Introduction

Schizophrenia is a complex psychiatric disorder characterized by severe emotional, cognitive and social dysfunction. It usually has delusions, auditory verbal hallucinations, visual hallucinations, paranoia, emotional withdrawal, difficulty focusing, etc. Neuroimaging studies have reported widespread regional alteration throughout the brain in schizophrenia. Overall, the volumes or thickness of cortices are significantly reduced in schizophrenia, which has been consistently reported by most published studies, including frontal [prefrontal1, superior frontal Brodmann Areas BA9 and BA102,3,4, middle frontal BA93,5, rostral middle frontal6, and inferior frontal (inferior frontal5, orbital inferior frontal7, and orbito-frontal cortices8, and pars triangularis9 and inferior frontal sulci10)], temporal (inferior temporal11,12,13, middle temporal7,11,12,14,15, and superior temporal3,11,12,16,17,18,19,20,21,22), parietal (precuneus23), occipital (fusiform7,11,13,24,25,26 and lingual gyri27,28), and insular29,30,31,32,33 cortices, and limbic system (anterior8,10 and posterior cingulate gyri8,10, and amygdala1,34,35,36, hippocampus1,7,34,36,37,38 and thalamus39). Surprisingly, the gray matter volumes (GMVs) of basal ganglia (BG) (caudate nucleus1,40,41,42,43, putamen40,44,45, pallidum40,46, nucleus accumbens47, and substantia nigra48,49) have been reported to significantly increase in schizophrenia by most published studies, with only a couple of exception34,39. Among the five BG structures, enlargement of caudate nucleus and putamen in schizophrenia has been consistently reported to be the two most significant and important neural markers of schizophrenia1,40,41,42,43,44,45,50,51,52,53. The patients with enlarged putamen are usually sensitive to and benefit from treatment with antipsychotics that block dopamine neurotransmission44,45,50, as evidenced by the findings that patients with good treatment outcomes have larger putamen than those with poor outcomes or healthy controls, in support of enlarged putamen as a physiological correlate of neuroleptic responsiveness and a predictor of treatment outcome44,45,50.

The distinct alteration between cortices and subcortical BG in schizophrenia suggests that some schizophrenia-risk genes might regulate these structures. A genome-wide association study (GWAS) on GMVs of seven subcortical structures, including -BG (nucleus accumbens, caudate nucleus, putamen, and pallidum) and limbic system (amygdala, hippocampus and thalamus), identified that five genetic variants significantly influenced the volumes of putamen (KTN1, DCC, BCL2L1 and DLG2) and caudate nucleus (FAT3)54. The strongest effect was found between putamen and rs945270 at 3′ flanking to KTN1. KTN1 encodes kinectin 1 receptor, which regulates neuronal cell shape and volume54,55,56,57, and thus kinectin 1 expression in BG may regulate the volumes of these structures.

KTN1 has been reported to play important roles in many neuropsychiatric or neurodegenerative diseases/phenotypes, including attention-deficit/hyperactivity disorder (ADHD)58,59, Parkinson’s disease (PD)60,61,62,63, heroin dependence64, marijuana dependence65, alcohol dependence66, and cognitive dysfunction in the elderly67. However, the role of KTN1 in schizophrenia has not been explored yet. In this study, we examined a dense set of SNPs across entire KTN1, in order to identify variants that were associated with schizophrenia consistently across independent samples, which reduced false positives and improved the reliability of findings. And then, we explored the regulatory effects of these schizophrenia-risk variants on KTN1 mRNA expression, GMVs of subcortical structures, and surface area (SA) and thickness (TH) of different cortical regions throughout whole brain.

Materials and methods

Subjects

Three independent samples, including two European and one African-American ones, were included for SNP-schizophrenia association analysis. Sample #1 included 1351 European-American patients with schizophrenia and 1378 healthy European-American controls. Sample #2 included a total of 1826 European parent–offspring trio subjects with 621 offspring with schizophrenia. Sample #3 included 1195 African-American patients with schizophrenia and 954 healthy African-American controls.

All subjects were at least 18 years old. Affected subjects met DSM-IV criteria for schizophrenia68. Patients with neurological disorders, substance use disorders, or mental retardation were excluded. Controls were free of schizophrenia, schizoaffective disorder, bipolar disorder, major depressive disorder, and psychotic symptoms including auditory hallucination and persecutory delusion. All subjects signed written informed consents before participating in the study. All study procedures were reviewed and approved by the Human Investigation Committee of Yale University (HIC#: 1007007175). Detailed demographic data for these samples have been published in previous studies69,70,71,72,73,74. All methods were carried out in accordance with relevant guidelines and regulations.

SNP-disease association analysis

The design of this study is illustrated in Fig. 1, which showed the links between different methods/datasets. The statistical associations between SNPs and diseases were tested first, then the biological functions of the disease-risk SNPs were examined including their regulatory effects on mRNA expression in brains and brain structural measurements, which could validate that the SNP-disease associations were biological but not only statistical ones.

Figure 1
figure 1

Regulatory pathway from SNP to schizophrenia. [Solid lines, pair-wise associations examined in the present study; dash lines, potential pair-wise associations predicted by literatures; arrows, regulatory effect directions; ICV, intracranial volume; GMV, subcortical gray matter volume; SA, cortical surface area; TH, cortical thickness; SZ, schizophrenia].

Full size image

Samples #1 and #3 were genotyped on AFFYMETRIX AFFY_6.0 microarray platform, and Samples #2 was genotyped on Agilent SureSelect Human All Exon v.2 microarray platform. To make the genetic marker sets consistent across different samples, we imputed the untyped SNPs across the entire KTN1 region using the same reference panels of 1000 Genome Project and HapMap3 Project data by the program IMPUTE275. A total of 849 imputed SNPs covering entire open reading frame (ORF) of KTN1, 137 kb regulatory region at 5′ flanking KTN1 that reaches the 3′-transcription termination site (TTS) of nearby FBXO34 gene, and 433 kb regulatory region at 3′ flanking KTN1 that reaches the 5′-transcription start site (TSS) of nearby PELI2 gene, were analyzed.

Before statistical analysis, we stringently cleaned the phenotype and genotype data, as described in detail previously74,76. In brief, subjects with missing diagnosis, missing race, or a missing genotype call rate ≥ 2% across all SNPs were excluded. Furthermore, we excluded SNPs with an overall missing genotype call rate ≥ 2% across all subjects. The SNPs with minor allele frequencies ≤ 0.05 in affected subjects, or in Hardy–Weinberg disequilibrium (p < 0.001) in unaffected subjects were excluded.

The allele frequencies of SNPs were compared between individuals with schizophrenia and controls using Fisher exact test, or between “transmitted” and “untransmitted” using “–dfam” as implemented in PLINK77, to identify the statistical SNP-disease associations. Multiple comparisons in each Sample were corrected by false discovery rate (FDR). Q-values for these analyses that adjusted p-values based on an optimized FDR approach were computed using the R package QVALUE. A q-value < 0.05 indicated statistical significance, and a p-value < 0.05 but q-value > 0.05 indicated nominal significance. A SNP-disease association with p < 0.05 across at least two samples was taken as a potential replicable association. These associations were also verified in other independent cohorts of Psychiatric Genomics Consortium (PGC) data78.

cis-acting expression quantitative trait locus (cis-eQTL) analysis

We examined the potential regulatory effects of schizophrenia-risk variants identified above on the KTN1 mRNA expression in human postmortem brains in a UK European cohort (n = 138) (i.e., BRAINEAC dataset)79 and in a European-American cohort (n = 210) (i.e., GTEx dataset)80 using cis-eQTL analysis. These subjects were free of neurodegenerative and neuropsychiatric disorders. In the UK European cohort, a total of 10 brain regions were analyzed, including cerebellar, prefrontal, occipital, and temporal cortices, hippocampus, medulla, putamen, substantia nigra, thalamus, and intralobular white matter. In the European-American cohort, a total of 11 brain regions were analyzed, including BG (putamen, caudate nucleus, nucleus accumbens, and substantia nigra), limbic system [anterior cingulate gyrus (BA24), amygdala, hippocampus, and hypothalamus], prefrontal cortex (BA9), and cerebellum. Normalized mRNA expression levels were compared between different alleles of each variant using t-test. Multiple comparisons in each brain region were corrected by FDR.

Regulatory effect of risk variants on the total intracranial volume (ICV) and the GMVs of subcortical structures

The ICV in 18,713 European subjects (17 CHAGE + 29 ENIGMA2 cohorts)81 and the GMVs of BG (caudate nucleus, putamen, pallidum, and nucleus accumbens) and limbic system (amygdale, hippocampus, and thalamus) in 38,258 European subjects (14 CHAGE + 35 ENIGMA2 + 1 UKBB cohorts)54,82 were measured by structural magnetic resonance imaging (MRI), following a standardized protocol procedure. GMVs were calculated using the brain segmentation software packages: FIRST83 or FreeSurfer84. All subjects were genotyped using microarray and imputed based on the 1000 Genome Project genotype panels. Genetic homogeneity was assessed in each subject using multi-dimensional scaling (MDS) analysis.

The potential regulatory effects of schizophrenia-risk variants identified above on ICV and GMVs were analyzed using multiple linear regression analysis, controlling for age, sex, 4 MDS components, ICV (for non-ICV phenotypes) and diagnosis (when applicable; most subjects were free of neurodegenerative and neuropsychiatric disorders). Multiple testing in each brain region was corrected by FDR.

Regulatory effect of risk variants on cortical SA and average TH

A total of 36,936 subjects were analyzed, including 33,992 Europeans (23,909 from 49 ENIGMA cohorts and 10,083 from the UK Biobank) and 2,944 non-European participants (eight cohorts)85. Measures of cortical SA and TH of these subjects were derived from in vivo whole brain T1-weighted MRI scans using FreeSurfer84. SA and TH were quantified for each subject across whole cortex and within 34 distinct gyral-defined regions in each brain hemisphere according to the Desikan-Killiany atlas86. SA was measured at the grey-white matter boundary, and TH was measured as the average distance between white matter and pial surfaces. The SA and TH of each cortex were analyzed separately, instead of volume, based on the radial unit hypothesis87 that different developmental mechanisms promoted SA expansion and TH increases. The total SA and average TH of each hemisphere was computed separately.

The potential regulatory effects of schizophrenia-risk variants identified above on a total of 70 traits (total SA, average TH, and SA and TH of 34 cortical regions averaged across right and left hemispheres) were analyzed by exploring the associations between these variants and traits. These associations were analyzed by multiple linear regression analyses, adjusting for the effects of sex, linear and nonlinear age effects, interactions between age and sex, ancestry (the first four MDS components), diagnostic status (when the cohort followed a case–control design), MRI acquisition orientation, and dummy variables for scanner (when multiple scanners were used at the same site). When regional SA and TH were analyzed, the global measure (total SA or average TH) was added as an additional covariate, to test for genetic influences specific to each region. Multiple testing in each brain region was corrected by FDR.

Results

Replicable associations between risk variants and schizophrenia

A total of 26 risk SNPs within the same variant block (r2 > 0.85), covering entire ORF of KTN1 (Supplementary Fig. S1), were nominally associated with schizophrenia across at least two samples (7.5 × 10–5 ≤ p ≤ 0.049; Fig. 2); 25 of them were associated with schizophrenia across two independent European samples (7.5 × 10–5 ≤ p ≤ 0.049), which were significant after FDR correction in at least one sample (3.6 × 10–5 ≤ q ≤ 0.049). Five of these SNPs were also associated with schizophrenia in an African-American sample (0.017 ≤ p ≤ 0.048); and eight of them were also associated with schizophrenia in a mixed European and Asian sample (PGC data: 0.013 ≤ p ≤ 0.030; Table 1). The risk allele of each variant, which had significantly higher frequency in cases (or “transmitted” group) than controls (or “untransmitted” group), was the same one across different European samples, but the opposite one in African-Americans. All risk alleles in Europeans were minor alleles (f < 0.5), but major alleles in Africans (f > 0.5; Table 1).

Figure 2
figure 2

Regional association plots in four samples [X-axis, chromosome positions; Y-axis, −log(p) values; Sample names labeled at the left top corners correspond to Table 1; The most significant SNP with the lowest p-value in each plot is pointed by an arrow; Red diamonds are those SNPs with p < 0.05 and in high LD with one another; Red vertical dashed lines define the replicable association variant block; Red horizontal dashed line correspond to p = 0.05].

Full size image
Table 1 p/q [allelic frequency] values for associations between KTN1 SNPs and schizophrenia in four samples.
Full size table

The schizophrenia-risk variants significantly regulated the KTN1 mRNA expression in brains

All of the 26 risk variants regulated the KTN1 mRNA expression in several brain regions (5.9 × 10–12 ≤ p ≤ 0.050; Table 2), including suggestively in occipital cortex and putamen in BRAINEAC (6.2 × 10–3 ≤ p ≤ 0.047; q > 0.05), and suggestively (p < 0.05 but q > 0.05) or significantly (q < 0.05) in prefrontal cortex (BA9), anterior cingulate gyrus (BA24), cerebellum, amygdala, hippocampus, hypothalamus, and BG (caudate nucleus, nucleus accumbens, putamen and substantia nigra) in GTEx (5.9 × 10–12 ≤ p ≤ 0.050). The regulatory effects were potentially significant in putamen in both BRAINEAC and GTEx (4.5 × 10–3 ≤ p ≤ 0.050), however, in opposite directions; both were not significant after FDR correction though. Interestingly, all of the effective alleles that increased KTN1 mRNA expression (t > 0) in GTEx were schizophrenia-risk alleles that increased risk for schizophrenia, and all effective alleles potentially increasing mRNA expression in BRAINEAC were schizophrenia-protective alleles. The effect direction for each SNP-mRNA association in BRAINEAC is illustrated in Fig. S2, and the effect size, i.e., normalized effect size (NES > 0), for each SNP-mRNA association in GTEx is presented in Fig. S3. Finally, no regulatory effects on KTN1 mRNA expression in other (25%) brain regions were found, including medulla, temporal cortex, thalamus, and intralobular white matter (p > 0.05).

Table 2 p values for associations between schizophrenia-risk SNPs and KTN1 mRNA expression in brain.
Full size table

The schizophrenia-risk variants significantly regulated the ICV and GMVs of subcortical structures

First of all, all of the 26 schizophrenia-risk variants regulated ICV in one sample (n = 18,713 Europeans; 6.1 × 10–4 ≤ p ≤ 0.008; Table 3); 16 of them were significant after FDR correction (q < 0.05). The effect size, i.e., Z-score (> 0), for each SNP-ICV association is presented in Fig. S4. Second, all or most of these variants significantly regulated the four BG GMVs (1.8 × 10–19 ≤ p ≤ 0.050; q < 0.05; Table 3), and many of these regulatory effects in each structure were replicable across two independent samples, including in putamen (1.8 × 10–19 ≤ p ≤ 1.0 × 10–4), caudate nucleus (3.9 × 10–5 ≤ p ≤ 0.033), pallidum (4.2 × 10–7 ≤ p ≤ 0.004), and nucleus accumbens (2.1 × 10–4 ≤ p ≤ 0.050). The regulatory effect on putamen was most significant among all structures. Consistently, all of the schizophrenia-risk alleles that increased risk for schizophrenia increased (z > 0) ICV, but decreased (z < 0) all GMVs of putamen, caudate nucleus, pallidum, and nucleus accumbens (Table 3). The effect size, i.e., Z-score or β (> 0), for each SNP-GMV association is presented in Fig. S5. Finally, no regulatory effects on limbic system (amygdale, hippocampus, and thalamus) were found (p > 0.05).

Table 3 p values for associations between schizophrenia-risk SNPs and ICV and GMVs of basal ganglia.
Full size table

The schizophrenia-risk variants regulated cortical SA

All or most of the schizophrenia-risk variants (excluding the missing rs7150285) potentially regulated the SA of whole brain, lingual gyrus, middle temporal cortex, precuneus, insula, and frontal pole (2.5 × 10–3 ≤ p ≤ 0.048; Table 4), although not significantly after FDR correction (q > 0.05). Overall, most of the schizophrenia-risk alleles that increased risk for schizophrenia potentially increased (β > 0) the SA of whole cortex (9.6 × 10–3 ≤ p ≤ 0.047). Regionally, all or most of these schizophrenia-risk alleles consistently potentially increased (β > 0) the SA of lingual gyrus and middle temporal cortex (2.5 × 10–3 ≤ p ≤ 0.042), but potentially decreased (β < 0) the cortical SA of precuneus, insula and frontal pole (0.010 ≤ p ≤ 0.048). Additionally, rs28393008 potentially decreased (β < 0) the SA of superior temporal cortex (p = 0.037; data not shown). Among these cortices, the precuneus, lingual gyrus, middle and superior temporal cortices are vicinages spatially. The effect size, i.e., β (> 0), for each SNP-SA association is presented in Fig. S6. Finally, no regulatory effects on other 28 regional cortices were found (p > 0.05).

Table 4 p values for associations between schizophrenia-risk SNPs and cortical surface areas and thickness.
Full size table

The schizophrenia-risk variants regulated the average TH of cortices

All or most of the schizophrenia-risk variants (excluding the missing rs7150285) potentially regulated the TH of cortices spatially surrounding the frontal pole (frontal pole, superior frontal, rostral middle frontal, lateral orbito-frontal, and pars opercularis cortices) or the isthmus cingulate gyrus (isthmus cingulate, precuneus, and fusiform gyri, and pericalcarine, transverse temporal, inferior temporal, and paracentral cortices) (0.006 ≤ p ≤ 0.050; Table 4), although not significantly after FDR correction (q > 0.05).

Specifically, these schizophrenia-risk alleles potentially increased (β > 0) the TH of frontal pole cortex (0.006 ≤ p ≤ 0.050; Table 4), which was replicable across two independent samples. They also potentially increased (β > 0) the TH of superior frontal (0.023 ≤ p ≤ 0.050), rostral middle frontal (0.017 ≤ p ≤ 0.046), lateral orbito-frontal (0.032 ≤ p ≤ 0.050), and inferior temporal cortices (0.023 ≤ p ≤ 0.040), pars opercularis (0.024 ≤ p ≤ 0.044), and isthmus cingulate gyrus (0.012 ≤ p ≤ 0.018),. In addition, rs28393008 also potentially increased (β > 0) the TH of fusiform gyrus (p = 0.019; data not shown) that was next to the inferior temporal cortex. The effect directions of these variants were consistent across these regions. On the other hand, these schizophrenia-risk alleles potentially decreased (β < 0) the TH of pericalcarine cortex (0.015 ≤ p ≤ 0.025) that was next to lingual gyrus. In addition, rs28393008 also potentially decreased (β < 0) the TH of transverse temporal cortex, which was replicable across two independent samples (p = 0.023 and 0.049, respectively), precuneus (p = 0.042) and paracentral cortex (p = 0.016) (data not shown). The precuneus, lingual gyrus, and pericalcarine and transverse temporal cortices are all conjunctive with or near the isthmus cingulate gyrus and insula spatially. The effect size, i.e., β (> 0), for each SNP-TH association is presented in Fig. S7. Finally, no regulatory effects on other 22 regional cortices were found (p > 0.05).

Summary

The minor alleles (f < 0.5) of all variants within the same block increased risk for schizophrenia in Europeans, but decreased risk in Africans. All or most of the schizophrenia-risk alleles suggestively or significantly increased (1) the KTN1 mRNA expression in the prefrontal cortex (BA9), anterior cingulate gyrus (BA24), cerebellum, amygdala, hippocampus, hypothalamus, and BG (caudate nucleus, nucleus accumbens, putamen and substantia nigra) in GTEx, (2) the ICV, (3) the SA of whole brain cortex, (4) the SA of lingual gyrus and middle temporal cortex, and (5) the TH of frontal pole, superior frontal, rostral middle frontal, lateral orbito-frontal, and inferior temporal cortices, pars opercularis, and isthmus cingulate gyrus. In addition, rs28393008 within this block also increased the TH of fusiform gyrus. On the other hand, these alleles decreased (1) the KTN1 mRNA expression in the occipital cortex and putamen in BRAINEAC, (2) all GMVs of putamen, caudate nucleus, pallidum, and nucleus accumbens, (3) the SA of precuneus, insula, and frontal pole and superior temporal cortices, and (4) the TH of precuneus, and pericalcarine, transverse temporal and paracentral cortices. After FDR correction, associations between SNPs and risk for schizophrenia in Europeans, KTN1 mRNA expression in brain in GTEx, and ICV and BG GMVs remained significant. Finally, the mRNA expression in 25% of brain regions examined, and the GMVs of limbic system (amygdale, hippocampus, and thalamus) and other 18 regional cortices (~ 50%) were not significantly affected at all.

Discussion

We identified 26 risk KTN1 variants that were located within the same block, spanned entire KTN1, and were significantly associated with schizophrenia across at least two independent samples. These schizophrenia-risk variants had potential or significant regulatory effects on the KTN1 mRNA expression in cortical or subcortical structures, the GMVs of subcortical structures, the SA and/or TH of about 50% cortices. The patterns of allele-schizophrenia associations and regulatory effects were highly consistent, replicable and robust, supporting that the KTN1 variants may be biologically functional and play a critical role in the development of schizophrenia. These findings were majorly identified in Europeans in this study, and genetic difference between Europeans and Africans were also identified.

Globally, the schizophrenia-risk alleles (minor alleles) increased ICV and the SA of whole cortex. Regionally, these schizophrenia-risk KTN1 alleles regulated the mRNA expression, GMVs, SA and/or TH widely across numerous brain regions, including both cortices and subcortical structures. The affected cortices spread over frontal lobe [frontal pole, superior frontal, middle frontal, inferior frontal (pars opercularis), paracentral, and lateral orbitofrontal cortices], temporal lobe (transverse temporal, superior temporal, middle temporal, and inferior temporal cortices), parietal lobe (precuneus), occipital lobe (occipital and pericalcarine cortices, and lingual and fusiform gyri), limbic system (anterior cingulate and isthmus of cingulate gyri), insula and cerebellum. The affected subcortical structures majorly included BG [striatum (putamen, caudate nucleus and nucleus accumbens), pallidum and substantia nigra] and limbic system (pallidum, nucleus accumbens, amygdala, hippocampus, and hypothalamus).

The most significant, consistent and reliable region regulated by the schizophrenia-risk alleles was frontal lobe, majorly prefrontal cortex, centering around the frontal pole. The schizophrenia-risk alleles increased the KTN1 mRNA expression in prefrontal cortex (BA9) and the TH of frontal pole, superior frontal, rostral middle frontal, and lateral orbito-frontal cortices, and pars opercularis. Interestingly, the schizophrenia-risk alleles potentially decreased the SA of frontal pole, which was opposite to the effects on the TH, supporting the radial unit hypothesis that SA and TH have distinct neurodevelopmental origins. The frontal lobe contains most of dopamine neurons in cerebral cortices and the dopaminergic pathways have long been associated with the development of schizophrenia. The frontal cortex is usually involved in attention88, memory89, planning90, motivation91, cognitive process of decision-making92, emotion93, processing of speech and language, and self-awareness94. The prefrontal cortex is the largest component of frontal cortex, and is responsible for internal, purposeful mental action, reasoning or prefrontal synthesis, being involved in the higher-order mental alteration in schizophrenia.

The second most significant, highly consistent and reliable region regulated by the schizophrenia-risk alleles was BG. The schizophrenia-risk alleles suggestively or significantly increased the KTN1 mRNA expression in BG (putamen, caudate nucleus, nucleus accumbens, and/or substantia nigra) in GTEx, but decreased the expression in BG (putamen) suggestively in BRAINEAC, and all four BG GMVs significantly. Among these four structures, the striatum (caudate nucleus, putamen and nucleus accumbens) was most affected, and the regulatory effects on the GMV of putamen were much more significant and robust than any other, consistent with the report that KTN1 was the most significant gene regulating the GMV of putamen.

BG, critical components of the “cortico-BG-thalamo-cortical” loop, play an important role in cognition95 and emotion. Striatum receives excitatory glutamatergic input from cortices, and then sends output to other components of BG including pallidum and substantia nigra pars reticulate (SNr). The external globus pallidus (GPe) receives input from striatum, and sends inhibitory GABAergic output to subthalamic nucleus (STN); the internal globus pallidus (GPi) receives input from striatum, and sends inhibitory GABAergic output to thalamus. STN receives input from GPe and projects excitatory glutamatergic output to GPi or SNr. SNr sends inhibitory GABAergic output to thalamus. Finally, thalamus relays these outputs to cortices, completing the projection loops. The loops with projections to BG from prefrontal, lateral orbitofrontal, and temporal cortices, and anterior cingulate gyrus are called cognitive/associative pathways, playing a role in cognitive processing and frontal lobe functioning. The dysfunction of BG in these loops may result in uncontrolled behaviors as well as cognitive deficits of schizophrenia, similar to those that result from damage to prefrontal cortex96. In addition, the loops with projections to the limbic part of BG (nucleus accumbens and ventral pallidum) from anterior cingulate gyrus, hippocampus, and insula are called limbic circuit97, playing a central role in reward learning as well as cognition and frontal lobe functioning too, via the mesolimbic pathway that uses the neurotransmitter dopamine, and the mesocortical pathway. Overactive dopaminergic projection of mesolimbic pathway has been implicated in schizophrenia98. Our findings that schizophrenia-risk alleles regulated KTN1 mRNA expression, and SA and TH of cortices in limbic system (amygdala, hippocampus, hypothalamus, and cingulate gyrus) and insula supported this implication.

Other three lobes, including the temporal, parietal, and occipital lobes, were also regulated by the schizophrenia-risk alleles. First, we found that the schizophrenia-risk alleles increased the SA of middle temporal cortex and the TH of inferior temporal cortex, but decreased the SA of superior temporal cortex and the TH of transverse temporal cortex. The alteration of these temporal regions may cause psychotic symptoms of schizophrenia in language processing, social cognition and auditory verbal hallucinations99,100. Second, we found that the schizophrenia-risk alleles decreased the SA and TH of precuneus. This parietal region has been involved in mental imagery concerning the self101, episodic memory102, and visuospatial imagery103. Finally, we found that the schizophrenia-risk alleles increased the SA of lingual gyrus and the TH of fusiform gyrus, but decreased the KTN1 mRNA expression in occipital cortex (in BRAINEAC), and the TH of pericalcarine cortex. These occipital cortices are where the primary visual cortex is normally located, and responsible for visual and sensory information processing104. The occipital lesions can cause visual hallucinations of schizophrenia.

We found that the schizophrenia-risk alleles decreased the volumes of some cortices, including frontal [frontal pole (SA) and paracentral], temporal (superior and transverse temporal), parietal (precuneus), occipital (pericalcarine), and insular cortices, which was consistent with the previous findings that the volumes of these cortices in schizophrenia were reduced, suggesting that KTN1 variants might play a dominant role in regulating the volumes in these areas. Conversely, we found that these schizophrenia-risk alleles increased the volumes of other cortices, including ICV, whole cortex, frontal [frontal pole (TH), superior frontal, rostral middle frontal, lateral orbito-frontal, and pars opercularis], temporal (middle and inferior temporal), and occipital (lingual and fusiform gyri) cortices, and limbic system (isthmus cingulate gyrus), which contradicted the previous findings that the volumes of these cortices were also reduced in schizophrenia. A hypothesis of compensation may be able to interpret these contradictory findings: First, other schizophrenia-risk non-KTN1 genes might be dominant in these areas, significantly decreasing the volumes of cortices and thus the neurotransmission in the “cortico-BG-thalamo-cortical” loop. Next, as a compensatory response to reduction of excitatory glutamatergic output from cortices to BG, the expression of cortical volume-expanding proteins may be activated, in order to maintain neural transmission within this loop. One of these proteins is kinectin 1, as would be evidenced by our finding that the schizophrenia-risk alleles increased the KTN1 mRNA expression in most cortices examined, which may then increase the kinectin expression in neurons, expand the cell sizes, and increase the volumes of the cortices. Finally, this expression activation of volume-expanding KTN1 in cortices did not appear to restore the reduced volumes of cortices and neurotransmission of the loop resulting from other dominant schizophrenia-risk genes, suggesting that KTN1 played a “recessive” role in these areas. This hypothesis well interpreted how the associations between schizophrenia-risk KTN1 alleles and increased cortical volumes co-existed with the associations between schizophrenia and reduced cortical volumes in some brain areas.

Unexpectedly, the schizophrenia-risk alleles decreased the BG GMVs, contradictory to the previous findings of BG GMV enlargement in schizophrenia. This was highly possibly because other, but not the current, set of variants in the regulatory region 3′ flanking to KTN1, including rs94527054, dominantly increased the BG GMVs (Luo et al., unpublished data). As a compensatory response to the expansion of BG GMVs, the expression of some volume-shrinking KTN1 alleles might be activated to shrink the BG GMVs, although this compensatory shrinkage did not restore the expansion of BG GMVs, suggesting a “recessive” role of these volume-shrinking KTN1 alleles in BG GMVs. The current set of schizophrenia-risk alleles, which increased the risk for schizophrenia, potentially decreased the KTN1 mRNA expression in BG (e.g., in BRAINEAC), was just a set of volume-shrinking alleles, which well interpreted how a “recessive” association between schizophrenia-risk KTN1 alleles and reduced BG GMVs co-existed with a “dominant” association between schizophrenia and expanded BG GMVs.

One limitation of our study is that the SNP-schizophrenia, SNP-mRNA and SNP-ICV/GMV/SA/TH associations were analyzed in separate cohorts, confounded potentially by cohort effect. We noticed that the potential regulatory effects of schizophrenia-risk alleles on KTN1 mRNA expression in putamen were opposite between BRAINEAC and GTEx, which may be attributed to cohort heterogeneity. Studying these associations in the same cohort in the future would address this limitation.

In summary, this is a study on the role of KTN1 in schizophrenia. We identified a significant, functional, and robust risk variant block at KTN1 for schizophrenia. It may regulate the risk for schizophrenia, the KTN1 mRNA expression in the brain (75%), the GMVs of subcortical structures, and the volumes of around 50% of cortices throughout the brain. It played a “dominant’ role in some specific cortices, but possible “recessive” roles in other cortices and BG in schizophrenia.

Ethics approval and consent to participate

All subjects signed written informed consents before participating in the study. All study procedures were reviewed and approved by the Human Investigation Committee of Yale University.