Khsrp depletion selectively eliminates AML cells that persist after venetoclax/azacitidine therapy via modulating SLC23A1-transported TET enzyme cofactors

Abstract

Abstract Despite improved targeted therapies and chemotherapy regimens for acute myeloid leukemia (AML), durable remissions remain elusive for most patients. A major barrier to curative treatment is therapy-persistent leukemia cells (residual disease) that evade initial treatment and drive relapse. We hypothesized that these persistent cells harbor distinct biological features that enable them to survive initial therapy. To identify selective vulnerabilities, we conducted a genome-wide CRISPR interference (CRISPRi) screen in AML cells that survived treatment. In parallel, we established an in vivo patient-derived xenograft (PDX) model of residual disease that models clinical disease scenarios and enables transcriptomic profiling at key treatment stages. We first performed CRISPRi screening in CAL-1 cells, derived from BPDCN, a representative high-risk AML subtype that responds to most initial therapies but rapidly relapses. Two regimens were tested, venetoclax plus azacitidine (VEN/AZA) and tagraxofusp, each at its IC90 dose, to explore shared targets in residual disease. Our screen identified several expected pathways — pro-apoptotic genes such as BAX and NOXA as mediators of sensitivity to VEN/AZA, and CD123 and diphthamide biosynthesis genes as sensitizers to tagraxofusp. Notably, KH-type splicing regulatory protein (KHSRP) was the top gene promoting therapy persistence across both regimens. In validation experiments, KHSRP depletion substantially reduced residual leukemia viability across several regimens in multiple AML cell lines derived from other high-risk disease subtypes, achieving up to 10 times more residual cell elimination after initial treatment vs. controls. Using the in vivo PDX model, we defined three time points: pre-treatment, residual disease after VEN/AZA treatment, and relapse. Pre-treatment and relapse had >80% marrow burden, whereas residual disease was <20%. RNA-seq revealed that residual cells were transcriptionally distinct from both primary and relapsed leukemia. We then integrated PDX expression analysis with gene expression in AML cell lines after KHSRP depletion. This turned our focus to SLC23A1, which encodes the sodium-dependent vitamin C transporter 1, as it was the only gene significantly downregulated in residual disease and upregulated by KHSRP depletion. Mechanistically, we found that KHSRP promotes SLC23A1 mRNA decay via recruitment of the NEXT nuclear RNA degradation complex. We thus hypothesized that residual leukemia cells reduce SLC23A1 to resist chemotherapy, and that KHSRP depletion restores its expression, sensitizing cells to treatment. This model was supported by rescue experiments in which SLC23A1 knockdown abolished the therapy-enhancing effects of KHSRP depletion. Vitamin C has been explored as an anti-leukemic agent via promoting the conversion of 5-methylcytosine (5-mC) to 5-hydroxymethylcytosine (5-hmC), facilitating DNA demethylation, particularly in TET2 mutant cells. Given that the active, reduced form of vitamin C is unstable in standard culture media, we tested the role of uric acid, the other molecule transported by SLC23A1 and that is present in the media at stable physiologic concentrations via FBS supplementation. Similar to vitamin C, we found that soluble uric acid also enhanced 5-mC to 5-hmC conversion in AML cells. Furthermore, exogenous vitamin C and uric acid were significantly synergistic in reducing viability of KHSRP-deficient residual cells after VEN/AZA treatment. Moreover, TET2 re-expression in TET2-null AML cells mimicked the effects of KHSRP depletion in eliminating the residual population, with no additional effect observed upon further KHSRP knockdown. Thus, uric acid and vitamin C may serve as alternative TET enzyme cofactors that mediate the therapeutic effects of KHSRP depletion. Importantly, in the Cancer Dependency Map, only a few solid tumor cell models are strongly dependent on KHSRP, and our data showed that KHSRP depletion had minimal effects on untreated leukemia cells. In contrast, the impact of KHSRP depletion on eliminating therapy-persistent leukemia cells was chemotherapy dose-dependent, indicating a selective sensitivity in residual disease. We are currently evaluating the effects of KHSRP depletion in prolonging survival of two independent AML in vivo models treated with VEN/AZA. These data nominate KHSRP as a selective vulnerability in therapy-persistent AML and warrant the development of KHSRP inhibitors to target residual disease.