Diverse alterations associated with resistance to KRAS(G12C) inhibition

Abstract

Inactive state-selective KRAS(G12C) inhibitors^1-8 demonstrate a 30-40% response rate and result in approximately 6-month median progression-free survival in patients with lung cancer⁹. The genetic basis for resistance to these first-in-class mutant GTPase inhibitors remains under investigation. Here we evaluated matched pre-treatment and post-treatment specimens from 43 patients treated with the KRAS(G12C) inhibitor sotorasib. Multiple treatment-emergent alterations were observed across 27 patients, including alterations in KRAS, NRAS, BRAF, EGFR, FGFR2, MYC and other genes. In preclinical patient-derived xenograft and cell line models, resistance to KRAS(G12C) inhibition was associated with low allele frequency hotspot mutations in KRAS(G12V or G13D), NRAS(Q61K or G13R), MRAS(Q71R) and/or BRAF(G596R), mirroring observations in patients. Single-cell sequencing in an isogenic lineage identified secondary RAS and/or BRAF mutations in the same cells as KRAS(G12C), where they bypassed inhibition without affecting target inactivation. Genetic or pharmacological targeting of ERK signalling intermediates enhanced the antiproliferative effect of G12C inhibitor treatment in models with acquired RAS or BRAF mutations. Our study thus suggests a heterogenous pattern of resistance with multiple subclonal events emerging during G12C inhibitor treatment. A subset of patients in our cohort acquired oncogenic KRAS, NRAS or BRAF mutations, and resistance in this setting may be delayed by co-targeting of ERK signalling intermediates. These findings merit broader evaluation in prospective clinical trials.

Extended Data Figure 1.. Characteristics of lung cancer patients with an exceptional response to sotorasib treatment.

a, Percent of patients with baseline alterations in the indicated genes. No significant differences by Fisher exact test. b, KRAS(G12C) allele frequency in baseline plasma specimens. c, As in b but only patients with available archival/baseline tissue are plotted. d, Baseline tumor burden, as determined by the sum of the longest diameter in RECIST target lesions. Two tailed p values from Mann-Whitney tests are shown in b-d. The lines denote median values. In a, b and d: n=8 (CR/LTR) and n=28 (Other) patients, whereas in c: n=5 (CR/LTR) and n=8 (Other) patients.

Extended Data Figure 2.. Treatment-emergent alterations in patient-derived xenograft models.

a, PDX-bearing mice were treated with sotorasib (100 mpk, Lu1, Lu10), adagrasib (100 mpk, Lu3, Lu7, Re1, Co1) or MRTX1257 (50 mpk, Lu36, Lu69). The doted line represents a tumor size of ~500 mm³, which was used to determine latency (n=4, mean ± s.e.m). b, Characteristics of patient derived xenograft models. c-d, VAFs or estimated copy numbers for the indicated variants (c) or genes (d). A large number of mutations were identified in Lu36, including several BRAF alterations. Re1 had a baseline BRAF(E26D) variant, which has been reported as germline. See Supplemental Data 3 for complete list of alterations.

Extended Data Figure 3.. Characterization of cell lines with acquired G12Ci resistance.

a, G12Ci-sensitive lung cancer cells (H358) were selected in the absence (H358T) or in the presence of G12Ci-treatment (see methods), either in cell-culture (1M) or in athymic mice followed by cell-culture (R1, R2). b, The indicated cell lines were treated with sotorasib (left) or adagrasib (right) for 72h to determine the effect on cell proliferation (n=3, mean ± s.e.m). c, d, Resistant (1M, R1, R2) or parental (P) cells were treated with the indicated inhibitors for two weeks to determine the effect of cell viability by crystal violet staining. A representative of two independent experiments is shown. e, Athymic mice bearing parental (H358T, Par) or resistant (R1, R2) cell line xenografts were treated with vehicle or MRTX1257 (50 mpk) to determine the effect on tumor growth (n=5, mean ± s.e.m). f-g, The indicated cell lines were treated with increasing concentrations of sotorasib for 2 hours (f) or with 1μM over time (g). The effect on KRAS signaling was determined by immunoblotting. A representative of at least two independent experiments for each cell line is shown. For gel source data, see Supplementary Figure 1.

Extended Data Figure 4.. Single-cell modeling of G12Ci-resistant models.

a, Boxplots (median, upper and lower quartiles, and outliers) showing the distribution of the allele frequency (VAF) for the indicated variants across subclonal populations in G12Ci-resistant models. b, A neighbor-joining tree showing the relationship of the single-cells originating from the indicated parental or resistant models. The circular heat map indicates the presence (blue) or absence (white) of the indicated variants in each single cell.

Extended Data Figure 5.. Secondary RAS mutations in the absence of KRAS-directed therapy.

a, Heat map of 304 KRAS-mutant biopsy specimens harboring multiple RAS variants. b, Alluvial plot showing the pairings of mutations across samples. Residues with cancer-associated hotspot mutations are labeled. c, As in a but only KRAS(G12C) mutant samples are shown. d, Frequency of 2oRAS mutations in samples with KRAS(G12C) (left) or any KRAS mutation (right). All specimens were sequenced using MSK-IMPACT.

Extended Data Figure 6.. Propensity of treatment-emerging alterations to attenuate KRAS(G12C) inhibition.

a-g, H358 cells expressing the indicated variants under dox-inducible promoters (a-f) or H358 cells with CRISPR/CAS9 mediated deletion of KEAP1 or STK11 (g, ref. 19), were treated as shown to determine the effect on signaling intermediates by immunoblotting (a) or cell viability (b-g) using cell titer glow. In b-f, n=3, in g, n=4. Mean ± s.e.m. are shown. A representative of at least two independent experiments is shown. For gel source data of a, see Supplementary Figure 1.

Extended Data Figure 7.. Progressive attenuation of KRAS(G12C) singling inhibition during drug selection.

a-b, KRAS(G12C) mutant cells were expanded in the presence of G12Ci-treatment to establish isogenic lineages (1M, R1 and R2) with acquired resistance. Serial passages (p) from the indicated lineages were assayed to determine the magnitude and duration of ERK inhibition (a) or cleaved PARP induction (b) after drug re-challenge for 0–72h. In the 1M series, p15 and p25 denote passages when the selection drug concentration was increased. The experiment was carried out several passages later. p0 denotes parental cells. pERK and cPARP immunoblots were quantified with imageJ and their expression level was normalized to time 0. c, Unlabeled parental (H358) cells and BFP-labeled derivatives expressing dox-inducible NRAS(Q61K) were co-cultured in the presence of dox and/or sotorasib for 72h (n=4, mean ± s.e.m). d, Parental H358 cells or their derivatives expressing NRAS(Q61K) (100%) were treated in the presence of sotorasib in the presence or absence of dox for 72h (n=4, mean ± s.e.m.). Norm: min-max normalization. e-f, RFP-labeled parental (H358) cells and GFP-labeled derivatives expressing dox-inducible NRAS(Q61K) were co-cultured in the presence of dox and/or sotorasib for 72h to determine the distribution of subpopulations (e) by FACS (n=20,000 independent single cells) and the effect of the minor 2oNRAS subclone on the major KRAS(G12C)-mutant parental subpopulation (f). A representative of two independent experiments is shown in e-f.

Extended Data Figure 8.. Selective vulnerabilities in resistant cells harboring KRAS and NRAS mutations.

a, Parental H358 (sensitive) and R1 (resistant) cells expressing CAS9 were transfected with a genome-wide sgRNA library. The cells were treated in triplicate with either DMSO or G12Ci (sotorasib, 1 µM) for 14 days. The scaled mean expression of four independent sgRNAs targeting the indicate genes is shown. b-c, Parental or resistant cells expressing control (sgNT) or SHOC2-specific sgRNAs were subjected to immunoblotting to determine the expression of SHOC2 (b) or the effect on the indicated signaling intermediates (c). d-f, Parental (d), R1 (e) and R2 (f) cells were treated with either DMSO or sotorasib for 10 days to determine the effect on cell number (n=4, mean ± s.e.m). g-h, R1 cells expressing NRAS-specific siRNAs were treated with sotorasib for the indicated times to determine the effect on signaling intermediates (g) or proliferation (h, n=6, mean ± s.e.m). A representative of at least two independent experiments is shown. For gel source data, see Supplementary Figure 1.

Extended Data Figure 9.. Co-targeting ERK-signaling enhances KRAS(G12C) inhibition in models harboring secondary RAS/BRAF mutations.

a-b, The indicated models were treated with sotorasib (1 µM) in combination with trametinib (25 nM, a) or LXH254 (2 µM; b) to determine the effect on ERK signaling intermediates. For gel source data, see Supplementary Figure 1. c. Resistant and parental cell lines were treated with G12Ci (sotorasib, 1 µM; adagrasib, 200 nM or MRTX1257, 200 nM) in combination with a RAFdi (LXH254, 2µM), MEKi (trametinib, 50 nM) or ERKi (SCH984, 500 nM) to determine the effect on cell viability. d-f, H358 cells expressing the dox-induced variants shown were treated with sotorasib (1 µM) alone or in combination with the noted inhibitors to determine the effect on cell viability over time, using cell titer glow (d, n=4 and e, n=6; mean is shown) or crystal violet staining (f, 10 days). A representative of two independent repeats is shown.

Extended Data Figure 10.. Targeting KRAS(G12C) in combination with MAPK intermediates in vivo.

Mice bearing the indicated cell line (a, n=5, mean ± s.e.m.) or patient-derived xenograft (b, n=4, mean ± s.e.m.) were treated with the inhibitors shown to determine the effect on tumor growth. Fractional differences in tumor volume over time are shown.

Figure 1.. Genetic alterations associated with resistance to sotorasib treatment.

a, Heatmap of baseline and treatment-emergent alterations. b, The number of emergent alterations in the indicated genes. c, Allele frequencies for the indicated variants before (B) and after (A) treatment. n.d.: not detected; mult: multiple.

Figure 2.. Treatment-emergent alterations in pre-clinical models.

a, Heatmap of notable alterations in vehicle or G12Ci-treated PDX models. b, VAFs of the indicated variants in Lu3. c, Single-cell and bulk DNA sequencing in parental and resistant cells. The number of single cells harboring the indicated alterations (top), as well as their VAFs by scDNA and bulk DNA sequencing (right) are shown. d, The composition of parental and resistant models. *: MRAS was not part of the single-cell sequencing panel, and its frequency was estimated through bulk sequencing.

Figure 3.. Temporal tracking of treatment-emergent alterations.

a, The allele frequency of RAS/BRAF variants in intermediate passages from the resistant lineages shown. b-c, Unlabeled parental (H358) cells and BFP-labeled derivatives expressing dox-inducible NRAS(Q61K) were co-cultured in the presence of dox and sotorasib for 72h to determine the distribution of subpopulations (b) by FACS (n=20,000 independent single cells) and the effect of the minor 2oNRAS subclone on the major parental subpopulation (c). A representative of three independent experiments is shown in b and c.

Figure 4.. Effect of co-targeting ERK-signaling intermediates.

a, Genome-wide sgRNA screens in sotorasib-resistant (R1) and parental (H358) cells. b, Xenograft-bearing mice were treated as shown to determine the effect on tumor growth (each biological replicate along with the mean ± s.e.m are plotted, two tailed t-test p values are shown).