Background

Obsessive-compulsive disorder (OCD) is characterized by distressing, intrusive obsessive thoughts and/or repetitive compulsive behaviours [1]. According to a survey of adult US citizens, OCD is a frequent psychiatric disorder with 2.3% of respondents meeting full OCD DSM-IV criteria for lifetime and 1.2% for 12-month prevalence [2]. Similar prevalence rates of 0.1 to 2.3% have been reported in European countries [3]. In up to 50% of the cases, OCD emerges already during childhood or adolescence [4], and a bimodal age of onset, during early puberty or in the early twenties [5], has been described. The definition of “early-onset” OCD (EO-OCD) differs among studies regarding the cut-off for age of onset; below 7 years of age was the lowest and under 18 years of age was the highest cut-off age, which thus marks the “minimal” consensus for definition criteria. Accordingly, patients with “late onset” (LO) must be older than 18 years at onset. EO-OCD may represent a more severe or more “biological” subtype of the disorder in comparison to LO [6], with an increased heritability as well as differences in gender distribution, symptomatology, and comorbidity. In addition, in OCD with early and childhood onset, triggering life events seem to be observed to a lesser extent [2, 7]. Therefore, it has been hypothesized that EO-OCD represents a specific highly heritable subtype of OCD [8].

Despite the strong heritability estimates between 40 and 80% [9,10,11], two recent genome-wide studies [12, 13] failed to detect association of common single nucleotide polymorphisms (SNPs) with OCD. However, in a meta-analysis [14] an association between OCD and polymorphisms in the serotonin-related genes (5-HTTLPR/ SLC6A4 and HTR2A) and in the catecholamine modulation genes (COMT and MAOA) have been reported. Similarly, we showed for the first time the association of 5-HTTLPR polymorphisms with EO-OCD [15, 16] as well as association to HTR2A polymorphism [17].

In order to further explain the heritability, it has been suggested that copy number variations (CNVs) play an important role in neurodevelopmental disorders [18,19,20]. There is rising awareness that common variants may explain a relatively small amount of common diseases, while rare variants with relatively large functional effects may add up to a significant disease contribution [21].

Up to now, five CNV studies in patients with OCD have been conducted [22,23,24,25,26]. The first study assessed the recurrent 15q11-q13 Prader-Willi syndrome and the 22q11.2 DiGeorge syndrome regions [22] in adults with OCD, since both disorders are often accompanied by OCD symptoms. However, in 236 OCD-patients no CNVs affecting the two regions were detected [22]. In a study, conducted by our group, an association between a ~100 base pairs (bp) deletion in the HTR2A promoter region and paediatric EO-OCD [23], as well as increased OCD-severity and an earlier age-of-onset, were reported. Recently, the first genome-wide analysis in a cohort of 1613 OCD and 1086 Tourette syndrome (TS) patients was performed. The study focused on CNVs larger than 500 kb and revealed aberrations in 16p13.11, 22q11.21 and in 6q25.2-q27 [24]. Additionally, a study on CNVs smaller than 500 kb screened 16 adults with EO-OCD and 12 controls [25] and a rare intragenic FMN1 microdeletion in 15q13.3 was detected. In a very recent study 307 EO-OCD (259 of European ancestry) have been investigated for small and rare CNVs [26]. Interestingly, the authors found CNVs in genes involved in neuronal migration, synapse formation and postsynaptic scaffolding, which might be relevant to the pathogenesis of OCD. Four cases had CNVs involving known genomic disorder loci (1q21.1, 15q11.2-q13.1, 16p13.11 and 17p12). [26].

In light of these current findings and the notion of a stronger genetic contribution in patients with EO-OCD, we conducted a genome-wide CNVs analysis, using high-resolution chromosomal microarray analysis (CMA) in 121 homogenous phenotyped paediatric EO-OCD patients.

Methods

Paediatric patients with OCD

121 paediatric patients with EO-OCD were recruited from the Departments of Child and Adolescent Psychiatry at the University of Würzburg, Germany and the University of Zürich, Switzerland (all Caucasian). The mean age ± SD of the patients was 12.99 ± 2.8; mean age of onset = 10.43 ± 3.2; 75 were males and 46 females. Although an EO-OCD sample could in principle also consist of adult patients with a retrospectively analysed onset, the present study includes only paediatric patients that guaranteed an early-onset of the disorder.

All patients fulfilled the diagnostic criteria for OCD according to the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) [27] and the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) [28]. Patients and parents from Würzburg were interviewed separately by senior clinicians with a semi-structured diagnostic interview of psychiatric disorders in children and adolescents (Kinder-DIPS; children and parents version) [29]. The patients and parents of Zürich underwent the German version of a semi-structured clinical interview (K-SADS-PL) [30] to investigate their phenotype, including psychiatric comorbidities. Autistic spectrum disorders were screened within an ascertainment for psychopathology (CASCAP-D) [31].

The Children’s Yale Brown Obsessive Compulsive Scale (CY-BOCS) [32] was used in all patients to assess severity and characteristics of OCD symptomatology. A summary score above 16 points was determined to be the cut-off for clinical impairment caused by OCD. In the present patients’ sample, the mean CY-BOCS score was 22.1, SD = 6.98. The DSM-IV and DSM-5 [1] do not require children to fulfil the criterion of insight into the irrationality of their symptoms. Therefore, it is possible that the CY-BOCS scores underestimated the severity of OCD in cases of reduced insight into the disorder. For the assessments of TS and tic-disorders, the German version [33] of the Child and Adult Schedule for Tourette and Other Behavioural Syndromes (STOBS) [34] was used in the Würzburg patients, the Yale Global Tic Severity Scale (YGTSS) [35] in the Zürich patients.

Patients with comorbid disorders (n = 75 /121) were included in the study when the OCD was the primary diagnosis, as assessed by two senior clinicians independently (see Additional file 1: Table S1).

Exclusion criteria were: lifetime history of psychotic disorders, Tourette’s disorder (TS), autistic spectrum disorder (ASD), alcohol dependence, intellectual disability (ID; IQ < 70). IQ was assessed by the Wechsler Intelligence Scale (WISC) [36] in the majority of cases (over 85%) while in few cases (less than 15%) the IQ was assessed by Culture Fair Intelligence Test (CFT) [37]. Since IQ < 70 was used only for exclusion criteria to avoid inclusion of any ID probands, using two scales do not affect the paediatric OCD population selection. Mean IQ score in our patient cohort was 108.27 (SD = 13.29).

The study was approved by the respective local ethics committees with the latest version of the Declaration of Helsinki, including an ethical permission granted by the Ethic Committees from Würzburg and the Cantonal Ethic Commission of Zürich (Ref. Nr. 39/97, 140/3 and EK: KEK-ZH-Nr. 2010–0340/3) and written informed consent was obtained in all cases from the participants or/and their parents.

Control sample for statistical comparison

We analysed the data of 124 random population samples of Caucasian origin (DNA was extracted from native tissue), recruited at the Institute of Medical Genetics, University of Zurich, in the same manner, on the Cytoscan HD Array as the patient samples (76 males and 48 females, X2 = 0.012, p > 0.999 compared to EO-OCD).

DNA extraction and chromosomal microarray analysis (CMA)

Constitutional genomic DNA was extracted from whole blood (EDTA-tubes; n = 74 cases and family) with the desalting Proteinase K methodology [38] or from saliva samples (Oragene DNA, DNA Genotek Inc., Ontario, Canada; n = 47 cases and family) using the manufacturer’s protocol. To assess genomic DNA purity, the ratio of absorbance at 260 nm and 280 nm was used. 74/121 cases were part of the previous reported genome-wide CNV (>500 kb) analysis [24], in which only one large CNV was reported and confirmed in the current study (patient #9025079001 7q21.11 deletion), without any previous findings that were not confirmed in the current study. DNA from available parents and/or siblings of 30 patients was investigated consecutively due to CNV findings in the patients.

DNA was analysed with the Cytoscan HD Array (containing about 750,000 SNPs and 1.9 million non-polymorphic probes) (Affymetrix Inc., Santa Clara, CA, USA) at a genome-wide resolution of 50 kb for both duplications and deletions. We have chosen this resolution in order to minimize the possibility of false positive CNVs, which according to our previous assessment have a median size of 19 kb [20]. Array hybridization and quality control was performed according to the manufacturer’s protocol. Data were analysed with Chromosome Analysis Suite (ChAS) software (Affymetrix) for changes of relative intensities. The CNV analysis was based on build 32.1. Genomic coordinates are based on GRCh37/hg19. In order to exclude common benign CNVs, we used a reference set of 820 in-house healthy controls and 1038 Affymetrix controls in combination with the Database of Genomic Variants (DGV) from the Centre for Applied Genomics (February 2009, hg19).

CNVs including coding sequences of genes that were absent in our in-house and Affymetrix primary control cohort and either unreported or reported only from limited sources in the DGV database were defined as rare CNVs in both the EO-OCD and the random control cohorts and used for statistical comparison.

Statistical analysis and enrichment analysis

Frequency analysis was conducted using Χ2 test and the power was calculated using G*Power v.3.1.9.2 [39]. For continuous measures, the two-tailed t-test was used.

The genes within the rare CNVs identified in the patients and controls were investigated for their enrichment in pathways and functional groups using the Enrichment Analysis of Selected Entities (EASE) tool, based on Gene Set Enrichment Analysis (GSEA) method [40, 41], with the Pathway Studio software v.11.2.04 (Elsevier; Mammal database). Significant enriched pathways were considered with p < 0.005. As additional validation, we used the Database for Annotation, Visualization and Integration Discovery (DAVID) [42] for functional annotation clustering. The genes within CNVs were entered using the identifier “Official_gene_symbol” to create gene lists for EO-OCD and for control samples. We reported clusters with enrichment score > 1.3 as suggested by Huang da W et al. [42]. Significant enriched pathways were considered with p < 0.05 following Bonferroni, Benjamini and FDR corrections. In addition, as a third confirmation analysis, we performed a CNV enrichment analysis using PLINK [43] version 1.07 and its --cnv-enrichment-test option on all CNVs called [44]. Following multiple testing correction (6 tests) p < 0.008 was considered as significant, and nominal significant for p < 0.05.

Statistical analysis was performed with SPSS v.21 (IBM) and StatView v.5.0 (SAS Inst.). The level of significance was alpha = 0.05.

Results

We detected rare CNVs (mean size = 258 ± 308 kb, min. 50 kb max. 1500 kb) in 41 out of 121 patients (Table 1; for details see Additional file 2: Table S2a) and in 39 out of 124 controls (mean size = 274 ± 234 kb, min. 50 kb max. 1027 kb; for details see Additional file 2: Table S2b). There was no significant difference in the number of all detected rare CNVs between patients (n = 41/121; 34%) and controls (n = 39/124; 31%; Χ2 = 0.165, p = 0.685) nor in the CNVs’ size (t(−74.47) = 0.255, p = 0.8). Moreover, no significant difference was observed in IQ scores or OCD severity (measured with CY-BOCS) between carriers and non-carriers of rare CNVs.

Table 1 Rare CNVs discovered in the paediatric obsessive compulsive disorder (EO-OCD) patients including inheritance pattern
Full size table

However, we observed a significantly higher number of CNV carrying brain/synaptic genes in the patients in comparison to the controls (Χ2 = 4.225, p = 0.0398; odds ratio (OR) = 1.98, 95% CI 1.02–3.84; Power 1-β=0.858). In particular, the number of rare deletion carrying brain/synaptic genes was significantly higher in patients (n = 13 vs. 108 non-carriers) compared to controls (n = 4 vs. 120 non-carriers; Χ2 = 5.360, p = 0.021; OR = 3.611, 95% CI 1.14–11.41; Power 1-β=0.8909).

In line with this finding, both enrichment cluster analyses (Pathway Studio and DAVID), revealed a higher number of synaptic and brain related functional pathways in the patients (10 brain/synapse clusters, 103 none brain/synapse clusters in Pathway studio; 6 brain/synapse clusters, 31 none brain/synapse clusters in DAVID) in comparison to the controls (2 brain/synapse clusters, 98 none brain/synapse clusters in Pathway studio; 0 brain/synapse clusters, 25 none brain/synapse clusters in DAVID; Χ2 = 4.682, p = 0.03 for Pathway Studio; Χ2 = 4.488, p = 0.033 for DAVID Additional file 3: Table S3a & b).

Both enrichment cluster analyses resulted in similar top clusters for the patients group (KRAB domain, zinc-finger protein family) and for the control group (serum amyloid A, high-density lipoprotein), confirming the integrity of both analyses. The most significant (p < 0.0001) GO functional group enrichment cluster in the patient cohort belong to: axon guidance (genes within the CNVs: RAF1, EPHB2, MYH11, SEMA3A, ITGA10, HFE2, NRXN1, COL4A1, COL4A2), axonal fasciculation (genes within the CNVs: EPHB2, NCAM2, SEMA3A), synapse (genes within the CNVs: EPHB2, GABRA4, GABRB1, NRXN1, MYO6, PIAS3, ANKS1B, PACSIN1) and neuron cell-cell adhesion (genes within the CNVs: NCAM2, NRXN1), as analysed by Pathway Studio (for details see Additional file 3: Table S3a). Similarly, DAVID analysis revealed in the patients several GO related to synaptic and brain related genes, like in cluster 5: axonogenesis, cell morphogenesis involved in neuron differentiation, and neuron projection morphogenesis (gene cluster: NCAM2, PARD3, NRXN1, SEMA3A, EPHB2) (for details see Additional file 3: Table S3b). Interestingly, the control group did not show any synaptic or brain related clusters when the DAVID software was used, while only 2 clusters (p < 0.005) were found using Pathway Studio analysis: catecholamine metabolic process (genes within the CNVs: SULT1A3, SULT1A4) and axonal fasciculation (genes within the CNVs: CNTN4, NRCAM; for details see Additional file 3: Table S3a).

We could further confirm a significant enrichment (p = 1.28 x 10−3) of CNVs in brain expressed genes by applying the PLINK enrichment analysis [44]. Notably, there was no significance for the enrichment analysis for overall genes. The significance threshold was 8.33 x 10−3 (0.05/6, because 6 tests were performed) (Additional file 4: Table S4).

Where possible, the heritability of rare CNVs was assessed by analysing the parents/siblings of the index patients. We detected two de-novo CNVs in two patients with EO-OCD (one male and one female), both of whom had a comorbid tic disorder (Table1 & Additional file 2: Table S2a). Twenty-six patients inherited the CNVs (16 from the mother and 10 from the father). In two additional patients (9025100001 & 9925022001) CNVs were not maternally inherited (paternal DNA was unavailable). Three patients carried CNVs located on the X-chromosome: two deletions (affecting the genes UBE2NL and MTMR8) and one duplication (affecting the genes KDM6A and CXorf36). Interestingly, both deletions are hemizygous aberrations.

Discussion

We did not detect a higher number of rare CNVs in paediatric patients with EO-OCD compared to controls, which is in line with previous publications [24, 26]. However, cluster analyses of the gene content of the rare CNVs revealed a significantly higher number of genes involved in synaptic and brain related pathways in the cases compared to controls, similarly to previous findings [24, 26]. Our results, therefore, further support the hypothesis that the aetiopathology of EO-OCD may be related to neurodevelopmental processes [9, 45]. In addition, we demonstrate that the OR of 3.6 (p = 0.021) for EO-OCD in carriers of rare small deletions is in the same range as that reported previously for larger deletions (OR = 4.4, p = 0.04) [24]. Therefore, as Gazzellone et al. [26] recently postulated, not only large CNVs >500 kb but also smaller rare CNVs (≥50 kb), particularly deletions in brain related genes, might represent a risk factor for paediatric EO-OCD.

We observed two large deletions (>500 Kb) in our cohort. The first one was a 1.5 Mb deletion in 16p13.11 with undeterminable inheritance in a male patient. The second was a 731 kb deletion in 7q21.11 (finding of this patient previously reported [24]) affecting the SEMA3A gene, which was inherited from the healthy father. While SEMA3A has not been described to be associated with human psychiatric disorders so far, the 16p13.11 deletion has previously been found to be associated with OCD [24] and with a variety of neurodevelopmental disorders, such as ASD, ID, epilepsy, and schizophrenia [46, 47], with reduced and male-biased penetrance [48]. Furthermore, a maternally inherited duplication overlapping the 16p13.11 was reported in an EO-OCD patient [26]. Therefore, the aforementioned findings support the possible causality of this aberration in OCD.

We detected another variant in a recurrent CNV locus in a male patient with EO-OCD and comorbid tics, who carries a 437 kb duplication encompassing the region 1q21.1. The duplication was inherited from the mother and present in the sister, both according to the CY-BOCS presenting sub-threshold OCD with similar phenotype including ordering behaviours. Notably, a duplication of that region has also been detected in a paediatric EO-OCD patient and his unaffected mother and dizygotic twin brother in the recently published cohort of Gazzellone et al. [26]. We, therefore, suggest that CNVs on the chromosome region 1q21.1, which have been reported to associate with congenital heart defects, developmental delay, ASD, and psychosis [49, 50], represent also a susceptibility locus for OCD. These phenotypes are subject to incomplete penetrance and variable expressivity, since in most of the cases they are inherited from apparently healthy parents [51].

We found several other smaller rare CNVs encompassing interesting candidate genes that have already been described in neurodevelopmental disorders, for which, however, additional evidence is needed to prove the possible relation with EO-OCD. One of these findings is the 258 kb duplication encompassing the gene GRM8, coding for a glutamate receptor, in 7q31.33, paternally inherited, observed in a female patient with EO-OCD and comorbid tics, but not in her sister with EO-OCD. CNVs encompassing the gene GRM8 have been reported in patients with attention-deficit hyperactivity disorder (ADHD) [52] and with developmental delay, hypotonia, and strabismus [20]. Recent evidence has highlighted the role of glutamatergic synaptic dysfunctions in the cortico-striatal-thalamo-cortical circuit in the aetiology of OCD and related disorders [9, 53, 54].

A further noteworthy CNV, 113 kb deletion encompassing the two genes GABRA4 and GABRB1 in 4p12, was found in a female patient of maternal origin that carried also a 66 kb duplication (EMILIN2 gene in 18p11.32). Involvement of GABA receptor subunit genes in the aetiology of autism has been reported [55,56,57]. Both GABRA4 and GABRB1 mRNA and protein were found to alter their expression in the parietal and frontal cortex and cerebellum of patients with ASD and in the lateral cerebellum of patients with schizophrenia and with affective disorders [58, 59]. In peripheral blood samples of patients with TS, mRNA expression of GABA receptors including GABRA4 correlated with tic severity. Furthermore, there was indication of GABRA4 being alternatively spliced in TS compared to healthy controls [60]. For the duplication, little is known about the EMILIN2 gene and neurodevelopment. However, patients carrying large deletions in 18p11.32 were reported to have developmental delay and mental retardation [61] while linkage study could show some evidence for schizophrenia susceptibility near this region [62].

A 279 kb deletion in Xq11.2 encompassing the gene MTMR8 was found in a patient with EO-OCD as well as in his mother and brother, who both suffer from anxiety disorder. Anxiety is very often a core feature of OCD symptomatology and in the DSM-IV [27], OCD was even classified among the anxiety disorders, which shows the close relationship between the two diseases. MTMR8 encodes a phosphatidylinositol kinase and reduced protein expression is associated with impaired survival of specific neuronal populations [63]. Loss-of-function mutations occurring in the phosphatidylinositol kinase gene family are known to cause different X-linked neurological diseases, including schizophrenia, bipolar disorder and age-related neurodegeneration, probably due to endosomal trafficking defects and accumulations of the lipid substrates [64, 65].

A de-novo deletion of 51 kb in 2p16.3 was detected in a male diagnosed with very EO-OCD (4 years old at age of onset) who suffered also from comorbid tics and hyperkinetic symptoms. The deletion encompasses the NRXN1 gene, for which an association with TS, ASD, ID, and schizophrenia has been described [24, 46, 66, 67]. Interestingly, the NRXN1-α knock-out mouse model supports the role of NRXN1 in neurodevelopmental disorders, since these mice displayed non-social behaviour as well as hyperactivity and learning deficits [68].

In another patient with comorbid tics we found a de-novo deletion of 310 kb in 12q23.1 harbouring the genes ANKS1B and UHRF1BP1 and a duplication of 80 kb in 19p13.11 harbouring the genes GMIP, ATP13A1, ZNF101 and ZNF14. ANKS1B is predominantly expressed in the brain and known to interact with the amyloid beta protein precursor that may play a role in normal brain development and in the pathogenesis of Alzheimer’s disease [69, 70]. CNVs affecting ANKS1B have been reported in ASD and ID including a de-novo deletion in a male with ASD and delayed early language development but average language abilities and IQ [47, 71]. In addition, a 190 kb deletion encompassing the ANKS1B gene was reported in a patient with EO-OCD [26], enhancing the possible role of the gene in neurodevelopmental disorders. Within the duplication, GMIP, coding for RhoA-specific GTPase-activating protein, was reported to be a key factor for neuronal migration in the postnatal brain [72] as well as regulating vesicular trafficking [73]. Moreover, in a large population-based twin-family study exploring genome-wide association of obsessive-compulsive symptoms, a significant association in 4 genes (MEF2BNB, RFXANK, MEF2BNB-MEF2B and MEF2B) located in proximity to GMIP gene was reported [74].

Conclusions

Our findings further support the role of rare CNVs in the aetiology of OCD and emphasize the role of rare small deletions encompassing brain genes as potential susceptibility factors in the aetiology of paediatric EO-OCD. Further studies are necessary to confirm the contribution of the individual variants in OCD.