Genetic variation affecting DNA methylation and the human imprinting disorder, Beckwith-Wiedemann syndrome

The human imprinting disorder Beckwith-Wiedemann syndrome (BWS) (OMIM 130650) is characterised by overgrowth in the prenatal and postnatal period, macroglossia, umbilical hernia or exomphalos, neonatal hypoglycaemia, ear lobe creases and pits, nevi flammei, hemihyperplasia and organomegaly, particularly of the kidney, liver and pancreas. One of its more serious complications is an increased predisposition to cancer in early childhood with tumour risk segregating with distinct genetic and epigenetic subtypes (reviewed in [1–4]).

The majority of the affected children have an isolated epigenetic abnormality at the 11p15.5 imprinting centre 2 (IC2), known as KCNQ1OT1 TSS-DMR, the location of the KCNQ1OT1/CDKN1C locus. Fifty percent of all children with features of BWS have either mosaic or complete loss of methylation (LOM) within the KCNQ1OT1 promoter and impaired expression of maternal CDKN1C, a critical regulator of growth during early development [5, 6]. For the majority of BWS cases with KCNQ1OT1 TSS-DMR imprinting disruption, the aetiology is unknown and the phenotype arises without any evidence of heritability. Rare cases have been reported with imprinting defects in KCNQ1OT1 TSS-DMR associated with inherited recessive mutations affecting NLRP2 on 19q13.42 [7], and more recently, maternal effect missense and truncating variants in other NLRP loci, including NLRP5 and NLRP7, were reported in the mothers of offspring with multi-locus imprinting disruption (MLID) that included loss of methylation at KCNQ1OT1 TSS-DMR [8, 9]. MLID cases with methylation loss at KCNQ1OT1 TSS-DMR frequently had characteristics of BWS. Animal studies have shown that defective methylation is set in the gametes or during the first cell divisions of the embryo, and genetic modifiers and environmental factors influence this process [10]. Indeed, in BWS, the increased incidence of BWS cases with KCNQ1OT1 TSS-DMR LOM after assisted reproduction (ART) points to the involvement of environmental factors that perturb methylation at KCNQ1OT1 TSS-DMR or alternatively suggest a link between therapeutic interventions for female infertility and defective methylation [11–15].

The one-carbon pathway is the major metabolic pathway through which dietary folate is converted to methyl donor groups that subsequently methylate DNA via catalysis of the universal methyl donor S-adenosyl-methionine in the presence of DNA methyltransferase 1 (DNMT1). The key enzymes of this pathway include methylenetetrahydrofolate reductase (MTHFR), methionine synthase reductase (MTRR), 5-methyltetrahydrofolate-homocysteine S-methyltransferase (MTR), cystathionine beta-synthase (CBS) and methionine adenosyltransferase (MAT1A), and the associated effects of genetic variants with impaired enzyme activity have been documented in several studies. A common SNV (previously commonly known as c.677C>T) in MTHFR, rs1801133: C>T, NP_005948.3: p.Ala222Val, increases protein thermolability [16] and has been linked to vascular thromobosis, hyperhomocysteinaemia and global hypomethylation in a setting of dietary folate depletion ([17–20]. A second MTHFR SNV, rs1801131: A>C, NP_005948.3: p.Gln429Ala, is linked to DNA hypomethylation, independently of folate availability and acts synergistically with rs1801133 to reduce MTHFR activity [21, 22]. MTR catalyses the remethylation of homocysteine to methionine and SNV rs1805087: A>G, NP_000245.2: p.Asp919Gly alters the helix structure in the substrate binding domain leading to impaired methionine biosynthesis and reduced methyl donor availability [23–25]. MTRR rs1801394: A>G, NP_002445.2 p.Ile22Met, acts synergistically with homozygous MTHFR rs1801133: C>T, NP_005948.3: p.Ala222Val, in the presence of low folate and/or vitamin B12 status [26].

The isoforms of Dnmt1 (Dnmt1o and Dnmt1s) play a critical role in the maintenance of DNA methylation at imprinted regions during pre-implantation mammalian development [27]. The absence of 118 N terminal amino acids allows Dnmt1o protein to accumulate to high levels in non-dividing growing oocytes [28], and expression is maintained in pre-implantation embryos in the morula and blastocyst. Howell et al. 2001 showed that Dnmt1o, while not essential for imprint establishment in oocytes, was required for the maintenance of methylation at imprinted loci, and heterozygous offspring of Dnmt1o-null mothers exhibited imprinting defects. The functions of Dnmt1o and Dnmt1 are interchangeable as Dnmt1 is itself able to maintain methylation during pre-implantation development [29, 30], and Dnmt1o can rescue Dnmt1−/− ES cells [31]. However, although these studies implicate DNMT1 in imprinting disorders, it has not previously been examined in the context of BWS or MLID.

In this study, we investigated the genetics of folate metabolism and maintenance of DNA methylation in cases of BWS with loss of methylation at KCNQ1OT1 TSS-DMR. This included an analysis of SNVs in MTHFR, MTRR, MTR, MAT1A and CBS predicted to alter the function and a full mutation screen of DNMT1. While a statistically significant association between deleterious SNVs in folate pathway genes could not be conclusively demonstrated, we identified three patients with rare missense amino acid substitutions in DNMT1 that are represented in dbSNP at a frequency of < 0.001. Functional characterization of these variants showed a significant reduction in complex formation with hemimethylated DNA, consistent with impaired DNMT1 activity. This is the first study examining the folate pathway and the major human methyltransferase, DNMT1, in the most common epigenetic subtype of BWS.

In this paper, we have made observations that have not been previously reported. Firstly, patients with BWS and KCNQ1OT1 TSS-DMR LOM are more likely to have the low-activity MTHFR variant commonly known as c. 677C>T (rs1801133: C>T) although when applying a stringent statistical test for multiple testing, this association was not found to be statistically significant. BWS is a comparatively rare disorder, and examination of this relationship in larger cohorts will be required to consolidate it definitively. Secondly, we identified three rare missense variants of DNMT1 and DNMT1o and characterised their capacity for maintaining DNA methylation in vitro. All variants had impaired enzymatic activity compared to wild-type DNMT1.

A large body of evidence has previously implicated MTHFR rs1801133: C>T in hyper-homocysteinaemia and DNA hypomethylation, that is potentiated by low dietary folate and homozygosity for the T allele [17–20, 22, 34–37]. MTHFR rs1801133: C>T causes an amino acid substitution of highly conserved alanine to valine at position 222 (p.Ala222Val) resulting in reduced enzyme activity due to increased thermolability [16]. Furthermore, BWS has been reported previously in association with homocysteinuria occurring with CBS enzyme deficiency [38]. Although we identified an association between the MTHFR rs1801133 T allele prevalence and BWS that was not statistically significant by a stringent statistical test, it remains to be determined whether a combination of low dietary folate and vitamin B6 and B12 deficiency, and this genotype may have a stronger association with BWS incidence. Future studies, including the prospective collection of dietary intake information, will require global collaborative efforts to examine these associations in large cohorts.

Our observations suggest that the identified low-activity DNMT1 variants may contribute to a subgroup of BWS cases with imprinting defects through the effects on the maintenance methylation of hemimethylated DNA during DNA replication prior to implantation. Given the observation of complete loss of methylation in all three cases with variant DNMT1, we further propose that methylation maintenance failure at KCNQ1OT1 TSS-DMR occurs at a very early stage of embryogenesis in the cleavage embryo. Evidence from mouse models suggests a maternal effect on the maintenance of DNA methylation in the pre-implantation embryo. High levels of oocyte-derived maternal Dnmt1o are present in the cleavage embryo and protect imprinted loci from methylation loss at imprinted sites as well as in the intergenic regions. Mothers lacking Dnmt1o have abnormal imprinting in their offspring at the pre-implantation stage, including abnormal X-inactivation, in addition to placental hypomethylation [39]. Human DNMT1o is similarly expressed in mature oocytes and in early-stage embryonic development [32].

A direct role for DNMT1 in imprinting disorders as a result of maternal effect mutation is also suggested by the associations between the Uhrf1 protein and Dnmt1. Dnmt1 is recruited by Uhrf1 to replicating DNA [40], and intriguingly, a potentially deleterious UHRF mutation has been recently described as having a maternal effect in twins affected by MLID with loss of methylation [9]. Furthermore, another maternal effect protein, NLRP2, implicated in MLID and BWS, causes re-localization of Dnmt1 in oocytes and pre-implantation embryos [41]. Hence, there is accumulating evidence for the potential involvement of DNMT1 as a maternal effect protein associated with loss of methylation at imprinted loci.

Of the three cases with rare missense DNMT1 alleles, maternal inheritance of DNMT1 NM_001130823.1: c.406C>T, NP_001124295 p.Arg136Cys, could be demonstrated. Modes of inheritance could not be examined in the remaining two cases due to the lack of availability of parental DNA; however, case B96 had a sibling who was stillborn with exomphalos who was not tested for BWS, and the possibility of an inherited predisposition in this individual cannot be discounted. Case B66 with the variant DNMT1 NM_001130823.1: c.3668G>A, NP_001124295 p.Arg1223His, was a female conceived by intracytoplasmic sperm injection, and she also had a greater number of the clinical features of BWS when compared with the other two cases (five versus three in the other DNMT1 variant cases). It is not known whether she had inherited the DNMT1 variant from either of her parents; however, maternal inheritance would be consistent with recent observations in the mouse showing that oocytes deficient in Dnmt1o are more susceptible to imprinting defects following ART [42].

There was no reported indication of parental or familial BWS in any case. However, this might not be unexpected in a scenario where a DNMT1 mutation has a maternal effect in the pre-implantation embryo. Mothers may be mutation carriers themselves but unaffected if they had inherited a deleterious DNMT1 SNV from their fathers or had acquired one de novo. In such circumstances, only offspring arising from an oocyte expressing abnormal DNMT1 protein would be affected.

Previous studies on mouse Dnmt1s have identified regions required for maintenance of imprinting that are localised to the region spanning amino acids 191-394; however, none of the variants in this study are located within this domain [27]. The variant DNMT1 NM_001130823.1: c.406C>T, NP_001124295 p.Arg136Cys, lies within an interaction site with DNMT3A suggesting the possibility of an effect on imprint establishment, given the evidence for coordination between DNMT1, DNMT3a and DNMT3b in both establishing and maintaining methylation [43, 44]. Recent studies have revealed insight into the structure of human DNMT1 protein, and conserved domains are described for DNMT1 NP_001124295.1 in the Conserved Domain Database hosted by NCBI [45–47]. None of the identified variants was located directly within the enzymatic active site at Cysteine 1242 in Ref Seq NP_001124295. Variant DNMT1: NP_001124295 p.His1118Arg is immediately adjacent to the second bromo-adjacent homology (BAH2) domain, and variant DNMT1: NP_001124295 p.Arg1223His lies within the catalytic domain. Histidine at 1118 is distant (> 45 Å away) from the DNA binding, SAM binding and catalytic cysteine within the catalytic domain. In the DNMT1 structures, this residue is in a short helix where it is solvent exposed and has few intra-molecular interactions (structure 3PTA in [47] and structure 4WXX in [46]). Residue Arg1223 is in the catalytic domain and distant from the DNA, SAM and Cys1242 sites (21 Å, < 15 Å, < 28 Å, respectively). This residue is leucine in mouse Dnmt1 (structure PDB ID 4DA4 [48], and both residues occupy similar structural space at the apex of a loop. In the human DNMT1 structures (PDB IDS 4WXX, 3PTA), the arginine adopts two radically side chain conformations. In both configurations, the arginine side chain does not engage in intra-molecular polar contacts and the residue is solvent exposed. The loop in which the DNMT1: NP_001124295 p.Arg1223His variant is found is added during the evolution of cytosine C5 methyltransferases. The loop is absent in the methyltransferase from Haemophilus parahaemolyticus (HhaI) [49] that has high structural similarity to the mouse and human DNMT1 catalytic domain, suggesting it is an acquired adaptation and not evolutionarily required for activity. Previous work found that DNMT1 mutations that cause hereditary sensory neuropathy were due to protein misfolding [50]. One residue was in a hydrophobic core region while the other was in a linker region between secondary structural elements. Both mutations were found to affect the folding and stability of DNMT1 protein. In contrast, the variants reported here are in surface-exposed regions and are unlikely to be involved in the intra-molecular interactions. Therefore, they are not proposed to affect folding or stability. However, despite sharing 80% sequence identity, recent subtle differences in function between the mouse and human structures of DNMT1 protein have been recently recognised [46]. Therefore, any definitive structural effects of the identified variants are currently unclear. It may be possible that mutations exert their effects by altering conformational flexibility that is not apparent in the crystal structure. It is interesting to note however that the effect of the DNMT1: NP_001124295 p.Arg1223His variant on the activity was the strongest in the trapping assay and comparable to the effects described for DNMT1 variants reported in the previous study examining hereditary neuropathies [50].

In a setting of maternal dietary folate deficiency or impaired embryonic folate metabolism, the observed decrease in DNMT1 enzymatic activity could be further exaggerated by reduced availability of the substrate S-adenosyl methionine (SAM) or increases in inhibitory S-adenosylhomocyteine (SAH). Interestingly, all cases with variant DNMT1 had at least two variants predicted to adversely affect folate metabolism via the effects directly on MTHFR and methylation of tetrahydrofolate or via the effects on the conversion of homocysteine to methionine via methionine synthase reductase (MTRR) and 5-methyltetrahydrofolate-homocysteine S-methyltransferase (MTR) using vitamin B12 as a cofactor.

Due to their rarity in the general population, presence in a subgroup of BWS and MLID cases, and functional characterization supporting reduced methyltransferase activity, these DNMT1 variants are worthy of reporting. Ultimately, mouse models examining the effect of variant DNMT1 in the context of imprinting disruption at specific loci including KCNQ1OT1 TSS-DMR will be required for proof of association; however, larger cohort studies on families with imprinting disorders are also likely to be fruitful.

In conclusion, the observations from this study suggest that novel genetic factors affecting DNA methylation may cause imprinting disruption in a subgroup of BWS cases. This study paves the way for larger prospective studies in BWS and other human imprinting disorders to examine the relationship between genotypes linked to DNA methylation in concert with environmental factors, to fully elucidate the causes of abnormal genomic imprinting.