Co-targeting SOS1 enhances the antitumor effects of KRASG12C inhibitors by addressing intrinsic and acquired resistance

Abstract

Combination approaches are needed to strengthen and extend the clinical response to KRAS^G12C inhibitors (KRAS^G12Ci). Here, we assessed the antitumor responses of KRAS^G12C mutant lung and colorectal cancer models to combination treatment with a SOS1 inhibitor (SOS1i), BI-3406, plus the KRAS^G12C inhibitor, adagrasib. We found that responses to BI-3406 plus adagrasib were stronger than to adagrasib alone, comparable to adagrasib with SHP2 (SHP2i) or EGFR inhibitors and correlated with stronger suppression of RAS-MAPK signaling. BI-3406 plus adagrasib treatment also delayed the emergence of acquired resistance and elicited antitumor responses from adagrasib-resistant models. Resistance to KRAS^G12Ci seemed to be driven by upregulation of MRAS activity, which both SOS1i and SHP2i were found to potently inhibit. Knockdown of SHOC2, a MRAS complex partner, partially restored response to KRAS^G12Ci treatment. These results suggest KRAS^G12C plus SOS1i to be a promising strategy for treating both KRAS^G12Ci naive and relapsed KRAS^G12C-mutant tumors.

Fig. 1. Efficacy of combination treatments in non-small cell lung cancer and colorectal cancer cell lines.

a, Comparison of combination scores (cScore) indicating the response of KRAS^G12C mutant NCI-H2122 cells to the combination treatment of AMG 510 or adagrasib with 179 other small molecules for 72 h. Positive scores indicate additive effects with scores >~10 generally indicating a ‘clear additive’ effect. Empty or orange/brown colored circles highlight selected compounds for reference. n = 1 independent culture per cell line, 2–72 technical replicates per treatment (Supplementary Table 1). b, Comparison of cScores indicating the response of KRAS^G12C mutant SW837 cells to combination treatment of AMG 510 (sotorasib) or adagrasib for 144 h. Positive scores and empty/colored circles are explained in a. n = 1 independent culture per cell line, 2–72 technical replicates per treatment (Supplementary Table 1). c, Response of indicated cell lines to the combination treatment of adagrasib with a SOS1 inhibitor (BI-3406), a SHP2 inhibitor (TNO155 or SHP099) or an EGFR inhibitor (cetuximab). Cell lines are indicated as CRC (orange), esophageal cancer (light blue) or NSCLC (dark blue). MPAS, KRAS copy number status, K/H/N/MRAS expression and co-occurring mutations for each cell line are indicated. A strong additive benefit is defined with a cell growth inhibition (CGI) ≥ 50 and a Bliss CGI value ≥ 10 in ≥5 neighboring wells; medium additive benefit is defined with a CGI ≥ 30 and a Bliss CGI value ≥ 10 in ≥5 neighboring wells. n = 2 independent cultures per cell line. d, Growth kinetic of SW837 (left) or NCI-H2122 (right) cells treated with indicated single or combination treatments or vehicle control. The y axis indicates change in confluency relative to t = 0 h. Inset shows boxed area for clarity. n = 1 culture per curve, three technical replicates per culture. Source data

Fig. 2. Efficacy of combination treatments in non-small cell lung cancer and colorectal cancer xenograft models.

a, Tumor volume (TV) of mice bearing NCI-H2122 cells. Mice were treated with vehicle, BI-3406 (50 mg kg⁻¹; bid, delta of 6 h; qdx5), TNO155 (10 mg kg⁻¹; twice daily; qdx5), adagrasib (100 mg kg⁻¹; qdx5) for 16 days, or with adagrasib plus BI-3406 or TNO155 for 30 days (left). n = 7 animals. On day 16, combination groups were compared to the adagrasib group using a one-sided Mann–Whitney–Wilcoxon U-tests adjusted for multiple comparisons using the Bonferroni–Holm Method within each subtopic. Mean ± s.e.m. shown. Relative NCI-H2122 TVs, up to one doubling (+100%), are indicated as percent change from baseline at day 16 (right). b, Tumor volumes of mice bearing SW837 cells (left). Mice were treated with vehicle, BI-3406 (50 mg kg⁻¹; bid, delta of 6 h; qdx5), cetuximab (20 mg kg⁻¹; q3 or 4d) or adagrasib (100 mg kg⁻¹; qdx5) or with adagrasib plus cetuximab or BI-3406 for 42 days. n = 7 animals for monotherapies; n = 10 animals per group for combination therapies. On days 28 and 42, combination groups were compared to the adagrasib group using one-sided Mann–Whitney–Wilcoxon U-tests adjusted for multiple comparisons with the Bonferroni–Holm Method within each subtopic. Mean ± s.e.m. shown. Relative SW837 TVs, up to one doubling (+100%), are indicated as percent change from baseline at day 28 (right). c, Change in TV in NSG mice implanted with KRAS^G12C F3008 CRC PDX fragments and treated with vehicle, adagrasib (100 mg kg⁻¹, daily), BI-3406 (50 mg kg⁻¹, twice daily), SHP099 (25 mg kg⁻¹, daily), cetuximab (15 mg kg⁻¹, twice weekly) or adagrasib plus BI-3406, SHP099 or cetuximab. n = 8 animals. d, Change in TV in NSG mice implanted with KRAS^G12C B8032 CRC PDX fragments and treated with vehicle, adagrasib (100 mg kg⁻¹, daily), BI-3406 (50 mg kg⁻¹, twice daily), cetuximab (20 mg kg⁻¹, twice weekly) or adagrasib plus BI-3406 or cetuximab. n = 5 animals. Comparison of combination groups to the adagrasib group was analyzed by a two-sided paired Student’s t-test (c,d) and resulting P values were adjusted for multiple comparisons with the Bonferroni–Holm method. Mean ± s.e.m. is shown. qd, once daily; bid, twice daily; q3, every 3rd day; 4d, every 4th day; qdx5, 5 days treatment per week. Source data

Fig. 3. Modulation of RAS-MAPK signaling by adagrasib monotherapy or combination treatments in a KRAS^G12C mutant non-small cell lung cancer cell line.

a, G-LISA plate analysis of total RAS-GTP levels across the indicated cell lines treated with vehicle (DMSO), adagrasib, BI-3406 or the combination of adagrasib plus BI-3406 for 2 h. Adagrasib concentrations used represent the IC₅₀ values for each cell line: 5 nM (MIA PaCa-2), 10 nM (NCI-H358), 20 nM (SW837) or 150 nM (H2122). BI-3406 concentration was 500 nM for all cell lines. Data are shown as percentage of the vehicle control at for each cell line; n = 2 independent cultures. b, NCI-H2122, NCI-H358 and SW837 cells were treated with vehicle (DMSO), adagrasib (150 nM), BI-3406 (1200 nM), TNO155 (1000 nM), adagrasib + BI-3406 or adagrasib + TNO155 for 2 or 48 h. The cell lysates were subjected to pull down using RAF-RBD beads followed by immunoblotting for KRAS, NRAS, HRAS and MRAS. GAPDH served as the loading control. Images are representative of two independent experiments. c, Western blot analysis of NCI-H2122 cells treated with adagrasib, BI-3406, SHP099 or TNO155 alone or adagrasib combined with BI-3406, SHP099 or TNO155 at indicated concentrations (nM) for either 6 or 24 h. KRAS shift indicates covalent binding of compound to KRAS^G12C resulting in a slower migrating form of KRAS. β-Actin served as the loading control. Ada, adagrasib; BI, BI-3406; SHP, SHP099; TNO, TNO155. The image is of one experiment with the indicated compounds. The western blot was repeated with SOS1i BI-1701963 and KRAS^G12Ci BI-1823911 with similar results. DMSO, dimethylsulfoxide. Source data

Fig. 4. Differential gene expression induced by single-agent or combination treatments in vivo in a KRAS^G12C-mutant non-small cell lung cancer xenograft model.

All treatments were administered for 7 consecutive days, and samples were collected after the last dose. a, Venn diagram showing the overlap of upregulated (left) or downregulated (right) differentially expressed genes in NCI-H2122 xenograft models treated with adagrasib (100 mg kg⁻¹) alone or combined with BI-3406 (50 mg kg⁻¹) or TNO155 (10 mg kg⁻¹) at 4 h post-last dose. n = 4 animals. b, Heatmap showing the normalized enrichment score of hallmark pathway gene sets in tumors from KRAS^G12C-mutant NSCLC xenograft model (NCI-H2122) xenograft models treated with indicated single or combination therapies at 4 h post-last dose. NES, normalized enrichment score. Stars on the heatmap indicate adjusted P value < 0.05. Nominal P values are obtained by GSEA and adjusted for multiple comparisons using the Benjamini–Hochberg method. n = 4 animals. Source data

Fig. 5. Differential modulation of single-agent or combination treatments in vivo on MAPK signaling and other pathways.

a, Modulation of overall MAPK pathway activity, as indicated by the MPAS, in NCI-H2122 xenograft models treated with indicated monotherapies or combination treatments at 4, 24 or 48 h post-last doses. n = 4 animals. Boxplots show low and upper quartiles and median line is indicated. Whiskers, 1.5 × interquartile range. Data were analyzed by a two-sided Wilcoxon rank-sum test. MPAS scores for adagrasib alone or in combination at 24 and 48 h post-last dose were compared against that of adagrasib at 4 h post-last dose. b, Differential modulation of select MAPK pathway genes in NCI-H2122 xenograft models treated with indicated single-agent or combination treatments. Tumors were collected at 4, 24 or 48 h post-last dose. n = 4 tumors per time point. Color legend, normalized expression. c, Percentage of positive tumor cells expressing DUSP6 or EGR1 mRNA analyzed by RNAscope in F3008 CRC PDX models treated with vehicle, adagrasib (100 mg kg⁻¹, daily), adagrasib plus BI-3406 (50 mg kg⁻¹, twice daily) or adagrasib plus cetuximab (15 mg kg⁻¹, twice weekly) for 5 days (top). Tumors were collected 4 h after the last dose. n = 5 animals. Boxplots show low and upper quartiles and median line is indicated. Whiskers indicate 1.5 × interquartile range. Data analyzed by two-sided Student’s t-test. Representative images from RNAscope analysis of DUSP6 and EGR1 expression in tumor tissue from models described in the top (bottom). Scale bar, 100 µm. d, Percentage of positive tumor cells expressing phosphor-ERK as analyzed by multiplex immunofluorescence in the same tumor tissues as described in c. n = 5 animals. Boxplots show low and upper quartiles and median line is indicated. Whiskers indicate 1.5 × interquartile range. Comparisons to vehicle analyzed by two-sided Student’s t-test. Representative images from multiplex immunofluorescence analysis of p-ERK and HLA in tumor tissue from models described in the top (bottom). Scale bar, 100 µm. NS, not significant. Source data

Fig. 6. Adagrasib-resistant models remained sensitive to an adagrasib plus SOS1i combination.

a, Tumor growth in the KRAS^G12C-mutant CRC xenograft model (SW837) treated with vehicle (orange) or adagrasib (blue, 50 mg kg⁻¹). n = 7 control mice and n = 30 out of 300 mice (Extended Data Fig. 8a) treated with adagrasib are shown. b, Outgrowing adagrasib-resistant tumors from a that had increased in size to ≤100 mm³ were treated with adagrasib (50 mg kg⁻¹, 5 times per week, orally), adagrasib plus BI-3406 (50 mg kg⁻¹, twice daily, orally), adagrasib plus cetuximab (20 mg kg⁻¹, twice weekly, i.p.) or adagrasib plus TNO155 (10 mg kg⁻¹, twice daily). n = 5 animals per group. Mean ± s.e.m. are shown. Combination groups were compared to adagrasib group on day 31 using a one-sided Mann–Whitney–Wilcoxon U-tests adjusted for multiple comparisons according to the Bonferroni–Holm Method within each subtopic. c, Modulation of overall MAPK pathway activity, as indicated by the MPAS, in tumors derived from SW837 xenograft models. Tumors were collected from models before adagrasib treatment (pretreatment) and during relapse as well as from adagrasib-resistant models treated with adagrasib plus BI-3406 or cetuximab. Boxplots show low and upper quartiles and median line is indicated. Whiskers indicate 1.5 × interquartile range. Data were analyzed by a two-sided Wilcoxon rank-sum test. Bold P values indicate comparison to adagrasib_4 h group; italicized red P values indicate comparison to relapse group. Vehicle, n = 4 animals; adagrasib_4 h, n = 5 animals; adagrasib_24 h, n = 5 animals; adagrasib_48 h, n = 4 animals; relapse, n = 16 animals; adagrasib plus cetuximab, n = 8 animals; adagrasib plus BI-3406, n = 3 animals. d, Volcano plot of differentially regulated genes between pre-adagrasib treatment (pretreatment) and relapsed SW837 tumors. The most significantly differential genes involved in MAPK pathway are highlighted. P values were obtained by two-sided Wald test and adjusted for multiple comparisons using Benjamini–Hochberg method. n = 6 and 16 tumors in pre-treatment and relapse groups, respectively. e, Cell viability of adagrasib-resistant NCI-H358 cells transfected with siRNA targeting MRAS (top) or SHOC2 (bottom) or non-targeting control (NTC) and treated with control or indicated single-agent or combination therapies. A dashed line was added to enable visual comparison to MRAS-/SHOC2-expressing cells treated with adagrasib monotherapy. Data were analyzed using a one-sided t-test with P values adjusted for multiple comparisons using Bonferroni method. n = 3 independent cell cultures. f, Schematic representing the putative mechanism of resistance to KRAS^G12Ci and overcoming resistance by KRAS^G12Ci and SOS1i combination. i.p., intraperitoneally. Source data

Extended Data Fig. 1. Molecular signatures of KRAS^G12C-driven tumor cell lines.

a, Comparison of combination scores (cScore) indicating the response of KRAS^G12C mutant NCI-H358 cells to combination treatment of AMG 510 (sotorasib) or adagrasib for 144 hours on a sub-panel of 61 representative compounds. Positive scores indicating additive effects with scores >~10 generally indicating a “clear additive effect.” Empty or orange/brown colored circles highlight selected compounds for reference. N = 1 independent culture per cell line. b, Comparison of combination scores (cScore) indicating the response of KRAS^G12C mutant NCI-H2122 cells and KRAS^G12C mutant SW837 cells to combination treatment of adagrasib. NCI-H2122 and SW837 cells were treated for 72 and 144 hrs, respectively. Positive scores and empty/colored circles are explained in (a). N = 1 independent culture per cell line. c, (left, top) Western blot analysis of the KRAS^G12D- and KRAS^G12C-mutant isogenic murine lung cancer cell line, LKR13, without (LKR13K) or with Lkb1 (LKR13KL) or Keap1 knockout (LKR13KK). (right, bottom) Effects of adagrasib or adagrasib plus BI-3406 or TNO155 in a cell viability assay in KRAS^G12C-mutant LKR13K, LKR13KL, and LKR13KK cells. Mean percentages of cell viability relative to the DMSO-treated control are shown. Error bars = ±SD; n = 3 independent cell cultures. d, Activity of hallmark cancer gene sets, as analyzed by RNA-seq, in untreated KRAS^G12C-driven tumor cell lines found generally responsive (yes; NCI-H2122, NCI-H1373, SW837, NCI-H358, LU65) or unresponsive (no; SW 1573, HOP-62, HCC-44, NCI-H2030, NCI-H1792, KYSE-410) to adagrasib combination treatments in Fig. 1c. e, KRAS expression, MRAS expression, and MAPK pathway activity (MAPK Pathway Activity Score [MPAS]) in responsive (yes) and unresponsive (no) cells described in (a). d-e: Each boxplot shows lower and upper quartile and median line is indicated. Range of whiskers: 1.5x interquartile range (two-sided Wilcoxon rank sum test, confidence interval 0.95; P value is indicated). Each cell line, characterized in Fig. 1c, was sequenced once. Source data

Extended Data Fig. 2. Efficacy of adagrasib single-agent and combination therapy in NCI-H358.

a, Relative tumor volume in NCI-H358 xenograft models treated with vehicle, adagrasib (100 mg/kg, daily), BI-3406 (50 mg/kg, twice daily), TNO155 (10 mg/kg, twice weekly), or the combination of adagrasib plus BI-3406 or TNO155 for 95 days. Tumor volumes were monitored until Day 157. N = 10 animals. Mean ± SEM shown. b, Tumor volume of NCI-H358 models given adagrasib (ada) alone or plus BI-3406 (Ada+BI) or TNO155 (Ada+TNO) on Day 1 (left) or Day 35 and Day 94 (middle) of treatment. All mice were taken off treatment on Day 95. Tumor volume was monitored until Day 157 (right). Differences from the adagrasib vs combination groups were analyzed using a two-sided non-parametric Mann–Whitney-Wilcoxon U-test. Day 94 adagrasib, adagrasib+TNO155; Day 157 adagrasib, adagrasib+TNO155: n = 9 animals; all other groups: n = 10 animals. Median indicated by horizontal line. Source data

Extended Data Fig. 3. Modulation of RAS-MAPK signaling by adagrasib monotherapy or combination treatments in a colorectal cancer (CRC) cell line.

Western blot analysis of the KRAS^G12C-mutant colorectal cancer cell line, SW837, treated with adagrasib, BI-3406, or TNO155 alone, or adagrasib combined with BI-3406 or TNO155 at indicated concentrations (nM) for either 6 or 24 hours. KRAS shift indicates covalent binding of compound to KRAS^G12C, resulting in a slower migrating form of KRAS. β-actin served as the loading control. For blots with ERK and cleaved PARP antibodies, 15 µg total protein/lane; for all other blots, 20 µg total protein/lane. Ada: adagrasib; BI: BI-3406; TNO: TNO155. Image is of one experiment with the indicated compounds. Western blot was repeated with SOS1i BI-1701963 and KRASG12Ci BI-1823911 with similar results.

Extended Data Fig. 4. Gene modulation by single-agent or combination treatments in KRAS^G12C–driven non-small cell lung cancer (NSCLC) and colorectal cancer (CRC) xenograft models.

a, Venn diagram showing the overlap of upregulated (left) or downregulated (right) differentially expressed genes in NCI-H2122 xenograft models treated with adagrasib (100 mg/kg) alone or combined with BI-3406 (50 mg/kg) or TNO155 (10 mg/kg) at 24 and 48 hours post-last dose. N = 4 animals. b, All treatments in SW837 xenograft models were administered over seven consecutive days. Overlap of upregulated (left) or downregulated (right) genes in tumors from SW837 xenograft models across three time points after receiving one of these treatments: 50 mg/kg bid, p.o. BI-3406 (BI); 20 mg/kg twice a week, i.p. cetuximab (Cet); 100 mg/kg qd, p.o. adagrasib (Ada); 50 mg/kg BI-3406 plus 100 mg/kg adagrasib (BI + Ada); 20 mg/kg cetuximab plus 100 mg/kg adagrasib (Cet + Ada). Samples were collected at 4, 24, and 48 hours after the last dose. Sample sizes are the same as Extended Data Fig. 5b. Source data

Extended Data Fig. 5. Differentially modulated pathways in a KRAS^G12C–driven colorectal cancer (CRC) xenograft model.

a, Heatmap showing the normalized enrichment score (NES) of hallmark pathway gene sets in tumors from SW837 xenograft models treated with indicated single or combination therapies at 4 hours post-last dose. Asterisks on the heatmap indicate adjusted p value < 0.05. Nominal p values are obtained by gene set enrichment analysis (GSEA) and adjusted for multiple comparisons using Benjamini–Hochberg method. Number of animals per group is same as Extended Data Fig. 5b. b, Modulation of overall MAPK pathway activity, as indicated by the MAPK Pathway Activity Score (MPAS), in tumors from SW837 xenograft models treated with indicate single-agent or combination treatments. Samples were collected at 4, 24 and 48 hours after the last dose. Boxplots show lower and upper quartiles and median line is indicated. Whiskers: 1.5 x interquartile range. Data analyzed by two-sided Wilcoxon rank sum test. Number of animals per group indicated on graph. c, Differential modulation of selected MAPK pathway genes in SW837 xenograft models treated with indicated single-agent or combination treatments. Samples were collected at 4, 24, or 48 hours post last dose. Sample sizes/group reported in b. Color legend: normalized expression. Number of animals per group is same as Extended Data Fig. 5b. d,e, RNAscope analyses of DUSP6, EGR1, and SPRY4 expression (d) as well as MRAS and KRAS expression (e) in CRC PDX models, F3008 and B8032, at the end of tumor growth curve experiments shown in Fig. 2c,d, respectively. Comparisons to vehicle (normal text) or adagrasib group (italicized text) analyzed by one-way analysis of variance followed with Holm-Šídák’s multiple comparisons test; p < 0.05 indicates significance; n.s. = non-significance. Boxplots show lower and upper quartiles and median line is indicated. Whiskers: 1.5 x interquartile range. F3008 or B8032: n = 8 or n = 5 tumors/group, respectively. Source data

Extended Data Fig. 6. Biomarker analysis in KRAS^G12C–driven non-small cell lung cancer (NSCLC) or KRAS^G12C–driven colorectal cancer (CRC) tumors treated with single-agent or combination treatments.

All treatments were administered over 7 consecutive days, with samples collected on Day 7 and 8 after the last dose. a, Biomarker analyses of cell proliferation biomarker, Ki-67 (top), and phosphor-ERK (p-ERK) (bottom), in NCI-H2122 xenograft models treated with vehicle (con), BI-3406 (50 mg/kg, bid, p.o.), adagrasib (100 mg/kg, qd, p.o.), TNO155 (10 mg/kg, bid, p.o.), adagrasib (100 mg/kg) plus BI-3406 (50 mg/kg), adagrasib (100 mg/kg) plus TNO155 (10 mg/kg) at 4 or 48 hours post-last dose. Control animals were treated with vehicle. Boxplots show low and upper quartiles and median line is indicated. Whiskers: 1.5 x interquartile range. Data analyzed by one-sided t-test with resulting p values adjusted for multiple comparisons using the Benjamini–Hochberg method within each subtopic. n = 5 tumors/group. b, Biomarker analyses of cell proliferation biomarker, Ki-67 (top), and phosphor-ERK (p-ERK) (bottom), in SW837 xenograft models treated with vehicle (con), BI-3406 (50 mg/kg, bid, p.o.), adagrasib (100 mg/kg, qd, p.o.), TNO155 (10 mg/kg, bid, p.o.), adagrasib (100 mg/kg) plus BI-3406 (50 mg/kg), adagrasib (100 mg/kg) plus TNO155 (10 mg/kg) at 4 or 48 hours post-last dose. Control animals were treated with vehicle. Boxplots show low and upper quartiles and median line is indicated. Whiskers: 1.5 x interquartile range. Data analyzed by one-sided t-test with resulting p values adjusted for multiple comparisons using the Benjamini–Hochberg method within each subtopic. n = 6 and n = 5 tumors/group. Source data

Extended Data Fig. 7. Evaluating mechanisms underpinning adagrasib resistance in vitro.

a, Growth response of BaF/3 cells harboring a KRAS^G12C single site variant library to KRAS^G12C-specific inhibitors alone, adagrasib (100 nM) or sotorasib (300 nM), or combined with 200 nM or 600 nM of BI-3406. n = 1 experiment. b, Cell viability following adagrasib treatment on MIA PaCa-2, NCI-H358, or SW837 parental cells or clones (cl) harboring additional mutations at KRAS position R68S, H95R or Y96D. Mean percentages of cell viability relative to the DMSO-treated control are shown. Error bars indicate mean ± SD; NCI-H358 KRAS H95R clones: n = 1 cell cultures; all other parental cell lines or clones: n = 2 independent cell cultures. Each group is representative of at least 2 independent cell cultures, with the exception of NCI-H358 KRAS H95R cl. 59440 (n = 1) c, Effects of adagrasib, BI-3406, or adagrasib plus BI-3406 (top) or adagrasib, TNO155, or adagrasib plus TNO155 (bottom) in a cell viability assay in adagrasib-sensitive (parental) or adagrasib-resistant NCI-H358 cells. Mean percentages of cell viability relative to the DMSO-treated control are shown. Each graph is representative of one of n = 2 independent cell cultures. d, MRAS expression (left) and protein levels (right) in parental (DMSO) and adagrasib-resistant NCI-H358 cell pool. Left, n = 2 individual cell cultures; right, n = 3 independent cell cultures. Error bars indicate mean ± SD. e, MAPK pathway activity, as indicated by MPAS score, in parental NCI-H358 cells treated with DMSO control or adagrasib-resistant NCI-H358 cells without drug (OFF) or treated with adagrasib (300 nM) (ON). DMSO group: n = 6 independent cell cultures; other groups: N = 4 independent cell cultures. f, Protein expression analysis of vimentin and E-cadherin, two EMT markers in parental and adagrasib-resistant NCI-H358 cells. Source data

Extended Data Fig. 8. Combination treatments elicit responses from KRAS^G12Ci-resistant colorectal cancer (CRC) and non-small cell lung cancer (NSCLC) models.

a, Growth of SW837 tumors treated long-term with adagrasib (50 mg/kg; n = 300 mice total) or vehicle (0.5% Natrosol + 0.5% DMSO; n = 7 mice) on a 5 days on/2 days off dosing schedule. Outgrowth indicating tumors that developed acquired resistance to adagrasib is indicated with a black rectangle. b, Growth of adagrasib-resistant SW837 tumors (n = 35 animals) from (a) that were further treated with the following for 31 days: adagrasib (50 mg/kg, qd, p.o.,), adagrasib plus BI-3406 (50 mg/kg, bid, p.o.), adagrasib plus cetuximab (20 mg/kg, q3 or 4d, i.p.), or adagrasib plus TNO155 (10 mg/kg, bid). Bar indicates median. Data were analyzed on Day 31 by an one-sided non-parametric Mann–Whitney-Wilcoxon U-tests adjusted for multiple comparisons according to the Bonferroni–Holm Method within each subtopic. Day 0 group, n = 35 mice; adagrasib group, n = 8 mice; other groups, n = 9 mice. ada = adagrasib; n.s. = non-significant. c, Individual tumor volumes of NCI-H2122 xenograft mice treated with vehicle or adagrasib (100 mg/kg, daily) for 15 days. On Day 15, adagrasib-resistant tumors (circled) were then re-randomized for second-line treatment of either continued treatment with adagrasib alone or adagrasib combined with BI-3406. Day 15 was selected for a 2nd randomization as, at this time point, several adagrasib-treated NCI-H2122 tumors started to grow again. N = 8 mice in the vehicle group, N = 16 mice in adagrasib-treated group. d, Tumor volume (Mean ± SD) of NCI-H2122 models given adagrasib alone or plus BI-3406 for 35 days after re-randomization from (b). N = 8 mice in adagrasib group, N = 7 mice in combination therapy group. Statistical differences were analyzed using two-sided t-test. Source data

Extended Data Fig. 9. Molecular response of KRAS^G12Ci-resistant colorectal cancer (CRC) models to combination treatments.

a, Gene set enrichment analysis (GSEA) plots show downregulated tyrosine kinase, EMT, and KRAS pathways between tumors collected from SW837 models prior to adagrasib treatment and during relapse after chronic adagrasib treatment. EMT = epithelial-mesenchymal transition. NES = normalized enrichment score; FDR = false discovery rate-adjusted q value. Upregulation (red) and downregulation (blue) are shown. N = 6 and 16 animals in before treatment and relapse groups, respectively. b, Heatmap showing expression z-score of genes belonging to apoptosis, epithelial-mesenchymal transition (EMT), or KRAS gene sets, as well as MAPK or tyrosine kinase signaling pathways from tumors collected from SW837 xenograft models prior to adagrasib treatment (Before Treatment, n = 6 sequenced tumors) and during relapse (Relapse, n = 16 sequenced tumors). Color legend: normalized expression. Source data

Extended Data Fig. 10. MRAS and SHOC2 regulation in adagrasib-sensitive and -resistant models.

a, MRAS expression in parental and adagrasib-resistant NCI-H358 cells transfected with siRNA targeting MRAS (siRNA) or with a non-targeting control (NTC). Error bars indicate mean ± SD. The dotted line represents the baseline level in sensitive (parental NTC) cells. All groups: n = 3 qPCR from independent cell cultures. b, SHOC2 expression parental and adagrasib-resistant (HD pool) NCI-H358 cells from (a) and transfected with siRNA targeting SHOC2 (siRNA) or non-targeting control (NTC). Dotted line represents the baseline level in sensitive (NTC parental) cells. n = 3 qPCR from independent cell cultures. Error bars indicate mean ± SD. c, MAPK activity (top) and MRAS expression (bottom) in SW837 xenograft tumors. SW837 models were treated with vehicle (n = 4 animals) or adagrasib (50 mg/kg) for 8 days (samples taken 4, 24, and 48 hours post-last dose, n = 4-5 animals/group) or adagrasib until relapse (n = 16 animals). Boxplots show low and upper quartiles and median line is indicated. Whiskers: 1.5 x interquartile range. Groups compared to vehicle using a two-sided Wilcoxon test; p values adjusted for multiple comparisons using Benjamini–Hochberg Method. FDR = false discovery rate-adjusted p value. Number of animals/group indicated on graph. d, MRAS expression in NCI-H2122 xenograft tumors treated with vehicle (Extended Data Fig. 8c) and in tumors following second-line treatment of adagrasib (100 mg/kg) alone or plus BI-3406 (50 mg/kg, bid) for 35 days (Extended Data Fig. 8d). Each boxplot shows lower and upper quartile; median line indicated. Whiskers: 1.5× interquartile range. Adagrasib group: N = 8 mice; other groups: N = 7 mice. e, MAPK activity (top) and MRAS expression (bottom) in indicated KRAS^G12C-mutated xenograft models treated with vehicle or adagrasib (100 mg/kg); data collected 24 hours post-last dose. Boxplots show low and upper quartiles; median line, indicated. Whiskers: 1.5 x interquartile range. Groups compared to vehicle using a two-sided Wilcoxon test; p values adjusted for multiple comparisons using the Benjamini–Hochberg Method. FDR = false discovery rate-adjusted p value. Number of tumors sequenced/group indicated on graph. Source data