Abstract
Resistance to inactive state-selective RASG12C inhibitors frequently entails accumulation of RASGTP, rendering effective inhibition of active RAS potentially desirable. Here, we evaluated the antitumor activity of the RAS(ON) multiselective tricomplex inhibitor RMC-7977 and dissected mechanisms of response and tolerance in KRASG12C-mutant non-small cell lung cancer (NSCLC). Broad-spectrum reversible RASGTP inhibition with or without concurrent covalent targeting of active RASG12C yielded superior and differentiated antitumor activity across diverse comutational KRASG12C-mutant NSCLC mouse models of primary or acquired RASG12C(ON) or RASG12C(OFF) inhibitor resistance. Interrogation of time-resolved single-cell transcriptional responses established an in vivo atlas of multimodal acute and chronic RAS pathway inhibition in the NSCLC ecosystem and uncovered a regenerative mucinous transcriptional program that supports long-term tumor cell persistence. In patients with advanced KRASG12C-mutant NSCLC, the presence of mucinous histologic features portended poor response to sotorasib or adagrasib. Our results have potential implications for personalized medicine and the development of rational RAS inhibitor-anchored therapeutic strategies.
INTRODUCTION
Activating mutations in KRAS constitute the most prevalent oncogenic driver event in nonsquamous non-small cell lung cancer (NSCLC) and frequently involve a glycine to cysteine substitution at residue 12 (G12C; ref. 1). The groundbreaking discovery and clinical development of irreversible, inactive state-selective allosteric inhibitors of KRASG12C, culminating in the regulatory approvals of sotorasib and adagrasib, have transformed the therapeutic landscape of KRASG12C -mutant advanced NSCLC (2–5). Nonetheless, the long-term efficacy of sotorasib and adagrasib is modest, with median progression-free survival (PFS) between 6 and 7 months, and is curtailed by both primary insensitivity in a fraction of patients and emergence of adaptive and acquired resistance (4–6). Cooccurring inactivating genomic alterations in key tumor suppressor genes—including KEAP1, SMARCA4, and CDKN2A— and coincident oncogenic alterations (encompassing both amplifications and somatic mutations) in RAS (KRAS, HRAS, NRAS) genes have been identified as drivers of early disease progression with inhibitors of GDP-bound KRASG12C and mark difficult-to-treat patient subgroups with poor prognosis and an unmet clinical need (7). On-treatment feedback reactivation of the RAS/MAPK pathway, as a result of synthesis and GTP loading of new KRASG12C and/or RTK-driven activation of wild-type (WT) RAS (KRAS, NRAS, or HRAS) can also occur rapidly and compromise the clinical efficacy of inactive state-selective KRASG12C inhibitors (8, 9). In addition, the emergence of secondary alterations in components of the RAS/MAPK pathway—often detected in multiple pathway genes simultaneously due to convergent tumor evolution—or trans-differentiation to squamous histology can foster disease progression and impose additional impediments to improving clinical outcomes (10–12). These critical challenges have spearheaded intense efforts to discover more potent KRASG12C inhibitors and identify rational combinations, with multiple agents currently in early or advanced stages of clinical development (13, 14).
Because accumulation of active, GTP-bound RAS (mutant and/or WT KRAS, NRAS, or HRAS) represents a convergent mechanism of resistance to inactive state-selective KRASG12C inhibitors that may ostensibly bypass even more potent KRASG12C inhibition, there is compelling rationale to develop broad-spectrum direct RAS inhibitors that tackle both mutant and WT active RAS oncoproteins. This is also critical in order to extend targeted therapeutic options to patients harboring KRASnon-G12C -driven tumors that account for approximately 50% of KRAS-mutant NSCLC and the vast majority of KRAS-mutant colorectal and pancreatic cancer cases, as well as to a smaller group of patients bearing HRAS or NRAS-mutant tumors. Previous efforts to broadly target RAS signaling indirectly with inhibitors of the SHP2 phosphatase or SOS1 have been hindered by modest efficacy and poor tolerability (15). Recently, RMC-7977, a reversible, active state-selective direct inhibitor of multiple mutant and WT canonical RAS isoforms (including KRAS, NRAS, and HRAS), a RAS(ON) multiselective inhibitor, was well tolerated and demonstrated robust antitumor activity across a spectrum of RAS-addicted tumors in preclinical models (16, 17). RMC-7977 exhibits a unique mechanism of action that depends on the formation of an RMC-7977-CYPA-RASGTP tricomplex with cyclophilin A (CYPA) and RASGTP that sterically hinders interactions with downstream effectors and impedes RAS signaling flux (16). Notably, RMC-6236, a related clinical stage RAS(ON) multiselective inhibitor (18), was safe and tolerable at clinically relevant drug exposures and yielded promising preliminary antitumor activity, with a 38% ORR across 40 response-evaluable patients with advanced KRASG12X(non-GI2C)-mutant NSCLC in an ongoing phase 1 study (19).
It is currently unclear how to best tailor distinct RAS-targeting approaches to individual patients with KRAS-mutant NSCLC, including difficult-to-treat, poorly prognostic subgroups. In addition, how to optimally deploy emerging broad-spectrum active RAS inhibitors such as RMC-6236 (18) with mutant-selective inhibitors of KRASG12C (including active state-selective inhibitors such as the recently described RMC-6291; ref. 20) or other mutant KRAS variants remains unknown. Here, we compared the activity of distinct RAS therapeutic strategies in multiple comutational models of KRASG12C-mutant NSCLC that recapitulate clinically challenging disease subgroups and evaluated mechanisms of response and tolerance to active RAS inhibition at both whole-tumor and single-cell resolution. Our results reveal superior and differentiated antitumor activity of the RAS(ON) multiselective inhibitor RMC-7977 as a monotherapy and in combination with the RAS(ON) G12C-selective inhibitor RMC-4998 (16) across a broad range of NSCLC comutational models with primary or acquired resistance to both inactive and active state-selective KRASG12C inhibitors. Critically, we unveil a regenerative/homeostatic mucinous transcriptional state that supports the persistence of tumor cells (TC) during treatment with RMC-7977 and other RAS pathway inhibitors across multiple murine and human models of both KRASG12C- and KRASnon-G12C-mutant NSCLC, with implications for clinical outcome prediction and the development of novel rational combination therapeutic strategies.
RESULTS
The RAS(ON) multiselective inhibitor RMC-7977 exhibits robust and durable antitumor activity as a monotherapy and in combination with the RAS(ON) G12C-selective inhibitor RMC-4998 in preclinical models of difficult-to-treat KRASG12C-mutant NSCLC
We devised a comprehensive experimental plan to dissect and compare the acute and long-term effects of distinct RAS inhibition strategies in a representative panel of immune-competent mouse models that reflect the comutational and phenotypic heterogeneities of human KRASG12C-mutated NSCLC and capture difficult-to-treat subsets with poor prognosis (Fig. 1A). These included (i) KL2 and KL5 (both KrasG12C-mutant and STK11-deficient KL) and KP2 (KrasG12C-mutant and p53-deficient KP) models, representing the prevalent KL and KP comutational subsets of KRASG12C-mutant NSCLC (1); (ii) isogenic derivatives of the KL5 cell line with stable shRNA-mediated knockdown of Keap1 (KL5shKeap1) and Smarca4 (KL5shSmarca4), reflecting poorly prognostic subgroups with coalterations in KEAP1 or SMARCA4 that exhibit impaired response to sotorasib or adagrasib (7); (iii) models with lack of thyroid transcription factor 1 (TTF1, encoded by the Nkx2.1 gene) expression, an established poor prognostic indicator (KL2, KL5), and high expression of HMGA2 (1, 21–23); and (iv) models with adenosquamous differentiation (KL1 and KL4, both KrasG12C mutant and Stk11 deficient; Fig. 1B; Supplementary Fig. S1A-S1H).
Figure 1.
Antitumor activity of RAS(ON) inhibition in preclinical models of difficult-to-treatKRASG12C-mutant NSCLC. A, Experimental strategy to evaluate acute and long-term effects of RAS inhibition in immune-competent co-mutational models of KRASG12C and KRASG12D-mutated NSCLC. Six to 13 mice were included in each treatment cohort. B, Histopathologic and IHC characterization of subcutaneous allografts established from murine KrasG12C and KrasG12D-mutated NSCLC cell lines that were utilized in the in vivo studies. C and D, RMC-7977 monotherapy and RMC-7977/RMC-4998 combination therapy induce deep and sustained tumor regressions in both STK11-deficient (KL2; C) and p53-deficient (KP2; D) co-mutational models of KRASG12C-mutant lung adenocarcinoma. Left, tumor volume growth curves in the study treatment arms. Curves are truncated at the time point that the first mouse in each cohort reached tumor volume ≥1,000 mm3. Top right, Kaplan-Meier estimate of TTD. Comparison of TTD between treatment groups is based on the log-rank test. Bottom right, Kaplan-Meier estimate of relapse-free survival (RFS), assessed as the time from treatment withdrawal on day 60 to TV ≥ 500 mm3. Pairwise comparisons of RFS between treatment groups is based on the log-rank test. The dotted red line indicates treatment withdrawal on day 60. E, RMC-7977 monotherapy/combination therapy overcomes primary resistance to inactive and active state-selective RASG12C inhibitors in KL5 derivative models with shRNA-mediated knockdown of Smarca4 or Keap1. F, Antitumor activity of RMC-4988, RMC-7977, and their combination in KEAP1 - and STK11 -mutated H2030 (top) and H2122 (bottom) KRASG12C-mutated human NSCLC CDX models. Tumor volume growth curves and Kaplan-Meier estimates of TTD are displayed for each model. Tumor volume data is represented as mean ± SEM and analyzed by two-way ANOVA followed by Bonferroni’s multiple comparison test. Comparison of TTD between treatment groups is based on the log-rank test. G, RMC-7977 exhibits robust antitumor activity in KrasG12C/+;Lkb1−/− GEMMs. BOR (top left) and PFS (top right) with RMC-4998 or RMC-7977 were determined based on mouse RECIST. Representative μCT images of primary lung tumors at the time of randomization and following 3 and 6 weeks of treatment. Note the emergence of on-treatment disease progression with RMC-4998, whereas treatment with RMC-7977 results in sustained response (bottom). P ≤ 0.05 was considered statistically significant for all comparisons. (A, Created with BioRender.com.)
Active RAS inhibition with RMC-7977 monotherapy or RMC-7977/RMC-4998 combination therapy elicited rapid, deep, and sustained tumor regressions in the KL2, KP2, and each of the KL5shCON, KL5shSmarca4, and KL5shKeap1 isogenic comutational models and resulted in significantly prolonged time to tumor doubling (TTD) compared with cohorts treated with either sotorasib or the RAS(ON) G12C-selective inhibitor RMC-4998 (P < 0.0001, log-rank test for each of the pairwise comparisons; Fig. 1C-E; Supplementary Fig. S1I; Supplementary Table S1). Treatment with RMC-7977 or RMC-7977 plus RMC-4998 further extended TTD compared with the combination of RMC-4998 (or sotorasib) and the SHP2 inhibitor, RMC-4550, in the KL2 and KP2 models (Fig. 1C and D; Supplementary Table S1). Critically, none of the mice that were treated with RMC-7977 or RMC-7977 plus RMC-4998 in the KL2, KP2, and KL5shCON, KL5shKeap1, and KL5shSmarca4 models developed on-treatment disease progression up to day 60, at which point treatment was withdrawn in order to evaluate recurrence-free survival (RFS) and determine the fraction of mice in each treatment arm that reach the “cure” endpoint. RMC-7977/RMC-4998 doublet RAS(ON) treatment significantly prolonged RFS compared with RMC-7977 monotherapy in the KL2 (<0.0001, log-rank test) and KP2 (P = 0.0008, log-rank test) models, although cures were observed in only 1/10 (10%) of KL2 and 0/7 KP2 mice that received combination therapy (tumor recurrence was noted in all mice treated with RMC-7977 alone). Strikingly, the addition of the RAS(ON) G12C-selective inhibitor RMC-4998 to the RAS(ON) multiselective inhibitor RMC-7977 in the KL5shCON, KL5shKeap1, and KL5shSmarca4 allograft models dramatically improved antitumor activity and elicited cures in 100% of treated mice (7/7 in the KL5shCon, 8/8 in the KL5shKeap1, and 7/7 in the KL5shSmarca4 models; Fig. 1E; Supplementary Table S1). Treatment with single-agent RMC-7977 induced cures in 50% (4/8) of KL5shCon mice, 28.6% (2/7) of KL5shSmarca4 mice, and 0% (0/8) of KL5shKeap1 mice (Supplementary Table S1). Improved activity of RMC-7977 in combination with RMC-4998 compared with either agent alone was also observed in the NCI-H2030 and NCI-H2122 and STK11- and KEAP1-deficient human cell line-derived xenograft (CDX) lung adenocarcinoma models (Fig. 1F).
We further evaluated the activity of (i) RMC-7977 and (ii) RMC-4998 in the setting of autochthonous lung tumors in the KrasLSL-G12C/+;Stk11FL/FL genetically engineered mouse model (GEMM; Supplementary Fig. S1J). Both RMC-4998 and RMC-7977 induced radiologic responses and significantly prolonged PFS, defined as ≥20% increase in the sum oflargest diameters of all target lesions, compared with vehicle-treated mice (P < 0.0001 for each pairwise comparison, log-rank test; Fig. 1G). Importantly, broad-spectrum inhibition of active RAS with the RAS(ON) multiselective inhibitor RMC-7977 elicited deeper and more durable responses compared with the RAS(ON) G12C-selective inhibitor RMC-4998 and significantly prolonged PFS (P = 0.0001, log-rank test; Fig. 1G).
Because squamous differentiation may impact clinical responses to mutant-selective KRASG12C inhibitors (10, 12), we also assessed whether RMC-4998 and RMC-7977 were active in the KL1 adenosquamous allograft model that exhibits primary resistance to sotorasib. Monotherapy with RMC-4998 resulted in short-lived tumor shrinkage followed by rapid disease progression (TTD 17 days; Supplementary Fig. S1I-S1K; Supplementary Table S1). In contrast, RMC-7977 induced rapid and deep tumor regressions with no apparent benefit from the addition of RMC-4998, resulting in significantly prolonged TTD, albeit with the eventual emergence of on-treatment resistance in this model (Supplementary Fig. S1K). Finally, to confirm that the antitumor activity of RMC-7977 extends to non-G12C-driven NSCLC models, we assessed the in vivo activity in the LKR10 KrasG12D-driven lung adenocarcinoma allograft model that is WT for both Stk11 and Trp53. In accordance with results in KrasG12C-driven models, treatment with RMC-7977 induced deep and durable responses with no tumor outgrowth observed until treatment discontinuation on day 60 and an overall 50% (4/8) cure rate (Supplementary Fig. S1L). Taken together, our results indicate that broad-spectrum inhibition of active RAS with RMC-7977 exhibits robust antitumor activity in KRAS - mutant NSCLC that can be further enhanced with the addition of the RAS(ON) G12C-selective covalent inhibitor RMC-4998, particularly in the context of difficult-to-treat, poorly prognostic subsets of KRASG12C-mutated NSCLC.
RMC-7977 Overcomes Acquired Resistance to Inactive and Active State-Selective RASG12C Inhibitors
Because increased activity of RAS proteins frequently underpins the emergence of adaptive or acquired resistance to inactive state-selective RASG12C inhibitors, we examined whether RMC-7977, which inhibits both mutant and WT alleles of GTP-bound KRAS, HRAS, or NRAS (16), can overcome acquired resistance to sotorasib. We first derived cell lines (KL2 SR1, KL5 SR1, and KL4 SR1) from tumors that developed in vivo resistance to sotorasib (at a dose of 100 mg/kg PO QD). Whole-exome sequencing (WES) revealed several acquired genomic alterations, including low copy number gains of genes in the RTK/RAS/MAPK and PI3K/AKT pathways, as well as cell cycle and immune-related genes (Supplementary Fig. S2A). Notably, KL5 SR1 allograft tumors exhibited squamous histology, consistent with adeno-to-squamous transition as the likely mechanism of resistance to sotorasib in this model (Supplementary Fig. S2B; ref. 12). Mice bearing subcutaneous KL2 SR1, KL5 SR1, and KL4 SR1 allograft tumors (200–250 mm3) were initially rechallenged with sotorasib to confirm in vivo resistance before treatment was switched (at a TV of 450–500 mm3) to RMC-7977 monotherapy or RMC-7977 plus RMC-4998 (Fig. 2A; Supplementary Fig. S2C). Both regimens induced rapid and sustained tumor shrinkage across most sotorasib-resistant models (Fig. 2A; Supplementary Fig. S2D), with only 2/9 mice bearing adenosquamous KL4 SR1 allograft tumors eventually developing on-treatment resistance (Fig. 2A; Supplementary Fig. S2E). Interestingly, RMC-7977-resistant KL4 SR1 tumors exhibited loss of TTF1 expression and lack of cellular cohesion, indicative of dedifferentiation and epithelial-to-mesenchymal transition (EMT) as putative mechanism(s) of acquired resistance to RMC-7977 (Supplementary Fig. S2E). Complete loss of E-cadherin and induction of vimentin expression were confirmed by Western blotting in a cell line derived from the KL4 SR1 RMC-7977-resistant tumor (Supplementary Fig. S2F). Loss of TTF1 and E-cadherin expression in this tumor was further validated by IHC (Supplementary Fig. S2G). Notably, loss of E-cadherin, TTF1, and p63 expression was also observed in KL1 tumors that developed on-treatment resistance to RMC-7977 (Supplementary Fig. S2H).
Figure 2.
RMC-7977 overcomes acquired resistance to inactive and active state-selective RASG12C inhibitors. A, RMC-7977 monotherapy and RMC-7977/RMC-4998 combination therapy elicit deep and sustained tumor regressions in KRASG12C-mutated NSCLC models with acquired resistance to sotorasib. B, RMC-7977 is active in NSCLC models with acquired in vivo resistance to the RASG12C(ON) inhibitor RMC-4998. C, RMC-7977 overcomes acquired resistance to RMC-4988 driven by secondary alterations in RAS genes. Individual tumors that developed on-treatment resistance to RMC-4988 (KL5shCon RMC-4988 R1-R3 and KP2A RMC-4988 R1-R3) were harvested, processed for WES, and propagated directly in C57BL/6 mice under continuous drug exposure. Mice harboring tumors 450 to 500 mm3 were subsequently treated with RMC-7977 or RMC-7977 plus RMC-4998. Tumor volume growth curves (left) and waterfall plot representation of individual mouse-level best % change in tumor volume (right) are shown. D, OncoPrint depicting somatic genomic alterations (SNVs, indels, and copy number alterations) in the RMC-4988-resistant models and the parental KL5 and KP2 cell lines. Individual tumors from the experiment depicted in D are indicated as R1, R2, and R3.
Next, we assessed whether RMC-7977 could overcome acquired resistance to active state-selective RASG12C inhibition with RMC-4998. Mice bearing KL5shCon and KP2 allograft tumors that developed disease progression with RMC-4998 (Fig. 1E and D) were switched to treatment with RMC-7977 at TV ≥ 1,000 mm3. Remarkably, RMC-7977 triggered profound and enduring tumor regressions, without the emergence of adaptive/acquired resistance for the duration of the experiment (Fig. 2B). It is notable that responses to RMC-7977 occurred despite high baseline tumor burden that has been linked with inferior responses to inactive state-selective KRASG12C inhibitors in patients with advanced KRASG12C - mutated NSCLC (6, 24), thus further underscoring the robust activity of RMC-7977 in this setting. To characterize the molecular mechanisms of acquired resistance to active RAS G12C inhibition that may retain sensitivity to RMC-7977, we transplanted individual RMC-4998-resistant tumors (KL5shCon RMC-4998 R1-R3 and KP2 RMC-4998 R1-R3) to the flanks of C57BL/6 recipient mice while retaining part of each tumor for WES, followed by treatment with either RMC-7977 or RMC-7977 plus RMC-4998 (at TV ≥ 500 mm3; Fig. 2C). Resistant tumors retained the lineage identity of parental cell lines (Supplementary Fig. S2I). High-level amplification of the KrasG12C allele (adjusted for TC content) was identified in the KL5 RMC-4998 R2 (>20 copies), KP2 RMC-4998 R3 (18 copies), and KP2 RMC-4998 R1 (six copies) tumors and was likely a key mechanism of acquired resistance to RMC-4998 (Fig. 2D; Supplementary Fig. S2J). In the KP2 RMC-4998 R2 and R3 tumors, we further detected the emergence of secondary somatic KrasG12D mutations, resulting in an oncoprotein that is not susceptible to inhibition by RMC-4998 but retains sensitivity to RMC-7977 (Fig. 2D). Acquired missense somatic mutations in Cul3, encoding an E3 ubiquitin ligase that is critical for NRF2 regulation, were also identified in the KP2 RMC-4998 R1 and R3 tumors, although their functional significance is unclear. Finally, the KL5 RMC-4998 R1 and R3 tumors exhibited low-level copy number gains of KrasG12C (three and five copies, respectively) and of several other genes in the RTK/RAS/MAPK, PI3K/AKT, developmental, and cell cycle control pathways, including Nras, Braf, Rafi, Ccnd2, Met, Ret, Notch2, Ntrkl, Pik3ca, Pik3c2g, and Sox2, frequently as a result of arm-level chromosome gain (Fig. 2D; Supplementary Fig. S2J). Of note, biallelic deletions of Cdkn2a/Cdkn2b, Amerl, and Med12 and hemizygous deletion of RbmlO were observed in the KL5 RMC-4998 R1-R3 tumors and may have also contributed to RMC-4998 resistance (Fig. 2D). The possible role of identified acquired somatic mutations and copy number alterations in these and other genes as drivers of RMC-4998 insensitivity is currently uncertain and requires further investigation.
Phenotypic Effects of Active RAS Inhibition in NSCLC
We interrogated both TC-intrinsic and nontumor cell-autonomous effects of RMC-7977 and RMC-7977 plus RMC-4998 to decipher the mechanisms that underpin their superior antitumor activity in KRASG12C-mutant NSCLC models. In the KL2 allograft model, treatment with RMC-7977 or RMC-7977 plus RMC-4998 for 7 days led to marked reduction of cytokeratin 7 (CK7) immunoreactivity (Fig. 3A) and potently suppressed phospho-ERK1/2 (Thr202/Tyr204) expression and TC proliferation (assessed as the fraction of TCs that are positive for Ki67), compared with vehicle- or sotorasib-treated mice [Fig. 3A and B (top)]. Similar results were observed in the KP2 [Supplementary Fig. S3A-S3C; Fig. 3B (lower)] and KL5shCon, KL5shKeap1, and KL5shSmarca4 models (Supplementary Fig. S4A-S4C), although the difference in the Ki67+ TC fraction between mice treated with sotorasib and RMC-7977-containing arms in the KP2 and KL5scCon models did not reach statistical significance [Fig. 3B (bottom); Supplementary Fig. S4B (top)]. In time-course experiments, RMC-7977 (alone or in combination with RMC-4998) prevented feedback MAPK pathway reactivation (that was mediated at least partially by HRAS and/or NRAS in response to treatment with sotorasib, in agreement with prior reports; ref. 8) and resulted in sustained phospho-ERK1/2 (Thr202/Tyr204) suppression in KL2 whole-tumor lysates 12 hours following a single treatment dose, whereas dual RMC-7977/RMC-4998 treatment was the most effective at suppressing the expression of phospho-S6 (Ser235/Ser236) compared with other treatment arms (Supplementary Fig. S4D and S4E), a finding that was also recapitulated in the KL5shCon model (Supplementary Fig. S4F). RMC-7977 and RMC-7977 plus RMC-4998 resulted in higher expression of cleaved caspase 3 and cleaved PARP at both 12 and 24 hours in the KL2 and KP2 models and at 12 hours in the KL5shCon model following single-dose treatment compared with vehicle, sotorasib, or RMC-4998 monotherapy, indicating more potent induction of apoptosis at early time points (Fig. 3C; Supplementary Figs. S3B and S3C and S4F). Robust suppression of phospho-ERK1/2 (Thr202/Tyr204) expression and induction of apoptosis were also observed in response to RMC-7977 in the KrasG12C/+;Lkbl−/− GEM model (Supplementary Fig. S5A-S5C). Importantly, treatment with diverse RAS inhibitors promoted the engagement of inflammatory cell death pathways, including both necroptosis [as evidenced by accumulation of phospho-RIP3 (Thr231/Ser232) and phospho-MLKL(Ser345)] and pyroptosis (characterized by increased expression of cleaved caspasel and cleaved gasdermin D; Fig. 3C; Supplementary Fig. S3B). Induction of inflammatory cell death was more prominent in mice treated with the active state-selective inhibitors RMC-4998 and RMC-7977 (either alone or in combination) compared with sotorasib in both the KL2 and KP2 models, particularly at the 7-day time point (Fig. 3C). Thus, potent induction of both apoptotic and nonapoptotic programmed cell death-including necroptosis and pyroptosis—may contribute to the enhanced activity of active state-selective RAS inhibitors in KRASG12C -mutated NSCLC.
Figure 3.
Tumor cell-intrinsic and microenvironmental effects of active RAS inhibition in NSCLC. A, Acute phenotypic responses to mutant-selective RASG12C and broad-spectrum RAS pathway inhibitors in the KL2 NSCLC model. Representative H&E, histochemical, and IHC images of subcutaneous KL2 tumor allografts following in vivo treatment for 7 days with the indicated inhibitors (N = 6 for all treatment groups). Scale bars, 100 μm. B, Violin plot representation of the Ki67+ TC fraction following 7-day treatment with the indicated RAS pathway inhibitors in the KL2 (top) and KP2 (bottom) allograft models. Ten representative fields in each tumor were selected by two experienced thoracic pathologists blinded to the treatment arm allocation and were analyzed by automated quantitative IHC using Halo software. Only comparisons with significant P values are indicated. C, Western blot analysis of (phospho)-protein abundance reflecting apoptosis (cleaved caspase3 and cleaved PARP), necroptosis (phospho-RIP3 and phospho-MLKL), and pyroptosis (cleaved caspasel and cleaved gasdermin D) in whole-tumor lysates collected at distinct time points following in vivo treatment with the indicated inhibitors. D, FACS-based enumeration of immune cell subsets in KL2 allograft tumors following 7-day treatment with the specified inhibitors. E, Comparison of tumor vascular density following treatment with the indicated inhibitors in the KL2 (left, 7-day treatment) and KP2 (right, 21-day treatment) allograft models. One-way ANOVA with Tukey’s multiple comparison test was used for all pairwise statistical comparisons between treatment groups. P value ≤ 0.05 was considered statistically signifiicant. Asterisks denote statistical signifiicance: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.
We subsequently assessed the impact of active state-selective RAS inhibitors on the tumor immune contexture in the KL2 model. In agreement with previous reports, the tumor immune microenvironment (TIME) of vehicle-treated STKll-deficient KL2 allograft tumors was characterized by paucity of CD8+ T cells and preponderance of neutrophils and polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs, representing ~70% of CD45+ cells; refs. 1, 25, 26). Treatment with active state-selective RAS inhibitors remodeled the TIME, resulting in decreased abundance of PMN-MDSCs (with the exception of the RMC-4998 plus RMC4550 arm) and an increase in populations of effector CD8+ T cells and NK cells (as a percentage of CD45+ cells), as previously reported for the inactive state-selective KRASG12C inhibitors sotorasib and adagrasib (Fig. 3A and D; refs. 27-29). Accumulation of NK cells and suppression of PMN-MDSCs were most prominent in mice that received RMC-7977/RMC-4998 combination therapy; in the latter case, this may reflect the role of TC-intrinsic RAS signaling on the establishment of a myeloid cell-rich suppressive TIME, through the secretion of proinflammatory cytokines (Fig. 3D; ref. 30). Thus, engagement of innate and adaptive antitumor immunity coupled with depletion of immune-suppressive cell subsets may contribute to the antitumor activity of active state-selective RAS inhibitors.
Finally, we determined the impact of RAS inhibitors on nonimmune components of the TME. Short-term treatment for 7 days induced profound changes in the architecture of KL2, KL5shCon, KL5shKeap1, and KL5shSmarca4 subcutaneous allograft tumors, marked by a prominent desmoplastic response (assessed by Mason’s trichome histochemical staining) that seemed to track with antitumor activity and was particularly striking in mice treated with RMC-7977 or RMC-7977 plus RMC-4998 (Fig. 3A; Supplementary Figs. S3A and S4A). Induction of desmoplasia was less pronounced in treated KP2 tumors, in accordance with a higher number of residual viable TCs at the 7-day time point compared with the KL2 and KL5 models (Supplementary Figs. S3A and S4A; Fig. 3A). Unexpectedly, treatment with the RAS(ON) multiselective inhibitor RMC-7977 (alone or in combination with RMC-4998) resulted in a significant reduction in tumor vascular density in the KL2 model (Fig. 3E). This effect was not observed with either sotorasib or RMC-4998 at this time point (Fig. 3E). Reduced tumor neovascularization with RMC-7977 monotherapy or combination therapy compared with vehicle, sotorasib, or RMC-4998 was further validated in the KP2 and KL5shCon models, albeit after a more prolonged treatment course (21 days; Fig. 3E; Supplementary Figs. S3A and S4C).
Drug-Tolerant Persister Cell Population Curtails the Antitumor Activity of RMC-7977
The stereotypical clinical course in response to RMC-7977 in these diverse comutational models, characterized by long-term tumor regression, rare on-treatment tumor growth, and eventual relapse upon compound withdrawal, suggests that persistence of a population of drug-tolerant TCs may be a barrier to the clinical efficacy of active RAS inhibition in KRAS-mutant NSCLC. To test this hypothesis, mice bearing resurgent subcutaneous allograft tumors following withdrawal of either RMC-7977 (in the KL2, KP2, KL5shCon, or LKR10 models) or RMC-7977 plus RMC-4998 (in the KL2 and KP2 models) were rechallenged with either RMC-7977 or RMC-7977/RMC-4998 when tumor volume exceeded 1,000 mm3 (Fig. 4A). Reexposure to RMC-7977 (blue curves) or RMC-7977 plus RMC-4998 (purple curves) rapidly induced tumor shrinkage in all treated mice that was durable for at least 45 days, consistent with reversion to a drug-sensitive state upon targeted therapy withdrawal, a key hallmark of the drug-tolerant persister (DTP) phenotype (Fig. 4A; ref. 31). Morphologically, RMC-7977- or RMC-7977 plus RMC-4998-rechallenged tumors frequently demonstrated cystic degeneration and were characterized by desmoplastic stroma with interspersed cancerous glands. Notably, residual TCs often displayed mucinous features with basally located nuclei and abundant supranuclear mucin and formed structures that were reminiscent of pancreatic intraepithelial neoplasia (Fig. 4B). Mucinous differentiation was also observed in DTPs to RMC-7977 in the KrasG12C/+,’Lkb1−/− GEMMs, whereas lung tumors from vehicle-treated mice exhibited a diverse phenotypic spectrum (Fig. 4C; Supplementary Fig. S6). DTP cells exhibited a low Ki67+ proliferative index, similar to the proliferative arrest induced in response to acute short-term treatment with RMC-7977 (Fig. 4B and D). Prolonged exposure to RMC-7977 also potently suppressed tumor neovascularization in the KL2 model, in agreement with the observed effects following short-term treatment (Fig. 4B and E).
Figure 4.
Emergence of a DTP cell population in response to RMC-7977. A, Resurgent tumors following drug withdrawal across diverse Kras-mutant NSCLC allograft models retain long-term sensitivity to rechallenge with RMC-7977 or RMC-7977/RMC4998, even in the context of high baseline tumor burden (≥1,000 mm3). Note that the single KL2 tumor with apparent on-treatment increase in tumor volume (highlighted with an asterisk) displayed prominent treatment-induced cystic degeneration [as illustrated in B (i)]. B, Representative H&E, histochemical, and IHC images of tumor allografts following long-term (6–8 weeks) rechallenge with RMC-7977 or RMC-7977+RMC4998. (i) Low magnification image from an intact KL2 residual tumor following long-term treatment with RMC-7977. (ii-vi) DTP TC populations in KL2 (ii, iii), KLsshCon (iv), KP2 (v), and LKR10 (vi) allograft models following long-term treatment with RMC-7977 (ii, iv, v, vi) or RMC-7977+RMC4998 (iii, KL2 model). (vii-xix) Representative images of Masson’s trichrome (vii, viii) and Alcian Blue (xix) staining in DTPs in the KL2 model. (ix, x) Ki67 immunostaining in KL2 and KP2, respectively. (xii) CD31 staining in DTPs in the KL2 model. Scale bars, 100 μm. C, Representative H&E images of tumor-bearing whole lung samples from KrasG12C/+;Lkb1−/− GEMMs following treatment with vehicle or RMC-7977. DTPs following long-term treatment with RMC-7977 (60 days) exhibit mucinous histologic features (high magnification image, right). D, RMC-7977-tolerant persister TCs in KL2 allograft tumors exhibit low Ki67+ proliferative index. E, Sustained inhibition of active RAS suppresses neovascularization in the KrasG12C-mutant NSCLC tumor microenvironment. Violin plot representation of vascular density in KL2 tumor allografts treated for 7 days with (i) vehicle, (ii) RMC-7977, (iii) following drug release (after 60 days of continuous treatment), and (iv) RMC-7977 rechallenge for a minimum of 6 weeks. One-way ANOVA with Tukey’s multiple comparison test was used for all pairwise statistical comparisons between treatment groups. P value ≤ 0.05 was considered statistically significant. Asterisks denote statistical significance: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.
Single-Cell Transcriptional Landscape of Acute and Chronic RAS Inhibition in KrasG12C-Mutant NSCLC
To dissect the temporal evolution of TC-intrinsic and microenvironmental responses to RAS inhibition in vivo, we profiled the transcriptomes of high-quality individual cells retrieved from 34 KL2 allograft tumor samples following treatment for 7 days with distinct RAS inhibitors [(i) vehicle (Vehicle_D7), (ii) sotorasib (G12C (OFF) i_D7), (iii) RMC-4998 (G12C (ON) i_D7), (iv) RMC-4998 plus RMC-4550 (G12C (ON) iSHP2i_D7), (v) RMC-7977 (RASMULTIi_D7), (vi) RMC-7977 plus RMC-4998 (G12C (ON) iRASMULTIi_D7); n = 4 mice per treatment arm]. In addition, we included residual tumors from eight mice that were initially treated with RMC-7977 for 60 days, followed by drug withdrawal and subsequent long-term rechallenge of resurgent tumors (at TV ≥ 1,000 mm3) with RMC-7977 for 6 to 8 weeks (RASMULTIiRechallenged). Finally, we also included renascent tumors from three mice following release from treatment with RMC-7977 plus RMC-4988 (4 weeks posttreatment cessation; DrugRelease). One sample (RASMULTIRechallenged_6) yielded fewer than 200 viable TCs and was excluded from further analyses. Following normalization, integration of cells across all samples using the fast mutual nearest neighbors correction (fastMNN) function to account for batch effects, dimensionality reduction, and clustering, we resolved 21 distinct cell types encompassing TCs, lymphoid cells, myeloid cells, fibroblasts, pericytes, and endothelial cells (Fig. 5A). Cluster identities were assigned based on the expression patterns of established marker genes (Supplementary Fig. S7A and S7B) and were further validated based on the reference dataset from the Immunologic Genome Project (ImmGen) using SingleR (Supplementary Fig. S7C; ref. 32).
Figure 5.
Single-cell transcriptional atlas of in vivo RAS inhibition in KRASG12C NSCLC. A, Uniform manifold approximation and projection (UMAP) plot of recovered individual cells, colored by cell-type identity. B, Stacked bar plot representation of absolute (top) and relative (bottom) abundance of distinct cell clusters in each tumor sample. Clusters are color-coded as in A. Note that part of the tumor in mice that were rechallenged with RMC-7977 was stored for IHC studies; therefore, only comparisons of relative cell cluster abundance is applicable. Whole treated tumors were collected and processed in all other treatment arms. C and D, Box plot representation of the scRNA-seq-derived NK cell fraction (C) and TIL to TC ratio (D) in each treatment arm. Dots indicate individual tumor samples. The Wilcoxon rank-sum test was used for statistical comparisons. E, UMAP plot of lymphoid cell subclusters (left) and projection of immune exhaustion marker gene expression (Pdcdl, Tox, Tigit, and Lag3; right). F, Violin plot representation of immune exhaustion signature scores in CD8+ T-cell subclusters. G, Relative abundance of CD8+ T-cell subclusters in individual treated tumor samples. H, UMAP view of neutrophils colored by cluster membership (left); absolute (center) and relative (right) abundances of neutrophil subclusters in each tumor sample.
Treatment with distinct RAS inhibitors significantly altered the absolute and relative abundances of multiple cell types in the tumor microenvironment of KL2 allograft tumors (Fig. 5B; Supplementary Figs. S7A-S7D, S8A-S8G, and S9A-S9F). The cellular composition of tumors from vehicle-treated mice was dominated by TCs and neutrophils, in line with our FACS data and previous reports of PMN-MDSC and SiglecFHigh neutrophil accumulation (bioRxiv 2023.07.15.549147) and paucity of CD8+ T cells in the TIME of STK11/LKB1-deficient NSCLC (Fig. 5B; ref. 25). RMC-7977 monotherapy and RMC-7977/RMC-4998 combination therapy potently suppressed both the absolute number and relative fraction of TCs at the 7-day time point, whereas sotorasib and RMC-4998 monotherapy had limited impact, in consonance with their differential antitumor activity in the KL2 model (Figs. 5B and 1C). We further observed treatment-induced accumulation of lymphoid cell subpopulations, including several effector T-cell subsets and NK cells, which was more prominent in cohorts with broader RAS pathway suppression (Fig. 5B and 5C; Supplementary Fig. S7D). Critically, exposure to RMC-7977 and RMC-7977 plus RMC-4998 for 7 days yielded the highest tumor-infiltrating lymphocyte (TIL) to TC ratio, whereas long-term rechallenge with RMC-7977 resulted in a lower ratio, which was nonetheless still higher than in vehicle-treated tumors (Fig. 5D). Furthermore, a higher fraction of CD8+ T cells in tumor samples from RMC-7977-rechallenged mice displayed features of exhaustion (Fig. 5E-G). Finally, both acute and long-term treatment with RMC-7977-containing regimens remodeled the myeloid cell compartment (Fig. 5B-H; Supplementary Fig. S8) and resulted in striking suppression and transcriptional remodeling of tumor-associated neutrophils (except for sample RASMULTIi_D7_4; Fig. 5H). Interestingly, the N1 neutrophil cluster that was apparently enriched in RMC-7977-treated tumors displayed lower MAPK signaling output (Supplementary Figs. S8F and S10A) and a transcriptional profile that was reminiscent of normal lung-resident neutrophils, including low expression of SiglecF (Supplementary Fig. S8G; ref 33). Lower MAPK output in response to RMC-7977 was also observed in other myeloid cell subsets—including monocytes and macrophages—raising the possibility that RAS pathway inhibition in innate immune cells may also contribute to immune modulation (Supplementary Fig. S10A). In agreement with this hypothesis, RMC-7977 selectively decreased the viability of bone marrow-derived macrophages (BMDM) polarized ex vivo toward an M2 phenotype (Supplementary Fig. S10B). These findings support the notion that modulation of both innate and adaptive branches of the antitumor immune response contributes to the preclinical activity of RAS inhibitors and highlights evasion of immune surveillance as a bottleneck for the emergence and survival of DTPs.
We also observed notable treatment-induced changes in the transcriptional profiles of nonimmune components of the tumor microenvironment (TME) encompassing fibroblasts and endothelial cells (Supplementary Fig. S9). Prolonged broad RAS pathway suppression with RMC-7977 resulted in decreased proportion of Lrcc15 + myofibroblasts and reciprocal increase in the relative fraction of Pi16 + universal/adventitial fibroblasts (Supplementary Fig. S9A-S9D). Treatment with active state-selective inhibitors, but not sotorasib, increased the relative proportions of endothelial cells in the VEC4, VEC5, and VEC6 states, which are characterized by increased expression of inflammation-related genes (such as Sele and Selp encoding for E- and P-selectin, respectively, in VEC5; Supplementary Fig. S9E and S9F).
Mucinous Regenerative-Homeostatic Transcriptional Program Underpins Active RAS Inhibitor Tolerance
Next, we interrogated the evolving single-cell transcriptional landscape of KL2 TCs following in vivo treatment with RAS pathway inhibitors. TC subclustering yielded eight distinct transcriptional states (TC1–8; Fig. 6A-C; Supplementary Fig. S11A-S11F). In vehicle-treated mice, TC2 constituted the dominant TC subpopulation (42.5%−56.1% of TCs), whereas the fraction of cells in the highly proliferative TC3 cluster ranged from 11.7% to 16.6% (Fig. 6B; Supplementary Fig. S11A). TC2 and TC3 displayed the highest MAPK pathway activity scores from PROGENy (Fig. 6D; ref. 34).
Figure 6.
A mucinous regenerative program supports RASGTP inhibitor tolerance. A, UMAP plot of cancer cells from KL2 allograft tumor samples, colored by cluster. B, Absolute (top) and relative (bottom) abundance of tumor cell clusters in individual tumor samples. C, Proportions and average expression levels of top differentially expressed TC marker genes. D and E, PROGENy pathway activity scores for cells pooled by tumor cluster (D) or tumor sample (E). F, Volcano plot representation of the top differentially expressed genes and selected mucinous/gastrointestinal marker genes in tumors from mice rechallenged with RMC-7977 or treated with vehicle. G, UMAP visualization of mRNA expression of select GI-related genes (Tff1, Tff2, Tff3, Spink4, Muc1, Mucl3, Gkn2, and Agr2). H, Representative IHC images showing enrichment of TFF1, TFF2, and MUC1-expressing tumor cells following long-term rechallenge with RMC-7977 in the KL2, KP2, KL5shCon, and LKR10 models. I, Representative IHC images showing TFF1 expression in mucinous tumor cells in lung specimens obtained from KrasG12C/+;Lkb1−/− GEMMs following long-term treatment with RMC-4998 (top) or RMC-7977 (bottom). Scale bars, 100 μm.
Treatment with RAS inhibitors induced profound changes in the TC transcriptional contexture that were dependent on both the extent and duration of RAS pathway inhibition (Fig. 6B). Exposure to the inactive state-selective RASG12C inhibitor sotorasib for 7 days exerted an overall minor impact on the transcriptome of KL2 cells, which predominantly reflected a shift in the relative proportions of the TC1 and TC2 subpopulations (Fig. 6B). This finding is consistent with the limited preclinical activity of sotorasib in this model, with short-lived tumor stabilization followed by disease progression due to the rapid emergence of adaptive resistance (Fig. 1C; Supplementary Fig. S1I). Interestingly, broad and sustained RAS pathway inhibition with (i) RMC-7977 monotherapy, (ii) RMC-7977 in combination with RMC-4998, or (iii) RMC-4998 plus RMC-4550 (Figs. 3A and 6E) induced a distinct transcriptional state (TC4, color-coded light green) in the majority of surviving TCs in these treatment arms, whereas TC2 and TC3 were potently suppressed (Fig. 6B). TC4 accounted for a small minority of TCs in vehicle- or sotorasib-treated KL2 allograft tumors and was modestly enriched in tumors exposed to RMC-4998. TC4 displayed the highest JAK-STAT, NF-κB, and TNFα PROGENy pathway activity scores and was further enriched in gene expression signatures pertaining to inflammation and antitumor immunity (Fig. 6C and D; Supplementary Fig. S11B and S11C). These results are in agreement with the observed prominent induction of inflammatory cell death and activation of both innate and adaptive antitumor immune responses following a 7-day treatment with RMC-7977 or RMC-7977 plus RMC-4998 (Figs. 3C and D and 5C and D). In addition, we observed increased WNT and TGFβ signaling in TC4 (Fig. 6D) as well as in treatment arms with maximal RAS pathway suppression (Fig. 6E) that may, together with JAK-STAT and NF-κB pathway activation, represent adaptive responses to broad RAS pathway inhibition, as previously reported (Fig. 6D and E; refs. 35-37).
Strikingly, long-term rechallenge with RMC-7977 uniquely reshaped the transcriptome of surviving DTPs, leading to pronounced enrichment of TC8 (color-coded maroon), which accounted for 76.3% ofRMC-7977-tolerant persister cells on average. This state was rare in vehicle-treated tumors or following short-term exposure to RAS pathway inhibitors, with the exception of tumors treated with RMC-7977 that contained a slightly higher fraction of TC8 cells even at this early time point, as well as resurgent tumors following RMC-7977 withdrawal (Fig. 6B). In concordance with the low Ki67 proliferative index of DTPs (Fig. 4C), TC8 was characterized by slow cycling activity (Supplementary Fig. S11A).
Next, we sought to elucidate the distinct molecular features of TC8 that was dominant in DTPs upon long-term rechallenge with RMC-7977. A survey of the top differentially expressed genes in TC8 indicated striking enrichment in transcripts encoding members of the trefoil family of secreted peptides (Tff1, Tff2, and Tff3), gastrokine paralogs ( Gknl, Gkn2, and Gkn3 ), and several mucins (including Muc1, Muc3, Muc3a, Mucl3, Muc4, Muc13, Muc16, and Muc20 ), which are collectively involved in GI regenerative/homeostatic responses and maintenance of barrier integrity (Fig. 6C, F, and G; Supplementary Fig. S11B). These findings were consistent with the frequently mucinous histologic appearance of RMC-7977-tolerant persister cells and were further validated by immunostaining for TFF1, TFF2, and MUC1, which revealed a dramatic enrichment of immune-reactive TCs following long-term rechallenge with RMC-7977 in the KL2 model (Fig. 6H). TFF1-, TFF2-, and MUC1-expressing DTPs also accumulated in the KP2, KL5shCon, and LKR10 models as well as in the KrasG12C/+;Lkb1−/− GEMMs, thus confirming engagement of this tolerogenic program across a broad range of NSCLC comutational models driven by either KrasG12C or KrasG12D (Fig. 6H and I). It should be noted that TFF1 expression was not uniform among DTPs, with both inter- and intra-tumoral heterogeneities that were sometimes apparent even among adjacent cells within a single glandular structure (Fig. 6H). Importantly, TFF1 expression in DTPs was also observed in the KL5shKeap1 and KL5shSmarca4 models, albeit to a lower extent (Supplementary Fig. S12A), as well as in the sotorasib-resistant KL2 SR1 and KL4 SR1 models (Supplementary Fig. S12B). Thus, it is likely that multiple distinct or partially overlapping transcriptional programs may contribute to the emergence of DTPs and that these may be further shaped by the comutational status of individual tumors. Baseline expression of TFF1 in allograft tumors from RMC-7977-naïve, vehicle-treated mice also exhibited significant variability, ranging from rare TTFl-positive TCs in the KL2, KL5shCon, and LKR10 models to ~10% to 15% immune reactivity in the KP2 model (Fig. 6H). Serial analysis of TFF1 expression over time in KL2 and KP2 tumors following treatment with RMC-7977 revealed progressive enrichment in the fraction of TFF1-expressing cells, which was most notable following prolonged drug rechallenge (Supplementary Fig. S12C).
We identified several transcription factor-encoding transcripts as among the most significantly enriched in TC8. These included (i) Hnf4a, encoding a master regulator of GI differentiation and function (38); (ii) Onecut2 and Onecut3, members of the ONECUT family of transcription factors; ONECUT2 has been identified as a master regulator of androgen signaling and mediator of neuroendocrine differentiation and metastasis and has been reported to promote KRAS-mutant NSCLC, whereas ONECUT3 has been linked with stemness and evasion of NK cell-mediated immune surveillance in pancreatic ductal adenocarcinoma (39–41); and (iii) Sox9, a major orchestrator of lung branching morphogenesis, which has also been identified as a driver of resistance to NK-mediated cytolysis in NSCLC (Fig. 6C; Supplementary Figs. S11B and S12D; ref. 42). Increased nuclear accumulation of HNF4α, ONECUT2, and SOX9 in DTPs was further confirmed by IHC, although both the baseline expression and degree of induction of individual TF seemed to vary across the different models (Supplementary Fig. S12E and S12F).
In agreement with the upregulation of a GI/mucinous transcriptional program in RMC-7977 DTPs, the PROGENy pathway analysis revealed increased WNT pathway signaling in TC8 and enrichment of a gastric/endodermal transcriptional module that was previously identified in a subpopulation of TCs from autochthonous tumors in the KP GEM model (Fig. 6D and E; Supplementary Fig. S11D; ref. 43). Gene set enrichment analysis (GSEA) further demonstrated a significant enrichment of xenobiotic metabolism genes from the mouse hallmark gene set collection, supporting the notion that enhanced drug metabolism/detoxification prowess may at least partially underpin RMC-7977 tolerance (Supplementary Fig. S11E). Signatures related to inflammation and antitumor immunity were increased in TCs from RMC-7977-rechallenged compared with vehicle control-treated samples but notably decreased compared with tumor samples from mice treated with RMC-7977 for 7 days (Supplementary Fig. S11E). Finally, although an increased expression of vimentin was observed in TC8 cells, other markers of EMT were not significantly enriched, and the TC8 state did not seem inherently more mesenchymal compared with either TC2 or TC1 (Fig. 6C; Supplementary Fig. S11B and S11D-S11F). Of note, we observed considerable overlap in patterns of gene set enrichment between our DTP program and KRAS-mutant lung adenocarcinoma (LUAD) with high expression of TFF1 and MUC1 from the The Cancer Genome Atlas (TCGA) dataset (Supplementary Fig. S13A).
Gastrointestinal/Mucinous Differentiation is Associated with Impaired Response to RAS Inhibition in Human KRAS-Mutant NSCLC
To determine whether the identified GI/mucinous persister transcriptional program also promotes tolerance to RAS inhibition in human KRAS-mutant NSCLC models, we initially exposed a panel of KRAS-mutant NSCLC cell lines (encompassing both KRAS G12C and non-G12C alleles) to treatment with 40 nmol/L of RMC-7977 (a concentration comparable to the steady-state Cmax in mouse following in vivo treatment with 25 mg/kg PO QD) or vehicle for 6 days and compared the expression of selected marker proteins, including MUC1 in surviving TCs. Remarkably, treatment with RMC-7977 resulted in a significant increase in the fraction of MUC1-high cells by FACS in seven out of the nine tested cell line models (Fig. 7A). We posited that longer exposure to RMC-7977 may be required to trigger the persister program in some models; to test this hypothesis, we further evaluated the expression of MUC1 following prolonged exposure to RMC-7977 in the A549, H460, and HCC44 cell lines and observed a significant enrichment of MUC1-high cells in all three models (Fig. 7B). Increased expression of MUC1 following a 6-day treatment with RMC-7977 was confirmed by Western blotting and immunofluorescence (Fig. 7C and D; Supplementary Fig. S13B) and was further validated by IHC in H460 and H1373 xenograft models, together with additional markers of the mucinous DTP program (TFF1, AGR2, CDX2; Fig. 7E). Importantly, enrichment of MUC1-high cells was also observed in response to a 6-day treatment with RMC-4998 or sotorasib in several KRASG12C -mutant NSCLC cell lines (including H23, HCC44, H358, and H1373; Fig. 7F). Thus, the emergence of the identified persister state may represent a more broadly applicable mechanism of tolerance to robust and sustained RAS pathway inhibition with either mutant-selective or broad-spectrum RAS inhibitors that target either the inactive or active state of RAS oncoproteins. Mechanistically, we observed high baseline expression of phospho-ERK1/2 in mucinous lung tumors in the KrasG12G/+;Lkb1−/− GEMM (Supplementary Fig. S14A), in concordance with a previous report in a mouse model of BrafV600E-driven NSCLC (44). Furthermore, we noted a relative preservation of phospho-ERK1/2 immune reactivity in clusters of mucinous cells in response to acute exposure to RMC-7977 (Supplementary Fig. S14B). Finally, in each of the H2122, HCC44, and H1373 KRASG12G -mutant NSCLC cell lines, we observed a significantly higher fraction of TCs that expressed high levels of phospho-ERK1/2 (pERKhigh) in the MUC1High compared with the MUC1low compartment, both in DMSO-treated cells and in response to a 6-hour treatment with RMC-7977 (Supplementary Fig. S14C and S14D). Thus, relative refractoriness to MAPK suppression in mucinous TCs may at least partially underpin their propensity to tolerate RAS inhibition.
Figure 7.
Mucinous differentiation is associated with RAS inhibitor tolerance in human KRAS-mutant NSCLC. A, Flow cytometry-based assessment of the fraction of MUC1High-expressing tumor cells following treatment with vehicle or RMC-7977 (40 nmol/L) for 6 days in a panel of nine KRASG12C (H1792, H23, H727, HCC44, H358, H1373, H441) and KRASnon-G12C-mutant (H460, A549) NSCLC cell lines. B, Long-term (>4 weeks) in vitro exposure to RMC-7977 promotes accumulation of MUC1High tumor cells. Error bars represent SEM from technical triplicates; a paired t test was used for comparisons between treatment arms in both A and B. Asterisks denote statistical significance: *, P ≤ 0.05; **, P ≤ 0.01; ***, P≤ 0.001; ****, P ≤ 0.0001. C, Western blot analysis of MUC1 protein abundance in whole cell lysates following in vitro treatment with vehicle or RMC-7977 (40 nmol/L) for 6 days. Similar results were observed with a second well-characterized antibody [MUC1-C (D5K9I) XP Rabbit mAb #16564, Cell Signaling Technology] against the C-terminal subunit of MUC1 (Supplementary Fig. S13B). D, Representative IF images depicting increased MUC1 expression (green fluorescence) in the H460 (top row) and HCC44 cell lines following long-term (>4 weeks) in vitro treatment with RMC-7977. DNA is stained with DAPI. Scale bars, 50 μm. E, IHC expression of TFF1, MUC1, AGR2, and CDX2 in the H460 and H1373 CDX models in response to treatment with vehicle (top) or RMC-7977 for 21 days (bottom). Scale bars, 100 μm. F, Flow cytometry-based assessment of the proportion of MUC1High tumor cells following treatment for 6 days with (i) vehicle, (ii) sotorasib (1 μmol/L), (iii) RMC-4998 (100 nmol/L), and (iv) RMC-7977 (40 nmol/L). The A549 and H460 cell lines represent negative controls for RASG12C-selective inhibitors. G and H, RECIST v1.1-based overall response rate (G) and Kaplan-Meier estimates of PFS and OS (H) in patients with advanced KRASG12C-mutant NSCLC treated with sotorasib or adagrasib monotherapy (MD Anderson clinical cohort), according to the presence or absence of mucinous features in baseline biopsy reports. The log-rank test was used for comparison of PFS and OS. HRs and corresponding CIs were estimated with the use of a stratified Cox proportional hazards model with adjustment for clinical variables [age, history of brain metastasis, prior lines of therapy for metastatic disease (0 vs. ≥1), PS (0–1 vs. 2)] and histologic subtype (adenocarcinoma vs. other histologic subtypes). P ≤ 0.05 was considered statistically significant.
Lastly, we examined whether the presence of mucinous histologic features at baseline impacted the clinical outcomes with sotorasib or adagrasib, in a cohort of 130 patients with metastatic KRASG12C -mutated NSCLC treated with sotorasib or adagrasib at MD Anderson Cancer Center. Strikingly, patients harboring KRASG12G-mutant tumors with mucinous features (as identified in the diagnostic pathology report) exhibited significantly lower ORR (35.1% vs. 7.7%; P = 0.038, Fisher’s exact test) and shorter PFS [median PFS 5.3 vs. 2.5 m, MV HR 3.23 (95% CI, 1.75–5.95), P = 0.002, log-rank test] and overall survival (OS) [median OS 9.5 vs. 2.9 m, MV HR 2.54 (95%CI, 1.35–4.77), P = 0.018, log-rank test] with sotorasib or adagrasib (Fig. 7G and H), even after adjustment for clinical covariates and histologic subtype (Fig. 7H; Supplementary Fig. S15A; Supplementary Table S2). The presence of mucinous features remained independently associated with shorter PFS even after adjustment for key adverse comutations (KEAP1, SMARGA4, and GDKN2A; Supplementary Table S2). Immune reactivity for CDX2, a commonly used marker of GI differentiation, was also associated with inferior PFS and OS with sotorasib and adagrasib, in a subset of patients with available IHC data (Supplementary Fig. S15B).
DISCUSSION
In this study, we evaluated the antitumor activity of the novel RAS(ON) multiselective tricomplex inhibitor RMC-7977 and dissected mechanisms of response and tolerance to active RAS inhibition across diverse KRASG12C-mutant NSCLC comutational models, which capture clinically challenging and poorly prognostic disease subsets. Through an integrated, temporally resolved analysis of clinical, pathologic, and single-cell transcriptional responses to distinct classes of RAS inhibitors, we established a comprehensive in vivo atlas of multimodal acute and chronic RAS pathway inhibition in the NSCLC ecosystem. Critically, we uncovered a regenerative mucinous transcriptional program that supports long-term TC persistence in response to RAS inhibition, with implications for clinical outcome prediction and the development of novel rational therapeutic strategies.
With multiple RAS-targeting assets in clinical development, an enduring question in the field is whether mutant-selective (covalent or reversible, inactive, active, or dual state) or broad-spectrum inhibitors of RAS oncoproteins (pan-KRAS or pan-RAS, inactive or active state) will yield superior therapeutic benefit. Our results indicate that broad-spectrum inhibitor targeting of GTP-bound RAS isoforms (including mutant and WT KRAS, HRAS, and NRAS variants) confers enhanced and differentiated antitumor activity compared with the G12C-selective covalent inhibitors sotorasib and RMC-4998 or their combinations with the SHP2 inhibitor, RMC-4550. This is supported by (i) significantly longer TTD or RFS with RMC-7977 or RMC-7977 plus RMC-4998 across syngeneic, autochthonous, and CDX models of difficult-to-treat KRAS-mutant NSCLC subsets, (ii) rare emergence of on-treatment adaptive/acquired resistance even after prolonged rechallenge at high tumor burden, and (iii) retained antitumor activity against models with acquired resistance to either sotorasib or RMC-4998. Notably, mechanisms of acquired resistance to RAS(ON) G12C-selective inhibitors have not hitherto been described and were dominated in our study by high-level amplification of KrasG12C, secondary KrasG12D mutations, or concurrent copy gains of KrasG12C, Hras, Nras, and multiple RTKs.
Our findings also highlight potential opportunities for targeted deployment of RAS(ON) multi- and mutant-selective inhibitor combinations. This is particularly relevant for difficult-to-treat disease subgroups with cooccurring alterations in KEAP1 or SMARCA4, which exhibit relative resistance to sotorasib and RMC-4998 but responded robustly to the doublet of RMC-7977 plus RMC-4998, in both syngeneic and CDX models. Confirmation of this tantalizing hypothesis may be informed by data from the ongoing clinical trials of the related investigational agents RMC-6236 or RMC-6236 plus RMC-6291.
Mechanistically, the diversified antitumor activity profile of RMC-7977 with or without RMC-4998 is underpinned by (i) potent and sustained suppression of phosphorylated ERK1/2 with quelling of feedback MAPK pathway reactivation, in agreement with previous reports, (ii) robust and early induction of apoptosis, (iii) more potent inhibition of TC proliferation, (iv) notable induction of inflammatory cell death programs including necroptosis and pyroptosis, (v) favorable remodeling of the TIME to boost antitumor immunity, and (vi) inhibition of neoangiogenesis. Necroptosis and pyroptosis constitute nonapoptotic regulated cell death modules that can trigger and amplify antitumor immunity (45, 46). Induction of pyroptotic cell death has previously been reported in response to EGFR TKIs, MEK inhibitors, and other targeted therapies but so far not with direct RAS inhibitors (47). Interestingly, induction of inflammatory cell death was more prominent with RMC-4998 and RMC-7977 either alone or in combination, particularly at the 7-day time point, raising the possibility that sustained suppression of RAS signaling output beyond a critical threshold may be required for engagement of inflammatory cell death pathways. This notion is further supported by significant enrichment of a distinct transcriptional state—TC4—in treatment arms with broad suppression of active RAS signaling. This state was characterized by dramatically increased JAK-STAT, NF-κB, and TNFα pathway signaling in PROGENy and exhibited the highest scores for multiple MSigB modules associated with inflammation. Taken together, these results support the prominent activation of TC-intrinsic inflammatory pathways in response to active RAS inhibition—particularly with broad-spectrum inhibitors such as RMC-7977—and nominate evasion of inflammatory death as a critical early evolutionary bottleneck, in addition to apoptosis resistance, for TC survival in this setting (48).
In accordance with earlier studies, treatment with diverse direct and indirect RAS inhibitors remodeled the TIME and triggered an influx of effector immune cell types including CD8+, CD4+, γδ T cells, and NK cells. Accumulation of TILs and γδ T cells/NK cells—which may be particularly relevant the for immune clearance of STK11/LKB1-deficient tumors that exhibit impaired antigen presentation—seemed to be most pronounced in response to RMC-7977-containing regimens, resulting in the highest TIL/TC and NK cell/total cell ratios. Intriguingly, RMC-7977 selectively recalibrated the myeloid cell compartment by triggering not only the dramatic depletion of tumor-infiltrating neutrophils but also a shift in their transcriptional landscape with enrichment of the more normal-like NC1 cluster (33). Given the observed potent suppression of MAPK signaling output in NC1, it is plausible that both direct effects on the RAS/MAPK axis in neutrophils and indirect effects (via reduced TC secretion of growth factors, cytokines, and chemokines) contribute to the distinctive reshaping of the tumor myeloid cell immune contexture in response to RMC-7977 (30). The extent to which specific quantitative or qualitative immune modulatory effects of broad-spectrum active RAS inhibitors contribute to their overall antitumor activity will require future dedicated studies.
Robust induction of an acute desmoplastic response was another salient feature of treatment with RMC-7977 in most models; this was evident as early as 7 days following therapy initiation. Interestingly, long-term rechallenge with RMC-7977 reshaped the transcriptional composition of tumor-associated fibroblasts and promoted the enrichment of a Pi16-high cluster that is reminiscent of universal/adventitial fibroblasts (49). The potential impact of CAF “normalization” and extracellular collagen deposition to the activity of ongoing RAS-GTP inhibition remains to be determined. Finally, we observed a reduced tumor vascular density in RMC-7977-containing treatment arms, particularly after longer treatment. This raises the possibility that suppression of neovascularization—either indirectly, through reduced TC secretion of VEGF, CXCLs, and other proangiogenic factors, or via direct effects (50) on endothelial cells—may also contribute to the antitumor activity of broad-spectrum active RAS inhibitors (51).
Crucially, our work identifies a regenerative mucinous/GI-like transcriptional state that supports long-term tolerance of broad-spectrum GTP-bound RAS inhibition in vivo. This program is characterized by both mucinous differentiation (marked by elevated expression of several mucins, including the prototypical MUC1, gastrokines, and trefoil factor family peptides) and progenitor activity and bears similitude to a high plasticity pre-EMT state identified in lineage tracing studies in GEM models and a SOX9-expressing WNT and FGF-responsive alveolar epithelial progenitor population previously implicated in the regeneration of severely injured lung and the establishment of macrometastases (42, 43, 52, 53). Indeed, RMC-7977 persister cells expressed high levels of SOX9 and several transcription factors involved in hepatopancreatobiliary, intestinal, and neural/neuroendocrine differentiation, such as HNF4a, ONECUT2, and ONECUT3. Although a degree of cross-species and intercell line heterogeneity was evident in the enrichment patterns of individual proteins— which may further reflect specific experimental conditions— we nonetheless observed an enrichment of key components of the persister program across the majority of tested human and murine models of both KRASG12C- and KRASnon-G12C-mutant NSCLC with either retained or absent TTF1 expression. Accumulation of MUC1High DTPs was also noted in response to high concentrations of sotorasib and RMC-4998 across several KRASG12C-mutant NSCLC cell lines. Interestingly, elevated expression of mucin-encoding transcripts and upregulation of WNT signaling in DTPs have also previously been reported in response to erlotinib and other targeted therapies in experimental in vivo models of oncogene-addicted NSCLC subsets as well as on-treatment biopsies of residual tumors from patients with advanced NSCLC treated with TKIs or neoadjuvant chemotherapy (42, 54, 55). Therefore, an alveolar epithelial progenitor-like program with mucinous lineage differentiation or transdifferentiation may constitute a universal tolerogenic state in the context of extreme tissue injury triggered by potent RAS pathway inhibitors as well as other effective anticancer therapies. This circuit may be especially relevant with broader-spectrum inhibitors of active, GTP-bound RAS that circumvent adaptive MAPK pathway reactivation. Delineation of shared and private components of DTP programs in response to distinct systemic therapies is an area of active study.
In accordance with these preclinical findings, we found inferior clinical outcomes with sotorasib and adagrasib in patients with advanced KRASG12C-mutated NSCLC bearing tumors with mucinous histologic features. Further work will be required to ascertain the predictive utility of mucinous differentiation in randomized clinical trials of mutant-selective and broad-spectrum RAS inhibitors and to distinguish prognostic from predictive effects. It will also be critical to determine whether the presence of mucinous histologic features per se or the baseline expression of more specific subprograms and markers of the alveolar epithelial progenitor/mucinous TC subpopulation can most accurately stratify clinical outcomes with RAS inhibitors. In this context, it is important to note that although KRASG12C inhibitors have so far consistently yielded lower ORR and shorter PFS in colorectal and pancreatic cancers when compared with lung cancer, deep and durable responses have also been observed in patients with KRASG12C-mutated GI tumors, supporting the need for further refinement of molecular indicators of RAS inhibitor persister biology (3, 56, 57). It should be further emphasized that our study does not distinguish between expansion of a preexisting rare population of DTPs under the selective pressure imposed by broad-spectrum active RAS inhibition versus treatment-induced plasticity with de novo establishment of the DTP program either stochastically or selectively in a subset of genetically or epigenetically “primed” TCs. Indeed, a lack of baseline TTF1 expression, which is associated with derepression of a latent gastric differentiation program, may predispose to the emergence of the mucinous DTP program (23, 58). It is indeed likely that both models are in operation, as evidenced by the heterogeneous and differential baseline expression and inducibility of distinct mucins and other DTP-enriched proteins. Differential propensities of individual TCs in untreated tumors to evolve as DTPs are further supported by prior findings of lineage promiscuity in surgically resected NSCLC (42).
How could mucinous differentiation impact in vivo RAS inhibitor tolerance mechanistically? Based on the robust engagement of inflammatory pathways in response to broad-spectrum active RAS inhibition, it is likely that evasion of inflammatory cell death and immune surveillance constitute important points of convergence of DTP programs. Construction of a TC glycocalyx can guard against immune cell recognition and attack by sterically shielding molecular epitopes and providing a physical and biochemical barrier to the establishment of immunologic synapses (59). Furthermore, MUC1 has been specifically linked with evasion of NK cell-mediated immune surveillance, which may be particularly relevant for antigenically heterogeneous tumors, including STK11/LKB1-deficient NSCLC; it is also notable that both ONECUT3 and SOX9—which are both significantly enriched in RMC-7977 persisters—have also been associated with NK cell immune escape and downregulation of immune-related and inflammation response genes (40, 42). Furthermore, trefoil factor family peptides, in their capacity as extracellular lectins, can block interactions between natural ligands and their cognate membrane receptors and exert antiinflammatory effects (60). A mechanical barrier could potentially also impose constraints on drug delivery and TC uptake, whereas intracellular TFFs and mucins could further act as molecular chaperones that quench intracellular stress pathways or reactive oxygen species/reactive nitrogen species (ROS/RNS) scavengers (61). Pleiotropic additional tumor-promoting roles have been previously ascribed to MUC1 (and its MUC1C subunit) and other mucins, encompassing evasion of apoptosis, EMT, epigenetic reprogramming, induction of stemness/pluripotency, and promotion of lipid metabolism; in addition, feed-forward engagement of WNT and JAK1-STAT3 signaling and activation of the PI3K/AKT and MAPK pathways as a result of MUC1 association with EGFR and other RTKs at the cell membrane have been previously reported (61, 62). Supporting the premise that mucinous tumors may exhibit relative refractoriness to MAPK pathway suppression, we observed higher baseline expression of phospho-ERK1/2 and relative retention of phospho-ERK1/2 immune reactivity in response to acute RAS inhibition in mucinous murine lung tumors in vivo and in subpopulations of KRASG12C-mutant NSCLC cells in vitro. Additional work will be required to delineate the precise mechanisms that underpin mucinous differentiation-related tolerance to RAS inhibition.
Our work provides important insights for developing novel rational therapeutic strategies that may maximize the efficacy of active RAS inhibitors. Ostensibly, RAS(ON) multiselective inhibitor (using the investigational agent RMC-6236)- anchored combination strategies may focus on thwarting early adaptive responses or eliminating DTPs. Our results nominate the JAK-STAT, WNT, NF-κB, TGFβ, and FGF pathways as candidate mediators of acute active RAS inhibitor tolerance in vivo that may be tractable pharmacologically. This is consistent with previous reports of transient cytoprotective feedback activation of STAT3 via an IL6-dependent autocrine loop in response to MEK1/2 inhibition in KRAS-mutant NSCLC and of macrophage-derived TGFβ as a driver of KRAS-independent growth in PDAC (35–37). An appealing alternative is to target the cell surfaceome of DTPs with antibody-drug conjugates (ADC), bispecific T-cell engagers, or CAR T cells. In this regard, DS3939, a tumor-associated MUC1 (TA-MUC1) ADC directed against aberrantly glycosylated MUC1, recently entered early stages of clinical development in solid tumors; in addition, both an autologous CAR T targeting the extracellular domain of cleaved MUC1 and an allogeneic CAR T against the C-terminal subunit of MUC1 (MUC1-C) anti-MUC1-C CAR T are undergoing early-phase clinical evaluation (NCT04020575 and NCT05239143), whereas MUC1 bispecific T-cell engagers have demonstrated antitumor activity in preclinical models (63). Multiple antibodies targeting alternative mucins are also being evaluated in the context of ADCs, CAR Ts, and T-cell engagers and may exhibit clinical utility for the extinction of DTPs in response to active RAS inhibitors.
Finally, our study provides the first report to date that acquired resistance to the RAS(ON) multiselective inhibitor RMC-7977 is associated with EMT and dedifferentiation with loss ofTTF1 expression. Further work is required to assess the relevance and prevalence of these putative mechanisms of resistance, in particular, following initial treatment with a RAS(ON) multiselective inhibitor alone or in combination with a RAS(ON) G12C-selective inhibitor.
In conclusion, our study establishes a comprehensive in vivo atlas of RAS inhibition in KRAS-mutant NSCLC and yields critical novel insights into the mechanisms of response and tolerance of broad-spectrum active RAS inhibitors, thus paving the way for optimized and tailored evaluation of this promising class of RAS-targeting agents in the clinic.
METHODS
Establishment of GEMM-Derived KrasG12C-Mutant NSCLC Cell Lines and Syngeneic Allograft Models
All animal studies were conducted according to protocols approved by the University of Texas MD Anderson Cancer Center Institutional Animal Care and Use Committee (IACUC). KrasLSL-G12C/+ (B6;129S4-Krasem1Ldow/J; The Jackson Laboratory) mice (on C57BL/6NJ background) were crossed with Stk11/Lkb1F/F (on a C57BL/6 background) to generate compound conditional KrasLSL-G12C/+;Stk11/Lkb1F/wt mice. KrasLSL-G12C/+;Stk11/Lkb1F/wt mice were crossed with Stk11/Lkb1Fl/F to generate KrasLSL-G12C/+;Stk11/Lkb1Fl/Fl (KG12CL) mice. Similarly, Trp53LSL-R172H/+ mice (on a C57BL/6 background) were crossed with KrasLSL-GG2C/+ to generate KrasLSL-G12C/+;Trp53LSL-R172H/+ (KG12CP) models. Six- to eight-week-old mice were transduced with 2.5 × 10E7 plaque-forming units (pfu) of Adeno-Cre (University of Iowa Viral Vector Core) via intranasal instillation, as previously described (64). Mice were monitored at least three times weekly and were euthanized humanely when they exhibited signs of ill health. Explanted autochthonous lung tumors were processed to a single-cell suspension using a TC isolation kit according to the manufacturer’s instructions (Miltenyi Biotec Inc., #130–110-187) and were propagated ex vivo in order to establish stable tumor-derived cell lines. The KrasG12C-mutant;Stk11-deficient KL1, KL2, KL4, and KL5 cell lines were derived from individual lung tumors arising in male KG12CL mice (KL1 and KL2 originated from distinct tumors in the same mouse), whereas the KrasG12C-mutant;p53-deficient KP2 model originated from a lung tumor in a female KG12CP mouse. STK11 loss in KG12CL models and stabilization of p53 (due to engineered dominant-negative and dominant gain-of-function Trp53R172H mutation) in the KP2 model were validated by Western blot analysis (Supplementary Fig. S1). We further established stable isogenic derivatives of the KL5 cell line with shRNA-mediated knockdown of Keap1 (clone TRCN0000099445, target sequence: 5′-CGGAAGCAAATTGAT-CAACAA-3′), Smarca4, (clone TRCN0000288151, target sequence: 5′-CGCCCGACACATTATTGAGAA-3′), or non-targeting shControl (TRC2 pLKO.5-puro shRNA Control Plasmid DNA, Sigma-Aldrich, #SHC202) by lentiviral transduction. Transduced cells were selected with puromycin 5 μg/mL (Thermo Fisher #A1113803) and were maintained under the same antibiotic concentration until implantation in mice. The LKR10 KraSG12D-mutant LUAD cell line was derived from a lung tumor in Dr. Tyler Jacks’ laboratory (129S4/SvJae genetic background).
In Vivo Treatment with RAS Inhibitors in Syngeneic and Xenograft Subcutaneous Models
For experiments in syngeneic models, 6 to 8-week-old male (for the KL1, KL2, KL4, KL5shCON, KL5shKEAP1, KL5shSMARCA4 models) or female (for the KP2 model) C57BL/6 mice or male 129/Sv mice (for the LKR10 model) were utilized. Approximately 2 × 106 TCs of the indicated models were injected subcutaneously in the flank of recipient mice in a 100 μL suspension, consisting of 50% Matrigel Basement Membrane Matrix (Corning, #47743–716) and 50% HBSS (Gibco #14175095). Mice bearing tumors with a volume ranging from 200 to 250 mm3 (for long-term therapeutic antitumor activity studies) and 450 to 500 mm3 (for scRNA-Seq, FACS, IHC, and Western blotting) were randomly assigned to receive treatment with the following: (i) vehicle, (ii) sotorasib [100 mg/kg via oral gavage (OG), QD], (iii) RMC-4998 (100 mg/kg, OG, QD), (iv) RMC-4550 (30 mg/kg, OG, QD), (v) sotorasib (100 mg/kg OG, QD) + RMC-4550 (30 mg/kg, OG, QD), (vi) RMC-4998 (100 mg/kg OG, QD) + RMC-4550 (30 mg/kg, OG, QD), (vii) RMC-7977 (25 mg/kg, OG, QD), or (viii) RMC-7977 (10 mg/kg, OG, QD) + RMC-4998 (100 mg/kg OG, QD). A lower dose of RMC-7977 (10 mg/kg, OG, QD) was chosen for studies in CDX models and in combination with RMC-4998 due to weight loss concerns. No significant weight loss (≥20% of body weight) was observed in any of the treatment cohorts with the specified doses. RAS(ON) inhibitors were obtained from Revolution Medicines under a sponsored research agreement and diluted in DMSO (Sigma, #900185)/PEG400 (Sigma, #25322–68-3)/Solutol HS15 (Sigma, #70142–34-6)/water, using the formulation made of 10/20/10/60 (%v/v/v/v) as previously described (16). RMC-4550 was provided by Revolution Medicines under a former collaboration agreement with Sanofi. The same vehicle formulation was used for the control group. Sotorasib (MedKoo, # 207085) was formulated in 2% hydroxypropyl methylcellulose (HPMC, Sigma, #H8384) and 1% Tween 80 (Sigma, #P4780), according to the manufacturer’s instructions. Treatment was administered daily until tumor volume exceeded 1,000 mm3 or until day 60 post randomization in mice without apparent on-treatment tumor outgrowth. Treatment was subsequently withdrawn for up to 30 days in order to evaluate time to recurrence (defined as the time from treatment withdrawal on D60 to TV ≥ 500 mm3) and determine the fraction of mice in each treatment cohort that reached the “cure” endpoint (defined as no increase in TV for 30 days posttreatment cessation). Resurgent tumors following drug withdrawal were allowed to reach a TV ≥ 1,000 mm3, at which point mice were rechallenged with RMC-7977 (25 mg/kg OG, QD) or RMC-7977 (10 mg/kg OG, QD) plus RMC-4998 (100 mg/kg OG, QD) for 6 to 8 weeks (long-term rechallenge cohort). Longitudinal bidirectional digital caliper tumor measurements were obtained at least thrice weekly and used to calculate tumor volume based on the formula V = W2L/2. Experiments involving the H460 and H1373 xenograft models were performed in NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. All animal studies with the exception of preclinical antitumor activity studies in the NCI-H2122 and NCI-H2030 xenograft models were conducted at the University of Texas MD Anderson Cancer Center, according to protocols approved by the IACUC.
Antitumor activity studies involving the NCI-H2122 and NCI-H2030 xenograft models were conducted by Revolution Medicines, under contract research agreements with WuXi AppTec Co., Ltd., and Pharmaron, respectively. All CDX mouse studies and procedures related to animal handling, care, and treatment complied with all applicable regulations and guidelines of IACUC at each facility with their approvals. Six- to 12-week-old female BALB/c nude and NOD/SCID mice were used for NCI-H2122- and NCI2030-based experiments, respectively. Animal vendors include Beijing AniKeeper Biotech Co. Ltd. and Shanghai Sino-British SIPPR/BK Laboratory Animal Co., LTD. For antitumor activity studies in CDX tumor models, mice were implanted subcutaneously at the right flank with 5 × 106 (NCI-H2122) or 1 × 107 (NCI-H2030) cancer cells in 0.2 mL of PBS supplemented with BD Matrigel (1:1). Mice were randomized to different treatment groups at an average tumor volume of 150 to 200 mm3. Animals were monitored daily, and body weight and tumor volume were recorded twice weekly.
Preclinical Studies in GEMMs
Six- to eight-week-old KrasLSL-G12C/+;Stk11/Lkb1Fl/Fl (KG12CL) mice underwent intranasal instillation of 2.5 × 10E7 plaque-forming units (pfu) of Adeno-Cre (University of Iowa Viral Vector Core). Weekly monitoring of lung tumor burden with micro-computed tomography (μCT) commenced 8 weeks postadenoviral Cre transduction with the use of the Bruker micro-CT SkyScan 1276 system at the MDACC imaging core facility. Reconstructed CT images were analyzed with the Bruker micro-CT Analyzer v.1.18.8.0., and the largest diameter of identified lung tumor lesions was recorded according to the principles of RECIST 1.1. Mice were randomized in treatment arms when the overall tumor burden, represented by the sum of the largest diameters of target lesions (1–3 tumors per mouse), reached 2 to 5 mm. GEMMs were treated daily with either vehicle, RMC-7977 (25 mg/kg OG QD), or RMC-4998 (100 mg/kg OG QD). Tumor burden was assessed with weekly μCT scans from the day of randomization until disease progression or completion of 60 days on-treatment without evidence of disease progression (study endpoint). Best overall response (BOR) was determined as the maximum percent tumor shrinkage recorded from the beginning of treatment until the conclusion of the experiment. PFS in the GEMM-based experiments was defined as the time from randomization to disease progression, defined as ≥20% increase in the sum of the largest diameters of target lesions or death/loss of clinical condition due to progressive lung cancer. Two mice with limited tumor burden died during anesthesia induction and were censored at the time of death.
Models with Acquired Resistance to Mutant-Selective Direct RASG12C Inhibitors
C57BL/6 mice bearing syngeneic KL2, KL4, or KL5 subcutaneous allograft tumors were treated with sotorasib (100 mg/kg OG QD) at a baseline tumor volume of 200 to 250 mm3 until eventual on-treatment tumor outgrowth (TV ≥ 1,000 mm3). Tumors were harvested, and individual tumor-derived sotorasib-resistant cell lines were established (referred to as KL2 SR1, KL4 SR1, and KL5 SR1) and used in subsequent experiments. Mice bearing sotorasib-resistant allograft tumors were initially retreated with sotorasib 100 mg/kg OD QD to confirm in vivo resistance before randomization (when the volume of individual tumors reached 450–500 mm3) to either RMC-7977 (25 mg/kg OG OD) or RMC-7977 (10 mg/kg OG OD) plus RMC-4998 (100 mg/kg OG OD). To investigate resistance to RMC-4998, individual tumors exhibiting on-treatment outgrowth with RMC-4998 (KL5shCon RMC-4998 R1-R3 and KP2 RMC-4998 R1-R3) were harvested when TV exceeded 1,000 mm3; part of the tumor was processed for DNA extraction for WES, and the remaining pieces were processed to a single-cell suspension under sterile conditions. Approximately 2 × 106 live cells mixed with 1:1 with Matrigel (Corning, #47743–716) were injected subcutaneously into the flank of syngeneic C57BL/6 recipient mice under continuous drug exposure. Mice harboring tumors measuring 450 to 500 mm3 were randomized to treatment with either RMC-7977 (25 mg/kg OR QD) or a combination of RMC-7977 (10 mg/kg OG OD) and RMC-4998 (100 mg/kg OG OD).
Cell Culture and In Vitro Treatments
Cells were cultured in a humidified incubator at 37°C under 5% CO2. Cancer cell lines were maintained in RPMI (HyClone RPMI 1640) and supplemented with 1% penicillin/streptomycin (Gibco #15140122) and 10% heat-inactivated FBS (HyClone, #SH30910.03). Cells were maintained in the logarithmic growth phase and were passaged using 0.25% Trypsin-EDTA (Thermo Fisher, #25300054). Human cell lines were obtained from the American Type Culture Collection and Dr. John Minna (UT Southwestern). Cell lines were authenticated by DNA fingerprinting and tested for Mycoplasma prior to use in experiments. Cells were treated with RMC-7977 (40 nmol/L), RMC-4998 (100 nmol/L), and sotorasib (1 pmol/L; all drugs were dissolved in DMSO) for 6 days or vehicle (DMSO), and drug-containing media was replenished every 48 hours.
Genomic DNA Extraction and WES Analysis
Cell lines (KL1, KL2, KL4, KL5, KP2) were harvested with trypsin, washed with PBS, centrifuged, and processed for genomic DNA extraction with the Blood & Cell Culture DNA Mini Kit (QIAGEN, #13323), according to the manufacturer’s instructions. Explanted allograft tumors were digested with Proteinase K, and genomic DNA was extracted using the QIAGEN Blood & Cell Culture DNA Mini Kit and according to the manufacturer’s instructions. DNA quality was evaluated using the Agilent TapeStation automated electrophoresis system. Samples meeting DIN > 6 criteria were processed for sequencing at the MDACC Advanced Technology Genomics Core facility. Libraries were constructed using the Twist NGS Library Preparation and Capture Kits (#102037) with the Mouse Twist Comprehensive Exome Panel (https://www.twistbioscience.com/resources/product-sheet/twist-mouse-exome-panel). WES was performed on the NovaSeq 6000 instrument (Illumina) utilizing the S4 flow cell, with NovaSeq Control Software v1.7.5 at 100 nt paired end with dual 10 index reads. The demultiplexing was executed using bcl2fastq v2.20.0.
Sequencing reads were quality controlled and trimmed by fastp (v0.23.0) and then mapped to the mouse reference sequence GRCm39 (mm39) using the Burrows-Wheeler Aligner (BWA)-MEM algorithm (v0.7.17). Duplicate reads were marked using Picard (v1.67) followed by realignment around known indels, and base quality recalibration was performed using GATK 3.7. Somatic mutation calls were performed using tumor (with or without treatment) and their derived cell line applying Mutect2, which combines the GATK’s local assembly and pair-HMM read-to-haplotype alignment (65). The resulting SNVs were annotated by multiple databases using Ensembl VEP (66) and filtered allowing at least 0.02 variant allele frequency, two variant allele counts, and coverage of ≥20 in tumor and up to a maximum of 0.01 allele frequency and coverage of ≥10 in normal samples and excluded from dbSNP146. Total somatic copy number alterations were detected using CNVkit (67). Copy number state segmentation was performed using “DNAcopy” R package that implements the circular binary segmentation and converted the overlapping segments at all gene levels using “CNTools” R package, except the mean segmentation for Stk11 gene, which was restricted to selected exons flanked by loxP sites.
Tumor purity was deduced based on the known clonal engineered biallelic deletion of STK11 in TCs in KL samples and the presence of clonal biallelic deletion of CDKN2A, CDKN2B, and MTAP in TCs in KP2 samples (confirmed by WES of the corresponding KL1, KL2, KL4, KL5, and KP2 cell lines, respectively).
Histochemistry and Immunohistochemistry
Tumors were harvested at the predefined experimental endpoints, fixed in 10% neutral buffered formalin (Sigma, #HT501128), transferred in 70% ethanol, embedded in paraffin, and sectioned into 5-pm sections. IHC assays were conducted using a BOND-MAX automated IHC staining system (Leica Microsystems) with BOND Polymer Refine Detection Kit (Leica, #DS9800). Briefly sections were deparaffinized and rehydrated with standard Leica protocol followed by antigen retrieval for 20 minutes at 100°C using BOND Epitope Retrieval Solution 1 (citrate based, pH 6, #AR9961) or BOND Epitope Retrieval Solution 2 (EDTA based, pH 9.0, #AR9640). Slides underwent peroxide blocking for 5 minutes, followed by protein block with IHC/ISH Super Blocking (Leica, #PV6122) for 10 minutes and incubation with the indicated rabbit primary antibodies for 15 minutes. Secondary detection was carried out using Refine Detection Kit Polymer (Leica, #DS9800) utilizing DAB as a chromogen (10 minutes), followed by counterstaining with hematoxylin for 8 minutes. Slides were dehydrated in a graded ethanol series and mounted using Clearium Mounting Media (Leica, #3801100). All wash cycles were performed with BOND Wash Solution (Leica, #AR9590) under standard Leica protocol. The complete list of antibodies, antigen retrieval protocols, and concentrations can be found in Supplementary Table S3. Hematoxylin and eosin (H&E), Masson’s trichrome, and Alcian Blue staining were performed by the MD Anderson Histology Research Core following standard protocols. Digitalization of slides was performed using Aperio AT2 system (Leica Biosystems) at 20× magnification. Digital image analysis was carried out using Halo software (Indica Lab) version 3.5, and regions of interest (ROIs) were determined by two independent pathologists. Proliferation quantification was performed using Ki67 in 10 500 × 500 μm ROI in representative tumor areas per tissue section. The identification of positive cells was accomplished using the Cytonuclear v2.0.9 module of the Halo software to detect nuclear positive expression in TCs, and the results were reported as the average percentage of TCs with positive Ki67 nuclear expression. Microvessel density was assessed using CD31 in the entire tumor tissue section, employing the Object colocalization v2.1.5 module of the Halo software to detect CD31-stained vessels. Vascular density was defined as the percentage of stained CD31/PECAM area compared with the total tumor area, following previously described literature (68). Necrotic areas, tissue artifacts, and non-tumoral tissue were excluded from all analyses.
Immunoblotting
Cultured cells or snap-frozen tumor tissue (40 mg) were lysed with RIPA buffer (Thermo Fisher, #PI89900) supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher, #78442). Tumor tissue samples underwent homogenization with a PowerGen 125 Homogenizer (Thermo Fisher). Lysates were centrifuged, and protein concentration was determined using Pierce BCA Protein Assay Kits (Thermo Fisher, #23225). Samples were loaded in 4% to 15% precast polyacrylamide gels (Bio-Rad #5671085), and proteins were transferred to PVDF membranes (Bio-Rad, #17042 72) using a Trans-Blot Turbo Transfer System (Bio-Rad, #1704270). Membranes were briefly washed in Tris-buffered saline (Bio-Rad, #1706435) supplemented with 0.1% Tween 20 (Sigma, #P7949; TBST) and then blocked with 5% blotting grade nonfat dry milk (Bio-Rad, # 1706404) in TBST for 1 hour at room temperature. Membranes were incubated with primary antibodies in TBST plus 1% BSA (Sigma, #A2153) overnight at 4°C, washed, and incubated for 1 hour with HRP-conjugated secondary antibodies (Bio-Rad) in 5% milk at room temperature. Protein visualization was performed using an enhanced chemiluminescence substrate (Thermo Fisher, #34577, #37071, #34095 or #A38556), and images were captured by a ChemiDoc Bio-Rad Imager and analyzed with Image Lab software. Primary antibodies used were LKB1 (CST, #3047), p53 (CST, #2524), vinculin (Sigma, #V9131), KEAP1 (CST, #7705), NRF2 (Proteintech, #163 9 6–1-AP), NQO1 (Abcam, #ab80588), BRG1 (Abcam, # ab110641), cleaved caspase3 (CST, #9664), cleaved PARP (CST, #94885), phospho-RIP3 (T231/S232; CST, #91702), phospho-MLKL (S345; CST, #37333), cleaved caspase1 (CST, #89332), cleaved gasdermin D (CST, #10137), phospho-ERK1/2 (T202/Y204; CST, #4370), phospho-S6 (S235/S236; CST, #4858), MUC1 (Thermo Fisher, #MA5–32265), GAPDH (CST, # 5174), NRAS (Santa Cruz, #sc31), HRAS (Abcam, #ab32417), E-cadherin (CST, #3195), KRAS (CST, #71835), and vimentin (Abcam, #ab92547). All primary antibodies were diluted at a ratio of 1:1,000, and HRP-conjugated secondary antibodies were diluted at a ratio of 1:3,000.
siRNA-Mediated Knockdown of Hras and Nras
KL2 cells were transfected with non-targeting control siRNA (siGENOME Non-Targeting siRNA Pool, Horizon, #D001206–13-05) or siRNAs targeting mouse Nras (ON-TARGETplus Mouse Nras siRNA SMARTpool, Horizon, #L044216–01-0005) and Hras siRNA (ON-TARGETplus Mouse Hras siRNA SMARTpool, Horizon, #L046324–00-0005) diluted in Opti-MEM medium containing DharmaFECT 1 Transfection Reagent (Horizon, #T2001–03) for 6 hours prior to treatment with 100 nmol/L of sotorasib for 24 or 72 hours. Lysates were collected and processed for Western blot analysis as previously described to assess NRAS, HRAS, KRAS, and phospho-ERK1/2 (T202/Y204) ERK1/2 protein expression.
Flow Cytometry
Harvested tumors were immersed in cold RPMI1640 and promptly processed into 2 to 4 mm3 pieces. Tumor pieces were transferred to a gentleMACS C Tube (Miltenyi, #130–096-334) containing a solution of Liberase (25 μg/mL), DNase 1 (30 U/mL), and hyaluronidase (0.01%) in serum-free RPMI1640 medium, followed by incubation for 45 minutes at 37°C and dissociation with a gentleMACS Octo Dissociator (Miltenyi). The enzymatic reactions were halted by adding cold RPMI1640 (10% FBS), and the suspensions were strained through a 70-μm cell strainer twice. Following red blood cell lysis using RBC lysis buffer (BioLegend, #420301), single-cell suspensions (approximately 1 × 106 cells in a total volume of 50 μL) were incubated with FcR-blocking reagent rat antimouse CD16/32 (2.4G2, BD Biosciences, #553142) for 15 minutes on ice. For extracellular staining, cells were treated with a mixture of conjugated antibodies in FACS buffer (PBS + 5% FBS + 2 mmol/L EDTA), including ghost dye violet 510 (Thermo Fisher, #13–0870-T100) for 1 hour at room temperature in the dark. For intracellular cytokine and transcription factor staining, single-cell suspensions underwent fixation and permeabilization using the eBioscience FoxP3/Transcription Factor Staining kit (Life Technologies, #00–5523-00), following the manufacturer’s instructions. Fluorochrome-conjugated monoclonal antibodies acquired from BioLegend were Pacific blue-anti-CD45 (30-F11), PE/Dazzle 594 anti-CD3 (17A2), APC/Cy7 anti-CD4 (RM45), PE/Cy7-anti-CD8 (53–6.7), APC-anti-T-bet (4B10), PE-anti-ICOS (7E.17G9), BV711-anti-CD44 (IM7), and BV605-anti-PD1 (29F.1A12), those acquired from BD Biosciences were BUV3 95-anti-CD25 (PC61) and BV605-anti-PD1 (RMP130), and those acquired from Life Technologies and Tonbo Bioscience were PerCP-Cy5.5-anti-FoxP3 (FJK16s) and FITC-anti-CD62L (MEL14), respectively. Flow cytometry data was acquired on LSR Fortessa flow cytometer (BD Biosciences) and analyzed using FlowJo 10.8.1 software (BD Biosciences).
For assessment ofMUC1 expression in human NSCLC cells, following treatment with RAS inhibitors for the indicated duration, TCs were collected by trypsinization, resuspended in FACS buffer (5% FBS, 2 mmol/L EDTA in PBS), fixed in 1% paraformaldehyde, and permeabilized. Cell suspensions were stained with MUC1 recombinant rabbit monoclonal antibody (Thermo Fisher Scientific, Cat# MA5–32265) at a dilution ratio of 1:100, followed by staining with goat anti-rabbit IgG (H + L) cross-adsorbed Alexa Fluor 568-conjugated secondary antibody (Life Technologies, Cat# A11004). The intracellular phospho-ERK1/2 (T202/Y204) staining protocol was adapted from Krutzik and colleagues (69). Briefly, TCs suspended in complete growth medium were treated with RMC-7977 (10 nmol/L) or trametinib (100 nmol/L) for 4 hours or stimulated with EGF (20 ng/mL) or PMA (10 ng/mL) for 15 minutes, followed by fixation with 1.5% formaldehyde and permeabilization in ice-cold methanol. The cells were subsequently incubated with mouse monoclonal MUC1 (Santa Cruz Biotechnology Cat# sc7313, 1:100 dilution) and rabbit monoclonal phospho-ERK1/2 (T202/Y204; Cell Signaling Technology Cat# 43 70, 1:900 dilution) primary antibodies, followed by incubation with goat antimouse IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor 488 (Life Technologies, Cat# A11001), and goat anti-rabbit IgG (H+L) cross-absorbed secondary antibody, Cyanine5 (Thermo Fisher Scientific Cat# A10523). Flow cytometry data was acquired in FACS Canto II (BD Biosciences) and analyzed using FlowJo 10.8.1 software (BD Bioscience).
Library Preparation and Single-Cell RNA Sequencing
Single-cell RNA sequencing (scRNA-seq) of all samples was conducted on the 10× Chromium system (10× Genomics), in collaboration with the MD Anderson Single Cell Genomics Center & Core facility. Briefly, cells were freshly isolated from mouse tumors, collected precisely 12 hours after the last dose of treatment with the corresponding RAS inhibitor at the defined experimental endpoints. Whole tumors (with the exception of tumors from mice that were rechallenged with RMC-7977 in which part of the tumor was reserved for IHC studies) were processed into small pieces and dissociated using a gentleMACS Octo Dissociator in a solution with Liberase (25 μg/mL), DNase 1 (30 U/mL), and hyaluronidase (0.01%) in serum-free RPMI1640 medium. Suspensions were strained twice through a 70-μm cell strainer. Dead cells and debris were removed using a Dead Cell Removal Kit (Miltenyi Biotec Inc., #130090101) according to the manufacturer’s protocol. Cells were resuspended in PBS with 0.04% BSA and underwent viability assessment using Trypan blue stain (0.4%) and the Countess II FL cell counter. Isolated cells were submitted to single-cell capture, barcoding, and library preparation, following the 10× Genomics Single-Cell Chromium 5′V2 protocols (CG000331, Rev E). Libraries were pooled to achieve a final concentration of 10 nmol/L. Pooled samples underwent qPCR for final concentration verification before submission for sequencing with the NovaSeq 6000 sequencer, utilizing the S4 200 cycles flow cell and sequencing with 26 cycles for read1, 10 cycles for i7 index, 10 cycles for i5, and 90 cycles for read 2 through the ATGC core at MD Anderson Cancer Center.
scRNA-seq Data Analysis
Count matrices were obtained aligning sequencing reads by running count from Cell Ranger (70) with default settings (version 7.1.0) using the mouse reference transcriptome, GRCm38 (mm10–3.0.0). For each sample, potential ambient RNA was removed using remove-background from CellBender (71). The resulting ambient corrected count matrices were analyzed with Seurat (version 5.0.0; ref. 72). Quality control was performed at the level of individual samples with tailored thresholds (selected based on the shape distribution of the quality control metrics for each sample) on the percentage of mitochondrial reads, percentage of ribosomal reads, number of features detected, and per cell library size. Neutrophils had overall lower library sizes compared with those of the rest of the cells, and therefore quality control filters were adjusted accordingly. We followed the Seurat procedure for normalization, dimensionality reduction, and clustering, choosing the number of principal components according to the talus plot method (73). Cells were normalized with Seurat using SCTransform with regression of the cell cycle scores using the gene list from Kowalczyk MS and colleagues (74) and the percentage of mitochondrial and ribosomal reads. Optimal clustering resolution was chosen with the help of the visualization tools included in the R package clustree (version 0.5.0; ref. 75). Clusters were assigned cell types using the expression of known marker genes from the literature, and cluster identities were validated and compared with the annotations from ImmGen (76) with SingleR (version 1.8.1; ref. 77). Keratinocytes and melanocytes were excluded from the analysis (based on the concerted expression of Krt5, Lgals7, Dsc3, and Lhx2 for keratinocytes and Tyr, Pmel, and Pax3 for melanocytes). Cells expressing high amounts of hemoglobin transcripts were also discarded. This clustering procedure was applied to each sample independently, thus enabling the potential removal of further artifacts. For the single-cell atlas, integration of cells across all samples in order to account for batch effects was performed using the fastMNN algorithm (78). Subclustering was conducted in the same manner for each major cell type independently (TCs, lymphoid cells, myeloid cells, vasculature, and fibroblasts) to determine the fine structure of these cell populations. Subclustering also enabled further removal of doublets and low-quality cells through inspection of the expression of population marker genes and quality control metrics. Plots were produced using a combination of native plotting capabilities R (v.4.1.3; November 3, 2023), as well as ComplexHeatmap (79) and scCustomize.
Pseudobulk counts were analyzed using the edgeR package (v.3.32.1; ref. 80). The count matrix was filtered to remove lowly expressed genes by keeping those with at least 15 counts, considering library size differences, in more than the minimum number of samples of the experimental design. The resulting filtered count matrix was then log2-transformed and trimmed mean of M-values normalized with the edgeR package (81). Heatmaps of the resulting data matrix were annotated by PROGENy (34) and were used to determine pathways of interest. Differential expression analyses were conducted using limma (v.3.46.0; ref. 82), and their respective volcano plots were produced using the EnhancedVolcano (https://github.com/kevinblighe/EnhancedVolcano; v.1.8.0) package. Pathways from the mouse hallmark gene set collection from the Molecular Signatures Database (MSigDB; ref. 83) were scored on the TCs by using the Add-ModuleScore function from Seurat.
Treatment of BMDMs with RAS Inhibitors In Vitro
BMDMs were isolated from the femurs of BALB/c mice and plated in complete RPMI media (Gibco) containing 10% heat-inactivated FBS (VWR), 1% Pen/Strep (Corning), and 1% HEPES (Corning) and supplemented with CSF1 (PeproTech) at 10 ng/mL. Cells were cultured for 6 days and then harvested using dissociation buffer (Gibco) and plated into 96-well plates at 15,000 cells/well. Macrophages were polarized (M1: IFNγ: 20 ng/mL; LPS: 100 ng/mL; M2: IL4: 20 ng/mL), and cells were cultured for an additional 24 hours. Polarized macrophages were treated with either RMC-7977, RMC-4998, BLZ945 (CSF1R inhibitor, Selleckchem), or staurosporine at different concentrations. Cell viability was measured 72 hours after compound treatment using the CellTiter-Glo assay (Promega). Caspase activity was measured using the Caspase-Glo 3/7 kit (Promega) 4 to 24 hours after treatment.
Immunofluorescence Analysis
Human KRAS-mutant NSCLC cell lines pretreated with RMC-7977 or DMSO were cultured in collagen-pre-coated glass coverslips overnight under continuous drug exposure. The following day, cells were washed with PBS and fixed with 4% paraformaldehyde for 15 minutes at room temperature and then permeabilized with ice-cold methanol for 10 minutes at −20°C and blocked in 2% BSA for 1 hour. Cells were incubated with MUC1 recombinant rabbit monoclonal antibody (Thermo Fisher Scientific, Cat# MA5–32265; 1:200, in 1% BSA PBS) overnight at 4°C. Cells were washed three times with PBS and incubated with secondary antibody Alexa Fluor 488 Donkey anti-rabbit IgG (Invitrogen, catalog# A21206, 1:500 dilution) for 2 hours at room temperature. Secondary antibody was washed out, and coverslips were counterstained by DAPI and mount on slides with ProLong Gold Antifade Reagent. Slides were counterstained with DAPI and mounted with ProLong Gold Antifade Reagent. A total of at least 10 randomly selected viewing fields per condition were captured using a Nikon confocal microscope with 60× oil objective, and images were processed using Nikon AR software.
TCGA Data Analysis
TCGA Pan Cancer LUAD cohort clinical outcomes and gene expression datasets were retrieved from cBioPortal (https://www.cbioportal.org/), accessed on February 10, 2024. Patients bearing KRAS WT tumors and/or with unknown M1 disease status were excluded, resulting in a total of 159 entries. mRNA expression and RS EM (batch normalized from Illumina HiSeq_RNASeqV2) data for key genes of interest were exported. Hallmark gene set (Human MSigDB) enrichment analysis was performed based on pre-ranked genes according to their Spearman’s rank correlation coefficient with TFF1 or MUC1 mRNA expression (batch normalized from Illumina HiSeq_RNASeqV2) utilizing GSEA software, v 4.3.2.
Graphical Illustrations
Graphical elements used to illustrate experimental design schemes were created with biorender.com, licensed by The University of Texas MD Anderson Cancer Center.
Statistical Analysis (In Vitro Data and Animal Studies)
Statistical analyses are detailed in each figure legend. The Kaplan-Meier method was used to estimate TTD, relapse-free survival, and PFS curves, and differences were assessed by the log-rank test. For quantitative measurements, a normal distribution was assumed, and two-way ANOVA followed by Bonferroni’s multiple comparison test was employed in experiments involving multiple groups, unless explicitly stated otherwise. A significance threshold at two-tailed P value < 0.05 was considered significant. All statistical analyses (except scRNA-seq and MD Anderson clinical cohort analyses) were executed using GraphPad Prism Software v 10.0.3.
MD Anderson Clinical Cohort of Patients with NSCLC Treated with Sotorasib or Adagrasib
The study was IRB approved at MD Anderson Cancer Center and included a waiver of patient informed consent and was conducted in accordance with ethical guidelines including the Declaration of Helsinki and US Common Rule. Patient information was collected through chart review of electronic medical records from patients with stage IV KRASG12C-mutant NSCLC who received treatment with sotorasib or adagrasib at MD Anderson Cancer between November 2018 and January 2024. The dataset was locked for survival analysis on February 20, 2024. Patients who received sotorasib or adagrasib as a single agent (without prior treatment with RAS-targeted therapies), were alive for ≥14 days after start of treatment, had ECOG performance status ≤2, and had available pathology reports and molecular profiling results prior to starting KRASG12C inhibitor treatment were considered eligible. The presence or absence of mucinous features in pretreatment tissue biopsies was determined based on formal pathology reports issued by a licensed clinical pathologist. CDX2 immune reactivity status was curated based on the pathology report for the subset of evaluable tumors. BOR was determined through investigator-assessed RECIST v1.1. Response rates were compared using Fisher’s exact test. The Kaplan-Meier method was used to estimate PFS and OS, and differences were assessed by the log-rank test. HRs and corresponding CIs were estimated with the use of a stratified Cox proportional hazards model adjusting for clinical variables: age, history of brain metastasis, prior lines of therapy for metastatic disease (≤1 vs. >1), PS (0–1 vs. 2), and histologic subtype (adenocarcinoma vs. other histologic groups). Further adjustment based on the presence of absence of key co-alterations (KEAP1, SMARCA4, CDKN2A) was also performed based on available next-generation sequencing-based comprehensive genomic profiling reports. Statistical significance was established at P ≤ 0.05. Statistical analysis was performed using IBM SPSS Statistics v 26.0.0.0.
Supplementary Material
Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).
SIGNIFICANCE:
Our work reveals robust and durable antitumor activity of the preclinical RAS(ON) multiselective inhibitor RMC-7977 against difficult-to-treat subsets of KRASG12C-mutant NSCLC with primary or acquired RASG12C inhibitor resistance and identifies a conserved mucinous transcriptional state that supports RAS inhibitor tolerance.
Acknowledgments
Work in F. Skoulidis’ laboratory was supported by the NIH/NCI 1R01 CA262469-01, 1R01 CA279194-01, and 1R01CA279452-01A1; Lung Cancer SPORE P50-CA70907-21; the Tammi Hissom Grant from Rexanna’s Foundation for Fighting Lung Cancer; and a sponsored research agreement with Revolution Medicines. M.V. Negrao acknowledges research funding from Rexanna’s Foundation for Fighting Lung Cancer. This work was supported by the generous philanthropic contributions to the University of Texas MD Anderson Cancer Center, MD Anderson Cancer Center Support Grant P30 CA016672, and CPRIT Single Core grant RP180684. The Advanced Technology Genomics Core, the Research Histology Core Laboratory, and the Flow Cytometry and Cellular Imaging Core Facility were supported in part by the University of Texas MD Anderson Cancer Center and P30CA016672. The authors would also like to acknowledge the digital Immune-Profiling Pathology Laboratory from the Translational Molecular Pathology-Immunoprofiling lab (TMP-IL) MoonShots Platform at the Department Translational Molecular Pathology, the University of Texas MD Anderson Cancer Center, for supporting shared resources. We further acknowledge the MD Anderson GEMINI Team, Jianling Zhou, and Mei Jiang for their technical contributions to this article and Matthew Holderfield from Revolution Medicines for valuable insights during review of this manuscript.
Authors’ Disclosures
H.A. Araujo reports a patent for technology related to this research pending to the Board of Regents, the University of Texas System. X. Pechuan-Jorge reports a patent for technology related to this research pending to Revolution Medicines. M.V. Negrao reports other support from Lilly, Mirati, Novartis, Checkmate, Alaunos, AstraZeneca, Pfizer, Genentech, and Navire during the conduct of the study; other support from Genentech, Mirati, Merck/MDS, Novartis, Sanofi, Pfizer, Lilly, and AstraZeneca outside the submitted work; speaker’s bureau from OncLive, Ideology, BIO Brasil, and Medscape; travel expenses from Ideology, DAVA Oncology; and writing support from ApotheCom and Ashfield Healthcare. D.L. Gibbons reports grants from the NIH during the conduct of the study; grants from Mirati, Boehringer Ingelheim, and NGM Biopharmaceuticals; and personal fees from Menarini Recherche, Eli Lilly, and Onconova outside the submitted work. D.S. Hong reports grants from Revolution Medicines during the conduct of the study; grants from Revolution Medicines outside the submitted work; research (Inst)/grant funding (Inst) from Abb Vie, Adaptimmune, Adlai Nortye, Amgen, Astellas, AstraZeneca, Bayer, Biomea, Bristol Myers Squibb, Daiichi-Sankyo, Deciphera, Eisai, Eli Lilly, Endeavor, Erasca, F. Hoffmann-LaRoche, Fate Therapeutics, Genentech, Genmab, ImmunoGenesis, Infinity, Kyowa Kirin, Merck, Mirati, Navier, NCI-CTEP, Novartis, Numab, Pfizer, Pyramid Bio, Quanta, Revolution Medicines, Seagen, STCube, Takeda, TCR2, Turning Point Therapeutics, and VM Oncology; travel, accommodations, and expenses from AACR, ASCO, CLCC, Bayer, Genmab, Northwestern, SITC, Telperian, and UNC; consultancy, speakership, or advisory role from 28Bio, AbbVie, Acuta, Adaptimmune, Alkermes, Alpha Insights, Amgen, Affini-T, Astellas, AUM Biosciences, Axiom, Baxter, Bayer, Boxer Capital, BridgeBio, CARSgen, CLCC, COG, COR2ed, Cowen, Ecorl, EDDC, Erasca, Exelixis, Fate Therapeutics, F. Hoffmann-La Roche, Genentech, Gennao Bio, Gilead, GLG, Group H, Guidepoint, HCW Precision Oncology, ImmunoGenesis, Incyte Inc., Inhibrx Inc., InduPro, Innovent, Janssen, Jounce Therapeutics Inc., Lan-Bio, Liberium, MEDACorp, Medscape, Novartis, Northwestern, Numab, Oncologia Brasil, ORI Capital, Pfizer, Pharma Intelligence, POET Congress, Prime Oncology, Projects in Knowledge, Quanta, RAIN, Ridgeline, Revolution Medicines, Seagen, Stanford, STCube, Takeda, Tavistock, Trieza Therapeutics, T-Knife, Turning Point Therapeutics, UNC, WebMD, YingLing Pharma, and Ziopharm; and other ownership interests from CrossBridge Bio (Advisor), Molecular Match (Advisor), OncoResponse (Founder, Advisor), and Telperian (Founder, Advisor). J.A. Roth reports grants from the NCI during the conduct of the study and grants, personal fees, and other support from Genprex outside the submitted work. J.V. Heymach reports other support from AbbVie, AnHeart Therapeutics, ArriVent Biopharma, AstraZeneca, BioCurity Pharmaceuticals, BioNTech, Blueprint Medicine, Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceuticals, Curio Science, DAVA Oncology, Eli Lilly, EMD Serono, Ideology Health, Immunocore, InterVenn Biosciences, Janssen, Mirati, Moffitt Cancer Center, ModeX, Nexus Health Systems, Novartis, Oncocyte, RefleXion, Regeneron, R.oche, Sandoz, Sanofi, Spectrum, and other support from Takeda outside the submitted work. J. Zhang reports other support from Merck; grants from Johnson and Johnson; other support from Novartis, Helius, and Summit; and personal fees from AstraZeneca, Takeda, Catalyst, Henrui, Varian, GenePlus, and personal fees from OncoHost outside the submitted work. F. Skoulidis reports grants from Revolution Medicines, NIH/NCI, and Rexanna’s Foundation for Fighting Lung Cancer during the conduct of the study; personal fees from Revolution Medicines, Amgen, Genentech/Roche, AstraZeneca, Novartis, BeiGene, BridgeBio, Guardant Health, Navire Pharma, Tango Therapeutics, Calithera Biosciences, Merck Sharp & Dohme, Novocure, Hookipa Pharma, ESMO, AACR, IASLC, Japanese Lung Cancer Society, Intellisphere LLC, Medscape LLC, PER LLC, Curio LLC, MI&T, IDEOLogy Health, MJH Life Sciences, VSPO McGill Universite de Montreal, RV Mais Promocao Eventos LTDS, DAVA Oncology, and BMS; other support from BioNTech SE and Moderna Inc.; grants from Amgen, Mirati Therapeutics/BMS, Pfizer, Novartis, Merck & Co., and Boehringer Ingelheim; and personal fees from Aimmune and HMP Global outside the submitted work. In addition, F. Skoulidis has a patent for technology related to this research pending to the Board of Regents, the University of Texas System. No disclosures were reported by the other authors.
Data Availability
The data generated in this study are available within the article and its supplementary data files. scRNA-seq raw and processed data reported in this article are available upon request and were also deposited at the NCBI Gene Expression Omnibus in accordance with AACR data availability policies under accession number GSE270541. The individual patient data generated in this study are governed by the University of Texas MD Anderson Cancer Center. To preserve patient confidentiality, to protect patient-related information, and to remain compliant with institutional regulatory requirements, aggregate and/or summary deidentified data may be made available upon reasonable academic request to the corresponding author. All other raw data are available from the corresponding author upon reasonable request.
REFERENCES
- 1.Skoulidis F, Byers LA, Diao L, Papadimitrakopoulou VA, Tong P, Izzo J, et al. Co-occurring genomic alterations define major subsets of KRAS-mutant lung adenocarcinoma with distinct biology, immune profiles, and therapeutic vulnerabilities. Cancer Discov 2015;5:860–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ostrem JM, Peters U, Sos ML, Wells JA, Shokat KM. K-Ras (G12C) inhibitors allosterically control GTP affinity and effector interactions. Nature 2013;503:548–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hong DS, Fakih MG, Strickler JH, Desai J, Durm GA, Shapiro GI, et al. KRASG12C inhibition with sotorasib in advanced solid tumors. N Engl J Med 2020;383:1207–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Skoulidis F, Li BT, Dy GK, Price TJ, Falchook GS, Wolf J, et al. Sotorasib for lung cancers with KRAS p.G12C mutation. N Engl J Med 2021;384:2371–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jänne PA, Riely GJ, Gadgeel SM, Heist RS, Ou S- HI, Pacheco JM, et al. Adagrasib in non-small-cell lung cancer harboring a KRASG12C mutation. N Engl J Med 2022;387:120–31. [DOI] [PubMed] [Google Scholar]
- 6.de Langen AJ, Johnson ML, Mazieres J, Dingemans A- MC, Mountzios G, Pless M, et al. Sotorasib versus docetaxel for previously treated non-small-cell lung cancer with KRASG12C mutation: a randomised, open-label, phase 3 trial. Lancet 2023;401:733–46. [DOI] [PubMed] [Google Scholar]
- 7.Negrao MV, Araujo HA, Lamberti G, Cooper AJ, Akhave NS, Zhou T, et al. Comutations and KRASG12C inhibitor efficacy in advanced NSCLC. Cancer Discov 2023;13:1556–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ryan MB, Coker O, Sorokin A, Fella K, Barnes H, Wong E, et al. KRASG12C-independent feedback activation of wild-type RAS constrains KRASG12C inhibitor efficacy. Cell Rep 2022;39:110993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ryan MB, Fece de la Cruz F, Phat S, Myers DT, Wong E, Shahzade HA, et al. Vertical pathway inhibition overcomes adaptive feedback resistance to KRASG12C inhibition. Clin Cancer Res 2020;26:1633–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Awad MM, Liu S, Rybkin II, Arbour KC, Dilly J, Zhu VW, et al. Acquired resistance to KRASG12C inhibition in cancer. N Engl J Med 2021;384:2382–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Zhao Y, Murciano-Goroff YR, Xue JY, Ang A, Lucas J, Mai TT, et al. Diverse alterations associated with resistance to KRASG12C inhibition. Nature 2021;599:679–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Tong X, Patel AS, Kim E, Li H, Chen Y, Li S, et al. Adeno-to-squamous transition drives resistance to KRAS inhibition in LKB1 mutant lung cancer. Cancer Cell 2024;42:413–28.e7. [DOI] [PubMed] [Google Scholar]
- 13.Sacher A, LoRusso P, Patel MR, Miller WH Jr, Garralda E, Forster MD, et al. Single-agent divarasib (GDC-6036) in solid tumors with a KRAS G12C mutation. N Engl J Med 2023;389:710–21. [DOI] [PubMed] [Google Scholar]
- 14.Punekar SR, Velcheti V, Neel BG, Wong KK. The current state of the art and future trends in RAS-targeted cancer therapies. Nat Rev Clin Oncol 2022;19:637–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Drilon A, Sharma MR, Johnson ML, Yap TA, Gadgeel S, Nepert D, et al. SHP2 inhibition sensitizes diverse oncogene-addicted solid tumors to Re-treatment with targeted therapy. Cancer Discov 2023;13:1789–801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Holderfield M, Lee BJ, Jiang J, Tomlinson A, Seamon KJ, Mira A, et al. Concurrent inhibition of oncogenic and wild-type RAS-GTP for cancer therapy. Nature 2024;629:919–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wasko UN, Jiang J, Dalton TC, Curiel-Garcia A, Edwards AC, Wang Y, et al. Tumour-selective activity of RAS-GTP inhibition in pancreatic cancer. Nature 2024. 629:927–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Jiang J, Jiang L, Maldonato BJ, Wang Y, Holder field M, Aronchik I, et al. Translational and therapeutic evaluation of RAS-GTP inhibition by RMC-6236 in RAS-driven cancers. Cancer Discov 2024;14: 994–1017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Arbour KC, Punekar S, Garrido-Laguna I, Hong DS, Wolpin B, Pelster MS, et al. 652O Preliminary clinical activity of RMC-6236, a first-in-class, RAS-selective, tri-complex RAS-MULTI(ON) inhibitor in patients with KRAS mutant pancreatic ductal adenocarcinoma (PDAC) and non-small cell lung cancer (NSCLC). Ann Oncol 2023;34:S458. [Google Scholar]
- 20.Jänne PA, Bigot F, Papadopoulos K, Eberst L, Sommerhalder D, Lebellec L, et al. Abstract PR014: preliminary safety and anti-tumor activity of RMC-6291, a first-in-class, tri-complex KRASG12C(ON) inhibitor, in patients with or without prior KRASG12C(OFF) inhibitor treatment. Mol Cancer Ther 2023;22:PR014. [Google Scholar]
- 21.Schilsky JB, Ni A, Ahn L, Datta S, Travis WD, Kris MG, et al. Prognostic impact of TTF-1 expression in patients with stage IV lung adenocarcinomas. Lung Cancer 2017;108:205–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Winslow MM, Dayton TL, Verhaak RGW, Kim-Kiselak C, Snyder EL, Feldser DM, et al. Suppression of lung adenocarcinoma progression by Nkx2–1. Nature 2011;473:101–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Snyder EL, Watanabe H, Magendantz M, Hoersch S, Chen TA, Wang DG, et al. Nkx2–1 represses a latent gastric differentiation program in lung adenocarcinoma. Mol Cell 2013;50:185–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dy GK, Govindan R, Velcheti V, Falchook GS, Italiano A, WolfJ, et al. Long-term outcomes and molecular correlates of sotorasib efficacy in patients with pretreated KRASG12C-mutated non-small-cell lung cancer: 2-year analysis of CodeBreaK 100. J Clin Oncol 2023;41:3311–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Koyama S, Akbay EA, Li YY, Aref AR, Skoulidis F, Herter-Sprie GS, et al. STK11/LKB1 deficiency promotes neutrophil recruitment and proinflammatory cytokine production to suppress T-cell activity in the lung tumor microenvironment. Cancer Res 2016;76:999–1008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Skoulidis F, Goldberg ME, Greenawalt DM, Hellmann MD, Awad MM, Gainor JF, et al. STK11/LKB1 mutations and PD-1 inhibitor resistance in KRAS-mutant lung adenocarcinoma. Cancer Discov 2018;8: 822–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Canon J, Rex K, Saiki AY, Mohr C, Cooke K, Bagal D, et al. The clinical KRAS(G12C) inhibitor AMG 510 drives anti-tumour immunity. Nature 2019;575:217–23. [DOI] [PubMed] [Google Scholar]
- 28.Briere DM, Li S, Calinisan A, Sudhakar N, Aranda R, Hargis L, et al. The KRASG12C inhibitor MRTX849 reconditions the tumor immune microenvironment and sensitizes tumors to checkpoint inhibitor therapy. Mol Cancer Ther 2021;20:975–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Mugarza E, van Maldegem F, Boumelha J, Moore C, Rana S, Llorian Sopena M, et al. Therapeutic KRASG12C inhibition drives effective interferon-mediated antitumor immunity in immunogenic lung cancers. Sci Adv 2022;8:eabm8780. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Hamarsheh S, Groß O, Brummer T, Zeiser R. Immune modulatory effects of oncogenic KRAS in cancer. Nat Commun 2020;11:5439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Cabanos HF, Hata AN. Emerging insights into targeted therapy-tolerant persister cells in cancer. Cancers (Basel) 2021;13:2666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Zemmour D, Goldrath A, Kronenberg M, Kang J, Benoist C. The ImmGen consortium OpenSource T cell project. Nat Immunol 2022;23:643–4. [DOI] [PubMed] [Google Scholar]
- 33.Zilionis R, Engblom C, Pfirschke C, Savova V, Zemmour D, Saatcioglu HD, et al. Single-cell transcriptomics of human and mouse lung cancers reveals conserved myeloid populations across individuals and species. Immunity 2019;50:1317–34.e10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Schubert M, Klinger B, Klünemann M, Sieber A, Uhlitz F, Sauer S, et al. Perturbation-response genes reveal signaling footprints in cancer gene expression. Nat Commun 2018;9:20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Lee HJ, Zhuang G, Cao Y, Du P, Kim H-J, Settleman J. Drug resistance via feedback activation of Stat3 in oncogene-addicted cancer cells. Cancer Cell 2014;26:207–21. [DOI] [PubMed] [Google Scholar]
- 36.Nagathihalli NS, Castellanos JA, Lamichhane P, Messaggio F, Shi C, Dai X, et al. Inverse correlation of STAT3 and MEK signaling mediates resistance to RAS pathway inhibition in pancreatic cancer. Cancer Res 2018;78:6235–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Hou P, Kapoor A, Zhang Q, Li J, Wu CJ, Li J, et al. Tumor microenvironment remodeling enables bypass of oncogenic KRAS dependency in pancreatic cancer. Cancer Discov 2020;10:1058–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Lv DD, Zhou LY, Tang H. Hepatocyte nuclear factor 4a and cancer-related cell signaling pathways: a promising insight into cancer treatment. Exp Mol Med 2021;53:8–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Ma Q, Wu K, Li H, Li H, Zhu Y, Hu G, et al. ONECUT2 overexpression promotes RAS-driven lung adenocarcinoma progression. Sci Rep 2019;9:20021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Shi H, Tsang Y, Yang Y, Chin HL. Identification of ONECUT3 as a stemness-related transcription factor regulating NK cell-mediated immune evasion in pancreatic cancer. Sci Rep 2023;13:18133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Rotinen M, You S, Yang J, Coetzee SG, Reis-Sobreiro M, Huang WC, et al. ONECUT2 is a targetable master regulator of lethal prostate cancer that suppresses the androgen axis. Nat Med 2018;24: 1887–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Laughney AM, Hu J, Campbell NR, Bakhoum SF, Setty M, Lavallée V-P, et al. Regenerative lineages and immune-mediated pruning in lung cancer metastasis. Nat Med 2020;26:259–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Yang D, Jones MG, Naranjo S, Rideout WM III, Min KHJ, Ho R, et al. Lineage tracing reveals the phylodynamics, plasticity, and paths of tumor evolution. Cell 2022;185:1905–23.e25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Zewdu R, Mehrabad EM, Ingram K, Fang P, Gillis KL, Camolotto SA, et al. An NKX2–1/ERK/WNT feedback loop modulates gastric identity and response to targeted therapy in lung adenocarcinoma. Elife 2021;10:e66788. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Gong Y, Fan Z, Luo G, Yang C, Huang Q, Fan K, et al. The role of necroptosis in cancer biology and therapy. Mol Cancer 2019;18:100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Wei X, Xie F, Zhou X, Wu Y, Yan H, Liu T, et al. Role of pyroptosis in inflammation and cancer. Cell Mol Immunol 2022;19:971–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Erkes DA, Cai W, Sanchez IM, Purwin TJ, Rogers C, Field CO, et al. Mutant BRAF and MEK inhibitors regulate the tumor immune microenvironment via pyroptosis. Cancer Discov 2020;10: 254–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.El-Kenawi A, Berglund A, Estrella V, Zhang Y, Liu M, Putney RM, et al. Elevated methionine flux drives pyroptosis evasion in persister cancer cells. Cancer Res 2023;83:720–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Mahadevan KK, McAndrews KM, LeBleu VS, Yang S, Lyu H, Li B, et al. KRASG12D inhibition reprograms the microenvironment of early and advanced pancreatic cancer to promote FAS-mediated killing by CD8+ T cells. Cancer Cell 2023;41:1606–20.e8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Serban D, Leng J, Cheresh D. H-Ras regulates angiogenesis and vascular permeability by activation of distinct downstream effectors. Circ Res 2008;102:1350–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Matsuo Y, Campbell PM, Brekken RA, Sung B, Ouellette MM, Fleming JB, et al. K-Ras promotes angiogenesis mediated by immortalized human pancreatic epithelial cells through mitogen-activated protein kinase signaling pathways. Mol Cancer Res 2009;7: 799–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Zacharias WJ, Frank DB, Zepp JA, Morley MP, Alkhaleel FA, Kong J, et al. Regeneration of the lung alveolus by an evolutionarily conserved epithelial progenitor. Nature 2018;555:251–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Marjanovic ND, Hofree M, Chan JE, Canner D, Wu K, Trakala M, et al. Emergence of a high-plasticity cell state during lung cancer evolution. Cancer Cell 2020;38:229–46.e13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Moghal N, Li Q, Stewart EL, Navab R, Mikubo M, D’Arcangelo E, et al. Single-cell analysis reveals transcriptomic features of drug-tolerant persisters and stromal adaptation in a patient-derived EGFR-mutated lung adenocarcinoma xenograft model. J Thorac Oncol 2023;18:49 9–515. [DOI] [PubMed] [Google Scholar]
- 55.Maynard A, McCoach CE, Rotow JK, Harris L, Haderk F, Kerr DL, et al. Therapy-induced evolution of human lung cancer revealed by single-cell RNA sequencing. Cell 2020;182:1232–51.e22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Fakih MG, Kopetz S, Kuboki Y, Kim TW, Munster PN, Krauss JC, et al. Sotorasib for previously treated colorectal cancers with KRASG12C mutation (CodeBreaK100): a prespecified analysis of a single-arm, phase 2 trial. Lancet Oncol 2022;23:115–24. [DOI] [PubMed] [Google Scholar]
- 57.Strickler JH, Satake H, George TJ, Yaeger R, Hollebecque A, Garrido-Laguna I, et al. Sotorasib in KRAS p.G12C-mutated advanced pancreatic cancer. N Engl J Med 2023;388:33–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Tata PR, Chow RD, Saladi SV, Tata A, Konkimalla A, Bara A, et al. Developmental history provides a roadmap for the emergence of tumor plasticity. Dev Cell 2018;44:679–93.e5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Park S, Colville MJ, Paek JH, Shurer CR, Singh A, Secor EJ, et al. Immunoengineering can overcome the glycocalyx armour of cancer cells. Nat Mater 2024;23:429–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Hoffmann W. Trefoil factor family (TFF) peptides and their links to inflammation: a Re-evaluation and new medical perspectives. Int J Mol Sci 2021;22:4909. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Kufe DW. Mucins in cancer: function, prognosis and therapy. Nat Rev Cancer 2009;9:874–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Chen W, Zhang Z, Zhang S, Zhu P, Ko JK, Yung KK. MUC1: structure, function, and clinic application in epithelial cancers. Int J Mol Sci 2021;22:6567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Fang J, Lai S, Yu H, Ma L. Suppression of MUC1-overexpressing tumors by a novel MUC1/CD3 bispecific antibody. Antibodies (Basel) 2023;12:47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.DuPage M, Dooley AL, Jacks T. Conditional mouse lung cancer models using adenoviral or lentiviral delivery of Cre recombinase. Nat Protoc 2009;4:1064–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, Sougnez C, et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat Biotechnol 2013;31:213–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.McLaren W, Gil L, Hunt SE, Riat HS, Ritchie GR, Thormann A, et al. The Ensembl variant effect predictor. Genome Biol 2016;17:122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Talevich E, Shain AH, Botton T, Bastian BC. CNVkit: genome-wide copy number detection and visualization from targeted DNA sequencing. PLoS Comput Biol 2016;12:e1004873. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Noguera-Troise I, Daly C, Papadopoulos NJ, Coetzee S, Boland P, Gale NW, et al. Blockade of Dll4 inhibits tumour growth by promoting non-productive angiogenesis. Nature 2006;444:1032–7. [DOI] [PubMed] [Google Scholar]
- 69.Krutzik PO, Nolan GP. Intracellular phospho-protein staining techniques for flow cytometry: monitoring single cell signaling events. Cytometry A 2003;55:61–70. [DOI] [PubMed] [Google Scholar]
- 70.Zheng GXY, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, et al. Massively parallel digital transcriptional profiling of single cells. Nat Commun 2017;8:14049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71.Fleming SJ, Chaffin MD, Arduini A, Akkad AD, Banks E, Marioni JC, et al. Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender. Nat Methods 2023;20:1323–35. [DOI] [PubMed] [Google Scholar]
- 72.Hao Y, Stuart T, Kowalski MH, Choudhary S, Hoffman P, Hartman A, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 2024;42:293–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 73.Henningsson R, Moratorio G, Borderia AV, Vignuzzi M, Fontes M. DISSEQT-DIStribution-based modeling of SEQuence space Time dynamics. Virus Evol 2019;5:vez028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 74.Kowalczyk MS, Tirosh I, Heckl D, Rao TN, Dixit A, Haas BJ, et al. Single-cell RNA-seq reveals changes in cell cycle and differentiation programs upon aging of hematopoietic stem cells. Genome Res 2015;25:1860–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 75.Zappia L, Oshlack A. Clustering trees: a visualization for evaluating clusterings at multiple resolutions. Gigascience 2018;7:giy083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76.Heng TS, Painter MW; Immunological Genome Project Consortium. The Immunological Genome Project: networks of gene expression in immune cells. Nat Immunol 2008;9:1091–4. [DOI] [PubMed] [Google Scholar]
- 77.Aran D, Looney AP, Liu L, Wu E, Fong V, Hsu A, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol 2019;20:163–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78.Zhang F, Wu Y, Tian W. A novel approach to remove the batch effect of single-cell data. Cell Discov 2019;5:46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79.Gu Z, Eils R, Schlesner M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 2016;32:2847–9. [DOI] [PubMed] [Google Scholar]
- 80.Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 2010;26:139–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 81.Fontes M, Soneson C. The projection score-an evaluation criterion for variable subset selection in PCA visualization. BMC Bioinformatics 2011;12:307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82.Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83.Liberzon A, Birger C, Thorvaldsdottir H, Ghandi M, Mesirov JP, Tamayo P The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 2015;1:417–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data generated in this study are available within the article and its supplementary data files. scRNA-seq raw and processed data reported in this article are available upon request and were also deposited at the NCBI Gene Expression Omnibus in accordance with AACR data availability policies under accession number GSE270541. The individual patient data generated in this study are governed by the University of Texas MD Anderson Cancer Center. To preserve patient confidentiality, to protect patient-related information, and to remain compliant with institutional regulatory requirements, aggregate and/or summary deidentified data may be made available upon reasonable academic request to the corresponding author. All other raw data are available from the corresponding author upon reasonable request.