De novo copy number variants associated with intellectual disability have a paternal origin and age bias
- Jayne Y Hehir-Kwa1,
- Benjamín Rodríguez-Santiago1,2,
- Lisenka E Vissers1,
- Nicole de Leeuw1,
- Rolph Pfundt1,
- Jan K Buitelaar3,
- Luis A Pérez-Jurado2,
- Joris A Veltman1
- 1Department of Human Genetics, Nijmegen Centre for Molecular Life Sciences and Institute for Genetic and Metabolic Disorders, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
- 2Unitat de Genètica, Universitat Pompeu Fabra, and Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER), Barcelona, Spain
- 3Department of Cognitive Neuroscience, Donders Institute for Brain Cognition and Behavior, Radboud University Nijmegen, The Netherlands
- Correspondence to Dr Jayne Y Hehir-Kwa, Department of Human Genetics-848, Radboud University Nijmegen Medical Centre, PO Box 9101, Nijmegen, The Netherlands;
Contributors JHK and BRS contributed equally to this work. JHK developed the algorithms used in this study. JHK, BRS, and LV performed the analysis, revised and drafted the manuscript. JKB collected and provided the control cohort. RP and NL analysed the microarray data, and LPJ and JV initiated the collaboration that led to this study and contributed to the manuscript.
- Received 26 April 2011
- Revised 27 July 2011
- Accepted 6 August 2011
- Published Online First 3 October 2011
Background De novo mutations and structural rearrangements are a common cause of intellectual disability (ID) and other disorders with reduced or null reproductive fitness. Insight into the genomic and environmental factors predisposing to the generation of these de novo events is therefore of significant clinical importance.
Methods This study used information from single nucleotide polymorphism microarrays to determine the parent-of-origin of 118 rare de novo copy number variations (CNVs) detected in a cohort of 3443 patients with ID.
Results The large majority of these CNVs (76%, p=1.14×10−8) originated on the paternal allele. This paternal bias was independent of CNV length and CNV type. Interestingly, the paternal bias was less pronounced for CNVs flanked by segmental duplications (64%), suggesting that molecular mechanisms involved in the formation of rare de novo CNVs may be dependent on the parent-of-origin. In addition, a significantly increased paternal age was only observed for those CNVs which were not flanked by segmental duplications (p=0.02).
Conclusion This indicates that rare de novo CNVs are increasingly being generated with advanced paternal age by replication based mechanisms during spermatogenesis.
The widespread use of high resolution genomic microarrays and whole exome sequencing has clearly demonstrated that rare de novo mutations at various genomic levels are a common cause of disease, particularly diseases that are associated with reduced reproductive fitness. It is therefore of the utmost importance to study the generation of these mutations and factors influencing their formation. Parental age and gender are known to affect both base substitutions as well as chromosomal aneuploidy.1–4 Base substitution mutations occur more frequently in males and more frequently with advanced age, which can be explained by the greater number of pre-meiotic cell divisions in males.5 In contrast, the majority of trisomic offspring is maternal in origin resulting from non-disjunction of homologous chromosomes during the first meiotic division, with also an increase in frequency with age.4 Until now, little has been known about the parental origin and the effect of increased parental age on the genesis of genomic copy number variation (CNV), including those which underlie a significant percentage of patients with cognitive disorders. Studies on a specific class of de novo CNVs—namely, those associated with recurrent microdeletion and microduplication syndromes—have indicated that there is no significant bias in the parent of origin6 or have only included a small number of individuals.7–9 Recurrent de novo CNVs are often flanked by segmental duplications that mediate the generation of these rearrangements through non-allelic homologous recombination (NAHR),10 which may be indifferent during both paternal and maternal meiosis. The majority of rare de novo CNVs associated with intellectual disability (ID), however, have non-recurrent breakpoints and do not generally involve known syndromes.11 Here, we report on a systematic parent-of-origin analysis of rare de novo CNVs identified in a large diagnostic cohort of individuals with ID. In addition, we assessed whether an elevated parental age may have influenced the generation of CNVs.
Rare de novo CNVs were detected by routine diagnostic screening of 3443 patients with ID between 2006 and 2010 in the department of human genetics, Radboud University Nijmegen Medical Centre. The consent of patients to perform this study was obtained. A total of 227 index patients showed one or more rare de novo CNVs after exclusion of large abnormalities such as derivative, ring, translocation, mosaic, or whole chromosome aberrations (including XYY). The overall process summarising the rare de novo CNV identification is illustrated in supplemental figure S1. Briefly, a 250 K single nucleotide polymorphism (SNP) mapping array (Affymetrix, Santa Clara, California, USA) was applied for CNV discovery using DNA extracted from whole blood using standard procedures. All microarrays were analysed in comparison to a pool of healthy reference samples. CNV losses larger than 150 kb and CNV gains larger than 200 kb spanning at least seven microarray SNP probes were included as previously described,9 with a calculated false negative rate of 0.95. A CNV was considered rare when present less than two times in the in-house database and less than three times in the Database of Genomic Variants.9 In all cases parental samples were studied using the same array technology.
The genome-wide mutation rate for these rare CNVs is 3.6×10−2 in this cohort of ID patients (227 rearrangements in 6258 transmissions; supplementary figure S1). This is significantly higher (p<10−4) than similar estimates obtained in a recent study on asthma patients (1.2×10−2) as well as autism patients (1.8×10−2 (supplementary table S1).12 This indicates that de novo CNVs are enriched in ID compared to asthma as well as autism, which is also shown by the odds ratio of 3.2 and 2.2 in comparison with these cohorts, respectively. This in line with the recent observation that de novo point mutations are also a common cause of ID13 and, together with de novo CNVs, explain why this disorder, associated with severely reduced fecundity, remains so common in the human population.
In the set of 227 rare de novo CNVs non-paternity was excluded as previously reported.9 Furthermore the SNP data on the array was used to identify any large (>10 Mb) regions of homozygosity, suggestive of uniparental disomy. None was found in the set of 227 rare de novo CNVs. Complete trio array data were available for 167 trios, with 154 occurring on autosomal chromosomes. The array data did not provide informative genotype SNP calls for 36 CNVs. These 36 CNVs were on average smaller, represented by fewer SNPs on the array, and were more often copy number gains. No difference was observed in the mean SNP QC call rate when comparing CNVs with determined and undetermined origin (supplemental table S2). We only considered autosomal rearrangements for further parent-of-origin analysis, as X chromosome CNVs associated with ID are over-represented in males. The parent-of-origin was conclusively determined by comparison of the informative SNP calls in the patient to the parents for 118 autosomal CNVs discovered in 108 patients (supplemental figure S1 and Supplemental table S3). For the CNV losses, this comparison indicates which allele is remaining, thus indicating that the other parent is the transmitting one. For CNV gains, the heterozygous genotype calls are combined with the allele contrast to determine the duplicated allele. Analysis of the informative genotype calls in the parents subsequently identifies who transmitted the duplicated allele.9 Using these two approaches, the parent-of-origin was determined for 118 autosomal rare de novo CNVs, as 36 CNVs did not encompass informative SNPs and 13 CNVs occurred on chromosome X (supplementary figure S1, supplementary table S2).
A significant paternal bias was observed with 90 of 118 rare de novo CNVs occurring in the paternal germ line (76% paternal origin, p=1.14×10−8, figure 1). Of the 90 paternal CNVs, 74% represented copy number loss and 26% a copy number gain (figure 1). Similarly, 75% of the 28 maternal origin CNVs represented copy number loss and 25% a copy number gain. Similar results were obtained when examining CNV size: 28% of paternal and 25% of maternal CNVs were smaller than 1 Mb in size, and the remainder was larger than 1 Mb (figure 1). These data indicate that there is no parental origin effect dependent on either the size or type of rare de novo CNV in our series.
Next, we identified the rare de novo CNVs with segmental duplications (>1 kb in size with more than 90% identity) in the direct vicinity of the CNV breakpoints (a flanking window size of 500 kb was used). Segmental duplications are important mediators of non-allelic homologous recombination (NAHR), a mechanism predominantly occurring in meiosis.10 14 In the absence of segmental duplications, non-homologous end joining (NHEJ) or microhomology mediated break induced repair (MMBIR) are the most likely mechanisms for CNV formation.10 15 16 Of 118 rare de novo CNVs, 25 CNVs showed characteristic segmental duplications at their breakpoints suggestive for the CNV being mediated by NAHR (figure 1 and supplementary figure S1). The proportion of CNVs not flanked by segmental duplications was twofold higher in paternal CNVs compared with maternal CNVs (figure 1), indicating that the paternal bias for CNV generation can largely be explained by an increase of replication based mechanisms such as NHEJ and MMBIR. These paternally derived rare de novo CNVs are mostly non-recurrent and are scattered throughout the genome. Our finding that paternally originating rare de novo CNVs occur more frequently is consistent with previous studies analysing larger chromosomal abnormalities,4 and single base pair mutations.1–3 The assessment of parental origin of chromosome abnormalities and parental age effects in genome-wide microarray based copy number has been performed in a small number of studies and have provided contrasting results.12 17–19 This discrepancy in previous studies on parent-of-origin might be explained by differences in material source, sample size studied, or in the frequency of NAHR mediated CNVs contained within the cohort being studied.
To study the presence of paternal age effects in relation to CNV formation we compared the paternal age at childbirth with a Dutch control cohort, consisting of 597 healthy trios.20 We matched this control cohort for ethnicity as well as for the era in which the children were born to minimise age effects depending on social and cultural differences. The average maternal and paternal age at childbirth in our series of rare de novo CNVs was 30.6 and 33.6 years, respectively (supplemental table S3). These ages were similar to parental ages at childbirth in our control cohort (31.4 and 32.1 years, respectively) as well as to those reported in literature.21 An analysis of variance (ANOVA) test was performed with post-hoc multiple comparison testing and Bonferroni correction. The paternal age of the ID cohort without rare de novo CNVs did not differ significantly from the mean paternal age of the control cohort. Similarly no increase in paternal age was seen for offspring with a rare de novo CNV when the breakpoints of the de novo CNVs were flanked by segmental duplications (p=1.00). Interestingly, a significant increase in paternal age was observed for de novo CNVs without segmental duplications in the direct vicinity of their breakpoints (p=0.02) (table 1). This indicates that increased paternal age has a major impact on the generation of CNVs with non-recurrent breakpoints.
In conclusion, our data provide for the first time convincing evidence that CNVs in ID are largely paternal in origin, similar to what has been described for single base pair mutations1 and large cytogenetically visible chromosomal aberrations.4 Both the paternal bias as well as the age effect that we observed can be explained by the continuation of cell divisions of self-renewing spermatogonia in males, whereas oogonia cease replication during fetal life after approximately 30 cell generations.22 The reduced fidelity of DNA replication and the inefficiency of repair mechanisms, both expected to increase with age,23 are also in concordance with these findings. The association between increased paternal age and non-NAHR events furthermore suggests that replication based mechanisms, such as NHEJ or MMBIR,15 play a major role in the generation of rare de novo CNVs in ID.
We thank Alejandro Arias-Vasquez for his useful discussions during the preparation of this manuscript. This work was supported by grants from the AnEUploidy project (EU-6FP, LSHG-CT-2006-037627) (to LAP-J and JAV), and the Netherlands Organization for Health Research and Development (ZonMW grants 916-86-016 to LELMV and 917-66-363 to JAV). B Rodríguez-Santiago was supported by a postdoctoral fellowship of the Fondo Investigación Sanitaria, Spain (FIS CD06/00019).
JY H-K and B R-S shared first authorship on this paper.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.