Germline RBBP6 mutations in familial myeloproliferative neoplasms

To the editor:

Myeloproliferative neoplasms (MPNs) comprise a heterogeneous group of hematologic disorders characterized by clonal overproduction of differentiated myeloid cells, propensity to thrombosis, hemorrhage, and increased risk of leukemia. Three MPN subtypes, polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF) are considered as the “classical BCR-ABL1–negative MPNs,” and they share many clinical and molecular features. Although most cases of MPN are sporadic, several previous studies have shown familial clustering of the disease^1-3  and increased risk of MPN in relatives.⁴  Somatic mutations have been associated with MPN, most notably disease causing, mutually exclusive mutations in Janus kinase 2 (JAK2), calreticulin (CALR), and myeloproliferative leukemia (MPL) genes. Somatic JAK2 mutations are present in ∼60% of MPN cases,^5,6 MPL mutations occur somatically in ∼1% to 5% of cases,⁷  and somatic CALR mutations are found in ∼20% to 30% of ET and PMF.^8,9  Familial MPN is clinically indistinguishable from sporadic MPN, and displays frequent somatic JAK2 and CALR mutations.^2,3,10-12  Other genes that frequently mutate somatically in MPN, such as TET2,¹³ DNMT3A,¹⁴  and ASXL1,¹⁵  do not significantly contribute to familial MPN and in the majority of familial cases the causative germline mutation is unknown.^16,17 

The role of common germline variants in MPN predisposition has been established. Specifically, the JAK2 “GGCC” haplotype increases the risk to develop JAK2-mutant MPN by several fold.¹⁸  Another germline variant in telomerase reverse transcriptase (TERT) (rs2736100) increases the risk of all molecular subtypes of MPN by approximately twofold.^19,20  Both of these single nucleotide polymorphisms (SNPs) contribute to familial clustering of the disease, although they can explain only a minor proportion of it.¹⁹  Recently several additional common variants conferring susceptibility to MPN have been identified.²¹ 

In order to identify the germline mutation predisposing to MPN, we studied an Australian pedigree of English ancestry (Northern England) with 5 members in 4 generations diagnosed with MPN. DNA was available from 3 affected members (Figure 1A). Each of the 3 members of the family had a different MPN-specific somatic mutation (JAK2-V617F, MPL-W515L, and CALR-type 1; Figure 1A). To map the candidate disease loci in the pedigree, we applied a nonparametric algorithm, segregation exclusion analysis (see supplemental Figure 1, available on the Blood Web site), and identified 12 shared genomic regions with a total size of 217.87 Mb among the 3 affected subjects (Figure 1B; supplemental Table 1). As one of these genomic regions was likely to carry the disease-causing mutation, we next applied exome sequencing. After reference alignment, a number of filtering criteria were applied to the detected variants (supplemental Figure 2). We performed Sanger sequencing of all the 18 final candidate variants and confirmed DNA variants segregating with the disease in three genes (RBBP6, C20orf3, and ARMC5) (Figure 1C; supplemental Table 2).

To identify which of the RBBP6, C20orf3, and ARMC5 variants is the one predisposing to MPN, we examined healthy subjects for the presence of the candidate variants. Based on this analysis, ARMC5-P507L was excluded due to 7% frequency in healthy controls, whereas RBBP6-R1569H and C20orf3-D395N were not found in any of the over 700 healthy controls.

Next, we sequenced the exons carrying the identified RBBP6 and C20orf3 mutations in an additional 66 MPN families. This analysis yielded two unique mutations in RBBP6 (E1654G and R1451T; Figure 1D) and 1 polymorphism in C20orf3 (P406L) present in 4% of the healthy subjects. Unfortunately, for both familial cases with the RBBP6 mutation, the DNA sample was available from only 1 affected member; therefore we could not obtain data on the mutation segregation with MPN for these 2 additional families. We used Polyphen 2 and SIFT mutation prediction tools to assess the possible effects of the mutations on protein function. RBBP6-R1569H is predicted to be damaging by both tools, whereas C20orf3-D395N mutation is benign (supplemental Table 3). In conclusion, C20orf3 is unlikely to be involved in familial MPN, whereas RBBP6 mutations remain strong candidates for familial predisposition to MPN.

Due to the low penetrance associated with RBBP6 mutations, establishing the family history of MPN may often be difficult. Therefore, we screened for RBBP6 mutations in 490 sporadic MPN cases. In this analysis, we identified 2 unique germline mutations (S1444F and A1673V) in 3 apparently unrelated subjects and 1 polymorphism (I1661V) present in 0.5% of healthy controls (Table 1; Figure 1E). Overall, we identified 5 different germline RBBP6 mutations associated with MPN and not detected in the general population (Table 1; Figure 1F).

Apart from the 3 affected members, 8 additional healthy family members carried RBBP6-R1569H in the Australian pedigree, consistent with the expected low penetrance.¹⁶  Because common germline SNPs have been shown to contribute to familial MPN,¹⁹  we checked all members of the family for JAK2 GGCC haplotype and the rs2736100_C risk variant in TERT (supplemental Table 4). As expected, from the affected members only the one with JAK2-V617F mutation was heterozygous for JAK2 GGCC haplotype, whereas 2 affected members were heterozygous for TERT risk variant and the third one was homozygous for the TERT risk allele. Although it is not possible to draw any solid conclusions based on a single family, we noticed a trend in JAK2 GGCC haplotype and TERT SNP distribution similar to the pattern we observed in our previous study.¹⁹  The risk allele frequencies for both JAK2 GGCC haplotype and TERT rs2736100 were higher in RBBP6-R1569H mutant MPN patients compared with RBBP6-R1569H mutant unaffected members (16.67% vs 6.25% for JAK2 GGCC haplotype, 66.67% vs 50.00% for TERT SNP).

We have identified germline RBBP6 mutations in ∼5% of familial MPN cases (3/67) and in ∼0.6% of sporadic cases (3/490) where family history is unknown. The low penetrance present in MPN pedigrees suggests that the disease is triggered by some stochastic factors, perhaps the acquisition of somatic mutations. In addition, common germline predisposition factors, such as JAK2 GGCC haplotype and TERT rs2736100 SNP, seem to have an additive effect on the MPN risk in RBBP6 mutation carriers.

RBBP6 is a RING finger E3 ubiquitin ligase located in the nucleus. It has been reported to ubiquitinate and degrade p53, in association with MDM2.²²  Because the RBBP6 mutations identified in our study were all located in the vicinity of its p53-binding domain (Figure 1F), they may affect p53 functions. It is likely that mutant RBBP6 causes an elevation in somatic mutagenesis rates through inhibition of p53 function and deregulation of cell cycle. The existence of somatic mutations in three hallmark MPN genes (JAK2, MPL, and CALR) in a single family might be due to elevated mutagenesis in RBBP6 mutation carriers. Alternatively, RBBP6 mutations might enhance JAK-STAT signaling by a yet unknown mechanism, providing “fertile ground” for MPN development.

We cannot completely exclude the possibility that another unidentified mutation might be segregating with MPN phenotype in the family we studied due to insufficient coverage of some genomic regions by exome sequencing or the mutation being located in a noncoding region. However, our data suggests that RBBP6 is a candidate gene for MPN susceptibility in a subset of pedigrees with familial MPN. The question how RBBP6 mutations predispose predominantly to MPN phenotype remains elusive. There are examples of germline mutations in cancer-associated genes causing a specific familial phenotype, eg, retinoblastoma²³  (RB1), neurofibromatosis²⁴  (NF1), melanoma²⁵  (CDKN2A), and others. Similarly, germline RBBP6 mutations may predominantly predispose toward myeloproliferative phenotypes.