Multimodal assessment of chronic myelomonocytic leukemia with extramedullary involvement reveals RAS pathway activation, adverse outcomes and epithelial-mesenchymal transition

Abstract Background: Extramedullary disease (EMD) in chronic myelomonocytic leukemia (CMML) patients (pts) remains poorly characterized with limited data on its genetic, phenotypic, and biological characteristics, and its clinical implications. Further understanding of EMD in CMML could inform therapeutic strategies aimed at preventing or treating EMD in CMML. Methods We evaluated 574 pts with CMML and further selected pts with biopsy-proven EMD. Clinically detected hepatosplenomegaly or biopsy-proven splenic EMD were not included. Histopathologic review was performed independently by two hematopathologists. Immunohistochemistry and/or flow cytometry (FC) were used to determine the immunophenotype of EMD. CD56 expression in bone marrow (BM) monocyte and blast populations was evaluated by FC. Next generation sequencing (NGS) was performed on BM specimens (n=452) at time of diagnosis and on EMD sites (n=5). Variant allele frequency (VAF) estimates were used to evaluate clonal relationships. Spatial transcriptomics (Visium HD) was applied to EMD tissue sites in two representative cases. Results Among 63 patients with EMD-CMML, 36 (57%) had myeloproliferative (MP) CMML, and 11 (18%) had CMML-2. EMD-CMML pts had similar baseline characteristics compared to those without EMD (n=511) except for male predominance (79% vs 69%, p=0.003). Renal failure at time of EMD diagnosis was observed in 18/52 (35%) evaluable pts. No significant differences in CD56+ monocyte or CD56+/CD34+ blast frequencies were observed in EMD, although numerically higher in MP-CMML and RAS pathway (RASp) mutant CMML pts. CMML with EMD had higher frequency of RASp mutations (69% vs 47%, p=0.009), particularly KRAS (36% vs 16%, p=0.002) and PTPN11 (11% vs 4%, p=0.055), and numerically higher STAG2 mutation frequency (11% vs 3%, p=0.060). Median VAF for KRAS mutations was numerically higher in EMD-CMML (35% vs 30%, p=0.4947). Among 26 (41%) EMD-CMML pts with evaluable BM mutation clonal hierarchy, RASp mutations were dominant in 13 (50%), with KRAS dominant in 55% of KRAS-mutant cases and the sole somatic mutation in one patient. NGS of EMD specimens (n=5) revealed RASp mutations in all cases (KRAS n=3; NRAS n=2; NF1 n=1; PTPN11 n=1). EMD sites included skin (62%), lymph nodes (13%), pleural fluid (8%), central nervous system (6%), soft tissues (5%), liver (3%), bones (3%), ascitic fluid (2%) and mediastinum (2%). Phenotypes were diverse: monocytic (48%), myelomonocytic (11%), neutrophilic (10%), blastic plasmacytoid dendritic cell neoplasm (8%), indeterminate cell (6%), Rosai-Dorfman like histiocytic (5%), stem-like myeloid (3%), mature plasmacytoid dendritic cell (pDC) (3%) and leukemia cutis not-otherwise specified (8%). EMD-CMML pts were treated with hypomethylating agents (63%), low-dose cytarabine-regimens (7%), intensive chemotherapy (5%), investigational agents (12%) or hydroxyurea (2%). With a median follow up of 54 months (95% CI 44-63 months), EMD-CMML pts had a shorter median OS than did pts without EMD (26 vs 44 months, p=0.002) and a higher likelihood of transformation to acute myeloid leukemia (46% vs 19%, p0.001). EMD manifestations emerged at time of transformation in 7 (11%) pts. Median OS from EMD onset was 9 months (95% CI 2-17 months). Spatial transcriptomics in skin tissue of one case with monocytic differentiation revealed dermal leukemic cell infiltrates with high expression of genes involved in TNF-a signaling via NF-kB, interferon and KRAS signaling (e.g MCL1, CD74, CXCL10 and IRF7) and epithelial-mesenchymal transition (EMT), including LGALS1. Leukemic clusters also expressed high levels of thymosin beta-4 (TMSB4X) RNA, known to mediate cell migration, modulation of tissue microenvironment and induction of EMT. Similar findings were observed in a case with myelomonocytic leukemic infiltration in which monocytic leukemic cells (expressing LYZ, CD63 and S100A6) and immature myeloid leukemic cells (expressing MPO and AZU1) also expressed high levels of TMSB4X. Immune cells characterized by CD4, CD8A and GZMB expression were adjacent to and admixed with the leukemic cells. Spatial topography of leukemic cells was confirmed by morphologic evaluation of serial hematoxylin & eosin-stained tissue sections. Summary/Conclusion EMD in CMML is associated with RAS pathway mutations, adverse outcomes and diverse phenotypes. Spatial transcriptomics suggests that EMT programs may drive tissue infiltration, highlighting potential therapeutic targets.