Editor—Beckwith-Wiedemann syndrome (BWS) is a human overgrowth disorder with a variable phenotype and genetic heterogeneity. Recent data indicate that the BWS locus is subject to genomic imprinting and current evidence shows that in many patients the disease is associated with epigenetic lesions of genes on 11p15.5. BWS is characterised by pre- and postnatal overgrowth, macroglossia, and anterior abdominal wall defects. Additional, but variable complications include organomegaly, hypoglycaemia, hemihypertrophy, genitourinary abnormalities, and a predisposition to embryonal tumours in about 5% of patients.1 The genetics of BWS are complex, but parent of origin effects, suggesting genomic imprinting, have been implicated in the pathogenesis of three major groups of patients2: (1) for patients (∼2%) with chromosome 11p15.5 abnormalities, duplications are of paternal origin and balanced translocations or inversion breakpoints of maternal origin; (2) in familial cases (∼15% of all cases) which exhibit more complete penetrance with maternal transmission; and (3) approximately 20% of sporadic cases have uniparental disomy (paternal isodisomy) for chromosome 11p15.5. Cloning of genes in the vicinity of BWSCR1, the most distal breakpoint cluster associated with BWS balanced cytogenetic anomalies, and within the area of minimal disomy present in cases of paternal isodisomy, has led to the identification of a group of genes as potential candidates in the aetiopathology of BWS. Thus, multiple imprinted genes bounded centromerically by NAP1L4 and telomerically by L23mrp (RPL23L) have been identified.3-6

For a gene to be a good candidate to account for a significant number of BWS cases, it should map to this region and constitute either a paternally expressed growth promoter or a maternally expressed growth inhibitor. However, because factors involved in maintaining or modifying genomic imprints may affect the epigenotype and therefore expression status of imprinted genes, it is possible that an underlying lesion in a non-imprinted gene may manifest itself as an imprinted trait. Extensive characterisation of IGF2,CDKN1C, andKVLQT1 in BWS patients has been undertaken already. With the exception of the involvement ofKVLQT1 in some balanced translocations and inversions,7 and IGF2 in paternal duplications,8 the only mutations within the coding sequence of a candidate gene described are for CDKN1C.9-14 In our series, germline CDKN1C mutations accounted for ∼40% of familial cases and 5% of sporadic cases.14This suggests that further BWS genes remain to be identified. Most sporadic BWS patients show LOI ofIGF2 15 16 and a candidate BWS gene might cause BWS by influencing IGF2imprinting status.

The H19 gene maps approximately 200 kb telomeric of IGF2 in humans; synteny is conserved in the mouse.17 Data from human tumours, BWS, and experimental manipulation of the mouse genome indicate that the regulation of H19 andIGF2 expression is closely and reciprocally linked.18 For some BWS patients withIGF2 LOI, biallelicIGF2 expression is associated with suppression of H19 expression and reversal of the normal (unmethylated) maternal allele methylation patterns, so that both parental IGF2 andH19 alleles display a paternal methylation pattern.19 20 However, in other cases with biallelicIGF2 expression,H19 and IGF2allelic methylation is normal.16 A possible explanation for these observations is that the maternalH19 RNA is functionally inactivated, not affecting its own imprinting status, but leading to loss of repression of the maternal IGF2 allele. This hypothesis is consistent with data from the mouse.21

The most centromeric imprinted gene in the 11p15.5 imprinted region is the candidate tumour suppressor geneTSSC3.22 23 TheNAP1L4 gene lies 15 kb 5′ ofTSSC3 and encodes a chaperone protein associated with chromatin assembly and has been shown to bind to core and linker histones facilitating transfer to the DNA template.24 25 Although NAP1L4has not been shown to be imprinted to date, this has not been extensively investigated and tissue specific or developmentally regulated imprinting cannot be excluded. NAP1L4lies within the interval associated with loss of heterozygosity in Wilms tumour (WT2) and centromeric to the BWSCR1 breakpoint cluster. It therefore fulfils one of the criteria for a BWS candidate gene. The possibility that chromatin structure affects the activity and imprinting status of genes is very strong26 and it is possible that mutations in NAP1L4 might appear to have an allele specific effect even ifNAP1L4 is not itself imprinted. Recent studies on a BWS family with a maternally inherited inversion of 11p15.5 suggest that relaxation of IGF2imprinting may result from an H19independent pathway. In the family reported by Brownet al,27 a BWSCR1 breakpoint in the region of NAP1L4 was associated withIGF2 LOI and normalH19 expression. Importantly,NAP1L4 is expressed in normal kidney and some Wilms tumours (WT) lacking NAP1L4expression show IGF2 LOI (Munroeet al, unpublished observations). While mutations in NAP1L4 have not been found in sporadic Wilms tumours,25 those associated with a genetic predisposition have not yet been examined, leaving open the possibility that NAP1L4 mutations in the germline or somatic mutations early in development may predispose to the changes seen in BWS and familial WT through an effect on the imprinting status of key genes such as IGF2.

Overexpression of Igf2 in mouse development mimics many features of BWS,28 further implicating IGF2in BWS. The IGF2 gene has a conserved differentially methylated region (DMR) in exon 9. This region has been shown in mice to be consistently methylated on the expressed paternal allele. It has been postulated that the DMR is a methylation sensitive site for silencer binding. Hence, IGF2 LOI could result from mutations that alter the sequence motifs for silencer binding on the maternal IGF2 allele.

To investigate the molecular mechanism of BWS, we have performed mutational analysis of the H19 andNAP1L4 genes and the DMR2IGF2 region in BWS patients. Up to 21 subjects (11 male, 10 female) with BWS were investigated. BWS was diagnosed according to previously defined criteria: (1) three major features (anterior abdominal wall defects, macroglossia, and pre-/postnatal growth >90th centile), or (2) two major features plus three or more of: characteristic ear signs (ear lobe creases or posterior helical ear pits), facial naevus flammeus, hypoglycaemia, nephromegaly, and hemihypertrophy.1 Peripheral blood samples were obtained from all patients and high molecular weight genomic DNA was extracted as described previously.20

The H19 genomic sequence was numbered as in Brannan et al.29 Sequencing was performed from nucleotides 650 to 3461. This included 170 bp of sequence 5′ to the transcription initiation site and all five exons and four introns. The H19 gene was sequenced using genomic DNA to derive overlapping template fragments approximately 500-700 bp long. Primer sequences used to cover the entire H19 genomic region in both forward and reverse directions (and PCR conditions) are available on request. The sequencing PCR reaction was run according to the ABI PRISM protocol and H19 sequence data were obtained using ABI software.

The H19 gene was sequenced in 15 BWS patients without uniparental disomy. These patients represented a variety of aetiologies of BWS; two patients were familial (without germline CDKN1C mutations) and 13 were sporadic. Of the latter, two had previously been identified as having aIGF2/H19imprinting centre defect (ICD) with H19promoter hypermethylation and silencing ofH19 expression.19 20 Of the remaining 11 sporadic BWS patients with normalH19 methylation, seven were informative for allele specific IGF2 mRNA expression analysis; five had biallelic IGF2 expression with normal H19 expression and two had biallelic IGF2 expression and absentH19 expression (but this was not associated with H19 hypermethylation as in the putative ICD cases).16 Southern analysis (PstI and SmaI digest) did not show evidence of a genomic rearrangement in any case.19 20

Comparison of the H19 sequences in BWS cases and controls to the sequence published by Brannanet al 29 (Genbank Accession number M32053) showed sequence differences from the published sequence at five sites that were present in all patients and controls. These were considered to represent sequencing errors in the published sequence (table 1). In addition to the consistent sequence changes identified in all patients and controls, 10 polymorphic sequence variants were noted in BWS patients. Two of these nucleotide substitution sites resulting in a RFLP have been described previously (AluI site at nt 2883 andRsaI site at nt 3241),34 35but eight novel sequence variants were identified (table 1): C→T at nt 1531, T→A at nt 1569, G→C at nt 2461, G→C at nt 2791, T→C at nt 2976, T→C at nt 2992, A→G at nt 3238, and T→C at nt 3281. In each instance, these changes were also identified in normal controls and were considered to represent simple polymorphisms.

The intron-exon structure of NAP1L4 has been reported previously25 and primers were designed to amplify each of the 14 exons (primer details and conditions are available on request). PCR products (5 μl) were denatured and electrophoresed through an 8% polyacrylamide gel containing 5% glycerol, using 0.5 × TBE as the running buffer. Electrophoresis was at 1 watt per gel for 12-18 hours depending on the size of PCR product. DNA bands were visualised by silver staining.

A total of 21 BWS patients were analysed forNAP1L4 gene mutations. These included the 15 patients analysed for H19 gene mutations and a further two familial and four sporadic cases with normalH19 methylation. Seven informative cases were known to have biallelic IGF2 expression analysis.16 The coding sequence of theNAP2 gene was amplified in 14 single exon fragments and analysed by SSCP and ABI sequencing. The only sequence variant identified was an A→C transversion at +7 of the 3′ exon 2 splice site. This sequence variant does not affect the splice site consensus sequence but causes gain/loss of anRsaI restriction enzyme cutting site (fig1). This finding was used to screen for the splice site variant and this was identified in both BWS cases and controls. We concluded that this change represented a simple polymorphism.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

H19 gene A/G polymorphism at nt 3238.

Primers used to analyse the DMR were forward: 5′ CCC TCT GCC CGT GGA CAT TAG 3′ and reverse: 5′ GGC GGG GTC TTG GGT GGG TAG 3′. The PCR conditions were as for SSCP of NAP1L4 with an annealing temperature of 61°C. Half of the PCR product from each patient was cloned into InVitrogen TA cloning vectors according to the manufacturer's instructions (InVitrogen, BV, The Netherlands). The rest of the PCR product from each patient was mixed with an equal volume of formamide loading dye, denatured at 95°C for five minutes, and electrophoresed on 8% PAGE. Sequencing of PCR clones obtained after the InVitrogen TA cloning was according to the ABI PRISM protocol described above. Thirteen patients with BWS (four withIGF2 LOI) were analysed by SSCP. The expected PCR product was 250 bp, and no differences in product were detected by SSCP between any of the patients. After cloning the PCR products, four clones per patient were sequenced. In total, 49 clones were successfully sequenced and all of these were normal.

It appears that mutations or epigenetic lesions inCDKN1C and IGF2respectively are involved in the pathogenesis of BWS. In certain cases, for example, uniparental disomy, other genes may also be involved. While a significant proportion of familial cases (40%) are caused by germline CDKN1C mutations, the vast majority of sporadic cases show either uniparental disomy or loss of imprinting of IGF2. 13 15 19 Manipulation of the mouse H19 gene shows that mutations in this locus may result in LOI ofIGF2, 30 suggesting that alterations in H19 expression could directly affect IGF2 imprinting in humans. BWS patients with LOI of IGF2 may have normal or absent H19 expression suggesting that a variety of mechanisms (including mutations) could inactivateH19 function and lead toIGF2 LOI.16 Despite frequent suggestions that H19 is a candidate BWS gene, H19 mutation analysis has not been reported previously. As germline H19deletions have not been detected in BWS patients withH19/IGF2imprinting centre defects (ICDs) (IGF2LOI, H19 hypermethylation, and silencing) or in “H19 null” BWS patients (IGF2 LOI, absentH19 expression, but normalH19 methylation), the possibility of mutations within the H19 gene itself required investigation. In mice, deletion of theH19 transcription unit leads to loss of imprinting of the adjacent, normally silent, maternal allele of theIGF2 gene, whereas the imprinting status of the replacement transcription unit is retained.21Similarly, deletion of sequences upstream and downstream ofH19, including an endoderm specific enhancer, which affect its transcription, also affect the IGF2 epigenotype both in cis and trans, together with a switch in the characteristic pattern ofIGF2 methylation.18 Taken together, these data suggest a role for H19and its surrounding sequences in the regulation ofIGF2 imprinting. We therefore analysed theH19 gene to determine if some BWS patients had H19 mutations causing a silencing of maternal H19 expression or an expressed but non-functional H19 RNA. AbsentH19 expression might be associated with large deletions, promoter mutations, or intragenic mutations which decreased RNA stability. Our molecular analysis would be expected to detect most intragenic mutations and the report of a single nucleotide substitution in the Xist promoter associated with skewed X inactivation31 raised the possibility thatH19 promoter mutations might account for the ICD or “H19 null” BWS patients. Most of the patients we studied had normal H19promoter methylation analysis, excluding the possibility of a large promoter deletion in these cases. Sporadic BWS cases with normalH19 methylation and absentH19 expression could haveH19 promoter mutations, but we did not detect any evidence of such changes in the 170 bp of sequence 5′ to the transcription initiation site that we sequenced. The finding that most patients were heterozygous for at least oneH19 intragenic polymorphic sequence variant excludes the presence of a complete H19 gene deletion in most cases where expression from both alleles is suppressed. With the reservation that small, more distal 5′ flanking region mutations may have been missed in some patients, we conclude that H19 germline mutations cannot account for the loss of IGF2 imprinting observed in most BWS cases.16 The mode of action ofH19 on IGF2allele specific transcriptional control is unclear, as it does not give rise to a translation product. Although it has been postulated thatH19 functions directly or indirectly as a modifier of chromatin structure in a way similar to that proposed for the action of XIST in X chromosome gene inactivation,32 deletion of the mouseH19 gene does not affect the imprinting status of Mash2, Cdkn1c, orKcnq1,33 suggesting that there are at least two imprinting control centres within this region.

In humans, the observation of biallelicIGF2 expression in a BWS family with a BWSCR1 breakpoint27 is compatible with anH19 independent pathway ofIGF2 imprinting control. BWSCR1 rearrangements associated with LOI IGF2 may show loss of a parental allele specific methylation pattern atKVLQT1, but not more distally atIGF2/H19. 36This observation is consistent with two imprinting control centres in 11p15.5. Thus, a candidate BWS gene might function as a cis acting repressor of maternal IGF2 expression and map centromeric to BWSCR1. NAP1L4 (hNAP2)lies at the centromeric boundary of the imprinted gene cluster on 11p1524 25 and maps centromeric to the BWSCR1 region in the candidate region for a cis acting regulator ofIGF2 imprinting. To date, evidence of imprinting has not been described forNAP1L4, but its biological function as a histone chaperone protein provides a possible mechanism for altering the imprinted status of one or more genes through potential effects on chromatin silencing or activation. Under this model, only the maternalIGF2 allele would be responsive (because of specific methylation or chromatin structure imprints) to a cis acting downregulator. Although NAP1L4 represented a strong BWS candidate gene, our failure to identifyNAP1L4 mutations in a large cohort of BWS patients strongly suggests that NAP1L4 is not a major BWS gene.

Having excluded coding sequence mutations inNAP1L4 and major changes in both the coding sequence of H19 and its immediate promoter region as being frequent pathogenic lesions in Beckwith-Wiedemann syndrome, we then considered that mutations in theIGF2 DMR2 would represent another cause ofIGF2 LOI. Consistent with its putativeIGF2 silencer function, DMR2 is more consistently methylated on the paternal allele than on the maternal in tissues in which Igf2 is expressed, and the unmethylated (maternal) DMR2 sequence is bound by specific nuclear proteins (AM and WR, unpublished observations). However, although DMR2 mutations represented a logical explanation for the subset of BWS patients withIGF2 LOI and normalH19 imprinting, no mutations were identified. Thus, the frequent IGF2 LOI found in BWS must therefore originate in other lesions, either genetic or epigenetic. It still remains to examine the candidature ofTSSC3, IMPT1, ASCL2, and other loci as they might emerge from this dense cluster of growth related genes. Such analysis will identify sequence mutations that lead to altered epigenetic modifications. However, epimutations (that is, changes in the epigenetic status without genetic modification) could be a mechanism for LOI which would be sporadic and potentially be reset in the germline. As mutations in further BWS candidate genes are excluded as a cause of LOIIGF2, the likelihood of epimutation being the major cause of BWS will increase.