Targeting HMGA1 Epigenetic Regulators and Downstream ETV5 Pathways with Histone Deacetylase Inhibitors in Combination Therapy for Relapsed, Pediatric B-ALL

Introduction: Despite advances in therapy for pediatric B-cell acute lymphoblastic leukemia (pB-ALL), relapse remains a leading cause of death in children with cancer. The gene encoding the High Mobility Group A1 (HMGA1) chromatin regulator is highly expressed in stem cells and hematologic malignancies where high levels portend poor outcomes (Li et al. Blood 2022). We discovered that transgenic mice overexpressing Hmga1 in lymphoid cells develop leukemic transformation by amplifying transcriptional networks involved in stem cell function, proliferation, and inflammation (Hillion et al, Cancer Res 2008, Schuldenfrei et al, BMC Genomics 2011). Further, HMGA1 is overexpressed in pB-ALL with highest levels after early relapse (Roy et al, Leuk Lymphoma 2013), suggesting that HMGA1 drives relapse. We therefore sought to: 1) test the hypothesis that HMGA1 is required for relapse in pB-ALL, and, 2) identify targetable mechanisms underlying HMGA1.

Methods: To elucidate HMGA1 function in pB-ALL, we silenced HMGA1 using CRISPR or short hairpin RNA (shRNA) in cell lines from relapsed pB-ALL, including REH, which harbor the TEL-AML1 fusion, and 697, which harbor the E2A-PBX1 fusion. We assessed leukemia phenotypes in vitro and leukemic engraftment in vivo. To dissect molecular mechanisms, we integrated multiomic sequencing (seq) approaches, including RNAseq and ATACseq. To identify pathways relevant to patients, we assessed HMGA1 gene networks and clinical outcomes in pB-ALL cohorts. Artificial intelligence (AI) was used to predict drugs that target HMGA1 networks which we tested in combination therapy in patient-derived xenograft (PDX) models.

Results: Silencing HMGA1 expression via CRISPR or shRNA in relapsed pB-ALL cell lines (REH, 697) disrupts proliferation and clonogenicity while increasing apoptosis. To assess HMGA1 function in vivo, we compared leukemogenesis following tail vein injection of pB-ALL cells with or without HMGA1 depletion in immunodeficient mice (NOD/SCID/IL2 receptor gamma^null). Strikingly, survival is prolonged in mice injected with either pB-ALL cell line (REH, 697) after HMGA1 depletion. Further, leukemic cells that ultimately engraft show increased HMGA1 expression relative to the pool of injected cells with HMGA1 silencing, suggesting that escape from HMGA1 silencing is required for engraftment. RNA/ATACseq revealed transcriptional networks governed by HMGA1 that regulate proliferation (G2M checkpoint, E2F), RAS/ERK signaling, hematopoietic stem cells, and ETS variant 5 transcription factor (ETV5) targets. We focused on ETV5 given its role in role in HSC self-renewal and association with other forms of aggressive leukemia. CRISPR-mediated ETV5 gene silencing recapitulates effects of HMGA1 silencing in pB-ALL cell lines, decreasing proliferation and clonogenicity in vitro, while delaying leukemogenesis in vivo. Further, restoring ETV5 expression in pB-ALL cell lines with HMGA1 silencing partially rescues anti-leukemogenic effects of HMGA1 depletion. Mechanistically, HMGA1 binds to AT-rich regions within the ETV5 promoter (-0.7 kb, -0.2 kb) and recruits activating histone marks (H3K27Ac, H3K4me3, H3K4me1) to induce ETV5. Moreover, HMGA1 and ETV5 are co-expressed and up-regulated in primary blasts from children with relapsed pB-ALL where high HMGA1 correlates inversely with time to relapse. AI predicted that the histone deacetylase inhibitor (HDACi), vorinostat, targets HMGA1-ETV5 gene networks. In PDX models of pB-ALL from children who relapsed, we found that vorinostat enhances responses to standard B-ALL therapy (dexamethasone, vincristine), decreasing leukemic burdens beyond that of conventional combination therapy.

Conclusions: We discovered a previously unknown epigenetic program whereby HMGA1 up-regulates ETV5 networks by binding to chromatin and recruiting active histone marks to ETV5 promoter-enhancer regions. Both HMGA1 and ETV5 are up-regulated at relapse and high HMGA1 levels predict decreased time to relapse. Finally, HMGA1-ETV5 networks can be targeted by the HDACi, vorinostat, which enhances responses to standard pB-ALL therapy in PDX models. Our findings reveal HMGA1 as an epigenetic switch that activates ETV5 and HSC gene networks in relapsed pB-ALL and promising therapeutic target to treat or prevent relapse.