Familial recurrent hydatidiform mole (FRHM) is a rare autosomal recessive condition in which affected women have a predisposition to abnormal pregnancies characterised by complete hydatidiform mole (CHM). In this condition other pregnancy loss is also common, but normal pregnancies are extremely rare.1 CHM (OMIM 231090) are characterised by hydropic swelling and trophoblastic hyperplasia of the placental villi.2 Most occur sporadically and are androgenetic in origin.3 In contrast, CHM in women with FRHM appear genetically normal, with a chromosome complement from each parent.4 5 These diploid biparental CHM not only have a similar pathology to androgenetic CHM, but also show similar imprinting defects. A paternal methylation pattern on both alleles of maternally transcribed genes has been reported in both types of CHM.6^–8 Diploid biparental CHM also have aberrant expression of maternally transcribed proteins, such as CDKN1C, similar to that seen in androgenetic CHM.9

Linkage analysis in families initially localised the gene for FRHM to 19q13.4.10 Refinement of the region using homozygosity mapping11 12 and subsequent screening of candidate genes in the region led to NLRP7 (NACHT, leucine rich repeat and PYD containing 7) being identified as the causative gene.13 To date, mutations in NLRP7 have been reported in only 12 families with CHM of biparental origin.8 13 14 Various mutations have been identified including splice-site mutations, point mutations and an intragenic duplication.8 13 14 The mechanisms by which mutations in NLRP7, a member of a gene family more usually involved in inflammatory and apoptotic pathways,15 result in the characteristic imprinting defects and abnormal development seen in CHM remain elusive.

To better understand the role of NLRP7 in early development, we performed mutation analysis in a large cohort of families from diverse ethnic backgrounds. We also examined the expression of NLRP7 in the ovary and compared this with the expression of other NLRP family member proteins shown to be important in early embryonic development.

METHODS

The study was approved by the Riverside Research Ethics Committee (RREC 2652).

Patients

Seven families, each with two sisters having recurrent CHM, were included in the study (table 1). Five of these have been described previously11 12 16 whereas families 9 and 16 are previously unreported families of South Asian and Caucasian origin respectively. A further 16 women, who had experienced at least 3 CHM pregnancies but had no family history of molar pregnancies, were identified from the databases of the Trophoblastic Tumour Screening and Treatment Centre at the Imperial College Healthcare NHS Trust or had been referred to the Centre for diagnosis during this study. A single woman with 2 CHM and a miscarriage, for which no tissue was available for pathological diagnosis, was also included. These 17 women were from small families and potentially represent single affected members from families with FRHM. Only those women subsequently shown to have molar pregnancies of diploid biparental origin were included in the study. In addition, 25 unaffected relatives from 7 of the families agreed to participate in the study.

Fluorescent microsatellite genotyping of complete hydatidiform moles

To confirm the diagnosis of FRHM, DNA from at least one molar pregnancy in the two previously unreported families and the 17 women with recurrent CHM was genotyped using fluorescent microsatellite genotyping. DNA was extracted from parental blood samples using a Qiagen blood mini kit (Qiagen, Sussex, UK). DNA from CHM was prepared from tissue microdissected from formalin-fixed, paraffin wax-embedded sections as previously described.5 PCR amplification of DNA was performed with a panel of 12 pairs of primers for microsatellite markers on different chromosomes. PCR products were resolved by capillary electrophoresis using a genetic analyser (ABI 310 Genetic Analyzer; Applied Biosystems, Warrington, UK) and genotypes determined with ABI 310 PRISM GeneScan software (Applied Biosystems). The genotype of the molar tissue was compared with that of the parents to determine whether the CHM was androgenentic or biparental.5

Mutation screening of NLRP7

DNA was extracted from blood samples of both affected and unaffected family members as described above. DNA was amplified by PCR, using a panel of 15 primers designed to cover the 11 exons and flanking intronic sequence of NLRP7 (supplementary table 1, online). The annealing temperature used in the PCR reaction was 58°C for all primer pairs except those for exons 2 and 7, for which annealing temperatures were 60°C and 55°C respectively. PCR products were purified using MicroSpin S-400 columns (GE Healthcare, Buckinghamshire, UK), sequenced in forward and reverse orientation using a genetic analyser (3100 or 3700 Genetic Analyzer; Applied Biosystems) and analysed with CodonCode aligner software (CodonCode Corporation, Massachusetts, USA).

SNP genotyping of missense mutations

To determine whether missense mutations identified in this study were likely to be pathogenic or were single-nucleotide polymorphisms (SNPs), SNP genotyping was performed for 192 Caucasian and 198 Asian control chromosomes using a Taqman SNP genotyping assay for each mutation. Each assay was designed using File Builder Software (Applied Biosystems) and submitted to the manufacturer for assay synthesis. Assays were performed and genotypes of the samples determined using the 7500 FAST Real-Time PCR System and SDS software V.1.3 (Applied Biosystems). Details of primers and probes are available upon request.

Expression of NLRP7, NLRP5 and NLRP9 in human ovary

Eight formalin-fixed, paraffin wax-embedded blocks of normal human ovarian tissue, from women aged 30–41 years, were provided by the Imperial College Healthcare NHS Trust Human Biomaterials Resource Centre. Use of the tissue was approved by COREC of Wales (07/MRE09/54). Immunohistochemistry was performed on 5 μm tissue sections using goat polyclonal antibodies against NLRP7, NLRP5 or NLRP9 (Santa Cruz Biotechnology, Santa Cruz, California, USA) at a dilution of 1:100 in phosphate-buffered saline. Appropriate epitope blocking controls were performed to confirm the specificity of each antibody. Goat IgG (1:200; Sigma, Poole, UK) staining was also included as a negative control to confirm that false-positive binding of antibodies was absent. Hydrogen peroxidase-conjugated donkey anti-goat secondary antibody (Santa Cruz Biotechnology) was added at a dilution of 1:200 and incubated for 1 hour, then the slides developed using diaminobenzidine (100 ng/ml; Sigma, Poole, Dorset, UK) as the chromogen. Slides were imaged using a brightfield microscope (Nikon Eclipse E600) and a Nikon digital camera (DXM 1200) (Nikon, Tokyo, Japan)

RESULTS

Fluorescent microsatellite genotyping of complete hydatidiform moles

In total, 11 CHM from the two previously unreported families with two affected sisters were genotyped and confirmed to be diploid biparental CHM. CHM from a further 16 women with ⩾3 pathologically confirmed CHM and no normal pregnancies were also genotyped. In 12 families, all CHM were found to be diploid biparental, confirming a diagnosis of FRHM, thus these 12 were included in the study. All CHM in the remaining four families were shown to be androgenetic and these cases were excluded. Finally, a woman with two CHM and a miscarriage of unknown pathology was included after both CHM were shown to be of diploid biparental origin.

Mutation screening of NLRP7

In total, 20 families with a confirmed diagnosis of FRHM were available for screening (table 1). In 17 of the families, 16 different mutations were identified, including 7 novel mutation, predicted to give rise to premature stop codons and a truncated protein, and 6 novel missense mutations (table 1, fig 1). In all families in which two affected sisters were screened, identical mutations were found in both sisters. The parents were screened in four of these families (4, 6, 9, 11) and shown to be heterozygous for the same mutation, confirming normal mendelian inheritance of the defective gene.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Schematic representation of NLRP7 mutations identified in the present study on a simplified domain model of the protein. The mutations identified in 17 families with FRHM are annotated with reference to NP_996611.2. Novel mutations are shown in bold. (A) Nine missense mutations identified in the present study. (B) The seven nonsense mutations identified. (C) Predicted protein domains PYRIN (PYD), NACHT and leucine-rich region (LRR) of NLRP7.

Only 3 of the 16 mutations identified in the present study (p.R693W, p.R693P and p.N913S), have been previously described.13 The proband in family 10 was of interest in that she was homozygous for the p.R693P mutation but also carried a single copy of the p.N913S mutation. Sequencing of DNA from other family members confirmed inheritance of the p.R693P mutation from her mother and both the p.R693P and p.N913S mutation from her father. Both these variants have been found in the same individual in a previous case of Asian origin.13 This woman was described as a compound heterozygote although no evidence was provided to show that the two mutations were inherited independently. One other mutation occurred in >1 case in the present study: a novel 14 base duplication in exon 4 (fig 2) that was found in three Caucasian families. This was homozygous in two families (2 and 3) and heterozygous in a third (16). In family 16, both sisters were also heterozygous for a single base deletion, c.2030delT, suggesting that they were compound heterozygotes. Two further families (15 and 17) were apparent compound heterozygotes. Screening showed that the proband in family 17 had inherited the p.R432X mutation from her father and the novel missense mutation, p.R693Q, from her mother. Parental DNA samples were not available for families 15 and 16. However, screening of DNA from all three CHM from family 15 and one CHM from family 16 identified only a single mutation in each CHM, confirming that the mutations were not carried on the same haplotype and that these cases were compound heterozygotes. To date, no mutations have been found in three families investigated (families 18–20).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Partial sequencing chromatogram of exon 4 from NLRP7. Genomic sequence of (A) control and (B) proband in family 3. The protein product of both sequences is represented by the first three letters of the amino acid encoded by the sequence. The affected woman has a homozygous insertion of 14 bp, resulting in a premature stop codon.

Five sisters with normal reproductive outcomes were screened in three families (4, 9 and 17). Two were found to be carriers and three had no mutations. Six brothers from three families ( 9, 10 and 11) were also screened. One man (from family 9) was found to be homozygous for the same p.R693P mutation as his two sisters. Although the sisters had a history of nine pregnancies, all of which resulted in a CHM, their brother has a normal son.

SNP genotyping of missense mutations

SNP genotyping of DNA from patients homozygous or heterozygous for missense mutations (p.L398R, p.P651S, p.R693W, p.R693P, p.R693Q, p.P716A, p.R721W, p.N913S, p.C761Y) confirmed the presence of the expected mutations in affectd and unaffected family members. None of these mutations was identified in 192 Caucasian or 198 Asian control chromosomes.

Expression of NLRP7, NLRP5 and NLRP9 in human ovary

Within the ovary, expression of NLRP7, NLRP5 and NLRP9 was confined to the oocytes. Strong cytoplasmic staining was identified in all stages of growing oocytes, including primordial, primary, secondary and tertiary follicles, in all tissue samples examined (fig 3). The three NLRP proteins had a similar pattern of expression within the oocytes. No expression was seen in the surrounding granulosa or stromal cells. Specific epitope peptides used to raise the antibodies successfully inhibited positive staining when added to the primary antibodies, confirming the specificity of the antibodies (fig 3). No false positive binding of NLRP antibodies to the tissue was found with goat IgG staining.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Comparative expression and localisation of NLRP7, NLRP5 and NLRP9 in human ovary. (A–F’) NLRP7 expression is cytoplasmic in all stages of growing oocytes. Protein expression can be seen in primordial oocytes (PrM), primary follicles (PF), secondary follicles (SF) and tertiary follicles (TF). The nucleus (ns) is clearly visible in all oocytes. The oocyte in (F′) is a higher magnification of the TF in (F). (G–I) NLRP5 protein expression in growing follicles. (J–L) NLRP9 protein expression in growing follicles showing similar cytoplasmic expression. (M–O) Negative control sections after epitope peptide competition staining for each antibody. (P) Negative staining with Goat IgG. Scale bars, 50 μm.

DISCUSSION

NLRP7 (NOD12, PYPAF3 and CLR19.4) is a cytoplasmic protein belonging to a group of proteins made up of an N-terminal pyrin (PYD) domain, a NACHT domain and a C-terminal leucine-rich repeat (LRR) domain.18 Mutations of NLRP7 have recently been described in a small number of families with FRHM. In the present study a further 20 families have been investigated, 7 with two affected sisters and 13 in which a single woman has had recurrent CHM of diploid biparental origin. Seventeen of these families were found to have mutations in NLRP7, confirming the importance of NLRP7 in the aetiology of FRHM. Sixteen different mutations were identified in the present study, demonstrating the wide diversity of mutations in NLRP7 that give rise to FRHM.

Despite the diversity of mutations and that consanguinity was reported in only one family (4), affected members in 14 of the 17 families were found to be homozygous for the mutation identified. Two of these families, from the same region of Turkey (7 and 8), were not only homozygous for the same p.R693W mutation but shared an identical haplotype across the 19q13.4 region flanking NLRP7 (unpublished observation) suggesting that, even in non-consanguineous families, mutations within families might be inherited from a common ancestor. Affected members of three families were found to be compound heterozygotes. One individual surprisingly had three NLRP7 mutations, being homozygous for the p.R693P mutation and heterozygous for the p.N913S mutation. Genotyping of other family members found that all three mutations were inherited from the patient’s parents and did not arise de novo. Although most mutations described in the present study were novel, both these mutations had been previously reported in women from the same ethnic background, suggesting that these particular mutations are more common in some populations.

No mutations were identified in three families in the present study. However, mutations in NLRP7 cannot be completely excluded. Deletions or intragenic duplications, involving whole exons, might not be identified by conventional sequencing. In a previous study, three Egyptian families, with no mutations in the coding sequence of NLRP7, were subsequently found to have an intragenic duplication involving several exons.8 It is also possible that the mutations might be in unidentified regulatory elements of the gene as only exonic sequences were screened in the present study. A more intriguing possibility is that a second gene might be associated with this condition in some families. In family 18, in which no NLRP7 mutations were identified, the two sisters inherited only one haplotype in common from their parents for the 19q13.4 region, suggesting that the causative gene in this family is unlikely to map to this region.16 The two other cases, without identifiable mutations, are also heterozygous across the 19q13.4 region and for several SNPs within NLRP7. Although the possibility that the women are compound heterozygotes for mutations missed using the current sequencing strategy, heterozygosity across the region is in favour of a mutation in a second locus outside the region.

Three mutations in the present study have been previously described.8 13 14 Two of these, p.R693W and p.R693P, occurred in five families in the present study and three of six previously reported families with missense mutations.8 13 14 A sixth patient in the present series had a novel mutation p.R693Q affecting the same codon. This codon, found in the LRR, would thus seem to be a hot spot for mutations in NLRP7. All other missense mutations in the present study, with the exception of the p.L398R mutation, were also located within the LRR of NLRP7. None of these mutations were found in 390 unrelated control chromosomes or described in the Infevers Autoinflammatory Mutation Online Registry19 which includes known variants of NLRP7. Two previously reported missense mutations p.A657V and p.L750V,8 14 not seen in this study, are also located in the LRR. In addition to missense mutations, the present study found six novel nonsense mutations occurring at intervals throughout the gene. None of these mutated genes would, if translated, give rise to a protein with a functional LRR. This provides very strong evidence that the LRR, which is involved in protein–protein interactions,20 is important for normal functioning of NLRP7 in reproduction.

NLRP7 has two major transcript variants. These are similar except that only variant 1 contains exon 10, whereas variant 2 has an additional 84 bases at the 5′ end of exon 5. In family 6, the missense mutation, p.P651S, occurred in the 5′ end of exon 5 found only in transcript 2. Similarly the p.A657V mutation previously described14 would only affect normal functioning of transcript 2. Thus it would seem to be the product of variant 2 that is important in normal reproduction. Identification of the ligand to which the LRR of this protein binds will be important to understanding the role of NLRP7 in pregnancy.

Most pregnancies in affected women were CHM although other reproductive losses including miscarriages, partial hydatidiform mole and a stillbirth were reported. Some pregnancies, reported as miscarriage, may not have been examined pathologically and may represent missed CHM. However, we have confirmed at least one miscarriage as a hydropic abortion, based on its pathology. p57^KIP2 immunostaining correlated with the pathology in this case, with the villous tissue showing positive staining. This, together with the observation that the methylation status of imprinted genes was normal in the live offspring of one affected woman,7 suggests that the variable phenotypes of the pregnancies in this condition reflect different degrees of aberrant imprinting. This may in turn be influenced by the specific mutation present in the family. In the present series, none of the affected women had pregnancies resulting in normal live births. However, none of the families had the unique splice-site mutation reported in the family in which affected women achieved occasional normal pregnancies.13 Owing to the large number of different mutations and the relatively small numbers of non-CHM pregnancies it was not possible to identify specific correlations between different mutations and reproductive outcome from the present series.

In the course of the study parental DNA was examined in six families (4, 6, 9, 10, 11 and 17) and the mothers shown to be heterozygotes. Two sisters with normal live births were also heterozygous for one of the mutations in their affected sister. There was no history of molar pregnancies in any of these heterozygous women and the only reproductive loss reported was a single miscarriage in the mother of two affected sisters who had six other pregnancies. Although it is possible that not all miscarriages were reported, in the present study there was no history of molar pregnancy or any evidence of increased reproductive wastage associated with being a heterozygote for any of the mutations identified.

Functionally, little is known about NLRP7. Some members of the NLRP family have been implicated in the assembly of inflammasomes and maturation of proinflammatory cytokines.21 Mutations in these genes, in particular gain-of-function mutations in the NACHT domain of NLRP3, have been shown to cause various autoinflammatory disorders.21 However, FRHM is an autosomal recessive disorder in which mutations occur predominantly in the LRR. Loss of NLRP7 does not seem to affect systemic immunity directly but instead results in aberrant imprinting similar to that seen in androgenetic conceptuses. Both androgenetic CHM and the diploid biparental CHM found in women with FRHM are characterised by abnormal expression of imprinted genes.9 In addition, several imprinted genes that are normally maternally imprinted have a paternal methylation pattern and remain unmethylated in both androgenetic and diploid biparental CHM.6^–8 These observations fit the hypothesis that the normal role of the defective gene in women with FRHM is in establishing or maintaining the maternal imprint. In both males and females, methylation marks are removed in the primordial germ cells and re-established during gametogenesis. In female mice this is achieved by sequential methylation of imprinted genes during oocyte growth.22 Factors important for differential methylation should therefore be present at all stages of oocyte development.

Although there is no mouse orthologue of NLRP7, some members of the NLRP family have been shown to be specifically expressed in the gonads and play a role in reproduction in mice.23^–25 Several of these, including NLRP5 (MATER) and NLRP9, are also expressed in the oocytes of both rhesus macaque monkeys and humans.26 27 In primates, transcripts of NLRP7 have been reported in ovary and in denuded oocytes at the germinal vesicle and metaphase I stages.13 15 27 28 However, the localisation of the protein has not previously been described. In the present study immunohistochemical staining of human ovarian tissue showed that NLRP7 expression is confined to the oocytes, is expressed at all stages from primary to tertiary follicles, and is similar to that of NLRP5 and NLRP9. This pattern of expression is consistent with a role for NLRP7 in the developing oocyte. In mice, NLRP transcripts are reportedly downregulated immediately after fertilisation,25 whereas in primates, NLRP transcripts (including NLRP7) seem to persist in the early cleavage embryos of both rhesus macaque monkeys and humans,13 27 28 and could therefore also play a role in establishment or maintenance of imprints. Further studies are needed to determine the precise mechanisms by which NLRP7 functions to control imprinting.

At present, counselling of families with FRHM is difficult. If the major role of NLRP7 is in setting the maternal imprint in the oocyte, egg donation should provide a strategy to avoid further CHM. However, if NLRP7 has other roles in the establishment and maintenance of pregnancy, egg donation might still result in pregnancy failure in women with FRHM. Women with this condition need to be aware that the most likely outcome of any subsequent pregnancy is an HM and the risk of persistent trophoblastic disease (PTD) if this proves to be the case. In the present series, there were an appreciable number of patients who experienced PTD requiring chemotherapy (table 1). In all cases, PTD followed a CHM pregnancy. Of the 94 CHM reported in the series, 15 (16%) were followed by PTD, an incidence similar to that seen after sporadic CHM.29

In conclusion, this study has identified 13 novel mutations in NLRP7: 7 nonsense and 6 missense mutations. Identification of a male homozygous for a mutation in NLRP7 but with normal reproductive outcomes confirms that mutations in NLRP7 affect only women. The demonstration that mutations occur in transcript variant 2 and cluster in the LRR define the LRR of transcript variant 2 as the critical domain for normal reproductive function of NLRP7.