The PIP5K2A and RGS4 genes are differentially associated with deficit and non-deficit schizophrenia

Abstract

Several putative schizophrenia susceptibility genes have recently been reported, but it is not clear whether these genes are associated with schizophrenia in general or with specific disease subtypes. In a previous study, we found an association of the neuregulin 1 (NRG1) gene with non-deficit schizophrenia only. We now report an association study of four schizophrenia candidate genes in patients with and without deficit schizophrenia, which is characterized by severe and enduring negative symptoms. Single-nucleotide polymorphisms (SNPs) were genotyped in the DTNBP1 (dysbindin), G72/G30 and RGS4 genes, and the relatively unknown PIP5K2A gene, which is located in a region of linkage with both schizophrenia and bipolar disorder. The sample consisted of 273 Dutch schizophrenia patients, 146 of whom were diagnosed with deficit schizophrenia and 580 controls. The strongest evidence for association was found for the A-allele of SNP rs10828317 in the PIP5K2A gene, which was associated with both clinical subtypes (P = 0.0004 in the entire group; non-deficit P = 0.016, deficit P = 0.002). Interestingly, this SNP leads to a change in protein composition. In RGS4, the G-allele of the previously reported SNP RGS4-1 (single and as part of haplotypes with SNP RGS4-18) was associated with non-deficit schizophrenia (P = 0.03) but not with deficit schizophrenia (P = 0.79). SNPs in the DTNBP1 and G72/G30 genes were not significantly associated in any group. In conclusion, our data provide further evidence that specific genes may be involved in different schizophrenia subtypes and suggest that the PIP5K2A gene deserves further study as a general susceptibility gene for schizophrenia.

Schizophrenia is a mental disorder that affects approximately 1% of the population worldwide. With a peak onset in late adolescence and early adulthood, it results in much individual suffering and large health-care expenses for society. Family studies have consistently shown that the susceptibility to schizophrenia is largely determined by heritable factors (approximately 80%) and that multiple genes and environmental factors must be involved (Sullivan et al. 2003). Recently, systematic fine-mapping efforts in regions of replicated genetic linkage have resulted in the identification of several putative schizophrenia susceptibility genes (Owen et al. 2005). However, their reported effect is typically modest, with estimated relative risks (RRs) of less than two, and the variations or haplotypes that increase the disease risk are not unique to schizophrenia patients. Interestingly, several of these genes appear to be associated with schizophrenia as well as with other psychiatric disorders (Chen et al. 2004b; Hattori et al. 2003; Hodgkinson et al. 2004; Green et al. 2005). Combined with the highly variable clinical presentation and disease course in individual patients, this suggests that particular genetic variations might play a role in specific disease symptoms. If so, genetic variations might distinguish patient subgroups with specific characteristics and ultimately guide the choice of treatment. Moreover, genetic heterogeneity could explain why the majority of genes involved in schizophrenia still seem to escape detection in unselected study samples.

So-called factor analyses have repeatedly shown that symptoms in schizophrenia can be grouped into several more or less independent dimensions. At least four symptom clusters are usually reported: positive, negative, disorganized and affective (McGrath et al. 2004; Serretti & Olgiati 2004). The presence of specific symptoms in individual patients varies considerably. Negative symptoms include loss of interest, social withdrawal and flattened emotions. Enduring negative symptoms and a generally poor prognosis are the hallmark of deficit schizophrenia, which is diagnosed using the Schedule for the Deficit Syndrome (SDS) (Kirkpatrick et al. 1989). Approximately, 15% of first-episode patients and 25–30% of chronic patients fulfil deficit criteria (Kirkpatrick et al. 2001). Several observations suggest that deficit schizophrenia represents a distinct biological entity (Kirkpatrick et al. 2001) with a high heritability (Kirkpatrick et al. 2000).

In a previous study, we found indications that the neuregulin 1 (NRG1) gene is associated with non-deficit schizophrenia but not with deficit symptoms (Bakker et al. 2004). This suggests that while some genes may be general susceptibility genes, variants in other genes may predominate in particular disease subtypes only. To further test this hypothesis, we have investigated the role of the DTNBP1 (dysbindin), G72/G30, RGS4 and PIP5K2A genes in deficit and non-deficit schizophrenia.

DTNBP1 was identified as a schizophrenia susceptibility gene after fine-mapping of a linkage region on chromosome 6p22. The first reported association in Irish multiplex families (Straub et al. 2002) has been replicated in multiple studies, although some have reported negative findings (for a review see Owen et al. 2005; Williams et al. 2005).

G72 and G30 are two overlapping genes, on chromosome location 13q32-33, which is a linkage region for both schizophrenia and bipolar disorder. Fine-mapping of a 5-megabase (Mb) critical interval resulted in the detection of significant association of schizophrenia with several single-nucleotide polymorphisms (SNPs) and SNP haplotypes (Chumakov et al. 2002). Association of the locus with schizophrenia has been confirmed in different populations (Addington et al. 2004; Korostishevsky et al. 2004, 2005; Schumacher et al. 2004; Wang et al. 2004; Zou et al. 2005), although negative results have also been reported (Mulle et al. 2005).

The first indications for involvement of the RGS4 gene came from gene expression studies, which showed a decreased expression in brains of schizophrenic patients (Mirnics et al. 2001). In addition, the gene is located in a schizophrenia linkage region on chromosome 1q21-q22 (Chowdari et al. 2002). A comprehensive association study of the region around the gene resulted in a group of four associated SNPs in three samples with a Caucasian, a mixed and an Indian background (Chowdari et al. 2002). Involvement of the same SNPs was subsequently confirmed in two Irish (Chen et al. 2004a; Morris et al. 2004) and a UK/Irish sample (Williams et al. 2004b). A large Scottish sample was associated with one of the SNPs only (Zhang et al. 2005), whereas a Han Chinese sample showed no association (Zhang et al. 2005).

Chromosome region 10p12 has repeatedly shown linkage with both schizophrenia and bipolar disorder. This region harbors the PIP5K2A gene, the product of which synthesizes phosphatidylinositol-4,5-bisphosphate, a membrane phospholipid that plays a central role in signal transduction and trafficking of synaptic vesicles. Lithium, which is used to treat bipolar disorder, is known to block inositol monophosphatases, enzymes that are part of the same phosphoinositide system (Stopkova et al. 2003). Different lines of evidence therefore indicate that PIP5K2A could be involved in both schizophrenia and bipolar disorder. Recently, an intragenic CT-repeat polymorphism was reported to be more frequent in patients with bipolar disorder than in controls (Stopkova et al. 2003). In a different study, several SNPs in the PIP5K2A gene were shown to be associated with schizophrenia (Sewekow et al. 2003) while a rare promoter variant may be a risk factor for schizophrenia (Stopkova et al. 2005).

We set out to investigate the association of previously reported SNPs in these genes with deficit and non-deficit schizophrenia in an ethnically homogeneous sample of 273 Dutch schizophrenia patients that was enriched for deficit schizophrenia (over 50% of patients).

Materials and methods

Sample collection

To enrich the sample for patients with the deficit syndrome, 308 schizophrenia patients were mainly recruited from psychiatric hospitals. Patients had at least three Dutch-born Caucasian grandparents. DSM-IV diagnosis of schizophrenia, excluding schizoaffective disorder, was made using the Comprehensive Assessment of Symptoms and History (CASH) (Andreasen et al. 1992) and information from medical records. The SDS (Kirkpatrick et al. 1989) was completed for 273 patients (89%), 146 (53%) of whom met deficit criteria. Patients without a completed SDS were excluded. In 29 patients with severe negative symptoms, it could not be ruled out that these symptoms were secondary to factors such as substance abuse. Following the SDS, they were classified as non-deficit schizophrenia. The Medical Ethical Committee of the UMC Utrecht approved the study and all patients gave written informed consent. The control panel (n = 580) consisted of 467 DNA samples from random Dutch individuals, obtained from the Immunogenetics and Transplantation Immunology Section of the Department of Immunohematology and Blood Transfusion, LUMC, Leiden, and 113 healthy controls from the Department of Biomedical Genetics, UMC Utrecht.

Marker selection

SNPs in the DTNBP1 and G72/G30 genes were selected based on publications and a previous screening of 28 SNPs in DNA pools (for details see Supplementary material on the website). SNPs for RGS4 were obtained from Chowdari et al. (2002) and for PIP5K2A from the Celera Database (Venter et al. 2001) and from a report by Sewekow et al. (2003).

Genotyping

Individual genotyping of SNPs in DTNBP1 and G70/G30 was performed on a 7900HT TaqMan system (Applied Biosystems, Foster City, CA, USA; further called ABI). SNPs in RGS4 and PIP5K2A were genotyped using an allele-specific PCR-assay with labelled primers, followed by separation of PCR products on a 3700 capillary sequencer (ABI). As controls, all four RGS4 and PIP5K2A SNPs were also blindly genotyped by direct sequencing in a random selection of 24 samples. Each 96-well PCR plate contained nine blinded duplicate samples.

Data analysis

Hardy–Weinberg equilibrium (HWE) was tested using a chi-square test. Linkage disequilibrium (LD) analyses were performed using gold software (Abecasis & Cookson 2000). Likelihood ratio tests for marker alleles and haplotypes were performed using the unphased program (Dudbridge 2003). Post hoc analyses of genotype frequencies were performed using a standard chi-square statistic with two degrees of freedom. Reported P-values are two-tailed, without correction for multiple testing.

Results

All tested markers were in HWE; resequencing of 24 random individuals for the four PIP5K2A and RGS4 SNPs confirmed genotyping results obtained by allele-specific PCR.

Patterns of LD between the SNPs in each gene are summarized in Table 1.

Table 1.  Patterns of linkage disequilibrium in DTNBP1, G72/G30, RGS4 and PIP5K2A

Marker		DTNBP1				G72/G30				RGS4		PIP5K2A
dbSNP	Alias*	P1655	P1635	P1765	SNP A	M14	M15	M23	M24	RGS4-1	RGS4-18	hCV11558870	hCV9591220
ns, not significant at the P = 0.05 level. D′ values for controls are shown below the diagonal, whereas r^2 values are shown above the diagonal. P numbers used by Straub et al. (2002); M numbers used by Chumakov et al. (2002); SNP A used by Williams et al. (2004a); RGS4-1 and RGS4-18, names according to Chowdari et al. (2002); hCV11558870 and hCV9591220, names in Celera database. * Marker names in different studies.
rs2619539	P1655		0.15	ns	0.06
rs3213207	P1635	0.94		0.49	0.08
rs2619528	P1765	ns	0.96		0.14
rs2619538	SNP A	0.21	0.80	0.76
rs3916967	M14						0.99	0.02	0.03
rs2391191	M15					1		0.02	0.03
rs3918342	M23					0.14	ns		0.91
rs1421292	M24					0.20	0.19	1
rs10917670	RGS4-1										0.55
rs2661319	RGS4-18									0.85
rs10828317	hCV11558870												0.55
rs746203	hCV9591220											0.78

DTNBP1

Table 2 summarizes allele frequencies and the resulting P-values of likelihood tests for alleles and marker haplotypes that were tested in the entire sample. None of the markers was significantly associated with deficit or non-deficit schizophrenia or the total schizophrenia sample. We also tested two- to four-marker haplotypes, none of which was significantly associated (data not shown).

Table 2.  Individual genotyping results

Gene	dbSNP*	Var^†	Control	Schizophrenia
				NDef^‡	P^§	Def^¶	P^§	Total**	p^§
Alleles and haplotypes with frequencies >5% are shown.P-values < 0.05 are shown in bold. * SNP names in dbSNP. ^† Var: alleles and haplotypes. Haplo: haplotypes, single and global; global, global P-value for haplotype. ^‡ NDef, non-deficit patients (n = 127). ^§ P, P-value of likelihood test of patient vs. control groups. ^¶ Def, deficit patients (n = 146). ^** Total, combined deficit and non-deficit samples (n = 273).
DTNBP1	rs2619539	C	53.3	53.9	0.873	57.9	0.183	56.1	0.307
	rs3213207	G	87.4	87.0	0.861	89.3	0.371	88.2	0.630
	rs2619528	G	21.8	22.6	0.766	21.5	0.932	22.0	0.898
	rs2619538	A	57.8	61.8	0.24	55.3	0.38	58.0	0.918
		Haplo			ns		ns		ns
G72	rs3916967	C	40.5	37.7	0.407	37.9	0.421	37.8	0.292
	rs2391191	C	61.0	62.6	0.649	62.4	0.677	62.5	0.574
	rs3918342	T	46.9	44.4	0.499	48.6	0.610	46.7	0.940
	rs1421292	A	53.9	55.2	0.747	54.9	0.782	55.0	0.696
		Haplo			ns		ns		ns
RGS4	rs10917670	G	37.8	45.4	0.033	38.7	0.790	41.9	0.131
	rs2661319	A	55.8	50.8	0.155	54.0	0.580	52.5	0.209
		A-A	52.2	49.0	0.336	54.5	0.568	51.9	0.832
		G-G	35.1	43.5	0.020	38.6	0.339	40.9	0.038
		A-G	9.2	5.2	0.047	6.5	0.192	5.9	0.034
		Global			0.022		0.269		0.025
PIP5K2A	rs10828317	A	62.1	70.3	0.016	71.9	0.002	71.2	0.0004
	rs746203	G	39.0	35.9	0.342	34.8	0.186	35.8	0.140
		A-G	7.2	7.7	0.884	8.8	0.408	8.4	0.305
		A-A	55.2	63.6	0.012	63.9	0.006	63.0	0.0045
		G-G	33.9	26.9	0.052	26.4	0.025	27.2	0.016
		Global			0.008		0.040		0.0003

G72/G30

Four SNPs that have repeatedly been associated with schizophrenia were genotyped. SNPs M14-M15 and M23-M24 formed two haplotype blocks, with hardly any LD between the blocks (Table 1). There was no significant association of single markers (Table 2) or any 2, 3 or 4-marker haplotype, single or global, with deficit or non-deficit schizophrenia, or with schizophrenia in general (data not shown). RGS4: There was significant linkage disequilibrium (LD) between the two tested markers (Table 1). Table 2 summarizes that the G-allele of SNP RGS4-1 was more frequent in non-deficit patients than in controls (P = 0.033), as were two-marker haplotypes (global P = 0.022 in the non-deficit patients; total sample P = 0.025). Post hoc analysis of genotypes for the associated SNP revealed a similar pattern, with the GA and GG genotypes increasing the risk mainly in non-deficit patients (Table 3). PIP5K2A: The two tested SNPs in PIP5K2A were in significant LD (Table 1). The A-allele of coding SNP rs10828317 was approximately 10% more frequent in the total sample (P = 0.0004), with a comparable contribution of the deficit and non-deficit groups. Haplotypes with SNP rs746203 were also associated (global P = 0.0003). Post hoc analysis by genotype suggests a recessive effect of the A-allele since the risk of disease seems to be highest for carriers of the AA genotype (Table 3). Genotype frequencies between deficit and non-deficit groups did not differ significantly, although heterozygotes appeared to be more frequent in the deficit group.

Table 3.  Genotype counts and analyses for associated SNPs in RGS4 and PIP5K2A^¶

Gene	SNP		Control		Schizophrenia
			Count	%	NDef*	%	Def^†	%	Total^‡	%
* NDef, genotype counts in non-deficit patients. ^† Def, genotype counts in deficit patients. ^‡ Total, genotype counts in combined deficit and non-deficit samples. ^§ Global P, P-value of chi-square test of genotype distributions of patients vs. controls.
RGS4	rs10917670	GG	82	16.4	21	17.2	19	14.1	40	15.6
		GA	214	42.7	68	55.7	67	49.6	135	52.5
		AA	205	40.9	33	27.0	49	36.3	82	31.9
		Global P^§			0.013		0.355		0.026
PIP5K2A	rs10828317	GG	73	14.5	13	11.0	6	4.3	19	7.4
		GA	247	49.0	44	37.3	64	46.4	108	42.2
		AA	184	36.5	61	51.7	68	49.3	129	50.4
		Global P^§			0.0099		0.0011		0.0002

Discussion

Several genes have been repeatedly shown to be associated with schizophrenia, but at present, the relevance of particular gene variants in clinical subtypes of schizophrenia remains unclear. We have investigated the association of variations in the DTNBP1, G72, RGS4 and PIP5K2A genes with deficit or non-deficit forms of schizophrenia. In the assessment of deficit/non-deficit states, it can be difficult to distinguish primary, disease-related negative symptoms from secondary symptoms caused by factors like depression or medication. The reliability of assessing these traits was not evaluated in this study. Misclassification, if present, would probably have resulted in an underestimation of differences between the groups.

The most convincing evidence for association with schizophrenia in our sample was found for a coding SNP (rs10828317) in the PIP5K2A gene (P = 0.0004 in the combined sample). Interestingly, we found this SNP to be strongly associated as a single marker, with both deficit and non-deficit schizophrenia. Two-marker haplotypes did not further increase the evidence, which is perhaps not surprising, because two markers probably capture only a fraction of possible haplotypes in the region. The genotype distribution in patients suggests a recessive effect of the A-allele. If this SNP is a causal variant, it would be a polymorphism with a high frequency in the general population. PIP5K2A may predispose to both schizophrenia and bipolar disorder because it is located in a linkage region common to both disorders. Moreover, it is part of a signalling pathway that is directly modulated by lithium (Stopkova et al. 2003), the most widely used medication in bipolar disorder. Our sample did not include patients with schizoaffective disorder, and one may speculate that the gene is more involved in psychotic than in mood symptoms. Taken together, the few available studies suggest that PIP5K2A is a promising susceptibility gene for psychiatric disorders, which deserves more attention.

Of the four repeatedly reported SNPs in RGS4, we have genotyped RGS4-1 and RGS4-18. These SNPs were in substantial LD with the other two SNPs and thus provide adequate coverage of previous work. The G-allele of SNP RGS4-1, as well as the two-marker G-G haplotype had a higher frequency in non-deficit patients. In post hoc analysis by genotype, SNP RGS4-1 was also associated with schizophrenia in the entire group. Others have recently reported stronger association in more narrowly defined schizophrenia groups (Chen et al. 2004a). Our data suggest that the gene is specifically involved in a further refined subgroup of schizophrenia patients without prominent negative (deficit) symptoms. Testing multiple markers and clinical phenotypes increases the chances of false–positive findings, and after correction for multiple testing, P-values are no longer significant at the P = 0.05 level. It should be noted, however, that the evidence in previous studies was also modest. The combined results from genetic and gene expression studies, together with its function in neurotransmission, suggest that RGS4 is involved in schizophrenia or in specific clinical aspects of the disorder.

For the dysbindin gene, we could not confirm the marginally significant association in DNA pools in the entire sample, which suggests that the pooling results in the smaller samples were chance findings. It is possible that our markers were not in strong LD with disease-related variants because there appears to be substantial molecular heterogeneity, even in geographically related patient samples. Our set of markers included SNPs that have repeatedly shown association as single markers or as (part of) associated haplotypes (see Supplementary material for a comparison of several published studies), but our set was limited and does not cover the entire gene. Therefore, we cannot exclude that yet other haplotypes are associated in the Dutch population. Small effects can also easily be missed by chance in the samples with modest sizes that have been studied so far. Under favorable conditions, the combined sample has ∼80% power to detect a locus with a RR of 1.6 (P = 0.05). For a locus with a RR of two, the power would be 98% (Purcell et al. 2003). Because previously reported RRs were typically less than two, we might have missed a truly associated locus due to insufficient power. The specific character of our sample provides yet another explanation for the lack of association. We have excluded patients with schizoaffective disorder, and specifically selected patients with a poor disease outcome and severe negative symptoms. Similar reasons may underlie our negative findings at the G72/G30 locus, where we have genotyped four markers that were repeatedly found to be associated with schizophrenia (see Supplementary material). Involvement of both DTNBP1 and G72/G30 in Dutch schizophrenia patients can therefore not be excluded based on these data.

In conclusion, our data suggest that the PIP5K2A gene increases susceptibility to schizophrenia in general while the RGS4 gene appears associated with the non-deficit subtype only. This is in concordance with our previous studies in which we showed that the NRG1 gene was associated with non-deficit schizophrenia only (Bakker et al. 2004). The schizophrenia genes identified so far necessarily play a substantial role in the unselected samples in which they were detected, but together these genes leave most of the genetic susceptibility to schizophrenia unexplained. Our findings clearly illustrate the feasibility and the need of defining particular patient groups with specific characteristics, not necessarily with the same clinical diagnosis, to identify additional susceptibility genes for psychiatric disorders.

Electronic database information

Celera: http://www.celera.com

DbSNP: http://www.ncbi.nlm.nih.gov/projects/SNP/

Acknowledgements

The authors thank N. M. Williams for kindly providing details of SNP A and associated haplotypes before their publication (Williams et al. 2004a). They also thank H. G. Otten for providing control DNA samples. This project was supported by the UMC Utrecht’s Genvlag program, ABC Neurogenomics and the Makaria Foundation. There are no conflicts of interest to declare.