM, Vainio S, Marjamäki P, Johansson J, Lehtiniemi P, Rokka J, et al. Detection of Microglial Activation in an Acute Model of Neuroinflammation Using PET and Radiotracers 11C-(R)-PK11195 and 18F-GE-180. J Nucl Med. 2014 Mar 10;55(3):466–72. 43. Dobos N, de Vries EFJ, Kema IP, Patas K, Prins M, Nijholt IM, et al. The role of indoleamine 2,3-dioxygenase in a mouse model of neuroinflammation-induced depression. J Alzheimers Dis. 2012 Jan;28(4):905–15. 44. Garvey LJ, Pavese N, Politis M, Ramlackhansingh A, Brooks DJ, Taylor-Robinson SD, et al. Increased microglia activation in neurologically asymptomatic HIV-infected patients receiving effective ART. AIDS. 2014 Jan 2;28(1):67–72. 45. Grover VPB, Pavese N, Koh S-B, Wylezinska M, Saxby BK, Gerhard a, et al. Cerebral microglial activation in patients with hepatitis C: in vivo evidence of neuroinflammation. J Viral Hepat. 2012 Feb;19(2):e89–96. 46. Wiley C a, Lopresti BJ, Becker JT, Boada F, Lopez OL, Mellors J, et al. Positron emission tomography imaging of peripheral benzodiazepine receptor binding in human immunodeficiency virus-infected subjects with and without cognitive impairment. J Neurovirol. 2006 Aug;12(4):262–71. 47. Cagnin A, Gerhard A, Banati RB. In vivo imaging of neuroinflammation. Eur Neuropsychopharmacol. 2002 Dec;12(6):581–6. 48. Doorduin J, de Vries EFJ, Willemsen ATM, de Groot JC, Dierckx R a, Klein HC. Neuroinflammation in schizophrenia-related psychosis: a PET study. J Nucl Med. 2009 Nov;50(11):1801–7. 49. Folkersma H, Boellaard R, Yaqub M, Kloet RW, Windhorst AD, Lammertsma A a, et al. Widespread and prolonged increase in (R)-(11)C-PK11195 binding after traumatic brain injury. J Nucl Med. 2011 Aug;52(8):1235–9. 50. van Berckel BN, Bossong MG, Boellaard R, Kloet R, Schuitemaker A, Caspers E, et al. Microglia activation in recent-onset schizophrenia: a quantitative (R)-[11C]PK11195 positron emission tomography study. Biol Psychiatry. 2008 Nov 1;64(9):820–2. 51. Kraguljac NV, Reid M, White D, Jones R, den Hollander J, Lowman D, et al. Neurometabolites in schizophrenia and bipolar disorder - a systematic review and meta-analysis. Psychiatry Res. 2013;203(2-3):111–25. 52. Tsai G, Coyle JT. N-Acetylaspartate in neuropsychiatric disorders. Progress in Neurobiology. 1995. p. 531–40. 53. Hemmer W, Wallimann T. Functional aspects of creatine kinase in brain. Dev Neurosci. 1993;15(3- 5):249–60. 54. MacDonald ML, Naydenov A, Chu M, Matzilevich D, Konradi C. Decrease in creatine kinase messenger RNA expression in the hippocampus and dorsolateral prefrontal cortex in bipolar disorder. Bipolar Disord. 2006 Jun;8(3):255–64. 55. Moore GJ, Galloway MP. Magnetic resonance spectroscopy: neurochemistry and treatment effects in affective disorders. Psychopharmacol Bull. 2002;36(2):5–23. 56. Otten M, Meeter M. Hippocampal structure and function in individuals with bipolar disorder: A systematic review. J Affect Disord. Elsevier; 2015;174:113–25. 57. Bearden CE, Thompson PM, Dutton R a, Frey BN, Peluso M a M, Nicoletti M, et al. Three- Dimensional Mapping of Hippocampal Anatomy in Unmedicated and Lithium-Treated Patients

166 with Bipolar Disorder. Neuropsychopharmacology. Elsevier B.V.; 2008 May;33(6):1229–38. 58. De Peri L, Crescini A, Deste G, Fusar-Poli P, Sacchetti E, Vita A. Brain Structural Abnormalities at the Onset of Schizophrenia and Bipolar Disorder: A Meta-analysis of Controlled Magnetic Resonance Imaging Studies. Current Pharmaceutical Design. 2012. p. 486–94. 59. Haarman BCM, Riemersma-Van der Lek RF, de Groot JC, Ruhé HG (Eric), Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder – A [11C]-(R)-PK11195 positron emission tomography study. Brain Behav Immun. Elsevier Inc.; 2014 Aug 3;40:219–25. 60. World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310(20)::2191–4. 61. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998 Jan;59 Suppl 2:22–33;quiz 34–57. Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder 62. Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck PE, Denicoff KD, et al. The Stanley Foundation Bipolar Treatment Outcome Network. I. Longitudinal methodology. J Affect Disord. 2001 Dec;67(1-3):33–44. 63. Young RC, Biggs JT, Ziegler VE, Meyer D a. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978 Nov 1;133(5):429–35. 64. Rush AJ, Weissenburger J, Giles E, Schlesser MA, Fulton CL. The Inventory Preliminary for Depressive Findings Symptomatology ( IDS ): Depression. 1986;65–87. 65. Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage. 1999;9(2):179–94. 66. Fischl B, Salat DH, Van Der Kouwe AJW, Makris N, Ségonne F, Quinn BT, et al. Sequence- independent segmentation of magnetic resonance images. In: NeuroImage. 2004. 67. Morey RA, Petty CM, Xu Y, Hayes JP, Wagner HR, Lewis D V, et al. A comparison of automated segmentation and manual tracing for quantifying hippocampal and amygdala volumes. Neuroimage. Elsevier B.V.; 2009 Apr 15;45(3):855–66. 68. Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med. 1993 Dec;30(6):672–9. 69. Cavassila S, Deval S, Huegen C, Van Ormondt D, Graveron-Demilly D. Cramer-Rao bounds: An evaluation tool for quantitation. NMR Biomed. 2001;14(4):278–83. 70. Svarer C, Madsen K, Hasselbalch SG, Pinborg LH, Haugbøl S, Frøkjaer VG, et al. MR-based automatic delineation of volumes of interest in human brain PET images using probability maps. Neuroimage. 2005 Feb 15;24(4):969–79. 71. Raz N, Lindenberger U, Rodrigue KM, Kennedy KM, Head D, Williamson A, et al. Regional brain changes in aging healthy adults: General trends, individual differences and modifiers. Cereb Cortex. 2005;15(11):1676–89. 72. Reyngoudt H, Claeys T, Vlerick L, Verleden S, Acou M, Deblaere K, et al. Age-related differences in metabolites in the posterior cingulate cortex and hippocampus of normal ageing brain: A 1H-MRS study. Eur J Radiol. 2012;81(3). 73. Kumar A, Muzik O, Shandal V, Chugani DC, Chakraborty P, Chugani HT. Evaluation of age-related 8 changes in translocator protein (TSPO) in human brain using 11C-[R]-PK11195 PET. Journal of Neuroinflammation. 2012. p. 232. 74. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med. 1990 Jul;9(7):811–8. 75. Kempton MJ, Salvador Z, Munafò MR, Geddes JR, Simmons A, Frangou S, et al. Structural neuroimaging studies in major depressive disorder. Meta-analysis and comparison with bipolar disorder. Arch Gen Psychiatry. 2011;68(7):675–90. 76. Wright IC, Rabe-Hesketh S, Woodruff PWR, David AS, Murray RM, Bullmore ET. Meta-analysis of regional brain volumes in schizophrenia. Am J Psychiatry. 2000;157(1):16–25.

167 77. Malhi GS, Tanious M, Das P, Coulston CM, Berk M. Potential mechanisms of action of lithium in bipolar disorder: Current understanding. CNS Drugs. 2013. p. 135–53. 78. van Erp TGM, Thompson PM, Kieseppä T, Bearden CE, Marino AC, Hoftman GD, et al. Hippocampal morphology in lithium and non-lithium-treated bipolar I disorder patients, non-bipolar co-twins, and control twins. Hum Brain Mapp. 2012;33(3):501–10. 79. Hajek T, Bauer M, Simhandl C, Rybakowski J, O’Donovan C, Pfennig A, et al. Neuroprotective effect of lithium on hippocampal volumes in bipolar disorder independent of long-term treatment response. Psychol Med. 2013;44(3):1–11. 80. Vita A, De Peri L, Sacchetti E. Gray matter, white matter, brain, and intracranial volumes in first- episode bipolar disorder: a meta-analysis of magnetic resonance imaging studies. Bipolar Disord. 2009 Dec;11(8):807–14. 81. Moore GJ, Bebchuk JM, Parrish JK, Faulk MW, Arfken CL, Strahl-Bevacqua J, et al. Temporal dissociation between lithium-induced changes in frontal lobe myo-inositol and clinical response in manic-depressive illness. Am J Psychiatry. 1999;156(12):1902–8. 82. Silverstone PH, Wu RH, O’Donnell T, Ulrich M, Asghar SJ, Hanstock CC. Chronic treatment with lithium, but not sodium valproate, increases cortical N-acetyl-aspartate concentrations in euthymic bipolar patients. International clinical psychopharmacology. 2003. 83. Deicken RF, Pegues MP, Anzalone S, Feiwell R, Soher B. Lower concentration of hippocampal N-acetylaspartate in familial bipolar I disorder. Am J Psychiatry. 2003 May;160(5):873–82. 84. Bertolino a. Neuronal pathology in the hippocampal area of patients with bipolar disorder: a study with proton magnetic resonance spectroscopic imaging. Biol Psychiatry. 2003 May 15;53(10):906–13. 85. Iosifescu D V., Moore CM, Deckersbach T, Tilley C a., Ostacher MJ, Sachs GS, et al. Galantamine- ER for cognitive dysfunction in bipolar disorder and correlation with hippocampal neuronal viability: A proof-of-concept study. CNS Neurosci Ther. 2009;15(4):309–19. 86. Scherk H, Backens M, Schneider-Axmann T, Kemmer C, Usher J, Reith W, et al. Neurochemical pathology in hippocampus in euthymic patients with bipolar I disorder. Acta Psychiatr Scand. 2008 Apr;117(4):283–8. 87. Colla M, Schubert F, Bubner M, Heidenreich JO, Bajbouj M, Seifert F, et al. Glutamate as a spectroscopic marker of hippocampal structural plasticity is elevated in long-term euthymic bipolar patients on chronic lithium therapy and correlates inversely with diurnal cortisol. Mol Psychiatry. Nature Publishing Group; 2009 Jul;14(7):696–704, 647. 88. Gigante AD, Lafer B, Yatham LN. (1)H-MRS of hippocampus in patients after first manic episode. World J Biol Psychiatry. 2014;15(2):145–54. 89. Paolicelli RC, Bolasco G, Pagani F, Maggi L, Scianni M, Panzanelli P, et al. Synaptic pruning by microglia is necessary for normal brain development. Science. 2011 Sep 9;333(6048):1456–8. 90. Schafer DP, Lehrman EK, Kautzman AG, Koyama R, Mardinly AR, Yamasaki R, et al. Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron. Elsevier Inc.; 2012 May 24;74(4):691–705. 91. Sierra A, Encinas JM, Deudero JJP, Chancey JH, Enikolopov G, Overstreet-Wadiche LS, et al. Microglia shape adult hippocampal neurogenesis through apoptosis-coupled phagocytosis. Cell Stem Cell. Elsevier Inc.; 2010 Oct 8;7(4):483–95. 92. Bagga D, Khushu S, Modi S, Kaur P, Bhattacharya D, Garg ML, et al. Impaired visual information processing in alcohol-dependent subjects: a proton magnetic resonance spectroscopy study of the primary visual cortex. J Stud Alcohol Drugs. United States; 2014 Sep;75(5):817–26. 93. Lee D-W, Nam Y-K, Kim T-K, Kim J-H, Kim S-Y, Min J-W, et al. Dose-dependent influence of short-term intermittent ethanol intoxication on cerebral neurochemical changes in rats detected by ex vivo proton nuclear magnetic resonance spectroscopy. Neuroscience. 2014;262:107–17. 94. Drzyzga L, Obuchowicz E, Marcinowska A, Herman ZS. Cytokines in schizophrenia and the effects of antipsychotic drugs. Brain Behav Immun. 2006 Nov;20(6):532–45.

168 95. Pollmächer T, Haack M, Schuld A. Effects of antipsychotic drugs on cytokine networks. J Psychiatr …. 2000;34(6):369–82. 96. Rybakowski JK. Antiviral and immunomodulatory effect of lithium. Pharmacopsychiatry. 2000 Sep;33(5):159–64. 97. Walker FR. A critical review of the mechanism of action for the selective serotonin reuptake inhibitors: Do these drugs possess anti-inflammatory properties and how relevant is this in the treatment of depression? Neuropharmacology. Elsevier Ltd; 2013;67:304–17. 98. Lavisse S, Guillermier M, Hérard A-S, Petit F, Delahaye M, Van Camp N, et al. Reactive astrocytes overexpress TSPO and are detected by TSPO positron emission tomography imaging. J Neurosci. 2012 Aug 8;32(32):10809–18. 99. Hostenbach S, Cambron M, D’haeseleer M, Kooijman R, De Keyser J. Astrocyte loss and astrogliosis in neuroinflammatory disorders. Neurosci Lett. Elsevier Ireland Ltd; 2013 Oct 12;10–2. 100. Bender R, Lange S. Adjusting for multiple testing - When and how? J Clin Epidemiol. Volume, metabolites and neuroinflammation of the in hippocampus bipolar disorder 2001;54(4):343–9.

169

CHAPTER 9 Diffusion tensor imaging in euthymic bipolar disorder A tract-based spatial statistics study

Bartholomeus C.M. Haarman, Rixt F. Riemersma – Van der Lek, Huibert Burger, Jan Cees de Groot, Hemmo A. Drexhage, Willem A. Nolen, Leonardo Cerliani

Journal of Affective Disorders. 2016 Oct 1;203:281–91. Abstract

Background In the current DTI study we compared euthymic bipolar I disorder (BD-I) patients and healthy controls (HC). We subsequently divided the total patient group into lithium-users and non-lithium-users and estimated differences across the three groups.

Methods Twenty-one euthymic BD-I patients and twenty-two HC participants were included in psychiatric interviews and MRI image acquisition (diffusion-weighted (DW) and T1-weighted scans). Fractional anisotropy (FA), radial, mean and axial diffusivity (RD, MD, AD) were estimated from the DW data, using DTI. These measures were then compared between groups using FSL Tract Based Spatial Statistics (TBSS). Correla- tions with age at onset, number of episodes and depression score were analyzed.

Results A difference in FA, MD, RD and AD between the whole sample of euthymic BD-I patients and healthy controls could not be detected. Amongst others, lithium- using patients demonstrated a higher FA and lower RD when compared to non-lithium-using BD-I patients in the corpus callosum and left anterior corona radiata. Widespread clusters demonstrated negative FA associations and positive RD and MD associations with minor depressive symptoms.

Limitations Patients were naturalistically treated. Although the sample size is comparable to several other DTI studies, a larger sample size would have been beneficial. TBSS and DTI have their own limitations.

Conclusion Our findings support the theory that previously described DTI-based microstructural differences between HC and BD patients could be less pronounced in euthymic BD patients. Differences in FA between patients using and not using lithium suggest a counteracting effect of lithium on white matter microstructural disturbances.

172 Introduction Over the years neuroimaging techniques have demonstrated important neuroanatomical differences between patients diagnosed with bipolar disorder (BD) and healthy controls (HC). Since group differences were reported in several brain regions1–3, recent neuroimaging research began to focus on the network level, to assess to what extent the symptomatology of BD might be related to an atypical functional and anatomical connectivity between sensible brain regions. In the context of anatomical connectivity, technical and methodological developments in magnetic resonance imaging (MRI) allowed to estimate various parameters related to the microstructural properties of the white matter bundles connecting different brain regions, providing additional substance and complexity to BD pathophysiology. Diffusion-weighted (DW) imaging is based on the observation that in white matter tracts, the coherent geometrical orientation of the axons and of the myelin sheet causes the water to diffuse more along the direction of the fibers than perpendicular to it. Water diffusion yields an attenuation of the net spin-echo signal in a white matter voxel. Therefore, by measuring this attenuation along different directions, it is possible to reconstruct a three-dimensional diffusivity profile reflecting the organization of white matter fibers in a voxel. Modelling the diffusivity profile using diffusion tensor imaging (DTI) allows to separately estimate the amount of water diffusion along the fiber tract (axial diffusivity, AD) and perpendicular to it (radial diffusivity, Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion RD), as well as the mean diffusivity (MD) of water in that voxel along all the recorded directions4. Another widely used measure of water diffusion in a white matter tract is fractional anisotropy (FA), which reflects the relationship between axial and radial diffusivity, quantifying how strongly directional is the diffusion of water in a voxel. While these measures are related to each other, their combined examination is particularly informative since they can be differentially modulated by several microstructural features, including axonal density or caliber, myelination, axonal membrane permeability, fiber complexity and orientation5–8. Differences in these microstructural features, and in the corresponding DTI measurement, across subjects can be associated with development, or relate to other stratifications of neurotypical participants. However, these measures can also reflect progressed or ongoing pathophysiological processes affecting white matter microstructure in different categories of neuro logical or neuropsychiatric patients. In BD, results from whole-brain DTI studies suggest widespread white matter abnormalities. Significant widespread decreases in FA have been predominantly reported so far in BD patients compared with healthy controls, with differences en- 9 compassing all major white matter tracts9. These widespread white matter diffusion abnormalities, including fronto-temporal and ventral striatal regions, were marked by differences in diffusivity measures other than FA as well, leading to the conclusion that myelination problems rather than axonal loss is implicated in BD10–12.

173 Despite these advancements, a general consensus is still lacking, since other studies reported either lower, higher or no difference in FA between BD patients and healthy controls9. Inconsistency across findings might reflect differences in data acquisition methods as well as in patient characteristics. Importantly, in several studies9,13–16 the BD sample includes patients in different mood states, which contributes to increase inter-individual variability, and is likely associated with differential activity within and across brain networks, as well as with different patterns of white matter connectivity 17. In addition to mood state heterogeneity, BD patients’ samples may differ for medication status, which could be considered important for interpreting DTI findings. Indeed, lithium seems to act as a promyelinating factor18,19 and the administration of lithium was found to be associated with increased axial connectivity20. In the current study we first compared estimates of white matter microstructure (FA, MD, RD, AD) between euthymic bipolar I disorder (BD-I) patients and HC. Based on previous studies pointing toward myelination problems in BD, we expected to find a widespread decrease in FA in several white matter tracts of BD-I patients compared to HC, associated with reciprocal alterations of other white matter microstructural parameters. Subsequently, we divided the patient group into lithium-users and non-lithium-users and analyzed the estimates of white matter microstructure across these three groups (non-lithium-users, lithium-users, healthy controls). In this analysis, supposing a restoring effect of lithium on myelination, we expected FA to be increased -and consistently a decrease of other white matter microstructural parameters- in lithium-using patients compared to non-lithium-using patients, possibly even attaining healthy control values. Finally, we investigated whether the estimates of white matter microstructure were associated with patient characteristics. As in some of the mentioned studies, we employed the analytical framework of tract-based spatial statistics (TBSS), which focuses on the centers of the major fiber bundles. With respect to whole-brain investigations, this approach increases the statistical power of the analysis and limits the chance of misalignment across subjects, as well as the issue of partial volume contamination 21.

174 Material and Methods

Participants For the present cross-sectional case-control study we included 22 BD-I patients and 24 HC frequency matched on age (in 5-year age groups) and gender that participated in the MOODINFLAME study (http://www.moodinflame.eu). All subjects underwent a MRI-scan and a subgroup of the subjects also received a PET-scan (15 patients, 12 controls)22,23. In the MOODINFLAME study we included adult male and female subjects who were free of inflammation-related symptoms including fever and current or recent infectious or inflammatory disease, uncontrolled systemic disease, uncontrolled metabolic disease or other significant uncontrolled somatic disorders known to affect mood. They did not use somatic medication known to affect mood or the immune system, such as corticosteroids, non-steroid anti-inflammatory drugs and statins. Female candidates who were pregnant or recently gave birth were excluded. Patients and controls did not have a contraindication for MRI scanning. Patients were allowed to continue their regular psychopharmacological treatment. They were euthymic at the time of scanning as indicated by an Inventory of

Depressive Symptoms - Clinician Version (IDS-C30) score <22 and a Young Mania Rating Scale (YMRS) score <12, respectively. Patients with any other current primary Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion major psychiatric diagnosis were excluded including: schizophrenia, schizoaffective disorder, anxiety disorder and substance use disorders. HC did not have any current or lifetime psychiatric diagnosis.

Ethical Considerations The Medical Ethical Review Committee of the University Medical Center Groningen approved the protocol, which was performed in accordance with the Helsinki Declaration of 201324. Written informed consent was obtained from all participants.

Assessments All subjects underwent a Mini-International Neuropsychiatric Interview (MINI) to confirm the diagnosis in the patient group and the absence of psychiatric disorders in the HC25. Clinical features were extracted from the interviews held according to the general MOODINFLAME protocol. This protocol included the Patient Questionnaire of the former Stanley Foundation Bipolar Network, the YMRS, the IDS-C30 and a somatic illness questionnaire26–28. The somatic illness questionnaire is a MOODINFLAME 9 specific checklist exploring all the organ systems for current and lifetime medical symptoms. In the event of a mismatch of results from the MINI in relation to the Patient Questionnaire, diagnoses were checked with the treating physician.

175 Magnetic Resonance Imaging acquisition Diffusion-weighted (DW) data were acquired using a single-shot pulsed gradient spin echo EPI sequence (TR = 8041 ms, TE = 59.98 ms) on a 3-T MR scanner (Intera, Philips Medical Systems, Best, the Netherlands) equipped with an eight-channel head coil. For each subject, 32 DW images were collected along non-collinear directions, with a maximum gradient strength of 40 mT/m and a b-value of 1000 s/mm2. In addition, one non-DW volume (b ≈ 0, referred in the following as a b0 image) was acquired. Each volume consisted of 55 transverse slices (FOV = 240 х 240 mm, voxel size = 2.5 mm isotropic, no gap). The total acquisition time for this sequence was less than 6 minutes. In addition, we acquired also one high-resolution (1 mm isotropic) T1-weighted image for each participant (SPGR, TR = 9758 ms, TE = 4.59 ms, flip angle 8°, FOV = 220 х 220 mm, voxel size 0.859 х 0.859, slice thickness 1.2mm) covering the entire cerebrum. The total acquisition time for this sequence was less than 5 minutes. PAR-REC files collected from scanning were transferred into NIFTI format data using MRIcroX software (version 1.2, http://www.mccauslandcenter.sc.edu/CRNL/tools/ mricro) which also provided the text files containing the diffusion vectors (bvecs) and the b-values (bvals) needed to fit the diffusion tensor to each image voxel. The acquisition of all MRI data was performed according to our protocol for limiting head movement: the participant’s head was restrained with foam pads on either sides, and further contained with a Velcro strap.

Image processing and analysis Data preprocessing and analysis were performed with the FSL FMRIB Software Library v5.0 (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/, FMRIB, Oxford Centre for Functional MRI of the Brain, University of Oxford, Department of Clinical Neurology, John Radcliffe Hospital, Oxford, United Kingdom), a comprehensive library of analysis tools for MRI data29–31.

Preprocessing and diffusion tensor estimation For every participant, and in the native space, DW images were first corrected for motion and eddy-current distortions using the eddy correct utility in FSL. The b0 image was used as the reference image for realignment of all the DW images. The log of this procedure provided information to estimate the mean displacement for each subject across the acquisition. A two-samples t-test failed to reveal a significant difference between the mean displacement between groups (t =-1.51, p=0.13). Sub- sequently, we extracted a brain mask on the b0 image by means of automated skull stripping32 and FSL Fdt was used to perform diffusion tensor estimation on the voxels encompassed by the brain mask33. This yielded an estimation of fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD) and axial diffusivity (AD) for each

176 brain voxel. Quality control was performed by visual inspection of the raw DTI and FA images34,35, performed independently by two researchers (BH, LC). Due to technical problems DW acquisition failed to provide workable images in one patient and two healthy controls, effectively yielding 22 HC DW scans and 21 BD-I DW scans.

Tract based spatial statistics Group differences in white matter measurements from DTI can be assessed for every voxel in the white matter, however they are particularly reliable for the inner section of large white matter tracts, where the effect of partial voluming (i.e. contamination of the diffusion signal by intravoxel grey matter or CSF) can be ruled out, and inaccu- racies in image registration across individuals are reduced with respect to the outer (radial) white matter, the latter being more problematic due to the highly variable gyral morphology. In this perspective, we estimated group differences in white matter microstructural parameters exclusively on the voxel of a white matter ‘skeleton’ approximating the center of the fiber tracts common to all the participants. To estimate the spatial location of this skeleton, we used Tract-Based Spatial Statistics to prepare our FA, MD, RD, AD images. The details of this method are provided in the article of Smith et al.21. In summary, each subject’s FA image is nonlinearly registered to the FMRIB58 FA Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion template and checked for distortions generated by the alignment procedure. A low degrees-of-freedom nonlinear registration is implemented, to prevent altering the subject-specific white matter topology (e.g. by merging two distinct fiber tracts into one). The mean FA image across all participants is calculated and thresholded to a minimum value of 0.2. The thresholded mean FA image defines the sample-specific white matter skeleton. The white matter skeleton is overlaid onto each subject’s registered FA image. The maximum FA value is searched along the normals to the skeleton - and weighted for increasing distance from it - and dragged onto the skeleton for group-level comparison. This step is required to take care of potential residual misalignments from step 1. TBSS has several advantages over whole-brain approaches when it comes to estimate between-group differences in white matter microstructure indices derived from diffusion tensor imaging: no smoothing is required either in the two-stages registration or in the creation of the white matter skeleton, thereby relaxing the requirement of prior hypotheses on the spatial extent of the effect, which is hard to define a priori. In addition, performing the group comparisons only on the white matter skeleton 9 increases the statistical power of the analysis and minimizes the chance that the results are driven by partial volume effects - another potential drawback of spatial smoothing - or confounding morphological differences such as ventricles enlarge- ment or atrophy.

177 Two separate group comparisons were performed. In the first, the mean FA, MD, RD and AD values for each voxel on the skeleton were compared between healthy control subjects (HC) and the whole group of bipolar patients (BD-I). To test a model in which lithium corrects white matter microstructure disturbances, we subsequently split up the BD-I subjects into two subgroups according to the presence or absence of lithium use. To maximize statistical power, we compared voxelwise values of white matter microstructure across the three groups (non-lithium-users, lithium-users, healthy controls). In both cases, parameter estimates for group-level analysis were obtained from a general linear model (GLM). Potential artifacts due to age36,37 or residual motion effects on DWI images were controlled in the group-level analysis, using a procedure similar to that described in Yendiki A. et al.38, that is by adding subject- specific estimates of motion - derived from the eddy_correct procedure - as a covariate of no interest in assessing differences in FA, MD, RD and AD. Within the clusters found by the second GLM model we compared the white matter estimates between the subgroups using a two-independent samples Student’s t-test. Finally, correlation analyses were performed to estimate the correlation between FA,

MD, RD, AD and age at onset, number of depressive and manic episodes, and IDS-C30 score within the patient group. These analyses were done using a GLM including age and head motion as a covariate. Inference was carried out by means of nonparametric permutation testing39,40 implemented in FSL’s randomise software. Five-thousands permutations were calculated for each group comparison, in order to estimate the null distribution of either FA, MD, RD and AD differences. Threshold-Free Cluster Enhancement (TFCE)41 was used to correct for multiple comparisons using the default values provided by the --T2 options of FSL randomize, which are optimized for TBSS analysis. Cluster size and MNI coordinates at signal peak were derived with FSL Cluster and the corresponding white matter tract retrieved from the ICBM-DTI-81 white-matter labels, JHU White-Matter Tractography and Harvard-Oxford cortical structural atlases42–44.

Demographic data analyses Statistical analyses on the demographic and FA data were performed using Stata Statistical Software, release 14 (StataCorp. 2015, College Station, Texas). The differences in demographic and FA data between the groups were investigated with Student’s t-test (age, FA), Pearson’s chi-squared test (gender, medication use) and

Wilcoxon-Mann-Whitney rank-sum test (IDS-C30 score, YMRS score). Pearson’s correlation coefficient was used in the analysis of the association between FA and

IDS-C30 score.

178 Results

Demographics

Subject characteristics are displayed in Table 1. With exception of the IDS-C30 score, there were no significant differences between BD and HC groups nor within the lithium use subgroups. Although within the normal (i.e. euthymic) range (between

0-14), mean IDS-C30 scores of BD-I patients and of both subgroups using and not using lithium were higher than in HCs (z=-3.68; p<0.001; z=2.90, p=0.004; and z=-3.33, p=0.001, respectively).

TABLE 1 Characteristics of the subjects

Healthy controls Bipolar disorder Bipolar disorder Bipolar disorder whole group non-lithium lithium Group size 22 21 9 12 Gender Male 11 (50%) 9 (43%) 5 (56%) 4 (33%) Female 11 (50%) 12 (57%) 4 (44%) 8 (67%) Age (mean (range), yr) 38.2 (19-67) 44.7 (24-61) 41.2 (24-59) 47.4 (36-61) * IDS-C30 score (mean (range)) 1.1 (0-4) 5.0 (0-14) 6.2 (0-14) 3.6 (0-12) YMRS score (mean (range)) 0 (0-0) 0.2 (0-2) 0.2 (0-2) 0.2 (0-2) Duration of illness (mean (range), yr) 25.2(2-39) 23.2 (2-39) 26.7 (12-37) Age at onset (mean (range), yr) 19.5 (12-43) 18.0 (12-30) 20.7 (12-43) Medication Citalopram 3 (14%) 1 (11%) 2 (16%) Trazodon 3 (14%) 1 (11%) 2 (16%)

Lithium 12 (57%) - (0%) 12 (100%) disorder bipolar euthymic in imaging tensor Diffusion Valproate 5 (24%) 3 (33%) 2 (17%) Carbamazepine 1 (5%) 1 (11%) 0 (0%) Lamotrigine 6 (29%) 2 (22%) 4 (33%) Levetiracetam 1 (5%) 1 (11%) 0 (0%) Quetiapine 6 (29%) 4 (44%) 2 (17%) L-Thyroxine 4 (19%) 1 (11%) 3 (25%) Benzodiazepines 1 (5%) 0 (0%) 1 (8%)

Characteristics of the subjects. With exception of the IDS-C30 score, there were no significant differences between BD and HC groups nor within the lithium use subgroups. * statistical significant p<0.05.

Group differences in white matter microstructure In the first model comparing the whole BD-I patient group to the HC group, while controlling for age and head motion, no significant clusters were found between the groups in the FA, MD, RD and AD analyses. In the second model analyzing the estimates of white matter microstructure across non-lithium-using bipolar disorder I (BD-I) patients, lithium-using BD-I patients and 9 healthy controls, while controlling for age and head motion, significant clusters with a positive FA association were found in the genu, body and splenium of the corpus collosum, the left anterior corona radiata, as well as two small peripheral tracts in the frontal orbital cortex (table 2, figure 1). Significant clusters with a negative

179 RD association in this model were found in the corpus callosum, left and right anterior corona radiata and the anterior thalamic radiation (table 2). Within the genu, body, splenium of the corpus collosum and the left anterior corona radiata, FA was found to be significantly lower in non-lithium-using BD-I patients compared to lithium-using BD-I patients, as well as healthy controls. Differences between lithium-using BD-I patients and healthy controls were not significant (figure 1). Significant differences were not found between any of the groups in MD and AD.

TABLE 2 Clusters with significant white matter microstructure estimate associations across non-lithium-using BD-I patients, lithium-using BD-I patients and healthy controls

MNI coordinates Number at signal peak of Minimal Direction of № White matter tract x y z voxels corrected p effect Genu and body of corpus callosum 1 78 156 81 6427 0.020 non-li < li < hc Anterior corona radiate R & L Body and splenium of corpus callosum

2 82 90 89 3993 0.019 non-li < li < hc

FA 3 Anterior corona radiata L 116 155 85 687 0.044 non-li < li < hc 4 White matter of the frontal orbital cortex L** 113 145 56 4 0.050 non-li < li < hc 5 White matter of the frontal orbital cortex L** 112 141 55 3 0.050 non-li < li < hc Genu and body of corpus callosum 1 78 156 81 7624 0.026 hc > li > non-li Anterior corona radiata R & L

RD 2 Body and splenium of corpus callosum 82 91 89 2419 0.038 hc > li > non-li 3 Anterior thalamic radiation* 119 164 92 222 0.049 hc > li > non-li

Clusters with significant white matter microstructure estimate associations across non-lithium-using bipolar disorder I (BD-I) patients, lithium-using BD-I patients and healthy controls. The clusters are ordered on size and presented with the MNI coordinates at signal peak and the corresponding white matter tract retrieved from the ICBM-DTI-81 white- matter labels, JHU White-Matter Tractography (*) and Harvard-Oxford cortical structural (**) atlases42–44. Voxelwise values of white matter microstructure were compared across non-lithium-users, lithium-users and healthy controls in a general linear model, including age and head motion as covariates. Voxelwise levels of significance, corrected for multiple comparisons, were calculated with a standard permutation testing (5000 permutations) by building up the null distribution (across permutation of the input data) of the maximum (across voxels) threshold-free cluster-enhancement (TFCE) scores and then using the 95th percentile of the null distribution to threshold signals at corrected p= 0.05. FA = fractional anisotropy, MD = mean diffusivity, RD = radial diffusivity, hc = healthy control, li = lithium-using BD-I patients, non-li = non-lithium-using BD-I patients, R = right, L = left.

180 FIGURE 1 .6

n.s.

.55 p=0.017 Fractional anisotropy .5

Cluster 1 p<0.001 .45 non-lithium BD lithium BD healthy control Subject group .65 .6 n.s. p=0.003 .55 Fractional anisotropy Cluster 2 .5 p<0.001 .45 non-lithium BD lithium BD healthy control Subject group .5

n.s. .45 p=0.004 .4 Fractional anisotropy p<0.001 Cluster 3 .35

non-lithium BD lithium BD healthy control Subject group Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion .5 .4

n.s. n.s. .3 Fractional anisotropy Cluster 4 .2 p=0.021 .1 non-lithium BD lithium BD healthy control Subject group .5

n.s. .4

.3 n.s. n.s. Fractional anisotropy Cluster 5 .2 .1 non-lithium BD lithium BD healthy control Subject group

Localization and fractional anisotropy plot of clusters demonstrating a significant association across non-lithium-using bipolar disorder I (BD-I) patients, lithium-using BD-I patients and healthy controls. Sagittal, coronal and axial MRI locations of the signal peak (cross hair) of the identified clusters with a significant positive fractional anisotropy (FA) association 9 across non-lithium-using bipolar disorder I (BD-I) patients, lithium-using BD-I patients and healthy controls. The mean FA skeleton has a green color. Voxels with a significant association (threshold corrected p=0.05) have a blue – light blue color. Comparisons between the subgroups were analyzed using Student’s independent samples t-test. A = anterior, P = posterior, S = superior, I = inferior, L = left, R = right, BD = bipolar disorder.

181 Correlation analyses within the patient group Analyses of the white matter microstructural estimates within the BD-I patient group while controlling for age and head motion, demonstrated a negative association between IDS-C30 and FA in 23 clusters, complemented with 2 clusters demonstrating a positive association with MD and 1 large cluster demonstrating a positive association with RD (table 3). The localization and an association plot of the five largest FA clusters are depicted in figure 2. Significant differences were not found between

AD and IDS-C30 and any of the estimates and age at onset, number of depressive or number of manic episodes.

Discussion In the present study we could not demonstrate a difference in white matter microstructure estimates between BD-I patients and healthy controls. Across non-lithium-using BD-I patients, lithium-using BD-I patients and healthy controls a positive FA association with a concomitant negative RD association was found in the corpus collosum and the (left) anterior corona radiata. Within these regions, FA was found to be significantly lower in non-lithium-using BD-I patients compared to lithium-using BD-I patients, as well as healthy controls, but differences between lithium-using BD-I patients and healthy controls were not significant. Within the BD-I patient group, negative associations between minor depressive symptoms and FA, partly overlapping with regions with a positive RD or MD association, were found to be widespread.

Euthymic patients As far as we are aware this is the first DTI study in BD that did not find a difference in DTI-measures between patients and HC. There are several explanations for this. First, as suggested by recent meta-analytic reviews on DTI in BD an effect of publication bias reporting positive but not negative findings cannot be ruled out9,45. However, our negative finding may also be related to the fact that in the present study all patients were euthymic, thereby presenting with less symptoms than patients do during an episode and could be less affected in a pathophysiological sense. Yet, some previous studies did report a difference between euthymic patients and HC, albeit some with a higher FA and others with a lower FA compared to HC 15,46–54. In this regard, it is conspicuous that the studies investigating BD patients while in an episode reported more regions with altered FA than studies investigating euthymic BD patients45, strengthening the case that white matter microstructure disturbances might indeed be more increased or more widespread in BD patients experiencing more mood symptoms than in euthymic patients.

182 TABLE 3 Clusters with a significant association between white matter estimates and

IDS-C30 score in BD-I patients

MNI coordinates at signal peak Number of Minimal Direction of № White matter tract x y z voxels corrected p effect Corpus callosum and 1 144 97 106 18810 0.017 negative Anterior corona radiata R & L Posterior thalamic radiation R and 2 70 43 107 6130 0.029 negative Sagittal stratum R 3 Sagittal stratum L 129 116 34 1706 0.029 negative 4 Cerebral peduncle R & L 102 101 66 1046 0.041 negative 5 Superior longitudinal fasciculus L 136 131 84 570 0.037 negative Inferior longitudinal fasciculus L and 6 106 67 120 195 0.049 negative Anterior thalamic radiation L* 7 Corticospinal tract L* 107 82 131 175 0.048 negative 8 Corticospinal tract L* 103 96 128 147 0.049 negative 9 Inferior longitudinal fasciculus L* 138 107 70 113 0.048 negative 10 Body of corpus callosum 81 145 99 110 0.049 negative

Posterior thalamic radiation R and 11 108 91 80 96 0.049 negative FA Retrolenticular part of internal capsule R 12 Cingulum* 104 73 128 46 0.049 negative 13 Cingulum* 79 69 129 29 0.044 negative 14 Cingulum* 102 75 121 27 0.050 negative Superior and Inferior longitudinal 15 138 98 52 26 0.049 negative fasciculus L 16 Inferior fronto-occipital fasciculus R* 57 84 86 19 0.050 negative 17 Anterior thalamic radiation L* 117 166 90 10 0.050 negative 18 White matter of Precentral Gyrus** 131 129 114 8 0.049 negative 19 Inferior longitudinal fasciculus L* 136 120 57 4 0.050 negative 20 Anterior thalamic radiation L* 122 168 89 3 0.050 negative * 21 Anterior thalamic radiation L 93 113 86 3 0.050 negative disorder bipolar euthymic in imaging tensor Diffusion 22 Anterior thalamic radiation L* 122 169 86 2 0.050 negative Cerebral peduncle R & L Superior cerebellar peduncle R & L 1 94 113 70 2674 0.018 positive Retrolenticular part of internal capsule R MD Fornix 2 Posterior limb of internal capsule R 79 123 68 23 0.046 positive Genu, body and splenium of corpus callosum Posterior thalamic radiation R & L 1 Internal capsule R & L 79 95 97 25445 0.006 positive RD Corona radiata R & L Cingulum Superior longitudinal fasciculus R & L

Clusters with a significant association between fractional anisotropy (FA), mean diffusivity

(MD), radial diffusivity (RD) and inventory of depressive symptoms (IDS-C30) score in bipolar disorder I (BD-I) patients. The clusters are ordered on size and presented with the MNI coordinates at signal peak and the corresponding white matter tract retrieved from the ICBM-DTI-81 white-matter labels, JHU White-Matter Tractography (*) and Harvard- Oxford cortical structural (**) atlases42–44. Voxelwise associations between values of white

matter microstructure and IDS-C30 score were analyzed in a general linear model, including age and head motion as covariates. Voxelwise levels of significance, corrected for multiple comparisons, were calculated with a standard permutation testing (5000 permutations) 9 by building up the null distribution (across permutation of the input data) of the maximum (across voxels) threshold-free cluster-enhancement (TFCE) scores and then using the 95th percentile of the null distribution to threshold signals at corrected p= 0.05. FA = fractional anisotropy, MD = mean diffusivity, RD = radial diffusivity, hc = healthy control, li = lithium- using BD-I patients, non-li = non-lithium-using BD-I patients, R = right, L = left.

183 FIGURE 2

15 r=-0.787 p<0.001 10 5 Cluster 1 0 .4 .45 .5 Fractional anisotropy

15 r=-0.837 p<0.001 10 5 Cluster 2 0

.35 .4 .45 Fractional anisotropy

15 r=-0.855 p<0.001 10 5 Cluster 3 0

.3 .32 .34 .36 .38 .4 Fractional anisotropy

15 r=-0.815 p<0.001 10 5 Cluster 4 0 .35 .4 .45 .5 Fractional anisotropy

15 r=-0.757 p<0.001 10 5 Cluster 5 0 -5 .3 .35 .4 .45 .5 Fractional anisotropy

Localization and fractional anisotropy plot of clusters demonstrating a significant negative

association between fractional anisotropy and IDS-C30 score within BD-I patients. Sagittal, coronal and axial MRI locations of the signal peak (cross hair) of the identified clusters with a significant positive fractional anisotropy (FA) association across

184 Indeed, this theory is supported by our finding of widespread regions with a negative association between FA and minor depressive symptoms, as well as partly overlapping regions with a positive association between RD, MD, and minor depressive symptoms. As far as we know, this correlation has never been described before in BD. Interesting- ly however, in a study investigating euthymic older patients a negative relation between FA and depression score has been reported55 and in another study with women with subclinical depression a positive relation with RD was described56, suggesting that the association between white matter microstructure aberrations and depression score may not be specific for BD. Additionally supporting this theory, Magioncalda et al. recently demonstrated white matter microstructure abnormalities to be more pronounced in groups of BD patients in a depressive or manic mood state, compared to euthymic patients52. Although this, together with the present study merely suggests a stronger association between mood state and white matter microstructure disturbances, it does emphasize the importance of longitudinal DTI studies in BD patients, while cycling through different mood states.

Lithium The major regions that we report to have a decreased FA and increased RD in non-lithium-using patients, compared to lithium-using patients and to healthy con- Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion trols, have all been reported to be implicated in BD DTI studies before9. It is important to mention that while alterations of DTI parameters (FA, MD, RD) account for microstructural differences, it is difficult to further elaborate on precisely what aspect of white matter microstructural integrity is affected in BD. In several of the anatomical locations where group differences were detected, such as the corpus callosum and the anterior corona radiata (see tables 2 and 3), the FA decrease is associated with an increase in perpendicular diffusivity (RD), while parallel diffusivity (AD) was not affected. This profile is usually associated with demyelination or dysmyelination8,12. From our results it seems that the commissural and projection fibers are more affected by lithium use. When affected, these tracts are known to be associated with impaired neurocognitive functioning. White matter leftward asymmetry of the anterior corona radiate is known to relate to the executive control function of attention57, a function that is frequently disturbed in BD. Implication of the corpus callosum is of specific interest, since this structure is thought to be important for adequate cognitive functioning58 and several studies describe cognitive functioning to be significantly associated with white matter integrity of the corpus callosum59–61, 9 whereas in another study lithium-use was found to be associated with lower prevalence of dementia in an elder BD population62. The regions demonstrating a higher FA in lithium-using BD-I patients compared to non-lithium-using BD-I patients is corresponding with a previous study by

185 Benedetti et al. demonstrating lithium-treatment to be associated with increased DTI measures20. In a recent study Gildengers et al. demonstrated a longer duration of lithium treatment to be associated with a higher integrity of the white matter micro- structure63. Building on this evidence and our results it may seem obvious to assume that lithium has a counteracting effect on white matter microstructural disturbances, perhaps being one of its therapeutic mechanisms64,65. From this point of view, in a prognostic way it may be promising that in the present study the FA of lithium-using patients in the determined clusters was indistinguishable from healthy controls. Besides being acknowledged for its efficacy in treating BD, lithium is also known for its multiple points of action66. Summarizing, lithium acts on a several second messenger systems that underpin its regulatory effects on neurotransmission and its neuroprotective properties. It modulates neurotransmission by moderating adenyl cyclase and cyclic adenosine monophosphate (cAMP) fluctuations, and by limiting myoinositol (mI), protein kinase C (PKC) and myristoylated alanine-rich c kinase substrate (MARCKS). Over time, these constraints modify gene transcription within the cells and yield long-lasting mood stabilization. Additionally, lithium reduces the oxidative burden caused by mood episodes and protects against apoptosis by promoting neuroprotective pathways such as Akt and facilitating the actions of neuroprotective proteins such as brain derived neurotrophic factor (BDNF) and bcl-2. Furthermore, it inhibits glycogen synthase kinase 3 beta (GSK-3β), which besides regulating glycogen synthesis, is also involved in gene transcription, synaptic plasticity, cell structure and resilience. Finally, lithium also inhibits pro-apoptotic proteins such as p53 and processes such as autophagy67. In white matter microstructure, lithium may primarily assert its effects by modera- tion of GSK-3β in glial cells, either directly or via the Akt pathway68. In mice, injection with GSK-3β-inhibitors was found to regulate oligodendrocyte differentiation and enhance myelination69. Compared to other species, the human brain is exceptionally myelinated70. This extensive myelination has imposed exceptionally high metabolic demands and is associated with vulnerabilities that are thought to make humans highly susceptible to brain disorders throughout their lifespan19,71. The present study, in addition to the above mentioned DTI studies by Benedetti et al. and Gildengers et al., provides in vivo support for the theory that lithium in BD has a counteracting effect on white matter microstructural disturbances by enhancing myelination of neuronal tracts via interaction with GSK-3β in glial cells. However, so far no longitudinal studies exist looking at DTI measures in patients both before and during or both during and after lithium treatment.

186 Limitations Several limitations have to be mentioned. First, all patients were naturalistically treated and none of them was ‘medication naïve’ possibly leading to other medication effects in the observations revealing false phenomena or obfuscating true ones (positive or false negative findings). In a review of structural and function imaging studies in BD, Hafeman et al. observed a limited effect of medication on functional fMRI and DTI study. Although rigorous analyses were typically not possible, they noted that in general the effects were normalizing in nature72. Second, the study would have taken advantage of a larger sample size thereby increasing its statistical power. However, the sample size of the whole BD-I patient group and the HC group is comparable to several other DTI studies in BD (ranging from 18 to 42 subjects per group) that did report differences between the groups45. Since the lithium-using BD and non-lithium-using BD subgroups are even smaller, we analyzed the effect of lithium across the three subgroups, instead of comparing it solely between lithium-using patients and non-lithium-using patients. Furthermore, a post-hoc power calculation of the study of Benedetti reveals that with the sample size used in that study, a 25% to 50% probability of type II error is still present (see supplemental information), which could also explain our null finding (see supplemental information). Third, although TBSS represents the state-of-the-art technique for voxel-wise DTI Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion analyses, it has its own limitations: (1) the mentioned advantages of considering only the center of the fiber tract limit the possibility to detect group differences in tract thickness; (2) the white matter skeleton does not include many white matter structures in the most radial (i.e. outer) white matter, closest to the cortex, where short- range connections subserved by U-shaped fibers are located. (3) precise hypotheses about the expected spatial location of the group difference would allow to perform a group comparison based on FA values averaged over a region of interest, thereby benefitting from the corresponding decrease of noise in the subject- and group-level parameter estimate. Finally, although in our study the more larger clusters display a FA decrease associated with an increase in perpendicular diffusivity, typical for demyelination or dysmyelin- ation8,12, the anatomical interpretation becomes more difficult for regions where only a difference in FA was detected. In this regions changes in FA can be due to a number of factors, including axonal diameter, axonal packing, or even to non-pathological situations (e.g. complex fiber organization such as crossing, kissing or fanning fibers) where the diffusion tensor is not able to adequately model the diffusion-weighted 9 signal73. The latter in particular represents a limitation of all in vivo studies on microstructural integrity in BD published so far. A better anatomical characterization of the group differences that can be estimated using diffusion-weighted imaging therefore awaits future methodological developments, which can better characterize the properties of the diffusion signal with respect to their anatomical origin.

187 Conclusions In conclusion, in the present study we did not find a difference in DTI-based white matter microstructural estimates beyond an alpha level of 0.05, for the comparison between euthymic BD-I patients and HC. Our null finding might suggest that white matter microstructure disturbances could be less outspoken in BD patients not experiencing mood symptoms, supported by our finding of widespread clusters demonstrating negative FA associations and positive RD and MD associations with minor depressive symptoms. Alternatively, our negative finding, as well as previous positive findings, could be explained by issues related to statistical power. Differences in FA and RD across non-lithium-using BD-I patients, lithium-using BD-I patients and healthy controls could be demonstrated in the the genu, body, splenium of the corpus collosum and the anterior corona radiata, suggesting a counteracting effect of lithium on myelination problems that are associated with BD. In our opinion, there is need for longitudinal DTI studies of BD patients that would allow for the investigation of white matter microstructural disturbances over the course of the illness and across mood episodes, as well as the evaluation of the effect of different treatments including lithium, deepening our understanding of this obvious important intermediate mechanism in BD pathophysiology.

Acknowledgements We thank Juliëtte Kalkman for accompanying the patients; Anita Sibeijn-Kuiper and Judith Streurman for their assistance with the acquisition of MRI images and Annemiek Oldenziel for het assistance with the TBSS analysis.

188 References

1. Kupferschmidt D a, Zakzanis KK. Toward a functional neuroanatomical signature of bipolar disorder: quantitative evidence from the neuroimaging literature. Psychiatry Res. Elsevier Ireland Ltd; 2011 Aug 30;193(2):71–9. 2. Gonul AS, Coburn K, Kula M. Cerebral blood flow, metabolic, receptor, and transporter changes in bipolar disorder: the role of PET and SPECT studies. Int Rev Psychiatry. 2009 Jan;21(4):323–35. 3. Haarman BCM, Riemersma - Van der Lek RF, Ruhé HG, Groot JC, Nolen WA, Doorduin J. Bipolar Disorders. In: Dierckx RAJO, Otte A, de Vries EFJ, Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Heidelberg: Springer Berlin Heidelberg; 2014. p. 223–51. 4. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J. 1994;66(1):259–67. 5. Beaulieu C. The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed. 2002;15(7-8):435–55. 6. Jones DK, Knösche TR, Turner R. White matter integrity, fiber count, and other fallacies: the do’s and don’ts of diffusion MRI. Neuroimage. 2013 Jun;73:239–54. 7. Lebel C, Benner T, Beaulieu C. Six is enough? Comparison of diffusion parameters measured using six or more diffusion-encoding gradient directions with deterministic tractography. Magn Reson Med. 2012 Aug;68(2):474–83. 8. Benedetti F, Giacosa C, Radaelli D, Poletti S, Pozzi E, Dallaspezia S, et al. Widespread changes of white matter microstructure in obsessive-compulsive disorder: Effect of drug status. Eur Neuropsychopharmacol. 2013;23(7):581–93. 9. Nortje G, Stein DJ, Radua J, Mataix-Cols D, Horn N. Systematic review and voxel-based meta- analysis of diffusion tensor imaging studies in bipolar disorder. J Affect Disord. Elsevier; 2013 Sep 5;150(2):192–200. 10. Harsan LA, Poulet P, Guignard B, Steibel J, Parizel N, de Sousa PL, et al. Brain dysmyelination and disorder bipolar euthymic in imaging tensor Diffusion recovery assessment by noninvasive in vivo diffusion tensor magnetic resonance imaging. J Neurosci Res. 2006 Feb 15;83(3):392–402. 11. Travers BG, Adluru N, Ennis C, Tromp DPM, Destiche D, Doran S, et al. Diffusion tensor imaging in autism spectrum disorder: a review. Autism Res. 2012 Oct;5(5):289–313. 12. Song S-K, Sun S-W, Ramsbottom MJ, Chang C, Russell J, Cross AH. Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water. Neuroimage. 2002 Nov;17(3):1429–36. 13. Versace A, Almeida JRC, Hassel S, Walsh ND, Novelli M, Klein CR, et al. Elevated left and reduced right orbitomedial prefrontal fractional anisotropy in adults with bipolar disorder revealed by tract-based spatial statistics. Arch Gen Psychiatry. 2008 Sep;65(9):1041–52. 14. Zanetti M V, Jackowski MP, Versace A, Almeida JRC, Hassel S, Duran FLS, et al. State-dependent microstructural white matter changes in bipolar I depression. Eur Arch Psychiatry Clin Neurosci. 2009 Sep;259(6):316–28. 15. Sussmann JE, Lymer GKS, McKirdy J, Moorhead TWJ, Muñoz Maniega S, Job D, et al. White matter abnormalities in bipolar disorder and schizophrenia detected using diffusion tensor magnetic resonance imaging. Bipolar Disord. 2009 Feb;11(1):11–8. 16. Liu J-X, Chen Y-S, Hsieh J-C, Su T-P, Yeh T-C, Chen L-F. Differences in white matter abnormalities between bipolar I and II disorders. J Affect Disord. Elsevier B.V.; 2010 Dec;127(1-3):309–15. 17. Bellani M, Boschello F, Delvecchio G, Dusi N, Altamura CA, Ruggeri M, et al. DTI and Myelin 9 Plasticity in Bipolar Disorder: Integrating Neuroimaging and Neuropathological Findings. Front Psychiatry. 2016;7(March):1–10.

189 18. Makoukji J, Belle M, Meffre D, Stassart R, Grenier J, Ghjacumu G. Lithium enhances remyelination of peripheral nerves. PNAS. 2012;2–7. 19. Bartzokis G. Neuroglialpharmacology: Myelination as a shared mechanism of action of psychotropic treatments. Neuropharmacology. Elsevier Ltd; 2012;62(7):2137–53. 20. Benedetti F, Bollettini I, Barberi I, Radaelli D, Poletti S, Locatelli C, et al. Lithium and GSK3-β Promoter Gene Variants Influence White Matter Microstructure in Bipolar Disorder. Neuropsychopharmacology. Nature Publishing Group; 2012;38(2):313–27. 21. Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE, et al. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage. 2006 Jul 15;31(4):1487–505. 22. Haarman BCM, Riemersma-Van der Lek RF, de Groot JC, Ruhé HG (Eric), Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder – A [11C]-(R)-PK11195 positron emission tomography study. Brain Behav Immun. Elsevier Inc.; 2014 Aug 3;40:219–25. 23. Haarman BCM, Burger H, Doorduin J, Renken RJ, Sibeijn-Kuiper AJ, Marsman J-BC, et al. Volume, metabolites and neuroinflammation of the hippocampus in bipolar disorder - A combined magnetic resonance imaging and positron emission tomography study. Brain Behav Immun. Elsevier Inc.; 2015 Sep 5;60(1):1–5. 24. World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310(20)::2191–4. 25. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998 Jan;59 Suppl 2:22–33;quiz 34–57. 26. Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck PE, Denicoff KD, et al. The Stanley Foundation Bipolar Treatment Outcome Network. I. Longitudinal methodology. J Affect Disord. 2001 Dec;67(1-3):33–44. 27. Rush AJ, Weissenburger J, Giles E, Schlesser MA, Fulton CL. The Inventory Preliminary for Depressive Findings Symptomatology ( IDS ): Depression. 1986;65–87. 28. Young RC, Biggs JT, Ziegler VE, Meyer D a. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry. 1978 Nov 1;133(5):429–35. 29. Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ, Johansen-Berg H, et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage. 2004;23 Suppl 1:S208–19. 30. Woolrich MW, Jbabdi S, Patenaude B, Chappell M, Makni S, Behrens T, et al. Bayesian analysis of neuroimaging data in FSL. Neuroimage. 2009;45(1 Suppl):S173–86. 31. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. FSL. Neuroimage. 2012 Aug 15;62(2):782–90. 32. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp. 2002 Nov;17(3):143–55. 33. Behrens TEJ, Woolrich MW, Jenkinson M, Johansen-Berg H, Nunes RG, Clare S, et al. Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magn Reson Med. 2003;50(5):1077–88. 34. Tournier J-D, Mori S, Leemans A. Diffusion tensor imaging and beyond. Magn Reson Med. 2011 Jun;65(6):1532–56. 35. Soares JM, Marques P, Alves V, Sousa N. A hitchhiker’s guide to diffusion tensor imaging. Front Neurosci. 2013;7(March):1–14. 36. Madden DJ, Spaniol J, Whiting WL, Bucur B, Provenzale JM, Cabeza R, et al. Adult age differences in the functional neuroanatomy of visual attention: a combined fMRI and DTI study. Neurobiol Aging. 2007 Mar;28(3):459–76. 37. Lebel C, Gee M, Camicioli R, Wieler M, Martin W, Beaulieu C. Diffusion tensor imaging of white matter tract evolution over the lifespan. Neuroimage. 2012 Mar;60(1):340–52.

190 38. Yendiki A, Koldewyn K, Kakunoori S, Kanwisher N, Fischl B. Spurious group differences due to head motion in a diffusion MRI study. Neuroimage. 2013 Nov 21;88C:79–90. 39. Winkler AM, Ridgway GR, Webster MA, Smith SM, Nichols TE. Permutation inference for the general linear model. Neuroimage. 2014 May 15;92:381–97. 40. Nichols TE, Holmes AP. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum Brain Mapp. 2002;15(1):1–25. 41. Smith SM, Nichols TE. Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage. 2009 Jan 1;44(1):83–98. 42. Wakana S, Caprihan A, Panzenboeck MM, Fallon JH, Perry M, Gollub RL, et al. Reproducibility of quantitative tractography methods applied to cerebral white matter. Neuroimage. 2007 Jul 1;36(3):630–44. 43. Hua K, Zhang J, Wakana S, Jiang H, Li X, Reich DS, et al. Tract probability maps in stereotaxic spaces: analyses of white matter anatomy and tract-specific quantification. Neuroimage. 2008 Jan 1;39(1):336–47. 44. Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006 Jul 1;31(3):968–80. 45. Vederine F-E, Wessa M, Leboyer M, Houenou J. A meta-analysis of whole-brain diffusion tensor imaging studies in bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry. Elsevier B.V.; 2011 May 23; 46. Barysheva M, Jahanshad N, Foland-Ross L, Altshuler LL, Thompson PM. White matter microstructural abnormalities in bipolar disorder: A whole brain diffusion tensor imaging study. NeuroImage Clin. The Authors; 2013;2(1):558–68. 47. Chaddock CA, Barker GJ, Marshall N, Schulze K, Hall MH, Fern A, et al. White matter microstructural impairments and genetic liability to familial bipolar I disorder. Br J Psychiatry. 2009 Jun;194(6):527–34. Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion 48. Haller S, Xekardaki A, Delaloye C, Canuto A, Lövblad KO, Gold G, et al. Combined analysis of grey matter voxel-based morphometry and white matter tract-based spatial statistics in late-life bipolar disorder. J Psychiatry Neurosci. 2011 Nov;36(6):391–401. 49. Mahon K, Burdick KE, Wu J, Ardekani BA, Szeszko PR. Relationship between suicidality and impulsivity in bipolar I disorder: a diffusion tensor imaging study. Bipolar Disord. 2012 Feb;14(1):80–9. 50. Wessa M, Houenou J, Leboyer M, Chanraud S, Poupon C, Martinot J-L, et al. Microstructural white matter changes in euthymic bipolar patients: a whole-brain diffusion tensor imaging study. Bipolar Disord. 2009 Aug;11(5):504–14. 51. Scholz V, Houenou J, Kollmann B, Duclap D, Poupon C, Wessa M. Dysfunctional decision-making related to white matter alterations in bipolar I disorder. J Affect Disord. 2016 Apr;194:72–9. 52. Magioncalda P, Martino M, Conio B, Piaggio N, Teodorescu R, Escelsior A, et al. Patterns of microstructural white matter abnormalities and their impact on cognitive dysfunction in the various phases of type I bipolar disorder. J Affect Disord. 2016 Mar 15;193:39–50. 53. Oertel-Knöchel V, Reinke B, Alves G, Jurcoane A, Wenzler S, Prvulovic D, et al. Frontal white matter alterations are associated with executive cognitive function in euthymic bipolar patients. J Affect Disord. 2014 Feb;155:223–33. 54. Canales-Rodríguez EJ, Pomarol-Clotet E, Radua J, Sarró S, Alonso-Lana S, Del Mar Bonnín C, et al. Structural abnormalities in bipolar euthymia: a multicontrast molecular diffusion imaging study. Biol Psychiatry. 2014 Aug 1;76(3):239–48. 9 55. Lamar M, Charlton RA, Morris RG, Markus HS. The impact of subcortical white matter disease on mood in euthymic older adults: a diffusion tensor imaging study. Am J Geriatr Psychiatry. 2010 Jul;18(7):634–42.

191 56. Hayakawa YK, Sasaki H, Takao H, Mori H, Hayashi N, Kunimatsu A, et al. Structural brain abnormalities in women with subclinical depression, as revealed by voxel-based morphometry and diffusion tensor imaging. J Affect Disord. 2013 Jan 25;144(3):263–8. 57. Yin X, Han Y, Ge H, Xu W, Huang R, Zhang D, et al. Inferior frontal white matter asymmetry correlates with executive control of attention. Hum Brain Mapp. 2013;34(4):796–813. 58. Hinkley LBN, Marco EJ, Findlay AM, Honma S, Jeremy RJ, Strominger Z, et al. The role of corpus callosum development in functional connectivity and cognitive processing. PLoS One. 2012;7(8). 59. Poletti S, Bollettini I, Mazza E, Locatelli C, Radaelli D, Vai B, et al. Cognitive performances associate with measures of white matter integrity in bipolar disorder. J Affect Disord. 2015 Mar 15;174:342–52. 60. Ajilore O, Vizueta N, Walshaw P, Zhan L, Leow A, Altshuler LL. Connectome signatures of neurocognitive abnormalities in euthymic bipolar I disorder. J Psychiatr Res. Elsevier Ltd; 2015;68:37–44. 61. Bauer IE, Ouyang A, Mwangi B, Sanches M, Zunta-Soares GB, Keefe RSE, et al. Reduced white matter integrity and verbal fluency impairment in young adults with bipolar disorder: a diffusion tensor imaging study. J Psychiatr Res. 2015 Mar;62:115–22. 62. Nunes P V, Forlenza O V, Gattaz WF. Lithium and risk for Alzheimer’s disease in elderly patients with bipolar disorder. Br J Psychiatry. 2007 Apr;190(step 5):359–60. 63. Gildengers AG, Butters MA, Aizenstein HJ, Marron MM, Emanuel J, Anderson SJ, et al. Longer lithium exposure is associated with better white matter integrity in older adults with bipolar disorder. Bipolar Disord. 2015;17(3):248–56. 64. Lenox RH, Hahn CG. Overview of the mechanism of action of lithium in the brain: fifty-year update. J Clin Psychiatry. 2000 Jan;61 Suppl 9(suppl 9):5–15. 65. Benedetti F, Bollettini I. Recent findings on the role of white matter pathology in bipolar disorder. Harv Rev Psychiatry. 2014;22(6):338–41. 66. Grunze H, Vieta E, Goodwin GM, Bowden C, Licht RW, Möller H-J, et al. The World Federation of Societies of Biological Psychiatry (WFSBP) guidelines for the biological treatment of bipolar disorders: update 2012 on the long-term treatment of bipolar disorder. World J Biol Psychiatry. 2013;14(3):154–219. 67. Malhi GS, Tanious M, Das P, Coulston CM, Berk M. Potential mechanisms of action of lithium in bipolar disorder: Current understanding. CNS Drugs. 2013. p. 135–53. 68. Marlinge E, Bellivier F, Houenou J. White matter alterations in bipolar disorder: Potential for drug discovery and development. Bipolar Disord. 2014;16(2):97–112. 69. Azim K, Butt AM. GSK3 negatively regulates oligodendrocyte differentiation and myelination in vivo. Glia. 2011;59(4):540–53. 70. Smaers JB, Steele J, Case CR, Cowper A, Amunts K, Zilles K. Primate prefrontal cortex evolution: Human brains are the extreme of a lateralized ape trend. Brain Behav Evol. 2011;77(2):67–78. 71. Haroutunian V, Katsel P, Roussos P, Davis KL, Altshuler LL, Bartzokis G. Myelination, oligodendrocytes, and serious mental illness. Glia. 2014;62(11):1856–77. 72. Hafeman DM, Chang KD, Garrett AS, Sanders EM, Phillips ML. Effects of medication on neuroimaging findings in bipolar disorder: an updated review. Bipolar Disord. 2012 Jun;14(4):375–410. 73. Jbabdi S, Behrens TEJ, Smith SM. Crossing fibres in tract-based spatial statistics. Neuroimage. 2010;49(1):249–56.

192 Diffusion tensor imaging in euthymic bipolar disorder bipolar euthymic in imaging tensor Diffusion

193

CHAPTER 10 Summary and general discussion

Bartholomeus C.M. Haarman

Adapted from Haarman BCM, Riemersma-Van der Lek RF, Burger H, Drexhage HA, Nolen WA. The dysregulated brain: consequences of spatial and temporal brain complexity for bipolar disorder pathophysiology and diagnosis. Bipolar Disord [Internet]. 2016 Dec 20. Preface

After describing the characteristics of bipolar disorder (BD), its burden of disease on afflicted persons and society, and the diagnostic and therapeutic struggles faced, this thesis started with a description of the main existent pathophysiological theories on BD. We explained that, based on the available knowledge, the immune system may play a key role in the pathophysiology of BD. Thus, this thesis aimed to clarify the role of the immune system in the pathophysiology of BD via several different approaches. In this discussion we will put our findings into perspective. We start by summarizing our main findings. Then, we consider some methodological strengths and limitations. Subsequently, we place the findings within the perspective of other recent neuroscientific developments and indicate the implications for clinical practice. And finally, we suggest starting-points for future research in this field and end with some concluding remarks.

Summary of main findings

This thesis includes two parts: the peripheral immune system and the neuroimmune system.

Part 1: Peripheral immune system The first part of the thesis centered around the function of the peripheral immune system in BD, focusing on monocyte pro-inflammatory gene-expression and C-reactive protein (CRP), using bio assay techniques.

Association between monocyte gene-expression and clinical features Initially, we presented our study on the associations between an extensive set of clinical features and quantitative PCR (qPCR) measured monocyte gene expression in BD (chapter 2). Our a-priori hypothesis that lifetime psychotic features would be associated with the pro-inflammatory monocyte gene expression could not be confirmed. However, based on our newly developed feature-expression heat map method (see also chapter 3) we visualized the following interesting findings in patients with: a possible relation between pro-inflammatory gene expression and manic symptomatology, a differential immune activation related to an earlier age at onset, an increased immune system dysregulation during the course of the disorder, and support for the concept of an immune suppressive action of some of the mood regulating medications. In chapter 3 we described the newly developed feature-expression heat map method: a combined presentation of effect size and statistical significance in a graphical

196 method, added to the ordering of the variables based on the effect-ordered data display principle. To visualize the associations of two sets of variables, adapted heat maps are drawn, displaying in the columns the preceding variable set, and in the rows the subsequent variable set. Each are ordered to facilitate the visual identification of meaningful clusters of association later on. An underlying cluster analysis tree may be added to one or both of the axes. In the feature-expression heat maps the associations between preceding and subsequent variables are represented by circles, visualizing the measure of effect size and the statistical significance of the analyses. This combination aids in the visual recognition of association patterns in complex systems, e.g. pathophysiological models.

Monocyte gene-expression: state or trait? To investigate whether the qPCR measured monocyte pro-inflammatory gene- expression is more related to mood state or a marker for disease (trait) we performed the next study, in which we presented the results of the bipolar cohort of the MOODINFLAME study1 (chapter 4). We demonstrated an elevated pro-inflammatory monocyte gene-expression in patients with experiencing a mood episode, as compared to both healthy controls (HC) and euthymic patients with. Furthermore, we found patients with experiencing a mood episode to have an increased inflammatory gene expression compared to when they were euthymic. This indicates that inflammatory gene expression in BD is related to the mood state, rather than being a trait marker.

Does CRP predict outcome in clinical practice? Subsequently, we examined whether higher CRP levels predicted a worse BD outcome in a clinical setting, defined as a shorter time to relapse or a longer time to recover, depending on the mood state at baseline (chapter 5). We found no statistically significant association between CRP and a more unfavorable BD prognosis, sug- Summary andSummary general discussion gesting that the application of CRP as a practical biomarker to predict outcome in a naturalistic outpatient care setting is not as straightforward as it may seem. In a first cross-sectional analysis, we could not distinguish a sub-group of patients with with an elevated baseline CRP level based on affective state. In the longitudinal analysis, no statistically significant association was found between higher CRP values and relapsing in either euthymic or non-euthymic patients, as well as when comparing them.

197 Part 2: Neuroimmune system In the second part of this thesis we investigated the function of the neuroimmune system, focusing on microglia activation, using positron emission tomography (PET), magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS), and diffusion tensor imaging (DTI) as neuroimaging techniques.

Previous PET/SPECT studies We first presented a literature review of the previous PET / single-photon emission computed tomography (SPECT) research efforts on BD (chapter 6). The earliest PET/ SPECT studies, mainly focusing on metabolism and blood flow, looked at various aspects of the metabolism based disease model in which prefrontal cortex (PFC) hypoactivity is accompanied by limbic hyperactivity. However, in its comprehensive form this model is probably not precise enough to account for most of the specific mood and cognitive disease features. Molecular imaging demonstrated the importance of serotonin transporter alterations in parts of the limbic system in BD and underscored the role of dopamine and cholinergic neurotransmission. We observed that most molecular imaging studies in BD have unique designs, extending our knowledge of the pathophysiological mechanisms, but also complicating comparisons between studies.

Microglial activation in the hippocampus We subsequently performed a neuroinflammation PET study in BD (chapter 7) and demonstrated a statistically significant increased binding potential of [11C]-(R)- PK11195 in the right hippocampus and a similar but trend level increased binding potential in the left hippocampus of bipolar I disorder (BD-I) patients as compared to healthy controls, indicative of microglial activation.

Associations between volume, metabolites and microglial activation Next, we investigated the relations between volume, metabolites and microglial activation of the hippocampus in a contemporaneously executed PET/MRI study (chapter 8). Using MRS, we demonstrated a decreased concentration of N-acetylaspartate (NAA) + N-acetyl-aspartyl-glutamate (NAAG) in the left hippocampus of BD-I patients as compared to HC. Using volumetric MRI, we were not able to prove decreased hippocampal volumes between these groups after correcting for individual whole-brain volume variations. In the subsequent explorative analyses that were executed in accordance with an a-priori analysis model, we identified a positive association between microglial activation and the NAA+NAAG concentration in the left hippocampus, indicating a positive relation between microglial activation and neuronal integrity in vivo. In these analyses, we furthermore found positive associations between alcohol use and NAA+NAAG concentration, and between microglial activation and the depression

198 score, and a negative relation between the creatine (Cr) + phosphocreatine (PCr) concentration and experienced occupational disability. Duration of illness was also bilaterally associated with hippocampal volume.

White matter microstructure disturbances and lithium usage Finally, using DTI we investigated white matter microstructure in BD-I and HC, and differences related to lithium usage among patients (chapter 9). In this study we could not demonstrate a difference in fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD) and axial diffusivity (AD) between BD-I patients and HC. Our null finding might suggest that white matter microstructure disturbances are less outspoken in patients with not experiencing mood symptoms. Lithium-using patients, when compared to non-lithium-using BD-I patients, amongst others demonstrated a higher FA and lower RD in the corpus callosum and left anterior corona radiata. Widespread clusters demonstrated negative FA associations and positive RD and MD associations with minor depressive symptoms. Differences in FA between patients using and not using lithium suggest a counteracting effect of lithium on white matter microstructural disturbances.

Methodological considerations: limitations and strengths

Each chapter of the thesis includes a discussion of the strengths and limitations of the specific issue addressed. Here we will highlight the most prominent limitations and strengths of the thesis as a whole.

Multiple testing and two-step analysis approach Since pathophysiologically oriented research into psychiatric illness is intrinsically Summary andSummary general discussion complex, often multiple variables are incorporated into the analysis models in an attempt to clarify the mechanisms involved. This can lead to an increased probability of false positive findings (type I error) in statistical analysis. This problem is especially important in confirmatory studies, where in the analyses a priori postulated hypo theses are either preserved or rejected2. Because the conclusions in a confirmatory study are assumed to be robust, the consequences of a false positive finding are generally more far-reaching than those of a false negative finding (type II error). In contrast, exploratory studies have a distinct, hypotheses-generating goal. There- fore, the consequences of type I and II errors weigh differently in exploratory studies than in confirmatory studies; in exploratory studies it is the false negative findings that pose the greatest problem. In other words, it is more of a problem to miss clues for a novel hypothesis. Even if this means one needs to accept that there is a chance 10

199 of developing hypotheses based on findings that are not actually true. This problem is dealt with in the scientific process whereby exploratory generated hypotheses must be evaluated by confirmatory studies, and with the reservation that the results of exploratory studies should be reported as providing only preliminary information. In the context of an exploratory study, one may argue that results should not be presented statistically as positive or negative findings but rather should be presented only as estimates with a confidence interval. Although the following viewpoint is not unanimously held, multiple sources state that multiplicity adjustment is not required in exploratory analyses (and in fact would increase the risk of type-II errors), provided that these results are reported as providing preliminary, indicatory information on relationships2–4. We reckon hypotheses-generating exploratory studies to have their own merits next to hypothesis-driven confirmatory studies. We therefore constructed several of our studies (chapters 2, 7, 8) in two parts: a hypothesis-driven group analysis part with confirmatory objectives and an exploratory part with results that do not claim to provide rigorous evidence, but have an exclusively hypothesis-generating goal. When dealing with the problem of multiplicity of data in our studies, we followed several somewhat different strategies. In most studies we applied correction for the false discovery rate (FDR), as described by Benjamini and Hochberg5,6 (chapters 2, 7, 8). The FDR method has several advantages over the Bonferonni method, which is regarded as being more conservative2,7. In chapter 4 we followed a different approach because the expression of the individual pro-inflammatory genes is not really independent; we have here instead measurements of a coherent pro-inflammatory cell function. To obtain a single measure of pro-inflammatory gene activation in monocytes the expression of multiple genes was reduced to a gene score, leav- ing further adjustments unnecessary. In chapter 5 we used simple measurements and applied no adjustment for multiple testing. The complex volumetric and DTI neuroimaging techniques in chapters 8 and 9 used specific methods to deal with multiplicity, based on cluster finding8,9.

Study design issues As with most pathophysiological research in humans, the studies in this thesis are limited by several design issues.

Cross-sectional case-control designs Cross-sectional case-control designs, as applied in most of the studies in this thesis, have the advantage that they are relatively simple to organize. They are, however, suboptimal for analyses of cause and effect and therefore in the scientific process should be followed by a prospective cohort study10. The distinctive feature of a prospective cohort study is that at the time of collecting baseline exposure information,

200 none of the subjects have developed any of the outcomes of interest and subjects are followed longitudinally. After a period of time one can investigate if and when they become diseased and whether their exposure status changes outcomes. In this way the prospective cohort study data can be used to answer many questions about the associations between “risk factors” and disease outcomes.

Sample size Another complicating factor is that the studies in this thesis are intrinsically limited in patient sample size because of careful ethical and economic considerations. Unargu- ably, the research would have profited from a larger sample size, thereby increasing its statistical power, while enabling comparisons between subgroups. This limitation is also related to the problem of the risk of increased type I-errors in multiple analyses, as described above.

Naturalistic design Naturalistic designs in which patients are treated regularly and no interventions are carried out by the investigators, have the advantage that in general they are easier to perform and are less burdensome for patients with this serious psychiatric disorder. However, they do not take into account the possible confounding effect of concomitant medication use. In the studies of this thesis, all patients were treated naturalistically and none of them was ‘medication naive’; this may have led to medi cation effects in the observations, revealing false findings or concealing true ones. For instance, and very relevant, it is known that most mood stabilizing medications, including lithium, anticonvulsants and antipsychotics, as well as several antidepressants (SSRI’s, clomipramine, imipramine, MAO inhibitors) have an effect on the immune system11–16. In general, their effects are thought to be immunosuppressive in nature. Thus, it can be argued that in the present study most medications would actually have reduced the effect of the observations or may even have concealed Summary andSummary general discussion them. In this light it must be noted, contrary to the communis opinio, studies with medication-naïve patients and patients in whom medication was stopped (washout studies) are also not without their flaws. The obvious advantage of studies with medication-naïve patients is the exclusion of these medication effects. The question arises, however, in how far the selection of these patients who are able to function without medication, interferes with the investigated mechanism (i.e. affecting the internal validity) and limits the generalizability (i.e. affecting the external validity). In washout studies one could argue that the withdrawal effects interfere with the investigated mechanism. 10

201 Pieces of the immunological puzzle The studies in this thesis focus, among other things, on monocyte gene-expression, CRP and TSPO receptor presentation, which are select parts of the complex immune system. Therefore, generalized statements should be considered in that regard, e.g. the gene-expression studies investigated only inflammatory gene expression of monocytes, which make up around 2-8% of the total white blood cell population. The original selection design of these gene-expression studies was based on the study of Padmos et al.13, who found these specific signature genes, possibly ruling out other important genes. In the PET studies, increased [11C]-(R)-PK11195 binding to the TSPO receptor in the brain is traditionally related to microglia activation17. It is important to note that the TSPO receptor can also be expressed in astrocytes, potentially influencing the [11C]-(R)-PK11195 binding potential signal18. However, because both cells are known to contribute to neuroinflammation19, it can be argued that regardless of whether activated microglia cells or astrocytes are responsible for the increased TSPO expression, the increased [11C]-(R)-PK11195 binding most likely represents a neuroinflammatory process.

Innovative techniques Despite these limitations, we do consider the studies in this thesis to have several key and innovative strengths and they provide important findings to guide further hypothesis-forming and serve as a starting-point for future studies. Our study of the associations between clinical features and monocyte gene expression was the first study to investigate the associations between psychotic, manic, and depressive symptoms and gene expression in such an extensive fashion and can be regarded as a next step in the converging approach between immunology and psychopathology. Especially, the feature-expression heat map method which we developed proved to be a useful graphical instrument to visually explore associations in complex biological systems where one-way direction is assumed; this method has also been adopted by other research groups20,21. Furthermore, using novel, innovative techniques we were able for the first time to investigate microglia activation in vivo, thereby providing direct evidence for the neuroinflammation theory, which has been reviewed extensively in the literature, but thus far solely based on studies using indirect measurements22–33. Moreover, we were also able to investigate associations between microglia activation, metabolites and volume of the hippocampus. Finally, we were one of the first research groups to focus on the effect of lithium on white matter microstructure in BD.

202 Main findings in perspective

The aim of this thesis was to clarify the role of the immune system in the pathophysiology of BD via several different approaches. In this section we will place those findings within the perspective of other recent neuroscientific developments and suggest starting-points for future research.

Biomarkers or dysfunctional processes: bankruptcy of the biomarker approach? The current psychiatric diagnostic systems are descriptive taxonomies based on classifications of phenomenology (Diagnostic and Statistical Manual, DSM34; International Statistical Classification of Diseases and Related Health Problems, ICD35). Psychiatric disorders thus traditionally lack the biological foundation that is an essential part of medical disease. Since the 1960s, in attempts to extend the diagnostic and treatment options for psychiatric patients, biological psychiatric research has sought incessantly to support the diagnostic system with robust psychopathophysiological models. Examples of such models for BD are described in chapter 1. Unfortunately, all these research efforts have yet to yield the promised clinical significance: clinical biomarker tests are still not available and novel pharmaceutical treatments appear only sparsely36–38. Back in 2009 the MOODINFLAME project started off with the development of biomarkers for mood disorders as one of its important goals1. Before the start of this project, Padmos et al. defined a promising monocyte pro-inflammatory gene- expression signature that was able to discriminate patients with BD from HC13. The study population in the Padmos study consisted of a mixed patient sample with euthymic, depressed and manic patients. In this study the signature was also found to be present in the offspring of patients with bipolar disorder, thereby fulfilling some of the criteria of an endophenotype. Endophenotypes are, by definition, more related Summary andSummary general discussion to the underlying genotype than to the ultimate phenotype. Endophenotypes should be consistently associated with the illness and represent persistent “trait” rather than episodic or “state” features. By definition, they also should be found at a higher rate in high-risk individuals, such as non-affected first-degree family members, than in the general population39. Further elaborating on this, endophenotypes in psychiatry could function as diagnostic biomarkers. Biomarkers are conventionally defined as characteristics that can be objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention40. Thus in general, biomarkers may be used as an aid to diagnose a disease, or to predict or follow treatment response. In medicine as a whole, biomarkers may include hormones, proteins and genetic markers at the levels of DNA and RNA, but also 10

203 structural and functional alterations that can be visualized with imaging techniques38. In the process of further validating the monocyte pro-inflammatory gene-expression signature as a biomarker, the first problems arose with regard to the specificity of the test. Drexhage et al. demonstrated that the gene-expression was also increased in schizophrenia (SZ), albeit a bit differently41. In a follow-up analysis we found indications that some of the gene-expression was associated with manic symptomatology (chapter 2). When analyzing the MOODINFLAME gene-expression results, already having abandoned the biomarker hypothesis, we found monocyte pro-inflammatory gene-expression not significantly different in euthymic MOODINFLAME BD patients, compared to HC (chapter 4)42. Instead, increase of the pro-inflammatory gene- expression was more associated with being in a mood episode (i.e. ‘state’). Further- more, the use of CRP, which is considered to be another promising candidate as a biomarker for psychiatric disorders43, for predicting prognosis in clinical practice also proved to be not as straightforward as it may have seemed (chapter 5). The process of initially promising biomarker candidates for psychiatric disorders that subsequently fail in validation steps is more a rule than an exception36,44. Although at first glance this may seem problematic and unfortunate, it may not be an entirely bad thing. Replication is not only important when findings are reproduced, it is equally vital when results are nuanced or not reproduced. In other words, while innovation points out possible new paths, replication points out likely paths, and progress relies on both44. Undeservedly, performing replication studies is underappreciated. In the scientific enterprise individual scientists are often prompted to prioritize novelty findings from explorative studies, even when not yet confirmed, over findings from replication studies44. Grant applications focusing on pioneering work instead of replication studies are known to have a higher chance of getting accepted45. And a related problem is getting the findings from replication studies published. This is one of the reasons behind publication bias: journal reviewers and editors may dismiss a new test of a published idea as uninspired44. Besides these problems, it has been argued that the development of neurobiological markers for mood disorders has been impeded by several factors: complexity of the brain (in both normal physiology and pathophysiology), difficulties with access to brain tissue, and relatively poor permeability of the brain to investigational neuroimaging ligands because of the blood-brain-barrier44,4642. Difficulties with regard to the blood-brain-barrier and with access to cerebral tissue are inherent to the biological nature of the structure and are deemed irresolvable. The notion that the brain is an extremely complex structure can, however, provide a way to understand why biomarker development for our current psychiatric disorder classifications is problematic. The dysfunctional processes in psychiatric disorders are not limited to one physiological level, e.g. the cellular level or neuronal circuit level. To the contrary, these dysfunc-

204 tional processes are involved in and influence all the physiological levels of the central nervous system, i.e. the genetic, molecular, cellular, neuronal circuits and phenomenological levels. This multilevel physiology of the brain is called spatial complexity47. An example of this multilevel pathophysiology for BD can be found in figure 1, emphasizing that for a complete understanding of the pathophysiology of BD, its neurobiology must be addressed at different physiological levels48. A problem with many of the simpler versions of the pathophysiological models described in chapter 1 is that they remain locked within one physiological level. In addition, the manifestations of the illness on each physiological level that are investigated as potential biomarkers are often dichotomized into either affected or not-affected, corresponding to a medical model for disease. Yet in the CNS functional

FIGURE 1 Multiple pathophysiological levels in bipolar disorder

Behavior Cognitive Affective Sensorymotor

• Synaptic connectivity Environmental Factors gonadal steroids) • Neuroplasticity ModifyingImprinting Genes Systems • Cytoskeletal Critical remodeling neuronal circuitry (including stressors, sleep deprivation, • Cell growth/survival

Neurotransmitter & neuropeptide s

• PKC & MARCKS Cellular • GSK-3 & substrates • CREB & BDNF • ERK MAP kinases Proteome • G proteins • G protein–coupled receptor kinases • Bcl-2 family of proteins • Neuronal cytoskeleton

• Transcription factors andSummary general discussion Molecular • mRNA stability Transcriptome • Nuclear import/export Susceptibility genes Protective genes

For a complete understanding of the pathophysiology of bipolar disorder, its neurobiology must be addressed at different physiological levels (i.e., molecular, cellular, systems, and behavioral). Bcl-2 = B-cell leukemia/lymphoma; BDNF = brain-derived neurotrophic factor; CREB= cAMP response element binding protein; ERK = extracellular receptor- coupled kinase; GSK-3 = glycogen synthase kinase-3; MAP kinase = mitogen-activated protein kinase; MARCKS = myristoylated alanine-rich C kinase substrate; PKC = protein kinase C; proteome = the population of cellular protein species and their expression level; transcriptome = the population of cellular messenger RNA species and their expression level. (Source: Manji and Lenox48, reprinted with permission) 10

205 pathophysiology, the illness manifestations are probably much more gradual due to the multiple complex interactions47, and such dichotomizations should be considered inadequate or false. As an example with regard to neuroinflammation, i.e. the activated, inflamed state of the immune system as a cause of psychiatric illness in general and BD specifically, this concept can be considered an oversimplification of a complex system in which in fact various (stimulating and inhibiting) aspects are dysregulated. Based on the spatial physiological complexity of the brain alone, one could argue that endeavors to find pure diagnostic biomarkers for the current psychiatric disorders will probably prove to be unfruitful. However, this does not necessarily mean that observing alterations in bioassays, thereby demonstrating the activity of a dysfunctional state, would not be helpful in predicting or following treatment response. Thus, bioassays would not replace clinical diagnosis, but support and supplement it in the process of treatment decision-making36,49.

Multidimensional psychopathology: Crossing the Kraepelinian dichotomy The false dichotomy problem in our present pathophysiological models is not limited to biological observations. Tracing back to the medical disease model, psychiatric disorders as diagnostic entities are also approached dichotomously in our current diagnostic systems. In other words: persons are either ill, and are called patients, or they are not. This oversimplification has some overlap with the Descartian dualism and the state of neuropathology in the late nineteenth century, when disorders were divided into those involving functional or organic pathology50. However, a vast amount of neuroscientific evidence has demonstrated the functional - organic dualism to be false51,52. One could even go as far as argue that all psychiatry is biological psychiatry. Or even more extreme: the most effective therapies should be those grounded in biology. In reality, psychiatric patients present with diverse psychological symptoms (phenomenology) that vary in form and severity, between patients and in the course of time. Historically based on observations, clinicians have recognized proto typic patterns in these symptom presentations, which have evolved into diagnostic classifications that are ordered into categories, i.e. the categorical diagnostic approach. Although improved over time, inter-operator validity is still a matter of debate 53. In addition, these diagnostic classifications still largely ignore the unique role of individual symptoms and, consequently, potentially important information is lost54. Besides oversimplified stratification of the symptoms, information about the course of the disorder is also only sparsely utilized in recent versions of the DSM (DSM-III, DSM-IV en DSM-5) and ICD (ICD-9 and ICD-10), respectively.

206 Starting more than a decade ago, researchers proposed a more dimensional diagnostic approach36,55,56, linking psychiatric symptoms more continuously to the underlying pathophysiology (see figure 2). In an attempt to expand on the dimensional diagnostic approach, the United States National Institute of Mental Health (US NIMH) has initiated the Research Domain Criteria (RDoC) project, which aims to “develop, for research purposes, new ways of classifying mental disorders based on dimensions of observable behavior and neurobiological measures”. Emil Kraepelin (1856-1926) was the first to make the distinction between the two disorders BD and SZ in what has since been known as the Kraepelinian dichotomy. This dichotomy was long clinically appreciated because it enabled psychiatrists to come to a clear diagnosis in complex clinical settings55 and guided them in decision making for pharmacological treatment: treatment with mood stabilizers in BD and antipsychotics in SZ. Yet, two problems arose. First, as many psychiatrists experienced, no point of rarity exists between these diagnoses57, which among other ‘solutions’ resulted in the intermediate classification of schizo-affective disorder. Secondly, the use of antipsychotics in particular has crossed the Kraepelinian dichotomy; these are nowadays also frequently used in BD for manic episodes, and some for depressive episodes58,59. Besides the clinical cracks in the Kraepelinian model, there is at present multiple evidence for shared pathophysiological mechanisms between BD and SZ. Most, if not all of the pathophysiological BD models described in chapter 1 have variants in SZ, and MDD51. Furthermore, genome-wide association studies (GWAS) have demonstrated the existence of genetic single nucleotide polymorphisms that influence the risk of both SZ and BD56,60. In addition, as described above, Drexhage et al. demonstrated that monocyte pro-inflammatory gene-expression was increased in both BD and SZ41. On the other hand, it would also be too simplistic to argue that both BD and SZ are the same disorder. There was some distinction between BD and SZ in the gene- Summary andSummary general discussion expression investigated by Drexhage et al.41. In addition, pre-schizophrenic children are characterized by cognitive and neuromotor impairments, which are not shared (at least not to the same extent) by children who later develop BD61. Finally, there are quite consistent neuroanatomical differences between BD and SZ, e.g. hippocampus volumes are typically decreased in patients with SZ, whereas in most studies hippocampus volumes in patients with BD are not distinguishable from those in HC62,63. Researchers in genetics, neurobiology and population epidemiology are increasingly inclined to adopt a continuous dimensional diagnostic approach to the variation in symptomatology in order to increase the validity of the diagnostic system. However, clinicians tend to maintain the categorical approach embodied in current classifications such as DSM-5 and ICD-1064. This should not be frowned upon. Current categorical classification systems have indisputably facilitated the development of 10

207 The Kraepelinian dichotomy – going, going . . . but still not gone

autism and other neurodevelopmental disorders and challenges influence on the risk of bipolar disorder,18,19 they appear to the view that these are completely unrelated diagnostic entities. contribute less to the susceptibility to bipolar disorder than to schizophrenia (to date, variants influencing bipolar disorder seem to be smaller, less likely to be deletions, and have smaller effect Findings suggesting that bipolar disorder sizes).19,20 Under the assumption that bigger structural genomic and schizophrenia do not have a single underlying variants, particularly involving DNA loss, are more likely to affect cause and are not the same clinical entity brain development, we note that these findings are consistent with the view that schizophrenia has a stronger neurodevelopmental Although we can reject a simple model of separate, unrelated component than bipolar disorder21 and suggest that it lies on a disease categories, the data do not support a model of a single- gradient of decreasing neurodevelopmental impairment between disease category that is undifferentiated with respect to the syndromes such as mental retardation and autism on one hand, relationship between clinical expression and genetic susceptibility, and bipolar disorder on the other (Fig. 1). and, hence, underlying biological mechanisms. For example, the Data suggesting a degree of specificity between pathophysiology same large family study2 that demonstrated a substantial overlap and phenotype come from work at the interface of the traditional in genetic susceptibility to bipolar disorder and schizophrenia also dichotomous categories. Cases with a rich mix of clinical features provided clear evidence for the existence of non-shared genetic of bipolar mood episodes and the psychotic symptoms typical of risk factors. These findings are fully consistent with earlier genetic schizophrenia (a broadly defined schizoaffective illness) may be FIGURE 2 22 data suggesting that there are relatively specific as well as shared particularly useful for genetic studies, and there is evidence that susceptibilityRelationships genes.1 Recent between studies suggest genotype that some and of this clinicalvariation phenotype within genes encoding across gamma-aminobutyric acid (A) specificity might be due to structural genomic variation (CNVs). receptor subunit genes may predispose relatively specifically to psychiatric disorders 23 Although there is emerging evidence that CNVs have some such mixed mood–psychosis clinical pictures. Although

Gradient of affective pathology Gradient of neurodevelopmental pathology

Mental Autism Schizophrenia Schizoaffective Bipolar/unipolar mood retardation disorder disorder Clinical syndromes

Cognitive impairment --—--—--—--—--—--—--— 5 5 55 57 5

--—--—- Negative symptoms --—--—--—--—- 8 7

--—--—--—--—--— Positive symptoms --—- Domains of 8 7 psychopathology --—--—--—--—--— Mood symptoms 8

Environmental influences & stochastic variation Neural modules 8 7 8 5 5 5

Biological systems 8 8 7 5

************** *** ** Genetic ** * ** * * variation

DNA structural variants 7

Fig. 1 Hypothesised model of the complex relationship between biological variation and some major forms of psychopathology.

This is a simplified model of a highly complex set of relationships between genotype and clinical phenotype. Starting at the level of genetic variation (lowest tier in figure), we haveThis represented is a DNAsimplified structural variation model (in purple) of as contributinga highly particularly complex to neurodevelopmental set of relationships disorders and associated between particularly with genotype enduring cognitive and and functional impairment. Single gene variants, of which there are many, are shown as asterisks. In general, even single base-pair changes in a gene may influence multiple biological systems becauseclinical genes typically phenotype. have multiple functions Starting and produce at proteinsthe level that interact of withgenetic multiple other variation, proteins. For simplicity, DNA westructural have shown only variation an example of a variantis that influences three biological systems (blue asterisk and arrows) and another that influences only one system (black asterisk and arrow). Variation in the relevant biological systems is influenced by genotype at many genetic loci and by environmental exposures/experiences both historically during development and currently to influence the dynamic state of the systems.represented The relevant biological (in systems purple) influence as the contributing neural modules that comprise particularly the key relevant to functional neurodevelopmental elements of the brain (shown as soliddisorders turquoise circles). and Typically, multiple biological systems influence each neural module. The (abnormal) functioning of the neural modules together influences the domains of psychopathology experienced and ultimatelyassociated the clinical syndromes. particularly We have ordered with some enduring important clinical cognitive syndromes along aand single functional major axis with a gradient impairment. of decreasing proportional Single neurodevelopmental gene contribution to causation and reciprocal increasing gradient of proportion of episodic affective disturbance (we use the term ‘mental retardation’ in the diagram because it is understoodvariants, internationally, of which but recognise there that the are terms many, intellectual disabilityare shown and learning as disability asterisks. are commonly In used general, in the UK). The even single axis single is a simplifying base-pair device – there is substantial individual variation and it is recognised that, for example, it is not uncommon for individuals diagnosed with autism to experience substantial mood pathology. Key featureschanges of the model in are a described gene within may the influence text. multiple biological systems because genes typically have multiple functions and produce proteins that interact with multiple other proteins. For simplicity, only an example of a variant has been shown that influences three biological 93 systems (blue asterisk and arrows) and another that influences only one system (black asterisk and arrow). Variation in the relevant biological systems is influenced by genotype at many genetic loci and by environmental exposures/experiences, both historically during development and currently, to influence the dynamic state of the systems. The relevant biological systems influence the neural modules that comprise the key relevant functional elements of the brain (shown as solid turquoise circles). Typically, multiple biological systems influence each neural module. The (abnormal) functioning of the neural modules together influences the domains of experienced psychopathology and ultimately the clinical syndromes. Some important clinical syndromes have been ordered along a single major axis with a gradient of decreasing proportional neurodevelopmental contribution to causation and reciprocal increasing gradient of proportion of episodic affective disturbance. (Source: Craddock and Owen56, reprinted with permission)

208 many psychiatric treatment options, thereby demonstrating almost indispensable clinical utility64. Therefore, altering the diagnostic system towards a dimensional approach in clinical practice would mean a major paradigm shift, requiring not only the investigation of connections between symptoms (phenomenology) and the underlying neuronal circuits, and cellular, molecular and genetic alterations, but also evidence for the prediction of treatment efficacy for the existing psychiatric treatments, relating them to this approach.

Bipolar disorder as a glial neurodevelopmental disorder with late clinical presentation Traditionally, biological psychiatric research and its pathophysiologic models focused mainly on the neuronal part of the CNS, e.g. brain region activation and neuronal neurotransmission, thereby ignoring the other predominant portion, namely the neuroglia, which consist of glial cells65. As indicated by their name, which translates from Greek as “glue”, glial cells were long thought to be of use mainly as structural supporting cells for the neurons: holding them in place, supplying them with nutrients and oxygen and destroying pathogens. However, at the turn of the century research began to demonstrate that glial cells have important functions in neurodevelopment and synaptic function66,67. The glial cell population consists mainly of oligodendroglia, astrocytes and microglia. Oligodendroglia create myelin sheaths around neuronal axons to give support and to increase the axonal transmission speed. In addition they provide trophic support by producing glial cell line-derived neurotrophic factor (GDNF), brain-derived neurotrophic factor (BDNF) and insulin-like growth factor-1 (IGF-1)68. In humans, astrocytes are now known to perform a multitude of functions such as, but not limited to: providing metabolic support as a lactate and glycogen energy buffer, vasomodulation by regulating blood flow69, promoting myelinating activity of oligodendroglia70,71, regulating nervous system repair72, facilitating long-term memory potentiation73 and Summary andSummary general discussion several kinds of signal transmission modulation, including modulation of synaptic transmission74 and regulation of ion concentration in the extracellular space75. Micro glia are the resident macrophages of the brain and spinal cord, and thus act as the first and main form of active immune defense in the CNS, constantly scavenging the CNS for plaques, damaged neurons and infectious agents. Besides functions relating to the immunoresponse, microglia play an important role in maintaining homeostasis. As with peripheral macrophages, microglial activation could be in an inflammatory sense (M1 macrophages), an anti-inflammatory sense (M2 macrophages), and a regenerating/tissue support sense (M2b macrophages). Animal models demonstrated that microglia are also involved in tissue regeneration and play an active role in neuronal support, i.e. the development of mature synapses during embryogenesis76, pruning synapses postnatally77, regulating neurogenesis78 and inducing apoptosis17. 10

209 It may well be the case that some microglial cells induce apoptosis, while others actively facilitate neurogenesis. Over the last few years these individual parts of the neuroglia have received increasing scientific attention in research on functional CNS disorders, including neurodegenerative disease79, neurodevelopmental disorders80 and psychiatric disorders in the strict sense81. In investigating BD in this thesis, we demonstrated activated microglia to be present in the (right) hippocampus of BD (chapter 6). We furthermore revealed a positive relationship between microglial activation and neuronal integrity, consistent with the concept of microglial function differentiation in vivo (chapter 7), thereby underpinning the concept of a key role for neuroglia in BD pathology. The concept of neuroplasticity, which is the ability of the brain to adapt during adult life, may be of use in understanding glial dysregulation in psychiatric disorders. Based on the principles of Darwinian selection82, the human brain displays an exceptional amount of plasticity, facilitating the adaptability characteristic of the human species 83. Recently, evidence arose linking astrocyte anatomical and functional exaptations, and genetic variations, to the complexity of the brain, which seems to be a distinctive feature in humans compared to other species84. In addition, a decreased density of glial cells across all layers was found in a post mortem morphometric study of the supragenual anterior cingulate cortex in mood disorder patients85. Elaborat- ing on this line of thought, it could be hypothesized that glial dysregulation in BD, among other things, leads to a diminished neuroplastic potential86,87. In this respect, it is noteworthy to emphasize that lithium, valproate and antidepressants indirectly regulate a number of pathways involved in cell survival and thereby bring about some long-term beneficial effects on this decreased neuroplastic potential88. As stated above, one of the essential functions of glial cells lies within the process of neurodevelopment66. Neurodevelopment, the process of creating a mature nervous system able to cope with its complex tasks, can be considered to be the temporal complexity of the brain, another form of CNS complexity in addition to the already mentioned spatial complexity47. In SZ research it has become clear that those patients demonstrate neurodevelopmental problems. For some time this was considered to be a distinctive feature of this disorder, as compared to BD56. However, more recently quite compelling scientific evidence became available which also indicated neurodevelopmental aberrations in BD89. The average total brain volume of patients with bipolar disorder was quite consistently found to be relatively small compared to HC, both in adults90 and adolescents91. Furthermore, there are reports of cell migration abnormalities in the neocortex of patients with BD92,93. Supporting this notion, in chapter 2 we demonstrated a possible association between an earlier age at onset and increased pro-inflammatory monocyte gene-expression. This is in agreement with several other studies demonstrating increased morbidity in BD patients with an earlier age at onset 94–96. In the Dutch bipolar offspring study, Mesman et al. demon-

210 strated that during adolescence, bipolar offspring showed increased inflammatory gene expression in monocytes, high serum pentraxin-related protein 3 (PTX3, an acute phase response protein) levels, but normal chemokine (C-C motif) ligand 2 (CCL2, a chemotactic protein) levels. In adolescence BDNF levels were decreased, while S100B levels, a marker for astrocytes, were normal. Interestingly at adulthood, circulating monocytes had lost their activation state, but CCL2 levels remained increased and both BDNF and S100B were now increased. The study suggests an aberrant neuro-immune state in bipolar offspring, following a dynamic course from adolescence into adulthood97. Based on these findings it can be hypothesized that neurodevelopmental disorganization, originating in glial dysregulation, may be a fundamental feature of BD which precedes the revelation of the clinical disorder. In BD, the brain probably temporarily overcomes this disorganization via compensation strategies, which are adequate for seemingly unaffected functioning during childhood, although in fact in hindsight many patients appear to have had prodromal symptoms98. Later on, these compensation strategies prove to be inferior to cope with excessive stress in adulthood, leading to psychic decompensation states such as manic, depressive and psychotic episodes (figure 3). Indeed, it has long been known that biological and psychosocial stress influence the nervous system and innate immunity through persistent activation of the hypothalamic-pituitary-adrenal (HPA) axis and the sympathetic-adrenal-medullary (SAM) axis, while on the other hand, chronic secre- tion of glucocorticoids and noradrenaline from the HPA and SAM axes can have pro- found effects on many immune functions99. In these stress-induced decompensation

FIGURE 3 Psychiatric and biological course of bipolar disorder Summary andSummary general discussion

Schematic representation of the activation of the inflammatory response system (IRS) and the development of psychiatric symptoms during the course of bipolar disorder. 10

211 states, additional dysregulation of specific neuronal circuits occurs, and several of the state dependent features mentioned in the pathophysiological models described in chapter 1 probably arise as a result.

Implications for clinical practice

Addition to current diagnostic and treatment options? As is intrinsic to pathophysiological research, this thesis did not give rise to direct diagnostic and treatment options, although the development of a diagnostic biomarker for BD was one of the initial goals, as discussed above. However, it is our hope that its main merit lies in its contribution of additional pieces of the puzzle that will eventually provide a clearer vision on the causes and mechanisms of psychiatric disorders. Based on the discussion above, some recommendations can be provided for clinical practice.

Gradual transformation to a more dimensional diagnostic system First, while many researchers are cautiously moving towards a more continuous dimensional diagnostic approach in order to increase the validity of the diagnostic system, clinicians prefer to hold on to the classical categorical approach, because of its accepted clinical utility64. Yet, reconciliation between validity and utility properties of the psychiatric diagnostic system is a vital goal to strive for, from the perspectives of both research and clinical practice. Following research, professionals in the clinical field should keep a close eye on dimensional diagnostic developments and be receptive for gradual, evolutionary dimensional alterations of the diagnostic system. This requires some re-evaluation of the status of our current diagnostic system, which is held in high esteem. The DSM and ICD are not nature-given, but man-made and have their utilitarian merits and their validation flaws. We should therefore be transparent about these uncertainties, both in our own stance towards the diagnostic system and in the communication with our patient. This needs to be done careful- ly without completely removing the basis of diagnosis, which is essential for both research and for clinical psychiatric practice. One move towards such a goal is to reappraise the vulnerability concept in our diagnostic systems100,101. In the current psychiatric population, amongst others, patients with psychotic and manic vulnerability can be recognized102. In this respect, patients with a manic vulnerability can be defined as patients that have had a previous manic or hypomanic episode, have a higher risk for recurrence (and ideally even first occurrence) of manic and depressive episodes, and can have cognitive problems, resulting in functional disabilities in daily life. Minimizing the effects of this vulnerability by

212 maximizing resilience is the target of the treatment approaches in these patients, requiring amongst others, psycho-educational programs, medication and psycho- therapy103. The vulnerability concept better grasps the developmental properties of the disorders and, because of its normalizing character, probably would facilitate the patient’s acceptance of having a psychiatric disorder.

Promising immune system related treatment options for bipolar disorder Based on the growing evidence for a role of the immune system and the neuroglia in BD pathophysiology, existing immune system targeting drugs have been investigated for treatment efficacy in BD and other disorders104. N-acetyl-cysteine (NAC), non-steroid anti-inflammatory drugs (NSAID) and omega 3 fatty acids (O3FA) have gained the most scientific attention. Although at this time there is inadequate evidence to support its widespread use, NAC shows the most promising results, being successful in lowering depressive symptomatology in four of five trials104,105. It is usually well tolerated and possesses anti-inflammatory effects, possibly via inhibiting NF-κB and modulating cytokine synthesis, and enhances the availability of glutathi- one, thereby affecting the NMDA and AMPA receptors105. Both NSAIDs and O3FAs produced conflicting results104. Some studies in psychosis did report a beneficial effect of celecoxib, a cyclooxygenase-2 inhibiting NSAID, especially in the early stage of illness onset104,106,107. Of the studies on O3FAs one neuroimaging study observed a short-term effect in membrane fluidity and neuronal activity in BD, but this effect was not reflected by a change in depression score108. O3FAs were, however, able to reduce the risk of progression to psychosis in another study with individuals at high risk for psychosis, emphasizing the importance of timely treatment109. It has to be noted that these findings are also interesting in light of the above discussion of the Kraepelinian dichotomy. Summary andSummary general discussion Future research perspectives When establishing the spatial and temporal complexity of the brain as important contributors hampering pathophysiological research in biological psychiatry, several important points for further research in BD need to be addressed.

Study design First, to elucidate the associations between various biological and phenomenological observations, aberrations on the individual physiologic levels of functioning need to be compared to each other, thereby stressing the importance of multimodal studies using multiple bioassay (genetics, cellular), neuroimaging (neuronal circuits) and interview (phenomenological) techniques. 10

213 Second, given the limitation of the dimensional diagnostic approach, one should invest in precise descriptions of the psychiatric phenomenology with predefined dimensions (e.g. manic, depressive, anxiety, and psychotic), using individual symptom profiles at the moment when the observational data are acquired. Besides profiling, stratification of the patients by means of staging techniques can help in coping with the (temporal) complexity of the brain, without falling into dichotomization110,111. Research has already demonstrated some prognostic biomarkers to be useful as an adjunctive tool in staging BD112. Third, longitudinal studies acquiring at least two time points (and preferentially more) are essential for elucidating cause and effect associations, both within and between the various pathophysiological levels. In addition, they are also essential for improving our understanding of the developmental aspects of BD. When attempting to elucidate the developmental aspects of BD, offspring studies are of special importance; these studies not only provide an elegant and valid method to study the familial transmission of BD, but they also make possible the identification of the early trajec- tory of BD and the earliest abberations97. Twin studies provide another perspective, revealing the absolute and relative importance of environmental and genetic influences on the development of BD113. Last but not least, randomized trials are longitudinal studies that can provide the highest level of causal relationships, as in many other areas of research. Finally, due to difficulties with accessing the brain tissue, complex cross-sectional and longitudinal studies in humans ultimately will not be able to answer all the research questions. Therefore, animal and laboratory models provide essential contributions, although they obviously must meet the highest ethical standards. Many of the more recent developments in biological psychiatric research would not have been possible without the scrutinizing research performed on these models to reveal the fundamental biological and biochemical processes involved.

Pharmacological considerations Focus on relapse prevention in psychoimmunological treatment As discussed above, based on current research it can be reasoned that glial cell dysregulation, together with inadequately coped neurodevelopmental disorganization, plays an important part in BD pathophysiology, where stress induced decompensation states give rise to neuronal circuit aberrations and metabolic alterations that are associated with mood episodes. Contrary to many of the current pharmacological treatments that have their main effect on these decompensation states, it can be argued that the goal of immune system targeting drug candidates should be to improve the stability of the dysregulated glial cells. In doing so these potential psychoimmunological treatment strategies should probably primarily target the prevention of relapses or exacerbation of the disorder.

214 Interestingly, this is something lithium already does114. Lithium is renowned for its efficacy in treating BD, where response to lithium can almost be considered pathog- nomonic, and for its multiple points of action114,115. Lithium acts on several second messenger systems that underpin its regulatory effects on neurotransmission and its neuroprotective properties. It modulates neurotransmission via several mechanisms. Over time, these processes modify gene transcription within the cells and yield long-lasting mood stabilization. Additionally, lithium reduces the oxidative burden caused by mood episodes and protects against apoptosis by promoting neuroprotective pathways and facilitating the actions of neuroprotective proteins. Furthermore, it inhibits glycogen synthase kinase 3 beta (GSK-3β), which besides regulating glycogen synthesis, is also involved in gene transcription, synaptic plasticity, cell structure and resilience. Finally, lithium also inhibits pro-apoptotic proteins and processes, e.g. autophagy114. It has been suggested that the unifying effect of all these functions lies within neurotrophic actions on neuronal and glial cells115. To increase our understanding of the multilevel BD pathophysiology, it is important to investigate longitudinally the versatile therapeutic actions of lithium not only on the cellular level, but also on the neuronal circuits level, and link these to the phenomenological alterations, both before and during or during and after lithium treatment.

Concluding remarks

As a consequence of the research techniques that have become available over the last few decades and the insights that are thereby provided, these are very exciting times to perform psychopathophysiological research, pushing our knowledge forward, always with the ultimate goal to further improve the lives of our patients. With this thesis we hope to contribute additional pieces of the puzzle, eventually providing a clearer vision of the causes and mechanisms of psychiatric disorders Summary andSummary general discussion than we have today. The neuroscientific research in this thesis adds to the important perception that glial cells play an important, and perhaps central, role in BD pathophysiology. Furthermore, increasingly evidence arises emphasizing the importance of looking at BD from a neurodevelopmental and transdimensional perspective to better understand its origins and its course. Understanding the spatial and temporal complexity of the brain, and the implications that these complexities pose for our thinking about BD pathophysiology, are essential to achieve further advancement in this field of research, unravelling the dysregulated brain. 10

215 References

1. MOODINFLAME website. http://moodinflame.eu. 2014. 2. Sainani KL. The problem of multiple testing. PM R. Elsevier Inc.; 2009;1(12):1098–103. 3. Bender R, Lange S. Adjusting for multiple testing - When and how? J Clin Epidemiol. 2001;54(4):343–9. 4. Schochet PZ. An approach for addressing the multiple testing problem in social policy impact evaluations. Eval Rev. 2009;33(6):539–67. 5. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med. 1990 Jul;9(7):811–8. 6. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B …. 1995; 7. Perneger T V. What’s wrong with Bonferroni adjustments. BMJ. 1998 Apr 18;316(7139):1236–8. 8. Hagler DJ, Saygin AP, Sereno MI. Smoothing and cluster thresholding for cortical surface-based group analysis of fMRI data. Neuroimage. 2006 Dec;33(4):1093–103. 9. Smith SM, Nichols TE. Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage. 2009 Jan 1;44(1):83–98. 10. Mann c J. Observational research methods . Research design II : Emerg Med J. 2003;(October 2008):54–61. 11. Drzyzga L, Obuchowicz E, Marcinowska A, Herman ZS. Cytokines in schizophrenia and the effects of antipsychotic drugs. Brain Behav Immun. 2006 Nov;20(6):532–45. 12. Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50. 13. Padmos RC, Hillegers MHJ, Knijff EM, Vonk R, Bouvy A, Staal FJT, et al. A discriminating messenger RNA signature for bipolar disorder formed by an aberrant expression of inflammatory genes in monocytes. Arch Gen Psychiatry. 2008 Apr;65(4):395–407. 14. Pollmächer T, Haack M, Schuld A. Effects of antipsychotic drugs on cytokine networks. J Psychiatr …. 2000;34(6):369–82. 15. Rybakowski JK. Antiviral and immunomodulatory effect of lithium. Pharmacopsychiatry. 2000 Sep;33(5):159–64. 16. Walker FR. A critical review of the mechanism of action for the selective serotonin reuptake inhibitors: Do these drugs possess anti-inflammatory properties and how relevant is this in the treatment of depression? Neuropharmacology. Elsevier Ltd; 2013;67:304–17. 17. Beumer W, Gibney SM, Drexhage RC, Pont-Lezica L, Doorduin J, Klein HC, et al. The immune theory of psychiatric diseases: a key role for activated microglia and circulating monocytes. J Leukoc Biol. 2012 Aug 8;92(September):1–17. 18. Lavisse S, Guillermier M, Hérard A-S, Petit F, Delahaye M, Van Camp N, et al. Reactive astrocytes overexpress TSPO and are detected by TSPO positron emission tomography imaging. J Neurosci. 2012 Aug 8;32(32):10809–18. 19. Hostenbach S, Cambron M, D’haeseleer M, Kooijman R, De Keyser J. Astrocyte loss and astrogliosis in neuroinflammatory disorders. Neurosci Lett. Elsevier Ireland Ltd; 2013 Oct 12;10–2. 20. Bushara NO, Hashim TH, Abraham A. Spatial Diversity of Monthly and Annual Rainfall in Sudan: Using Rainfall Predictions and Geographical Information System ( GIS ) to Produce Rainfall Maps. Int J Comput Inf Syst Ind Manag Appl. 2015;7(1):151–72. 21. Röper A-C, Orabi J, Lütken H, Christensen B, Thonning Skou A-M, Müller R. Phenotypic and Genotypic Analysis of Newly Obtained Interspecific Hybrids in the Campanula Genus. PLoS One. 2015;10(9):e0137537.

216 22. Naaldijk YM, Bittencourt MC, Sack U, Ulrich H. Kinins and microglial responses in Bipolar Disorder: a neuroinflammation hypothesis. Biol Chem. 2016 Feb; 23. Brites D, Fernandes A. Neuroinflammation and Depression: Microglia Activation, Extracellular Microvesicles and microRNA Dysregulation. Front Cell Neurosci. Switzerland; 2015;9:476. 24. Furtado M, Katzman MA. Examining the role of neuroinflammation in major depression. Psychiatry Res. Ireland; 2015 Sep;229(1-2):27–36. 25. Kim Y-K, Na K-S, Myint A-M, Leonard BE. The role of pro-inflammatory cytokines in neuroinflammation, neurogenesis and the neuroendocrine system in major depression. Prog Neuropsychopharmacol Biol Psychiatry. England; 2016 Jan;64:277–84. 26. Walker AK, Kavelaars A, Heijnen CJ, Dantzer R. Neuroinflammation and comorbidity of pain and depression. Pharmacol Rev. United States; 2014;66(1):80–101. 27. Sublette ME, Postolache TT. Neuroinflammation and depression: the role of indoleamine 2,3-dioxygenase (IDO) as a molecular pathway. Psychosomatic medicine. United States; 2012. p. 668–72. 28. Hurley LL, Tizabi Y. Neuroinflammation, neurodegeneration, and depression. Neurotox Res. United States; 2013 Feb;23(2):131–44. 29. Maes M, Smith R, Scharpe S. The monocyte-T-lymphocyte hypothesis of major depression. Psychoneuroendocrinology. 1995 Jan;20(2):111–6. 30. Smith RS. The macrophage theory of depression. Med Hypotheses. 1991 Aug;35(4):298–306. 31. Pascoe MC, Crewther SG, Carey LM, Crewther DP. What you eat is what you are -- a role for polyunsaturated fatty acids in neuroinflammation induced depression? Clin Nutr. England; 2011 Aug;30(4):407–15. 32. Mondelli V. And the BBI 2014 award for lead actor in “neuroinflammation and depression” goes to.... Brain Behav Immun. United States; 2014 Nov;42:48–9. 33. Wohleb ES, McKim DB, Sheridan JF, Godbout JP. Monocyte trafficking to the brain with stress and inflammation: a novel axis of immune-to-brain communication that influences mood and behavior. Front Neurosci. Switzerland; 2014;8:447. 34. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Arlington, VA: American Psychiatric Publishing, Inc; 2013. 35. World Health Organisation. ICD-10 Classifications of Mental and Behavioural Disorder: Clinical Descriptions and Diagnostic Guidelines. Geneva, Switzerland: World Health Organisation; 1992. 36. Kapur S, Phillips AG, Insel TR. Why has it taken so long for biological psychiatry to develop clinical tests and what to do about it? Molecular Psychiatry. Nature Publishing Group; 2012. p. 1192–3. 37. Friedman RA. A Dry Pipeline for Psychiatric Drugs. The New York Times. New York; 2013 Aug 20;D3. 38. Jentsch MC, Van Buel EM, Bosker FJ, Gladkevich A V, Klein HC, Oude Voshaar RC, et al. Biomarker Summary andSummary general discussion approaches in major depressive disorder evaluated in the context of current hypotheses. Biomark Med. 2015;9(3):277–97. 39. Gottesman II, Gould TD. The endophenotype concept in psychiatry: etymology and strategic intentions. Am J Psychiatry. 2003 Apr;160(4):636–45. 40. Biomarkers Definitions Working Group. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin Pharmacol Ther. 2001 Mar;69(3):89–95. 41. Drexhage RC, van der Heul-Nieuwenhuijsen L, Padmos RC, van Beveren N, Cohen D, Versnel M a, et al. Inflammatory gene expression in monocytes of patients with schizophrenia: overlap and difference with bipolar disorder. A study in naturalistically treated patients. Int J Neuropsychopharmacol. 2010 Nov;13(10):1369–81. 42. Becking K, Haarman BCM, van der Lek RFR, Grosse L, Nolen WA, Claes S, et al. Inflammatory monocyte gene expression: trait or state marker in bipolar disorder? Int J Bipolar Disord. 2015 Dec 17;3(1):20. 10

217 43. Dickerson F, Stallings C, Origoni A, Boronow J, Yolken R. Elevated serum levels of C-reactive protein are associated with mania symptoms in outpatients with bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry. 2007 May 9;31(4):952–5. 44. Open Science Collaboration. Estimating the reproducibility of psychological science. Science (80- ). 2015;349(6251):aac4716–aac4716. 45. Trouble at the lab: Scientists like to think of science as self-correcting. To an alarming degree, it is not. The Economist. 2013 Oct 19; 46. Niciu MJ, Mathews DC, Nugent AC, Ionescu DF, Furey ML, Richards EM, et al. Developing biomarkers in mood disorders research through the use of rapid-acting antidepressants. Depress Anxiety. 2014;31(4):297–307. 47. Bassett DS, Gazzaniga MS. Understanding complexity in the human brain. Trends Cogn Sci. 2011 May;15(5):200–9 48. Manji HK, Lenox RH. Signaling: Cellular insights into the pathophysiology of bipolar disorder. Biol Psychiatry. 2000;48(6):518–30. 49. Kapczinski F, Dias VV, Kauer-Sant’Anna M, Brietzke E, Vázquez GH, Vieta E, et al. The potential use of biomarkers as an adjunctive tool for staging bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry. Elsevier Inc.; 2009 Nov 13;33(8):1366–71. 50. Kendler KS. The dappled nature of causes of psychiatric illness: replacing the organic- functional/hardware-software dichotomy with empirically based pluralism. Mol Psychiatry. 2012 Apr;17(4):377–88. 51. Dierckx RAJO, Otte A, de Vries EFJ, van Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Berlin, Heidelberg: Springer Berlin Heidelberg; 2014. 52. Dehaene S, Naccache L. Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework. Cognition. 2001 Apr;79(1-2):1–37. 53. Freedman R, Lewis DA, Michels R, Pine DS, Schultz SK, Tamminga CA, et al. The Initial Field Trials of DSM-5: New Blooms and Old Thorns. Am J Psychiatry. 2013 Jan;170(1):1–5. 54. Boschloo L, Van Borkulo CD, Rhemtulla M, Keyes KM, Borsboom D, Schoevers RA. The network structure of symptoms of the diagnostic and statistical manual of mental disorders. PLoS One. 2015;10(9):1–12. 55. Craddock N, Owen MJ. The beginning of the end for the Kraepelinian dichotomy. Br J Psychiatry. 2005 May 1;186(5):364–6. 56. Craddock N, Owen MJ. The Kraepelinian dichotomy - Going, going... but still not gone. Br J Psychiatry. 2010;196(2):92–5. 57. Kendell RE. Diagnosis and classification of functional psychoses. Br Med Bull. 1987 Jul;43(3):499– 513. 58. Grunze H, Vieta E, Goodwin GM, Bowden C, Licht RW, Möller H-J, et al. The World Federation of Societies of Biological Psychiatry (WFSBP) guidelines for the biological treatment of bipolar disorders: update 2012 on the long-term treatment of bipolar disorder. World J Biol Psychiatry. 2013;14(3):154–219. 59. Yatham LN, Goldstein JM, Vieta E, Bowden CL, Grunze H, Post RM, et al. Atypical antipsychotics in bipolar depression: potential mechanisms of action. J Clin Psychiatry. 2005 Jan;66 Suppl 5:40–8. 60. Bramon E, Sham PC. The common genetic liability between schizophrenia and bipolar disorder: a review. Curr Psychiatry Rep. 2001 Aug;3(4):332–7. 61. Murray RM, Sham P, Van Os J, Zanelli J, Cannon M, McDonald C. A developmental model for similarities and dissimilarities between schizophrenia and bipolar disorder. Schizophr Res. 2004 Dec 1;71(2-3):405–16. 62. Otten M, Meeter M. Hippocampal structure and function in individuals with bipolar disorder: A systematic review. J Affect Disord. Elsevier; 2015;174:113–25. 63. Wright IC, Rabe-Hesketh S, Woodruff PWR, David AS, Murray RM, Bullmore ET. Meta-analysis of regional brain volumes in schizophrenia. Am J Psychiatry. 2000;157(1):16–25.

218 64. Jablensky A. Psychiatric classifications: validity and utility. World Psychiatry. 2016 Feb;15(1):26–31. 65. Virchow R. Die Cellular pathologie in ihrer Bergundung auf Physiologie und pathologischeGewebelehre. Hirschwald: Berlin. 1858; 66. Nadarajah B, Brunstrom JE, Grutzendler J, Wong RO, Pearlman AL. Two modes of radial migration in early development of the cerebral cortex. Nat Neurosci. 2001 Feb;4(2):143–50. 67. Araque A, Parpura V, Sanzgiri RP, Haydon PG. Tripartite synapses: glia, the unacknowledged partner. Trends Neurosci. 1999 May;22(5):208–15. 68. Bradl M, Lassmann H. Oligodendrocytes: biology and pathology. Acta Neuropathol. 2010 Jan;119(1):37–53. 69. Parri R, Crunelli V. An astrocyte bridge from synapse to blood flow. Nat Neurosci. 2003 Jan;6(1):5– 6. 70. Ishibashi T, Dakin KA, Stevens B, Lee PR, Kozlov S V, Stewart CL, et al. Astrocytes promote myelination in response to electrical impulses. Neuron. 2006 Mar 16;49(6):823–32. 71. Bartzokis G. Neuroglialpharmacology: Myelination as a shared mechanism of action of psychotropic treatments. Neuropharmacology. Elsevier Ltd; 2012;62(7):2137–53. 72. Sofroniew M V. Reactive astrocytes in neural repair and protection. Neuroscientist. 2005 Oct;11(5):400 –7. 73. Han X, Chen M, Wang F, Windrem M, Wang S, Shanz S, et al. Forebrain engraftment by human glial progenitor cells enhances synaptic plasticity and learning in adult mice. Cell Stem Cell. 2013 Mar 7;12(3):342–53. 74. Piet R, Vargová L, Syková E, Poulain DA, Oliet SHR. Physiological contribution of the astrocytic environment of neurons to intersynaptic crosstalk. Proc Natl Acad Sci U S A. 2004 Feb 17;101(7):2151–5. 75. Walz W. Role of astrocytes in the clearance of excess extracellular potassium. Neurochem Int. 2000 Apr;36(4-5):291–300. 76. Paolicelli RC, Bolasco G, Pagani F, Maggi L, Scianni M, Panzanelli P, et al. Synaptic pruning by microglia is necessary for normal brain development. Science. 2011 Sep 9;333(6048):1456–8. 77. Schafer DP, Lehrman EK, Kautzman AG, Koyama R, Mardinly AR, Yamasaki R, et al. Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron. Elsevier Inc.; 2012 May 24;74(4):691–705. 78. Sierra A, Encinas JM, Deudero JJP, Chancey JH, Enikolopov G, Overstreet-Wadiche LS, et al. Microglia shape adult hippocampal neurogenesis through apoptosis-coupled phagocytosis. Cell Stem Cell. Elsevier Inc.; 2010 Oct 8;7(4):483–95. 79. Mrak RE, Griffin WST. Glia and their cytokines in progression of neurodegeneration. Neurobiol Aging. 2005 Mar;26(3):349–54. Summary andSummary general discussion 80. Zeidán-Chuliá F, Salmina AB, Malinovskaya NA, Noda M, Verkhratsky A, Moreira JCF. The glial perspective of autism spectrum disorders. Neurosci Biobehav Rev. 2014 Jan;38:160–72. 81. Frick LR, Williams K, Pittenger C. Microglial dysregulation in psychiatric disease. Clin Dev Immunol. 2013 Jan;2013:608654. 82. Darwin C. On the origin of species. London: John Murray; 1859. 502 p. 83. Gomez-Robles A, Hopkins WD, Schapiro SJ, Sherwood CC. Relaxed genetic control of cortical organization in human brains compared with chimpanzees. Proc Natl Acad Sci. 2015;112(48):14799–804. 84. Robertson JM. Astrocytes and the evolution of the human brain. Med Hypotheses. 2014 Feb;82(2):236–9. 85. Gittins RA, Harrison PJ. A morphometric study of glia and neurons in the anterior cingulate cortex in mood disorder. J Affect Disord. Elsevier; 2011;133(1):328–32. 86. Eyre H, Baune BT. Neuroplastic changes in depression: A role for the immune system. Psychoneuroendocrinology. Elsevier Ltd; 2012;37(9):1397–416. 10 87. Drevets WC. Neuroplasticity in mood disorders. Dialogues Clin Neurosci. 2004;6(2):199–216.

219 88. Manji HK, Moore GJ, Rajkowska G, Chen G. Neuroplasticity and cellular resilience in mood disorders. Mol Psychiatry. 2000;5(6):578–93. 89. O’Shea KS, McInnis MG. Neurodevelopmental origins of bipolar disorder: iPSC models. Mol Cell Neurosci. Elsevier Inc.; 2016 Jun;73:63–83. 90. De Peri L, Crescini A, Deste G, Fusar-Poli P, Sacchetti E, Vita A. Brain Structural Abnormalities at the Onset of Schizophrenia and Bipolar Disorder: A Meta-analysis of Controlled Magnetic Resonance Imaging Studies. Current Pharmaceutical Design. 2012. p. 486–94. 91. Najt P, Wang F, Spencer L, Johnston JAY, Cox Lippard ET, Pittman BP, et al. Anterior Cortical Development During Adolescence in Bipolar Disorder. Biol Psychiatry. 2016 Feb 15;79(4):303–10. 92. Tabares-Seisdedos R, Escamez T, Martinez-Gimenez JA, Balanzá V, Salazar J, Selva G, et al. Variations in genes regulating neuronal migration predict reduced prefrontal cognition in schizophrenia and bipolar subjects from mediterranean Spain: a preliminary study. Neuroscience. Elsevier; 2006;139(4):1289–300. 93. Uribe E, Wix R. Neuronal migration, apoptosis and bipolar disorder. Rev Psiquiatr y salud Ment. 2012;5(2):127–33. 94. Bellivier F, Golmard JL, Henry C, Leboyer M, Schürhoff F. Admixture analysis of age at onset in bipolar I affective disorder. Arch Gen Psychiatry. 2001 May 1;58(5):510–2. 95. Leboyer M, Henry C, Paillere-Martinot M-L, Bellivier F. Age at onset in bipolar affective disorders: a review. Bipolar Disord. 2005 Apr;7(2):111–8. 96. Coryell W, Fiedorowicz J, Leon AC, Endicott J, Keller MB. Age of onset and the prospectively observed course of illness in bipolar disorder. J Affect Disord. Elsevier; 2013 Mar 20;146(1):34–8. 97. Mesman E, Hillegers MH, Ambree O, Arolt V, Nolen W a., Drexhage H a. Monocyte activation, brain-derived neurotrophic factor (BDNF), and S100B in bipolar offspring: A follow-up study from adolescence into adulthood. Bipolar Disord. 2014;39–49. 98. Hafeman DM, Merranko J, Axelson D, Goldstein BI, Goldstein T, Monk K, et al. Toward the Definition of a Bipolar Prodrome: Dimensional Predictors of Bipolar Spectrum Disorders in At-Risk Youths. Am J Psychiatry. 2016 Jul 1;173(7):695–704. 99. Kawli T, He F, Tan M-W. It takes nerves to fight infections: insights on neuro-immune interactions from C. elegans. Dis Model Mech. 2010;3(11-12):721–31. 100. Müller-Spahn F. Individualized preventive psychiatry: syndrome and vulnerability diagnostics. Eur Arch Psychiatry Clin Neurosci. 2008 Nov;258 Suppl:92–7. 101. Nuechterlein KH, Dawson ME. A heuristic vulnerability/stress model of schizophrenic episodes. Schizophr Bull. 1984;10(2):300–12. 102. Stone MH. Assessing vulnerability to schizophrenia or manic-depression in borderline states. Schizophr Bull. 1979;5(1):105–10. 103. Stassen HH, Angst J, Hell D, Scharfetter C, Szegedi A. Is there a common resilience mechanism underlying antidepressant drug response? Evidence from 2848 patients. J Clin Psychiatry. 2007 Aug;68(8):1195–205. 104. Ayorech Z, Tracy DK, Baumeister D, Giaroli G. Taking the fuel out of the fire: Evidence for the use of anti-inflammatory agents in the treatment of bipolar disorders. J Affect Disord. 2014 Dec 12;174C:467–78. 105. Deepmala, Slattery J, Kumar N, Delhey L, Berk M, Dean O, et al. Clinical Trials of N-acetylcysteine in Psychiatry and Neurology: A Systematic Review. Neurosci Biobehav Rev. Elsevier Ltd; 2015;55:294–321. 106. Müller N, Riedel M, Scheppach C, Brandstätter B, Sokullu S, Krampe K, et al. Beneficial antipsychotic effects of celecoxib add-on therapy compared to risperidone alone in schizophrenia. Am J Psychiatry. 2002 Jun;159(6):1029–34. 107. Müller N, Krause D, Dehning S, Musil R, Schennach-Wolff R, Obermeier M, et al. Celecoxib treatment in an early stage of schizophrenia: results of a randomized, double-blind, placebo- controlled trial of celecoxib augmentation of amisulpride treatment. Schizophr Res. 2010 Aug;121(1-3):118–24.

220 108. Frangou S, Lewis M, Wollard J, Simmons A. Preliminary in vivo evidence of increased N-acetyl-aspartate following eicosapentanoic acid treatment in patients with bipolar disorder. J Psychopharmacol. 2007 Jun;21(4):435–9. 109. Amminger GP, Schäfer MR, Schlögelhofer M, Klier CM, McGorry PD. Longer-term outcome in the prevention of psychotic disorders by the Vienna omega-3 study. Nat Commun. 2015;6:7934. 110. Berk M, Hallam KT, McGorry PD. The potential utility of a staging model as a course specifier: a bipolar disorder perspective. J Affect Disord. 2007 Jun;100(1-3):279–81. 111. Hetrick SE, Parker AG, Hickie IB, Purcell R, Yung AR, McGorry PD. Early identification and intervention in depressive disorders: towards a clinical staging model. Psychother Psychosom. 2008;77(5):263–70. 112. Kapczinski F, Dal-Pizzol F, Teixeira AL, Magalhaes PVS, Kauer-Sant’anna M, Klamt F, et al. Peripheral biomarkers and illness activity in bipolar disorder. J Psychiatr Res. Elsevier Ltd; 2010 Jun 10;1–6. 113. Vonk R, van der Schot AC, Kahn RS, Nolen W a, Drexhage H a. Is autoimmune thyroiditis part of the genetic vulnerability (or an endophenotype) for bipolar disorder? Biol Psychiatry. 2007 Jul 15;62(2):135–40. 114. Malhi GS, Tanious M, Das P, Coulston CM, Berk M. Potential mechanisms of action of lithium in bipolar disorder: Current understanding. CNS Drugs. 2013. p. 135–53. 115. Machado-Viera R, Manji H, Zarate. The role of lithium in the treatment of bipolar disorder: convergent evidence for neurotrophic e ff ects as a unifying hypothesis. BiPolar Disord. 2009;11(Supplement):92–109. Summary andSummary general discussion

221

Dutch summary / Nederlandstalige samenvatting

Bartholomeus C.M. Haarman Achtergrond

“Toen ik vijftien jaar oud was, schreef ik in mijn dagboek dat ik liever dood was. Ik vond er helemaal niets aan. Ik wilde er een einde aan maken. Echter, er waren ook momenten dat het leven mij toelachte: ik was zeer energiek en alles was een groot feest. Dit eindigde vaak weer met verdriet, huilbuien en somberheid. Ik werd door mijn familie ‘hysterisch’ en ‘een aansteller’ genoemd. Ik was heel eenzaam. Zo gingen de jaren voorbij en ik wist niet wat er was en begon zeer aan mijzelf te twijfelen. Vaak werd aan de depressie een fysieke oorzaak gegeven. Het was vallen en weer opstaan, het waren bergen met hele diepe dalen. Uiteindelijk ben ik op eigen initiatief weer naar de huisarts gegaan. Toen begon een moeilijk traject waarin de diagnose van een bipolaire stoornis pas na lange tijd gesteld werd. Ik was toen vijftig jaar oud en had meer dan dertig jaar met deze stoornis geleefd, zonder therapie. Het was verdrietig, eenzaam, slecht voor de fysieke gezondheid en ook mijn sociale leven leed eronder. Ik kon na zesendertig jaar als verpleegkundig specialist helaas mijn werk niet meer doen. Er is veel onbegrip, stigmatisering en onwetendheid over de bipolaire stoornis. Nu, acht jaar later, gaat het soms goed, maar er zijn ook periodes met veel suïcidaliteit. Ook was er veel strijd met de naaste omgeving, een- zaamheid vanwege het onbegrip en natuurlijk acceptatie van de bipolaire stoornis en wat de gevolgen en bijwerkingen van medicatie met zich mee kunnen brengen. Het bezoeken aan gelegenheden met veel mensen, veel lawaai, zware druk en emotionele gebeurtenissen - maar ook aangename gebeurtenissen - kunnen leiden tot een hernieuwde manie of depressie of deze verergeren. Ik heb jarenlang veel medicatie gebruikt, met wisselend resultaat of geen resultaat. Ook heeft jarenlange psychotherapie bijgedragen aan een verbetering van mijn kwaliteit van leven. Uiteindelijk hebben ECT’s (electroconvulsietherapie) een geweldig succes opgeleverd en leid ik weer een volledig bestaan, rekening houdend met triggers en de mogelijkheid van recidief.”

- Riekje Venema, 58 jaar

224 Naaststaand citaat beschrijft de impact die een bipolaire (ofwel manisch- depressieve) stoornis kan hebben op het leven van patiënten, de worsteling om er toch het beste van te maken en de blijvende kwetsbaarheid. Een bipolaire stoornis manifesteert zich door stemmingsklachten, waarbij patiënten episodes hebben waarin hun stemming ernstig verhoogd is (manie) of verlaagd (depressie). Daarnaast hebben ze moeite om hun dagelijkse taken uit te voeren, waarbij ze vaak ook in meer of mindere mate problemen ervaren met het geheugen en het vermogen om na te denken (cognitie). Er bestaan twee vormen van de bipolaire stoornis: de bipo laire I-variant, waarbij vooral manische periodes optreden en de bipolaire II-stoornis, waarbij sprake is van hypomane episodes. Het verschil tussen manie en hypomanie is als volgt: manische episodes beperken patiënten ook om te functioneren in hun werk, relatie of sociale leven en kunnen aanleiding geven tot een klinische opname, terwijl mensen bij hypomane episodes deze functionele beperkingen niet hebben en ook niet hoeven worden opgenomen. De bipolaire stoornis komt relatief vaak voor, bij ongeveer 2% van de algemene bevolking. Deze heeft impact op iemands persoonlijke leven en vormt bovendien een aanzienlijke maatschappelijke last. De periode tussen het moment waarop de eerste klachten optreden en waarop de diagnose gesteld wordt, duurt relatief lang: gemiddeld ongeveer zes jaar. Dit komt vooral doordat de ziekte vaak begint met een depressieve episode en de karakteristieke (hypo)manische klachten pas later optreden. De start met de juiste behandeling kan vertraging oplopen, omdat de behandeling van een bipolaire stoornis verschilt met die van een ‘gewone’ depressieve stoornis (major depressive disorder; MDD). Om sneller een diagnose te kunnen stellen en om nieuwe behandelmethoden te vinden, is een volledig begrip van de pathofysiologie (oorzaak/ontstaanswijze) van deze stoornis essentieel. Door de jaren heen zijn daartoe verschillende theoretische modellen ontwikkeld. Het oudste nog gangbare model is de monoaminetheorie, maar dankzij andere (nieuwe) onderzoeken zijn er in de laatste jaren een aantal nieuwe theorieën bijgekomen. Monoamine neurotransmitters zijn neurotransmitters die dezelfde groep amino Dutch / summary samenvatting Nederlandstalige zuren bevatten, zoals serotonine, noradrenaline en dopamine. De monoaminetheorie gaat ervan uit dat bij ernstige psychiatrische aandoeningen de activiteit van deze neurotransmitters in de verbindingspunten tussen zenuwen (synapsen) verstoord is. Met de geneesmiddelen die toegepast worden bij psychiatrische stoornissen (antidepressiva of antipsychotica), wordt geprobeerd om de activiteit van deze stoffen te beïnvloeden.

225 FIGUUR 1 Het limbisch systeem

Onderdelen van het limbisch systeem in het brein. (Bron: aanpassing van originele tekening van OpenStax College op de Anatomy & Physiology, Connexions Website http://cnx.org/content/col11496/1.6/ onder de Creative Commons Attribution 3.0 Generic licentie (CC BY 3.0))

Uit onderzoek is gebleken dat het zogenaamde limbisch systeem, dat bestaat uit een aantal gebieden in het brein die verantwoordelijk zijn voor emotionele beleving, actiever is bij patiënten met een stemmingsstoornis (figuur 1). Volgens de cortico limbische theorie vertoont de prefrontale hersenschors (cortex) hierbij ook een verminderde activiteit; dit is een gebied in het brein dat planning en sociaal gedrag regelt en dat impulsief gedrag afremt. Deze theorie heeft deels overlap met be vindingen dat verschillende neurologische netwerken die emotieverwerking regelen, bij psychiatrische patiënten verstoord zijn. Tijdens de behandeling van patiënten met een bipolaire stoornis viel op dat deze mensen verhoudingsgewijs meer auto-immuunziekten hebben dan de doorsnee bevolking. Hieruit voortvloeiend onderzoek heeft aanwijzingen gevonden dat de witte bloedcellen, die het grootste deel van het immuunsysteem vormen, bij patiënten met een bipolaire stoornis op sommige vlakken anders actief zijn dan bij gezonde mensen. Bovendien bleek de concentratie van verschillende immuuntransmitterstoffen, cytokines, te verschillen. Ook bleken bij deze mensen de monocyten, een groep witte bloedcellen, een verhoogde ontstekinggerelateerde genexpressie te hebben. Onderzoeken met geavanceerde diffusietensor beeldvorming (diffusion tensor imaging , DTI) hebben uitgewezen dat de microstructuur van verbindingsbanen tussen verschillende hersengebieden bij patiënten met een bipolaire stoornis verstoord is. DTI is een MRI-techniek (magnetische resonantie beeldvorming) waarbij in bepaalde

226 gebieden naar de voorkeursrichting van waterbeweging wordt gekeken (diffusiviteit). Hiermee kan worden onderzocht of de microstructuur van de witte stof van het brein, die de verbindingsbanen bevat, intact is. DTI levert verschillende parameters, waar onder een schatting van de waterbeweging dóór de vezelbaan (axiale diffusiviteit, AD), haaks op de vezelbaan (radiale diffusiviteit, RD) en de gemiddelde diffusiviteit (MD). Een andere veelgebruikte maat is fractionele anisotropie (FA), die de relatie tussen axiale en radiale diffusiviteit weergeeft en aangeeft hoe sterk de waterbeweging in een richting is. De meeste onderzoeken die tot dusver uitgevoerd zijn, tonen verstoringen in de prefrontaalkwab, de hersenbalk en verschillende associatievezels. Waar deze verstoringen zich precies voordoen in het algehele pathofysiologische proces is nog onduidelijk, maar ze hebben bij de bipolaire stoornis mogelijk te maken met het immuunsysteem, ongeveer net zoals bij de ziekte multiple sclerose (MS). Uit onderzoek van de zenuwcellen van overleden patiënten met een bipolaire stoornis, is gebleken dat bij hen de mitochondriën, de energiecentrales van de cellen, kleiner zijn en dichter rond de celkern liggen dan bij gezonde personen. Daarnaast werd duidelijk dat sommige hersengebieden van patiënten een bepaald enzym minder aanmaken, dat een rol speelt in de energiehuishouding (creatinekinase). Mogelijk komt dit door overactiviteit in die gebieden, maar ook zou verstoring in de celenergiehouding de problemen kunnen veroorzaken. Hoewel meerdere pathofysiologische modellen de oorzaak van de bipolaire stoornis proberen te verklaren, bestaat er nog steeds geen omvattend model waarin alle verschillende afwijkingen samenkomen die ontdekt zijn. De cellen van het immuunsysteem staan waarschijnlijk centraal in een dergelijk omvattend model, omdat ze naast een voor het lichaam onmisbare bescherming tegen ziektekiemen ook in het brein voorkomen (de zogenaamde microglia) en daar houden ze het interne milieu van het brein in evenwicht (homeostase). Dit proefschrift wil de rol onderzoeken van het immuunsysteem in de pathofysiologie van de bipolaire stoornis, via verschillende benaderingen. Het immuunsysteem bestaat uit twee delen: het neuroimuunsysteem in het brein en het perifere immuun- Dutch / summary samenvatting Nederlandstalige systeem. In tegenstelling tot de immuuncellen van het perifere immuunsysteem bestaan die van het neuroimmuunsysteem voornamelijk uit gliacellen, in het bij zonder microglia. Het neuroimmuunsysteem beschermt het zenuwstelsel en regelt de structuren en processen die de interactie tussen het immuunsysteem en het centrale zenuwstelsel verzorgen.

227 Het perifere immuunsysteem Het eerste deel van dit proefschrift hebben we de functie van het perifere immuunsysteem onderzocht. In hoofdstuk 2 bekeken we of er een verband is tussen een uitgebreide lijst van klinische kenmerken en ontstekinggerelateerde genexpressie in monocyten, die gemeten is door middel van kwantitatieve polymerasekettingreactie (quantitative polymerase chain reaction, qPCR). De hypothese hierbij, die gebaseerd was op eerder onderzoek, was dat er een verband zou zijn tussen het voorkomen van psychotische symptomen (ooit in het leven) en ontstekinggerelateerde genexpressie in deze cellen. Hoewel we dit verband niet hebben aangetoond, hebben we een nieuwe onder- zoeksmethode ontwikkeld: de feature-expression heat map-methode. Deze leverde wel aanwijzingen dat er een relatie is tussen manische symptomen en ontsteking gerelateerde genexpressie. Daarnaast is er een verschil in immuunactivatie te zien, die te maken zou kunnen hebben met de beginleeftijd van de patiënt. Bovendien is ondersteunend bewijs gevonden dat sommige geneesmiddelen, die een effect hebben op de stemming, het immuunsysteem remmen; dit hadden andere onder zoekers ook al gerapporteerd. De nieuwe, door ons ontwikkelde, feature-expression heat map-methode, die in detail aan de orde komt in hoofdstuk 3, laat bij complexe verbanden in één grafische weergave van verschillende verbanden het effect zien en tegelijkertijd in welke mate het verband statistisch betekenisvol (significant) is. Deze zogenoemde heatmaps tonen verbanden tussen twee groepen kenmerken, waarbij de groep met voorafgaande (onafhankelijke) kenmerken in kolommen wordt weergegeven en de groep met erop volgende (afhankelijke) kenmerken in rijen. Beide groepen worden in een volgorde geplaatst waarbij kenmerken die vergelijkbaar zijn naast elkaar staan. Dit kan bijvoorbeeld op basis van hun eigenschappen (fenomenologie) of een cluster analyse, die meer datagestuurd te werk gaat. Deze rangschikking maakt het mogelijk om associatieclusters te herkennen, volgens het effect-ordered data display-principe. In de feature-expression heat map worden de verbanden tussen onafhankelijke en afhankelijke kenmerken aangegeven met cirkels, waarbij de intensiteit en de kleur van de cirkel (rood is positief, blauw is negatief) de effectgrootte weergeeft en de diameter de mate waarin het statistisch betekenisvol is (zie figuur 2). Deze com binatie maakt het mogelijk om associatieverbanden in complexe systemen, zoals pathofysiologische modellen, in een oogopslag te zien. Vervolgens voerden we binnen het Europese MOODINFLAME-project een vervol- ganalyse uit onder de groep met patiënten met een bipolaire stoornis die deelnamen aan deze studie. Hierbij was het doel om te onderzoeken of de ontstekinggerelateerde genexpressie in monocyten te maken heeft met de stemmingstoestand van mensen of eerder een kenmerkend iets van de ziekte op zich is. Hoofdstuk 4 bevat hiervan de resultaten: een verhoogde ontstekinggerelateerde genexpressie komt

228 FIGUUR 2 Overzicht van een feature-expression heat map

Voorafgaande groep variabelen A (volgorde door cluster analyse of fenomenologisch)

Variabele 1 Variabele 2 Variabele 3 ... Variabele n kleur intensiteit = effect grootte Variabele 1 diameter cirkel = statistische significantie Variabele 2 blauw = negatief effect

Variabele 3 rood = positief effect

Variabele 4 midden stip = significante associatie na ‘false discovery rate’ correctie

of fenomenologisch) ... samenvallende associaties Volgende groep variablelen B variablelen groep Volgende (volgorde door cluster analyse door cluster analyse (volgorde Variabele n

bij patiënten met een bipolaire stoornis meer voor wanneer zij zich in een zoge noemde stemmingsepisode (met name somberheid) bevinden. Dit kwam naar voren in vergelijking met zowel gezonde mensen als patiënten met een neutrale stemming (de zogeheten euthyme periode). Daarnaast was bij patiënten tijdens een episode de ontstekinggerelateerde genexpressie hoger, vergeleken bij dezelfde patiënten wanneer zij zich in een euthyme periode bevonden. Daarna behandelden we in hoofdstuk 5 een retrospectief medisch-dossieronderzoek naar de vraag of C-reactive protein (CRP), een algemene biomarker voor ontsteking, Dutch / summary samenvatting Nederlandstalige een ongunstige prognose aangeeft; een biomarker is een stof die wordt gebruikt om een bepaalde biologische toestand te markeren. In dit onderzoek (in een poliklini- sche setting) veronderstelden we dat patiënten met een hogere CRP-meting en die euthym waren, na een kortere tijd een terugval kregen en dat patiënten die in een stemmingsepisode zaten, een langere tijd nodig zouden hebben om te herstellen. Echter, in tegenstelling tot wat eerdere studies beweerden, was hiertussen in onze praktijkstudie geen verband te zien: er was in de longitudinale analyse geen verband zichtbaar tussen de CRP en de prognose in de verschillende groepen. Daarnaast bleek er geen verschil te zijn in de beginmeting tussen de verschillende groepen. Al met al lijkt de CRP als eenvoudige biomarker in de praktijk niet zo geschikt om het beloop van de stoornis te voorspellen, zoals eerder wel gesuggereerd was.

229 Neuroimmuunsysteem In het tweede deel van dit proefschrift hebben we het neuroimmuunsysteem onderzocht aan de hand van de volgende beeldvormingsmethoden: positron emissietomografie (PET), magnetische resonantiebeeldvorming (MRI), magnetische resonantiespectroscopie (MRS) en diffusietensorbeeldvorming (DTI). Hierbij ging speciale aandacht uit naar de microglia, de immuuncellen van het brein. Als eerste beschreven we deze technieken met een overzicht van de beschik bare literatuur over eerdere studies met PET en single photon emission computed tomography (SPECT) bij patiënten met een bipolaire stoornis (hoofdstuk 6). PET is een beeldvormende techniek waarbij de patiënt een radioactief isotoop (een radionuclide, een radioactief deeltje) krijgt toegediend. Gebonden aan een specifieke stof, een tracer, verzamelen deze radionuclides zich op bepaalde plaatsen in het lichaam, bijvoorbeeld met FDG (fludeoxyglucose) op een plek waar het lichaam relatief veel energie verbruikt, zoals in een tumor. Wanneer de radionuclides gebonden zijn aan een andere tracer, bijvoorbeeld met een onderzoeksstof genaamd PK11195 verzamelen ze zich op een plek waar het immuunsysteem actief is, zoals bij een ontsteking. De radionuclides zijn in hun atoomkernen instabiel en bij het uiteen- vallen produceren zij een positron en een neutrino. Positronen zijn de antideeltjes (tegenovergestelde deeltjes) van elektronen: ze hebben dezelfde massa, maar een positieve lading. Een gevormd positron zal na een kort pad botsen met een electron en bij deze annihilatie (vernietiging) een paar gammafotonen produceren, die worden gemeten als de patiënt zich in een ringdetector bevindt. Zo kan vastgelegd worden waar elke annihilatie heeft plaatsgevonden en wordt duidelijk hoe de radionuclides in het lichaam verdeeld zijn. Bij SPECT-onderzoek komen direct gammastralen vrij, in plaats van protonen. De eerste PET/SPECT-studies richtten zich met name op het metabolisme en de bloeddoorstroming van verschillende hersengebieden. Zij vonden overwegend een overactiviteit in het limbisch systeem, die samenhangt met een verminderde activiteit in de prefrontaalkwab: de eerdergenoemde corticolimbische theorie. Echter, in zijn beknopte vorm is dit model niet volledig genoeg om alle specifieke symptomen te verklaren die zich voordoen bij stemmingwisselingen en het denkvermogen van patiënten. Daarnaast toonde moleculaire beeldvorming aan hoe relevant verande ringen van de serotoninetransporter in delen van het limbisch systeem zijn, net als veranderingen in de dopaminerge en cholinerge neurotransmissie (monoamine theorie). Opmerkelijk genoeg gebruiken de meeste moleculaire beeldvormingsstudies veel op zichzelf staande onderzoeksopzetten; hiermee ontstaat weliswaar meer inzicht in verschillende pathofysiologische mechanismen, maar wordt het lastiger om de uitkomsten onderling te vergelijken.

230 Vervolgens onderzochten we microglia-activatie in de hippocampi bij de bipo laire stoornis in een neuroinflammatie PET-studie (hoofdstuk 7). De hippocampi (linker en rechter hippocampus) zijn gebieden in het brein die een onderdeel zijn van het limbisch systeem. We maakten daarbij gebruik van radioactief gelabeld [ 11C]-(R)-PK11195, een tracer om geactiveerde microglia weer te geven. Vergeleken met gezonde personen was bij patiënten met een bipolaire stoornis een statistisch verhoogde [11C]-(R)-PK11195 bindingspotentiaal in de rechter hippocampus zichtbaar en een soortgelijke, maar statistisch niet betekenisvolle, verhoging in de linker hippocampus. Deze bevinding wijst op een verhoogde activiteit van het immuunsysteem (neuroinflammatie) op deze locatie in het brein van deze patiënten. Daarna hebben we de hippocampi verder onderzocht, waarbij het volume, de concentratie van verschillende biochemische stoffen (metabolieten) en opnieuw microglia-activatie aan de orde kwamen. Met behulp van specifieke software (Freesurfer) kon met klassieke MRI-scans het volume bepaald worden van beide hippocampi (links en rechts). Deze zogeheten volumetrische analyses toonden geen verschil in volumes tussen de hippocampi van beide groepen proefpersonen, na correctie voor individuele variaties in de totale breingrootte. Met behulp van magnetic resonance spectroscopy (MRS), een MRI-techniek, werd de concentratie van specifieke metabolieten in een vooraf bepaald deel van de hippocampus onder zocht. Hierbij zijn met name de concentraties gemeten van de volgende stoffen: N-acetylaspartaat en N-acetyl-aspartyl-glutamaat (NAA+NAAG, een marker die aangeeft of zenuwcellen intact zijn), creatine en creatinefosfaat (Cr+PCr, een marker voor de celenergiehuishouding) en cholinefosfaat (PCh) en glycerofosfocholine (PCh+GPC, een marker voor de afbraak van celmembranen). Deze analyses toonden aan dat de linker hippocampus van patiënten een lagere NAA+NAAG-concentratie bevatte dan die van gezonde mensen. De daaropvolgende verkennende analyses zijn uitgevoerd volgens een a priori (van tevoren) vastgesteld analysemodel en toonden een positief verband tussen microglia-activatie en de NAA+NAAG-concentratie in de linker hippocampus. Dit Dutch / summary samenvatting Nederlandstalige gaf voor het eerst aanwijzingen uit een onderzoek bij levende patiënten met een bipolaire stoornis dat er een relatie bestaat tussen microglia-activatie en het intact zijn van de zenuwcellen. Daarnaast was een positief verband zichtbaar tussen alcoholgebruik en de NAA+NAAG-concentratie en tussen microglia-activatie en de mate van depressiviteit. Er was een negatief verband tussen de Cr+PCr-concentratie en arbeidsongeschiktheid bij patiënten. In beide hippocampi was ook een positief verband te zien tussen ziekteduur en het hippocampusvolume. In de laatste studie onderzochten we de microstructuur van de witte stof van het brein; dit is het deel van het centraal zenuwstelsel dat axonen bevat, die de informatieoverdracht in de hersenen verzorgen. Deze structuur is onderzocht door middel van DTI. In deze studie is gekeken naar de verschillen hierin tussen patiënten

231 met een bipolaire stoornis en gezonde mensen en naar verschillen die te wijten zijn aan lithiumgebruik. Er bleek geen verschil te bestaan in FA, MD, RD en AD bij patiën- ten met een bipolaire stoornis en gezonde mensen, terwijl in andere studies wel afwijkingen zijn gevonden; dit kan mogelijk verklaard worden doordat de patiënten op het moment van onderzoek geen of weinig stemmingsklachten hadden. Bovendien zijn er wel wijdverspreide clusters gevonden die een verband toonden met (beperkte) depressieve symptomen: negatieve met FA en positieve met RD en MD. Verdere analyses lieten onder meer zien dat in bepaalde hersengebieden, namelijk de hersenbalk en de linker corona radiata, van patiënten die lithium gebruikten een hogere FA en een lagere RD bevatten, vergeleken met patiënten die geen lithium gebruikten. Deze verschillen zouden verklaard kunnen worden doordat lithium mogelijk een herstellend effect heeft op de microstructurele afwijkingen van de verbindingsbanen bij mensen met een bipolaire stoornis.

Discussie

Psychiatrische stoornissen hebben niet dezelfde biologische basis als medische diagnoses. Sinds de jaren zestig van de vorige eeuw hebben biologisch-psychiatrische onderzoekers voortdurend geprobeerd om de diagnostische systemen waarmee zij psychiatrische stoornissen vast konden stellen, een robuuste biologische basis te geven. Al deze inspanningen hebben echter nog niet waargemaakt wat ze leken te beloven: er zijn nog steeds geen biomarkers voor diagnostiek en ook zijn er nog niet of nauwelijks nieuwe geneesmiddelen ontwikkeld op basis van de pathofysiologische modellen, die hiervoor beschreven zijn.

Biomarker of disfunctionele processen: faillissement van de biomarkeraanpak?

Huidige biomarkerbenaderingen schieten tekort In 2009 is het Europese MOODINFLAME-project gestart met als voornaamste doel om diagnostische biomarkers voor stemmingstoornissen te ontwikkelen. Vooraf- gaand hadden Padmos et al. (2008) melding gemaakt van een veelbelovende ‘hand tekening’ van een ontstekinggerelateerde genexpressie in monocyten, die uniek was bij patiënten met een bipolaire stoornis. De deelnemende proefpersonen aan deze studie vormden een gemengde groep met euthyme, depressieve en manische patiënten.

232 Tijdens nadere analyses om het bestaan van deze genexpressiehandtekening verder te bewijzen, ontdekten we het volgende: tussen euthyme MOODINFLAME-patiënten met een bipolaire stoornis en gezonde mensen vertoonde deze genexpressie geen opvallende verschillen. Bovendien bleek dat de genexpressie samenhing met stemmingsepisodes (hoofdstuk 4). Ander onderzoek liet zien dat CRP, een andere veelbelovende mogelijke biomarker die de aanwezigheid van psychiatrische aandoeningen vast kon stellen, ook niet geschikt was om het verdere beloop van de stoornis te voorspellen (hoofdstuk 5).

De complexiteit van het brein als verklaring Er zijn verschillende mogelijke verklaringen denkbaar waarom er nog geen neurobiologische markers zijn voor stemmingsstoornissen: de complexiteit van het brein (zowel wat betreft fysiologie als pathofysiologie), moeilijkheden om hersenweefsel te verkrijgen voor onderzoek en technische/praktische beperkingen in het onderzoek. Het brein is bijvoorbeeld slechts beperkt toegankelijk voor stoffen vanuit de rest van het lichaam vanwege de bloed-hersenbarrière, waardoor tracers van de moleculaire beeldvorming (PET/SPECT) moeilijk doordringen. Vanwege de biologische aanleg van het brein zijn deze moeilijkheden waarschijnlijk onoplosbaar. Naast deze onderzoeks- moeilijkheden biedt de complexe structuur van het brein een verklaring waarom het in de psychiatrie tot nu toe zo lastig was om biomarkers te ontwikkelen.

Ruimtelijke complexiteit en de valkuil van de onjuiste dichotomie De processen die bij psychiatrische aandoeningen ontregeld zijn en onvoldoende functioneren, hebben betrekking en invloed op alle fysiologische niveaus van het centrale zenuwstelsel, niet enkel op een onderdeel ervan. Het centrale zenuw stelsel kent vijf fysiologische niveaus: het genetische niveau, het moleculaire niveau, het cellulaire niveau, de neuronale circuits en het fenomenologische niveau. Deze multilevelfysiologie kan ook wel beschouwd worden als de ruimtelijke complexiteit van het brein. Manji en Lenox (2000) gaven hierbij een voorbeeld: om de pathofysio Dutch / summary samenvatting Nederlandstalige logie van de bipolaire stoornis volledig te kunnen begrijpen, moet worden gekeken naar de neurobiologie van alle fysiologische niveaus (figuur 3). Een probleem bij veel van de eenvoudigere varianten van de hierboven beschreven pathofysiologische modellen is dat ze zich vaak beperken tot één fysiologisch niveau in plaats van dat ze een verklaring geven die geldig is op alle of meerdere fysiologische niveaus. Daarnaast worden de ziekteverschijnselen die mogelijke biomarkers zouden kunnen zijn, op elk fysiologisch niveau ook nog eens opgedeeld in ‘wel’ of ‘niet aangedaan’, zoals in het klassieke medische ziektemodel. Echter, omdat het brein zo complex is, zijn ziekteverschijnselen waarschijnlijk meer gradueel van aard en variëren bijvoorbeeld van gezond, licht afwijkend, matig afwijkend tot ernstig afwijkend (onjuiste tweedeling).

233 FIGUUR 3 Fysiologische niveaus bij de bipolaire stoornis

Behavior Cognitive Affective Sensorymotor

Neurotransmitter & neuropeptide s

• Transcription factors Molecular • mRNA stability Transcriptome • Nuclear import/export Susceptibility genes Protective genes

Om de pathofysiologie van de bipolaire stoornis volledig te begrijpen, moet de neurobiologie bekeken worden op verschillende fysiologische niveaus (moleculair, cellulair, neuronale systemen en symptoomniveau). Bcl-2 = B-cell leukemia/lymphoma; BDNF = brain-derived neurotrophic factor; CREB = cAMP response element binding protein; ERK = extracellular receptor-coupled kinase; GSK-3 = glycogen synthase kinase-3; MAP kinase = mitogen- activated protein kinase; MARCKS = myristoylated alanine-rich C kinase substrate; PKC = protein kinase C). (Bron: Manji and Lenox, met toestemming om te herdrukken)

Neuroinflammatie betekent gewoonlijk dat het immuunsysteem in het brein ge activeerd is, wat verondersteld wordt voor te komen bij psychiatrische stoornissen in het algmeen en de bipolaire stoornis in het bijzonder. Ook dit begrip kan beschouwd worden als een vereenvoudiging van een complex systeem, waarin verschillende (stimulerende en remmende) onderdelen ontregeld zijn. Alleen al vanwege de ruimtelijke fysiologische complexiteit, is het wellicht onmoge- lijk om diagnostische biomarkers voor psychiatrische stoornissen te vinden. Toch is het mogelijk zinvol om veranderingen te meten in biologische markers bij patiënten om het effect van een behandeling te voorspellen of te volgen. Op die wijze zouden biologische bepalingen niet zozeer het diagnostische proces vervangen, maar veeleer een aanvulling vormen om de meest geschikte behandeling te kiezen.

234 Multidimensionale psychopathologie: Oversteken bij de Kraepeliniaanse tweedeling De validiteit van huidige en eerdere benaderingen van psychopathologie Het probleem van de onjuiste tweedeling in de huidige pathofysiologische modellen geldt niet alleen voor de biologische observaties zoals hierboven beschreven. Net zoals in het medische ziektemodel worden psychiatrische stoornissen ook geschei- den benaderd in de huidige diagnostische systemen: mensen zijn ziek en worden dan ‘patiënten’ genoemd, ofwel ze zijn niet ziek, dus ‘gezond’. Door patiënten te obser- veren, hebben clinici prototypische (karakteristieke) patronen herkend in de manier waarop ziekteverschijnselen zich voordoen, waarmee ze vervolgens prototypische diagnoses konden beschrijven. Deze zijn geordend in verschillende categorieën, die uiteindelijk de basis hebben gevormd voor diagnostische classificatiesystemen, zoals de Diagnostic Statistical Manual (DSM-III, DSM-IV en DSM-5) en de International Classification of Diseases (ICD-9 en ICD-10). Hoewel nieuwere versies van deze systemen wel betrouwbaarder zijn dan eerdere, blijft de onderlinge betrouwbaarheid tussen verschillende gebruikers van deze systemen nog steeds moeizaam. Daarnaast houden ze geen rekening met de mate en ernst van individuele ziekteverschijnselen en het beloop van de klachten.

De Kraepeliaanse tweedeling Emil Kraepelin (1856-1926) was de eerste die het onderscheid maakte tussen schizofrenie en de bipolaire stoornis, een onderscheid dat later bekend is geworden als de Kraepeliaanse tweedeling (of dichotomie). Deze tweedeling werd lang gehanteerd voor de behandeling van patiënten, omdat de psychiater ermee tot een heldere diagnose kon komen. Bovendien gaf dit onderscheid sturing aan de behandeling: er werden antipsychotica voorgeschreven bij schizofrenie en stemmingsstabilisatoren bij de bipolaire stoornis. Toch waren er problemen. In de eerste plaats bestaat er geen zeldzaamheidspunt tussen beide diagnoses, dat wil zeggen dat er ook veel patiënten zijn die symptomen van beide stoornissen hebben, waardoor er een extra diagnose Dutch / summary samenvatting Nederlandstalige kwam voor het overgangsgebied tussen schizofrenie en de bipolaire stoornis: de schizo-affectieve stoornis. Ten tweede hebben de antipsychotica de Kraepeliaanse tweedeleing ‘overgestoken’: zij bleken ook effectief bij de bipolaire stoornis. Naast deze klinische barsten in de Kraepliaanse tweedeling is er tegenwoordig ook wetenschappelijk bewijs dat de twee aandoeningen gezamenlijke pathofysiologische mechanismen hebben. Bijna alle pathofysiologische modellen bij de bipolaire stoornis kennen ook varianten bij schizofrenie en de depressieve stoornis. Grote genetische genoomwijde associatiestudies hebben aangetoond dat er enkelvoudige nucleotide polymorfismen (SNP’s) bestaan, die het risico op zowel schizofrenie als de bipolaire stoornis beïnvloeden. Drexhage et al. (2010) hebben bovendien laten zien dat de ontstekinggerelateerde genexpressie in monocyten niet alleen bij mensen met een

235 bipolaire stoornis toegenomen is, maar ook bij mensen met schizofrenie. Het is echter te kort door de bocht om te veronderstellen dat de bipolaire stoornis en schizofrenie dezelfde aandoening zijn. Zo bleek uit het onderzoek van Drexhage et al. dat er wel enig onderscheid is in de genexpressie. Daarnaast vertoonden kinderen die later schizofrenie ontwikkelden al voordat hun stoornis zichtbaar was cognitieve en neuromotorische beperkingen, die niet (of in elk geval niet in dezelfde mate) voor komen bij kinderen die later een bipolaire stoornis ontwikkelden. Ten slotte zijn ook herhaaldelijk neuroanatomische verschillen tussen de bipolaire stoornis en schizofrenie aangetoond; de volumina van de hippocami bijvoorbeeld zijn bij schizofrenie meestal verkleind, terwijl ze bij mensen met een bipolaire stoornis meestal niet te onderscheiden zijn van die van gezonde mensen.

Dimensionale diagnostiek in het onderzoeksveld en de klinische praktijk Meer dan tien jaar geleden kwamen onderzoekers met het voorstel om de psychia trische diagnostiek dimensionaler te benaderen, in plaats van de huidige benadering die categoriaal is (opgedeeld ‘in hokjes’). Dat wil zeggen dat in een dimensionaal model de psychiatrische symptomen op meerdere dimensies zoals manie, depressiviteit, psychose, enz., ingeschaald worden. Hierdoor kunnen psychiatrische symptomen beter gelinkt worden aan de onderliggende pathofysiologie. Om deze benadering kracht bij te zetten, organiseerde het nationale onderzoeksinstituut van de Verenigde Staten (NIMH) het Research Domain Criteria (RDoC)-project om psychische aan doeningen dimensionaal te classificeren. Inmiddels proberen onderzoekers in de genetica, de neurobiologie en de epidemiologie het gebruik van een dimensionale diagnostische benadering uit, onder meer om de validiteit (correctheid) van het diagnostische systeem te vergroten. In de praktijk geven clinici vooralsnog de voorkeur aan de huidige categoriale benadering van DSM-5 en ICD-10, wat verklaarbaar is. De huidige categoriale benadering werkt met een internationaal uniform en breed gedragen classificatiesysteem en heeft ontegenzeggelijk geholpen bij de ontwikkeling van en de kennis over de meeste, zo niet alle huidige psychiatrische behandelmogelijkheden. Wanneer het diagnostische systeem zou moeten veranderen van de huidige categoriale naar een dimensionale benadering, zou dat een grote paradigmaverandering betekenen. De verbanden tussen symptomen en de onderliggende genetische, moleculaire en cellulaire niveaus en neuronale circuits moeten dan opnieuw worden onderzocht. Daarnaast moet ook opnieuw worden bekeken in hoeverre de dimensionale diagnose kan voorspellen wat het effect is van zowel nieuwe, toekomstige behandelingen als de huidige, bestaande behandelingen, om ze aan te laten sluiten op de dimensionale benadering. Net als de onderzoekers dienen ook de professionals in de patiëntenzorg de ontwik- kelingen in de dimensionale diagnostiek in het oog te houden en ervoor open te staan dat het diagnostische systeem op een graduele, geleidelijke manier dimensionaler

236 van aard wordt. Een herwaardering van het kwetsbaarheidsconcept (vulnerability) in de diagnostische systemen kan hierbij helpen. Met de bestaande systematiek worden in de psychiatrie onder meer patiënten met een psychotische en met manische kwetsbaarheid herkend. Aan de hand van het kwetsbaarheidsconcept worden patiënten met een manische kwetsbaarheid bijvoorbeeld als volgt herkend: het zijn mensen met eerstegraads familieleden die een bipolaire stoornis hebben, maar die nog geen klachten hebben en die eerder in hun leven manische symtomen (maar nog geen manische of hypomane episodes) hebben doorgemaakt. Zij hebben een verhoogd risico om manische of depressieve episodes te ontwikkelen. Het voornaamste doel van de preventieve behandeling van deze patiënten is om de effecten van deze kwetsbaarheid zoveel mogelijk te beperken, door hun weerbaarheid (resilience) zoveel mogelijk te vergroten, bijvoorbeeld in de vorm van psycho-educatie of training om beter met stress om te kunnen gaan. Het kwetsbaarheidsconcept omvat de ontwikkelingskenmerken van de stoornissen beter en maakt het door het begrip ‘kwetsbaarheid’ waarschijnlijk voor mensen ook gemakkelijker om te accepteren dat ze een psychiatrische aandoening kunnen krijgen (of al hebben).

De bipolaire stoornis als een gliale ontwikkelingsstoornis met een laat-klinische presentatie

Neuroglia, de vergeten werkers van het brein Van oudsher is de biologische psychiatrie vooral gericht op het neuronale deel van het centraal zenuwstelsel, dus met name op breinactivatie en neurotransmissie tussen zenuwcellen(neuronen), waarbij de andere cellen, de neuroglia, over het hoofd worden gezien (figuur 4). Er werd lang gedacht dat gliale cellen enkel ondersteunende cellen waren die de neuronen op hun plaats hielden, hen verzorgden met voedingsstoffen en zuurstof en die ziektekiemen bestreden. Echter, rond de laatste eeuwwisseling is aangetoond dat gliale cellen van belang zijn voor de ontwikkeling van het brein en het Dutch / summary samenvatting Nederlandstalige functioneren van synapsen. Gliale cellen bestaan voornamelijk uit zogeheten oligodendroglia, astrocyten en microglia. Oligodendroglia vormen myelineschedes rond neuronale axonen om deze te ondersteunen en de neuronale overdrachtssnelheid te verbeteren. Daar- naast bevorderen ze neuronale groei, doordat ze verschillende groeistimulerende stoffen (groeifactoren) afgeven. Astrocyten hebben uiteenlopende functies in het menselijk brein. Ze spelen een rol als energiebuffer, bij het reguleren van de bloed- toevoer, bij het stimuleren van de myelinisatie door oligodendroglia, bij het herstel van het zenuwstelsel, de versterking van de langetermijnpotentiëring van synapsen (functie bij geheugen) en bij de versterking of verzwakking van verschillende soorten signaaloverdracht. Microglia zijn de lokale macrofagen (opruimende witte

237 FIGUUR 4 Neuroglia

Verschillende gliacellen in het brein: microglia, astrocyten en oligodendrogliacellen. (Bron: aanpassing van originele tekening van OpenStax College op de Anatomy & Physiology, Connexions Website https://cnx.org/contents/[email protected]:fEI3C8Ot@10/Preface onder de Creative Commons Attribution 4.0 Generic licentie (CC BY 4.0))

bloedcellen) van het brein en het ruggenmerg en vormen de voornaamste immuun- afweer van het centraal zenuwstelsel. Daarnaast zijn microglia ook van belang om het interne evenwicht (homeostase) te behouden. Uit diermodellen blijkt dat, net zoals bij perifere macrofagen, ook microglia geactiveerd kunnen worden op een ontstekingsreactiebevorderende wijze (M1-macrofagen), een ontstekingsreactie- remmende wijze (M3-macrofagen) en een herstellende/weefselondersteunende wijze (M2b-macrofagen). Ook kunnen sommige microglia aanzetten tot apoptose (geprogrammeerde celdood), terwijl andere juist actief neurogenese (celvorming) stimuleren.

Gliale ontregeling en neuroplasticiteit bij neurodegeneratieve ziekten en psychiatrische stoornissen In de afgelopen jaren zijn de afzonderlijke onderdelen van het neurogliasysteem uit- gebreid onderzocht bij neurodegeneratieve aandoeningen, ontwikkelingsstoornissen en psychiatrische aandoeningen.

238 In dit proefschrift hebben we aangetoond dat bij patiënten met een bipolaire stoornis de microglia in de rechter hippocampus meer geactiveerd zijn dan bij gezonde con- trolepersonen (hoofdstuk 7). Daarnaast werd duidelijk dat er een positief verband is tussen microglia-activatie en het intact zijn van de neuronen; dit onderzoek onder proefpersonen sloot aan bij het microgliale functiedifferentiatieprincipe wat reeds bekend was uit proefdiermodellen (hoofdstuk 8). Neuroplasticiteit is het vermogen van het brein om zich tijdens de volwassen- heid toch aan te kunnen passen en is een bruikbaar begrip om gliale ontregeling bij psychiatrische aandoeningen beter te begrijpen. Het menselijk brein is uitzonderlijk plastisch, waardoor het aanpassingsvermogen tot stand komt dat voor de mens zo kenmerkend is. Recent onderzoek heeft laten zien dat evolutionaire aanpassingen en genetische variatie van astrocyten gekoppeld is aan de complexiteit van het menselijk brein, in vergelijking met andere (dier)soorten. Daarnaast bleken patiënten met een stemmingsstoornis een verminderde gliale celdichtheid te hebben in de cortex cingularis anterior supragenualis; onder andere gliale ontregeling zou daarom kunnen leiden tot een verminderde neuroplasticiteit bij mensen met een bipolaire stoornis. Het is in dit geval opmerkelijk dat lithium, valproïnezuur en antidepressiva indirect celoverleving stimuleren en mede daardoor op positieve wijze bijdragen aan de langetermijneffecten op de neuroplasticiteit van de hersenen.

Temporele complexiteit en de bipolaire stoornis als een ontwikkelingsstoornis Zoals hierboven uiteengezet werd, sturen gliacellen de neuronale ontwikkeling aan. Neuronale ontwikkeling is het proces waardoor in een groeiend jong lichaam tot een volwassen zenuwstelsel gekomen wordt, dat om kan gaan met complexe taken. Deze ontwikkeling kan beschouwd worden als de temporele complexiteit van het brein, die een andere vorm van breincomplexiteit is dan de eerder genoemde ruimtelijke complexiteit. Over schizofrenie is bekend dat deze aandoening gepaard gaat met neuronale ontwikkelingsproblemen en lange tijd is gedacht - en als onderscheidend kenmerk Dutch / summary samenvatting Nederlandstalige gehanteerd - dat mensen met een bipolaire stoornis deze niet hebben. De laatste tijd is echter overtuigend wetenschappelijk bewijs geleverd dat er ook bij hen ontwikkelingsproblemen spelen. Zo is bij deze patiënten het totale breinvolume gemiddeld kleiner dan bij gezonde mensen, zowel bij volwassenen als bij adolescenten. Daarnaast zijn er aanwijzingen voor afwijkingen in de celmigratie in het brein van patiënten met een bipolaire stoornis. Het door ons gevonden mogelijk verband tussen een vroegere beginleeftijd en een toegenomen ontstekinggerelateerde genexpressie (hoofdstuk 2) is in overeenstemming met andere onderzoeken die lieten zien dat patiënten met een vroegere beginleeftijd meer last hebben van ziekteverschijnselen, wat dit verder ondersteunt. Mesman et al. (2015) onderzocht kinderen van ouders met een bipolaire stoornis, die zelf een verhoogde kwetsbaarheid hebben

239 om dezelfde stoornis te ontwikkelen. Zij toonde aan dat de ontstekinggerelateerde genexpressie in monocyten en de PTX3- en CCL2-concentraties in het bloed van deze kinderen dynamisch verlopen in de periode van adolescentie naar volwassen- heid. Uit deze bevindingen kan verondersteld worden dat een ontregelde neuronale organisatie, die voortkomt uit gliale ontregeling, kan veroorzaken dat mensen kwets- baar zijn voor de bipolaire stoornis en dat deze voorafgaat aan het ontstaan ervan. Bij de bipolaire stoornis slaagt het brein er waarschijnlijk in om deze ontregeling tijdelijk te overwinnen, door compensatiemechanismen die de weerbaarheid in stand houden. Later, wanneer een patiënt volwassen wordt, blijken deze compensatiemechanismen echter steeds meer tekort te schieten, waardoor manische en depressieve – en mogelijk ook psychotische – episodes ontstaan (figuur 5).

FIGUUR 5 Psychiatrisch en biologisch beloop bij de bipolaire stoornis

Schematische weergave van de activatie van het inflammatoir response systeem (IRS) en de ontwikkeling van psychiatrische symptomen tijdens het beloop van de bipolaire stoornis.

240 Slotopmerkingen

Door de onderzoekstechnieken die de afgelopen decenia beschikbaar zijn gekomen en door de resultaten en inzichten die ze al hebben opgeleverd, is het momenteel een zeer interessante tijd om psychopathofysiologisch onderzoek te verrichten en onze gezamenlijke kennis voort te stuwen met het uiteindelijke doel om de pathofysiologie en daarmee het ontstaan van de bipolaire stoornis beter te begrijpen en beter te weten hoe de bipolaire stoornis te behandelen. Dit met als uiteindelijk doel het om leven van patiënten met deze stoornis te verbeteren. De afgelopen jaren heeft wetenschappelijk bewijs duidelijk gemaakt dat de bipolaire stoornis bekeken moet worden vanuit een neuronaalontwikkelings- en transdimen- sionaal perspectief, om daarmee het ontstaan, de behandeling en het beloop ervan beter te begrijpen. Begrip van de ruimtelijke en temporele complexiteit van het brein en de gevolgen hiervan op het denken over de pathofysiologie van de bipolaire stoornis, is noodzakelijk om verdere vooruitgang te boeken in dit onderzoeksveld en daarmee het ontregelde brein te ontwarren. Dutch / summary samenvatting Nederlandstalige

241

List of publications Peer reviewed publications

Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation-related monocyte gene expression in bipolar disorder - towards a better understanding of psychoimmunological interactions. Bipolar Disord. 2014 Mar 29;16(2):137–50.

Becking K, Boschloo L, Vogelzangs N, Haarman BCM, Riemersma-van der Lek R, Penninx BWJH, et al. The association between immune activation and manic symptoms in patients with a depressive disorder. Transl Psychiatry. Nature Publishing Group; 2013 Jan;3(10):e314.

Haarman BCM, Riemersma-Van der Lek RF, Nolen WA, Mendes R, Drexhage HA, Burger H. Feature-expression heat maps – A new visual method to explore complex associations between two variable sets. J Biomed Inform. 2015 Oct 14;53:156–61.

Haarman BCM, Riemersma-Van der Lek RF, de Groot JC, Ruhé HG, Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder - A [(11)C]-(R)-PK11195 positron emission tomography study. Brain Behav Immun. 2014 Aug 3;40:219–25.

Haarman BCM, Burger H, Doorduin J, Renken RJ, Sibeijn-Kuiper AJ, Marsman J-BC, et al. Volume, metabolites and neuroinflammation of the hippocampus in bipolar disorder - A combined magnetic resonance imaging and positron emission tomo graphy study. Brain Behav Immun. Elsevier Inc.; 2015 Sep 5;60(1):1–5.

Becking K, Haarman BCM, van der Lek RFR, Grosse L, Nolen WA, Claes S, et al. Inflammatory monocyte gene expression: trait or state marker in bipolar disorder? Int J Bipolar Disord. 2015 Dec 17;3(1):20.

Haarman BCM, Riemersma-Van der Lek RF, Burger H, de Groot JC, Drexhage HA, Nolen WA, et al. Diffusion tensor imaging in euthymic bipolar disorder - A tract-based spatial statistics study. J Affect Disord. 2016 Oct 1;203:281–91.

Balukova SM, Haarman BCM, Riemersma-van der Lek RF, Schoevers RA. Does CRP predict outcome in bipolar disorder in regular outpatient care? Int J bipolar Disord. 2016 Dec;4(1):14.

Haarman BCM, Riemersma-Van der Lek RF, Burger H, Drexhage HA, Nolen WA. The dysregulated brain: consequences of spatial and temporal brain complexity for bipolar disorder pathophysiology and diagnosis. Bipolar Disord [Internet]. 2016 Dec 20.

244 Book chapter

Haarman BCM, Riemersma - Van der Lek RF, Ruhé HG, Groot JC, Nolen WA, Doorduin J. Bipolar Disorders. In: Dierckx RAJO, Otte A, de Vries EFJ, Waarde A, den Boer JA, editors. PET and SPECT in Psychiatry. Heidelberg: Springer Berlin Heidelberg; 2014. p. 223–51.

Frey BN, Minuzzi L, Haarman BCM, Sassi RB. Neuroimaging and illness progression. In: Kapczinski F, Berk M, editors. Staging and Neuroprogression in Bipolar Disorder. OUP Oxford; 2014. p. 145–74. Conference presentations

Haarman BCM, Riemersma–Van der Lek RF, Burger H, Horňáková Š, Renken RJ, Sibeijn-Kuiper AJ, Nolen WA. Creatine alterations in the hippocampus in bipolar I disorder – A 3T 1H-MRS study. Poster presentation on the biennial conference of the International Society for Bipolar Disorders (ISBD), Istanbul, Turkey 2012.

Haarman BCM, Riemersma-Van der Lek RF, Burger H, Netkova M, Drexhage RC, Bootsman F, et al. Relationship between clinical features and inflammation related monocyte gene expression in bipolar disorder. Oral presentations on the biennial International Conference for Bipolar Disorder (ICBD), Miami, Florida, USA 2013 and the annual Spring congress of Dutch Society of Psychiatry, Maastricht, The Netherlands 2013.

Haarman BCM. Neurobiology and imaging techniques of bipolar disorder. Oral presenation on the 10th Autumn Symposium Bipolar Disorders, Utrecht, The Netherlands 2013.

Haarman BCM, Riemersma-Van der Lek RF, de Groot JC, Ruhé HG, Klein HC, Zandstra TE, et al. Neuroinflammation in bipolar disorder – a [11C]-(R)-PK11195 PET study. Oral presentations on the annual conference of the International Society for Bipolar Disorders (ISBD), Seoul, Korea 2014 and the annual conference of the European College of Neuropsychopharmacology (ECNP), Berlin, Germany 2014.

Haarman BCM, Burger H, Doorduin J, Renken RJ, Sibeijn-Kuiper AJ, Marsman J-BC, et al. Volume, metabolites and neuroinflammation of the hippocampus in bipolar disorder. Poster presentation on the annual conference of the International Society List of publications of List for Bipolar Disorders (ISBD), Toronto, Canada, 2015 and oral presentation on the annual conference of the International Society for Bipolar Disorders (ISBD), Amsterdam, The Netherlands 2016.

Haarman BCM. Neurobiology of bipolar disorder – the dysregulated brain. Oral presenation on the 13th Autumn Symposium Bipolar Disorders, Utrecht, The Netherlands 2016.

245

Acknowledgements / Dankwoord In oktober 2008 ging mijn lang gekoesterde wens in vervulling om opnieuw wetenschappelijk onderzoek te gaan doen. Deze keer in de vorm van een promotie traject binnen het destijds nieuwe MOODINFLAME-project. Bij de kick-off meeting in november van dat jaar leerde ik direct al meerdere bevlogen collega’s kennen en, hoewel er zoals bij elk project hoogte- en dieptepunten waren, is mijn enthousias- me om op samenwerkende wijze wetenschappelijk onderzoek te bedrijven niet meer weggegaan.

Terugkijkend voel ik vooral veel dankbaarheid. Dankbaarheid voor de kans om dit te mogen doen, voor de plezierige samenwerking, voor de voldoening die het geeft om soms op de cutting edge nieuwe kennis te vergaren en dat ik dit succesvol heb kunnen afronden. Met dit proefschrift hoop ik enkele stukjes aan de neurobiologische puzzel van de bipolaire stoornis toe te hebben gevoegd, waarmee uiteindelijk een duidelijker beeld ontstaat van de oorzaken en werkingsmechanismen van deze ernstige aan doening.

Om te beginnen gaat mijn dank uit naar alle patiënten en gezonde controles die deel hebben genomen aan de studies die ten grondslag liggen aan dit proefschrift. Het is geen sinecure om zonder eigen belang zoveel informatie over jezelf te geven en om, soms moeilijke, onderzoeken te ondergaan.

Daarnaast heb ik zeer veel steun gehad van mijn promotores, prof. dr. Willem Nolen en prof. dr. Hemmo Drexhage. Willem gaf me het vertrouwen en de unieke kans om aan dit project te beginnen. Hij daagde me op verschillende domeinen uit om te groeien, niet alleen als wetenschapper, maar ook als mens. Ook zorgde hij voor de stabiele ondersteuning om het onderzoeksproject tot een succes te kunnen maken. Van Hemmo leerde ik om scherp na te denken over het doel en de boodschap van verschillende geschreven stukken en ik kon gebruik maken van zijn uitgebreide kennis over de immunologie. Beide promotores leerden mij over de verschillende facetten van wetenschappelijk discussiëren: wanneer rekkelijkheid op zijn plaats is en wanneer zorgvuldige precisie wenselijker is.

Mijn twee copromotores, dr. Rixt Riemersma – Van der Lek en dr. Huib Burger, hebben mij geholpen om na te denken over het wetenschappelijk werk en de medische statistiek, maar juist ook over andere zaken zoals het politieke bedrijf en filosofische thema’s. Rixt ben ik gaan beschouwen als een echte vriendin en ik denk met plezier terug aan onze verschillende bezoeken aan congressen en symposia. Huib heeft me aanzienlijk geholpen om inzichtelijk en kritisch naar medische statistiek te kijken en zijn kritische en onafhankelijke reflectie waren daarbij onmisbaar, bovendien kon ik dankzij zijn toegankelijkheid met hem sparren over nieuwe, soms wat wilde, analyse-ideeën.

248 Prof. dr. Robert Schoevers is een belangrijke initiator geweest in de totstandkoming van het CRP-onderzoek, samen met Sonya Balukova. Ook was hij met Karlijn Becking nauw betrokken bij de MOODINFLAME-analyses, tevens onderdeel van dit proefschrift en waar momenteel nog aan een vervolg gewerkt wordt. Met Robert heb ik ook veel prettige gesprekken over het werk als klinisch-wetenschappelijk georiënteerd academisch medisch specialist.

Wetenschappelijk onderzoek komt tot stand met medewerking van veel collega’s. Juliëtte Kalkman heeft een belangrijke bijdrage geleverd aan de inclusie van de deelnemers aan de MOODINFLAME-studie in Groningen en heeft de deelnemers op steunende en betrokken manier begeleid bij de vaak intensieve scanonderzoeken. De medewerkers van het Neuroimaging Center: Anita Sibeijn - Kuiper, dr. Remco Renken, Judith Streurman en dr. Jan Bernard Marsman hebben intensief geholpen bij de totstandkoming van het MRI-gedeelte van het onderzoek. Dr. Janine Doorduin en dr. Erik de Vries, met behulp van medewerkers van de afdeling Nucleaire genees kunde en Moleculaire beeldvorming van het UMCG, hebben het PET-gedeelte tot een succes gemaakt. Dr. Jan Cees de Groot van de afdeling Radiologie heeft geholpen om de wetenschappelijke vraagstelling en de methoden van de verschillende beeld- vormingsonderzoeken aan te scherpen. De bijdrage van dr. Leonardo Cerliani aan de DTI-analyses was van bijzondere waarde. Het is een waar genoegen om met deze kundige collega’s samen te kunnen werken.

Binnen MOODINFLAME heb ik met verschillende collega’s uit Nederland en Europa gewerkt, wat er van het begin af aan bijdroeg om mijn horizon te verbreden. Behalve de betrokken medewerkers van universiteiten en instellingen uit onder andere München, Innsbruck, Leuven, Londen, Dublin, Milaan en Parijs wil ik met name Angelique van Rijswijk, Harm de Wit en Annemarie Wijkhuijs van de Erasmus Universiteit Rotterdam noemen, die een aanzienlijke bijdrage hebben geleverd aan de organisatie van MOODINFLAME en de analyse van de bioassays. Gesprekken met dr. Veerle Bergink en dr. Manon Hillegers van de afdelingen psychiatrie van respectievelijk de Erasmus Universiteit en UMC Utrecht leerden mij over de

ontwikkelingsproblematiek bij de bipolaire stoornis. Met Laura Grosse van de West- / Dankwoord Acknowledgements faalse Wilhems-Universiteit uit Münster werkten Karlijn Becking en ik goed samen aan vergelijkende analyses tussen de bipolaire stoornis en de unipolaire depressieve stoornis. Naast andere genoemden, zijn prof. dr. Volker Arolt van diezelfde universiteit, dr. Francesco Benedetti van het San Raffaele ziekenhuis Milaan, prof. dr. Stephan Claes van de Katholieke Universiteit Leuven en prof. dr. Dirk de Ridder, momenteel werkzaam bij de Universiteit van Otago, voor mij een bron van inspiratie geweest.

249 Mijn dank gaat tevens uit naar de overige co-auteurs van de stukken in dit proefschrift: dr. Florian Bootsman, dr. Roos Drexhage, dr. Erik Hoencamp, Richard Mendes, dr. Esther Mesman, Mina Netkova, dr. Eric Ruhé, dr. Annette Spijker en Tjitske Zandstra. Marieke van der Vliet heeft op uiterst zorgvuldige en vakkundige wijze het grafisch ontwerp van dit proefschrift verzorgd.

Naast het onderzoekswerk, bestaat een belangrijk deel van mijn werk uit klinische patiëntenzorg. Zeer plezierig vind ik de samenwerking met mijn betrokken collega-psychiaters. Dr. Marrit de Boer, dr. Chris van der Gaag, dr. Marieke Eldering, Jérôme Schuch en Onno Habekotté wil ik hier in het bijzonder noemen. Marrit en Chris zijn mijn paranimfen. Zij hebben mij de afgelopen jaren bijgestaan en waren getuige van ingrijpende gebeurtenissen in mijn leven: ik ben dan ook ontzettend verheugd dat ik mijn proefschrift mag verdedigen met hen aan mijn zijde. Met Chris, Marieke en Dorien Laan, nu werkzaam in de kinderen jeugdpsychiatrie, heb ik de nodige tijd doorgebracht tijdens de specialisatie en ik beschouw hen als zeer goede vrienden. Nog steeds koesteren we onze jaarlijkse weekendjes weg en etentjes samen. Marieke is mij daarnaast met raad en daad tot steun geweest tijdens het begin van mijn onderzoek.

Op de afdeling Opname Depressie werk ik samen met een uiterst betrokken en professioneel team. Mijn waardering gaat in het bijzonder uit naar de arts-assistenten die ik de afgelopen jaren heb mogen begeleiden en de (waarnemend) regieverpleeg- kundigen: Rachel Oziël, Petra Homan - Stokker, Annelies Luining – Nieman en Sonja Boerma.

De periode van mijn promotieonderzoek omspande daarnaast een cruciale fase in mijn leven. Vrijwel in het begin, in 2009, werd mijn lieve moeder onverwacht ernstig ziek en kwam te overlijden. Samen met mijn vader heeft zij ervoor gezorgd dat ik terug kan kijken op een zeer warmhartige en waardevolle jeugd en dat ik ben geworden wie ik ben, en wil zijn. In mijn werk als psychiater zie ik vaak dat het ook anders kan. Ik heb het ontzettend getroffen met mijn lieve, spontane zussen Marlies en Sabine, op wie ik altijd kan rekenen. De afgelopen jaren ben ik getrouwd, heb ik twee prachtige kinderen gekregen en heb ik mijn specialisatie afgerond. Het is een voorrecht om getrouwd te mogen zijn met mijn liefdevolle, mooie en verstandige vrouw, Karin Heijneman. Onze prachtige kinderen, Jelte en Yfke, te mogen zien opgroeien is werkelijk bijzonder en het meest waardevolle in ons leven.

250 Acknowledgements / Dankwoord Acknowledgements

251

Curriculum vitae 254 Bartholomeus Cornelius Maria (Benno) Haarman (1978) is a Researcher at the University of Groningen, the Netherlands, and a Psychiatrist and Head of the Mood and Anxiety Disorders Program of the Department of Psychiatry, University Medical Center Groningen. Before obtaining his psychiatry license in 2012, he took his residency training in psychiatry at this hospital.

In October 2008, he started his PhD candidacy under the supervision of Prof. Willem Nolen, Prof. Hemmo Drexhage, Dr. Rixt Riemersma–Van der Lek and Dr. Huib Burger at the University of Groningen, the Netherlands. His research focuses on the pathophysiological mechanisms of bipolar disorder with an emphasis on neuroinflammation. He participated in the MOODINFLAME study, where he coordinated the inclusion of the bipolar cohort, performed a combined positron emission tomography/magnetic resonance imaging study and analyzed monocyte gene expression. His publications include articles on the first neuroinflammation positron emission tomography study in bipolar disorder and on the association between monocyte gene expression and specific psychiatric symptomatology in bipolar disorder, using a novel approach. This last publication earned him the Samuel Gershon award of the Inter national Society for Bipolar Disorders (2013).

Benno graduated from his medical studies with great honor in 2006 at the University Antwerp, Belgium. While studying medicine, he performed his master thesis on the topic of microvascular compression of the cochleovestibular nerve in the Neuro surgery Department of the Antwerp University Hospital, under the supervision of Prof. Dirk de Ridder. He also obtained his teaching certificate in biology.

Besides being passionate in his work, Benno enjoys playing guitar and listening to music, doing photography, running, and reading news, opinion pieces and general history. Benno is married to Karin Heijneman. Together they have two children, Jelte (2010) and Yfke (2014). Curriculum vitae

255