Transcriptome of DNA Repair Genes in CMML: Rationale for Targeting ERCC1 and CDKN1A As a Synthetic Lethality Approach

Background: Chronic myelomonocytic leukemia (CMML) is characterized by features of both myeloproliferative neoplasm and myelodysplastic syndrome (MDS). Despite the main role of DNA repair pathways in cancer, in myelodysplastic disorders, somatic mutations are not found in these pathways. The novel discovery of non-classical leukaemogenesis mediated by splicing defects, profound influence of epigenetic anomalies, deeper knowledge of the DNA repair network, and development of unbiased high-throughput sequencing approaches, prompted us to revisit whether these processes might be disrupted by non-mutational mechanisms in CMML and related myeloid disorders. Study aims: i) to screen the transcriptome of DNA repair genes in a CMML discovery cohort; ii) to validate their misregulation in an independent large CMML cohort; and iii) determining how those candidates behave through the spectrum of myeloid malignancies. Methods: RNA-seq was performed in the discovery cohort: 27 CMML bone marrow samples (BM) at diagnosis and 9 healthy BM samples, and in 7 patients treated with Azacitidine in BM collected before and after 4 cycles. RT-qPCR was used to validate 9 DEGs through the myeloid spectrum and in a CMML validation cohort: 74 additional CMML, 70 MDS, 66 AML , and 25 CML patients. Candidates were chosen based on clinical considerations: i) druggable oncogenes highly overexpressed ii) oncogenes infra-expressed with inhibitory molecules already being tested in myeloid neoplasms. Global pattern of DNA repair gene expression was compared with MDS and AML extracted from MILE study data. The DEGs methylation status was studied using available public data and we confirmed by pyrosequencing the methylation status of our main candidates. The CMML validation cohort was mutationally characterized by targeted sequencing of 18 CMML recurrently mutated genes. Alternative splice sites, skipped exons, mutually exclusive exons and retained introns were quantified using rMATS.

Results: Of 27 CMML patients and 9 healthy donors, the expression of 30 genes was significantly different between the two groups (8 genes up-regulated and 22 down-regulated). Defects on genes predominantly unique to a single strand breaks repair pathway included: NEIL1 and OGG1 in base excision repair, XPA and MSS19 in nucleotide excision repair and RPA4 in mismatch repair. XRCC4 and MSH4 overexpression and PRKDC down-regulation were found as significant changes among the genes associated exclusively to double strand breaks repair. We next extracted the DNA repair transcriptional components from MILE study: 206 MDS, 47 ckAML and 73 healthy donors. MDS misregulation was characterized by upregulated genes while ckAML showed a global disruption, mostly downregulation of Homologous Recombination genes. Some genes showed opposite sense of misregulation depending on the myeloid entity: TDP1 was upregulated in CMML cases and downregulated in ckAML, and viceversa for BAP1 . CDK1 and EXO1 were upregulated in MDS cases but downregulated in ckAML cases. Statistically significant and direction-concordant dysregulation of CDKN1A, ERCC1, XRCC4, NEIL1, MSH4, PARP1 and BAP1 were confirmed in our independent CMML validation cohort. The methylation study showed a global demethylated status of our misregulated DNA repair genes in CMML. Either responder s (n=3) or non-responders (n=2) to azacitidine failed to achieve a substantial expression change in their DNA Repair DEGs. PARP1 infraexpresion was associated with TET2 mut cases. Lower CDKN1A and BAP1 expression was related to chromatin regulator mutations. SRSF2 mutations induced widespread aberrant splicing events in DNA repair genes in CMML

Conclusions: We have identified by means of an unbiased high-throughput approach, and validated in an independent cohort, a subset of DNA repair genes consistently misregulated in CMML. The genes identified warrant further study as potential novel therapeutic targets, both directly and through modulation of associated compensatory pathways. CDKN1A and ERCC1 in particular emerge as realistic candidates for a synthetic lethality approach. Our findings suggest that the promise of PARP1 inhibition appears less pertinent to CMML than for other myeloid diseases. These differences might also partially explain the different genomic and phenotypic manifestations of different myeloid neoplasms, and shed insights into their distinctive biology.