The presence of two mutually exclusive recurrent chromosomal rearrangements, t(1;17) and inv(7), in BPOP have been documented by several cytogenetic studies.9–12 However, the exact genomic breakpoints or fusion genes have not previously been identified. In this study, we identified COL1A1::MIR29B2CHG and COL1A2::LINC-PINT as the fusion genes corresponding to t(1;17) and inv(7), respectively.
In 2004, Nilsson et al. first identified t(1;17)(q32;q21) as the sole cytogenetic abnormality in a case of BPOP.6 Further FISH analyses of 4 additional cases showed that all cases had a 1q32 split, while all but one had a 17q21 split, suggesting a recurrent genetic event involving these loci in BPOP (Table 1). Two subsequent reports supported this finding, describing t(1;17)(q42;q23) in a BPOP of the toe and t(1:17)(q32;q21) in a BPOP in the distal ulna.7,8 The genomic breakpoints were determined by FISH mapping, and approximated to be within BAC clone RP11-99A19 on 1q32 and RP11-219F9 on 17q21.6 The 1q32 locus covered by RP11-99A19 encompasses 3 genes: CR1L, CD46 (formerly known as MCP), and MIR29B2CHG. Our data confirmed MIR29B2CHG rearrangement as the critical genetic event in BPOP at the RNA level. As for RP11-219F9 on 17q21, we were not able to find this BAC clone in the current UCSC hg19 and hg38 database to verify its location. In the original study, it was described to be distal to RP11-243D13 on 17q21.32,6 compatible with the location of COL1A1 (17q21.33) as identified in our case 4. Although the authors stated that RP11-219F9 covered genes such as NBR2 and NSF, but not COL1A1, we speculate that the sequence alignments back then were imprecise, as the NBR2 and NSF genes lie proximal to RP11-243D13 in the current human genome assembly hg19 and hg38.
The other fusion gene identified in this study, COL1A2::LINC-PINT, matched the cytogenetic abnormality of inv(7)(q21.1-22q31.3-32) in BPOP. Intriguingly, in 2 of the 4 reported BPOPs with inv(7), the lesions showed central continuity with the underlying medullary cavity of bone, an uncommon finding in BPOP.9,10 Our case 1 with available radiographic information also showed corticomedullary continuity with the underlying bone. Given the limited case number and clinical information available in this study, it remains to be clarified whether this radiographic finding is predictive of the genotype. In the literature, two of the BPOPs reported with inv(7) also had inv(6)(p25q15),9,11 which was not inferred from our RNA sequencing data.
The pathogenetic mechanisms of these novel fusions remain to be studied. COL1A1 and COL1A2 gene rearrangements are most commonly seen in fibroblastic tumors, such as COL1A1::PDGFB and rarely COL1A2::PDGFB in dermatofibrosarcoma protuberans, COL1A1::USP6 in aneurysmal bone cyst, myositis ossificans, and fibro-osseous pseudotumor of digits, and rare COL1A2::USP6 in nodular fasciitis, as well as occasionally in non-fibroblastic tumors such as COL1A2::PLAG1 fusions in lipoblastoma.18–24 In these fusions, COL1A1 and COL1A2 often contribute their active promoters, which result in the overexpression of the 3’ fusion partner genes. Indeed, in case 4 the expression of MIR29B2CHG was markedly elevated in comparison with all other mesenchymal tumors subjected to the same RNA exome platform (Supplementary Table 3). Meanwhile, however, LINC-PINT expression approached zero in case 1. Moreover, the expression levels of COL1A1/2 exceeded their fusion partners by at least 3 orders of magnitude in both cases. Collectively, we doubt that the tumorigenic mechanisms mediated by the two novel fusions could be simply or even correctly attributed to the aforementioned “active promoter” model.
Instead, we hypothesize that these gene fusions resulted in the disruption of coding regions of COL1A1 and COL1A2, and hence the production of truncated and dysfunctional type I collagen peptides. For each fusion transcript detected, the collagen proteins are predicted to lose a significant portion from either the N-terminus or C-terminus (Supplementary Table 2). Furthermore, in the fusion transcripts where COL1A2 and COL1A1 were the 5’ fusion partners, stop codons were observed in the 3’ fusion partner genes, not long after the fusion junction (Figs. 2 and 3, and Supplementary Table 2). This suggests a tumorigenic mechanism contributed by the non-coding RNA genes, when serving as the 3’ partner, not much more than providing the stop codons. Importantly, the frameshift mutations of COL1A1 identified in case 5 further supported the possible role of type I collagen disruption in BPOP formation. Although these fusions and mutations are predicted to be heterozygous, the altered collagen products might exert dominant-negative effects that lead to abnormal assembly of collagen fibrils, similar to COL2A1 mutations in a subset of chondrosarcoma.25 In case 5 with COL1A1 mutations, we were unable to determine the germline genotype of the patient because normal DNA material was unavailable; however, since this patient has no known phenotypic association with syndromic diseases, it is most likely that these frameshift mutations are somatic.
Interestingly, rearrangements of another collagen gene, COL12A1, have also been identified in subungual exostosis, another bone surface osteochondromatous lesion bearing some histomorphologic similarity to BPOP. 26 Earlier cytogenetic studies revealed recurrent t(X;6)(q25-26;q15-21) in subungual exostosis.3, 27 Using FISH mapping, Storlazzi et al. identified COL12A1 as the breakpoint on chromosome 6, and the breakpoint on chromosome X were found to be COL4A5 in 4 cases and the adjacent IRS4 in 1 case.26 Subsequently, up-regulation of IRS4 was observed in subungual exostosis compared to other chondromatous/osteochondromatous tumors, including BPOP and osteochondroma.28 Together with our findings, it is plausible that BPOP and subungual exostosis are distinct entities with different genetic abnormalities, although both involve alterations in collagen genes.
Both fusion partners, LINC-PINT and MIR29B2CHG, are non-coding RNA genes. LINC-PINT (long intergenic non-protein coding RNA, p53 induced transcript) is a long non-coding RNA functioning as a tumor suppressor relevant in various cancers, including osteosarcoma.29, 30 MIR29B2CHG (also known as C1orf132) is the host gene of miR-29b2 and miR-29c, and has been shown to be down-regulated as an alternative longer transcript in triple-negative breast cancer.31 We speculate that the gene fusions might also disturb the tumor-suppressing function of these non-coding RNAs. Of potential importance, exon 1 of LINC-PINT contains a highly conserved sequence element essential for its ability to inhibit tumor invasiveness,32 which would be lost in the COL1A2::LINC-PINT fusion. It is noteworthy that the COL1A2::LINC-PINT transcript was not detected in case 3 by RNA sequencing using oligo(dT) selection, implying that the reciprocal transcript LINC-PINT::COL1A2 might be more critical in the tumorigenesis of BPOP. However, it has been known that the 3’ poly-A tails might be very short or even absent in some lncRNAs, thus excluding these lncRNAs from the sequencing libraries with oligo(dT) selection. Whether this could account for the failure to detect the COL1A2::LINC-PINT fusion remains to be determined.
Other gene fusions involving non-coding RNA genes have also been reported, such as MALAT1::GLI1 fusion in plexiform fibromyxoma and gastroblastoma of stomach,33, 34 MALAT1::TFEB in renal cell carcinoma,35, 36 and MIR143::NOTCH1/2/3 fusion in glomus tumor.37 It has been suggested that these fusions could use the active promoters of the 5’ non-coding RNA genes to upregulate the 3’ oncogenes. Nonetheless, we do not find it quite as plausible that this could apply to the fusions identified in our current study, given the much lower expression levels of these RNA genes than COL1A1/2 and the predicted truncation of the collagen genes, as mentioned above.
In summary, our study revealed that a majority of BPOP cases, including all those subjected to next-generation sequencing, had COL1A1/2 fusions or mutations, which was in keeping with the high prevalence of t(1;17) or inv(7) described in the original paper.6 It is possible that a small subset of fusion-negative BPOP might have a cryptic fusion or mutations of either gene, which could elude detection by FISH or even RNA sequencing and therefore might require appropriate next-generation sequencing methods to identify. In addition, the frequently suboptimal tissue quality, probably due to previous decalcification, has prevented us from genotyping a significant subset of cases. The prevalence of these genetic events remains to be further elucidated in a more systematic fashion.
In conclusion, we identified novel COL1A1::MIR29B2CHG and COL1A2::LINC-PINT fusions in BPOP. The locations of these genes were consistent with the previous cytogenetic reports of t(1;17) and inv(7) in BPOP. The current findings further solidify the neoplastic nature of BPOP and shed light on the potential roles of damaged type I collagen in BPOP formation. Further investigations are required to examine the prevalence of these alterations, the contribution of the non-coding RNA fusion partners, other alternative tumorigenic mechanisms in fusion-negative BPOP, and the correlation between genotypes and various clinicopathologic features.