Common functional polymorphisms of DISC1 and cortical maturation in typically developing children and adolescents

Abstract

Disrupted-in-schizophrenia-1 (DISC1), contains two common non-synonymous single-nucleotide polymorphisms (SNPs)—Leu607Phe and Ser704Cys—that modulate (i) facets of DISC1 molecular functioning important for cortical development, (ii) fronto-temporal cortical anatomy in adults and (iii) risk for diverse psychiatric phenotypes that often emerge during childhood and adolescence, and are associated with altered fronto-temporal cortical development. It remains unknown, however, if Leu607Phe and Ser704Cys influence cortical maturation before adulthood, and whether each SNP shows unique or overlapping effects. Therefore, we related genotype at Leu607Phe and Ser704Cys to cortical thickness (CT) in 255 typically developing individuals aged 9–22 years on whom 598 magnetic resonance imaging brain scans had been acquired longitudinally. Rate of cortical thinning varied with DISC1 genotype. Specifically, the rate of cortical thinning was attenuated in Phe-carrier compared with Leu-homozygous groups (in bilateral superior frontal and left angular gyri) and accelerated in Ser-homozygous compared with Cys-carrier groups (in left anterior cingulate and temporal cortices). Both SNPs additively predicted fixed differences in right lateral temporal CT, which were maximal between Phe-carrier/Ser-homozygous (thinnest) vs Leu-homozygous/Cys-carrier (thickest) groups. Leu607Phe and Ser704Cys genotype interacted to predict the rate of cortical thinning in right orbitofrontal, middle temporal and superior parietal cortices, wherein a significantly reduced rate of CT loss was observed in Phe-carrier/Cys-carrier participants only. Our findings argue for further examination of Leu607Phe and Ser704Cys interactions at a molecular level, and suggest that these SNPs might operate (in concert with other genetic and environmental factors) to shape risk for diverse phenotypes by impacting on the early maturation of fronto-temporal cortices.

Introduction

Disrupted-in-schizophrenia-1 (DISC1) has become the focus of intense investigation in psychiatric genetics since it was first described as spanning the chromosome 1 breakpoint of a balanced t(1:11) translocation¹ that had been found to segregate with schizophrenia (SCZ), bipolar affective disorder and major depressive disorder (MDD) in a large Scottish family.² This initial observation of apparent diagnostic nonspecificity has since been supplemented by cytogenetic, linkage and association studies implicating the DISC1 locus, and variants within it to diverse psychiatric phenotypes including SCZ (reviewed in Chubb et al.³), schizoaffective disorder,⁴ affective disorders (both bipolar affective disorder and MDD),⁵ autism spectrum disorders,^6,7 cognitive and symptom profile among people with SCZ,^8–11 and both personality traits¹² and cognitive ageing^13,14 in non-clinical samples.

A potential explanation for how DISC1 variants might confer susceptibility to such a wide range of conditions comes from cell and animal studies that firmly establish DISC1 as having a key role in development of the cerebral cortex through its influence on core processes such as progenitor cell division, radial migration of neurons into the cortical plate, neurite outgrowth and arborization, and postnatal neurogenesis.^3,15 Just as DISC1 seems to be intimately involved in cortical development, post mortem^16–18 and in vivo structural magnetic resonance imaging (sMRI)^19,20 studies indicate that aberrant cortical development before adulthood is likely to be a key component of the developmental neurobiology of many conditions that have been linked to DISC1. Furthermore, these conditions often have prodromal or frank symptom onset during childhood and adolescence—when the cortex is known to undergo dramatic structural remodeling in typical development.^21–23 These observations raise the as yet untested hypothesis that variations in DISC1 functioning may impart risk for psychopathology by modulating cortical maturation during these crucial developmental phases.

Available studies regarding the role of DISC1 in development draw heavily on inferences from the mouse.^24–26 One strategy for better understanding how genetic variation in DISC1 relates to cortical development in humans would be to focus on polymorphisms within DISC1 that (i) are sufficiently common to allow meaningful statistical comparisons of brain sMRI images between groups of individuals bearing different genotypes, (ii) have shown some (although not necessarily unequivocal²⁷) association with mental disorder in association studies, (iii) are known to alter protein expression, posttranslational modification or function and (iv) for which there is preferably some previous evidence linking genotype to differences in cortical structure or function in humans. Only two genetic variants within DISC1 are currently known to meet all these criteria—the single-nucleotide polymorphisms (SNPs) Leu607Phe (rs6675281) and Ser704Cys (rs821616). Table 1 details the manner in which each SNP fulfills the above criteria.

Table 1.

Summary of studies supporting the use of Leu607Phe and Ser704Cys in DISC1 imaging-genetic investigations

Characteristic	Leu607Phe (rs6675281)	Ser704Cys (rs821616)
Common	Minor allele frequency: 0.14	Minor allele frequency: 0.31
Possible association with disorder/symptom profilea	Schizophrenia8,28,29,30	Schizophrenia/schizoid traits10–12,35
	Schizoaffective disorder4	Major depressive disorder36
	Major depressive disorder31	Bipolar affective disorder37
	Bipolar affective disorder31	Autism spectrum disorder6
Allele most commonly linked to increased risk for disorder	Phe	Ser
Alters DISC molecular behavior	Reduced DISC1-PCM co-localization, and neurotransmitter release in Phe vs Leu.32,33 Disruption of DISC1-PCM co-localization in animal models causes delayed migration of neurones into the developing cortex.32 Increased brain and lymphocyte expression of short splice variants in Phe vs Leu. These shorter variants are enriched in the brains of some individuals with schizophrenia, and result in the omission of exons mediating important DISC1-partner-protein interactions.34	Higher rates of DISC1 oligomerization in Ser vs Cys.38 DISC1 oligomerization is important for normal DISC1–NDEL1 interactions involved in nuerite outgrowth.39,40 Reduced activation in Eph receptor B2 (ERK) and Akt pathways in Cys vs Ser. These pathways have been implicated in cytoskeletal remodeling, neurite outgrowth and cell survival.36
Positive previous imaging-genetic studies	Reduced cortical volume within frontal and temporal lobes in Phe-carriers vs Leu-homozygotes.8,30	Associated with structural and functional variation within frontal cortices and hippocampus.10,36,41–43

Abbreviations: DISC1, disrupted-in-schizophrenia-1; ERK, extracellular signal-regulated kinase; NDEL1, nuclear distribution gene E homolog-like 1; PCM, pericentriolar material 1;SNP, single-nucleotide polymorphism.

^aNegative associations studies exist for both SNPs.^5,28

Available DISC1 ‘imaging-genetic’⁴⁴ studies have made important first steps in relating genotype at Leu607Phe and Ser704Cys to cortical anatomy in humans. However, these studies have related DISC1 genotype to cross-sectional sMRI measures in adults, using sample sizes that have precluded Leu607Phe and Ser704Cys being related to cortical anatomy concurrently, within the same sample. Therefore, it remains unknown if Leu607Phe and Ser704Cys influence cortical maturation during childhood and adolescence, and the extent to which these two SNPs—which fall within distinct regions that encode different DISC1 partner–protein interaction domains—will show overlapping vs unique relationships with cortical anatomy.

In this study, we related genotype at both Leu607Phe and Ser704Cys to sMRI-derived measures of cortical thickness (CT) change over time, in a large longitudinal sample of typically developing children and adolescents. We wished to test the hypotheses that fronto-temporal cortical maturation varies as a function of genotype at Leu607Phe and Ser704Cys, and that these SNPs show both overlapping and unique regional associations with CT. To facilitate comparison with previous imaging-genetic investigations of these variants and to optimize sample size in each genotype group, participants were grouped as Phe-carriers vs Leu-homozygotes (PheCar/LeuLeu), and Cys-carriers vs Ser-homozygotes (CysCar/SerSer). We chose CT as the morphometric index of interest because it is a highly heritable aspect of cortical anatomy,⁴⁵ and its developmental trajectories (i) follow a well characterized pattern,²² (ii) are known to be altered in many of the disorders with which DISC1 has been associated,^19,46,47 (iii) show sensitivity to genotype at other functional genetic polymorphisms⁴⁸ and (iv) can be characterized in an accurate and reliable manner at several thousand points (vertices) across the cortical sheet using high-throughput automated approaches, which allows for spatially non-biased examination of cortical maturation in large samples.⁴⁹ There are also several methodological and theoretical advantages to using CT rather than voxel-based measures of ‘gray matter density’ as a spatially non-biased index of cortical anatomy.⁵⁰

Materials and methods

Participants

Please see Table 2 for full details of participant demographics and genotype composition. We included 255 healthy Caucasian children and adolescents (108 females) from 179 families, on whom a total of 598 sMRI scans had been acquired between the ages of 9 and 22 years. Participants were recruited through local advertisement. The absence of neurological or psychiatric illness was established through completion of a screening questionnaire (childhood behavior checklist), and a structured diagnostic interview administered by a child psychiatrist.⁵¹ Participants were of mixed handedness (handedness established using physical and neurological examination of soft signs).

Table 2.

Participant characteristics

Feature	By Leu607Phe genotype				By Ser704Cys genotype
	LeuLeu	PheCar	Test statistic	P-Value	CysCar	SerSer	Test statistic	P-Value
Participants, no.	193	62			131	124
Sex, no.
M	107	40	χ2 = 1.6	0.13	80	67	χ2 = 1.2	0.3
F	86	22			51	57
Estimated IQ, mean (s.d.)	113 (12.3)	114 (11.7)	t = −0.7	0.5	114 (12.2)	113 (12.2)	t = 0.7	0.5
SES, mean (s.d.)	42 (18.9)	45 (19.1)	t = −1.3	0.2	43 (18.4)	43 (19.6)	t = 0.2	0.9
Right handed, no. (%)	160 (83)	57 (91)	χ2 = 2.0	0.2	117 (89)	100 (81)	χ2 = 2.7	0.1
Number of scans, no.
1	53	69			33	40
2	60	18			47	31
≥3	85	24			51	53
Age at each scan in years, mean (s.d.)
First scan	12.3 (2.8)	12.2 (2.9)	t = 0.1	0.9	12.7	11.8	t = 2.3	0.02
Second scan	14.8 (2.6)	15 (3.1)	t = −0.7	0.5	15.1	14.5	t = 1.4	0.2
Third scan	17.1 (2.6)	16.9 (2.5)	t = 0.3	0.8	16.7	17.3	t = −1.2	0.2
Genotype group at other SNP, no.
CysCar	97	34	χ2 = 0.4	0.5	—	—
SerSer	96	28			—	—
Scans, no.	454	144			313	285
Age distribution of scans, years
Mean (s.d.)	14.5 (3.5)	14.5 (3.6)			14.7 (3.5)	14.2 (3.6)
Range	9–22.8	9–22.8			9–22.8	9–22.8

Abbreviations: IQ, intelligence quotient; SES, socioeconomic status; SNP, single-nucleotide polymorphism.

All participants had a full-scale intelligence quotient (IQ) of > 80 (IQ was estimated using age-appropriate Wechsler intelligence scales⁵²). Socioeconomic status was quantified using Hollings-head scales.⁵³ The institutional review board of the National Institutes of Health approved the research protocol used in this study and written informed consent and assent to participate in the study were obtained from parents and children respectively.

Genotyping

For each participant, DNA was extracted from previously prepared lymphoblastoid cell lines using standard methods (Qiagen, Alameda, CA, USA). It has been established that conversion of cells into lymphoblastoid lines does not cause errors into SNP genotyping.⁵⁴ Genotyping was performed by Prevention Genetics (Marshfield, WI, USA), using sub-microliter allele-specific polymerase chain reactions.⁵⁵ DNA sequencing of positive controls was conducted to ensure correct assignment of genotypes. Allele frequencies were Leu 0.88, Ser 0.70—in keeping with reference Centre d’Etude Polymorphism Humain (CEPH) data for populations of European descent. Genotype frequencies were LeuLeu 0.76, LeuPhe 0.23, PhePhe 0.01/SerSer 0.49, SerCys 0.43 and CysCys 0.01. Genotype frequencies did not deviate from Hardy–Weinberg equilibrium for either SNP regardless of whether related individuals were included or not (Leu607Phe P = 0.5, Ser704Cys P = 0.6). Genotype at one SNP was independent of genotype at the other (P = 0.5).

Neuroimaging

Of all 225 participants with at least one brain sMRI scan, 60% had two or more, and 15% had three or more scans. Scans were acquired at approximately 2-year intervals. All sMRI scans were T-1 weighted images with contiguous 1.5 mm axial slices and 2.0 mm coronal slices, obtained on the same 1.5-T General Electric (Milwaukee, WI, USA) Signa scanner using a three-dimensional spoiled gradient recalled echo sequence with the following parameters: echo time, 5 ms; repetition time, 24 ms; flip angle 45°; acquisition matrix, 256 × 192; number of excitations, 1; and field of view, 24 cm. Head placement was standardized as described previously. Native MRI scans were submitted to the CIVET pipeline (version 1.1.8) (http://wiki.bic.mni.mcgill.ca/index.php/CIVET) to generate separate cortical models for each hemisphere as described previously.⁵⁶ Briefly, this automated set of algorithms begins with linear transformation, correction of non-uniformity artifacts, and segmentation of each image into white matter, gray matter and CSF.⁵⁷ Next, each image is fitted with two deformable mesh models to extract the white/gray and pial surfaces. These surface representations are then used to calculate CT at approximately 40 000 vertices per hemisphere (MacDonald et al., 2000).⁵⁸ A 30 mm bandwidth blurring kernel was applied, the size of which was selected to maximize statistical power while minimizing false positives—as determined by population simulation.⁵⁹ The validity of these vertex-based estimated of CT is well established.²²

Statistical analysis

Demographic characteristics between genotype groups were compared using analysis of variance and two-sample t-tests for continuous variables, and χ² test for categorical variables. General linear models with age at scan as a dependent variable were used to confirm that interactions of demographic characteristics (for example, IQ and handedness in combination) were not unevenly distributed across the age range studied.

In neuroimaging analyses, we modeled—at each vertex—the fixed effects of age, Leu607Phe genotype, Cys704Ser genotype and interactions between these terms, while controlling for gender differences in CT. Mixed model regression was used as it permits the inclusion of multiple measurements per person at different ages, and irregular intervals between measurements, thereby increasing statistical power.⁶⁰ We included a nested random effects term that modeled within family and within person dependence of observations. All models were run with and without handedness, IQ and socioeconomic status as main effects, and in interaction with genotype terms. These terms were not included in the final model because (i) their inclusion did not add significantly to the predictive power of the model (as determined using a likelihood ratio test), (ii) their addition did not alter the distribution of significant findings for DISC1 genotype terms and (iii) handedness, IQ and socioeconomic status were evenly distributed between genotype groups and across age. Therefore, in the final model, at each vertex, CT for ith family’s jth individual’s kth time-point was modeled as:

$${\text{CT}}_{\mathit{\text{ijk}}}=\text{Intercept}+d_i+d_{\mathit{\text{ij}}}+{\beta }_1(\text{sex})+{\beta }_2(\text{mean age})+{\beta }_3(\text{Leu}607\text{Phe})+{\beta }_4(\text{Ser}704\text{Cys})+{\beta }_5(\text{Leu}607\text{Phe}\times \text{Ser}704\text{Cys})+{\beta }_6(\text{mean age}\times \text{Leu}607\text{Phe})+{\beta }_7(\text{mean age}\times \text{Ser}704\text{Cys})+{\beta }_8(\text{mean age}\times \text{Leu}607\text{Phe}\times \text{Ser}704\text{Cys})+e_{\mathit{\text{ijk}}}$$

Leu607Phe and Ser704Cys were binary categorical variables with levels LeuLeu, PheCar and CysCar, SeSer, respectively. Age was centered to allow for interpretation of genotype group differences at the average age (14.5 years) rather than age 0. A decision was made to include a linear (rather than non-linear) trajectory because (i) preliminary analyses within this data set established that higher order age terms were not better able to predict variance in CT than linear age, and (ii) our previous work has shown that over the age range included in this study—the predominant effect of age is linear.²² Results are reported after the application of a q = 0.05 false discovery rate⁶¹ threshold across P-values for all fixed-effect terms excluding the intercept. For any vertex in which an interaction term between two or more main effects terms was significant, significant constituent main effects are not presented.

Results for each fixed-effect term of interest were visualized by projection of the t-statistics for that term at all vertices onto a standard brain template. T-statistic maps (Figures 1 and 2) were also presented as ‘Binarized-maps’ (Figure 3 and Supplementary Figure 1) with significant clusters for any one term (that is, LeuPhe × age or SerCys × age) corresponding to one color, thus allowing multiple terms to be shown alongside each other on the same brain template.

Figure 1.

Vertex maps of regions showing statistically significant fixed differences in CT between genotype groups. Phe607Leu top panel. Ser704Cys bottom panel. Phe carriers (PheCar) and Ser (SerSer) homozygotes had thinner cortices than their counterparts in the regions shown. ‘Warmer’ colors indicate thickness differences of greater statistical significance. Results have been corrected for multiple comparisons using a false discovery rate correction with q = 0.05. Note the marked overlap in the effect of both SNPs in right lateral temporal cortices.

Figure 2.

Vertex maps of regions showing statistically significant differences in the rate of cortical thinning between genotype groups. Phe607Leu top panel. Ser704Cys bottom panel. Phe carriers (PheCar) showed a significant attenuation of cortical thinning relative to Leu homozygotes (LeuLeu) in the colored regions shown. The inset plot illustrates estimated genotype group trajectories for the left superior frontal focus. Ser homozygotes (SerSer) showed a significant acceleration of cortical thinning in the colored regions shown. The inset plot illustrates estimated genotype group trajectories for the left posterior superior temporal focus. In all instances ‘Warmer’ colors indicate thickness trajectory differences of greater statistical significance. Results have been corrected for multiple comparisons using a false discovery rate correction with q = 0.05. Note the more extensive distribution of Leu607Phe effects compared with Ser704Cys, and the absence of overlap between regions wherein each SNP is associated with rate of CT loss.

Figure 3.

A summary vertex map. Colored regions indicate where either (Leu607Phe only in green, Ser704Cys only in purple) or both (red) SNPs showed a statistically significant association with either the rate of cortical thinning or fixed differences in CT that did not change with age. In all, 20% of all vertices fall within colored regions. The inset bar chart plots mean CT for each of the four possible ‘haplotypes’ in the region of overlapping SNP effects on fixed differences in CT.

Results

Demographic characteristics

LeuLeu and CysCar groups did not differ significantly from PheCar and SerSer groups by gender, handedness or IQ, respectively. Neither were these demographic characteristics significantly different across the four ‘haplotype’ groups (results not shown). Age at scan was not significantly predicted by any interactions between genotype, gender, handedness and IQ. Full-scale IQ did not differ significantly between genotype groups. Full demographic details are provided in Table 2.

DISC1 effects on CT that did not vary significantly with age

Leu607Phe genotype was uniquely associated with CT in left middle temporal, left primary sensory, superior parietal and right inferior parietal cortices, in which PheCar were always thinner than LeuLeu (see upper panel of Figure 1). Ser704Cys genotype was uniquely associated with CT in bilateral supplementary motor cortices, in which SerSer were always thinner than CysCar (see Figure 1 lower panel). Leu607Phe and Ser704Cys genotype were both significantly associated with CT (PheCar thinner than LeuLeu, SerSer thinner than CysCar) in a large confluent right lateral temporal region encompassing inferior, middle and superior temporal gyri. Across haplotypes groups (see Figure 3), CT was greatest in the LeuLeu/CysCar group, and smallest in the PheCar/SerSer group.

DISC1 effects on the rate of CT change

Leu607Phe influenced the rate of CT change with age in bilateral superior and medial frontal cortices (although more so on the left), and left angular and middle temporal cortex. In all these regions CT thinning with age was attenuated in PheCar relative to LeuLeu. In the left superior frontal cortex for example, this translated into LeuLeu undergoing a 7% reduction in CT between ages 9 and 23 years, whereas no CT change was observed in PheCar (see upper panels of Figure 2). At age 9 years, CT in this region was significantly less in PheCar compared with LeuLeu (t = −4.3, P = 0.0005), but by age 23 there was a significant group difference in the opposite direction (t = 2.0, P = 0.05). The spatial extent of Ser704Cys influences on the rate of CT change was less pronounced than that for Leu607Phe, and was restricted to the left anterior cingulate, and regions within the left middle and superior temporal cortices. In all these areas, rate of CT thinning with age was increased in SerSer compared with CysCar. In the anterior cingulate, for example, CT reductions in SerSer between 8 and 22 were twice that observed in CysCar (see lower panels of Figure 2). At age 9 years, CT in this region was significantly less in CysCar compared with SerSer (t = −2.3, P = 0.03), but by age 23 the direction of this group difference had inverted, and there was no longer a statistically significant difference between the two groups in CT (t = −1.7, P = 0.1). Maps of regions in which Leu607Phe and Ser704Cys influenced the rate of CT change did not overlap.

Interaction between Leu607Phe and Ser704Cys

Significant interaction between Leu607Phe and Ser704Cys in predicting age-invariant differences in CT was restricted to a small region within the left primary sulcus and adjacent sensory cortex. In this region, PheCar was only associated with significant thinning relative to LeuLeu in the presence of SerSer (see Supplementary Figure 1—right-hand panel).

Significant interactions between Leu607Phe and Cys704Ser in predicting the rate of CT change with age were restricted to small regions within left orbitofrontal, right lateral parietal and right middle temporal cortices. In all these regions, CT changed little with age among those with the PheCar/CysCar genotype combination, while CT reduced robustly with age in all three other genotype combinations (see Supplementary Figure 1—left-hand panel).

Summary

Approximately 20% of all cortical vertices showed a significant relationship with Leu607Phe and/or Ser704Cys genotype (see Figure 3). These relationships (either in terms of the rate of CT change over time, or in terms of static differences between genotype groups in CT) adopted a prominent front-temporal distribution (bilateral superior frontal, medial frontal, superior temporal and middle temporal cortices), as well as including bilateral supplementary motor, and right lateral parieto-occipital cortices. Other than a single large right lateral temporal region in which there was overlap between maps of Leu607Phe and Ser704Cys influence on CT, genotype at each SNP seemed to influence CT in different cortical regions (see Figure 3).

Discussion

Our study design enabled us to conduct the first longitudinal examination of the relationship between polymorphisms in DISC1 and cortical structure in humans, as well as to consider both Leu607Phe and Cys704Ser in parallel within the same sample. In keeping with our first hypothesis, the relationships between these SNPs and CT were most marked in fronto-temporal cortical regions of demonstrated relevance to the neurobiology of many of the psychiatric disorders for which DISC1 is a candidate risk gene. With regards to our second hypothesis, we observed both overlap and dissociation between the spatial distribution of SNP effects on CT. The majority of SNP effects were spatially dissociated, but overlap was observed in right lateral temporal cortex (in which genotype at both Leu607Phe and Ser704Cys were associated with static differences in CT in an additive manner), and smaller regions in the right orbitofrontal, right parietal and left sensorimotor cortices (in which the effect of genotype at one SNP on the rate of CT loss was dependent on genotype at the other).

Available imaging-genetic studies of Leu607Phe and Ser704Cys find these SNPs to influence cortical anatomy in regions that overlap with those highlighted in our study. Both of the available studies relating genotype at Leu607Phe to cortical anatomy^8,30 report reduced gray matter volume in PheCar adults compared with LeuLeu adults within superior frontal regions that overlap with those in which we report PheCar as showing attenuated rates of CT loss with age relative to LeuLeu during childhood and adolescence. For Ser704Cys, Hashimoto et al.³⁷ report gray matter volume reductions in CysCar adults compared with SerSer, within cingulate cortices that include the anterior cingulate region in which we find CysCar to have attenuated rates of CT loss with age compared with SerSer. However, closer comparison between our findings and existing DISC1 imaging-genetics studies is complicated by several methodological differences. First, previous studies examined the main effect of each SNP on cortical anatomy without controlling for genotype at the other, and as such the SNP main effects reported are not equivalent. Second, sample size differences between our study and previous work has implications for the risk of sample bias and type II error. Third, we explicitly examined longitudinal trajectories of CT in youth, whereas previous studies have been cross-sectional studies of cortical volume in adults. Two false assumptions would be made in attempting to infer cross-sectional genotype group differences in cortical volume during adulthood by extrapolating genotype group differences in longitudinal CT change during youth; that cortical volume and CT are neurobiologically equivalent,¹⁹ and that the linear trajectories we describe between 9 and 22 years of age can be extended into adulthood.²² Finally, in some previous DISC1 genetic-imaging studies, the effect of genotype has been examined while combining patients and controls in the same analyses,^8,30 whereas we focused on typically developing children and adolescents (although it is possible that some of the participants in our study may go on to manifest psychiatric symptoms when older). The relationship between genotype and cortical anatomy in clinical groups may be confounded by genotype-dependent differences in neuropsychological⁸ or symptom³⁰ profile, and treatment responsiveness. These features could alter cortical maturation in their own right, and the confounding effects of such secondary changes may well become more pronounced over time. For these reasons, examining gene–brain relationships in younger samples of healthy, asymptomatic controls may provide a more confound-free means of identifying the neurostructural correlates of genetically determined variations in DISC1 functioning. Our findings on adopting this strategy have a number of implications for our understanding of DISC1 neurobiology and its role in shaping brain development and risk for psychopathology.

First, multiple sMRI studies have reported differences in CT between typically developing controls and individuals suffering from conditions linked to DISC1 within those regions in which we show cortical anatomy to be influenced by Leu607Phe and Ser704Cys genotype (SCZ—bilateral superior frontal, superior temporal and angular gyrus,^19,48 bipolar affective disorder—lateral temporal cortices⁶²). Our findings also overlap with regions of altered CT in individuals at increased risk for, but not manifesting symptoms of disorders linked to DISC1 (SCZ—left prefrontal and bilateral temporal cortices,⁶³ MDD—right superior frontal and supplementary motor cortex⁶⁴), suggesting that altered structural cortical maturation in such regions is not purely a consequence of illness, but indexes heritable risk for disorder. Therefore, the associations we report between genotype at Leu607Phe and Ser704Cys and CT suggest a neurodevelopmental mechanism through which these variants might act to alter risk for psychiatric disorder. It remains unclear if Leu607Phe and Ser704Cys influence cortical maturation by affecting on (i) the way DISC1 functions in the cortex during adolescence or (ii) ‘foundational’ roles had by DISC1 during earlier cortigogenesis, which then have the delayed consequence of derailing adolescent cortical development. Although not mutually exclusive, these hypotheses can be contrasted in animal models.⁶⁵ Future neuroimaging studies using methods for image analysis that permit assessment of earlier-established morphometric properties of the cortex such as sulcal patterning may also help to further parse the developmental timing of DISC1 influences on cortical development.

Second, our findings in a sample of typically developing individuals indicate that genotype differences at Leu607Phe and Ser704Cys, and the associated variation in fronto-temporal cortical maturation clearly are not, in isolation, sufficient for psychiatric illness (or prodromal developmental problems) to become apparent. This is consistent with the fact that these are relatively common SNPs, and (based on available genetic association study data) not highly penetrant for psychiatric phenotypes, implying that other genetic and environmental factors must interact with genotype at Leu607Phe and Ser704Cys, and presumably the intermediate phenotypes they impart, to influence behavioral phenotype. In support of this model, (i) both SNPs impact on interactions between DISC1 and specific molecular systems (for example, pericentriolar material-1 and extracellular signal-regulated kinase), which are themselves known to regulate cortical development and harbor genetic variants conferring risk for SCZ and MDD,^33,66 (ii) genotype at Ser704Cys conditions genetic associations between SCZ and polymorphisms within NDEL1 (nuclear distribution gene E homolog-like 1, formerly known as NUDEL or endooligopeptidase A (EOPA)) which interacts with DISC1 through a domain containing residue 704, and has its own well-established role in cortical maturation⁶⁷ and (iii) the recent finding that Leu607Phe and Ser704Cys can alter DISC1 posttranslational splicing greatly increases the scope for interaction between these SNPs and other genetic and environmental factors.³⁵ Our findings highlight the processes underlying fronto-temporal cortical maturation during adolescence as a potentially ‘common pathway’ on which such interactions might converge, and within which the search for DISC1 interactors may prove fruitful.

Third, DISC1 influences on CT are observed in 20% of the cortical surface, and although there are some cortical regions in which both SNPs influence CT in an additive (right lateral temporal), or interacting (left primary sulcus, right middle temporal, orbitofrontal and parietal) manner, there is otherwise remarkably little overlap between the regions in which each SNP is associated with CT. Spatial or temporal differences in cortical DISC1 expression could potentially distinguish those regions in which either SNP is associated with CT from those in which neither SNP is associated with CT, and further studies characterizing these expression patterns are required. As neither Leu607Phe nor Ser704Cys seem to influence DISC1 expression,⁶⁸ then the presence or absence of overlap between Leu607Phe and Ser704Cys effects within a given cortical region may reflect convergence or divergence (respectively) in that region of the molecular pathways influenced by each SNP. The potential for divergence is apparent in the fact that Leu607Phe and Ser704Cys cause functionally relevant changes in DISC1 amino acid sequence within C-terminal sub-domains that are distinct in the molecular interactions they mediate,³ meaning that regional differences in the expression of these interactors could lead to regions differences in the consequence of genotype at each Leu607Phe and Ser704Cys. Conversely, convergence is allowed for by the fact that the DISC1 ‘interactome’ is extensive, and includes many partner–proteins that not only interact richly with each other, but also have a role in common cellular processes.⁶⁹ Further studies characterizing the spatio-temporal distribution of DISC1 partner–protein expression in the cortex, and examining the molecular consequences of allelic variation at Leu607Phe and Ser704Cys in tandem would help test these hypotheses.

Our study suffers from certain limitations. First, we have modeled age using a linear term only despite having previously shown CT change to follow more complex cubic or quadratic trajectory. However, a linear model for CT change is appropriate over almost all the cortex between the upper and lower age range of our study.²² Given the dramatic inter-individual variation in cortical anatomy, very large longitudinal samples are required to have the statistical power required to describe group differences in CT change using non-linear terms. Second, as in previous imaging-genetic studies of Leu607Phe and Ser704Cys, allele frequencies at each SNP meant that we did not have sufficient subjects to compare dominant vs recessive models. Third, the degree of overlap and dissociation between SNP effects is partly dependent on the statistical threshold applied. Fourth, we were not able to test for interaction between the common functional DISC1 SNPs we studied and polymorphism within key DISC1 interactors such as pericentriolar material-1 and NDEL1. Finally, the cellular events underlying CT change with age as measured by sMRI remain poorly understood, although data exist suggesting that dendritic pruning⁷⁰ or encroachment of white matter into the cortical mantle⁷¹ may contribute to CT reduction with age. Further studies will be required to examine how the relationship between DISC1 genotype and brain maturation is modulated by other genetic and environmental factors, and to extend the neuroanatomical phenotype beyond the cerebral cortex.

In spite of these limitations, our study represents the first longitudinal assessment of the relationship between DISC1 genotype and cortical development in humans, and suggests that the molecular consequences of allele variation at Leu607Phe and Ser704Cys may modulate risk for conditions such as SCZ, bipolar affective disorder and MDD, by affecting on the maturation of front-temporal cortical systems.