|Home | About | Journals | Submit | Contact Us | Français|
Dystrobrevin binding protein 1 (DTNBP1) has been identified as putative schizophrenia susceptibility gene, but it remains unknown whether polymorphisms relate to altered cerebral structure. We examined relationships between a previously implicated DTNBP1 risk variant [P1578] and global and segmented brain tissue volumes and regional cortical thickness in schizophrenia (n=62; 24 risk carriers) and healthy subjects (n=42; 11 risk carriers), across ethnic groups and within Caucasians. Schizophrenia patients showed similar brain volumes, but significantly reduced brain-size adjusted gray matter and CSF volumes and cortical thinning in a widespread neocortical distribution compared to controls. DTNBP1 risk associated with reduced brain volume, but not with tissue sub-compartments. Cortical thickness, which was weakly associated with brain size, showed regional variations in association with genetic risk, although effects were dominated by highly significant genotype by diagnosis interactions over broad areas of cortex. Risk status associated with regional cortical thinning in patients, particularly in temporal networks, but with thickness increases in controls. DTNBP1 effects for brain volume and cortical thickness appear driven by different neurobiological processes. Smaller brain volumes observed in risk carriers may relate to previously reported DTNBP1/cognitive function relationships irrespective of diagnosis. Regional cortical thinning in patient, but not in control risk carriers, may suggest that DTNBP1 interacts with other schizophrenia-related risk factors to affect laminar thickness. Alternatively, DTNBP1 may influence neural processes for which individuals with thicker cortex are less vulnerable. Although DTNBP1 relates to cortical thinning in schizophrenia, morphological changes in the disorder are influenced by additional genetic and/or environmental factors.
Schizophrenia has a substantial heritable component (~80%) where genetic in addition to harmful environmental factors appear to account for disease processes (Harrison and Weinberger 2005; Sullivan, et al. 2003). Epidemiological research suggests that multiple genes of small to moderate influence underlie the genetic transmission of schizophrenia (Sullivan, et al. 2003). Accordingly, a growing number of loci and positional susceptibility genes have been implicated in the search for factors that affect disease inheritance. However, findings across laboratories lack strong correspondence, and the role of any specific gene has not been established with certainty (Harrison and Weinberger 2005; Norton, et al. 2006; Sanders, et al. 2008; Sullivan, et al. 2003). Imaging data also suggests that some brain abnormalities observed in schizophrenia may serve as markers for disease-related genetic predisposition in at-risk, but as of yet unaffected, individuals (Bearden, et al. 2007; Glahn, et al. 2007). Understanding how genes shown to convey a risk for schizophrenia affect changes in brain structure or function may elucidate the role of putative risk genes in the etiology of the disorder and/or provide information concerning the underlying biological mechanisms of specific genetic factors in general. At present, the relationships between schizophrenia susceptibility genes and structural imaging phenotypes are largely uncharacterized.
Of the susceptibility genes identified, moderate evidence exists to suggest dystrobrevin binding protein 1 or dysbindin (DTNBP1) (Straub, et al. 2002), located on chromosome 6p22.3, is a risk factor for schizophrenia. DTNBP1 has been implicated in many independent genetic linkage and association studies, e.g., (Duan, et al. 2007; Riley and Kendler 2006; Tochigi, et al. 2006; Tosato, et al. 2007; Vilella, et al. 2008), although associations at the genome-wide level have not yet been confirmed and there are failures to replicate across different samples (Li and He 2007; Peters, et al. 2008; Sanders, et al. 2008; Turunen, et al. 2007). The molecular mechanisms of DTNBP1, a protein that binds alpha and beta dystrobrevin, are not completely understood. Studies implicate its involvement in postsynaptic density functions and presynaptic glutamatergic transmission (Numakawa, et al. 2004). DTNBP1 also appears to influence neuron viability, and via glutamate signaling, to modulate the activity of dopamine and other neurotransmitters (Weickert, et al. 2004). Through biochemical disturbances involving dopaminergic and glutamatergic systems (Pearlson 2000), DTNBP1 may also relate to neuropil abnormalities reported in schizophrenia (Rajkowska, et al. 1998; Selemon and Goldman-Rakic 1999; Selemon, et al. 2003). DTNBP1 is widely distributed in the central nervous system (Harrison and Weinberger 2005) with expression reported in frontal and temporal cortices and within the hippocampus, thalamus and basal ganglia (Talbot, et al. 2004; Weickert, et al. 2004), regions highly implicated in the structural and functional neuropathology of schizophrenia (Clinton and Meador-Woodruff 2004; Honea, et al. 2005; Mamah, et al. 2007; Shenton, et al. 2001). Notably, existing data support that variation in the DTNBP1 gene acts to increase risk for schizophrenia through reduced expression (Bray, et al. 2005), which has been demonstrated in prefrontal (Weickert, et al. 2004) and hippocampal (Talbot, et al. 2004; Weickert, et al. 2008) regions.
Schizophrenia imaging studies report structural alterations in brain regions contributing to many aspects of sensorimotor and higher cognitive function. In the presence of sub-threshold brain size reductions, relative deficits of gray matter are most reliably documented in lateral and medial temporal cortices and in frontal and parietal association regions (Narr, et al. 2005a; Narr, et al. 2005b; Shenton, et al. 2001). Some of these brain abnormalities appear to serve as intermediate phenotypes (or endophenotypes) for schizophrenia-related genetic predisposition, e.g., (Baare, et al. 2001b; Bearden, et al. 2007; Narr, et al. 2002; Seidman, et al. 2003). Therefore, genes conferring risk for schizophrenia may also modulate disease-associated changes in brain structure. While some prior studies have examined relationships between imaging endophenotypes and putative genetic risk factors, to our knowledge associations between DTNBP1 and variations in brain structure in schizophrenia and healthy subjects have not been addressed. However, at least two electrophysiological studies have shown relationships between DTNBP1 risk variants and brain function that suggest a role for this gene in prefrontal (Fallgatter, et al. 2006) and early sensory processing (Donohoe, et al. 2008). Evidence from behavioral studies further suggest that DTNBP1 influences prefrontal spatial working memory (Donohoe, et al. 2007), results that are consistent with recent evidence from animal data supporting that this gene is involved in working memory mediated by prefrontal cortices and long term memory mediated by the medial temporal lobe (Takao, et al. 2008). It thus appears plausible that DTNBP1 may also affect changes in the structural integrity of cortical networks shown as disturbed in schizophrenia. We therefore set out to investigate to examine relationships between DTNBP1 risk status and altered cerebral structure in the disorder.
A number of different DTNBP1 DNA polymorphisms have been linked with schizophrenia where discrepancies exist concerning the risk alleles and/or haplotypes genotyped and/or shown to associate with the disorder across studies (Mutsuddi, et al. 2006, also see: http://www.ncbi.nlm.nih.gov/entrez/dispomim.cgi?id=607145). However, several implicated single nucleotide polymorphisms (SNPs) appear in high allelic association and share a strong linkage disequilibrium (LD) structure (Li et al., 2007). In this current investigation we focused on examining genotype/imaging phenotype relationships using the DTNBP1 [P1578] variant previously shown to associate with schizophrenia (Funke, et al. 2004) and to associate with cognitive function in an overlapping sample (Burdick, et al. 2007; Burdick, et al. 2006), where general cognitive ability itself is shown to relate to brain size and regional variations of cortical thickness (Narr et al., 2007). We assessed associations between the DTNBP1 risk genotype with overall brain volume and the volume of each brain tissue compartment. Relationships between DTNBP1 risk status and gray matter thickness measured at high spatial resolution were examined to isolate more regional effects of genotype. Although DTNBP1 variants have been shown to associate with schizophrenia across racial categories, population stratification may account for prior discrepancies in results (Li and He 2007). We thus assessed DTNBP1/imaging phenotype relationships in schizophrenia patients (n=62) and healthy controls (n=42) while controlling for ethnic group, as well as in within patient (n=26) and control (n=25) Caucasian subjects only.
Study participants included 62 patients with schizophrenia (mean age: 27.6±6.2 SD; 46m/16f) and 42 healthy control subjects (mean age: 28.5±7.5; 16m/26f). Of these subjects, 26 patients (mean age: 27.5±7.9, 19m/7f) and 25 healthy controls (mean age: 26.9±6.5, 10m/15f) were categorized as Caucasian. Other ethnic categories included African Americans (25 patients; 11 controls), Hispanics (4 patients; 5 controls), Asian Americans (5 patients; 1 control) and 2 patients of mixed ethnicity. The gender and age composition of subjects by diagnosis and genotype are shown in Table 1 across ethnic categories and within Caucasians only. The participants of this study overlap with, but are not identical to those included in prior investigations of the structural imaging correlates of first-episode schizophrenia (Narr, et al. 2005a; Narr, et al. 2005b) and with prior studies of DTNBP1 and schizophrenia and cognition (Burdick, et al. 2007; Burdick, et al. 2006).
All patients were recruited through admissions to the Zucker Hillside Hospital in Glen Oaks, New York and were participating in clinical trials comparing the efficacy of antipsychotic medications. Of the 62 patients, 33 were receiving treatment at the time of the MRI exam where antipsychotics included (number in parentheses) clozapine (4), fluphenazine (1), haldol (3), olanzapine (17) or risperidone (7). One patient was receiving olanzapine and haldol. Medication data were unavailable for 7 patients. The diagnostic status of patients was confirmed with the Structured Clinical Interview for Axis I DSM-IV Disorders (SCID) (First, et al. 1997) and the Schedule for Affective Disorders and Schizophrenia – Change Version (SADS-C) (Endicott and Spitzer 1978). To characterize the sample clinically, two symptom cluster scores were computed based on items from a modified version of the Scale for the Assessment of Negative Symptoms (SANS) (Andreasen 1984) and the SADS-C with psychosis and disorganization items (SADS-C + PD). The positive symptom cluster score was computed as the average of the severity of delusions and hallucinations items from the SADS-C +PD. The negative symptom cluster score was computed as the average of the global ratings of affective flattening, alogia, avolition-apathy, and asociality from the SANS. In this sample the mean level of positive symptoms was 3.9± 1.7 SD and mean level of negative symptoms was 2.6±1.0. Data were unavailable for one subject.
Healthy comparison subjects were recruited through local newspaper advertisements and by community word of mouth. Control subjects had no history of psychiatric illness as determined by clinical interview using the non-patient SCID (SCID-NP). Study exclusion criteria for all participants included serious neurological or endocrine disorders, any other medical condition known to affect the brain, or meeting DSM-IV criteria for mental retardation. Education level differed significantly between patients (mean years of education: 12.89± 1.94 SD) and controls (mean: 15.04±2.47) (F(1,100)=23.47, p<.001). Information was missing for 3 patients. However, parental socio-economic status (SES) was similar between patient (mean: 3.09±1.94) and control groups (mean: 2.63±.955 SD) (p>.16). The North Shore – Long Island Jewish Health System IRB approved all procedures and informed written consent was obtained from all subjects.
Venous blood samples were used for the extraction of genomic DNA. Genotyping protocols have been detailed previously (Burdick, et al. 2006; Funke, et al. 2004). DTNBP1 SNPs were genotyped using the MALDI-TOF mass spectrometry by the Sequenom system where markers were checked to ensure criteria for Hardy–Weinberg equilibrium. Only subjects with alleles that could be assigned with a confidence of ≥ 95% were included for study. Our analyses focused on P1578 (rs1018381) shown to be a tagging SNP for a haplotype (CTCTAC) previously associated with increased risk for schizophrenia (Funke, et al. 2004) and to demonstrate relationships with cognitive phenotypes in patients and controls (Burdick, et al. 2007; Burdick, et al. 2006). Subjects were categorized according to whether they possessed at least one copy of the risk allele. That is, homozygote risk carriers were grouped with heterozygote carriers (Burdick, et al. 2007; Burdick, et al. 2006) and compared with non-carriers for analyses of genotype-imaging phenotype relationships.
High-resolution 3D SPGR MR images were obtained on a 1.5 Tesla scanner (General Electric, Milwaukee, WI) as a series of 124 contiguous 1.5 mm coronal brain slices (256×256 matrix, 0.86 mm × 0.86 mm in-plane resolution). Image processing procedures were performed as previously detailed (Narr, et al. 2005a; Narr, et al. 2005b). Image volumes were corrected for head tilt and alignment using a six parameter rigid-body transformation, edited to remove non-brain tissue and the cerebellum (scalp editing inter-rater reliability, rI= .99), corrected for signal intensity non-uniformities (Sled and Pike 1998), and segmented into voxels most representative of gray matter, white matter and CSF using a partial volume correction method (Shattuck, et al. 2001). Estimates of total brain size and of brain tissue compartments were obtained from the scalp-edited tissue classified volumes.
The cortical surfaces of each hemisphere were extracted from scalp edited image volumes (MacDonald, et al. 1994) and used to manually identify 29 sulcal/gyral landmarks by employing validated anatomic protocols for which inter and intra-rater reliability procedures have been reported (Narr, et al. 2005a; Narr, et al. 2005b). Cortical pattern matching methods (Narr, et al. 2005a; Narr, et al. 2005b; Thompson, et al. 2004; Thompson, et al. 2000) were then used to spatially relate homologous regions of cortex across subjects to allow relationships between DTNBP1 and gray matter thickness to be examined at high spatial resolution. Briefly, surface-warping algorithms use the manually derived sulcal/gyral landmarks as anchors to drive the same cortical anatomy into correspondence. This warping procedure thus assigns the same coordinate values to matching anatomic regions across subjects at high spatial density (65,536 coordinate locations) without imposing any scaling to individual brain volumes. Cortical thickness measures, defined as the shortest 3D distance without crossing CSF voxels from the cortical white-gray matter interface to the hemispheric surface, were then obtained at each spatially homologous cortical coordinate point.
Genotype/imaging phenotype relationships were assessed using the General Linear Model (GLM). Diagnosis (schizophrenia patients; healthy controls) and Genotype (carriers of the risk allele; non-carriers) were included as independent variables and interactions between Diagnosis and Genotype were examined. Whole brain and segmented brain tissue volumes (gray matter, white matter and CSF) were used as dependent variables. Sex and Age were included in the model and total brain volume was included as an additional covariate for analyses of segmented brain tissue volumes. Regional variations of cortical thickness were assessed using the same statistical model implemented using R (http://www.r-project.org/). However, the cube root of brain volume was included as a covariate given that cortical thickness represents a distance rather than a volume measure. Analyses of cortical thickness measures were performed at 65,536 coordinate locations to provide spatially detailed maps of results. These results were mapped to the average surface representation of the cortex where color indicates the probability values for significant regional effects. Analyses were performed in all subjects including ethnic category as a covariate and in Caucasian subjects separately.
A two-tailed alpha level of p<.05 was used as the threshold for statistical significance. However, since cortical thickness measures were made at thousands of spatially correlated surface locations, permutation testing was used to confirm the overall significance of regional results (Anderson and Legendre, 1999; Anderson and ter Braak, 2003). For permutation analyses, the number of surface points within the hemisphere that were significant using a more conservative threshold of p<.01 (two-tailed) were compared to the number of significant surface points that occurred by chance when subjects were randomly assigned to groups in 10,000 new randomized analyses.
Kolmogorov–Smirnov statistics determined that brain tissue volume measures were normally distributed. A main effect of Genotype was observed for whole brain volume, F(1,103)=3.78, p<.05, across all ethnic groups and within Caucasians only, F(1,50)=4.20, p<.05, where carriers of the risk allele exhibited smaller brain volumes compared to non-carriers [Fig 1.]. Effects of Genotype were not significant for any tissue compartment when controlling for whole brain volume (p<.10). However, main effects of Diagnosis were observed for brain size adjusted gray matter, F(1,103)=8.31, p<.005 (controls larger) and CSF F(1,103)=8.42, p<.005 (patients larger) across all races, although these effects were at trend level significance within Caucasian subjects only (gray matter, F(1,51)=3.20, p<.08; CSF, F(1,51)=2.50, p<.12). Interactions between Genotype and Diagnostic group were not observed for brain volume or brain tissue volume measures within Caucasians or across races. Means and standard deviations for brain tissue volumes are provided in Table 1 in groups defined by Diagnosis and Genotype. For each tissue compartment, values have been residualized for sex, age, and brain volume. The residuals were added to the mean volume of each measure for the entire sample to allow relative differences between groups to be compared in units comparable to the actual volumes. Figure 1. plots brain volume measures by risk status and diagnosis.
Cortical thickness results are mapped in Figures 2–4. Figure 2 shows the main effect of Diagnosis across all races (right) and within Caucasian subjects (left). Figure 3 shows the main effects of Genotype, where the bottom row shows significant interactions between Genotype and Diagnosis, mapped within all races and Caucasians separately. Figure 4 shows the effects of Genotype mapped within patients and control subjects separately.
For the main effect of schizophrenia mapped in Figure 2, red and pink indicate regions where healthy control subjects show significantly increased cortical thickness compared to schizophrenia patients; blue shades indicate regions where patients show increases compared to controls. Patterns of schizophrenia-related cortical thinning are similar to those previously reported in an overlapping sample (Narr, et al. 2005a; Narr, et al. 2005b) and apparent in supragenual cingulate and paracingulate cortices and the occipital pole medially. Laterally, pronounced cortical thinning is evident in heteromodal association and sensorimotor cortices including dorsolateral, parietal association areas and regions of the superior and middle temporal gyri. Cortical thinning effects appear similar when mapped across all races and within Caucasians separately. A discrete region in the vicinity of the medial gyrus rectus showed increased cortical thickness in patients with respect to controls. Permutation tests thresholded at p<0.01 confirmed the significance of schizophrenia-related cortical thinning within the left (corrected p-value=.002 and .001) and right hemisphere (corrected p-value=.006 and .015) for Caucasians and across all races respectively.
Main effects of Genotype mapped across all subjects (left) and within Caucasian subjects only (right) are shown in Figure 3. Red and pink indicate regions where carriers of the DTNBP1 risk allele show decreased cortical thickness with respect to non-carriers. Blue shades indicate where risk carriers show increased cortical thickness compared to non-risk carriers. Contrary to our expectations, Caucasian risk carriers showed some significant regional increases of cortical thickness in sensorimotor cortices (the paracentral lobule medially), the posterior superior and middle frontal gyrus, and the occipital poles compared to Caucasian non-carriers. Only relatively subtle reductions of cortical thickness were apparent in risk carriers compared to non-carriers. Main effects of Genotype mapped across all races were less pronounced: risk carriers showed increased thickness in the posterior middle frontal gyrus laterally and left anterior prefrontal cortex. Permutation analyses thresholded at p<.01 confirmed the significance of regional effects within Caucasian subjects only (corrected p-value=.017 and .02 for the left and right hemisphere respectively).
The interactions between Diagnostic Group and Genotype are mapped in the second row of Figure 3. The color bar represents directional effects that can be interpreted by examining the simple effects of Genotype mapped in Fig. 4. Pronounced interaction effects are apparent over much of the lateral left hemisphere both in Caucasians and across all races encompassing dorsolateral, sensorimotor, parietal, temporal and occipital association regions. Interaction effects in the right hemisphere were slightly less prominent and observed within dorsolateral prefrontal cortex, parietal association regions, and the superior temporal gyrus. Medially, effects were concentrated within left paracentral lobule and paracingulate cortices. Permutation analyses confirmed the overall significance of regional effects (left hemisphere corrected p-value=.002 for both Caucasians and across races, right hemisphere corrected p-value=.02 for Caucasians and .007 across all races).
Figure 4 shows the simple effects of Genotype mapped separately in healthy subjects (top row) and in patients with schizophrenia (bottom row). In healthy Caucasian risk carriers cortical thickness increases were observed within sensorimotor, premotor and dorsolateral prefrontal cortices, and within parietal-occipital association regions and the right superior temporal gyrus compared to Caucasian non-carriers. These effects, though less pronounced in sensorimotor cortices, showed a similar pattern when mapped across all races. Decreases of cortical thickness in risk carriers were apparent only in concentrated orbitofrontal clusters. Effects of Genotype mapped within patients with schizophrenia showed a very different pattern of results to those exhibited by healthy subjects. Risk carriers showed patterns of decreased--as opposed to increased--laminar thickness in distributed cortical regions, particularly within temporal cortices bilaterally in Caucasian patients and frontal, parietal and temporal association regions across all races. Increases in cortical thickness were mostly absent within patient risk carriers, with the exception of some clusters localized in the frontal and occipital poles. Permutation analyses for simple effects of genotype within diagnostic groups were significant for control Caucasians only (corrected p-value=.01 and .02 for the left and right hemispheres respectively). Corrected p-values for effects of genotype mapped across all healthy subjects irrespective of race and within patients were slightly above the threshold of significance (p>.10).
To reconcile the volumetric and cortical thickness findings reported above, post-hoc correlation analyses were performed to examine the relationships between brain tissue volume measures and cortical thickness measurements averaged across the entire cortical mantle in each subject. Results showed that although whole brain volume is highly correlated with gray matter (r=.89, p<.0001), white matter (r=.83, p<.0001) and CSF volume (r=.51, p<.0001), and average cortical thickness is significantly associated with both gray (r=.35, p<.001) and white matter volume (r=−.36, p<.001), average cortical thickness is not significantly associated with CSF (r=−.05, p=.61) or whole brain volume (r=.03, p=.81).
DTNBP1 is one of the more studied schizophrenia susceptibility genes, although it remains unknown whether variants of this gene relate to alterations of cerebral structure observed in disorder. This investigation thus sought to address relationships between brain tissue volumes and cortical thickness and a DTNBP1 risk variant. The primary results were as follows: (1) schizophrenia patients showed significantly reduced gray matter and increased CSF volumes (after correction for whole brain volume) and pronounced cortical thinning over large areas of sensorimotor and frontal, temporal, and parietal association cortices compared to healthy subjects; (2) carriers of the DTNBP1 risk allele showed significantly smaller whole brain volumes compared to non-carriers across diagnostic groups irrespective of race, although effects of genotype were not significant for any particular brain tissue compartment; (3) risk carriers showed some localized brain size-corrected increases in laminar thickness in frontal, sensorimotor and occipital cortex that survived permutation testing in Caucasian subjects only; however, (4) highly significant interactions between diagnostic group and DTNBP1 risk status were observed in wide areas of cortex within Caucasians and across racial categories, where (5) follow-up analyses of simple effects revealed that schizophrenia patients who are DTNBP1 risk carriers exhibit regional cortical thinning in distributed brain regions including temporal cortices, but healthy subjects who are risk carriers show some regional increases in cortical thickness.
Since carriers of the DTNBP1 risk allele showed significantly smaller whole brain volumes compared to non-carriers, but did not vary significantly in gray matter, white matter or CSF volume, results suggest the affects of genotype are not specific to a particular tissue class. In contrast, main effects of schizophrenia were absent for overall brain volumes, but were observed for brain size-adjusted gray matter and CSF volumes suggesting that different disease-related genetic or environmental factors account for global gray matter reductions and CSF enlargements widely observed in the disorder (Shenton et al., 2001). Even so, disease processes accounting for gray matter deficits in schizophrenia may overlap with those associated with DTNBP1 risk factors at a more local level. Since the thickness of the cortex reflects the size, density, and arrangement of neurons, neuroglia and nerve fibers, cortical thickness may more closely relate to disturbances in neural function and may point to systems-level changes in morphology potentially associated with specific genetic factors. Indeed, cortical thickness averaged across the cortex did not associate with overall brain volume, although was strongly associated with gray matter and white matter volume as expected. These findings are in line with prior imaging (Im, et al. 2008) and postmortem findings (Pakkenberg and Gundersen 1997) showing that brain size and cortical thickness do not scale proportionately, where brain size is more strongly associated with cortical surface area and convolution. It is thus plausible that brain size and cortical thickness are structural imaging phenotypes influenced by different neurobiological processes although gray and white matter volumes share some overlapping variance with both measures and cortical thickness relationships may vary at the regional level. Understanding the nature of the relationships between these measures may explain the differential influences of DTNBP1 risk on overall brain volume and regional changes in cortical thickness observed in this study.
In cortical thickness analyses, schizophrenia patients showed regional thinning in bilateral sensorimotor and heteromodal association cortices encompassing frontal, temporal and parietal cortical regions with respect to healthy subjects [Fig 2]. These results are similar to those documented in an overlapping sample of first-episode schizophrenia patients (Narr, et al. 2005a; Narr, et al. 2005b) and to findings observed in chronic schizophrenia (Kuperberg, et al. 2003), with the exclusion of sensorimotor cortex. Notwithstanding, a study of early onset schizophrenia reported prominent gray matter changes encompassing sensorimotor cortices that were shown to progress to relatively widespread areas of cortex across development (Thompson, et al. 2001b). Only a spatially discrete area within the vicinity of gyrus rectus showed increased thickness in patients. Since a prior longitudinal voxel-based study of gray matter density showed progressive gray matter decline in bilateral temporal and right frontal and parietal regions in subjects with high genetic risk of developing schizophrenia while healthy subjects showed gray matter reductions within the gyrus rectus only, it is possible that normal developmental processes affecting the gyrus rectus specifically may be delayed in schizophrenia (Job, et al. 2005). Few studies have addressed whether regional changes of laminar thickness may represent intermediate markers of genetic predisposition in schizophrenia. One study showed prominent gray matter deficits in left prefrontal, bilateral temporal cortices and more subtle effects in right prefrontal and inferior parietal cortices in nonpsychotic siblings of early onset patients that were shown to attenuate with age (Gogtay, et al. 2007). Another study utilizing an automated parcellation method showed reversed hemispheric asymmetry of cingulate thickness and a bilateral reduction of superior temporal sulcal thickness in healthy relatives (siblings and parents) of schizophrenia patients compared to controls (Goghari, et al. 2007). These studies and others examining gray matter volume or density (Cannon, et al. 2002; Cannon, et al. 1998) suggest that gray matter changes are influenced by schizophrenia-related genetic factors, particularly in frontal and temporal regions. However, interactions of genetic factors with the timing of neurodevelopmental processes (Gogtay, et al. 2007) and environmental factors including exposure to antipsychotic medication (Dorph-Petersen, et al. 2005) may additionally act to affect changes of brain morphology in the disorder.
Since patients with schizophrenia show significant cortical thinning over large areas of cortex and the DTNBP1 risk allele has been associated with schizophrenia in previous studies, we predicted that risk carriers would exhibit regional reductions in cortical thickness. Contrary to our expectations, DTNBP1 risk status was associated with increased thickness while decreases were subtle and only present in small, non-localized clusters [Fig. 4]. However, highly significant and broadly spread interactions were observed between the DTNBP1 variant and diagnosis in Caucasians and across all races [Fig. 4]. Subsequent analyses of the effects of genotype within each diagnostic group showed that the DTNBP1 risk-allele associates with some discrete increases of cortical thickness in healthy subjects, but is associated with regional thinning patients with schizophrenia, particularly in temporal cortices in Caucasians. These findings suggest that the DTNBP1 risk allele may be protective in controls, but contribute towards reductions of laminar thickness in particular brain systems in patients with schizophrenia. This argument, although speculative, may help explain why from an evolutionary standpoint schizophrenia risk genes are protected to a degree from natural selection. Available data suggest that multiple genes account for the heritability of schizophrenia (Harrison and Weinberger 2005; Sullivan, et al. 2003). Observations that the DTNBP1 risk allele associates differently with cortical thickness in patients than in healthy subjects may also suggest the presence of epistatic effects. That is, the effect of DTNBP1 risk on cortical thickness may depend on additional genetic factors conferring genetic risk for schizophrenia contributing to regional gray matter deficits in patients but not in healthy individuals who carry the DTNBP1 risk allele. Altered DTNBP1 expression has been documented in schizophrenia postmortem data (Talbot, et al. 2004; Weickert, et al. 2008; Weickert, et al. 2004). DTNBP1 expression is influenced by cis-acting variation at the DTNBP1 locus. However, a polymorphic trans-acting locus on chromosome 8p, which contains the widely implicated schizophrenia risk gene neuregulin-1 (NRG1), is also suggested to affect DTNBP1 expression (Bray, et al. 2008). These findings could thus support the hypothesis that a genetic interaction may account for the differential effects of the DTNBP1 risk allele on cortical thickness in schizophrenia patients and controls. It is also possible that genetic risk may interact with environmental factors to contribute to regional gray matter deficits in patients but not in healthy subjects. Alternatively, DTNBP1 may control other neurobiological processes that pose some risk for schizophrenia. Thus individuals who by virtue of these other genetic factors have a thicker cortex may be less vulnerable to the DTNBP1 effect, while those with thinner cortices are more vulnerable.
DTNBP1 associates with cognitive function at the phenotypic level. Notably, a large genome-wide study investigating the genetic contributors of intelligence showed linkage with chromosomal regions encompassing DTNBP1 (6p22.3–6p25.3) in healthy subjects (Posthuma, et al. 2005). DTNBP1 variants and/or haplotypes are further reported to associate with level of educational achievement in schizophrenia (Williams, et al. 2004), with cognitive ability in both schizophrenia and healthy individuals (Burdick, et al. 2007; Burdick, et al. 2006), and with intelligence in unaffected siblings of schizophrenia patients (Zinkstok, et al. 2007). Brain size and intelligence are highly heritable (Baare, et al. 2001a; Posthuma, et al. 2002; Thompson, et al. 2001a; Tramo, et al. 1998) and are demonstrated to be modestly related (McDaniel 2005; Narr, et al. 2007). Since overlapping genes appear to influence both phenotypes, it is possible that the main effect of the DTNBP1 risk allele on overall brain size also relates to general intellectual ability.
There are several limitations to the current study. Firstly, statistical power issues may have obscured the detection of more subtle regional effects of DTNBP1 risk status across and within groups. Moreover, small cell sizes precluded us from examining homozygote carriers of the DTNBP1 risk allele separately where effects may be more pronounced. Secondly, different risk and protective haplotypes and SNPs have been implicated across prior studies that could be influenced by ethnic background. Although the DTNBP1 marker examined in this study has shown associations with schizophrenia in studies including Caucasian, Hispanic, and Asian subjects (Funke, et al. 2004; Tang, et al. 2003; Tosato, et al. 2007; Van Den Bogaert, et al. 2003), this variant has not been associated with schizophrenia in African American samples specifically (Duan, et al. 2007; Funke, et al. 2004) although only three studies to our knowledge focused on this minority group (Duan, et al. 2007; Funke, et al. 2004; Pedrosa, et al. 2007). Unfortunately small cell sizes again precluded the analysis of DTNBP1 risk status within all ethnic groups separately. Thirdly, we did not examine other implicated DTNBP1 single or multi-marker variants in this investigation, but instead focused on one marker that has been implicated in several prior schizophrenia association studies (Funke, et al. 2004; Tang, et al. 2003; Tosato, et al. 2007; Van Den Bogaert, et al. 2003) and that has shown relationships with general intellectual ability (Burdick, et al. 2007; Burdick, et al. 2006). However, even in spite of the potential complications of population stratification, available information suggests that linkage disequilibrium likely exists between the variant examined in this study and other implicated DTNBP1 risk markers (Bray, et al. 2005). However, it is possible that more than one DTNBP1 liability variant exists or that the actual variant conferring risk for schizophrenia has not yet been isolated (Riley and Kendler 2006). These questions can potentially be addressed by future studies as new information concerning DTNBP1 risk factors for schizophrenia become available. Finally, it is necessary to note that some recent larger scale studies suggest that DTNBP1 does not account for a large measurable portion of the genetic risk for schizophrenia (Peters, et al. 2008; Sanders, et al. 2008).
In spite of some controversy concerning the role of DTNBP1 in the heritability of schizophrenia, our results suggest that DTNBP1 risk factors are linked with brain morphology in general and with regional changes in cortical thickness that differ between diagnostic groups. Localized cortical thinning in patient risk carriers, but not in control carriers, may suggest the interactions of other genetic factors or protective affects of the risk allele in controls and/or in individuals with thicker cortex. However, schizophrenia-associated cortical thinning in widely distributed neocortical areas is clearly also under the control of additional genetic and/or harmful environmental variables. Although the biological mechanisms underlying DTNBP1-brain morphometry relationships are not clear, our findings complement observations of genotype-neurocognitive relationships to further suggest DTNBP1 susceptibility variants affect overall brain size and specific cortical networks in schizophrenia.
This work was generously supported by research grants from the National Center for Research Resources (P41 RR13642), the National Institute of Mental Health (MH079800, MH001760, to AKM, K01 MH073990, to KLN, and R01 MH060374, to RMB), the NIH Roadmap for Medical Research (U54 RR021813, Center for Computational Biology (CCB); and UL1 RR024911, Consortium for Neuropsychiatric Phenomics (CNP)).