Genome-wide association study of atypical psychosis


  • H.Y. and N.I. are joint last authors.
  • Conflicts of interest: nothing to declare.


Atypical psychosis with a periodic course of exacerbation and features of major psychiatric disorders [schizophrenia (SZ) and bipolar disorder (BD)] has a long history in clinical psychiatry in Japan. Based upon the new criteria of atypical psychosis, a Genome-Wide Association Study (GWAS) was conducted to identify the risk gene or variants. The relationships between atypical psychosis, SZ and BD were then assessed using independent GWAS data. Forty-seven patients with solid criteria of atypical psychosis and 882 normal controls (NCs) were scanned using an Affymetrics 6.0 chip. GWAS SZ data (560 SZ cases and 548 NCs) and GWAS BD (107 cases with BD type 1 and 107 NCs) were compared using gene-based analysis. The most significant SNPs were detected around the CHN2/CPVL genes (rs245914, P = 1.6 × 10−7), COL21A1 gene (rs12196860, P = 2.45 × 10−7), and PYGL/TRIM9 genes (rs1959536, P = 7.73 × 10−7), although none of the single-nucleotide polymorphisms exhibited genome-wide significance (P = 5 × 10−8). One of the highest peaks was detected on the major histocompatibility complex region, where large SZ GWASs have previously disclosed an association. The gene-based analysis suggested significant enrichment between SZ and atypical psychosis (P = 0.01), but not BD. This study provides clues about the types of patient whose diagnosis lies between SZ and BD. Studies with larger samples are required to determine the causal variant. © 2013 Wiley Periodicals, Inc.


Schizophrenia (SZ) and bipolar disorder (BD) are chronic and debilitating mental illnesses with serious emotional, cognitive, behavioral, and financial consequences, not only for affected individuals, but also for their families and society as a whole. With a typical onset in early adulthood, these disorders often have devastating effects on the patient's entire adult life. They are also common, each affecting almost 1% of the population worldwide, with high heritability (∼80%) [Tsuang et al., 1997; Merikangas et al., 2007; Saha et al., 2008]. For these reasons, SZ and BD are considered major public health concerns [Wyatt et al., 1995; Rice, 1999; Goldner et al., 2002].

SZ and BD have been classified as separate illnesses since the late 19th century following the long-term clinical observations of E. Kraepelin and his group [Kraepelin, 1921, 1971], and modern nosological systems, such as the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision (DSM-IV TR) [American Psychiatric Association, 2000] and the International Classification of Diseases, 10th revision (ICD-10) [World Health Organization, 1993], have assimilated the concept, known as the Kraepelinian dichotomy. In fact, it has been concluded from comprehensive epidemiological research, such as the IOWA500 study led by M.T. Tsuang [Tsuang and Winokur, 1975; Tsuang et al., 1980; Goldstein et al., 1990], that relatives of SZ sufferers have a higher risk of the condition relative to the general population (by 3.2 ± 0.9%, mean ± SD), but not of BD and other mood disorders (e.g., mania, by 1.0 ± 0.7%; depression, by 0.9 ± 0.5%). Similarly, the relatives of BD patients have a higher risk of BD (mania, by 13.1 ± 0.7%; depression, by 12.9 ± 1.8%), but not of SZ (by 7.0 ± 1.5%).

However, the validity of this distinction has come under increasing scrutiny. Their core features differ and each can be reliably diagnosed to the exclusion of the other, but the commonly encountered secondary features of each illness resemble the primary features of the other (i.e., affective disturbance in SZ and psychosis in BD). This suggests that the disorders exist as different extremes on a multidimensional spectrum that encompasses both affective and psychotic syndromes. Clinically, the boundaries between the two are often breached in the form of psychotic BD or the bipolar type of schizoaffective disorder. Furthermore, the disorders may share some biological substrates, as suggested by the responsiveness of both SZ and BD to atypical antipsychotics.

Beyond their clinical, behavioral, and biological similarities, there is recent evidence that these disorders also share genetic underpinnings [Berrettini, 2004; Murray et al., 2004; Sullivan et al., 2012]. At each level of analysis, from family [Van Snellenberg and de Candia, 2009] and twin [Cardno et al., 2002] studies to candidate-gene analyses [Craddock et al., 2001, 2005; Glatt et al., 2003] and unbiased linkage [Badner and Gershon, 2002; Potash et al., 2003] and genome-wide association studies (GWASs) [Green et al., 2009; International Schizophrenia Consortium et al., 2009; Williams et al., 2011a], the evidence for an extensive genetic overlap between BD and SZ has become increasingly persuasive. Indeed, the title of a recent publication is quite telling: “Most genome-wide significant susceptibility loci for schizophrenia and bipolar disorder reported to date cross traditional diagnostic boundaries” [Williams et al., 2011b]. According to Berrettini [Berrettini, 2004], “the current nosology must be changed to reflect the new knowledge concerning the shared etiologies of bipolar disorder and schizophrenia”. Thus, the recent biological findings bring into focus a new perspective, such that the disorders are now regarded as two prominent summits over the widespread “endogenous psychoses pathology” [Owen et al., 2007]. This pathology, which encompasses a wide area, has been recognized as “amorphous”, and therefore little intensive research employing biological methods has been performed throughout the history of research on psychiatry.

On the other hand, in the history of taxonomy of endogenous psychoses, there have been some attempts to classify the rare group from the group of “amorphous” patients. Such research has resulted in the introduction of a third summit featuring the characters of both SZ and BD as an independent entity (Fig. 1). Zykloide Psychose [Leonhard, 1957, 1961], or bouffée déliriante [Magnan, 1893], has the features of acute onset, hallucination and/or mood disturbance at the acme phase. Patients with this type of the disorder worsen periodically, but they can live a normal life after such periods of short-duration exacerbation, and may often exhibit a memory disturbance about the duration. On adopting the recent nosological classification criteria, these patients are classified using the DSM-IV TR as having schizoaffective disorder (295.7) or brief psychotic disorder (298.8) [American Psychiatric Association, 2000], whereas they would be diagnosed with acute and transient psychotic disorders (F23) or schizoaffective disorder (F25) using ICD-10 criteria [World Health Organization, 1993]. It is of note that the original concept of schizoaffective disorder has features of both acute and transient psychotic disorders (F23) and schizoaffective disorder (F25) [Kasanin, 1994].

Figure 1.

The schema of atypical psychosis between schizophrenia and bipolar disorder. The patients with Atypical psychosis show characteristics of Schizophrenia (delusion or hallucination) and bipolar disorder (mood disturbance).

In Japan, the affected group was named “Mitsuda psychosis” or “atypical psychosis” [Mitsuda, 1965], and various types of research have been conducted to clarify the phenotype [Mitsuda, 1950; Mitsuda and Sakai, 1968; Hatotani and Nomura, 1983; Hayashi et al., 2001]. Using the candidate gene approach, psychiatric geneticists have found that STAT6 (Gene ID: 6778, signal transducer and activator of transcription 6, interleukin-4 induced) is involved in the etiology of this disorder [Kawashige et al., 2008]. However, atypical psychosis, as the term indicates, has become a target for criticism because it is ambiguous as to which patients can be classified into this disorder. In response to this criticism, new clinical diagnostic criteria have been established by clinical specialists of atypical psychosis in Japan (Table 1). However, even according to these criteria, some data are still not available, such as morbidity, prevalence, and differences in the suicide rate, mortality rate, profit, and loss for the society and disability-adjusted life years between similar disorders such as SZ and BD.

Table 1. Diagnostic Criteria for Atypical Psychosis
  • aCharacterized by intense feelings of happiness or ecstasy, overwhelming anxiety or marked irritability.
  • bCharacterized by puzzlement over perceived disorganization of thought, misidentification of people or places, or incoherence of the train of thought, resulting in impairment in cognition of the environment and intellectual performance, leading to confused thoughts and behaviors.
  • cAt least one of the following must be prominent: (1) Motoric immobility as evidenced by catalepsy or stupor. (2) Excessive motor activity. (3) Extreme negativism or mutism. (4) Peculiarities of voluntary movement as evidenced by posturing, stereotyped movement, prominent mannerisms or prominent grimacing. (5) Echolalia or echopraxia.
A.A sudden onset of psychotic symptoms (2 weeks or shorter from a mentally healthy state to a clearly psychotic state that fulfils criterion B)
B.Symptoms from at least two of the following categories, occurring simultaneously
 1. Emotional turmoila
 2. Perplexity and confusion of memoryb
 3. Catatonic behaviorc or hallucinations or delusions
C.The total duration of the disorder does not exceed 3 months, and there is almost complete recovery to the premorbid level of functioning
 The diagnosis should be qualified as “provisional” until 3 months after the onset
D.The disturbance is not due to the direct physiological effects of a substance or a general medical condition

Similarly, it essential to determine the etiology based on biological research. In particular, the susceptibility genes must be found because the estimated heritability in atypical psychosis and other similarities is slightly higher than for SZ and BD [Leonhard, 1957; Das et al., 2001; Marneros and Pillmann, 2004]. If the highly penetrant causative genes could be established in this particular patient group, there would be important consequences for etiological research on similar disorders. Thus, we explored the causative genes by employing the latest gene-chip platform for GWASs. In addition, comparison of the resulting GWAS data of similar disorders (i.e., SZ and BD) rendered it possible to determine the degree of overlap that exists between the three groups (i.e., atypical psychosis, SZ, and BD), or whether atypical psychosis is more similar to either SZ or BD.



We selected 47 patients with atypical psychosis (males 35.4%) according to the diagnostic criteria (Table 1) and 882 psychiatrically unscreened healthy controls (males 48.9%) for GWAS analysis. All subjects were unrelated, living in Japan, and self-identified as Japanese. The subjects provided written informed consent to participate after receiving a complete description of the study. This study was approved by the ethics committees of each university participating in this project.

For the enrichment analyses, we used an independent Japanese sample of SZ GWAS (560 SZ and 548 controls) and BD GWAS (107 BD type I and 107 controls). The detailed data of the SZ cases are described elsewhere [Ikeda et al., 2011], and the GWAS dataset for BD was obtained from the “Japanese Genetics Initiative of Mood Disorders data of GWAS for Bipolar Disorder” ( [Hattori et al., 2009].

GWAS and Quality Control

Genotyping was performed using the Affymetrix Genome-Wide Human single-nucleotide polymorphism (SNP) Array 6.0 (Santa Clara, CA) according to the manufacturer's protocol. After applying several quality control (QC) criteria [e.g., call rate ≥95%, autosomal chromosomes, Hardy–Weinberg equilibrium ≥0.0001 and minor allele frequency (MAF) ≥5%], the final GWAS comprised 929 samples (47 cases and 882 controls) and 545,513 SNPs (MAF ≥5%). Q-Q plots were generated on the basis of an allele-wise analysis of SNPs that passed QC criteria (Fig. 2); our observed value of λ (=1.027) was consistent with those generally reported in well-matched samples.

Figure 2.

Quantile-quantile (QQ) plot for association results (high-resolution image is also in Supplementary Information). The QQ plot shows the distribution of expected p values against the observed distribution. The empirical and theoretical distributions are shown as black dots and red line, respectively.

Gene-Based Comparison With SZ and BD Using Versatile Gene-Based Association Study (VEGAS) Software

Further investigation was conducted with the aid of Versatile Gene-based Association Study (VEGAS) software [Liu et al., 2010], which enables the comparison of different GWAS data with a higher power than individual SNP-based analyses. The software annotates SNPs according to their position in genes, produces a gene-based test statistic, and then uses simulation to calculate an empirical gene-based P-value. The HapMap Asian sample (Tokyo, Japan, and Beijing, China) was used as a reference and the 3′/5′ boundaries of the genes were set at 50 kb from the initial and last exons for each gene.

Gene enrichment was evaluated by hypergeometric analysis and the significance level was set at P < 0.05. This test enables us to evaluate whether suggested genes with P-values <0.05 in one disorder may correspond to the genes implicated in another disorder. The significance of the hypergeometric test represents the gene enrichment, with a higher number indicating more genes significantly matched than expected at random.


Single-Marker Association Analysis

Our atypical psychosis GWAS data did not detect an association with a genome-wide significance level (Fig. 3), although some susceptibility genes were suggested (Table 2). The top-ranked was CHN2 (chimerin 2) [Hashimoto et al., 2005], which interacts with both AKT1 [Ikeda et al., 2004] and ERBB3 [Kanazawa et al., 2007] genes, which encode a protein with a phorbol-ester/DAG-type zinc finger (rs245914, P = 1.6 × 10−7). Another top-ranked CPVL (carboxypeptidase) exhibits strong sequence similarity to serine carboxypeptidases (rs12196860, P = 2.5 × 10−7), although the detailed function and its full-length nature have yet to be determined.

Figure 3.

Manhattan plot of atypical psychosis (high resolution image is also in Supplementary Information). The −log10 P values (y axis) of all SNPs in 47 cases with atypical psychosis and 882 healthy controls are shown relative to their chromosomal positions (x axis). The red horizontal line represents a genome-wide significance level (P = 5 × 10−8).

Table 2. Top SNPs Based on GWAS of Atypical Psychosis
RankSNPChromosomePositionReference alleleP-valueaClosest geneb
  • SNP, single-nucleotide polymorphism; GWAS, Genome-Wide Association Study.
  • The notational convention, such as Position and Reference allele, are in accordance with the National Center for Biotechnology Information SNP database.
  • aP-value was calculated on the basis of the allele-wise test (two-tailed).
  • bIdentified using the National Center for Biotechnology Information SNP database.
1rs245914729218159G1.61 × 10−7CHN2, CPVL
2rs12196860655950374A2.45 × 10−7COL21A1
3rs121054212103576088C5.31 × 10−7 
4rs19595361451446771C7.73 × 10−7PYGL, TRIM9
5rs60815412019212890A1.24 × 10−6SLC24A3
6rs4619807323193928A1.55 × 10−6UBE2E2, RPL24P7
7rs169024608128920197G2.61 × 10−6PVT1
8rs157259113102060076A2.99 × 10−6ITGBL1, NALCN
9rs80299891538733847A3.54 × 10−6FAM98B, RASGRP1
10rs2736172631590898G3.60 × 10−6MICB, TNF

The third-ranked gene, COL21A1 (collagen, type XXI, alpha 1), was significantly associated with an SZ GWAS analysis led by another group [Stefansson et al., 2009]. One of the highest peaks was detected on the major histocompatibility complex region (MHC), which is rs2736172 on Chromosome 6 (P = 3.60 × 10−6) closed to MICB, TNF, and HLA genes.

Gene-Based Comparison With SZ and BD

The gene-based analysis with employing the hypergeometric test revealed significant enrichment between SZ and atypical psychosis (P = 0.014), but not BD (P = 0.93; Fig. 4).

Figure 4.

The schema of overlapping percentages between three disorders (P < .05). Based on the result of VEGAS software analysis, the number of the genes with its P-value < 0.05 is shown in this schema (see Table 3). The overlapping percentage within Atypical psychosis is 7.10% (70/986) with SZ, and 5.29% (27/510) with BD. The enriched relationship of atypical psychosis with SZ was statistically significant (P = 0.011).


The scientific approach of beginning with an accurate classification, and ideally the precise nosology based on the symptomatology, represents a fundamental contribution to the scientific progress of clinical psychiatry. The feedback from new knowledge obtained through biological research improves clinical psychiatry, and brings crucial benefits to patients suffering from a psychiatric disorder. Since cases that are intermediate between SZ and BD are “amorphous”, they are not usually a target for scientific research. Many researchers have struggled with this subgroup of patients, and no unequivocal evidence is available [Marneros and Tsuang, 1986; Andreasen and Carpenter, 1993; Marneros et al., 1995]. The objective selection of patients in the present study based on new atypical psychosis criteria, and the latest GWAS analysis and comparison with other similar disorders have revealed new clues as to the genetic features of these amorphous cases.

First of all, compared with the normal controls, the most-associated SNP did not reach a genome-wide significance level despite selected samples based on the relatively “pure” phenotype. Additionally, the SNPs in the STAT6 genes positively suggested as a susceptible gene by the haplotype method were not positively associated with this GWAS analysis. However, it is of note that the SNPs in the major histocompatibility complex (MHC) region on chromosome 6 exhibited a nominal association with atypical psychosis (Supplementary Fig. 5). This area is well established as being associated with the robust risk of SZ by several GWAS, as well as more recently by mega-analysis conducted by the Psychiatric GWAS Consortium. However, the MHC region comprises large blocks with a very high linkage disequilibrium [Jeffreys et al., 2001]. Thus the particular “risk” gene being difficult to reliably identify at the present time.

Comparison of our findings with independent GWAS datasets in the same population should only be conducted with the caveat that GWAS platforms differ between the groups. If there was no discrepancy between the platforms, the results from the different groups can be compared using polygenic component analysis [International Schizophrenia Consortium et al., 2009]. However, we used the latest version of the Affymetrics chip (version 6.0), even though the Japanese GWAS BD data available for comparison were obtained using another platform (Affymetrics 100 K). Since the SNPs on the arrays differed between atypical psychosis, SZ, and BD, it was not possible to make direct comparisons based on individual SNPs due to the use of different platforms (i.e., Affymetrics 6.0, 5.0, and 100 K, respectively). This meant that VEGAS software was applied [Liu et al., 2010], which calculates the P value for each gene even for data from different platforms (of course, results based on denser SNPs can generate more-accurate P values). The obtained data indicated that atypical psychosis is closer to SZ than to BD; 7.10% (70/986) of genes for which P < 0.05 overlapped with SZ, while 5.29% (27/510) of such genes overlapped with BD. The enriched relationship of atypical psychosis with SZ was statistically significant (P = 0.011) by hypergeometric analysis, although this is a nominal level of significance, it appears that both disorders (SZ and AP) may have a shared genetic risk (Table 3). It should be noted that a lower total number of genes in common with BD is partly due to the smaller BD sample size than SZ, therefore the failure to attain shared significance with BD could be partly explained by lower statistical power.

Table 3. Gene-Based Analysis of Atypical Psychosis Compared With SZ and BD
  atypical psychosis 
  P < 0.05P ≥ 0.05Total 
SZP < 0.0570916986 
 P ≥ 0.0585815,37116,229 
 Total92816,28717,215P = 0.014
  atypical psychosis 
  P < 0.05P ≥ 0.05Total 
  1. SZ, schizophrenia; BD, bipolar disorder.
  2. Using the VEGAS software, a P-value was calculated from the sum of χ statistics of all GWAS SNPs in a certain gene. The number in the table is for the genes divided by P-value calculated by VEGAS software. The P-value on the bottom-right corner represents the result of the hypergeometric analysis for evaluating the gene enrichment between two disorders.
BDP < 0.0527483510 
 P ≥ 0.0562811,02911,657 
 Total65511,51212,167P = 0.93

The present findings should be interpreted with caution, mainly because of the small sample, especially for atypical psychosis. The estimated prevalence of the disease is 11.4% within SZ and similar patients [Regier et al., 1994; Sartorius et al., 1995], which represents a prevalence of approximately 0.1% in the general population. The present study adopted specific criteria to determine the solid causative gene, which is why our sample was so limited in this preliminary study. Studies of more samples chosen based on the same diagnostic criteria will be essential for conclusive results. Moreover, larger samples of the Japanese BD cohort with a denser DNA chip are also required—this would enable direct comparisons at the individual SNP level in addition to using polygenic component analysis.

Because of the newly established criteria for atypical psychosis, another limitation is the lack of inter- and intra-rater reliability. Moreover, as described in the Introduction section, we still need to determine the morbidity, prevalence, and differences in the suicide rate, mortality rate, profit and loss for the society and disability-adjusted life years between similar disorders such as SZ and BD, the replication work in different countries and areas is warranted.

In the GWAS era, a larger sample (∼10,000) is essential in order to satisfy statistical requirements. However, we believe it is important to shed light on specific groups with certain phenotypes or courses, with the realization that it will be difficult to obtain a sufficient large number of patients that fit the criteria for those specific groups. The current method adopted in mainstream research under the name of SZ or BD may overlook the importance of clinical nosology. To see the patients in detail in order to make the appropriate diagnosis and in order to give the suitable treatment is another essential part of psychiatry, and these two disorders appear to be empirically heterogeneous. At the very least, the subgroups of SZ and BD should be considered before sampling. Data obtained in future investigations similar to the current study will provide clues that will reduce the heterogeneity of both SZ and BD.


The diagnostic criteria for “atypical psychosis” were established by a working group comprising Doctors Takaaki Abe, Hirohiko Harima, Akira Iwanami, Kosuke Kanemoto, Jun Koh, Kazuhiko Nakayama, Kaoru Sakamoto, Hiroshi Yoneda and Hidemichi Suga (the Chairman). Dr. Hirohiko Harima's insightful comments and suggestions regarding translation of the manuscript into English are particularly appreciated. Finally, we owe an intellectual debt to Professor Stephen V. Faraone and Professor Ming T. Tsuang, since this work could not have been published without their support.