Microduplications Disrupting the MYT1L Gene (2p25.3) are Associated with Schizophrenia

Abstract

Childhood-onset schizophrenia (COS) is a rare severe form of schizophrenia that may have greater salient genetic risk. Despite evidence for high heritability, conclusive genetic causes of schizophrenia remain elusive. Recent genomic technologies in concert with large case-control cohorts, have led to several associations of highly penetrant rare copy number variants (CNVs) and schizophrenia. We previously reported two patients with COS who carried a microduplication disrupting the PXDN and MYT1L genes at 2p25.3. This is a significantly higher rate of duplications within our COS population (N=92) when compared to 2,026 healthy controls (p=0.002). As a replication, we report a meta-analysis of four recently published studies that together provide strong evidence for association between variably-sized microduplications involving the MYT1L gene and schizophrenia. None have reported this separately. Altogether, among 5,325 cases and 9,279 controls, 10 microduplications were observed; 9 in cases and 1 in a control (OR=15.7, p=0.001). Further, the 2% rate observed in our COS patients is also significantly higher than the rate in adult onset cases (0.14%, OR=16.6, p=0.01). This report adds to the growing body of literature implicating rare CNVs as risk factors for schizophrenia, and that some risk CNVs are more common among extreme early-onset cases.

INTRODUCTION

Schizophrenia is a chronic and severe disorder that affects ~1% of the population and ranks among the top 10 causes of disability worldwide(Tandon et al. 2009). Childhood-onset schizophrenia (COS), a rare and severe form of the condition exhibits a prevalence of less than 1 in 30,000(McKenna et al. 1994). Previous work has established the phenotypic and neurobiologic continuity between COS and the more typical adult onset form of the disorder (AOS)(Nicolson and Rapoport 1999; Rapoport et al. 1999). For 20 years, the child psychiatry branch of the National Institutes of Mental Health has studied COS to understand the etiology and neurobiology of the illness.

Recently, high density microarrays have allowed us to perform genome-wide assessments of previously unknown submicroscopic genomic deletions and duplications. These technologies have identified many rare Copy Number Variants (CNVs) that impact known genes in the COS population(Walsh et al. 2008). Since 2008, several studies have provided substantial evidence that there is a greater burden of CNVs in schizophrenia. For example, The International Schizophrenia Consortium (ISC) conducted one of the largest studies with 3,391 cases and 3,181 controls from seven European populations; and found cases had a 1.15 fold excess of rare CNVs which increased to 1.67 fold for deletions >500 Mb(Consortium 2008). Also, the Wellcome Case Control Consortium (WCCC) assessed the involvement of rare (<1%) CNVs in 471 cases of schizophrenia and 2,792 controls and determined large CNVs (>1 Mb) were 2.26 times more common in cases (p=0.00027)(Kirov et al. 2009). Lastly, CNV burden has drawn attention from the most recurrent CNV deletions (1q21.1, 3q29, 15q11, 15q13.3, and 22q11.2) and duplications (7q36.3(VIPR2) and 16p11.2)) which are now considered to be strong risk factors for schizophrenia(Levinson et al. 2011; Magri et al. 2010).

In 2008, Vrijenhoek et al. first reported two schizophrenia patients with microduplications that disrupted MYT1L at 2p25.3. We subsequently reported two COS patients also carrying microduplications impacting the MYT1L and PXDN genes(Addington and Rapoport 2009). Contrary to other recurrent CNVs associated with schizophrenia, this region does not contain an obvious mechanism for non-allelic homologous recombination (NAHR), i.e. low copy flanking repeats(Ingason et al. 2011). Given that this was the only recurrent CNV besides 22q11.2 and 16p11.2 observed in the COS cohort, we tested the hypothesis that microduplications in 2p25.3 are associated with schizophrenia through a meta-analysis of public datasets.

MATERIALS AND METHODS

Five published case-control studies were combined from publicly available validated CNV calls. Cohorts, ancestry and microarray platforms are detailed in Table 1. Published .bed files from large population cohorts were collected from each study to examine CNVs at 2p25.3. We performed association tests by comparing the number of variants reported in cases versus controls using Fisher’s exact test.

Table 1.

Study	cases n	# of dups.	%	controls n	# of dups.	%	size (kb)	UCSC Build HG18	Odds Ratio	p-val
NIMH COS1	92	2	2.17%	2026*2	0	0.00%	143-216	chr2:1618945-1835426	NA	0.002
								chr2:1713636-1857129
other studies
Vrijenhoek et al 20083	806	2	0.25%	706	0	0.00%	967-3800	chr2:859616-1826716	NA	0.5
								chr2:1400000-4500000
ISC 20084	3391	3	0.09%	3181	1	0.03%	103-238	chr2:1712998-1815909	2.8	0.63
								chr2:1712998-1922382
								chr2:1716437-1826717
Kirov et al 20095	471	1	0.21%	2792	0	0.00%	484	chr2:1785886-2269413	NA	0.14
Ikeda et al 20106	575	1	0.17%	564	0	0.00%	855	chr2:987335-1842013	NA	1
Totals
total other studies	5233	7	0.13%	7253	1	0.01%			9.7	0.01
total all 5 studies	5325	9	0.17%	9279	1	0.01%			15.7	0.001
COS vs. other studies
COS vs. AOS7	92	2	2.17%	5233	7	0.13%			16.6	0.01

¹Affymetrix 500K arrays and Agilent 185K or 244K CGH (for validation); mixed ancestry; Walsh et al. 2008

^*2Illumina HumanHap 550 arrays; 2/3 Caucasian, 1/3 African-American ancestry; controls from Children’s Hospital Of Philadelphia (CHOP) CNV database (cnv.chop.edu)

³Affymetrix 250K or Illumina HumanHap550v3 arrays; Dutch ancestry from GROUP (Genetic Risk and Outcome I Psychosis) Consortium

⁴Affymetrix 5.0 or 6.0 arrays; 7 populations of European ancestry

⁵Affymetrix 500K arrays; WTCCC (Wellcome Trust Case Control Consortium)

⁶Affymetrix 5.0 arrays, Japanese ancestry

⁷COS=childhood onset schizophrenia; AOS=adult onset schizophrenia

RESULTS

The number of CNVs affecting the coding region of the MYT1L gene in cases and controls are listed (Table 1). We compared our NIMH COS patient sample to the 2,026 healthy children sample published by Shaikh et al 2009(Shaikh et al. 2009). Comparing CNVs in the 2p25.3 region, there was a significant excess in the COS sample (p=0.002). While no single study from the published literature revealed a trend for association with duplications in this region, in aggregate, there was a clear excess of duplications in subjects with schizophrenia compared to controls (OR=9.7, p=0.01). Combining all 5 studies, a total of 10 duplications (9 cases; 1 control, Figure 1), and zero deletions were reported (OR=15.7, p=0.001). Further, COS subjects had a significantly higher rate of duplications when compared to AOS subjects (OR= 16.6, p=0.01). The microduplications associated with schizophrenia varied in size from 103 kb to 3.8 Mb, though the minimal region of overlap was in the 1,800,000 to 1,900,000 region of 2p25.3 (UCSC hg18), disrupting the 3′ end of MYT1L. Lastly, all but one of the microduplications observed in the subjects with schizophrenia also impacted the neighboring PXDN gene. Interestingly, a qPCR analysis revealed significant over-expression of MYT1L exon junction 21/22 in patient COS534 (carrying the most restricted duplication), but no differential expression for patient COS1358 versus unaffected controls. Neither patient demonstrated detectable expression for the remaining exons (MYT1L 22/23; 22/24; PXDN exon 1) after 40 cycles under optimized conditions (Supplementary table 1).

Figure 1.

Combination of all 5 studies, a total of 10 duplications (9 cases; 1 control), and zero deletions were reported (OR=15.7, p=0.001).

In terms of inheritance patterns, patient COS-NSB1358 inherited the microduplication from his unaffected mother, but was not transmitted to a healthy sibling. Inheritance for COS-NSB534 was undeterminable as the mother lacked the CNV, and the father was unavailable.

DISCUSSION

We report that microduplications disrupting the MYT1L gene confer a clear increase in risk for schizophrenia. This study provides strong evidence for an association between schizophrenia and MYT1L, whose increase is reported here for the first time and was made possible by meta-analysis of four published studies. The overall rate of 0.17% for this duplication in patients with schizophrenia is comparable with rates observed for other rare CNVs associated with high risk (i.e. 1q21.1 and 22q11.2)(Kirov 2010). In addition, our data suggests even greater prevalence of this microduplication in COS patients as compared to other published large population studies which include adult onset cases (OR=16.6). This increased rate of specific rare CNVs concurs with work from our group on 22q11.2 deletions(Sporn et al. 2004) and 16p11.2 duplications(Addington and Rapoport 2009; McCarthy et al. 2009), which suggests more salient genetic influences in COS compared to AOS(Addington and Rapoport 2009).

The reported microduplications varied in size and breakpoint, likely due to the absence of flanking segmental duplications. Thus nonallelic homologous recombination is the unlikely CNV promoting mechanism for 2p25.3 (though common for other recurrent CNVs associated with neuro-developmental disorder risks (Carvalho et al. 2010; Lupski and Stankiewicz 2005)). Aside from the association with deletions in NRXN1, the affected region is also unique because the duplications affect only 2 genes (MYT1L and PXDN), limiting the number of implicated risk genes. In contrast, other schizophrenia-associated CNV regions contain as many as 28 (16p11.2 duplication) or 40 genes (22q11.2 deletion), making identification of the most relevant genes difficult. However, inference about the functional impact of duplications is more challenging than deletions because of the unknown location and orientation of the duplicated DNA. Thus long-range epistatic effects on gene dosage and aberrant transcription are implicated but not confirmed.

MYT1L is known to be involved in proliferation and differentiation of oligodendrocytes, thus disruption of the coding region could lead to aberrant transcriptional activity throughout the CNS(Kim et al. 1997). Work within our group suggests COS patients exhibit up to 2.2% slower white matter growth rates per year than healthy controls; thereby relating abnormalities in oligodendrocyte function to potential deficits in schizophrenia(Gogtay et al. 2008). Law et al. suggest that MYT1L family genes may regulate Neuregulin1 and modulate synaptic development and function(Law et al. 2006). Work by Vierbuchen et al. also demonstrate that MYT1L, along with ASCL1 and BRN2, can convert mouse fibroblasts directly into functional neurons in vitro(Vierbuchen et al. 2010). The authors hypothesize that changes in transcriptional activity could lead to genome-wide alterations of epigenetic mechanisms affecting DNA methylation and histone modifications. Thus, there is a clear putative role for MYT1L in affecting neuronal cell identity, differentiation and plasticity. The PXDN gene is an extracellular matrix-associated peroxidase thought to affect peroxide driven oxidations, phagocytosis and immune defense. Concurrently, much literature suggests the presence of oxidative stress playing an important role in the physiopathology of schizophrenia(Do et al. 2009), therefore PXDN may be a potential risk gene as well.

In conclusion, current studies suggest that the 2p25.3 locus gives rise to diverse neurodevelopmental phenotypes(Buizer-Voskamp et al. 2011). Further resequencing efforts will be critical in elucidating the role of this gene as a risk factor for schizophrenia. Our findings provide convincing evidence for the role of microduplications involving MYT1L that provide further insight into the genetic basis of schizophrenia.