Introduction

Beckwith–Wiedemann Syndrome (BWS) is a congenital overgrowth disorder with an incidence of about one in 13 000. Phenotypic expression of BWS is variable, but the three major features are pre- and/or postnatal overgrowth, macroglossia and anterior abdominal wall defects ranging from diastasis recti to exomphalos.1 Less frequent (minor) features include ear creases and helical pits, neonatal hypoglycaemia, hemihypertrophy, facial naevus flammeus and a predisposition to embryonal tumours, particularly Wilms’ tumour. The frequency of embryonal neoplasms in BWS is generally considered to be 5–10%, but there are no standard clinical diagnostic criteria1, 2 and estimates of tumour frequency have varied between studies. Hence, molecular genetic diagnosis would facilitate the diagnosis of BWS and comparison of different BWS cohorts.

The genetics of BWS are complex, but all causes to date are associated with alterations in the expression or function of one or more imprinted genes in the 11p15.5 imprinted gene cluster.3 Chromosome 11p15.5 was first implicated by the finding of paternally derived duplications of 11p15.5 in BWS patients.4, 5, 6, 7 Subsequently, maternally inherited balanced rearrangements of 11p15 were also demonstrated to be associated with BWS.8, 9 In contrast, maternally derived 11p15.5 duplication was associated with growth retardation.10 Overall, it is estimated that up to 3% of BWS patients have a duplication (BWSDUP11) or a balanced rearrangement (inversion BWSINV11; translocation BWSTRANS11). The finding of chromosome 11 paternal uniparental disomy (BWSUPD) in a subset of sporadic BWS patients provided further evidence that BWS is an imprinting disorder.11, 12 About 20% of sporadic BWS patients have UPD that is invariably a mosaic paternal isodisomy and includes the 11p15.5 gene cluster.13, 14 This cluster contains more than eight imprinted genes, but those most strongly linked to BWS include the paternally expressed growth promoter IGF2, and the maternally expressed candidate tumour suppresser genes CDKN1C (p57KIP2, a cyclin-dependent kinase inhibitor) and H19 (an untranslated RNA). Whereas BWSUPD cases are predicted to have increased IGF2, reduced H19 and reduced CDKN1C expression, only the expression of paternally expressed genes such as IGF2 should be altered in BWSDUPL11 patients. Nevertheless, CDKN1C was unequivocally implicated in the pathogenesis of BWS by the finding of germline CDKN1C mutations in a subset of patients (BWSMUTCDKN1C subtype15, 16, 17). Approximately 15% of BWS cases are familial, and germline CDKN1C mutations are found in about half of familial cases and 5% of sporadic cases.16, 17 Extensive analysis of potential imprinting control elements for the reciprocally imprinted IGF2 and H19 genes led to the definition of a paternally methylated intergenic CTCF binding region that is proposed to function as an imprinting control centre (IC1) regulating imprinting of IGF2 and H19, but not more centromeric genes such as CDKN1C.18, 19, 20 Approximately 5% of BWS patients (BWSICD1) are proposed to have a defect of this distal imprinting centre as the maternal IGF2 and H19 alleles display a paternal epigenotype (hypermethylation and silencing of H19 and biallelic IGF2 expression). Recently, some patients with IC1 defects have been demonstrated to have germline deletions within the intergenic CTCF binding region.21 Finally, a second more centromeric imprinting centre (IC2) has also been proposed.3, 22, 23 IC2 is associated with a maternally methylated differentially methylated region (KvDMR1) and a paternally expressed untranslated RNA (KCNQ1OT1 (LIT1)).22 Loss of maternal allele methylation at KvDMR1 (and biallelic KCNQ1OT1 expression) occurs in 50–60% of BWS patients (BWSICD2),23, 24 and is associated with downregulation in cis of CDKN1C expression25 and variable loss of imprinting (LOI) of IGF2.23 However, H19 imprinting and methylation is unchanged consistent with IC1 and IC2 as two distinct imprinting centres within the 11p15.5 gene cluster. Most patients with IC2 defects appear to have epimutations, but germline deletions may occur.26

Previously, we looked for ‘genotype–phenotype’ correlations in BWS and reported associations between hemihypertrophy and BWSUPD, and exomphalos and BWSMUTCDKN1C and BWSICD2 subtypes.16, 24 In addition, we noted that embryonal tumours occurred in the BWSUPD and BWSICD1 but not BWSICD2 subgroups, but these differences were not statistically significant. Subsequently, DeBaun et al27 confirmed the associations with anterior abdominal wall defects and demonstrated significant associations of BWSUPD with hemihypertrophy, cancer and hypoglycaemia. However, Weksberg et al28 found that cancer was not limited to the BWSUPD and BWSICD1 groups, but non-Wilms’ tumours were frequent in their BWSICD2 group. The identification of clear genotype–phenotype correlations, which led to different management strategies according to the underlying molecular diagnosis, would have important clinical implications. Hence, we have undertaken further analysis on a large (n=200) group of BWS cases with a defined molecular subtype. This analysis demonstrates that BWS should be considered as a heterogeneous disorder with different molecular subgroups of BWS posing different genetic and clinical management problems.

Methods

Patients

In all, 193 patients were ascertained via a national research study or following referral to the West Midlands Regional Genetics Service for diagnostic testing (a minority were included in previous publications16, 24). In order to avoid ascertainment bias, all patients referred for molecular genetic diagnosis on suspicion of BWS underwent molecular analysis. In order to obtain meaningful genotype–phenotype correlations, we then analysed only those cases with a proven molecular diagnosis of BWS. Having ascertained our cohort of 193 patients with a confirmed molecular genetic diagnosis of BWS, we recruited into the study an additional seven BWSICD1 cases in order to provide sufficient power for statistical analysis of the BWSICD1 subgroup.21 Sufficient DNA was available for analysis of the extent of UPD in 48 cases, and in order to increase statistical power, DNA samples from a further two patients with UPD and Wilms’ tumour were studied to evaluate more fully whether the extent of segmental disomy influenced Wilms’ tumour risk.29 However, these two additional BWSUPD patients were excluded from the genotype–phenotype analysis, as their tumour status was known upon enrolment. Clinical information was collected by a standard questionnaire, inspection of hospital notes or direct examination. Patients were included in the study following a molecular diagnosis of BWS irrespective of the original indication for referral, but we excluded patients from multiple pregnancies for the prenatal growth analyses.

Molecular analysis

DNA was extracted from peripheral blood lymphocytes or tissue samples by standard procedures. CDKN1C was sequenced as described previously.16 UPD analysis was performed on 25 ng of genomic DNA in a 10 μl total reaction volume containing fluorescently labelled forward primer (0.4 pM), unlabelled reverse primer (0.4 pM), dNTP (0.16 mM) Ampli Taq (0.75 U) in 1 × Ampli Taq buffer containing 1.5 mM MgCl2. After an initial denaturation step (5 min at 94°C), 24 cycles of amplification were performed (94°C 1 min, TA°C 1 min, 72°C 1 min) followed by a final extension for 5 min at 72°C. The microsatellite markers tested were those at tyrosine hydroxylase,11 TA=62°C, D11S1318 (TA=57°C) and D11S1984 (TA=57°C). Alleles were resolved through a 6% polyacrylamide gel on an ABI377, and the ratio of paternal to maternal allele intensity was determined using the Genotyper software (ABI), a ratio of greater than 1.3:1 in favour of the paternal allele indicated UPD.

Methylation analysis was performed either by Southern blotting as described previously (IC1),30 (IC2)22 or following bisulphite conversion of unmethylated cytosines to uracil. Bisulphite modification of DNA was according to Agathanggelou et al,31 the modified DNA was amplified by PCR (using a fluorescently labelled forward primer); for H19 methylation analysis, the primers used were: forward GTA GGG TTT TTG GTA GGT ATA GAG; and reverse CTT AAA TAA CCC RAA ACR TTT CCA C (TA=55°C, and a hot start was required). For KvDMR1 methylation analysis the primers used were W2 GTT ATT TTA TAT TTA GTT AGT GTT TTA TG and W4 TCT TAC TAA AAA ACT CCC TAA AAA TC (TA=53°C). The PCR products were then digested with the restriction enzyme BstU1, which has the recognition site CGCG and hence will cut only modified DNA derived from previously methylated DNA. The fragments were separated on an ABI377 or 3730 and the ratio of methylated to unmethylated DNA determined.

Statistical analysis

Fisher's exact testing, Wilcoxon–Rank sum, t-testing and Kaplan–Meier analysis were used as appropriate. Statistical significance was taken at the 5% level.

Results

In all, 200 BWS patients with a defined molecular subtype (BWSMUTCDKN1C n=16 (from seven families), BWSICD2 n=116, BWSICD1 n=14 and BWSUPD n=54) were analysed for genotype–phenotype correlations. Although we did not restrict molecular analysis to patients satisfying specific clinical criteria, 97% of patients had at least two ‘major’ features of BWS (eg macrosomia; macroglossia; exomphalos or umbilical hernia; visceromegaly, anterior linear ear lobe creases/posterior helical ear pits; embryonal tumour, hemihypertrophy, adrenocortical cytomegaly or renal anomaly or one or more family members with a clinical diagnosis of BWS; see http://www.geneclinics.org/). Interestingly, of two patients with hemihypertrophy and an ICD2 defect, there were no other features of BWS in one case and only neonatal hypoglycaemia in the other.

Considering all patients together, there were significant associations between the presence of ear pits and/or creases and facial naevus flammeus (P=0.015), macroglossia and facial naevus flammeus (P=0.022), hemihypertrophy and embryonal tumour (P=0.044), and inverse relationships between the presence of exomphalos and hemihypertrophy (P<0.001) and hemihypertrophy and macroglossia (P=0.006).

The severity of anterior abdominal wall defects differed significantly between subgroups (see Figure 1a). Thus, the frequency of exomphalos was much higher in the BWSMUTCDKN1C and BWSICD2 subtypes than in the BWSICD1 and BWSUPD subtypes (BWSMUTCDKN1C vs BWSICD1 P=0.006, BWSMUTCDKN1C vs BWSUPD P=0.001, BWSICD2 vs BWSICD1 P⩽0.001, BWSICD2 vs BWSUPD P⩽0.001). In contrast, BWSUPD and BWSICD1 subtypes were more likely to have no defect or an umbilical hernia (see Figure 1a).

Figure 1
figure 1

The frequencies of clinical features in BWS stratified by molecular pathology. The frequency of umbilical hernia and exomphalos (a), (the more severe defect, exomphalos, is depicted by a black bar whereas umbilical hernias are depicted by grey bars), hemihypertrophy (b) and ear creases or pits (c) was determined in individuals with BWS and a known molecular defect.

Full size image

Birth weight and gestation data were used to calculate birth weight centiles and these were compared for the four molecular subgroups. BWSICD1 were the heaviest (mean 99.8±0.45 centile), followed by BWSUPD (mean 90.7±21.4 centile), BWSICD2 (mean 85.6±28.0 centile) and BWSMUTCDKN1C (mean 82.9±27 centile). The birth weight centile differences between the BWSICD1 and BWSMUTCDKN1C groups, and the BWSICD1 and BWSICD2 groups were statistically significant (P=0.018 and 0.018, respectively). All other comparisons were P>0.09.

Hemihypertrophy was present in 31% of the patients analysed, but whereas the majority of the BWSUPD group (71%) had hemihypertrophy, it was detected in only a minority of the other subtypes (Figure 1b): BWSMUTCDKN1C=7%, BWSICD2=15%, BWSICD1=23% (BWSUPD vs BWSMUTCDKN1C P=0.06, BWSUPD vs BWSICD2 P<0.001, BWSUPD vs BWSICD1 P=0.001).

The overall frequency of ear creases and pits was 61% (BWSMUTCDKN1C=73%, BWSICD2=67%, BWSICD1=38%, BWSUPD=53%), but these were significantly more common in BWSICD2 cases than BWSICD1 cases (P=0.035) (see Figure 1c). There were no significant differences between BWS subgroups for the frequencies of macroglossia (overall frequency 91%, BWSMUTCDKN1C=93%, BWSICD2=92%, BWSICD1=100%, BWSUPD=86%), facial naevus flammeus (overall frequency 43%, BWSMUTCDKN1C=33%, BWSICD2=48%, BWSICD1=31%, BWSUPD=38%) or neonatal hypoglycaemia (overall frequency 54%, BWSMUTCDKN1C=56%, BWSICD2=51%, BWSICD1=62%, BWSUPD=55%).

In all, 5% of patients in our cohort developed a neoplastic lesion (see Table 1) including one BWSUPD patient who had both a hepatoblastoma and nephroblastoma (Wilms’ tumour), this is similar to the frequency observed by Wiedemann in his study of nearly 400 patients.32 Neoplasia was more frequent in individuals with BWSICD1 or BWSUPD (13%) than in those with BWSMUTCDKN1C or BWSICD2 (1%) (P=0.007, Figure 2). In order to perform more accurate subgroup analysis, we combined our data with those reported by others.27, 28, 29, 33 The results are depicted in Figure 2. Neoplasia was observed in all molecular subgroups, and the overall frequency of tumours was 11%; however, the frequencies in patients with BWSUPD or BWSICD1 were 22 and 28%, respectively, which is significantly greater than the 2.9% and 9.5% observed in patients with BWSICD2 and BWSMUTCDKN1C cases (BWSUPD vs BWSICD2 P<0.0001; BWSICD1 vs BWSICD2 P<0.0001; ICD1 vs BWSMUTCDKN1C P<0.05; BWSUPD vs BWSMUTCDKN1C P=0.06). Wilms’ tumours were not observed in BWSMUTCDKN1C or BWSICD2 subgroups; however, they are the most frequently reported tumour in BWSICD1 and BWSUPD subgroups (compare Figure 2; ‘All studies, all tumours’ with ‘All studies, Wilms’’). The risk of neoplasia in our series was lower than in some other series.27, 28, 29, 33 This might reflect ascertainment bias in some series or differences in age distributions of various cohorts. In our series, most children were aged <8 years (see Figure 3). To allow for this, we performed a Kaplan–Meier plot for all patients and the BWSICD2 and BWSUPD subgroups. The age-related risks of neoplasia in the three groups at 5 years were 9, 0 and 24%, respectively (see Figure 4).

Table 1 Clinical information on BWS patients in whom neoplasia was observed. BWS121 and BWS361 were initially diagnosed as Wilms’ tumour (nephroblastoma)
Full size table
Figure 2
figure 2

The frequencies of neoplasia observed in this study and four smaller studies. The frequency of neoplasia was determined in this study, and four smaller studies. Toronto indicates the study of 65 patients by Weksberg et al,28 Baltimore refers to the study of 58 patients by DeBaun et al,27 Amsterdam refers to the study of 52 patients by Bliek et al,29 and France refers to the study of 71 patients by Gaston et al.33 ‘All studies, Wilms’’ refers to the data for Wilms’ tumours (WT) taken from this study, the Canadian, Dutch and French studies. CDKN1C mutation analysis was not carried out on the American and Dutch cohorts. In total, 45 tumours were reported from 411 children with BWS (21 WT from 353 BWS children), 24 tumours from 107 children (13 WT from 98) with UPD, 12 tumours from 43 children (eight WT from 33) with imprinting centre 1 defects, seven tumours from 240 children (0 WT from 201) with imprinting centre 2 defect and two tumours from 21 children (0 WT from 21) with a germline CDKN1C mutation.

Full size image
Figure 3
figure 3

The age intervals of children at latest clinical follow-up or first tumour incidence. The proportion of the cohort falling into different age groups is indicated.

Full size image
Figure 4
figure 4

Kaplan–Meier plot of risk of neoplasia in all patients with BWS, and in BWSUPD and BWSICD2 subgroups. For details of neoplastic lesions see Table 1.

Full size image

In view of the evidence of imprinted transcripts at the Wilms’ tumour suppresser gene, WT1 locus, 11p13,34, 35, 36, 37 we investigated whether the extent of segmental UPD influenced the risk of renal neoplasia in BWSUPD patients. Of 50 BWSUPD cases analysed, disomy extended to WT1 in 28 patients. Disomy at WT1 was present in 7/8 patients with renal neoplasia (six with Wilms’ tumour and two with severe nephroblastomatosis) and in 21/42 not known to have renal neoplasia (P=0.054). UPD included WT1 in 5/6 patients with Wilms’ tumour.

Discussion

We analysed genotype–phenotype correlations in a series of 200 BWS patients with defined molecular abnormalities. There are no generally agreed diagnostic criteria for BWS. By focusing on patients with a molecular genetic diagnosis of BWS, we were able to consider patients for inclusion irrespective of whether they satisfied ‘strict’ or ‘loose’ clinical diagnostic criteria (see http://www.geneclinics.org). A molecular diagnosis is possible in at least 70% of BWS children, so although in some cases a diagnosis of BWS has to be made on clinical findings, in most cases diagnosis is possible by a combination of clinical and molecular genetic findings. Some patients with molecular findings characteristic of BWS have an incomplete phenotype (eg we found two patients with hemihypertrophy and an IC2 defect in which there were no other features of BWS in one patient and only neonatal hypoglycaemia in the other), but in our series such cases were uncommon. Nevertheless, Martin et al38 have reported five patients with hemihypertrophy and an IC2 defect with no or minimal additional features of BWS (eg single episode of hypoglycaemia, glabellar naevus flammeus). These findings suggest that some patients presenting primarily with hemihypertrophy form part of the BWS spectrum and further follow-up is required to see how the natural history and prognosis of these patients differs from other patients with hemihypertrophy and no clinical or molecular features of BWS.

We found that there were significant differences between molecular subtypes of BWS in tumour risk, severity of anterior abdominal wall defect, birth weight, hemihypertrophy and ear signs of BWS. Differences in tumour risk are important for the clinical management of BWS. We have studied the largest cohort of BWS patients with defined molecular pathology yet analysed, and the overall frequency of neoplasia was 5%, which is lower than in other large studies from USA, Canada and Europe.27, 28, 29, 33 These differences probably reflect differences in ascertainment procedures (eg we tried to avoid preferentially ascertaining patients with tumours and did not use cancer registries) but might also reflect differences in the ages of children in the different cohorts. Kaplan–Meier analysis allows age-related risks of neoplasia to be calculated and we found that the risk for all patients was 9% at age 5 years but was higher in BWSUPD (24%) and lower in BWSICD2 (0% subgroups). Overall, we found a significantly higher risk of tumours in BWSICD1 and BWSUPD cases than in those with CDKN1C mutations or IC2 defects. While this is consistent with previous studies,27, 28, 29, 33 in most individual studies statistical significance was not reached. Therefore, to obtain the best estimates of tumour risks in different molecular subtypes of BWS, we combined our data with those of other centres. This confirmed that overall risk of tumour is highest in the BWSICD1 and BWSUPD groups and lowest in the BWSMUTCDKN1C and BWSICD2 groups. Although we did not identify any tumours in our BWSICD2 group, Weksberg et al28 observed a 16% tumour risk in such individuals, however, these were hepatoblastomas, rhabdomyosarcomas and a gonadoblastoma, and not Wilms’ tumour, which is the most frequently occurring BWS associated tumour in Caucasian populations. Tumours also appear to be relatively uncommon in germline CDKN1C mutation carriers. We have identified a mesoblastic nephroma in one patient and two cases of neuroblastoma have been reported previously.17, 33

Although all children with BWS should be considered at increased risk of tumours compared to the general population, the value of routine tumour surveillance programmes is controversial.39, 40 While many authorities have suggested that all children with BWS should be offered 3-monthly renal ultrasound scans to age 7 years, no effect of screening on survival from Wilms’ tumour has been detected. Another controversial question is whether α-fetoprotein monitoring should be instigated for early detection of hepatoblastoma. Our findings suggest that, among children with a molecular diagnosis of BWS, screening for Wilms’ tumour should be targeted to those at greatest risk, the BWSUPD and BWSICD1 subgroups. In contrast, screening for hepatoblastoma would be of most benefit in BWSICD2 cases and intensive screening for Wilms’ tumour is not indicated for this subgroup. The high risk of Wilms’ tumours in the BWSICD1 and BWSUPD groups compared to the BWSMUTCDKN1C group suggests that increased IGF2 or decreased H19 expression (rather than CDKN1C inactivation) has a key role in the pathogenesis of these tumours. Accordingly, LOI of IGF2 and H19 hypermethylation and repression are frequently observed in sporadic Wilms’ tumours.41, 42 LOI of IGF2 may also occur in patients with IC2 defects, but there is no information as to whether tumour risk in BWSICD2 cases correlates with IGF2 imprinting status. As only a subset of BWSUPD patients develop tumours, we investigated whether the extent of UPD, in particular, whether disomy extended to WT1, influenced the risk of neoplasia. Intriguingly the frequency of Wilms’ tumour and severe nephroblastomatosis was higher in those with WT1 disomy, although it did not reach statistical significance. While WT1 is not thought to be imprinted, it has been suggested that an alternative WT1 transcript (AWT1) and the antisense WT1 transcript (WT1-AS which overlaps the 5′-end of WT1) may be imprinted.35, 36, 43 In foetal tissues, transcription of WT1-AS occurs from both alleles; however, in adult tissues, the maternal allele has been silenced by methylation, whereas in Wilms’ tumours the foetal state is retained.36 Paternal UPD for 11p13 would also have the effect of maintaining the chromosome in the foetal state associated with biallelic expression of WT1-AS. WT1 protein levels are high in foetal kidney and Wilms’ tumours, but low in adult kidney and it has been shown by Moorwood et al35 that WT1-AS transcription can elevate WT1 protein levels in vitro. This may imply that individuals who have UPD extending beyond 11p13 and whose mosaicism encompasses the cells of the kidney are those at highest risk of developing Wilms’ tumour.

We confirmed our earlier finding that exomphalos is much more frequent in BWSMUTCDKN1C and BWSICD2 cases than BWSICD1 or BWSUPD children, and in mouse models of BWS, mice overexpressing Igf2 exhibit overgrowth without exomphalos,44 whereas exomphalos in the absence of organ overgrowth is observed in the Cdkn1c knockout mouse.45, 46 The phenotypic similarity between patients with CDKN1C mutations and those with BWSICD2 (ie loss of methylation (LOM) of KvDMR1) suggested that the unmethylated DMR may act to silence CDKN1C expression, and consistent with this, patients with KvDMR1 LOM do indeed show reduced fibroblastic CDKN1C expression in culture.25 Our novel observation that mean birth weight centile was higher in the BWSICD1 group than in the BWSMUTCDKN1C and BWSICD2 groups is also consistent with the results of mouse models of BWS (see above) and suggests that although IGF2 and CDKN1C may regulate a common growth control pathway, there are subtle differences in the phenotypic consequences of IGF2 over expression and CDKN1C inactivation/downregulation. Although not all cases of BWS will have a detectable molecular abnormality, molecular diagnosis is possible in most cases. As the results of molecular analysis influence both genetic counselling and surveillance of BWS children, optimum medical management should include molecular genetic analysis.