The Emerging Immunogenetic Architecture of Schizophrenia

Abstract

Schizophrenia is a severe psychiatric disorder of complex etiology. Immune processes have long been proposed to contribute to the development of schizophrenia, and accumulating evidence supports immune involvement in at least a subset of cases. In recent years, large-scale genetic studies have provided new insights into the role of the immune system in this disease. Here, we provide an overview of the immunogenetic architecture of schizophrenia based on findings from genome-wide association studies (GWAS). First, we review individual immune loci identified in secondary analyses of GWAS, which implicate over 30 genes expressed in both immune and brain cells. The function of the proteins encoded by these immune candidates highlight the role of the complement system, along with regulation of apoptosis in both immune and neuronal cells. Next, we review hypothesis-free pathway analyses which have so far been inconclusive with respect to identifying immune pathways involved in schizophrenia. Finally, we explore the genetic overlap between schizophrenia and immune-mediated diseases. Although there have been some inconsistencies across studies, genome-wide pleiotropy has been reported between schizophrenia and Crohn’s disease, multiple sclerosis, rheumatoid arthritis, systemic lupus erythematosus, type 1 diabetes, and ulcerative colitis. Overall, there are multiple lines of evidence supporting the role of immune genes in schizophrenia. Current evidence suggests that specific immune pathways are involved—likely those with dual functions in the central nervous system. Future studies focused on further elucidating the relevant pathways hold the potential to identify novel biomarkers and therapeutic targets for schizophrenia.

Immune Hypothesis of Schizophrenia

Immune processes have long been considered a potential initiating insult in schizophrenia.¹ Accumulating evidence from epidemiological² and clinical^3–6 studies suggests that immune disturbances are present in at least a subset of patients with schizophrenia, garnering further support for this immune hypothesis.⁷ Advances in neuroimmunology have provided new insights into the mechanisms by which the immune system influences brain development and function,^8–11 generating diverse hypotheses about the potential role of the immune system in schizophrenia. Neuronal autoantibodies,¹² priming of microglia,¹³ and altered T-cell distribution¹⁴ have all been proposed as immune mechanisms predisposing to the disease.

One approach to refine these hypotheses is via human genetic association studies; such studies identify genetic variants associated with schizophrenia case status, thereby providing a window into the biological processes that cause the disease. Here, we review the immunogenetic architecture of schizophrenia. The focus is on evidence from genome-wide association studies (GWAS), given that GWAS (1) are currently the largest studies available in schizophrenia and therefore have the greatest statistical power, and (2) allow direct comparison to other diseases with available GWAS data. Importantly, GWAS provide an incomplete picture of immunogenetic variation—common single nucleotide polymorphisms (SNPs) are well captured, but rare and complex structural variants are not. When GWAS findings are driven by rare or complex structural variants, sequencing studies can help identify the causal variant(s) and shed new light on disease biology.¹⁵ The utility of sequencing studies is particularly relevant for immune genes, which are enriched for complex genetic variation due to strong pathogen-driven selective pressure.¹⁶ For instance, targeted sequencing in a region of association identified by schizophrenia GWAS identified complex structural variation at complement component 4 (C4) as a risk factor for the disease.17 Subsequent functional studies revealed increased C4 expression was associated with schizophrenia, and implicated synaptic pruning as a potential biological mechanism by which C4 variants contribute to the disease.¹⁷ Thus, sequencing studies hold the potential to move beyond regions of association identified in schizophrenia GWAS toward the identification of causal immune variants and their functional roles. Moving forward, the integration of sequencing data will be an important step to further understand the immunogenetic architecture of schizophrenia and other human diseases.

Association of Immune Genes in Schizophrenia

Major Histocompatibility Complex Genes

The extended major histocompatibility complex (xMHC) region is a 7.7 Mb stretch of DNA on the short arm of chromosome 6 (25.7–33.4 Mb) containing over 400 genes.¹⁸ There is extensive linkage disequilibrium (LD) across the xMHC, posing a challenge for studies of genetic variants in this region.¹⁹ Important immune genes in the xMHC region include the highly polymorphic human leukocyte antigen (HLA) genes, as well as genes encoding tumor necrosis factor family (TNF, LTA, LTB) and complement components (BF, C2, C4A, C4B). There are also many nonimmune genes in the xMHC region, including candidates for central nervous system disorders such SYNGAP1, MOG, NOTCH4, histones, and tRNAs.

In light of a suspected immune-related etiology, more than 60 hypothesis-driven candidate gene studies of functional HLA variants have been conducted over the years.²⁰ However, the sample sizes of these candidate gene studies were modest (less than 300 cases), and the implicated HLA variants were inconsistent. Overall, there was some evidence that DQB1*0602 and DRB1*04 alleles may protect against schizophrenia in a systematic analysis by Wright et al.²⁰ In European populations, in more recent hypothesis-free GWAS of schizophrenia, SNP alleles in the xMHC region have consistently shown the strongest association (reviewed by Kodavali et al²¹).

Fine-mapping of the xMHC revealed 3 independent associations across the region^17,22 in a large Psychiatric Genomics Consortium (PGC) schizophrenia GWAS²⁸ of over 35000 cases and 46000 controls (referred to here as the ‘PGC2’ study).^17,22 Interestingly, HLA variants do not appear to drive the xMHC²⁸ region association in schizophrenia (in contrast to many immune-mediated diseases).^17,22 The strongest association with schizophrenia corresponds to an extended class I region association spanning ~2 Mb which remains to be further localized.^17,22 The second region of association in the xMHC corresponds to C4, and has been well elucidated in genetic and functional work.¹⁷ Genetic variation at the C4 locus containing C4A and C4B is complex, with both structural and copy number variation influencing C4 expression. Genetic variants conferring increased C4 expression were associated with schizophrenia in the PGC GWAS.¹⁷ Furthermore, C4 knockout mice demonstrated significantly reduced synaptic pruning,¹⁷ suggesting that increased expression of C4 in schizophrenia may contribute to disease via excessive synaptic pruning. The third region of association in the xMHC is just downstream of the extended class II region.^17,22 The strongest associated SNP in this region (rs210133) was identified near BAK1 by Sekar et al¹⁷ when accounting for the extended class I region association and genotype-predicted C4 expression. In conditional analysis using GWAS data alone without accounting for C4 expression, the strongest associated SNP (rs9461856) was approximately 140 kb upstream within SYNGAP1.²² While these SNPs are not in significant LD (r² = .15 based on 1000 Genomes Phase 3 CEU European reference panel),²³ it is notable that conditional analyses using 2 different approaches provided converging evidence for an independent third region of association in the extended class II region. Interestingly BAK1 encodes a pro-apoptotic mitochondrial protein required to prevent autoimmunity,²⁴ while SYNGAP1 encodes a postsynaptic density protein previously implicated in schizophrenia²⁵ and other neurodevelopmental disorders.^26,27

Overall, robust evidence has emerged for the immune gene C4 as a susceptibility gene in schizophrenia.¹⁷ Follow-up studies combining genetic and functional approaches, similar to those undertaken by Sekar et al,¹⁷ are needed to delineate the underlying causal variants for the extended class I and extended class II major histocompatibility complex (MHC) associations in schizophrenia.

Non-MHC genes

Outside of the xMHC region, over 100 loci reached genome-wide significance (P < 5 × 10⁻⁸) in the PGC2 GWAS.²⁸ Follow-up investigations to identify the causal variants and/or genes in these regions are currently underway. Among the genome-wide significant variants in the PGC2 GWAS, our group highlighted 6 SNPs near immune genes CLU, DPP4, EGR1, ESAM, HSPD1, and NFATC3.²² All 6 of these immune candidates are expressed in human brain tissue,²⁹ and have established roles in the brain^30–34 in addition to their roles in immunity. An integrative analysis of the PGC and other schizophrenia GWAS incorporating eQTL and pathway analysis identified variants annotated to 28 non-MHC genes highly expressed in B lymphocytes and T lymphocytes (Box 1).³⁵ Another study identified 18 genes located in genome-wide significant regions of the SCZ-52 GWAS which encoded proteins targeted by approved drugs, and an additional 21 genes whose products were targeted by drugs currently in clinical trials.³⁶ Among these genes, 11 have known immune functions and are expressed by immune cells in addition to brain: AKT3, CD14, CLU, DPP4, EGR1, FURIN, HSPD1, LRP1, MAPK3, SERPING1, and TNFRSF13C (Human Protein Atlas available from www.proteinatlas.org, accessed March 16, 2018).³⁷ Overall, these 3 studies^22,35,36 identified 39 candidate immune genes for schizophrenia outside of the MHC region (table 1). Notably, HSPD1 was identified as a top schizophrenia gene in all 3 independent secondary analyses of schizophrenia GWAS data^22,35,36 while CD14,^35,36DPP4,^22,36CLU,^22,36 and EGR1^22,36 were identified in 2 independent secondary analyses (table 1).

Table 1.

Schizophrenia Immune Gene Candidates

Box 1. Schizophrenia susceptibility genes expressed in B and T lymphocytes

ARHGAP4, ARNTL, BTG1, CD14, CD46, DOCK4, EP300, FES, FLII, FMR1, GCA, GNL3, HBEGF, HSPD1, IK, IRAK1, MCL1, MSH6, NUCB2, PIK3C2A, PLA2G15, PLCB2, PRKCD, PRMT1, PTGS2, STAT6, TPR, and ZEB2.

There is accumulating evidence of shared genetic risk across schizophrenia and other neuropsychiatric phenotypes.^51–55 Given this genetic sharing, it is notable that 12 of the 39 non-MHC immune candidate gene regions identified in previous studies^22,35,36 have been associated with other neuropsychiatric traits in the NHGRI-EBI GWAS Catalogue³⁸ including bipolar disorder, anxiety, neuroticism, cognitive abilities, and amyotrophic lateral sclerosis (table 1). In addition to these brain-related phenotypes, schizophrenia immune candidate gene regions have been implicated in a number of immune and inflammatory phenotypes (supplementary table 1). While these findings raise the possibility that certain immune genes implicated in schizophrenia also contribute to other brain- and immune-related phenotypes, there are some important caveats. Firstly, not all of the immune candidate genes reviewed here achieved genome-wide significance in the PGC2 GWAS (supplementary table 1). Thus, there may be some false positive findings, and further validation of these regions of association in larger samples is needed. Secondly, the immune genes reviewed here are located in broad regions of association containing additional genes (supplementary table 1). Until fine-mapping and functional investigations are completed to identify the causal variant(s) in each region, it remains unclear whether the association signals are driven by variants in the candidate immune genes or other variants in the region. Finally, although the same genomic region containing an immune gene of interest is associated with schizophrenia and other traits (table 1), the underlying causal genes or polymorphisms may be different (supplementary table 1). For instance, a susceptibility locus on chromosome 11 containing SERPING1—an inhibitor of the complement system—has been implicated in both neuroticism (rs10896636; P_discovery = 4.52 × 10⁻¹⁰, P_replication = .02)⁵⁰ and schizophrenia (rs9420; P = 2.24 × 10⁻⁹).²⁸ Although these SNPs are in high LD (r² = .90 based on 1000 Genomes Phase 3 CEU European reference panel)²³ the nearest gene to rs10896646 is not SERPING1, but instead ZDHHC5—a gene that has been shown to regulate hippocampal-dependent learning in ZDHHC5-deficient mice.⁵⁶ Thus, before any conclusions about shared immune pathways across neuropsychiatric phenotypes can be made, future investigations must confirm a causal role for immune candidate genes in each phenotype of interest. Nevertheless, available data suggest that some immune genes may play a role in predisposing to schizophrenia specifically (eg, HSPD1, NFATC3, TNFRSF13C), while others may contribute to a number of different neuropsychiatric and immune-related phenotypes (eg, AKT3, DPP4, CLU).

Keeping in mind the important limitations described above, we performed exploratory protein–protein interaction (PPI) analysis of previously identified schizophrenia candidate immune genes (table 1) using STRING, an online tool providing functional association of proteins based on evidence from a variety of data sources.⁵⁷ We included functional interactions with total scores above 0.400 (medium confidence) based on evidence from 3 different scores: textmining, experiments, and co-expression. Our PPI analysis highlighted the roles of both the adaptive and innate immune response (see figure 1 for details of the PPI analysis, and supplementary table 2 for complete list of enriched biological pathways and gene members). Interestingly, regulation of cell death—including neuronal death—was a recurring biological process in the PPI network.

Fig. 1.

Functional protein association network for immune candidates implicated in schizophrenia. The 39 immune gene candidates outside of the xMHC region identified in this review were included in a PPI analysis using STRING.⁵⁷ We included functional interactions with total scores above 0.400 (medium confidence) based on evidence from 3 different scores: textmining, experiments, and co-expression. Line thickness connecting the nodes indicates the strength of data support for a functional interaction between those proteins. Large nodes have known 3D structure, small nodes do not.

Current findings suggest that immune genes implicated in schizophrenia may be involved in both brain- and immune-related phenotypes, and may have dual roles in immunity and central nervous system (CNS) function (eg, regulation of cell death). The co-expression of these candidate immune genes in CNS is perhaps unsurprising given the vast transcriptome of immune cell types; an estimated 76% of all human proteins are expressed by at least one type of immune tissue (Human Protein Atlas available from www.proteinatlas.org, accessed March 16, 2018).³⁷ Thus, further studies are needed to clarify whether specific immune pathways or types of immune cells may play a role in schizophrenia. For example, in autism large-scale RNA sequencing identified a gene co-expression network implicating dysregulated microglial responses in the disease.⁵⁸ Given that our understanding of the genetic architecture of schizophrenia remains incomplete, it will be important to revisit these analyses in the future as new discoveries are made.

Immune Pathways in Schizophrenia

Schizophrenia is a polygenic disorder involving thousands of genetic risk variants.⁵⁹ To identify the biological processes implicated by these risk variants, genetic pathway analyses evaluate whether groups of functionally related genes (eg, calcium channels) are jointly associated with the disorder. Several pathway analysis approaches have been developed and applied to schizophrenia, each with their own set of advantages and limitations.¹⁹ One major disadvantage of these methods is that they rely on existing research to assign biological functions to a gene set or variant set to define a pathway, typically based on known gene functions. To overcome this barrier, the recently developed Data-driven Expression-Prioritized Integration for Complex Traits (DEPICT) method integrates gene co-expression data with existing knowledge of the biological functions of genes.⁶⁰ Unlike traditional pathway analyses where understudied genes with unknown functions are left out of the study, these genes are assigned a function in DEPICT based on similar expression to other genes with known function.⁶⁰

Analyses of early schizophrenia GWAS suggested a role for immune pathways. For instance, in gene-set enrichment analysis of the Genetic Association Information Network (GAIN) schizophrenia GWAS (N = 1158 cases and 1378 controls), 3 of the 7 overrepresented pathways were related to the immune system (TGF-β, TNFR1, and TOB1 pathways).⁶¹ A meta-analysis of 18 schizophrenia GWAS including the ISC, SGENE, MGS, CATIE, and Swedish samples (N = 11185 cases and 10768 controls) also reported enrichment of immune pathways (antigen processing, T-cell adhesion molecules, and translocation to immunological synapse pathways).⁶² However, both of these pathway analyses were undertaken without special consideration of the xMHC region.^61,62 Their findings must therefore be interpreted in light of the fact that extensive LD in the xMHC region can bias standard enrichment approaches and lead to false-positive results,¹⁹ particularly given the strong association of SNPs in this region with schizophrenia.

The Network and Pathway Analysis Subgroup of the PGC performed a rigorous analysis of 19752 pathways obtained from 6 sources (KEGG, GO, PANTHER, TARGETSCAN, Reactome, OMIM), testing for enrichment of these pathways in 5 psychiatric disorders—including schizophrenia, major depressive disorder (MDD), bipolar disorder (BP), attention deficit hyperactivity disorder (ADHD), and autism spectrum disorder (ASD)—by combining the results of 5 different pathway analysis methods (FORGE, MAGENTA, ALIGATOR, INRICH, and set screen testing). No pathway enrichment was detected in the schizophrenia sample (N = 9379 cases and 7736 controls), whether including or excluding the xMHC region.⁶³ After pooling data across adult psychiatric disorders (schizophrenia, MDD, BP), TGF-β signaling and B-cell activation pathways were significantly enriched.⁶³

In the primary analysis of the SCZ-52 GWAS by the PGC (N = 35476 cases and 46839 controls), all pathway analyses excluded the xMHC region and no immune pathway enrichment was observed.²⁸ These analyses were not as comprehensive as that of the Network and Pathway Analysis Subgroup; a total of 9016 pathways obtained from 7 sources (KEGG, GO, PANTHER, Reactome, BioCarta, MGI, NCI) were tested for enrichment using 2 different methods (ALIGATOR and INRICH).²⁸ Genome-wide significant schizophrenia associations were enriched at enhancers (defined by histone H3K27ac signal) active in CNS and, interestingly, B-lymphocytes.²⁸ However, follow-up analyses by Finucane et al⁶⁴ using Stratified LD Score Regression (sLDSC) to analyze all SNPs in the SCZ-52 GWAS, rather than restricting to the set of genome-wide significant SNPs, only found enrichment of CNS enhancers (defined by H3K27ac, H3K4me1, H3K4me3, or H3K9ac signal) with no enrichment of immune cell enhancers. In contrast, autoimmune diseases such as Crohn’s disease and rheumatoid arthritis showed enrichment of immune cell enhancers.⁶⁴ Using sLDSC, our group tested for enrichment of SNPs capturing variation in genes with known immune and CNS function; we observed significant enrichment of the CNS genes in schizophrenia, but no significant enrichment of immune genes.²² In contrast, immune genes were enriched among 5 autoimmune diseases.²² Pers et al⁶⁰ applied DEPICT to 109 genome-wide significant loci outside of the xMHC region from the same PGC data set and found enrichment among genes expressed in the brain as well as 15 canonical pathways, but no enrichment of immune-related gene sets.

With these contrasting observations, it remains unclear whether genetic data support a role for immune cell subsets or immune signaling pathways in the pathogenesis schizophrenia. Differences in gene sets and enrichment methods across studies as well as insufficient statistical power may contribute to the discrepant findings to date, particularly if there is true enrichment of immune-related pathways within specific cell subsets (eg, B-lymphocytes) or signaling pathways (eg, complement cascade). The availability of increasingly massive genetic data sets for schizophrenia presents an important opportunity for follow-up studies to investigate specific immune cell subsets and functional pathways, in order to further clarify their potential role in schizophrenia.

Genetic Correlation Between Schizophrenia and Immune Diseases

Beyond pathway analysis, GWAS of schizophrenia allow comparison to other human diseases with available GWAS data (cross-trait analysis). Methods to detect genome-wide sharing of genetic risk across disorders (genetic correlation) include polygenic risk scoring (PRS),⁶⁵ restricted maximum likelihood (REML),⁶⁶ and cross-trait LDSC.⁶⁷ Similar to pathway analysis methods, each has their own advantages and limitations.⁶⁸ Already cross-trait analyses have illustrated strong genetic correlation between schizophrenia and other adult onset psychiatric disorders such as bipolar disorder and major depressive disorder,^51,52,67 suggesting that these diseases are not as distinct at a pathophysiological level as current diagnostic criteria suggest.

Of interest with respect to the immunogenetic architecture of schizophrenia are cross-trait analyses investigating genetic overlap between schizophrenia and immune-mediated disorders. At least 10 published studies have reported such cross-trait analyses (table 2), although in some earlier studies immune diseases were analyzed as a negative control.^51,70 Despite available GWAS data for more than 19 immune diseases in ImmunoBase (www.immunobase.org, accessed March 16, 2018), published studies have only investigated pleiotropy between schizophrenia and nine immune diseases (celiac disease, Crohn’s disease, multiple sclerosis, psoriasis, rheumatoid arthritis, systemic lupus erythematosus, systemic sclerosis, type 1 diabetes, and ulcerative colitis).

Table 2.

Previous Investigations of Genetic Overlap Between Schizophrenia and Immune-Mediated Disease

Immune Disease	Study	GWAS Samples	N Cases (% Male)a	Method	Resultb
CEL	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	n.s.
		CEL69	4533
CRO	Purcell et al70	Schizophrenia ISC70training	3322 (66%)	PRS	n.s.
		CRO71target	1748 (39%)
	Cross-Disorder Group of the PGC51	Schizophrenia PGC1–shared samples51	9074 (66%)	REML	n.s.
		CRO–shared samples72	4793
	Stringer et al73	Schizophrenia PGC1–caws28training	8922 (66%)	PRS	+*
		CRO71target	1664 (39%)
	Bulik-Sullivan et al67	Schizophrenia PGC2–Asian28	33332 (61%)	LDSC	n.s.
		CRO74	5956 (45%)
	Wang et al75	Schizophrenia PGC Cross-Disorder51	9379 (66%)	GPA	+*
		CRO72	6333
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	+
		CRO72	6333
MS	Andreassen et al76	MS77training	9772 (30%)	CFDR	+*
		Schizophrenia PGC178target	9394 (66%)
	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	n.s.
		MS79	931 trios (25%)
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	n.s.
		MS77	9772 (30%)
PBC	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	+
		PBC80	2764
PSO	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	n.s.
		PSO81	1359 (48%)
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	+
		PSO82	2178
RA	Purcell et al70	Schizophrenia ISC70training	3322 (66%)	PRS	n.s.
		RA71target	1860 (25%)
	Stringer et al73	Schizophrenia PGC1–caws28training	8922 (66%)	PRS	+*
		RA71target	1835 (25%)
	Lee et al83	Schizophrenia PGC1+Swe59,78	12793 (64%)	REML	−*
		RA84,85	8064 (32%)
	Euesden et al86	Schizophrenia PGC1.559	5001 (60%)	PRS	n.s.
		RA71	1989 (25%)
	Bulik-Sullivan et al67	Schizophrenia PGC2–Asian28	33332 (61%)	LDSC	n.s.
		RA84	5539 (37%)
	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	+
		RA84	5539 (37%)
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	n.s.
		RA85	14361 (23%)
SLE	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	+
		SLE87	720 (0%)
	Daniel Tylee (personal communication)	Schizophrenia PGC228	35476 (60%)	LDSC	+
		SLE88	4036
SSC	Daniel Tylee (personal communication)	Schizophrenia PGC228	35476 (60%)	LDSC	n.s.
		SSC89	2296 (11%)
T1D	Purcell et al70	Schizophrenia ISC70training	3322 (66%)	PRS	n.s.
		T1D71 target	1963 (51%)
	Stringer et al73	Schizophrenia PGC1–caws28training	8922 (66%)	PRS	+*
		T1D71target	1894 (51%)
	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	n.s.
		T1D90	7514
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	n.s.
		T1D91	9934
UC	Bulik-Sullivan et al67	Schizophrenia PGC2–Asian28	33332 (61%)	LDSC	n.s.
		UC74	6968 (52%)
	Wang et al75	Schizophrenia PGC Cross-Disorder52	9379 (66%)	GPA	+*
		UC92	6687
	Daniel Tylee (personal communication)95	Schizophrenia PGC228	35476 (60%)	LDSC	+
		UC92	6687

Note: n.s., not significant; +, positive genetic correlation between schizophrenia and the immune disease of interest; −, negative genetic correlation between schizophrenia and the immune disease of interest; *, finding survives multiple testing correction; CEL, celiac disease; CFDR, conditional false discovery rate⁹³; CRO, Crohn’s disease; GPA, Genetic analysis incorporating pleiotropy and annotation⁹⁴; ISC, International Schizophrenia Consortium; LDSC, cross-trait LD Score regression⁶⁷; MS, multiple sclerosis; PBC, primary biliary cirrhosis; PRS, polygenic risk scoring⁶⁵; PSO, psoriasis; RA, rheumatoid arthritis; REML, restricted maximum likelihood⁶⁶; SLE, systemic lupus erythematosus; SSC, systemic sclerosis; T1D, type 1 diabetes; UC, ulcerative colitis.

^aPercent of male cases in the GWAS is provided for all studies where it was reported.

^bOnly those results that were robust to exclusion of the major histocompatibility complex (MHC) region are presented as significant in this table.

Studies to date have reported evidence of pleiotropy between schizophrenia and Crohn’s disease,⁷³,^75,95 multiple sclerosis,⁷⁶ primary biliary cirrhosis,⁹⁵ psoriasis,⁹⁵ rheumatoid arthritis,^73,75 systemic lupus erythematosus,^75,95 type 1 diabetes⁷³, and ulcerative colitis.^75,95 For rheumatoid arthritis negative genetic correlation (whereby genetic risk for rheumatoid arthritis confers protection against developing schizophrenia) has also been reported,⁸³ a result that is more in keeping with the epidemiological observation that schizophrenia patients have lower rates of rheumatoid arthritis than the general population.⁸⁶ To further complicate the picture, some studies have reported no evidence of shared genetic risk between schizophrenia and these immune diseases (table 2).

There are a number of potential reasons for these apparently conflicting results. False negative findings may be driven by differences in GWAS sample sizes and cross-trait analysis methodologies (table 2), which influence statistical power to detect true genetic correlation. Alternatively, false positive findings may be driven by confounders that are inadequately adjusted for such as population stratification, LD, or nonindependence of samples in the schizophrenia and immune GWAS. With respect to LD, the xMHC is the strongest region of association in schizophrenia GWAS as well as many of the immune diseases. Given the extensive LD in this region, its inclusion in cross-trait analyses can lead to spurious findings. Regarding nonindependence of samples, Wellcome Trust Case Control Consortium (WTCCC) samples were used in the control group for the majority of the immune disease and schizophrenia GWAS, creating the potential for false positives if not accounted for. Another consideration is that, due to the significant sex bias of autoimmune diseases (which have a greater overall prevalence in women⁹⁶), there may be sex-specific genetic correlations which have not been formally evaluated in existing studies; given the different sex proportions across GWAS samples analyzed to date (% male cases ranging from 0% to 66%; table 2), sex-specific effects could have either diluted or increased true signal leading to different results across studies.

Additional studies are needed to reconcile the inconsistencies in existing cross-trait analyses, with a focus on including all immune diseases with available GWAS data, comparing results across methodological approaches, applying rigorous control for population stratification and overlapping samples, and evaluating sex-specific effects. To this end, our group has undertaken a more comprehensive cross-trait analysis of all immune diseases curated in ImmunoBase, using the largest available GWAS for both schizophrenia and immune diseases, and comparing results from PRS and cross-trait LDSC approaches.⁹⁷

Overall, there is accumulating evidence of shared genetic risk for schizophrenia and at least a subset of immune-mediated diseases. While it remains unknown what is driving this genetic overlap, it may represent a pathological process shared by immune cells and neurons (eg, a disturbance in calcium channel function or synaptic signaling). Future studies focused on identifying shared biological pathways between schizophrenia and genetically correlated immune diseases are needed to evaluate this hypothesis. Interestingly, genetic overlap with immune diseases has also been reported in Parkinson’s disease⁹⁸ and Alzheimer’s disease⁹⁹ using the conditional false discovery rate approach.⁹³ If these findings are replicated, they raise the possibility that a genetic relationship with immune diseases may not be a phenomenon specific to schizophrenia but may instead represent an important etiological pathway for neuropsychiatric disorders in general.

Conclusion

The genomics era provides a new opportunity for evaluation of the immune hypothesis of schizophrenia using large-scale genetic data. Already GWAS have identified several immune candidates in association with schizophrenia. Within the xMHC region, strongest evidence has emerged for C4. Outside of the xMHC, more than 30 immune candidates have been identified including CD14, CLU, DPP4, EGR1, and HSPD1. Proposed immune candidates are expressed in both immune cells and the brain, suggesting dual roles in immunity and CNS function. The best current evidence for this is C4, which encodes a protein responsible for pathogen clearance in the immune system and synaptic pruning in the CNS. Additional fine-mapping and functional studies are needed to validate and understand the role of other immune candidate genes in schizophrenia.

Studies investigating immune pathway enrichment in schizophrenia GWAS have been less conclusive. While some studies have implicated B lymphocytes, others have not identified any immune pathway involvement. Results from cross-trait analyses of schizophrenia and immune-mediated diseases have also been somewhat inconsistent. Genetic correlation with a number of immune diseases has been reported, including Crohn’s disease, multiple sclerosis, rheumatoid arthritis, systemic lupus erythematosus, and type 1 diabetes. However, negative genetic correlation and non-significant findings have also been reported.

Overall, there is strong evidence supporting the involvement of certain immune variants in schizophrenia. However, there is conflicting evidence for enrichment of specific immune cell types or pathways, and for pleiotropy with specific immune diseases. Differences in sample size and methodological approach, clinical heterogeneity, and the strong xMHC association in schizophrenia are all potential factors contributing to these discrepancies across studies. Nevertheless, current data suggest that certain biological pathways with dual roles in immunity and CNS—such as cell death regulation or calcium channel signaling—are involved in both immune diseases and schizophrenia. To further support this hypothesis and identify the specific immune genes/cells/pathways involved in schizophrenia, there is a need for future studies to integrate methods, analyze more clinically homogenous patient samples or use less “noisy” phenotypes (eg, cognitive symptoms, negative symptoms), perform analyses both including and excluding the xMHC region, and follow-up on robust findings with functional studies.

Funding

This work was supported by the Canadian Institutes of Health Research (CIHR) MD/PhD Studentship, Fulbright Canada, the Weston Foundation, and Brain Canada through the Canada Brain Research Fund - a public-private partnership established by the Government of Canada.

Supplementary Material

Supplementary data are available at Schizophrenia Bulletin online.