Background In a subset of imprinting disorders caused by epimutations, multiple imprinted loci are affected. Familial occurrence of multilocus imprinting disorders is rare.
Purpose/objective We have investigated the clinical and molecular features of a familial DNA-methylation disorder.
Methods Tissues of affected individuals and blood samples of family members were investigated by conventional and molecular karyotyping. Sanger sequencing and RT-PCR of imprinting-associated genes (NLRP2, NLRP7, ZFP57, KHDC3L, DNMT1o), exome sequencing and locus-specific, array-based and genome-wide technologies to determine DNA-methylation were performed.
Results In three offspring of a healthy couple, we observed prenatal onset of severe growth retardation and dysmorphism associated with altered DNA-methylation at paternally and maternally imprinted loci. Array-based analyses in various tissues of the offspring identified the DNA-methylation of 2.1% of the genes in the genome to be recurrently altered. Despite significant enrichment of imprinted genes (OR 9.49), altered DNA-methylation predominately (90.2%) affected genes not known to be imprinted. Sequencing of genes known to cause comparable conditions and exome sequencing in affected individuals and their ancestors did not unambiguously point to a causative gene.
Conclusions The family presented herein suggests the existence of a familial disorder of DNA-methylation affecting imprinted but also not imprinted gene loci potentially caused by a maternal effect mutation in a hitherto not identified gene.
- Clinical genetics
- hypomethylation syndrome
Statistics from Altmetric.com
Genomic imprinting leads to parent-of-origin-specific DNA-methylation and gene expression.1 Imprinting defects in humans contribute to several recognisable syndromes such as Beckwith–Wiedemann (BWS), Silver–Russell (SRS), Prader–Willi (PWS) or Angelman (AS) syndrome. These imprinting disorders show DNA-methylation changes at the disease-specific imprinted locus; however, more recent studies suggest that a subset of individuals with imprinting disorders shows changes of DNA-methylation at multiple imprinted loci. These multilocus imprinting disorders seem to be particularly prevalent in individuals with BWS, SRS and transient neonatal diabetes mellitus, the latter has been associated with a ‘maternal hypomethylation syndrome’.2–4
Familial occurrence of true imprinting defects, that is, changes of DNA-methylation at imprinted loci without causative mutations at the imprinted locus itself in cis, is rare. An autosomal-recessive trait has been described in a subset of the ‘maternal hypomethylation syndromes’, in which affected individuals carry biallelic mutations in the ZFP57 gene.5 An alternative mechanism leading to familial occurrence of imprinting defects is the presence of ‘maternal effect mutations’ in genes important for establishing or maintaining genomic imprints in early development. Indeed, two evolutionary closely related genes located head-to-head in chromosomal region 19q13, NLRP2 and NLRP7, have been associated with such ‘maternal effect mutations’. A homozygous mutation in NLRP2 has been described in a woman giving birth to siblings with BWS due to an imprinting defect in 11p15 and partial loss of methylation in PEG1 in one child.6 In women with biallelic mutations of NLRP7, pregnancies completely fail to develop properly but instead result in hydatidiform moles, which also show alterations at multiple imprinted loci (for an overview, see ref.7). Recently, in women with familial biparental hydatidiform moles without NLRP7 mutations, changes in KHDC3L were observed.8–10
Here we describe a family in which two fetuses and one child of a couple showed altered DNA-methylation patterns not only of maternally and paternally imprinted genes but also of various other genes, suggesting a more widespread disorder of DNA-methylation.
Patients and methods
We report on two fetuses (III-1, III-3) and one child (III-2) of a healthy non-consanguineous Turkish couple (II-3, II-4). Family histories were unremarkable on both sides. In the first pregnancy (III-1), omphalocele and shortened femora were noticed at 21 weeks of gestation. At 33 weeks, hypoplastic thorax and clover leaf skull were noted and the pregnancy was terminated. On pathological examination marked lung hypoplasia was confirmed and abnormal lung lobulation, gall bladder agenesis, hydronephrosis and further abnormalities were noted. As X-rays showed no signs of skeletal dysplasia or craniosynostosis, the tentative diagnosis of SRS was suggested. At that time only maternal UPD7 was known as the cause for SRS, but microsatellite analyses on cultured amniocytes ruled this out. The second pregnancy (III-2) was complicated by early diagnosis of molar changes in the placenta, asymmetrical fetal growth restriction, omphalocele and massively elevated β-human chorionic gonadotropin (β-HCG). The child was born at 32 weeks. After birth coarse facial features, facial haemangioma, omphalocele and asymmetric growth restriction were suggestive both of SRS and BWS, which prompted methylation studies. Postnatally body proportions harmonised. The child is developmentally delayed. In the third pregnancy (III-3), early asymmetric growth retardation, elevated β-HCG and molar changes of the placenta were noted. The pregnancy ended by spontaneous fetal demise 1 week after chorionic villi sampling at 12 weeks. For further details, see figure 1A, online supplementary results, table S1 and figure S1.
Various specimens of the affected individuals (III-1, III-2, III-3) and peripheral blood samples of further family members were investigated (see online supplementary table S2). Detailed information on analysed materials and used methods (including PCR conditions and primer sequences) are provided in the online supplementary appendix. The study has been approved by the Institutional Review Board of the Medical Faculty of the Christian-Albrechts University Kiel (AZ B305/08). All family members investigated and for minors the respective parents gave written informed consent for their participation.
Karyotyping, FISH and molecular karyotyping were performed according to standard techniques and manufacturers’ instructions.
Mutation analysis of NLRP2 (NM_001174081), NLRP7 (NM_001127255), ZFP57 (NM_001109809), KHDC3L (NM_001017361) and of the oocyte-specific variant of DNMT1 (NM_001130823) was performed by Sanger sequencing. Segregation of NLRP7/NLRP2 alleles was verified by microsatellite analysis on chromosome 19.
Exome enrichment using the NimbleGen Human SeqCap EZ V.3.0 Kit followed by sequencing on an Illumina HiSeq 2000 system and data analysis was carried out on DNA from two affected children (III-1 and III-2), the parents (II-3 and II-4) and the maternal grandparents (I-3, I-4) as detailed in the online supplementary appendix.
Locus-specific DNA-methylation analysis was performed using methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA), methylation-specific PCR, sequence-based quantitative methylation analysis and bisulfite pyrosequencing. Global DNA-methylation was analysed using LUminometric Methylation Assay (LUMA). For array-based DNA-methylation quantification of 27 578 CpG sites, the HumanMethylation27 DNA Analysis BeadChip (Illumina) was applied. Raw hybridisation signals were analysed using GenomeStudio software (GSE47879).11 A detailed description of the bioinformatic analysis of array-based methylation data is provided in the online supplementary appendix.
Results and discussion
In order to rule out a chromosomal aberration as the cause of phenotype, we performed conventional chromosome analyses in both parents and all three offspring, molecular karyotyping on both parents, as well as individuals III-1 and III-2, and FISH analyses of chromosomes X, Y, 13, 18 and 21 on amniotic fluid samples of III-1 and III-2. All these analyses lacked evidence for any kind of chromosomal aberration as the cause of phenotype in the offspring III-1, III-2 and III-3 (see online supplementary table S3).
As several phenotypic characteristics of the affected individuals resembled features of imprinting disorders, DNA-methylation analyses of known imprinted loci were initiated. Results of locus-specific methylation analyses are shown in table 1. In all three offspring, DNA-methylation changes at multiple imprinted loci were identified. Remarkably, these affected both maternal and paternal imprints. The only hypermethylation was observed for the NESP somatic DMR, but this is most likely caused by hypomethylation of the GNAS DMR. Remarkably, DNA-methylation at the imprinted loci was not fully lost and the extent of DNA-methylation changes varied between the different individuals as well as between the tissues from the same individual (table 1). These observations strongly suggested a DNA-methylation defect to underlie the phenotype in the offspring. Moreover, the extent and variation of DNA-methylation levels at imprinted loci suggested mosaicism for the DNA-methylation changes, which could explain the phenotypic variability of the three offspring.
The DNA-methylation analyses excluded a maternal hypomethylation syndrome but instead proved that the underlying disorder affects both parental imprints.4 ,5 Such familial occurrence of a multilocus imprinting disorder affecting both paternal and maternal imprints has yet been rarely described.12 ,13
We next wondered whether the DNA-methylation changes extended over differentially methylated regions of imprinted loci. Thus, in order to rule out a global disturbance of the DNA-methylation, we performed LUMA in two of the three affected offspring and the parents. By this approach, no overall aberrations of DNA-methylation were detected (see online supplementary figure S2).
As LUMA is biased towards the determination of DNA-methylation levels in repeat regions, we additionally aimed at investigating single-copy locus DNA-methylation. To this end, we performed array-based DNA-methylation quantification of 27 578 CpG sites in the human genome. DNA-methylation levels in different accessible tissues from the affected individuals were compared with matched tissue samples from healthy controls (for details, see online supplementary appendix). In line with previous studies demonstrating the validity of the array,14 there was a good agreement of the array-based results and the above results of locus-specific analyses of imprinted loci. In order to get an initial estimation of the extent of the DNA methylation changes in the offspring, we performed a supervised comparison of all samples of the three affected individuals compared with all control samples independent of the tissue origin. By this global approach, we identified 95 CpG loci corresponding to 87 genes to be differentially methylated (t test, FDR <0.01; online supplementary figure S3 und table S4). In line with the molecular analyses described above, these included CpGs from several imprinted genes, such as GNAS, H19, KCNQ1, MEG3, NNAT and PLAGL1. Formal testing showed known or previously suggested imprinted genes to be strongly enriched among the affected genes (OR=13.90, RR=12.12, p<0.001). Moreover, GATHER maploc analysis15 revealed an enrichment for loci mapping to chromosome band 11p15 (p<0.01, Bayes factor >14.3) containing the imprinted gene cluster associated with BWS and SRS.
Obviously, DNA-methylation patterns can be tissue-specific.16 Thus, in a second approach to estimate the extent of the DNA methylation changes, we aimed at identifying CpG loci differentially methylated between corresponding tissue of affected and control individuals. The sum of tissues showing aberrant DNA-methylation between affected and control individuals was calculated for each CpG locus. In total, 287 genes accounting for 2.1% of the genes analysed by the array were affected recurrently, that is, in at least two comparisons (see online supplementary figure S4 and table S5). These 287 genes were again significantly enriched for known or previously suggested imprinted genes (OR=9.49, RR=8.17, p<0.0001). Moreover, 18 of these 287 genes overlapped with the 87 genes identified by the first approach, which included imprinted genes such as GNAS, KCNQ1, L3MBTL, MEG3, NAP1L5, NNAT, PLAGL1 and ZNF597. These 18 genes also included the RB1 gene, which we could previously prove to be imprinted based on the findings in this family.17
Both approaches concordantly showed clearly strong and significant enrichment of imprinted loci among the aberrantly methylated loci in the affected individuals of the family. However, the vast majority of the aberrantly methylated genes (75/87, 86.2% and 259/287, 90.2%, respectively) have yet not been described to be imprinted. In line with this, we could recently prove by targeted deep bisulfite sequencing that the aberrant methylation at least of some of the aberrantly methylated genes in III-2 is independent of the parental allele.11 Moreover, the aberrant methylation was not restricted to hypomethylation as we observed both increased and decreased methylation in the affected individuals compared with the controls (see online supplementary figure S3).
As the above-described findings indicated that the DNA-methylation disorder in the family presented is not restricted to imprinted loci and moreover might have features different from previously described multilocus methylation disorders, we performed in silico analysis of the aberrantly methylated genes. The aberrantly methylated genes derived from both approaches were enriched for genes with low CpG content promoters (OR=2.59, RR=2.57, p<0.001 and OR=1.89, RR=1.86, p<0.0001, respectively), while they were significantly depleted for high CpG content promoters (OR=0.40, RR=0.40, p<0.001 and OR=0.43, RR=0.44, p<0.0001) and CpG-islands (OR=0.29, RR=0.29, p<0.001 and OR=0.37, RR=0.38, p<0.0001) (figure 1B). Furthermore, the aberrantly methylated genes were significantly enriched for target genes of several transcription factors, including SMAD3 (p<0.01; Bayes factor >8.0, online supplementary table S6). Remarkably, SMAD3 has recently been shown to colocalise with CTCF to the H19 imprinting control region and has been proposed to play a role in chromatin cross-talk organised by the H19 ICR.18
Next, we aimed at investigating the mechanism that might lead to the DNA-methylation defect in the family. Besides NLRP2, NLRP7, ZFP57 and KHDC3L, which have yet been predominately linked to changes of maternal marks at imprinted DMRs with high CpG content, we considered DNMT1o as potential candidates. We failed to identify potentially pathogenic coding mutations in any of these genes in the family. Nevertheless, a heterozygous missense variant c.2156C>T (p.A719V) in NLRP7 was identified in the mother (II-4) of the affected individuals (figure 1C, online supplementary figure S5 and table S7).19 NLRP7 mutations in the sense of so-called ‘maternal effect mutations’ have been previously associated with recurrent hydatidiform moles, which show a disturbance of imprinting at multiple loci. Indeed, the very same variant p.A719V has been detected in heterozygous state by Messaed and colleagues in a woman with four spontaneous abortions, one of which led to a gestational trophoblastic disease.20 Nevertheless, this variant is also listed in the 1000 Genome (http://www.1000genomes.org) and in the dbSNP database (build 138, rs104895526) with very low frequencies and always in a heterozygous state. This is in line with a recent observation that gave a minor allele frequency of 0.017 for this change.21 Moreover, in the family presented herein, the maternal grandmother also carried the mutated allele. Together, considerable evidence suggests that this maternal change alone is not sufficient to cause the phenotype in the offspring, or that a stochastic process is involved. In this context, it is remarkable that in the array-based methylation analysis we noticed the mother (II-4) of the affected offspring to show a strong hypomethylation (normalised methylation value −1.2) at one CpG (cg16106497) in the region containing the 5′ ends of both genes NLRP2 and NLRP7 compared with controls (mean 2.0, range −0.2 to 3.9, see online supplementary figure S6). Moreover, custom tiling array CGH provided evidence for a small deletion (∼300 bp) in this region in the index patient II-4, but analysis was hampered by a simple tandem repeat consisting of two highly similar (98%) sequence parts (see online materials). Though it is tempting to speculate that these potential variants reduce NLRP7 expression so that the mother is functionally compound heterozygous at this locus, a maternal dysfunction of NLRP7 could not ultimately be proven.
To explore whether changes in genes other than those known to be related to imprinting disorders might cause the phenotype, we performed exome sequencing of two affected children (III-1 and III-2), the parents (II-3 and II-4) and the maternal grandparents (I-3, I-4). Since the inheritance pattern could be most likely explained by a mutation in a maternal effect gene, we searched for (i) de novo mutations in the mother by comparing exome data of the mother and her parents; (ii) homozygous and compound heterozygous mutations in the mother and (iii) paternally inherited mutations in the mother not present in the maternal grandmother and the cohort of healthy individuals of our in-house database. The query for de novo, compound heterozygous and homozygous mutations yielded no hit. In the third approach, 110 mutations including 83 missense mutations were discovered (see online supplementary table S8). The unexpected high number of mutations/variations is likely due to the ethnic differences between the family and our in-house healthy controls. The variant in NLRP7 was confirmed, but as it is listed in dbSNP it has not been included in online supplementary table S8. We also analysed the exome data in regard to a paternal contribution, although we think it is unlikely that this inheritance model would be compatible with the observed effects. The query showed 35 variants present in the father and the two siblings investigated by exome sequencing (III-1 and III-2; see online supplementary table S9). These variants were not present in the mother or in the maternal grandparents. No overlapping de novo mutations were identified in both siblings rendering germline mosaicism in one of the parents unlikely. Thus, the exome sequencing approach in this single family did not unambiguously point to a causative gene.
The family presented herein along with the molecular studies suggests the existence of a hitherto unrecognised familial disorder of altered DNA-methylation, which besides paternally and maternally imprinted loci also affects a considerable number of loci not known to be associated with parent-of-origin-specific DNA-methylation. Whether the disorder is linked to allelic methylation or other mechanisms remains speculative at this stage. The same holds true for the pathogenic role of the NLRP7 variant. Nevertheless, based on the findings presented herein, future studies of individuals with multilocus imprinting disorders should also include the analysis of DNA-methylation of loci not known to be imprinted.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
AC, JR, OA and DK contributed equally to this study. KB and RS share senior authorship.
Acknowledgements We thank Prof. Paul-Martin Holterhus (University Medical Center Schleswig-Holstein, Kiel, Germany) for providing DNA of primary scrotal fibroblast tissue culture for controls. The support of the technical staff of the involved institutes is gratefully acknowledged. We thank Melanie Heitmann for expert technical assistance and Dr Ludger Klein-Hitpaß for supervising exome sequencing experiments. Special thanks go to the family for supporting the analyses and the clinical colleagues involved in the care and follow-up of the family, particularly Drs Susanne Metzger, Felix Riepe, Uta Siebert and Alexander Claviez. Members of the authorship are part of the European Network of Imprinting Disorders; COST Action BM1208.
Contributors RS, AC, BH and KB designed the study. NJS and LS performed histopathological review and provided material. JR, AH, JB, DK, BK, SB, DM, IN, IV and KB performed the experiments. JR, OA, JIM-S and DK performed data analysis. AC, EJ, CSvK and RS performed clinical characterisation and provided material of the family members. IKT and BH were involved in data interpretation. AC, JR, OA and RS wrote the manuscript. All authors approved the manuscript. AC, JR, OA and DK contributed equally to this work. KB and RS share senior authorship.
Funding This study has been supported by the German Ministry of Science and Education (BMBF) in the framework of the project ‘Disorders caused by imprinting defects’) Imprinting Network (01GM0886 and 01GM1114).
Competing interests None.
Patient consent Obtained.
Ethics approval The study has been approved by the Institutional Review Board of the Medical Faculty of the Christian Albrechts University Kiel (AZ B305/08).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.