Genetic Vulnerability to Psychosis and Cortical Function: Epistatic Effects between DAAO and G72

Genetic Vulnerability to Psychosis and Cortical Function:

Epistatic Effects between DAAO and G72

Mechelli Andrea1, Prata Diana1,2, Papagni Sergio Alessandro1,3, Tognin Stefania1, Kambeitz

Joseph1, Fu Cynthia1, Picchioni Marco1,4, Walshe Muriel1, Toulopoulou Timothea1, Bramon

Elvira1, Murray Robin1, McGuire Philip1

1. Department of Psychosis Studies, Institute of Psychiatry, King’s College London, De

Crespigny Park, SE5 8AF U.K.

2. Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, King's College

London, De Crespigny Park, SE5 8AF U.K.

3. Section of Psychiatry and Clinical Psychology, Department of Medical Sciences, University

of Foggia, Foggia, Italy.

4. St Andrews Academic Centre, Kings College London, Institute of Psychiatry, Northampton,

NN1 U.K.

Corresponding Author: Stefania Tognin, PO BOX 67, Department of Psychosis Studies, Institute of

Psychiatry, King’s College London, De Crespigny Park, London SE5 8AF, Email:

[email protected]

Keywords: G72, DAAO, glutamate, schizophrenia, bipolar disorder, functional magnetic resonance

imaging, vulnerability.

Abstract

2

Recent studies have described G72 and DAAO as susceptibility genes for schizophrenia and bipolar

disorder. Both genes modulate glutamate neurotransmission, which plays a key role in neurocognitive

function and is thought to be altered in these disorders. Moreover, in vitro transcription studies indicate

that the two genes interact with each other at the molecular level. However, it is unclear how these

genes affect cortical function and whether their effects interact with each other. The aim of this study

was therefore to examine the impact of G72 rs746187 and DAAO rs2111902 genotypes on brain

function in schizophrenia, bipolar disorder and healthy volunteers. We used functional magnetic

resonance imaging and an overt verbal fluency paradigm to examine brain function in a total of 120

subjects comprising 40 patients with schizophrenia, 33 patients with bipolar I disorder and 47 healthy

volunteers. A significant 3 way interaction between G72, DAAO and diagnosis was detected in the

right middle temporal gyrus (x=60 y=-12 z=-12; z-score: 5.32; p<0.001 after family-wise error

correction), accounting for 8.5% of the individual variance in activation. These data suggest that there

is a non-additive interaction between the effects of variations in the genes implicated in glutamate

regulation that affects cortical function. Also, the nature of this interaction is different in patients and

healthy controls, providing support for altered glutamate function in psychosis. Future studies could

explore the effects of DAAO and G72 in individuals with prodromal symptoms of psychosis, in order

to elucidate glutamate dysfunction in this critical phase of the disorder.

3

Introduction

Inheritance accounts for vulnerability to psychosis by 40% to 70% [1, 2]. The effects of risk genes

may be expressed in the brain at molecular and macroscopic level even when they are not evident at

behavioural level. The combination of molecular genetics and neuroimaging (“imaging genetics”)

allows the investigation of the impact of genetic polymorphisms of interest on brain structure and

function, with the aim of understanding the neurobiological basis of vulnerability to psychosis. Several

recent linkage and association studies have identified the brain-expressed genes for G72 and D-Amino

Acid Oxidase (DAAO), located in chromosomal regions 13q32-33 and 12q24 respectively, as probable

susceptibility genes for schizophrenia and bipolar disorder [3-10]. The product of G72 is an activator of

DAAO, which is the only enzyme that oxidises D-serine, a co-agonist for the NMDA glutamate

receptor [11]. Glutamate is the most abundant excitatory neurotransmitter in the human brain and

glutamate neurotransmission plays a key role in neurocognitive function [12]. In healthy volunteers, the

administration of glutamatergic antagonists results in impaired performance on tests of verbal [13], and

nonverbal [14] declarative memory, verbal fluency and problem solving [15]. There is also a growing

body of evidence to suggest that glutamate neurotransmission is altered in psychosis [11, 16-18]. For

instance, glutamate function is perturbed in people with prodromal signs of psychosis, and

glutamatergic dysfunction is associated with a reduction in gray matter volume in brain regions thought

to be critical to the pathogenesis of the disorder [17]. It has been proposed that hypofunction of

glutamate in cortico-striatal projections may lead to the changes in striatal dopamine concentration

which are thought to underlie the emergence of psychotic symptoms [19]. Consistent with this

hypothesis, NMDA receptor antagonists, which are potent activators of dopamine release, can cause

psychotic symptoms in healthy participants and exacerbate psychotic symptoms in patients [20].

To date, the impact of the glutamate-regulating G72 and DAAO genes on neurocognitive function have

been assessed independently. Three studies have found evidence that the G72 gene moderates neuronal

responses in the medial temporal cortex during verbal working memory [21], working memory [22],

episodic memory [23], and semantic memory [24] tasks. Moreover, Prata and colleagues [25]

demonstrated that a variation in G72 genotype modulates the activation of the left postcentral and

supramarginal gyri during a verbal fluency task in a group of healthy participants. The DAAO gene has

been associated with individual differences in spatial working memory in young male military

4

conscripts [26] but did not appear to have a significant impact on cognitive function in a different study

which examined healthy controls, schizophrenic patients and their unaffected siblings [23]. We

recently demonstrated that a genetic variation in DAAO is associated with differences in regional

activation and functional connectivity during a verbal fluency task in schizophrenia patients compared

to healthy participants, suggesting a diagnosis-dependent pattern of gene action [27].

Chumakov and colleagues [3] presented evidence of epistatic interaction between variants at the G72

and DAAO loci in schizophrenia susceptibility using French-Canadian and Russian samples; the

authors suggested that variation at the G72 and DAAO loci might influence efficiency of glutamate

gating of the NMDA ion channel contributing to schizophrenia susceptibility. However, two

subsequent studies have failed to replicate this epistatic effect [4, 28]. More recently, Corvin and

collagues [7] found evidence for epistasis between the G72 and DAAO markers most strongly

associated with schizophrenia in an Irish sample, although these did not correspond to the single

nucleotide polymorphisms (SNPs) reported in the original article by Chumakov and colleagues [3]. The

existence of altered glutamate neurotransmission in psychosis suggests that any epistatic effects

between the G72 and DAAO loci might differ in psychotic patients compared with healthy volunteers.

To date, no neuroimaging studies have investigated epistasis between the G72 and DAAO loci, or the

extent to which this may be altered in psychosis. We therefore examined (i) the existence of a non-

additive interaction between the G72 and DAAO genes on brain activation during a verbal fluency

task, and (ii) the extent to which this interaction is modified in schizophrenia and bipolar disorder. We

used functional Magnetic Resonance Imaging (fMRI) to study healthy volunteers and patients with

schizophrenia or bipolar disorder, with genotype subgroups of sufficient size to detect interactive

effects of G72 and DAAO on activation. Subjects were scanned while they performed an overt

phonological verbal fluency task, which is associated with activation in a distributed network including

prefrontal, cingulate and medial temporal regions in healthy volunteers and with impaired performance

and altered activation in schizophrenia and bipolar disorder. On the basis of evidence that both G72 and

DAAO regulate glutamate neurotransmission, which plays a key role in neurocognitive function [21-

23, 26] and the topographic distribution of altered glutamate levels in patients with psychotic disorders

[29], and prodromal symptoms [16], we hypothesized that there would be an epistatic interaction

5

between the two genes on regional activation. In addition, in view of for the putatively altered

glutamate neurotransmission in schizophrenia and bipolar disorder [5], and evidence that patients with

these disorders carry other genes that alter glutamate transmission [30], we predicted that the

hypothesised epistatic effects on brain function would be altered in patients relative to healthy

volunteers. Finally, on the basis that patients with schizophrenia and bipolar disorder are likely to share

the risk variants of several other genes, we hypothesized that epistatic effects between our

polymorphisms of interest would be similar in the two groups as a result of a common genetic context.

Method

Some of the functional neuroimaging data examined in the present investigation have been included in

previous studies which investigated brain dysfunction in psychosis or the impact of other candidate

genes [25, 27, 31-34].

Subjects. A total of 120 subjects were investigated, including 47 healthy volunteers, 40 patients with

schizophrenia and 33 patients with bipolar I disorder. All participants were native English speakers and

gave written informed consent in accordance with protocols approved by the Local and Multicentre

Research Ethics Committee. Healthy volunteers were recruited through local advertisement and had no

family history of psychiatric illness as assessed using the FIGS (Family Interview for Genetic Studies).

Patients with schizophrenia and bipolar disorder were recruited through the South London and

Maudsley NHS Mental Health Trusts and met the relevant DSM-IV criteria, as determined by a

detailed clinical interview augmented where necessary by a systematic review of their medical records.

The mean duration of illness (defined as time since the first episode) for patients with schizophrenia

was 12.22 years (SD=9.54). These patients were taking regular doses of antipsychotic medication; the

mean dose in chlorpromazine equivalents was 571.25 (SD=460.47). The mean duration of illness for

patients with bipolar disorder (defined as time since diagnosis) was 13.39 years (SD=11.03). Only a

minority of these patients (n=9) were taking regular doses of antipsychotic medication; within this

subgroup, the mean dose in chlorpromazine equivalents was 300.00 (SD= 278.66). In addition, eleven

bipolar patients were taking lithium medication (mean dose:836.36 mg/day; SD:196.33). Demographic

data (including age, full-scale IQ, years of education, handedness, gender and ethnicity) are

summarised in Table 1 and described in Supplementary Material S1. Within each diagnostic group,

6

participants were classified based on genotyping of the SNPs rs746187 for G72 and rs2111902 for

DAAO. The exact number of subjects within each genotype group and their demographic and clinical

characteristics are reported in Supplementary Material S2.

Genotyping. Genomic DNA was extracted from blood or cheek swabs following standard

methodology [35], and was resuspended in TE (Tris/EDTA) buffer (10 mM Tris/HCl, pH 7.6; 1 mM

EDTA). Genotyping of G72 SNP rs746187, DAAO SNP rs2111902 and DAAO SNP rs3918346, was

performed blind to status under contract by KBioscience (Herts, UK; http://www.kbioscience.co.uk/).

The rs2111902 and rs3918346 SNPs were chosen for DAAO because they have been previously

associated with schizophrenia and bipolar disorder studies either individually or in haplotype form [4,

7, 36, 37]. However in the present article we report the results for DAAO SNP rs2111902 only, since

we did not detect any epistatic effects involving DAAO SNP rs3918346. The G72 SNP rs746187 was

also chosen because it was previously associated, in haplotype and/or individual form, with

schizophrenia in case-control and transmission disequilibrium test designs [3, 38, 39]. In particular this

SNP was found to be associated with both schizophrenia and bipolar disorder in a case-control

investigation carried out by our research group [37]. The genotyping results of our sample were in

Hardy Weinberg equilibrium (p>0.05) for both DAAO SNP rs2111902 (X2=0.34; p=0.6291) and G72

SNP rs746187 (X2=0.01; p=0.7125) .

- - - Table 1 around here - - -

Verbal Fluency Task and Image Acquisition. The task and image acquisition was performed as

described before [40], see Supplementary Material S3 for details. In brief, during a “generation”

condition, subjects were visually presented with a series of letters and required to overtly articulate a

word beginning with the presented letter. This condition was contrasted with a “repetition” condition,

in which subjects were presented with the word “rest” and were required to say rest out loud. The

demands of the generation condition were manipulated experimentally by presenting different sets of

cue letters that have previously been found to make the task relatively “easy” or “hard” [40].

Behavioural Analysis. The effect of task load, genotype (G72 and DAAO), diagnosis, and their

interaction on the level of accuracy of verbal responses (measured by the number of incorrect responses

7

during scanning) were assessed by using a 3 x 2 x 2 x 2 ANOVA in SPSS (Statistical Package for

Social Sciences; version 15.0), with diagnosis, G72 genotype and DAAO genotype as between-subjects

factors and task load as a within-subject factor.

Neuroimaging Analysis. Analysis was performed with Statistical Parametric Mapping (SPM5)

software (www.fil.ion.ucl.ac.uk/spm) [41], running under Matlab 6.5 (Mathworks). All inferences were

made within a single statistical model, see Supplementary Material S4 for details. In order to reduce the

number of experimental groups in the statistical model, we combined individuals with one or two

copies of the less frequent alleles within the same group for both G72 and DAAO genotypes (see Table

2). We examined the main effects of task and diagnostic category using a standard threshold of p<0.05

after voxel-level correction for multiple comparisons across the whole brain with family-wise error rate

(FWE) rate. Because we originally explored the impact of two DAAO SNPs, we examined the impact

of genotype and its interaction with diagnostic category using a further Bonferroni correction resulting

in a statistical threshold of p<0.025 after voxel-level FWE correction for multiple comparisons across

the whole brain. To assess how much of the inter-individual variance in blood-oxygen-level-dependent

activation was explained by the genetic variation, we used the ηp2 measure of effect size in SPSS. To

confirm that demographic variables (gender, ethnicity, years of education and IQ) and medication

variables (dose, type and duration of antipsychotic treatment) did not bias our results, we repeated the

statistical analysis modelling them as covariates of no interest and also performed a regression analysis

with each medication variable as a covariate. Coordinates are reported in Montreal Neurological

Institute (MNI) space.

- - - Table 2 around here - - -

Results

Performance

Performance data are reported in Table 1. The number of errors significantly differed as a function of

diagnostic group (F=4.368; df=2; p=0.015). Post hoc t-tests revealed that patients with schizophrenia

made significantly more errors than healthy volunteers (F=11.108; df=1; p=0.001). Patients with

bipolar disorder made an intermediate number of errors and did not significantly differ (p>0.05) from

either healthy volunteers or patients with schizophrenia. The number of errors did not differ as a

8

function of G72 or DAAO genotype (p>0.05), irrespective of whether the 3 diagnostic groups were

considered separately or in combination. As expected, there was also a significant main effect of task

demand on the number of errors (F=55.049; df=1; p<0.001). Finally, there were no significant 2- or 3-

way interactions.

Main Effect of Task. In all three diagnostic groups, word generation relative to repetition (irrespective

of task difficulty or genotype) was associated with activation in a bilateral network that included the

inferior and middle frontal gyri, the insula, the dorsal anterior cingulate cortex, the caudate, the

thalamus, the middle and superior temporal gyri and the inferior parietal cortex . Conversely, repetition

relative to word generation was associated with greater activation of the rostral anterior cingulated

gyrus, precuneus and occipital cortex. These data were reported in detail in an earlier study [31].

Main Effect of Diagnostic Group. Patients with schizophrenia expressed greater activation relative to

controls in the left angular gyrus (x=-48 y=-60 z=36 Z-score =4.7 p=0.002 after FWE correction). In

this region, there was also a trend for greater activation in patients with bipolar disorder than controls

(x=-48 y=-60 z=36 Z-score=2.8 p=0.062 after FWE correction). In contrast, direct comparison of

patients with schizophrenia and with bipolar disorder did not reveal any significant difference. These

data were reported in detail in an earlier study [31].

Individual main effects of G72 and DAAO genotypes. There were no regions showing a significant

effect of either G72 or DAAO that was expressed consistently across the three diagnostic groups.

Individual diagnosis-dependent effects of G72 and DAAO genotypes.

A significant diagnosis by G72 genotype interaction was detected in the left precuneus (Figure 1).

Plotting of the parameter estimates revealed that, in this region, the AA genotype was associated with

greater deactivation (i.e. repetition > verbal fluency) during task performance than the AG&GG

genotype in patients with schizophrenia and in patients with bipolar disorder, but not in healthy

volunteers. This interaction effect, which accounted for 6% of the variance in activation, was most

significant for the two patient groups combined (x=-18 y=-52 z=24 Z-score=5.50 p<0.001 after FWE

correction; cluster size=133), but was also evident when the schizophrenia group (x=-14 y=-54 z=24 Z-

9

score=5.03 p=0.002 after FWE correction) and the bipolar group (x=-18 y=-52 z=24 Z-score=4.18

p=0.066 after FWE correction) were contrasted against the control group separately. The effect of G72

in this region did not differ between the schizophrenia and bipolar groups, even at trend level (p>0.001

uncorrected). There were no regions showing a significant diagnosis by DAAO genotype interaction.

- - - Figure 1 around here - - -

G72 x DAAO Interaction Irrespective of Diagnostic Group. There were no regions showing

epistatic effects which were expressed consistently across all three diagnostic groups.

G72 x DAAO x Diagnostic Group Interaction. A significant interaction between G72, DAAO and

diagnosis was detected in the right middle temporal gyrus (x=60 y=-12 z=-12; z-score=5.32; p<0.001

after FWE correction; cluster size=22 voxels). In this region, the GG&GT DAAO genotype was

associated with less activation than the TT DAAO genotype in patients with bipolar disorder and

patients with schizophrenia, but not in healthy volunteers; furthermore, this DAAO x diagnostic group

interaction was more pronounced in individuals with the AA genotype for G72 than in those with one

or two copies of the G allele, resulting in a 3 way interaction (Figure 2). This G72 x DAAO x

diagnostic group interaction accounted for 8.5% of the variance in activation in this region. Although

this 3 way interaction was significant when the two patient groups combined were contrasted against

healthy controls, plotting of the parameter estimates suggested that it was more pronounced in the

patients with bipolar disorder than in those with schizophrenia (Figure 2). Consistent with this

observation, the 3 way interaction survived correction for multiple comparisons when the bipolar group

alone was contrasted against healthy controls (x=60 y=-12 z=-12; z-score=4.78; p=0.006 after FWE

correction), but was only expressed at an uncorrected level when the schizophrenia group alone was

contrasted against healthy controls (x=60 y=-12 z=-10; z-score=3.78; p<0.001 uncorrected). However a

direct comparison which contrasted the strength of the G72 x DAAO interaction in one patient group

against the other was not significant, even at trend level (p>0.001 uncorrected), suggesting that the

strength of the interaction between G72 and DAAO did not differ significantly between the two patient

groups.

10

- - - Figure 2 around here - - -

Effects of Potentially Confounding Factors on Activation. When the dose, type (first vs. second

generation) and duration of antipsychotic treatment were entered into the statistical analysis as

covariates of no interest, they did not change the foci of maximal significance or reduce the associated

Z-scores. Furthermore, whole brain analysis indicated that the activation in the left precuneus and the

right middle temporal gyrus, where significant effects of genotype were detected, was not related to

either the dose, type (first vs. second generation), or the duration of antipsychotic treatment, even at a

liberal statistical threshold (p<0.05 uncorrected). Likewise, the inclusion of dose of lithium medication

as covariate of no interest did not affect the significance of the results and the amount of activation in

the left precuneus and the right middle temporal gyrus was not related to this variable (p<0.05

uncorrected).

Discussion

Psychotic disorders are likely to result from the interaction of multiple genes, each of which has a small

effect on its own [42]. Thus, looking for interactions between the effects of genes implicated in

psychotic disorders may be more useful than focussing on the effects of a given gene in isolation.

Similarly, because patients with these disorders are likely to carry the risk variants of several other

genes, it is potentially useful to examine the effect of gene-gene interactions on brain function in

patients, as well as in healthy controls, as these effects may differ as a result of altered genetic context

[42]. Previous studies have implicated G72 and DAAO in the aetiology of schizophrenia and bipolar

disorder. The product of G72 is thought to activate DAAO, which in turn is the only enzyme that

oxidises D-serine, an important co-agonist for the NMDA glutamate receptor, which is implicated in

the pathogenesis of schizophrenia [3, 42]. We therefore examined the interaction between the G72 and

the DAAO polymorphisms on neurocognitive function in healthy participants and patients with

schizophrenia and bipolar disorder.

We first characterized the individual diagnosis-dependent effects of G72 and DAAO genotypes

separately. While we found no evidence for an interaction between DAAO genotype and diagnosis, we

detected a significant G72 genotype by diagnosis interaction in the left precuneus. In this region, the G

allele was associated with greater activation than the T allele in patients with schizophrenia and in

11

patients with bipolar disorder, but not in healthy volunteers, irrespective of DAAO genotype. The left

precuneus is implicated in executive and working memory processes and is structurally and

functionally altered in patients with schizophrenia and their non-psychotic relatives [43-45]. The

finding of a G72 genotype by diagnosis interaction in this region indicates that this gene has a

diagnosis-dependent impact on brain function. The mechanisms that lead to a different effect of G72 in

controls and patients, or in different diagnostic categories, are unclear. One possibility is that the

effects of variation in a given gene depend on the genetic context [42]. Patients with schizophrenia and

bipolar disorder are likely to carry a number of different risk genes in addition to the gene of interest,

and these may interact with the gene of interest, such that its effect is modified. Similarly, the effect of

a gene may also vary with differences in environmental exposure [46], which again may differ between

patients and controls. The effects of a given gene may also interact with the effects of schizophrenia

and bipolar disorder on the function of the brain.

We then examined the interaction between the G72 and DAAO genes, irrespective of, and dependent

on diagnostic group. There were no significant interaction effects between G72 and DAAO

independent of diagnostic group. However, we detected a significant interaction between G72, DAAO

and diagnosis in the right middle temporal gyrus. In this region, the G allele for DAAO was associated

with less activation than the T allele for DAAO allele in patients with bipolar disorder, but not in

patients with schizophrenia or healthy volunteers; critically, this interaction effect was more

pronounced in individuals with the AA genotype for G72 than in those with one or two copies of the G

allele. The right middle temporal gyrus is thought to play a key role in multimodal and higher sensory

processing, and has also been implicated in the processing of complex, socially relevant stimuli

including the human voice [47] and audiovisual speech [48]. Neuroimaging studies have provided

evidence of reduced gray matter in the middle temporal gyrus of patients with bipolar disorder [49] and

first episode schizophrenia [50] although there have also been inconsistencies in the results [51].

The observation of 2-way and 3-way interactions between G72, DAAO and diagnostic group is

consistent with the idea that the two genes interact with each other at molecular level, as suggested by

in vitro transcription [11]. As both these genes are thought to influence glutamate neurotransmission,

the finding also provide indirect support for the hypothesis that glutamate dysfunction contributes to

the pathophysiology of psychotic disorders [52]. The NMDA receptor is known to play an important

12

role in synaptic plasticity, neurodevelopment and excitotoxicity [53]. It is characterized by two distinct

sub-units known as NR1 and NR2; NR1 is a binding site for co-agonists glycine and D-serine while

NR2 is the agonist binding site for glutamate. NR1 must be occupied for glutamate to be able to open

the channel; this therefore depends on availability of glycine and, to a greater extent, D-serine.

Production and breakdown of D-serine is in part moderated by the DAAO enzyme, which in turn

depends on its activator G72. Selective degradation of D-serine by the DAAO enzyme results in

reduced NMDA neurotransmission. Furthermore, D-serine levels are decreased in the cerebrospinal

fluid and serum of patients with schizophrenia [54], and administration of D-serine may reduce

negative, positive and cognitive symptoms in schizophrenia [55]. It has therefore been proposed that

increased activity of the DAAO enzyme may degrade D-serine, resulting in relative NMDA

hypofunction in psychosis [11]. However, the exact mechanism by which G72 and DAAO may interact

to moderate the availability of D-serine is not fully understood; furthermore it is still unclear how this

mechanism becomes altered in schizophrenia and bipolar disorder [11].

There is preliminary evidence that glutamate regulation is altered not only in patients with full-blown

psychosis but also in individuals with prodromal symptoms [17, 56]. It would therefore be of great

interest to examine the effects of G72 and DAAO on cortical activation in individuals with prodromal

signs of psychosis. The observation of effects of G72 and DAAO similar to those observed in patients

who have developed full-blown psychosis, would provide support to the notion of altered glutamate

regulation in the prodromal phase. Conversely the finding of effects of G72 and DAAO similar to those

observed in healthy volunteers, would suggest that glutamate dysregulation may be a marker of

transition to full-blown psychosis. It has also been hypothesized that glutamate hypofunction in

cortico-striatal projections may lead to the changes in striatal dopamine concentration which are

thought to underlie the emergence of psychotic symptoms [19]. Thus, it would interesting to examine

epistatic interactions between genes implicated in glutamate and dopamine function on cortical

activation in the prodromal phase of the disease, and whether these interactions are predictive of long-

term clinical outcome.

It should be noted that, although a number of genetic studies have associated the G allele of G72

rs746187 and the G allele of DAAO rs2111902 with increased risk of schizophrenia and bipolar

13

disorder, the results have not always been consistent as to which of these alleles confers the higher risk

[5-7]; this could be due to false positive results or reflect allelic heterogeneity (i.e. different alleles in

the same marker being associated with disease); [57]. In the present investigation, we have therefore

avoided reference to the terms low or high risk, and have referred to the alleles instead. Assuming that

G72 rs746187 and DAAO rs2111902 do confer an increased risk of schizophrenia and bipolar disorder,

it remains to be established whether or not the diagnosis-dependent epistatic effects identified in the

present investigation lie upon the pathway between genes and clinical phenotype [58]. This is an

important question, since neuroimaging endophenotypes may mediate the increased risk conferred by

genes, but could also reflect gene effects which do not necessarily result in increased risk, consistent

with the notion of pleiotropy [58].

The present investigation has a number of limitations. First, since all the patients with schizophrenia

and some of those with bipolar disorder had been treated with antipsychotic medication, our results

might have been affected by medication effects. However, within our patient samples, neither the dose,

type nor duration of treatment differed between the genotype subgroups; furthermore, none of these

variables was significantly correlated with brain activation in the right middle temporal gyrus, as

revealed by a series of correlation analyses; finally, the modelling of these variables as covariates of no

interest in the statistical analysis did not alter the peak foci of activation, or the Z scores. A second

limitation of the present study is that the less frequent alleles for the two genes under investigation (i.e.

G for G72 and G for DAAO) are found in a small fraction of the Caucasian population. Thus, in the

present investigation, individuals with one or two copies of the less frequent alleles were combined

within the same group; this means that our data cannot reveal whether the action of the risk allele on

brain function is best described by a dominant or an additive model. A third limitation is that the size of

some experimental groups was relatively small (Table 2). It is therefore important that the three-way

interaction identified in the present study is replicated using a larger sample. The relatively small

number of subjects in some experimental groups may have also limited our sensitivity and prevented us

from detecting additional effects to the ones reported. A fourth limitation is that reaction times were not

measured during scanning and could not be modelled in the statistical analysis; however we modelled

correct and incorrect trials separately thereby minimizing the potential confounding impact of

performance accuracy. A final limitation is that, within the DAAO and G72 genes, several SNPs have

14

been identified by genetic association, and there is no agreement as to which is the marker with the

most significant effect on disease-risk; functional polymorphism with plausible causality has not been

identified.

In conclusion, these data suggest that there is a non-additive interaction between the effects of

variations in the genes implicated in glutamate regulation that affects cortical function. Also, the nature

of this interaction is different in patients and healthy controls, providing support for altered glutamate

function in psychosis. Future studies could explore the effects of DAAO and G72 in individuals with

prodromal symptoms of psychosis, in order to elucidate glutamate dysfunction in this critical phase of

the disorder.

Acknowledgement

Dr Mechelli was supported by a project grant from the Wellcome Trust and an Independent

Investigator Award from NARSAD. Dr Prata was funded by the Fundacao para a Ciencia e Tecnologia,

Lisbon, Portugal. Dr Fu was supported by a Travelling Fellowship from the Wellcome Trust. Dr

Picchioni was funded by a Training Fellowship from the Wellcome Trust (064971) and Dr Elvira

Bramon-Bosch by a New Investigator Research Grant from the MRC (G0901310).

Supportive/Supplementary material

S1. Demographic data.

S2. Number of subjects, demographic and clinical characteristics within each genotype group.

S3. fMRI data acquisition and verbal fluency task.

S4. Neuroimgaing analysis.

15

References

[1] Moldin SO, Gottesman, II. At issue: genes, experience, and chance in schizophrenia--

positioning for the 21st century. Schizophr Bull 1997; 23: 547-61.

[2] Bigos KL, Weinberger DR. Imaging genetics--days of future past. Neuroimage 2010; 53: 804-

9.

[3] Chumakov I, Blumenfeld M, Guerassimenko O, Cavarec L, Palicio M, Abderrahim H, et al.

Genetic and physiological data implicating the new human gene G72 and the gene for D-amino acid

oxidase in schizophrenia. Proc Natl Acad Sci U S A 2002; 99: 13675-80.

[4] Schumacher J, Jamra RA, Freudenberg J, Becker T, Ohlraun S, Otte AC, et al. Examination of

G72 and D-amino-acid oxidase as genetic risk factors for schizophrenia and bipolar affective disorder.

Mol Psychiatry 2004; 9: 203-7.

[5] Detera-Wadleigh SD, McMahon FJ. G72/G30 in schizophrenia and bipolar disorder: review

and meta-analysis. Biol Psychiatry 2006; 60: 106-14.

[6] Kapoor R, Lim KS, Cheng A, Garrick T, Kapoor V. Preliminary evidence for a link between

schizophrenia and NMDA-glycine site receptor ligand metabolic enzymes, d-amino acid oxidase

(DAAO) and kynurenine aminotransferase-1 (KAT-1). Brain Res 2006; 1106: 205-10.

[7] Corvin A, McGhee KA, Murphy K, Donohoe G, Nangle JM, Schwaiger S, et al. Evidence for

association and epistasis at the DAOA/G30 and D-amino acid oxidase loci in an Irish schizophrenia

sample. Am J Med Genet Part B, Neuropsychiatric Genetics 2007; 144B: 949-53.

[8] Madeira C, Freitas ME, Vargas-Lopes C, Wolosker H, Panizzutti R. Increased brain D-amino

acid oxidase (DAAO) activity in schizophrenia. Schizophr Res 2008; 101: 76-83.

[9] Opgen-Rhein C, Lencz T, Burdick KE, Neuhaus AH, DeRosse P, Goldberg TE, et al. Genetic

variation in the DAOA gene complex: impact on susceptibility for schizophrenia and on cognitive

performance. Schizophr Res 2008; 103: 169-77.

16

[10] Bass NJ, Datta SR, McQuillin A, Puri V, Choudhury K, Thirumalai S, et al. Evidence for the

association of the DAOA (G72) gene with schizophrenia and bipolar disorder but not for the

association of the DAO gene with schizophrenia. Behav Brain Funct 2009; 5: 28.

[11] Boks MP, Rietkerk T, van de Beek MH, Sommer IE, de Koning TJ, Kahn RS. Reviewing the

role of the genes G72 and DAAO in glutamate neurotransmission in schizophrenia. Eur

Neuropsychopharmacol 2007; 17: 567-72.

[12] Zahr NM, Mayer D, Pfefferbaum A, Sullivan EV. Low striatal glutamate levels underlie

cognitive decline in the elderly: evidence from in vivo molecular spectroscopy. Cereb Cortex 2008; 18:

2241-50.

[13] Parwani A, Weiler MA, Blaxton TA, Warfel D, Hardin M, Frey K, et al. The effects of a

subanesthetic dose of ketamine on verbal memory in normal volunteers. Psychopharmacology (Berl)

2005; 183: 265-74.

[14] Newcomer JW, Craft S, Fucetola R, Moldin SO, Selke G, Paras L, et al. Glucose-induced

increase in memory performance in patients with schizophrenia. Schizophr Bull 1999; 25: 321-35.

[15] Krystal JH, D'Souza DC, Karper LP, Bennett A, Abi-Dargham A, Abi-Saab D, et al.

Interactive effects of subanesthetic ketamine and haloperidol in healthy humans. Psychopharmacology

(Berl) 1999; 145: 193-204.

[16] Hashimoto T, Volk DW, Eggan SM, Mirnics K, Pierri JN, Sun Z, et al. Gene expression

deficits in a subclass of GABA neurons in the prefrontal cortex of subjects with schizophrenia. J

Neurosci 2003; 23: 6315-26.

[17] Stone JM, Day F, Tsagaraki H, Valli I, McLean MA, Lythgoe DJ, et al. Glutamate

dysfunction in people with prodromal symptoms of psychosis: relationship to gray matter volume. Biol

Psychiatry 2009; 66: 533-9.

[18] Verrall L, Burnet PW, Betts JF, Harrison PJ. The neurobiology of D-amino acid oxidase and

its involvement in schizophrenia. Mol Psychiatry 2010; 15: 122-37.

17

[19] Lang UE, Puls I, Muller DJ, Strutz-Seebohm N, Gallinat J. Molecular mechanisms of

schizophrenia. Cell Physiol Biochem 2007; 20: 687-702.

[20] Krystal JH, Karper LP, Seibyl JP, Freeman GK, Delaney R, Bremner JD, et al. Subanesthetic

effects of the noncompetitive NMDA antagonist, ketamine, in humans. Psychotomimetic, perceptual,

cognitive, and neuroendocrine responses. Arch Gen Psychiatry 1994; 51: 199-214.

[21] Jansen A, Krach S, Krug A, Markov V, Eggermann T, Zerres K, et al. A putative high risk

diplotype of the G72 gene is in healthy individuals associated with better performance in working

memory functions and altered brain activity in the medial temporal lobe. Neuroimage 2009; 45: 1002-

8.

[22] Hall J, Whalley HC, Moorhead TW, Baig BJ, McIntosh AM, Job DE, et al. Genetic variation

in the DAOA (G72) gene modulates hippocampal function in subjects at high risk of schizophrenia.

Biol Psychiatry 2008; 64: 428-33.

[23] Goldberg TE, Straub RE, Callicott JH, Hariri A, Mattay VS, Bigelow L, et al. The G72/G30

gene complex and cognitive abnormalities in schizophrenia. Neuropsychopharmacology 2006; 31:

2022-32.

[24] Krug A, Markov V, Krach S, Jansen A, Zerres K, Eggermann T, et al. Genetic variation in

G72 correlates with brain activation in the right middle temporal gyrus in a verbal fluency task in

healthy individuals. Hum Brain Mapp 2011; 32: 118-26.

[25] Prata DP, Papagni SA, Mechelli A, Fu CH, Kambeitz J, Picchioni M, et al. Effect of D-amino

acid oxidase activator (DAOA; G72) on brain function during verbal fluency. Hum Brain Mapp 2011.

[26] Stefanis NC, Trikalinos TA, Avramopoulos D, Smyrnis N, Evdokimidis I, Ntzani EE, et al.

Impact of schizophrenia candidate genes on schizotypy and cognitive endophenotypes at the population

level. Biol Psychiatry 2007; 62: 784-92.

[27] Papagni SA, Mechelli A, Prata DP, Kambeitz J, Fu CH, Picchioni M, et al. Differential effects

of DAAO on regional activation and functional connectivity in schizophrenia, bipolar disorder and

controls. Neuroimage 2011; 56: 2283-91.

18

[28] Williams NM, Williams, H., Norton, N., Spurlock, G., Kirov, G., Morris, D. W., Waddington,

J. L., Gill, M., Corvin, A., Owen, M. J., O'Donovan, M. C. Replication of schizophrenia susceptibility

loci. Am J Med Genet Part B, Neuropsychiatric Genetics 2003; 122B: 16-7.

[29] Stone JM, Morrison PD, Pilowsky LS. Glutamate and dopamine dysregulation in

schizophrenia--a synthesis and selective review. J Psychopharmacol 2007; 21: 440-52.

[30] Harrison PJ, Weinberger DR. Schizophrenia genes, gene expression, and neuropathology: on

the matter of their convergence. Mol Psychiatry 2005; 10: 40-68.

[31] Mechelli A, Prata DP, Fu CH, Picchioni M, Kane F, Kalidindi S, et al. The effects of

neuregulin1 on brain function in controls and patients with schizophrenia and bipolar disorder.

Neuroimage 2008; 42: 817-26.

[32] Prata DP, Mechelli A, Fu CH, Picchioni M, Kane F, Kalidindi S, et al. Effect of disrupted-in-

schizophrenia-1 on pre-frontal cortical function. Mol Psychiatry 2008; 13: 915-7.

[33] Prata DP, Mechelli A, Fu CH, Picchioni M, Toulopoulou T, Bramon E, et al. Epistasis

between the DAT 3' UTR VNTR and the COMT Val158Met SNP on cortical function in healthy

subjects and patients with schizophrenia. Proc Natl Acad Sci U S A 2009; 106: 13600-5.

[34] Prata DP, Mechelli A, Fu CH, Picchioni M, Kane F, Kalidindi S, et al. Opposite effects of

catechol-O-methyltransferase Val158Met on cortical function in healthy subjects and patients with

schizophrenia. Biol Psychiatry 2009; 65: 473-80.

[35] Freeman B, Smith N, Curtis C, Huckett L, Mill J, Craig IW. DNA from buccal swabs

recruited by mail: evaluation of storage effects on long-term stability and suitability for multiplex

polymerase chain reaction genotyping. Behav Genet 2003; 33: 67-72.

[36] Wood LS, Pickering EH, Dechairo BM. Significant support for DAO as a schizophrenia

susceptibility locus: examination of five genes putatively associated with schizophrenia. Biol

Psychiatry 2007; 61: 1195-9.

19

[37] Prata D, Breen G, Osborne S, Munro J, St Clair D, Collier D. Association of DAO and

G72(DAOA)/G30 genes with bipolar affective disorder. Am J Med Genet Part B, Neuropsychiatric

Genetics 2008; 147B: 914-7.

[38] Hall D, Gogos JA, Karayiorgou M. The contribution of three strong candidate schizophrenia

susceptibility genes in demographically distinct populations. Genes Brain Behav 2004; 3: 240-8.

[39] Shinkai T, De Luca V, Hwang R, Muller DJ, Lanktree M, Zai G, et al. Association analyses of

the DAOA/G30 and D-amino-acid oxidase genes in schizophrenia: further evidence for a role in

schizophrenia. Neuromolecular Med 2007; 9: 169-77.

[40] Fu CH, Morgan K, Suckling J, Williams SC, Andrew C, Vythelingum GN, et al. A functional

magnetic resonance imaging study of overt letter verbal fluency using a clustered acquisition sequence:

greater anterior cingulate activation with increased task demand. Neuroimage 2002; 17: 871-9.

[41] Friston KJ. Introduction: experimental design and statistical parametric mapping. In:

Frackowiak RSJ, Ashburner, J., Penny, W. D., Zeki, S., Friston, K. J., Frith, C., Dolan, R., Price, C. J.,

editor. Human Brain Function. San Diego: Academic Press; 2003. p. pp. 599-632.

[42] Weinberger D. On the matter of neuroimaging in the context of schizophrenia genetics.

Schizophr Res 2010; 117: 109-10.

[43] Fusar-Poli P, Perez J, Broome M, Borgwardt S, Placentino A, Caverzasi E, et al.

Neurofunctional correlates of vulnerability to psychosis: a systematic review and meta-analysis.

Neurosci Biobehav Rev 2007; 31: 465-84.

[44] Shimizu E, Hashimoto K, Ochi S, Fukami G, Fujisaki M, Koike K, et al. Posterior cingulate

gyrus metabolic changes in chronic schizophrenia with generalized cognitive deficits. J Psychiatr Res

2007; 41: 49-56.

[45] Koch K, Wagner G, Schultz C, Schachtzabel C, Nenadic I, Axer M, et al. Altered error-related

activity in patients with schizophrenia. Neuropsychologia 2009; 47: 2843-9.

[46] Caspi A, Moffitt TE. Gene-environment interactions in psychiatry: joining forces with

neuroscience. Nat Rev Neurosci 2006; 7: 583-90.

20

[47] Belin P, Zatorre RJ, Lafaille P, Ahad P, Pike B. Voice-selective areas in human auditory

cortex. Nature 2000; 403: 309-12.

[48] Wright TM, Pelphrey KA, Allison T, McKeown MJ, McCarthy G. Polysensory interactions

along lateral temporal regions evoked by audiovisual speech. Cereb Cortex 2003; 13: 1034-43.

[49] Chen X, Wen W, Malhi GS, Ivanovski B, Sachdev PS. Regional gray matter changes in

bipolar disorder: a voxel-based morphometric study. Aust N Z J Psychiatry 2007; 41: 327-36.

[50] Wilke M, Kaufmann C, Grabner A, Putz B, Wetter TC, Auer DP. Gray matter-changes and

correlates of disease severity in schizophrenia: a statistical parametric mapping study. Neuroimage

2001; 13: 814-24.

[51] Kuroki N, Shenton ME, Salisbury DF, Hirayasu Y, Onitsuka T, Ersner-Hershfield H, et al.

Middle and inferior temporal gyrus gray matter volume abnormalities in first-episode schizophrenia: an

MRI study. Am J Psychiatry 2006; 163: 2103-10.

[52] Olney JW, Farber NB. Glutamate receptor dysfunction and schizophrenia. Arch Gen

Psychiatry 1995; 52: 998-1007.

[53] Tuominen HJ, Tiihonen J, Wahlbeck K. Glutamatergic drugs for schizophrenia: a systematic

review and meta-analysis. Schizophr Res 2005; 72: 225-34.

[54] Hashimoto K, Engberg G, Shimizu E, Nordin C, Lindstrom LH, Iyo M. Reduced D-serine to

total serine ratio in the cerebrospinal fluid of drug naive schizophrenic patients. Prog

Neuropsychopharmacol Biol Psychiatry 2005; 29: 767-9.

[55] Heresco-Levy U. Glutamatergic neurotransmission modulators as emerging new drugs for

schizophrenia. Expert Opin Emerg Drugs 2005; 10: 827-44.

[56] Fusar-Poli P, Stone JM, Broome MR, Valli I, Mechelli A, McLean MA, et al. Thalamic

Glutamate Levels as a Predictor of Cortical Response During Executive Functioning in Subjects at

High Risk for Psychosis. Arch Gen Psychiatry 2011.

21

[57] Moskvina V, O'Donovan MC. Detailed analysis of the relative power of direct and indirect

association studies and the implications for their interpretation. Hum Hered 2007; 64: 63-73.

[58] Owen M. Imaging genetics in psichiatry: disease path or garden path? Schizopr Res 2010;

117: 123.

22

Table captions

Table 1. Degrees of freedom (df), F or X2 test value and p-value are reported for comparisons between

diagnostic groups that reached significance at p<0.05. C = controls; S = schizophrenic patients; BD =

bipolar patients;. SAPS = Scale for the Assessment of Positive Symptoms; SANS = Scale for the

Assessment of Negative Symptoms; BDI = Beck Depression Inventory; ASRM = Altman Self-Rated

Mania Scale; n.s. = not significant.

Table 2. C = controls; S = schizophrenic patients; BD = bipolar patients; n = number of subjects.

Figures legend

Figure 1. Significant G72 genotype by diagnosis interaction (p<0.05 after FWE correction) in the left

precuneus. Parameter estimates refer to brain activation during performance of the verbal fluency task

relative to baseline with negative values indicating deactivation during the task performance; error bars

refer to standard error. C = controls; S = schizophrenic patients; BD = bipolar patients.

Figure 2. Significant interaction between G72, DAAO and diagnosis in the right middle temporal

gyrus. Amongst individuals with the AA genotype for G72, the effect of DAAO genotype differed

across diagnostic groups; however, such difference was not evident amongst individuals with one or

two copies of the G allele for G72. Parameter estimates refer to the direction and size of the DAAO

effect, with positive values indicating GG&GT>TT and negative values indicating TT>GG&GT; error

bars refer to standard error. C = controls; S = schizophrenic patients; BD = bipolar patients.

Table 1. Demographics, clinical scores and behavioural performance in control, schizophrenic and bipolar samples as a function of G72 and DAAO genotypes.

Gene G72 rs746187 DAAO rs2111902 Group Comparisons Diagnosis C S BD C S BD C vs S C vs BD S vs BD Genotype AA AG

&GG AA AG

&GG AA AG

&GG TT TG

&GG TT TG

&GG TT TG

&GG

N. of subjects 22 25 17 23 10 23 27 20 18 22 16 17 Age: mean (sd) 32.9

(8.2) 34.8 (11.8)

35.9 (9.3)

34.6 (13.1)

38.5 (13.5)

39.0 (11.7)

34.5 (10.3)

33.2 (10.3)

35.0 (12.4)

35.4 (11.0)

39.1 (12.5)

38.5 (12.0)

n.s. n.s. n.s.

IQ: mean (sd) 114.9 (12.0)

116.9 (10.1)

101.8 (9.0)

95.4 (18.3)

105.7 (12.3)

107.9 (16.3)

116.3 (12.5)

115.6 (8.6)

97.7 (15.0)

98.4 (15.8)

104.4 (16.4)

110.0 (13.5)

F=40.295 df=1 p<0.001

F=18.070 df=1 p<0.001

F=6.543 df=1 p=0.013

Gender: male/female 11/11 13/12 17/0 16/7 4/6 8/15 13/14 11/9 16/2 17/5 6/10 6/11 χ2=9.452 df=1 p=0.002

n.s. χ2=16.278 df=1 p=0.000

Ethnicity: cauc/black/carib/ black-afric/mixed

21/0/0/0/1

24/0/1/0/0

11/5/1/0/0

22/0/0/0/1

8/1/1/0/0 23/0/0/0/0

26/0/1/0/0

19/0/0/0/1

15/1/1/0/1

18/4/0/0/0

16/0/0/0/0

15/1/1/0/0

n.s. n.s. n.s.

Handedness: right/left/mixed 22/0/0 22/3/0 17/0/0 19/4/0 8/2/0 22/1/0 25/2/0 19/1/0 16/2/0 20/2/0 13/3/0 17/0/0 n.s. n.s. n.s. Year of education: mean (sd) 15.2

(2.6) 15.0 (2.9)

14.9 (13.5)

13.5 (2.4)

14.6 (2.9)

15.3 (3.2)

14.7 (2.3)

15.6 (3.2)

14.5 (2.0)

13.8 (2.3)

14.4 (3.0)

15.8 (3.2)

n.s. n.s. n.s.

Duration of illness: mean (sd) 12.3 (8.8)

12.2 (10.3)

13.8 (12.7)

13.2 (10.5)

15.4 (9.7)

11.6 (11.6)

10.8 (11.2)

11.5 (10.1)

n.s. n.s.

CPZ rate: mean (sd) 564.7 (426.4)

567.1 (493.6)

162.5 (263.3) [4 sub.]

46.7 (149.7) [5 sub.]

610.0 (437.5)

437.5 (406.0)

175.0 (187.5) [5 sub.]

207.7 (375.0) [4 sub.]

Years of antipsychotic medication: mean (sd)

12.3 (8.8)

12.2 (10.3)

8.4 (13.5)

7.0 (9.6) 15.4 (9.7)

11.6 (11.6)

7.8 (12.0)

1.8 (2.2)

SAPS: mean (sd) 6.9 (7.0) 7.9 (7.4) 9.5 (8.9) 5.9 (4.9) SANS: mean (sd) 7.1 (3.5) 9.1 (5.4) 9.1 (5.0) 7.5 (4.5) BDI: mean (sd) 15.9

(12.8) 7.7 (6.5) 6.2 (5.2) 13 (10.5)

ASRM: mean (sd) 3.5 (2.4) 4.2 (2.8) 3.2 (2.4) 4.6 (2.6) N. of errors: mean (sd) N. of errors “easy”: mean (sd) N. of errors “hard”: mean (sd)

9.4 (5.2)

3.4 (2.7)

6.0 (3.5)

9.4 (8.2)

3.2 (3.9)

6.2 (4.6)

14.7 (7.1)

5.9 (4.0)

8.8 (4.4)

14.0 (6.9)

5.6 (4.3)

8.3 (3.8)

10.2 (9.2)

3.2 (2.9)

7.0 (6.5)

11.8 (9.9)

4.7 (5.2)

7.2 (6.1)

10.6 (7.9)

4.0 (3.6)

6.6 (4.6)

7.7 (4.9)

2.3 (2.8)

5.5 (3.3)

13.0 (5.8)

4.7 (3.1)

8.3 (3.8)

15.3 (7.6)

6.6 (4.8)

8.7 (4.3)

12.1 (8.9)

3.8 (3.5)

8.3 (6.3)

10.6 (10.4)

4.6 (5.6)

6.1 (5.9)

F=10.932 df=1 p=0.001

n.s. n.s.

Table 2. Number of subjects included in each experimental group after combining individuals with one or two copies of the less frequent alleles within the same group for both G72 and DAAO genotypes.

C S BD

G72 rs746187 AA AG&GG AA AG&GG AA AG&GG

DAAO rs2111902 TT TG&GG TT TG&GG TT TG&GG TT TG&GG TT TG&GG TT TG&GG

n 12 10 15 10 7 10 11 12 5 5 11 12

Supplementary Material S1

There were no significant differences across the 3 diagnostic groups with respect to age, ethnicity,

handedness and year of education. Full-scale IQ was assessed using the WAIS-III (Wechsler Adult

Intelligence Scale-III) [1], the WAIS-R (Wechsler Adult Intelligence Scale-R) [2], the WASI-FSIQ-4

(Wechsler Abbreviated Scale of Intelligence) [3], or the Quick Test [4], a passive response picture-

vocabulary test. Previous studies have shown that the WAIS-III correlates highly with both the WAIS-

R (93.9%) [1] and the WASI-FSIQ-4 (92%) [3]. The Quick test has also been shown to yield

comparable results to the WAIS-R (91%) [5]. The proportion of subjects assessed with each method

was comparable across diagnostic as well as genotypic groups. The 3 diagnostic groups differed in

terms of IQ (F=18.570; df=2; p<0.001) and male:female ratio (X2 =17.060; df=2; p<0.001). Post hoc t-

tests revealed that the group of healthy volunteers had a higher IQ than both groups of patients with

schizophrenia and bipolar disorder and that the group of patients with schizophrenia had a higher

male:female ratio than both healthy volunteers and patients with bipolar disorder; the difference in IQ

is consistent with previous studies [6-8].

References

[1] Wechsler D. Wechsler Adult Intelligence Scale —Third Edition Manual. New York: The

Psychological Corporation 1997.

[2] Wechsler D. Manual for the Wechsler Intelligence Scale — Revised. New York: The

Psychological Corporation 1981.

[3] Wechsler D. Wechsler Abbreviated Scale of Intelligence. New York: The Psychological

Corporation 1999.

[4] Ammons RB, Ammons CH. The Quick test: Psychological Test Specialists. Missoula 1962.

[5] Frith CD, Leary J, Cahill C, Johnstone EC. Performance on psychological tests. Demographic and

clinical correlates of the results of these tests. Br J Psychiatry Suppl 1991; 13:44-6.

[6] Krabbendam L, Arts B, van Os J, Aleman A. Cognitive functioning inpatients with schizophrenia

and bipolar disorder: a quantitative review. Schizophr Res 2005; 80:137–149.

[7] Daban C, Martinez-Aran A, Torrent C, Tabares-Seisdedos R, Balanza-Martinez, V, Salazar-Fraile

J, Selva-Vera G, Vieta E. Specificity of cognitive deficits in bipolar disorder versus schizophrenia. A

systematic review. Psychother Psychosom 2006; 75:72-84.

[8] Burdick KE, Gunawardane N, Woodberry K, Malhotra AK. The role of general intelligence as an

intermediate phenotype for neuropsychiatric disorders. Cogn Neuropsychiatry 2009; 14:299-311.

Supplementary Material S2

Within the control group, there were 22 subjects with the AA G72 genotype including 12, 7 and 3 with

the TT, TG and GG DAAO genotype respectively; 15 subjects with the AG G72 genotype including 9,

6 and 0 with the TT, TG and GG DAAO genotype respectively; 10 subjects with the GG genotype

including 6, 4 and 0 with the TT, TG and GG DAAO genotype respectively. Within the schizophrenic

group, there were 17 subjects with the AA G72 genotype including 7, 7 and 3 with the TT, TG and GG

DAAO genotype respectively; 20 subjects with the AG G72 genotype including 9, 9 and 2 with the TT,

TG and GG DAAO genotype; 3 subjects with the GG G72 genotype including 2, 1 and 0 with the TT,

TG and GG DAAO genotype. Within the bipolar group, there were 10 subjects with the AA G72

genotype including 5, 4 and 1 with the TT, TG and GG DAAO genotype respectively; 18 subjects with

the AG G72 genotype including 7, 10 and 1 with the TT, TG and GG DAAO genotype; 5 subjects with

the GG G72 genotype including 4, 1 and 0 with the TT, TG and GG DAAO genotype respectively.

Age, IQ, gender, ethnicity, handedness, and years of education did not differ significantly as a function

of either G72 or DAAO genotype within each diagnostic group (p>0.05). Medication variables

including dose of antipsychotic medication (in chlorpromazine equivalent), duration of illness and

duration of medication did not differ as a function of G72 or DAAO genotype within the schizophrenia

or bipolar groups (p>0.05). In the schizophrenia group, symptom profile was assessed using the Scale

for the Assessment of Positive Symptoms (SAPS) and the Scale for the Assessment of Negative

Symptoms (SANS) which measure symptoms experienced within the month prior to the interview; in

the bipolar group, symptom profile was examined using the Beck Depression Inventory (BDI) and the

Altman Self-Rated Mania Scale (ASRM). Within the schizophrenia sample, SAPS and SANS scores

did not differ as a function of G72 or DAAO genotype. Within the bipolar sample, scores on the BDI

were associated with both G72 (F=5.176; df=1; p=0.031) and DAAO (F=4.94; df=1; p=0.034)

genotypes (see Table 1); in contrast, scores on the ASRM were not associated with either G72 or

DAAO genotype.

Supplementary Material S3

Verbal fluency task. During fMRI scanning, subjects performed an overt verbal fluency task involving

two main conditions: generation and baseline [1]. In the generation condition, subjects were presented

with a series of letters on a computer screen; the task required them to respond to each letter by

generating a word that started with that letter. Letter cues were presented in blocks of seven with a

stimulus onset asynchrony of 4000ms; all cues in a given block were of the same letter but each block

involved a different letter. The paradigm thus resembles the classical version of the task used in

neuropsychological studies except that each response is cued at regular intervals, rather than the subject

responding freely as many times as they can following a single cue. A paced paradigm is more

compatible with an fMRI study, as it reduces variation in the timing of overt verbal responses within

and between subjects, and reduces the risk that only a small proportion of the block is associated with

performance of the task, as would occur if a subject rapidly articulated all their responses in the initial

part of the block and then disengaged from the paradigm. A further benefit of using a paced task with a

relatively long inter-stimulus interval was that there was less risk of large between-subject and

between-group variation in task performance, which could confound interpretation of differences in

activation, particularly between the control and patient groups. In the baseline condition, subjects were

presented with the visual word “rest” and were required to say “rest” out loud; “rest” cues were also

presented in blocks of seven with a stimulus onset asynchrony of 4000 ms. Functional MRI data were

acquired during two separate acquisition runs, each including 5 blocks of letters alternating with five

blocks of “rest” trials. This resulted in a total of 70 letter stimuli and 70 “rest” trials for each subject.

Verbal responses were recorded by means of a microphone that was compatible with the MRI

apparatus; this allowed us to identify "incorrect" trials in which the subject did not generate any

response or generated repetitions, derivatives or grammatical variations of the previous word.

fMRI data acquisition. The T2*-weighted gradient-echo single-shot echo-planar images were

acquired on a 1.5-T, neuro-optimized IGE LX System (General Electric, Milwaukee) at the Maudsley

Hospital, London, U.K. Twelve noncontiguous axial planes (7-mm thickness, slice skip: 1 mm) parallel

to the anterior commissure–posterior commissure line were collected. A "clustered" acquisition

(TE=40 ms, flip angle=70°) was used in order to minimize the impact of head movement during

verbalization [2, 3]. A clustered acquisition sequence capitalizes on the delay of the haemodynamic

response, which reaches its peak about 3–5 s after stimulus onset [4]. A letter cue was presented for

750 ms and an overt verbal response could be made over a silent period of 2900 ms; an image was then

acquired over 1100 ms resulting in a total repetition time (TR) of 4000 ms.

References

[1] Fu CH, Morgan K, Suckling J, Williams SCR, Andrew C, Vythelingum GN, McGuire PK. A

functional magnetic resonance imaging study of overt letter verbal fluency using a clustered acquisition

sequence: greater anterior cingulate activation with increased task demand. Neuroimage 2002; 17:871–

879.

[2] Bullmore ET, Brammer MJ, Rabe-Hesketh SR, Curtis VA, Morris RG, Williams SCR, Sharma T,

McGuire PK. Methods for diagnosis and treatment of stimulus-correlated motion in generic brain

activation studies using fMRI. Hum Brain Mapp 1999; 7:38-48

[3] Amaro EJ, Williams SC, Shergill SS, Fu CH, MacSweeney M, Picchioni MM, Brammer MJ,

McGuire PK. Acoustic noise and functional magnetic resonance imaging: current strategies and future

prospects. J Magn Reson Imaging 2002; 16:497-510.

[4] Glover GH. Deconvolution of impulse response in event-related BOLD fMRI. Neuroimage 1999;

9:416-429.

Supplementary Material S4

To minimize movement-related artifacts, all volumes from each subject were realigned and unwarped

using the first as reference [1], normalized to a standard MNI-305 template, and spatially smoothed

with an 8-mm FWHM isotropic Gaussian kernel. First, the statistical analysis of regional responses was

performed in a subject-specific fashion by convolving each onset time with a synthetic haemodynamic

response function (HRF). To minimize performance confounds, we modeled correct and incorrect trials

separately by using an event-related model. This design resulted in a total of 4 experimental conditions:

(i) easy generation, (ii) hard generation, (iii) repetition, and (iv) incorrect responses. The latter

condition was excluded from the group analysis to control for effects of group differences in task

performance on brain activation. To remove low-frequency drifts, the data were high-pass filtered by

using a set of discrete cosine basis functions with a cut-off period of 128 s. The parameter estimates

were calculated for all brain voxels by using the general linear model, and contrast images for “easy

generation > repetition” and “hard generation > repetition” were computed in a subject-specific

fashion. Second, to permit inferences at the population level [2], the subject-specific contrast images

were entered into a second level analysis using the general linear model. The less frequent alleles for

the two genes under investigation (i.e. G for G72 and G for DAAO) are found in a small fraction of the

Caucasian population; thus, in the present investigation, individuals with one or two copies of the less

frequent alleles for G72 and DAAO were combined within the same group. We avoided using a 3 x 2 x

2 ANOVA with diagnostic group, G72 genotype and DAAO genotype as factors, as this would have

resulted in some cells with as few as 5 subjects each. Instead we used an ANCOVA model in which

diagnostic group (controls, schizophrenic patients, bipolar patients) and G72 genotype (AA, AG/GG)

were modelled as between-subject factors, and DAAO genotype (TT, TG/GG) was modelled as an

interactive covariate. Modelling DAAO genotype as an interactive covariate involved entering 6

regressors made of −1 (for individuals who were TT homozygotes) and 1 (for individuals with one or

two copies of the G allele), one for each of the 6 experimental groups that resulted from modelling

diagnostic group and G72 genotype as factors. This statistical model, which has previously been used

to examine three-way interactions [3], allowed us to test for the main effect of the task, the main effect

of diagnostic group, the main effects of G72 and DAAO genotypes, and any non-additive interactions

between the two genes, either diagnosis-dependent or diagnosis-independent. Task load (easy, hard)

was also modelled in same the statistical model as a within-subject factor to minimize error variance;

however this manipulation was irrelevant to the hypotheses of the present study, and we therefore

report results for the hard and easy conditions combined. Estimation of the model included correction

for non-sphericity to account for possible unequal variance between experimental groups [4]. The t-

images for each contrast at the second level were transformed into statistical parametric maps of the Z-

statistic.

References

[1] Ashburner J, Friston KJ. In High-Dimensional Image Warping. In: Frackowiak RSJ, Ashburner J,

Penny WD, Zeki S, Friston KJ, Frith C, Dolan R, Price CJ, editor. Human Brain Function. San Diego:

Academic Press. 2003; pp 673–694.

[2] Friston KJ. Introduction: experimental design and statistical parametric mapping. In: Frackowiak

RSJ, Ashburner J, Penny WD, Zeki S, Friston KJ, Frith C, Dolan R, Price CJ, editor. Human Brain

Function. San Diego: Academic Press. 2003; pp 599-632.

[3] Prata DP, Mechelli A, Fu CH, Picchioni M, Toulopoulou T, Bramon E, Walshe M, Murray RM,

Collier DA, McGuire P. Epistasis between the DAT 3' UTR VNTR and the COMT Val158Met SNP on

cortical function in healthy subjects and patients with schizophrenia. Proc Natl Acad Sci U S A 2009;

106:13600-5.

[4] Glaser DE, Friston KJ. Variance Components. In: Frackowiak RSJ, Ashburner J, Penny WD, Zeki

S, Friston KJ, Frith C, Dolan R, Price CJ, editor. Human Brain Function. San Diego: Academic Press.

2003; pp 781-91.