Functional genomics studies identify determinants of response vs. resistance to pharmacological inhibitors of KRAS in multiple myeloma

Abstract

Abstract Up to 50% of multiple myeloma (MM) patients harbor mutations of KRAS or NRAS. Pharmacologic inhibitors targeting specific KRAS mutations have clinical activity in solid tumors. We evaluated Ras inhibitors in preclinical models of MM, with emphasis on genome-scale CRISPR studies to define the molecular determinants of response and resistance to these agents. We studied selective inhibitors of KRAS G12C (MRTX-1257) or G12D (MRTX-1133) mutants; the broader spectrum mutant-KRAS inhibitor BI-2865; or the tricomplex pan-Ras inhibitor RMC-6236. These agents were active against human MM lines at clinically relevant concentrations comparable to those achieved in solid tumor patients. RMC-6236 is active against a spectrum of MM lines with KRAS or NRAS mutations, as well as wildtype lines, some of which have upstream lesions (e.g. FGFR3 mutations) that can activate Ras signaling. In vivo treatments (e.g. with MRTX-1257 or MRTX-1133 against diffuse lesions of XG-7 or KP-6 cells, respectively, after IV injections in NSG mice) were active, but tumor escape eventually ensued. Notably, tumor cells harvested at full-blown relapse after KRAS inhibitor treatment as well as cells from vehicle-treated mice were similarly responsive in vitro upon retreatment with the same KRAS inhibitor suggesting that the in vivo microenvironment contributes to relapse. DNA barcode-based clonal tracking studies showed lack of significant barcode enrichment in KRAS inhibitor-treated (vs. DMSO control) MM cells, suggesting that treatment escape was not primarily due to selection of pre-existing drug-resistant cells. Co-culture of human MM cells with bone marrow stromal cells attenuated their response to KRAS G12C inhibitors; and genetically engineered VQmyc (NRASQ61R) mouse MM cells respond to RMC-6236, though less so in the presence of IL-6, again suggesting that the microenvironment and its cytokines can affect sensitivity to Ras inhibitors. To functionally map the landscape of genomic lesions that can influence Ras inhibitor response, we conducted 36 genome-scale CRISPR knockout (KO) or activation (CRISPRa) screens across 11 genotypically diverse MM lines. These studies revealed a heterogeneous, cell line-specific landscape of gene perturbations influencing sensitivity and resistance to Ras inhibition, without obvious association between the patterns of CRISPR “hits” in individual MM cell lines and their respective molecular subtype (e.g. t(4;14), t(14;16), t(11;14)). CRISPR KO vs. CRISPRa studies with the same inhibitor in a given cell line yielded complementary and orthogonal results. For example, CRISPRa identified upstream surface receptors (e.g. EGFR, MET), ABCB1, KRAS itself, or positive regulators of Ras/MAPK signaling (e.g. SHOC2) among the top resistance-associated perturbations. In a complementary way, recurrent resistance-associated hits in CRISPR KO screens included diverse negative regulators of Ras/MAPK (e.g. LZTR1); NF-κB (e.g. TRAF3), or PI3K/Akt (e.g. PTEN) signaling; as well as the oxidative stress sensor KEAP1. Many gene perturbations were identified as hits in CRISPR studies across all 3 classes of Ras inhibitors, but others had distinct roles, e.g. PPIA (cyclophilin A) loss decreased MM cell responses to RMC-6236, consistent with this protein forming a complex with Ras and RMC-6236 to block downstream Ras signaling. Many hits from our screens were shared with those in publicly available CRISPR studies of KRAS inhibitors in solid tumor lines, while others were distinct for MM. Bulk RNA-seq of 6 Ras inhibitor-treated MM lines identified a transcriptional signature of genes concordantly up- or down-regulated across all MM lines tested: most of these genes, including some known Ras pathway inhibitors (e.g. SPRED1, DUSP6) were not prominent hits in our CRISPR studies, indicating that the most pronounced/recurrent regulators of response to these agents may not be directly inferred from transcriptional profiling. Our study reveals that MM cell responses to pharmacological Ras inhibitors can be governed by an interplay of non-genomic adaptations, microenvironment-derived cues, and a diverse and heterogeneous landscape of genomic perturbations. Notably, even MM cells with the same KRAS mutation can display distinct “resistomes”, highlighting the complex functional genomic landscape underlying Ras inhibitor responses. We envision that these results will inform personalized uses of Ras inhibitors in future clinical studies in MM.