Original Article

Molecular Psychiatry (2009) 14, 416–428; doi:10.1038/mp.2008.127; published online 9 December 2008

Gene discovery through imaging genetics: identification of two novel genes associated with schizophrenia

S G Potkin1, J A Turner1, J A Fallon1, A Lakatos1, D B Keator1, G Guffanti2 and F Macciardi2 FBIRN1

  1. 1Department of Psychiatry and Human Behavior, Brain Imaging Center, University of California, Irvine, CA, USA
  2. 2Department of Science and Biomedical Technology, University of Milan, Milan, Italy

Correspondence: Professor SG Potkin, Department of Psychiatry and Human Behavior, Brain Imaging Center, University of California, Irvine, 5251 California Ave, Suite 240, Irvine, CA 92617, USA. E-mail: sgpotkin@uci.edu

Received 18 February 2008; Revised 6 October 2008; Accepted 9 October 2008; Published online 9 December 2008.

Top

Abstract

We have discovered two genes, RSRC1 and ARHGAP18, associated with schizophrenia and in an independent study provided additional support for this association. We have both discovered and verified the association of two genes, RSRC1 and ARHGAP18, with schizophrenia. We combined a genome-wide screening strategy with neuroimaging measures as the quantitative phenotype and identified the single nucleotide polymorphisms (SNPs) related to these genes as consistently associated with the phenotypic variation. To control for the risk of false positives, the empirical P-value for association significance was calculated using permutation testing. The quantitative phenotype was Blood-Oxygen-Level Dependent (BOLD) Contrast activation in the left dorsal lateral prefrontal cortex measured during a working memory task. The differential distribution of SNPs associated with these two genes in cases and controls was then corroborated in a larger, independent sample of patients with schizophrenia (n=82) and healthy controls (n=91), thus suggesting a putative etiological function for both genes in schizophrenia. Up until now these genes have not been linked to any neuropsychiatric illness, although both genes have a function in prenatal brain development. We introduce the use of functional magnetic resonance imaging activation as a quantitative phenotype in conjunction with genome-wide association as a gene discovery tool.

Keywords:

cognition, fMRI, GWAS, case–control study, quantitative trait loci, prefrontal cortex

Top

Introduction

The genetic aspect of neuropsychiatric disorders, such as schizophrenia, bipolar disorder and autism, is often initially determined through linkage studies which can point to areas of the genome linked to the disorder (for example, references1, 2, 3, 4, 5). Combining this approach with known molecular functions of the genes in the linkage areas identifies likely genes that can then be specifically tested in case–control designs. Although useful, this method depends on current knowledge regarding genes found in the linkage area; it greatly limits the potential genes studied, both in number and mechanism.

Genome-wide association studies (GWAS) allow the identification of genes whose relationship with the disease phenotype has not even been hypothesized. Genome-wide association techniques have developed exponentially in the past few years; current chips allow 100000–1000000 single nucleotide polymorphisms (SNPs) to be assessed in an individual. GWAS offer enormous promise in identifying genetic variation involved with illness and its response to treatment (for example, Ozaki et al.;6 Klein et al.7) by allowing all areas of the genome to be considered. However, as the number of potential genetic variations under study increases, making it more likely to find the important variations, so does the likelihood of spurious findings. Solutions to this statistical problem have been to increase the sample size to tens of thousands, or more; to increase the significance threshold astronomically; or to limit the number of SNPs considered to a handful of a priori determined candidates.8, 9 We suggest that use of a quantitative trait (QT) as the phenotype, rather than a categorical case–control design, has considerably greater statistical power, thereby reducing the sample size needs in a GWAS by many fold.10

Functional MRI reflects neural activation while performing tasks and has been clearly shown to reflect neuropsychiatric dysfunction.11, 12, 13, 14, 15 Imaging Genetics is the emerging field that integrates genetic and neuroimaging data, usually through candidate gene approaches.16, 17, 18, 19, 20 The integration of the quantitative imaging phenotype analysis with GWAS is the basis of our approach to Imaging Genetics. This approach both directly addresses the candidate gene limitations with a genome-wide scan, and has the increased statistical power of QT analysis coupled with the sensitivity of imaging to reflect brain function.

We use differential brain imaging activation patterns as the starting point in Imaging Genetics analyses, based on the assumption that brain imaging will reveal important pathophysiological differences. We then determine the impact of genetic variation on these brain activation phenotypic patterns, to identify genetic influences potentially key to understanding the pathophysiology. In this way, we use a QT as a marker for the neuropsychiatric dysfunction, and gain the statistical power of using a QT while targeting a phenotype that separates the population of interest from the control population. The differential brain imaging patterns can be either be chosen from the literature or empirically determined within the study. We chose activation in the dorsolateral prefrontal cortex (DLPFC) for this study because of the extensive literature implicating it in schizophrenia.21, 22, 23, 24, 25, 26, 27, 28, 29, 30

The use of QTs with a comprehensive genome-wide scan has not been commonly applied to neuropsychiatric disorders, perhaps because of difficulty in determining the QT. A notable exception was the discovery of Kidney and Brain Protein (KIBRA) using memory performance as the QT, based on quartile ranking in verbal episodic memory, in a genomics scan of pooled DNA.31 Recently, Almasy et al.32 used cognitive function as a QT in a family study of schizophrenia with 386 microsatellite markers. Neuroimaging has been used to reveal the function of candidate genes, for example COMT,33, 34 using studies designed to begin with a specific gene and explore its effects on various phenotypes. Brain imaging has been used to study the function of a number of other genes such as 5HTTR transporter, DRD4, DRD1, HTR3A, TPH2 and MAOA20, 35, 36, 37, 38 and genes associated with schizophrenia including NRG1, RGS4, COMT, GRM3, G72, DISC1 and BDNF.16, 18, 19, 39 In functional neuroimaging studies of neuropsychiatric patients and healthy controls, differential activation in regions of interest (ROI) or putative circuits can be identified. We limit our imaging phenotypes to these areas—in this case, left DLPFC (BA46)—and then examine the function of individual genetic variation on these phenotypes at an individual level, that is, how each SNPs predicts activation in the DLPFC. Our approach reverses the candidate gene strategy: Rather than beginning with a specific candidate gene as a grouping factor and searching for differences in neuroimaging results within groups, we begin with brain imaging as a phenotype and determine the SNPs that influence that phenotype.

We implemented this Imaging Genetics approach in a small, discovery sample. This combination of genome-wide exploration with a quantitative imaging phenotype led to the identification of two genes, RSRC1 and ARHGAP18, not previously associated with schizophrenia or other brain disorders. To support our findings and to validate our method of gene discovery we conducted a case–control association study in a new sample of schizophrenia patients and healthy controls. SNPs representing these two genes were verified in this sample.

Top

Materials and methods

Subjects

The discovery sample consisted of 28 chronic schizophrenic patients diagnosed according to Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria by a Structured Clinical Interview for Diagnosis (SCID).40, 41 A total of 75% of the patients were right handed. The average age was 43 years (range 27–60 years old). The mean duration of illness was 13.6 years (range 1–32 years). All were treated with stable doses of atypical antipsychotic drugs, all except two with conventional antipsychotic agents. See the Supplementary Materials for additional patient characteristics.

The corroborative sample consisted of 82 subjects with chronic schizophrenia and 91 controls overall matched for gender and age, also diagnosed according to DSM-IV criteria with a SCID interview. This was part of a larger cohort collected by the Functional Imaging Biomedical Informatics Research consortium (FBIRN). A total of 89% of both groups were right handed. The average age was 38 (range 18–61 years) and 36.2 (range 18–65 years) years of age for the subjects with schizophrenia and the controls, respectively. The mean duration of illness was 14.3 years (range 2–43 years). All were treated with stable doses of antipsychotic drugs. This sample is typical of chronic schizophrenic patients in treatment with a moderate degree of stable symptoms (Supplementary Materials).

fMRI methods (also see Supplementary Materials)

fMRI scanning
 

All subjects in the discovery sample were scanned on a 1.5T Phillips (Picker Eclipse) scanner using a T2*-weighted gradient echo sequence, during which the subjects performed three runs of a Serial Item Recognition Paradigm (SIRP).27 The SIRP task included two memory loads (2 digits and 5 digits to remember) and a control condition (left and right pointing arrows, to control for movement activations), each in a blocked design. In the baseline condition blocks, subjects were presented a series of arrows and asked to indicate the direction in which the arrow pointed (left or right). In both memory load conditions blocks, subjects were presented with a set of numbers (presented simultaneously for 5s) then presented with a series of 10 probe trials each consisting of a single number presented for 2s. Subjects indicated whether the probe was in the memory set of numbers or not. The memory sets were different in every block and every run. Each run was 240s long and consisted of nine blocks, beginning and ending with a baseline condition block.

fMRI analyses

The first two images of each scan were deleted to allow for saturation effects. The remaining functional magnetic resonance imaging (fMRI) data for each scan were motion-corrected, normalized to a standard space, smoothed using an 8-mm FWHM Gaussian filter, and analyzed using SPM2 (http://www.fil.ion.ucl.ac.uk/spm/), using the canonical HRF. The General Linear Model (GLM) modeled the effects of the low and high memory load relative to the control condition. The contrast of interest compared the high memory load against the low memory load. The voxel-wise analysis was performed on each subject to provide a subject-specific measure of BOLD signal change.

The primary ROI, the left DLPFC, and 11 areas with neuroanatomical connections to the DLPFC and 2 control areas were defined by sampling the digital Talairach atlas developed by Lancaster et al.42 in coordination with the NIH/NIMH Human Brain Project. A summary statistic for each atlas determined region was calculated for each subject (a mean β-value across all the voxels in each ROI for the high memory load >low memory load contrast). These summary statistics, reflecting differential imaging signals, were used as the initial imaging phenotypes. The left DLPFC was defined a priori as the primary dependent variable given the extensive literature implicating it in schizophrenia.21, 22, 23, 24, 25, 26, 27, 28, 29 The 11 other ROI and 2 control ROIs were secondary dependent variables chosen to address the biological plausibility of the DLPFC findings.

Genotyping methods

Discovery sample
 

Genotyping of the discovery sample was performed by the Broad Institute (http://www.broad.mit.edu/gen_analysis/genotyping/) with the Illumina Infinium Human1 chip. This provided a total of 105950 autosomic SNPs. Samples successfully genotyped in less than 90% of markers on either array were excluded from analysis. We removed 5297 SNPs with missing genotypes >10% across subjects and 2117 SNPs with minor allele frequency (MAF) <1% that partially overlapped with the previous category. After removal of SNPs that did not pass the quality assurance measures we had 98648 SNPs to analyze, with a global mean call rate of 98.2%.

This GWAS analysis identified two genes described in ‘Results’. To further characterize these two genes additional SNPs were genotyping with the HumanCNV370-Duo Bead Chip performed by the Laboratory for Genetic Epidemiology of Complex Traits of the University of Milan. The choice of this platform was based on cost-effective considerations.

Corroborative sample
 

Corroborative samples for RSRC1 and ARHGAP18 genes were assayed with the Illumina Infinium HumanHap300 Bead Arrays for cost effectiveness and Taqman SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA) for those RSRC1 and ARHGAP18 markers selected for confirmation that were not present on the HumanHap300. Quality assurance criteria were the same as applied to the discovery sample WGS. Six RSRC1 SNPs and one ARHGAP18 SNP were genotyped by Custom Taqman SNP Genotyping Assay (Applied Biosystems; see Supplementary Materials for details).

Stratification
 

To correct for possible population stratification in our sample(s) we used the program EIGENSTRAT43 that controls for the risk of stratification by performing a principal component analysis with the highest possible number of available SNPs. We thus used the entire set of SNPs (described above), as suggested by the method, and additionally performed a parallel analysis using a subset of SNPs 18036 that (1) were not in linkage disequilibrium (LD) to each other, (2) showed a MAF >0.30 and (3) were not in chromosomal regions previously known to be related to schizophrenia. With both approaches, our samples did not show evidence of stratification.

QT statistical analyses on the discovery sample

For each of the SNPs of the Human1 Bead Array and the additional RSRC1 and ARHGAP18 SNPs on the HumanCNV370-Duo Bead Chip, we performed a QT regression on the imaging phenotype in the discovery sample. Using the PLINK permutation procedure for QTs we identified statistically significant results. PLINK (http://pngu.mgh.harvard.edu/purcell/plink/)44 performs a regression-like approach for QTs, using either an asymptotic (likelihood ratio or Wald tests) or an empirical significance test for each SNP analyzed. To control for the risk of false positives, we calculated the empirical P-value for association significance with permutations; for each individual SNP we exchanged labels across the different quantitative values, and using the adaptive permutation algorithm of PLINK, we determined the empirical significance for each SNP in the genotype analyses. With this procedure, the number of permutations performed is different for each SNP and is constrained to the observed significance ‘raw’ value for each SNP. The empirical permuted P-value refers to the proportional number of times that the permuted test passed the original significance value. We also permuted genotype assignments 550000 times for all the SNPs that were included in the discovery sample analysis as a check on the individual SNP results, and observed that the smallest P-value by chance was P=10−5, which is consistent with the individual SNP permutation results.

Case–control association analyses on the corroborative sample

To determine the support of the results of the discovery sample, we focused only on the SNPs mapping the RSRC1 and ARHGAP18 genes, with 35 and 61 SNPs, respectively. A logistic regression analysis for small case–control samples, BINREG,45 was used. BINREG incorporates a permutation procedure to control for false positives. Tables 1 and 2 provide the case–control results for comparison with the discovery sample.



Top

Results

QT analyses on the discovery sample

Using the left hemisphere DLPFC summary statistic as the imaging phenotype, one gene, RSRC1 (arginine/serine-rich coiled-coil 1), was identified as having at least one SNP whose QT analysis was significant at P<10–7, with an empirical P-value of 10–6 by permutation. We found that SNPs rs12696067, rs6803630 and rs1915935, all mapping within the RSRC1 gene on chromosome 3, significantly affected activation of DLPFC (BA46), with P-values of 10–7 (empirical by permutation 10–6; Table 1). Of the 19 SNPs associated with RSRC1 in the Human 1 Array based on Build 35, 14 SNPs were significant by permutation testing of each SNP. An additional 31 SNPs for RSRC1 were evaluated using the HumanCNV370-Duo; 13 of those were also statistically significant. Thus, for the studied SNP data set, 27 of the 48 RSRC1 SNPs tested were significant by permutation testing of each SNP.

This BA46 imaging phenotype was also affected by SNP rs11154490 at a permutation significance level of P<10–5. rs11154490 is one of nine SNPs on the Human 1 array that map the gene ARHGAP18 on chromosome 6 (Rho GTPase activating protein 18). Of the nine ARHGAP18 SNPs in the Human 1 Array (Build 35), four were significant by permutation testing. An additional 49 SNPs for ARHGAP18 were evaluated using the Human370CNV; 15 of those were also statistically significant by permutation testing (Table 2).

Thus, 27 SNPs in RSRC1 and 19 in ARHGAP18 were significantly associated with activation in the DLPFC. Tables 1 and 2 show the results of the QT analysis using PLINK, as well as the corroborative results (see below). See Supplementary Tables S1 and S2 for full results and Figures 1 and 2 for physical maps of the two genes.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Physical map of the single nucleotide polymorphisms (SNPs) associated with RSRC1 in the Discovery sample produced by WGAViewer.72 The topmost sector is the ideogram of Chromosome 3; the vertical line shows the relative of location of RSRC1. Below this is the graph showing the -log(P) significance values of the individual SNPs on the imaging phenotype. The vertical lines in red indicate the SNPs that are significant in both the Discovery and Corroborative samples. The small blue lines below this indicate the location of the exons in the transcripts annotated (translated region of the DNA). The vertical lines above the accompanying triangular matrix indicate the SNP locations, and demonstrate the linkage disequilibrium (LD) pattern between SNPs (r2). The warmer colors on the flame scale indicate greater LD whereas the blue indicates absence of LD. These patterns in this figure suggest minimal LD between the studied SNPs in the Discovery sample.

Full figure and legend (264K)

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Physical map of the single nucleotide polymorphisms (SNPs) associated with ARHGAP18 in the discovery sample produced by WGA Viewer.72 See Figure 3 legend for description. The clustering pattern of primarily blue and yellow squares indicate minimal linkage disequilibrium (LD) between the SNPs studied. This indicates relative statistical independence between the SNPs analyzed in the discovery sample. The association of these SNPs with the quantitative phenotype strengthening the proposed relationship of the ARHGAP18 gene with schizophrenia.

Full figure and legend (204K)

Circuitry exploration

Our analysis based on a priori assumptions focused on measures of the left DLPFC activation as the primary imaging phenotype, and then considered imaging measures in other regions of interest in exploring biological plausibility. Examining the effect of the significant genetic locus (using permuted P-values) across other brain regions determines if the effect of that locus follows the pattern of known brain circuitry, or if it appears randomly across the brain. We measured the same summary statistic, a change in BOLD signal between the harder and easier memory load in the SIRP task, over each of an additional 11 ROIs for each subject, as for the DLPFC (BA46), and two control areas. The 11 ROIs were chosen based on the literature documenting their relationship to working memory and executive and cognitive functions, as well as anatomically connected circuitry.15, 24, 46, 47 The hypothalamus and right uncus served as the control areas.

The brain areas significantly related to the polymorphic variation in SNPs that represent both genes appear to be involved in circuitry previously implied in schizophrenia, and not in the control areas. The circuitry areas include the left cortical areas BA46, 9, 8, 6, 7, the limbic lobe cortex, amygdala, as well as the right cerebellar cortex and dentate nucleus and, subcortically, the left thalamus, putamen and globus pallidus. Although BA7 is not part of the prefrontal system formally, it has strong direct interconnections with the dorsal prefrontal areas BA46, 9, 8. These brain areas are involved in several circuits relevant to schizophrenia including the ‘prefrontal system’ linking prefrontal and limbic cortices with associated crossed cerebellar, and thalamic and basal ganglia loops.46, 48, 49 Thus, there is a consistent association between both genes and the prefrontal and dorsal neocortical circuits relevant to both the memory task and schizophrenia deficits.

In Figure 3 the distribution of permuted P-values is shown across a single portion of chromosome 3 and chromosome 6, by brain area. This figure shows multiple imaging phenotype results simultaneously. The pattern of peaks (low P-values) is localized to one area of chromosome 3 and one area of chromosome 6, and appears strongly in BA46 and functionally related brain areas, but not in the two control areas (right uncus and hypothalamus). The number of statistically significant SNPs in the depicted regions of approximately 7 and 2 million base pairs, respectively, is generally limited to the ARHGAP18 and RSRC1 genes, rather than randomly distributed.

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Imaging Genophenotypes in the discovery sample. The relationship between activation in brain areas and single nucleotide polymorphisms (SNPs) on chromosomes 6 and 3 from the quantitative trait analysis are depicted. Each tube represents activation in a different brain area. The brain areas are grouped to explore similarities within the significant genes (SNPs) in related areas forming anatomical circuits. Peaks in the tubes represent permuted P-values (plotted as –log(P)) for all SNPs represented over an approximately 7 million base pair region of chromosome 6 around ARHGAP18 (lower left) and 2 million base pairs on chromosome 3 around RSRC1 (lower right). The height of each tube depicts the significance level for those SNPs related to brain activation in each brain area. The brain areas are represented by their Brodmann area designation or their commonly used anatomical labels. The specific RS numbers for SNPs bracketing the main peaks are listed in their approximate locations and mark the boundaries of RSRC1 and ARHGAP18. The directional arrows on the anatomical template demonstrate the implied anatomical circuitry for brain areas represented in figure. GP, globus pallidus; hyp, hypothalamus; BA46, dorsolateral prefrontal cortex.

Full figure and legend (198K)

Gene annotation

Gene ontology, canonical pathway and functional network analyses were executed using Ingenuity Pathways Analysis (Ingenuity Systems, Mountain View, CA, USA), Ensemble and SWISSPROT (http://ca.expasy.org/sprot/) tools; these indicated a potential DNA binding property and physical interaction with growth differentiation factor 9 precursor (GDF9) of the RSRC1 gene product. The dbSNP database (http://www.ncbi.nlm.nih.gov/projects/SNP/index.html) revealed intronic localization in the gene for the SNPs, respectively.

Exploring the same genetic databases revealed poor annotation for the three most significant SNPs that related to BA46 within ARHGAP18, rs11154490, rs9372944 and rs9388724. We found all to be intronic. Given that ARHGAP18 belongs to the RhoGAP family, members of this family may control aspects of synapse function. The ARHGAP18 gene products such as Rho GTPases are linked to RAS and EGFR-mediated proliferation, migration and differentiation of forebrain progenitors (EGFR, epidermal and transforming growth factor-α growth factor receptors (erb B)50). In addition, based on in silico analyses both genes have a function in dopamine, glutamate and fibroblast growth factor receptor signaling.

Corroboration in an independent case–control sample

To establish if our method of gene discovery using brain imaging as a quantitative phenotype could successfully identify genes related to schizophrenia, we tested the most promising genes discovered (RSRC1 and ARHGAP18) in an independent case–control study collected by the FBIRN consortium (www.nbirn.net). The statistically significant RSRC1 and ARHGAP18 SNPs from the discovery sample were analyzed in a case–control design, using a logistic regression analysis for small samples that included a permutation procedure to determine significance threshold P-value for each SNP (GLM BINREG45). Genotyping data were available for 17 of the 27 positive discovery sample SNPs for RSRC1 and for 18 of the 19 positive SNPs for ARHGAP18. Of the 17, 13 RSRC1 SNPs tested and 6 of the 18 ARHGAP18 SNPs tested were significantly associated with a diagnosis of schizophrenia in the corroborative sample, substantiating a function of RSRC1 and ARHGAP18 in the genetics of schizophrenia (see Figure 4 for summary; Tables 1 and 2). These significant SNPs in the logistic regression (P-values shown in the table) demonstrate the value of the imaging genetics approach in identifying genes associated with the disorder. These findings highlight the ability of our method using a quantitative brain imaging approach to identify genes related to schizophrenia.

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Summary of the statistically significant single nucleotide polymorphisms (SNPs) associated with RSRC1 and ARHGAP18 in the discovery and corroborative samples. The top rectangles indicate the number of statistically significant SNPs from those that were tested in the discovery sample using brain imaging as a quantitative phenotype. Seventeen of the positive RSRC1 SNPs and eighteen of the positive ARHGAP18 SNPs from the discovery sample were measured in the Corroborative sample. The lower rectangles indicate the number of these SNPs that were also statistically significant in the case–control Corroborative study. The results for all SNPs in RSRC1 and ARHGAP18 are in Tables 1 and 2.

Full figure and legend (69K)

Top

Discussion

Imaging genetics provides a powerful method for identifying genetic variations that affect a QT of interest in neuropsychiatric populations. Admittedly, this method excludes genes or polymorphisms that do not influence differences in brain area activation, or the particular imaging phenotypes chosen. On the other hand, structural and functional brain imaging provide sensitive quantitative and objective measures of brain function in neuropsychiatric illness. Brain development, structure and function are heritable and have strong genetic influences. Using an imaging phenotype has face validity and biological relevance as it constrains the GWAS analyses.

Studies of schizophrenia must consider potential medication effects. Manoach21 in reviewing fMRI studies in schizophrenia pointed out that working memory deficits are observed in medicated, unmedicated and medication–naïve patients as well as their siblings, concluding that such deficits are present regardless of medication22, 23 and have at least, in part, a genetic basis. Using the Sternberg working memory task as activation in an fMRI twin study of schizophrenia and healthy controls, evidence was found for both the behavioral and fMRI measures being heritable traits.24 Using the same task in schizophrenia, BOLD activation patterns involving the DLPFC differed by DRD1 genotype despite equivalent memory performance.25 The degree of heritability has yet to be defined although working memory performance decreased with increasing genetic load among schizophrenic twins discordant for schizophrenia and control twins.26 The Sternberg task-related abnormalities are found in the relatives of persons with schizophrenia as well as in the patients both on medication, and with no medication, which is also consistent with a heritable trait.21, 27

We address the issue of false positives, always a concern in such high-dimensional data, from both statistical and biological viewpoints. The imaging data are summarized by averaging BOLD activation over an atlas-determined ROI. The primary ROI of interest, the DLPFC was chosen a priori. The hundreds of thousands of imaging-genetic analyses are corrected for chance findings by permutation tests.51 The anatomically and/or functionally connected regions in the brain should show a similar pattern of genotype influence. This is not a statistical consideration, but a biologically driven one. Finally, these SNPs identified in an Imaging Genetics analysis become candidates which must be replicated in an independent sample.

We have provided a proof of concept of this novel approach using activation in the left DLPFC as a quantitative phenotype. The Imaging Genetics analyses included massively parallel analyses of all 105950 SNPs in conjunction with summary imaging results. We picked an ROI based on known differences between patients and controls from the published literature.22, 23, 28, 29, 33, 48 This ROI, used as a quantitative phenotype, identified genes which were previously unlinked to schizophrenia through molecular hypotheses—and these genes showed case–control differences in an independent sample, where they would never have been considered using a standard candidate gene approach.

The QT can be determined based on the literature as in this study or can be empirically derived. In the latter application, the brain imaging patterns between the patient population and normal healthy controls in the data set are contrasted to allow a choice of ROIs that best distinguish the groups. We then generate summary measures on activation patterns in those ROIs for each patient and each control subject. The SNPs that influence these brain activation quantitative phenotypes are identified through a series of GLM calculations. The statistical approach is built upon a GLM that combines imaging phenotypes, disease diagnosis and genetic data in a single model:

Imaging phenotype=genotype effect+diagnosis effect+genotype–diagnosis interaction effect

The value of this latter application is that it includes the diagnosis-by-genotype interaction, as well as the ability to add additional terms for gene–gene interactions. The GLM models the effect of each SNP with the brain activation measure as the dependent variable. Significance of the interaction effect is determined through permutation testing or other appropriate methods. The genotype–diagnosis interaction represents disease effects on the neuroimaging phenotype that are influenced by normal genetic variation, thus, the conceptual fusion of the interaction between imaging and genetics by diagnosis. It is more than the additive effects of these domains, but rather their interaction, modeled explicitly in the analysis.

The correction for multiple testing is an ongoing point of research. The BOLD activation is a summary score of the activation of all imaging voxels in the designated ROI chosen for the dependent variable or QT. The most conservative Bonferroni correction is not appropriate to SNP data as they typically do not meet the assumption of independence of multiple tests. Other methods used to correct for the risk of falsely concluding for a positive association, that is increasing the risk of the frequency of false positives, range from the Benjamini–Hochberg proposal52 adapted for genome analyses by Storey and Tibshirani53 with their FDR ‘correction’, to the Nyholt54 and Meng et al.55 methods that consider the dependency across SNPs. Some methods however, like Nyholt's and Meng's, are well-suited for a ‘small’ SNP set, for example as SNPs across a gene or in a chromosomal region, but are not easily generalizable to whole genome association studies. Other methods establish a sample-based significance threshold by a permutation approach.8, 56 Any correction for type I errors should be balanced with the risk of increasing the false-negative results. We used permutation testing, as it makes no assumptions regarding independence of tests and uses the sample being tested to empirically determine appropriate statistical thresholds.8, 29 Permutation is a preferred method for determining the likelihood of chance findings; the advantages are discussed in Hirschhorn and Daly,8 Balding,57 Dudbridge and Koeleman.58

There are six published GWAS in schizophrenia.32, 59, 60, 61, 62, 63 For the most part these studies have not studied the SNP panels that we studied and were not sufficiently powered to confirm our findings. Only the study by Kirov et al. used the same SNPs profile and of the other studies only two of our positive ARHGAP18 SNPs were overlapping. Each of these GWAS has identified 1 or 2 genes (SNPs) that have passed genome-wide significance. None of the studies provide data or P-values below the genome-wide significance levels for the overlapping RSRC1 and ARHGAP18 SNPs that were positive in our study. The actual P-values are necessary to confirm our finding for these two genes that do not require the genome-wide threshold of significance for testing the entire genome.

Up to half of all genes are expressed in the brain (assuming 22–25K genes). Any or all of the SNPs tested in this GWAS could correlate with brain activation. We expected to find some clearly spurious results identifying genes that are not expressed in the brain. Empirically, we have determined that this is a relatively rare event in this data set. The top eight most significant results, implicate the same two genes. Most of the analyzed SNPs were not related to task-specific brain activation used in this study. Further, investigation of gene annotation shows all the significant SNPs represent genes that are expressed in the brain, providing additional face validity.

The sign of the regression coefficient is positive for 17 of the 19 β-coefficients for the 19 SNPs in RSRC1 or AHRGAP18 that are significant for both the discovery and the case–control samples. A positive coefficient means that the minor allele increases the phenotype mean, in the QT analysis. An increase in the phenotype mean implies that those subjects with the minor allele activate their DLPFC more than those without the minor allele. Schizophrenia subjects are expected to have a greater deviant activation, that is, less activation with a comparatively high memory load than controls; therefore having less of the minor allele than healthy controls. This is precisely what was observed in the case–control analyses. Plenge et al.64 interpreted similar observations as ‘the minor allele was associated with protection’, but this interpretation is premature in our study. It is important to emphasize that SNPs on the Illumina chips used in our study were primarily chosen because they are tagging SNPs and not necessarily causal or coding SNPs. More detailed investigations of the areas identified by these tagging SNPs are needed to investigate which SNPs are causally linked to schizophrenia and to determine their functionality.

The results of the discovery and corroborative sample analyses are intriguing in several ways. Brain regions connected to left DLPFC (BA46) also showed a significant influence of RSRC1 and ARHGAP18 SNPs on brain activation measures, as shown in the figures, whereas control brain areas do not. These anatomically connected areas have several interesting features in common; all are neocortical regions that receive a dense dopamine innervation, all are highly interconnected with each other, and participate in a dorsal cortical circuitry that is consistently implicated in the etiology of schizophrenia, especially the DLPFC.48 Interestingly, these areas are associated with dopamine function especially of the D1 receptor class. We have not, however, tested the effects of these genotypes on brain activation in healthy controls; that will have to be the topic of future studies.

ARHGAP18 gene products belong to the human RhoGAP family with approximately 80 RhoGAP proteins known to be encoded in the human genome. The RhoGAPs, GTPase-activating proteins have the ability to modulate Rho-mediated signaling pathways by controlling the balance between active and inactive Rho proteins. Rho proteins belong to the Ras superfamily that is composed of over 50 members divided into 6 families, including Ras, Sar, Rho, Ran, Rab and Arf.65 They participate in an array of physiological processes, such as cell migration, intercellular adhesion, cytokinesis, proliferation, differentiation and apoptosis.66 Rho GTPases are important regulators of the actin cytoskeleton and consequently influence the shape and migration of cells. GTPases of the Rho family are strong regulators of signaling pathways that link growth factors and/or their receptors to adhesions and associated structures.67 One signaling pathway mediated by Ras is initiated by the EGF receptor (erb B epidermal growth factor receptor, EGFr), leading to cell proliferation. EGFr signaling can induce mitosis, proliferation, cell motility, differentiation and protein secretion.68 EGFr is localized on subventricular neural progenitor (svz) cells in the fetal and adult lateral ventricles, and these progenitors give rise to forebrain neurons in development and after injury in the adult.50 Thus, the ARHGAP18 gene products (Rho GTPases) are linked to Ras, and thus, to EGFr-mediated proliferation, migration and differentiation of forebrain progenitors. Therefore, our finding of ARHGAP18 SNPs related to DLPFC activation in schizophrenia is promising because schizophrenia has been linked to altered prenatal neurogenesis of cortical neurons, including those in dorsal prefrontal cortex.30 In addition, ARHGAP18 is precisely contained within 6q22-24, which has been shown to be linked to schizophrenia.69

RSRC1 is a unique marker of bone marrow-derived stem cells also expressing the cd34 marker.70 The cd34 marker is also found in a proliferating population of the subventricular zone (svz) stem cell population in the human fetal forebrain.71 Circulating bone marrow-derived cd34 stem cells can take up residence in subventricular niches in the developing and postnatal forebrain. These neural stem cells of the svz give rise to EGFr-responsive progenitors which can be induced to massively proliferate and migrate to lesion sites in the forebrain in the presence of transforming growth factor-α, which binds to the erb B EGFr receptor.50 Thus, there may be an association between RSRC1 variants, and the cd34-positive hematopoietic and neural stem cells in the progenitors that give rise to forebrain neural development.

Taken together, the present finding of a highly significant association between ARHGAP18 and RSRC1 SNPs with schizophrenia is intriguing as both genes have a function in prenatal brain development linking hematopoietic and neural stem cell proliferation in the svz and migration to forebrain structures such as the dorsal cortical stream and associated limbic lobe, striatal and amygdaloid circuitry.

The approach described is a screening method that makes GWAS data usable and exploratory in preparation for future studies, for example molecular studies, expression and transgenic studies, and all other functional genomic approaches. It allows for completely novel SNPs to be identified as having a function in the disease phenotype.

Top

Notes

Author contributions: The fMRI task, imaging data from the discovery sample and imaging analyses for these results were programmed and implemented by Jessica Turner; the neuroanatomical and neuroscience expertize and genetic annotation was contributed by James Fallon; the genetic data analysis, PLINK and Eigenstrat analyses and genetic annotation were performed by Guia Guffanti and Fabio Macciardi; the in silico annotations were performed by Anita Lakatos; the visualization and gene viewer methods were developed by David Keator; the design and oversight of the experiments and analyses were the responsibility of Steven Potkin. Article preparation was a joint effort of all authors.

Top

References

  1. McQueen MB, Devlin B, Faraone SV, Nimgaonkar VL, Sklar P, Smoller JW et al. Combined analysis from eleven linkage studies of bipolar disorder provides strong evidence of susceptibility loci on chromosomes 6q and 8q. Am J Hum Genet 2005; 77: 582–595. | Article | PubMed | ISI | ChemPort |
  2. Vicente AM, Macciardi F, Verga M, Bassett AS, Honer WG, Bean G et al. NCAM and schizophrenia: genetic studies. Mol Psychiatry 1997; 2: 65–69. | Article | PubMed | ChemPort |
  3. Kwasnicka-Crawford DA, Roberts W, Scherer SW. Characterization of an Autism-Associated Segmental Maternal Heterodisomy of the Chromosome 15q11-13 Region. J Autism Dev Disord 2006; 37: 694–702. | Article |
  4. Stein CM, Millard C, Kluge A, Miscimarra LE, Cartier KC, Freebairn LA et al. Speech sound disorder influenced by a locus in 15q14 region. Behav Genet 2006; 36: 858–868. | Article | PubMed |
  5. Chagnon YC. Shared susceptibility region on chromosome 15 between autism and catatonia. Int Rev Neurobiol 2006; 72: 165–178. | Article | PubMed |
  6. Ozaki K, Ohnishi Y, Iida A, Sekine A, Yamada R, Tsunoda T et al. Functional SNPs in the lymphotoxin-alpha gene that are associated with susceptibility to myocardial infarction. Nat Genet 2002; 32: 650–654. | Article | PubMed | ISI | ChemPort |
  7. Klein RJ, Zeiss C, Chew EY, Tsai JY, Sackler RS, Haynes C et al. Complement factor H polymorphism in age-related macular degeneration. Science 2005; 308: 385–389. | Article | PubMed | ISI | ChemPort |
  8. Hirschhorn JN, Daly MJ. Genome-wide association studies for common diseases and complex traits. Nat Rev Genet 2005; 6: 95–108. | Article | PubMed | ISI | ChemPort |
  9. Wang WY, Barratt BJ, Clayton DG, Todd JA. Genome-wide association studies: theoretical and practical concerns. Nat Rev Genet 2005; 6: 109–118. | Article | PubMed | ISI | ChemPort |
  10. Liu B. Statistical Genomics: Linkage, Mapping, and QTL Analysis. CRC Press: Boca Raton, 1997.
  11. Ebmeier K, Rose E, Steele D. Cognitive impairment and fMRI in major depression. Neurotox Res 2006; 10: 87–92. | PubMed | ChemPort |
  12. Dickerson BC. Functional MRI in the early detection of dementias. Rev Neurol (Paris) 2006; 162: 941–944. | PubMed | ChemPort |
  13. Yurgelun-Todd DA, Ross AJ. Functional magnetic resonance imaging studies in bipolar disorder. CNS Spectr 2006; 11: 287–297. | PubMed |
  14. Davis CE, Jeste DV, Eyler LT. Review of longitudinal functional neuroimaging studies of drug treatments in patients with schizophrenia. Schizophr Res 2005; 78: 45–60. | PubMed |
  15. Kircher TT, Thienel R. Functional brain imaging of symptoms and cognition in schizophrenia. Prog Brain Res 2005; 150: 299–308. | PubMed | ISI |
  16. Lawrie SM, Hall J, McIntosh AM, Cunningham-Owens DG, Johnstone EC. Neuroimaging and molecular genetics of schizophrenia: pathophysiological advances and therapeutic potential. Br J Pharmacol 2008; 153(Suppl 1): S120–S124. | Article | PubMed | ChemPort |
  17. Turner JA, Smyth P, Macciardi F, Fallon JH, Kennedy JL, Potkin SG. Imaging phenotypes and genotypes in schizophrenia. Neuroinformatics 2006; 4: 21–49. | Article | PubMed |
  18. Roffman JL, Weiss AP, Goff DC, Rauch SL, Weinberger DR. Neuroimaging-genetic paradigms: a new approach to investigate the pathophysiology and treatment of cognitive deficits in schizophrenia. Harv Rev Psychiatry 2006; 14: 78–91. | Article | PubMed | ISI |
  19. Meyer-Lindenberg A, Zink CF. Imaging genetics for neuropsychiatric disorders. Child Adolesc Psychiatr Clin N Am 2007; 16: 581–597. | Article | PubMed |
  20. Glahn DC, Thompson PM, Blangero J. Neuroimaging endophenotypes: strategies for finding genes influencing brain structure and function. Hum Brain Mapp 2007; 28: 488–501. | Article | PubMed |
  21. Manoach DS. Prefrontal cortex dysfunction during working memory performance in schizophrenia: reconciling discrepant findings. Schizophr Res 2003; 60: 285–298. | Article | PubMed | ISI |
  22. Barch DM, Carter CS, Braver TS, Sabb FW, MacDonald III A, Noll DC et al. Selective deficits in prefrontal cortex function in medication-naive patients with schizophrenia. Arch Gen Psychiatry 2001; 58: 280–288. | Article | PubMed | ISI | ChemPort |
  23. Callicott JH, Egan MF, Mattay VS, Bertolino A, Bone AD, Verchinksi B et al. Abnormal fMRI response of the dorsolateral prefrontal cortex in cognitively intact siblings of patients with schizophrenia. Am J Psychiatry 2003; 160: 709–719. | Article | PubMed | ISI |
  24. Karlsgodt KH, Glahn DC, van Erp TG, Therman S, Huttunen M, Manninen M et al. The relationship between performance and fMRI signal during working memory in patients with schizophrenia, unaffected co-twins, and control subjects. Schizophr Res 2007; 89: 191–197. | Article | PubMed |
  25. Tura E, Turner JA, Fallon JH, Kennedy JL, Potkin SG. Multivariate analyses suggest genetic impacts on neurocircuitry in schizophrenia. Neuroreport 2008; 19: 603–607. | Article | PubMed |
  26. Cannon TD, Keller MC. Endophenotypes in the genetic analyses of mental disorders. Annu Rev Clin Psychol 2006; 2: 267–290. | Article | PubMed |
  27. Manoach DS, Press DZ, Thangaraj V, Searl MM, Goff DC, Halpern E et al. Schizophrenic subjects activate dorsolateral prefrontal cortex during a working memory task, as measured by fMRI. Biol Psychiatry 1999; 45: 1128–1137. | Article | PubMed | ISI | ChemPort |
  28. Perlstein WM, Carter CS, Noll DC, Cohen JD. Relation of prefrontal cortex dysfunction to working memory and symptoms in schizophrenia. Am J Psychiatry 2001; 158: 1105–1113. | Article | PubMed | ISI | ChemPort |
  29. Manoach DS, Gollub RL, Benson ES, Searl MM, Goff DC, Halpern E et al. Schizophrenic subjects show aberrant fMRI activation of dorsolateral prefrontal cortex and basal ganglia during working memory performance. Biol Psychiatry 2000; 48: 99–109. | Article | PubMed | ISI | ChemPort |
  30. Akbarian S, Bunney Jr WE, Potkin SG, Wigal SB, Hagman JO, Sandman CA et al. Altered distribution of nicotinamide-adenine dinucleotide phosphate-diaphorase cells in frontal lobe of schizophrenics implies disturbances of cortical development. Arch Gen Psychiatry 1993; 50: 169–177. | PubMed | ISI | ChemPort |
  31. Papassotiropoulos A, Stephan DA, Huentelman MJ, Hoerndli FJ, Craig DW, Pearson JV et al. Common Kibra alleles are associated with human memory performance. Science 2006; 314: 475–478. | Article | PubMed | ISI | ChemPort |
  32. Almasy L, Gur RC, Haack K, Cole SA, Calkins ME, Peralta JM et al. A genome screen for quantitative trait loci influencing schizophrenia and neurocognitive phenotypes. Am J Psychiatry 2008; 165: 1185–1192. | Article | PubMed |
  33. Egan MF, Goldberg TE, Kolachana BS, Callicott JH, Mazzanti CM, Straub RE et al. Effect of COMT Val108/158 Met genotype on frontal lobe function and risk for schizophrenia. Proc Natl Acad Sci USA 2001; 98: 6917–6922. | Article | PubMed | ChemPort |
  34. Meyer-Lindenberg A, Nichols T, Callicott JH, Ding J, Kolachana B, Buckholtz J et al. Impact of complex genetic variation in COMT on human brain function. Mol Psychiatry 2006; 11: 867–877, 797. | Article | PubMed | ISI | ChemPort |
  35. Glahn DC, Paus T, Thompson PM. Imaging genomics: mapping the influence of genetics on brain structure and function. Hum Brain Mapp 2007; 28: 461–463. | Article | PubMed |
  36. de Geus E, Goldberg T, Boomsma DI, Posthuma D. Imaging the genetics of brain structure and function. Biol Psychol 2008; 79: 1–8. | Article | PubMed |
  37. Aleman A, Swart M, van Rijn S. Brain imaging, genetics and emotion. Biol Psychol 2008; 79: 58–69. | Article | PubMed |
  38. Shaw P, Gornick M, Lerch J, Addington A, Seal J, Greenstein D et al. Polymorphisms of the dopamine D4 receptor, clinical outcome, and cortical structure in attention-deficit/hyperactivity disorder. Arch Gen Psychiatry 2007; 64: 921–931. | Article | PubMed |
  39. van Haren NE, Bakker SC, Kahn RS. Genes and structural brain imaging in schizophrenia. Curr Opin Psychiatry 2008; 21: 161–167. | Article | PubMed |
  40. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition. (SCID-I/P). Biometrics Research New York State Psychiatric Institute: New York, 2002.
  41. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Non-patient Edition. (SCID-I/NP). Biometrics Research New York State Psychiatric Institute: New York, 2002.
  42. Lancaster JL, Woldorff MG, Parsons LM, Liotti M, Freitas CS, Rainey L et al. Automated Talairach atlas labels for functional brain mapping. Hum Brain Mapp 2000; 10: 120–131. | Article | PubMed | ISI | ChemPort |
  43. Price A, Patterson N, Plenge R, Weinblatt M, Shadick N, DA R. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 2006; 38: 904–909. | Article | PubMed | ISI | ChemPort |
  44. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D et al. PLINK: a tool set for whole-genome association and population-based linkage analysis. Am J Hum Genet 2007; 81: 559–575. | Article | PubMed | ISI | ChemPort |
  45. StataCorp. Stata Statistical Software Release 10. Stata Corporation: College Station, Texas, 2007.
  46. Heimer L. Basal forebrain in the context of schizophrenia. Brain Res Brain Res Rev 2000; 31: 205–235. | Article | PubMed | ChemPort |
  47. Petrides M. Lateral prefrontal cortex: architectonic and functional organization. Philos Trans R Soc Lond B Biol Sci 2005; 360: 781–795. | Article | PubMed |
  48. Fallon JH, Opole IO, Potkin SG. Neuroanatomy of schizophrenia: circuitry and neurotransmitter systems. Clin Neurosci Res 2003; 3: 77–107. | Article | ChemPort |
  49. Potkin SG, Alva G, Fleming K, Anand R, Keator D, Carreon D et al. A PET study of the pathophysiology of negative symptoms in schizophrenia. Positron emission tomography. Am J Psychiatry 2002; 159: 227–237. | Article | PubMed |
  50. Fallon J, Reid S, Kinyamu R, Opole I, Opole R, Baratta J et al. In vivo induction of massive proliferation, directed migration, and differentiation of neural cells in the adult mammalian brain. Proc Natl Acad Sci USA 2000; 97: 14686–14691. | Article | PubMed | ChemPort |
  51. Nichols T, Hayasaka S. Controlling the familywise error rate in functional neuroimaging: a comparative review. Stat Methods Med Res 2003; 12: 419–446. | Article | PubMed |
  52. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Stat Soc B 1995; 57: 289–300. | ISI |
  53. Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci USA 2003; 100: 9440–9445. | Article | PubMed | ChemPort |
  54. Nyholt DR. A simple correction for multiple testing for single-nucleotide polymorphisms in linkage disequilibrium with each other. Am J Hum Genet 2004; 74: 765–769. | Article | PubMed | ISI | ChemPort |
  55. Meng Z, Zaykin DV, Xu CF, Wagner M, Ehm MG. Selection of genetic markers for association analyses, using linkage disequilibrium and haplotypes. Am J Hum Genet 2003; 73: 115–130. | Article | PubMed | ISI | ChemPort |
  56. Dudbridge F, Gusnanto A, Koeleman BP. Detecting multiple associations in genome-wide studies. Hum Genomics 2006; 2: 310–317. | PubMed | ChemPort |
  57. Balding DJ. A tutorial on statistical methods for population association studies. Nat Rev Genet 2006; 7: 781–791. | Article | PubMed | ISI | ChemPort |
  58. Dudbridge F, Koeleman BP. Efficient computation of significance levels for multiple associations in large studies of correlated data, including genomewide association studies. Am J Hum Genet 2004; 75: 424–435. | Article | PubMed | ISI | ChemPort |
  59. Kirov G, Zaharieva I, Georgieva L, Moskvina V, Nikolov I, Cichon S et al. A genome-wide association study in 574 schizophrenia trios using DNA pooling. Mol Psychiatry 2008; DOI:10.1038/mp.2008.33. | Article |
  60. Lencz T, Morgan TV, Athanasiou M, Dain B, Reed CR, Kane JM et al. Converging evidence for a pseudoautosomal cytokine receptor gene locus in schizophrenia. Mol Psychiatry 2007; 12: 572–580. | Article | PubMed | ISI | ChemPort |
  61. Mah S, Nelson MR, Delisi LE, Reneland RH, Markward N, James MR et al. Identification of the semaphorin receptor PLXNA2 as a candidate for susceptibility to schizophrenia. Mol Psychiatry 2006; 11: 471–478. | Article | PubMed | ISI | ChemPort |
  62. Shifman S, Johannesson M, Bronstein M, Chen SX, Collier DA, Craddock NJ et al. Genome-wide association identifies a common variant in the reelin gene that increases the risk of schizophrenia only in women. PLoS Genet 2008; 4: e28. | Article | PubMed | ChemPort |
  63. Sullivan PF, Lin D, Tzeng JY, van den Oord E, Perkins D, Stroup TS et al. Genomewide association for schizophrenia in the CATIE study: results of stage 1. Mol Psychiatry 2008; 13: 570–584. | Article | PubMed | ChemPort |
  64. Plenge RM, Cotsapas C, Davies L, Price AL, de Bakker PI, Maller J et al. Two independent alleles at 6q23 associated with risk of rheumatoid arthritis. Nat Genet 2007; 39: 1477–1482. | Article | PubMed | ChemPort |
  65. Takai Y, Sasaki T, Matozaki T. Small GTP-binding proteins. Physiol Rev 2001; 81: 153–208. | PubMed | ISI | ChemPort |
  66. Symons M. Rho family GTPases: the cytoskeleton and beyond. Trends Biochem Sci 1996; 21: 178–181. | Article | PubMed | ISI | ChemPort |
  67. Kozma R, Ahmed S, Best A, Lim L. The Ras-related protein Cdc42Hs and bradykinin promote formation of peripheral actin microspikes and filopodia in Swiss 3T3 fibroblasts. Mol Cell Biol 1995; 15: 1942–1952. | PubMed | ISI | ChemPort |
  68. Wells A. EGF receptor. Int J Biochem Cell Biol 1999; 31: 637–643. | Article | PubMed | ISI | ChemPort |
  69. Cao Q, Martinez M, Zhang J, Sanders AR, Badner JA, Cravchik A et al. Suggestive evidence for a schizophrenia susceptibility locus on chromosome 6q and a confirmation in an independent series of pedigrees. Genomics 1997; 43: 1–8. | Article | PubMed | ISI | ChemPort |
  70. Zhang QH, Ye M, Wu XY, Ren SX, Zhao M, Zhao CJ et al. Cloning and functional analysis of cDNAs with open reading frames for 300 previously undefined genes expressed in CD34+ hematopoietic stem/progenitor cells. Genome Res 2000; 10: 1546–1560. | Article | PubMed | ISI | ChemPort |
  71. Rakic S, Zecevic N. Early oligodendrocyte progenitor cells in the human fetal telencephalon. Glia 2003; 41: 117–127. | Article | PubMed |
  72. Ge D, Zhang K, Need AC, Martin O, Fellay J, Urban TJ et al. WGAViewer: software for genomic annotation of whole genome association studies. Genome Res 2008; 18: 640–643. | Article | PubMed | ChemPort |
Top

Acknowledgements

This research was supported by grants to the Transdisciplinary Imaging Genetics Center (TIGC-P20 RR020837-01) and to the Functional Imaging Biomedical Informatics Research Network (FBIRN-1 U24 RR021992) from the National Center for Research Resources (NCRR) at the National Institutes of Health (NIH) and by grants POCEMON (FP7-ICT-2007-216088), FIRB Italia-Israele (RBIN04SWHR) and HYPERGENES (HEALTH-F4-2007-201550) and by an anonymous foundation. The Broad Institute Center for Genotyping and Analysis is supported by grant U54 RR020278-01 from the NCRR. We acknowledge the help and support of Mita Mancini and Yann Legros from Illumina, as well as Cristina Barlassina, Chiara Dal Fiume, Alessandro Orro and Federica Torri (University of Milan) for performing the HumanHap300 Bead Array procedures, as well as Liv Trondsen and Divya Rajpoot (UCI) for editorial support. We also acknowledge the recruitment, evaluation and SCID-based diagnostic assessment of healthy controls and schizophrenic subjects by FBIRN investigators: John Lauriello and Juan Bastillo, University of New Mexico; Daniel O'Leary, University of Iowa; Kelvin Lim, University of Minnesota; Gregory McCarthy, Judith Ford, Yale University; Arthur Toga, Tyrone Cannon, UCLA; Randy Gollub, Harvard University; Aysenil Belger, University of North Carolina; Dana Nguyen, Diane Highum, University of California, Irvine. We acknowledge the helpful comments of William E Bunney and Hal Stern.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)