Acquired resistance to the RAS(ON) multi-selective inhibitor daraxonrasib guides rational combination therapy strategies in pancreatic cancer.

abstract cc by

Daraxonrasib is an orally bioavailable RAS(ON) multi-selective tri-complex inhibitor of the oncogenic mutant and wild-type variants of N, H and KRAS. We previously reported encouraging efficacy in a phase 1/2 clinical trial evaluating daraxonrasib monotherapy at clinically active dose levels in patients with previously treated, RAS mutant metastatic pancreatic adenocarcinoma (PDAC), providing the basis for confirmatory evaluation in the randomized phase 3 RASolute 302 clinical trial. Here we report mechanisms of acquired resistance to daraxonrasib monotherapy observed through targeted sequencing of over 800 genes in paired pretreatment and end of treatment circulating tumor DNA samples from 44 patients in the phase 1/2 clinical trial. Treatment-emergent genomic alterations in the RAS signaling pathway were observed in more than half (26 of 44; 59%) of these patients, including, most notably, mutant KRAS amplifications in one-third (16 of 44; 36%), as well as alterations in receptor tyrosine kinase (RTK) (4 of 44; 9%), MAPK (11 of 44; 25%) and PI3K (4 of 44; 9%) pathways. Notably, no acquired secondary KRAS mutations were observed, distinct from resistance profiles of mutant-selective KRAS G12C(OFF) inhibitors. To corroborate these clinical findings, we found, or mechanistically established, concordant mechanisms of daraxonrasib resistance in human and murine preclinical models of PDAC, including mutant KRAS and MYC amplification and RTK upregulation, with these alterations guiding various combination therapy concepts. Notably, daraxonrasib combined with agents targeting DNA damage response, RTKs or the mutant-selective RAS(ON) G12D inhibitor zoldonrasib averted resistance in preclinical models. Collectively, these results show that most daraxonrasib genomic resistance mechanisms drive reactivation of RAS pathway signaling and guide potential combination strategies in PDAC for further investigation.

main cc by

Hotspot RAS mutations are among the most prevalent oncogenic drivers in cancer, including in PDAC, non-small cell lung cancer (NSCLC) and colorectal cancer (CRC). Mutated RAS is an oncogenic driver in over 90% of PDAC cases1,2. These hotspot mutations stabilize RAS in its active GTP-bound (‘ON’) state, driving increased RAS/MAPK signaling, proliferation, metastasis and immune evasion3–5. Therefore, broad-spectrum RAS(ON) inhibition holds the potential to benefit a large population of patients with cancers driven by activating RAS mutations6,7. Daraxonrasib is a RAS(ON) multi-selective inhibitor that targets the active, GTP-bound forms of KRAS, HRAS and NRAS, with affinity for both mutant and wild-type (WT) variants8,9. Daraxonrasib is currently being investigated in multiple clinical trials, including an ongoing phase 1/2 monotherapy study in patients with previously treated advanced RAS mutant tumors (NCT05379985)10,11. We previously reported the antitumor activity of daraxonrasib in this single-arm study in patients with metastatic PDAC receiving daraxonrasib as second-line treatment at 300 mg daily12). In the second-line RAS mutant population (n = 38), we observed a median progression-free survival (PFS) of 8.1 months (95% CI 5.9–10.1) and a median overall survival of 15.6 months (95% CI 10.9–not evaluable) with a median follow-up of 28 months (range 17.1–32.6)12. We recently reported confirmatory results in this second-line population in the randomized phase 3 RASolute 302 clinical trial13. While the clinical activity of daraxonrasib monotherapy is encouraging, many patients eventually experience disease progression indicating the development of acquired resistance on treatment. Mechanisms of resistance to RAS(ON) inhibitors in PDAC have not been elucidated and may be distinct from those for daraxonrasib in other solid tumors14 given the exquisite RAS-dependency of this disease15,16. Mechanisms of acquired resistance to KRAS G12C(OFF) inhibitors have been described in NSCLC, CRC and, to a more limited degree, in PDAC, which include secondary oncogenic KRAS mutations and mutant KRAS allele amplification (KRASamp)17,18. Identifying treatment-emergent resistance mechanisms has the potential to guide rational combination therapies. This study presents the first insights into the genomic mechanisms of acquired resistance to daraxonrasib monotherapy in PDAC and identifies potential combination therapy strategies to overcome resistance.

acquired genomic alterations in patients with metastatic pdac following progress cc by

We profiled paired circulating tumor DNA (ctDNA) samples (pretreatment and end of treatment (EOT) at the time of progression) from 44 patients with RAS mutant metastatic PDAC treated at a dose range of 160–300 mg daraxonrasib daily as second or later line therapy in a single-arm phase 1/2 trial (NCT05379985)12. Previously reported translational modeling predicted this dose range drives RAS pathway inhibition ≥90% and antitumor activity in humans8. Each patient exhibited either a complete response (CR), partial response (PR) or stable disease (SD) on treatment with a PFS of >3 months followed by subsequent progression on therapy (Extended Data Fig. 1a). Consistent with daraxonrasib activity, each patient showed a relative reduction in mutant KRAS variant allele frequency (VAF) in ctDNA from pretreatment time point (cycle 1, day 1; C1D1) to early on treatment (cycle 2, day 1 (C2D1) or cycle 3, day 1 (C3D1)). We subsequently observed an increase in mutant KRAS VAF at EOT in most patients (Extended Data Fig. 1b). We detected acquired oncogenic alterations in 26 of 44 (59%) patients, including several putative genomic resistance mechanisms resulting in aberrant RAS signaling: alterations in RAS, MAPK, PI3K and RTK components (Fig. 1a and Supplementary Table 1). Among the acquired oncogenic alterations in RAS pathway signaling genes, KRASamp was the most frequent, identified in 36% (16 of 44) of the patient samples profiled, accounting for 62% (16 of 26) of all putative RAS pathway genomic resistance mechanisms. Of note, we did not observe acquired secondary mutations in KRAS, in contrast to previous observations of these as resistance drivers to mutant-selective KRAS G12C(OFF) inhibition19,20.Fig. 1Putative acquired genomic resistance mechanisms to daraxonrasib treatment in patients with PDAC.a, Putative acquired genomic resistance mechanisms were identified in a majority of the 44 patients profiled. Each column represents a single patient, and each row represents a specific genomic alteration. First four rows depict confirmed best overall response by RECIST v.1.1, planned dose, RAS mutation at enrollment and pretreatment TP53 alteration status (based on pretreatment ctDNA). Supplementary Table 3 lists the KRAS CN for each patient. Gene names not italicized per standard oncoprint conventions. b, Summary of putative acquired genomic resistance mechanisms (RAS, MAPK, RTK, PI3K/mTOR and other (KEAP1 and MYC) pathways) in patients with PDAC. Eleven patients acquired more than one alteration in this pathway. c, Association of dose level of daraxonrasib (x axis) with acquisition of RAS signaling pathway genomic alterations (all genes shown in the oncoprint) at progression. P = 0.11, two-sided Fisher’s exact test. SV, short variant, defined as single-nucleotide variant (SNV) or short indel; CNV, copy-number variant, defined as ‘focal amplification’ (for oncogenes) or ‘homozygous deletion’ (for tumor-suppressor genes); copy number (CN) above 2 is defined as amplification; Fusion, requires at least one of the partner genes to be in pathway of interest; TRUNC, variant resulting in predicted truncated protein, defined as nonsense mutation, splice site or indel; qd, once daily. Panel b created in BioRender; Nayak, B. https://BioRender.com/23cpkzi (2026).Source data We detected multiple acquired alterations in ctDNA from 11 of 44 patients (25%). Putative genetic mechanisms of acquired resistance are summarized in the context of the RAS signaling pathway in Fig. 1b. We documented acquired alterations in RAF genes (5 of 44; 11%), including multiple single-nucleotide ARAF, BRAF and RAF1 variants along with co-occurring KRASamp in one patient (patient 11; Fig. 1a). RTK alterations (4 of 44; 9%), included HER2 (ERBB2) amplification (copy number of 2.31), a gene fusion in MET, each co-occurring with mutant KRASamp; a FGFR1 gene fusion, and a single-nucleotide variation in MET observed in two other patients. PI3K/mTOR signaling pathway genes (PIK3CA, AKT1, AKT2, RICTOR and PIK3R1) were altered in a minority of patient samples (4 of 44; 9%) with only an AKT2 amplification found without co-occurring KRASamp. We also observed truncating mutations in NF1 (2 of 44; 4%), KEAP1 (2 of 44; 4%), a PTPN11 mutation or MYC amplification in one patient each (1 of 44; 2%). Of 18 of 44 patients with no putative genomic resistance mechanism identified in ctDNA, 5 had no detectable acquired alterations in any of the 800+ genes examined (Supplementary Table 1). The remaining 13 of 18 patients acquired one or more alterations in genes either associated with clonal hematopoiesis of indeterminate potential, such as TP53, ASXL1, PPM1D and TET2 (refs. 21–23) or non-recurrent oncogenic alterations in tumor-suppressor genes not detected in pretreatment ctDNA, potentially due to low shedding (CDKN2A, SMAD4, ARID1A and BRCA1; Supplementary Table 4). Additionally, 2 of these 18 patients had pre-existing TP53 alterations and acquired additional loss of function mutations in this gene. Some patients were treated at doses lower than 160 mg in the dose-escalation phase of the aforementioned phase 1/2 trial12 (Extended Data Fig. 1). We found that acquired RAS pathway alterations were more frequent at the higher dose range of 160–300 mg daily than in ctDNA of 13 patients treated with daraxonrasib ≤120 mg daily (59% versus 31%), although this difference did not reach statistical significance (P = 0.11, two-sided Fisher’s exact test; Fig. 1c). Of note, the types of acquired alterations seen in 4 of 13 patients at lower doses were consistent with those seen in the higher dose range (KRASamp, and SNVs in BRAF, KEAP1, PTPN11, NF1, RIT1 and PIK3CA; Supplementary Table 2).

acquired mutant cc by

We further examined genomic features in ctDNA from 16 patients with acquired KRASamp, the most frequently observed resistance mechanism (Fig. 1a and Supplementary Table 3). While the acquired KRASamp copy number observed at progression showed a wide range from 2.21 to 28.36 copies, we found no relationship between this copy number and daraxonrasib dose or baseline KRAS mutant allele (KRAS G12D, KRAS G12V, KRAS G12R and KRAS Q61H). Next, we dissected the dynamics of KRASamp acquisition in ctDNA from four patients, one who experienced a CR, and three with PR or SD, followed by progression on treatment (Fig. 2 and Extended Data Fig. 2). Each patient had a baseline KRAS copy number of 2 in pretreatment ctDNA, and as noted above, showed a relative reduction in mutant KRAS VAF in ctDNA at C2D1. We then observed a marked increase in mutant KRAS VAF at progression in conjunction with detection of KRASamp in ctDNA. In patient 2, we confirmed acquired KRASamp in EOT ctDNA via next-generation sequencing from a tumor biopsy collected at progression (Fig. 2b). In this patient, we detected KRASamp in ctDNA one cycle immediately before (C12D1) radiological tumor progression at C13, consistent with KRASamp as a putative resistance mechanism. EOT ctDNA from patient 1 showed acquisition of MYCamp as an additional putative resistance mechanism. Using a previously reported method to infer whether amplification observed in ctDNA occurred in the mutant or WT KRAS allele24, we calculated the ratio of mutant KRAS VAF relative to non-RAS tumor mutation VAFs (KRASMUT VAF/non-KRAS max VAF). We observed a relative ratio increase of >1 for mutant KRAS VAF relative to other somatic mutations in all 16 patients with KRASamp at progression, indicating that the mutant KRAS allele present before therapy is amplified in each case (Supplementary Table 3 and Fig. 2c–f).Fig. 2KRASamp is a recurrent mechanism of acquired genomic resistance to daraxonrasib in patients with PDAC.a,b, Computed tomography scans for two patients who had complete (a) or partial responses (b) to daraxonrasib (21-day cycles) followed by disease progression. ctDNA profiling in both patients showed complete clearance of KRASG12D or KRASG12R early on-treatment preceding the clinical response followed by a large increase in KRASG12D or KRASG12R VAF at progression (EOT) and newly acquired KRAS amplification (a,b, bottom tables). Additionally, patient 1 was also found to have acquired MYC amplification as detected in the EOT ctDNA. In patient 2, a paired pretreatment and EOT liver biopsy was sequenced with comprehensive genomic profiling which confirmed the acquired KRAS amplification at progression detected in ctDNA. The patient also had an oncogenic TP53 L194R alteration before therapy detected in pretreatment ctDNA and liver biopsy. c–f, Representative ctDNA dynamics in individual patients with varying acquired KRAS amplification copy numbers at EOT. VAF at pretreatment (Pre), early on-treatment at cycle 2 day 1 (On), and EOT at progression (End) with daraxonrasib. RAS mutant allele represented in dark green and the highest somatic mutation in a non-RAS gene shown in light green. Other somatic oncogenic alterations are shown in gray. In each patient, KRAS amplification was not detected at Pre and On time points (CN = 2) and was detected at EOT at the CN shown. g, Association of acquired KRAS amplification at progression (x axis) with pretreatment TP53 oncogenic mutation. P = 0.03, two-sided Fisher’s exact test. One patient with acquired KRAS amplification at progression did not have a pretreatment TP53 alteration detected in pretreatment ctDNA (low shedding with max VAF = 0.59%); however, a baseline TP53 alteration was detected in archival tumor tissue testing (13.5% VAF). ctDNA dynamics for all patients with KRAS amplification are shown in Supplementary Tables 1 and 3. ND, not detected.Source data Finally, we evaluated the association of mutant KRASamp at acquired resistance with common genomic co-alterations occurring at > 10% frequency in ctDNA from the 44 patients profiled herein. TP53 mutations were the most common co-alterations found (Fig. 2g; present in 31 of 44; 70%), consistent with the expected frequency in KRAS mutant PDAC1,2. Furthermore, we observed a significant association between a pretreatment TP53 alteration and acquired KRASamp at progression: 15 of 32 (47%) patients with pretreatment oncogenic TP53 mutation in ctDNA showed acquired KRASamp at progression in comparison to 1/12 (8%) patients with pretreatment WT TP53 (P = 0.03, two-sided Fisher’s exact test). We did not find any significant associations between the acquisition of KRASamp at progression and pretreatment co-alterations in SMAD4, CDKN2A, ARID1A and CHEK2 (Supplementary Table 4), with each of these occurring at frequencies near the expected prevalence in PDAC1.

absence of acquired secondary‑site mutations in cc by

No secondary gene mutations were detected in ctDNA from the 44 patients with PDAC profiled in this study. Of note, Sang et al. recently demonstrated that daraxonrasib treatment-emergent mutations at KRAS Y64 in non-PDAC patients disrupt the affinity of tri-complex formation14. In parallel, we performed a deep mutational scanning (DMS) screen in the Ba/F3 mouse pro-B cell line to evaluate such possible second-site mutations in KRAS that confer resistance to daraxonrasib. This screen identified mutations at multiple KRAS residues that conferred resistance, including at E37, Y64, Y71, L56, T58 and R68, among others; many of these mutations were also identified via base-editing screens in mutant Kras25. Analysis of each of these second-site KRAS alterations in the X-ray crystal structure of the tri-complex containing daraxonrasib bound to KRAS G12C mutant protein and cyclophilin A (CYPA) revealed that most of these changes reside at the inhibitor-binding interface and are predicted to attenuate tri-complex formation by daraxonrasib14 (Extended Data Fig. 3). Daraxonrasib treatment-emergent mutations at Y64 disrupt RAS binding, whereas changes at Y71 enhance RAF-binding as potential resistance mechanisms14. Consistent with Sang et al., showing that BRAF mutations can confer resistance to RAS(ON) multi-selective inhibition14, the observation of RAF gene alterations in PDAC in this study suggests that these acquired mutations may attenuate the activity of daraxonrasib and reactivate downstream RAS pathway signaling. Consistent with the clinical observations described herein, we did not identify second-site mutations in KRAS in any preclinical PDAC model with acquired resistance to long-term RAS(ON) inhibitor treatment (in vitro or in vivo, see below).

modeling acquired resistance to daraxonrasib monotherapy cc by

To study the causal role of each candidate resistance mechanism and to evaluate potential combination approaches to mitigate resistance, we adopted two complementary approaches to develop preclinical models: (1) targeted genetic engineering introducing candidate genomic resistance drivers into treatment-naive models, and (2) isolation of resistant clones from initially treatment‑naive tumor models that outgrew under sustained in vivo daraxonrasib (or a relevant RAS(ON) inhibitor) treatment pressure with subsequent molecular characterization and serial passaging under continued RAS(ON) inhibition in vitro and/or in vivo (Fig. 3, Extended Data Figs. 4 and 5, Supplementary Table 5 and Methods). The majority of key acquired resistance mechanisms identified in the EOT ctDNA samples (for example, mutant KRASamp, RTK alterations and MYCamp) were recapitulated in daraxonrasib-resistant preclinical models derived following prolonged RAS(ON) inhibition.Fig. 3Acquired resistance to daraxonrasib monotherapy in preclinical models.a, Five-day two-dimensional (2D) viability assays were performed using HuP-T3 cells with doxycycline (DOX)-inducible KRASG12R overexpression. One of two independent experiments is shown; technical duplicates were included. Data represent mean ± s.d. b, Five-day 2D viability assays were performed in HuP-T3 cells with DOX-inducible KRASG12R expression and AP009 murine pancreatic organoids engineered to overexpress KrasG12D. RMC-7977 dose–response curves and immunoblots of KRAS, pERK and ERK following KRAS induction are shown. c, HPAC xenograft tumors that regrew after initial response to the RAS(ON) G12D-selective inhibitor RMC-9945 were used to generate resistant cell lines. Sensitivity to daraxonrasib was assessed by 5-day 2D viability assays in parental and resistant HPAC cells. One of two independent experiments is shown; technical duplicates were included, and data represent mean ± s.d. Immunoblot analysis of pERK, ERK and RAS is shown. d, Digital PCR was used to quantify KRASG12D and WT KRAS CN in parental and resistant HPAC cells. Relative CN was normalized to RNase P (mean ± s.e.m., n = 2). e,f, Mice bearing parental (e) or daraxonrasib-resistant Capan-1 (f) CDX tumors were treated with vehicle or daraxonrasib at 25 mg kg−1 orally, once daily (po qd). Tumor volumes were assessed throughout the treatment period (mean ± s.e.m.). Vehicle control and treatment groups were compared by two-way repeated-measures analysis of variance (ANOVA) on the last measurement day of the control group (**P < 0.01). g, Representative images of immunohistochemistry (IHC) staining of HER2 in Capan-1 parental and daraxonrasib-resistant tumors collected at 4 h post-dosing of vehicle control or daraxonrasib on day 7. h,i, Mice bearing parental (h) and resistant HPAF-II (i) CDX tumors were treated with vehicle control or RMC-7977 at 10 mg kg−1 po qd. Tumor volumes were monitored throughout treatment (mean ± s.e.m.) and compared by two-way repeated-measures ANOVA (**P < 0.01; ***P < 0.001). j, HPAF-II parental tumors treated with vehicle control or RMC-7977 and RMC-7977-resistant HPAF-II tumors were collected for immunoblot analysis of the indicated markers. DMSO, dimethylsulfoxide.Source data

ras pathway hyperactivation drives resistance to daraxonrasib cc by

We then tested the hypothesis that acquired mutant KRASamp, the most frequently observed genomic alteration in ctDNA, would result in hyperactivation of RAS pathway flux and thereby cause attenuation of daraxonrasib response in tumors. To model KRAS overexpression resulting from mutant KRASamp, we engineered the human PDAC cell line HuP-T3 (KRASG12R) with a doxycycline-inducible KRASG12R construct and assessed doxycycline concentration-dependent response to daraxonrasib. We also engineered a murine PDAC organoid-derived cell line model AP009 with different levels of cytomegalovirus promoter-driven KrasG12D overexpression. We found that increasing levels of KRASG12R in HuP-T3 cells or of KrasG12D in AP009 proportionally modulated pERK levels and attenuated response to daraxonrasib, resulting in decreased antiproliferative potency and depth of response in vitro (Fig. 3a,b and Extended Data Fig. 5a). We also generated a RAS(ON) multi- and RAS(ON) G12D-selective inhibitor cross-resistant cell line derived from HPAC KRASG12D xenograft tumors26 (Fig. 3d and Extended Data Fig. 5b). Consistent with the in vivo model, this cell line was cross-resistant to daraxonrasib and zoldonrasib (Fig. 3c and Extended Data Fig. 5c) and exhibited acquired KRASG12D amplification resulting in increased KRAS protein levels and proportional modulation of pERK (Fig. 3c), as well as acquired MAP2K1 and PIK3CA mutations. Together, these data link mutant KRAS amplification to increased RAS signaling and attenuated daraxonrasib response. We generated a series of xenograft models of PDAC to interrogate the impact of RAS pathway hyperactivation on daraxonrasib sensitivity in vivo. A serially passaged daraxonrasib-resistant Capan-1 KRASG12V homozygous cell line-derived xenograft (CDX) (Fig. 3e,f) revealed elevated HER2 protein expression localized to the cell membrane (Fig. 3g). This observation is consistent with previous studies reporting the upregulation of ERBB2/HER2 following genetic or pharmacological downregulation of RAS pathway signaling27 and with acquired resistance driven by RTK genomic alterations identified in EOT ctDNA of patients 8, 14, 15 and 21. Recently, we characterized the HPAF-II CDX model that exhibits baseline amplification of KRASG12D (copy number of 6)26. In this model, RMC-7977, a RAS(ON) multi-selective inhibitor preclinical tool representative of daraxonrasib28, treatment drove initial tumor regressions with eventual relapse observed after long-term dosing (Fig. 3h), providing a useful model with baseline KRASamp in which to examine combination therapies. In HPAF-II tumors growing out under constant therapeutic pressure with RAS(ON) inhibition (Fig. 3i) we observed reactivation of RAS signaling, as demonstrated by elevated pERK and pMEK levels (Fig. 3j). We also utilized two relatively insensitive models in which daraxonrasib monotherapy drives an initial response followed by relapse on treatment. The KRASG12D KP-4 CDX model of PDAC (Extended Data Fig. 5d) also harbors MYC amplification and elevated EGFR and c-MET expression, the latter previously reported to hyperactivate RTK signaling via an HGF/c-MET autocrine loop29 (Extended Data Fig. 5e). The KRASG12R mutant PSN-1 xenograft model showed evidence for mutant KRASamp and MYC amplification upon relapse (Extended Data Fig. 5f,g). MYC amplification was observed in ctDNA of patient 1, whereas elevated c-MET expression models genomic alterations in patients 9 and 16. Collectively, this curated series of preclinical models consistently mirrored the findings in the ctDNA of patients with PDAC progressing on daraxonrasib monotherapy. We then used these models to test combination therapeutic strategies aimed at averting or overcoming daraxonrasib resistance.

combination approaches to treat monotherapy resistance to daraxonrasib cc by

Where possible, we tested whether combination regimens could be effective in models with acquired resistance to daraxonrasib as well as in treatment-naive models, the latter to evaluate the potential of the upfront combination to delay or prevent said resistance. All treatment regimens included in this study and described below were tolerated in mice based on body weight assessment (Extended Data Figs. 6 and 7).

orthogonal combinations with daraxonrasib cc by

We observed in clinical samples that mutant KRASamp, as a frequent potential resistance driver, was associated with oncogenic LOF TP53 alterations in pretreatment ctDNA (Fig. 2g). Loss of p53 function and/or KRAS alteration can induce replication stress, creating vulnerability to DNA damage response (DDR) inhibitors30. We therefore evaluated daraxonrasib in combination with inhibitors of the ATR–CHK1–WEE1 axis. We evaluated BBI-355 (a CHK1 inhibitor)31 and azenosertib (a WEE1 inhibitor)32 in the parental and daraxonrasib-resistant KRASG12D; TP53WT HPAC cell line models. Both models were sensitive to single-agent DDR inhibition and combination treatment (Extended Data Fig. 8a–d). Notably, in the resistant cell line with mutant KRASamp antiproliferative effects were driven primarily by DDR inhibition, indicating increased DDR dependence (Extended Data Fig. 8c,d). Daraxonrasib reduced pERK and pCDK1 in both parental and resistant lines, albeit with more rapid restoration of pERK in the resistant model suggesting RAS hyperactivation driven by KRASG12D amplification. Combination treatment increased DDR pathway activation marker γH2AX and induced apoptosis (cleaved caspase-3, PARP) in parental cells, with only modest effects in resistant cells (Extended Data Fig. 8c,d). We then assessed the combination of daraxonrasib and the WEE1 inhibitor azenosertib in vivo in the parental KRASG12D amplified; TP53R273H HPAF-II CDX PDAC tumor model. Azenosertib showed minimal single-agent activity, while daraxonrasib induced transient tumor stasis followed by regrowth; the combination achieved deeper and more durable regressions, though not exceeding optimal-dose daraxonrasib monotherapy (Extended Data Fig. 8e,f). Together, these data support DDR machinery as a therapeutic vulnerability in RAS mutant PDAC. We next explored strategies targeting MYCamp-mediated resistance, observed in patient 1 and previously in the KPC genetically engineered mouse model (GEMM)33. Although prevalent in GEMMs, MYCamp is less frequent in KRAS G12C(OFF) inhibitor-treated PDAC and in this cohort19. We previously discovered that the combination of TEA domain transcription factor and RAS(ON) multi-selective inhibition was effective in the context of MYCamp in vitro33; we now evaluated the combination of daraxonrasib with the pan-TEA domain transcription factor inhibitor IAG933 in MYCamp KP-4 CDX tumors (along with additional candidate resistance drivers). IAG933 as a single agent was inactive, while daraxonrasib induced transient response (Extended Data Fig. 8g). The combination achieved deeper responses but did not prevent on-treatment relapse, suggesting the combinatorial potential of this approach may be limited (at least in this model).

engaging rtks in combination with ras(on) multi-selective inhibition cc by

Based on the prevalence of RTK alterations in our ctDNA analyses and daraxonrasib‑induced RTK upregulation in the Capan‑1 model, we hypothesized that combining RAS(ON) multi-selective inhibition with therapies that leverage RTKs as targets could mitigate resistance and restore tumor sensitivity. Consistent with previous studies with RAS pathway inhibitors27, we observed that daraxonrasib treatment resulted in a concentration-dependent increase in surface HER2 expression, though the magnitude of induction varied between cell lines (Extended Data Fig. 9a,b)34. We hypothesized that the collateral vulnerability of HER2 upregulation at the tumor cell membrane would drive internalization of the antibody–drug conjugate (ADC) trastuzumab deruxtecan (T-DXd)-modified receptor and delivery of a cytotoxic payload, which would synergize with daraxonrasib-mediated RAS pathway inhibition. In the MIA Paca-2 cell line27, daraxonrasib further enhanced surface HER2 levels, and subsequent treatment with T-DXd led to rapid and efficient HER2 internalization, suggesting functional target engagement (Extended Data Fig. 9c). We then assessed the in vivo antitumor activity of daraxonrasib in combination with T-DXd in four xenograft models of PDAC harboring distinct KRAS mutations: MIA Paca-2 (KRASG12C), WU-0022 (KRASG12R), WU-0002 (KRASG12D) and Pa01C (KRASG12D)27. In the first three models, the treatment resulted in combinatorial activity and sustained tumor regressions (Fig. 4a,b and Extended Data Fig. 9d). We did not observe a meaningful combinatorial effect in the Pa01C model (Extended Data Fig. 9e), which shows very low initial levels of surface HER2 with minimal enhancement following daraxonrasib treatment, and no response to T-DXd monotherapy, indicating that the combination may require a threshold level of HER2 expression, robust ADC internalization, and/or baseline sensitivity to topoisomerase inhibition (Extended Data Fig. 9b). We further evaluated this combination in the parental and daraxonrasib-resistant Capan-1 CDX model pair described above, simulating upfront and sequential combination treatment paradigms, respectively. In both models, the combination led to dramatic complete and durable tumor regressions in Capan-1 parental tumors, which were sustained even after treatment cessation (Fig. 4c,d). Consistent with these findings, the combinatorial activity of daraxonrasib and T-DXd was also observed in the high-bar KP-4 CDX model, where the combination achieved more durable tumor responses (Fig. 4e).Fig. 4Combinations of daraxonrasib and RTK targeting agents overcome tumor growth in models with RTK activation.a,b, Mice bearing WU-0002 (a) or WU-0022 patient-derived (b) xenograft (PDX) tumors were treated with vehicle, daraxonrasib at 25 mg kg−1 po for repeating cycles of 5 days on and 2 days off, T-DXd intravenously (i.v.) at 10 mg kg−1 per week, or the combination. Group sizes were: WU-0002, n = 5 for vehicle and combination and n = 6 for daraxonrasib and T-DXd; WU-0022, n = 4 for vehicle, n = 5 for daraxonrasib, n = 6 for T-DXd and n = 7 for combination. Tumor volumes were assessed through the treatment period (mean ± s.e.m.). Vehicle control and treatment groups were compared by two-way repeated-measures ANOVA on the last measurement day of the control group (**P < 0.01, ***P < 0.001). c,d, Mice bearing parental (c) or daraxonrasib-resistant Capan-1 (d) CDX tumors (n = 6 per group) were treated with vehicle, daraxonrasib (25 mg kg−1, po qd), T-DXd (4 mg kg−1, i.v., weekly) or the combination. Waterfall plots show the percentage change in individual tumor volumes after the indicated treatment duration. e, Mice bearing KP-4 CDX tumors (n = 6 per group) were treated with vehicle control, daraxonrasib at 25 mg kg−1 po qd, T-DXd i.v. at 4 mg kg−1 per week, or the combination. # indicates that one mouse was euthanized. f, Mice bearing KP-4 CDX tumors (n = 9 per group) were treated with vehicle, daraxonrasib (25 mg kg−1 po qd), amivantamab (10 mg kg−1, intraperitoneally (i.p.) twice weekly) or the combination. Mean tumor volumes and individual percentage changes across the course of treatment are shown. For all studies, tumor volumes are shown as mean ± s.e.m.; vehicle and treatment groups were compared by two-way repeated-measures ANOVA on the final control-group measurement day (**P < 0.01, ***P < 0.001). biw, twice weekly; R, regression; qw, once weekly.Source data Given the detection of c-MET alterations in the ctDNA of some patients who progressed on daraxonrasib monotherapy and the reciprocal relationship of c-MET and EGFR in the context of KRAS G12C(OFF) inhibitor resistance in PDAC35, we hypothesized that the combined inhibition of c-MET and EGFR may effectively block acquired RTK-mediated daraxonrasib resistance. We therefore modeled the combination of daraxonrasib and amivantamab, a bispecific inhibitory antibody against EGFR and MET36, in the KP-4 CDX model which expresses high levels of c-MET33. Daraxonrasib monotherapy drove initial tumor regressions with tumor outgrowth on treatment observed after 14 days, while the combination of daraxonrasib and amivantamab resulted in a significant improvement in both the depth and duration of tumor responses as compared to either agent alone (Fig. 4f). In sum, these results support the combination of RTK targeting therapeutic modalities such as ADCs and bispecific agents with daraxonrasib to overcome resistance driven by RTK overexpression and activation.

doubling down on ras pathway signaling by combining ras(on) multi- and mutant-se cc by

Next, we hypothesized that daraxonrasib-resistant models with dependency on elevated RAS signaling could benefit from a combination of multi- and mutant-selective RAS(ON) inhibitors, maximizing direct inhibition of RAS signaling through both mutant and WT RAS, analogous to our work demonstrating this combination concept in KRAS G12C‑driven lung cancer models37. In the HPAC parental cell line model, the combination resulted in a strong antiproliferative effect and cell killing in vitro26 (Fig. 5a). In the daraxonrasib (and zoldonrasib)-resistant HPAC line, we observed a combination benefit that was evident but somewhat reduced as compared to that in the parental line, consistent with increased oncogenic flux through the RAS pathway in this context (Fig. 5b). Therefore, we hypothesized that the upfront combination therapy of a RAS(ON) inhibitor doublet would be more effective in forestalling the development of RAS-driven resistance rather than sequential treatment in the context of established monotherapy resistance. We tested the upfront combination therapy in a series of treatment-naive KrasG12D driven KPC/Y GEMM-derived models in vivo38. The RAS(ON) inhibitor doublet had a profound combination effect on antitumor activity, resulting in greater depth of response and significant extension of PFS (as measured by time to tumor doubling from baseline) compared to each monotherapy, with the upfront combination durably preventing the emergence of resistance (Fig. 5c,d and Extended Data Fig. 10a–f). In each of these models, single-agent daraxonrasib and zoldonrasib substantially suppressed RAS pathway signaling as indicated by decreased pERK signal compared with vehicle controls and the RAS(ON) inhibitor doublet induced deeper and more durable reduction of pERK at 4 h and Dusp6 expression at 24 h post-dose compared to single agents (Fig. 5e,f and Extended Data Fig. 10g–i).Fig. 5Doubling down on RAS pathway signaling by combining RAS(ON) multi- and mutant-selective inhibitors.a,b, Five-day 2D viability assays were performed in parental (a) and resistant (b) HPAC cell lines treated with daraxonrasib, zoldonrasib or the combination. Data are shown as heatmaps (% of DMSO control) and dose–response curves. One of two independent experiments is shown; technical duplicates were included, and data represent mean ± s.d. c, Mice bearing KPC/Y GEMM-derived tumors (KPCYc3, n = 10 per group except combination n = 9; KPCYc4, n = 7 per group except combination n = 6; KPCYc5, n = 10 per group) were treated with vehicle, daraxonrasib (25 mg kg−1 po qd), zoldonrasib (100 mg kg−1 po qd) or the combination. Mean tumor volume change from baseline on day 21 is shown (mean ± s.e.m.). d, Combined Kaplan–Meier curves from experiments in c representing PFS, with progression defined as tumor doubling from baseline on treatment (n = 27 mice per treatment group). log-rank test *P = 0.0467, ****P < 0.0001. e,f, Pharmacodynamic effects were assessed in KPCYc3 tumors following a single dose of daraxonrasib, zoldonrasib or the combination. Mean pERK positivity in pan cytokeratin (panCK)-positive tumor cells was quantified by IHC relative to vehicle (mean ± s.e.m., n = 3 mice per group per time point; e) and representative pERK IHC images (f) at 8 h post-dose are shown. g, HPAF-II tumor-bearing mice were treated with vehicle (n = 3), daraxonrasib (25 mg kg−1; n = 7), zoldonrasib (100 mg kg−1; n = 6) or daraxonrasib (10 mg kg−1) plus zoldonrasib (100 mg kg−1; n = 7). Tumor growth and individual tumor responses after 28 days are shown. h, Mice bearing resistant HPAF-II tumors (n = 9 per group) received the same treatments. Tumor growth and individual responses after 24 days are shown. For g and h, data represent mean ± s.e.m.; groups were compared by two-way repeated-measures ANOVA on the final vehicle-group measurement day (*P < 0.05, **P < 0.01, ***P < 0.001).Source data Finally, we examined the zoldonrasib and daraxonrasib combination in the paired parental RAS(ON) multi-selective inhibitor naive and treatment-resistant KRASG12D HPAF-II CDX models. In the parental model, zoldonrasib or daraxonrasib as single agents drove initial tumor regressions following relapse and outgrowth on treatment, while the RAS(ON) inhibitor doublet drove deep and durable tumor regressions (Fig. 5g) and averted resistance development, consistent with the results in the KPC/Y models above. In the corresponding resistant model, the RAS(ON) inhibitor doublet showed combinatorial antitumor activity that was significantly improved as compared to either monotherapy activity (Fig. 5h). Taken together, these results imply that a RAS(ON) inhibitor doublet can be deployed either to forestall or to overcome RAS hyperactivation-driven resistance to daraxonrasib monotherapy.

discussion cc by

The development of direct RAS inhibitors, beginning with KRAS G12C(OFF) targeting compounds and now continued by the advent of RAS(ON) inhibitors beginning with daraxonrasib, has transformed the treatment of RAS-driven cancers18,39. The encouraging monotherapy activity for daraxonrasib in patients with PDAC following standard-of-care chemotherapy12 was recently confirmed by results from the RASolute 302 trial13. Timely interrogation of monotherapy resistance mechanisms informs rational combination strategies co-targeting key drivers and/or taking advantage of treatment-induced vulnerabilities18,40. Acquired genomic alterations detected in this study converged on RAS pathway hyperactivation, indicating increased oncogenic flux as a dominant escape mechanism. Of note, resistance mechanisms appeared more frequently at higher daraxonrasib doses, suggesting a dose-dependent bottleneck in RAS-driven tumors41. Acquired secondary KRAS mutations were notably absent in this PDAC resistance study, albeit a parallel study by Sang et al. in non-PDAC tumors found secondary mutations in 2 of 40 patients14, distinct from secondary KRAS mutations that abrogate the binding of KRAS G12C(OFF) inhibitors18; however, acquired resistance to KRAS G12C(OFF) inhibitors in PDAC altogether shows fewer secondary KRAS mutations compared with those in NSCLC and CRC19,39. Notably, no compensatory oncogenic alterations in NRAS or HRAS following daraxonrasib treatment were identified, consistent with the broad and concomitant inhibition of active state mutant and WT canonical RAS proteins by daraxonrasib28. These observations support a resistance pattern dominated by pathway reactivation rather than on-target KRAS second-site mutations. We demonstrate combinatorial antitumor activity for RTK-targeted biologics combined with daraxonrasib in PDAC models. In the case of T-DXd, RAS inhibition-driven HER2 upregulation may enhance delivery of a cytotoxic payload19,20. We also demonstrated that the RAS(ON) inhibitor doublet of daraxonrasib and zoldonrasib (under clinical evaluation; NCT06040541) was well tolerated and drove combinatorial pathway suppression sufficient to prevent resistance in a series of naive PDAC models and to overcome acquired RAS(ON) inhibitor monotherapy resistance. These results support clinical evaluation of this RAS(ON) inhibitor doublet in KRAS G12D-mutant PDAC. Collectively, these data nominate two clinically actionable strategies in PDAC: (1) combinations with RTK-directed biologics (for example, T-DXd and amivantamab) and (2) intensified direct RAS blockade with a RAS(ON) inhibitor doublet (daraxonrasib + zoldonrasib). Additionally, orthogonal combinations of daraxonrasib and DDR inhibitors in mutant KRASamp models enhanced antitumor activity albeit the feasibility of such targeted combination approaches will rely on the development of agents targeting the DDR axis with favorable safety profiles. In tumors with TP53 inactivation, it is possible that chemotherapy-induced DNA damage could mimic replicative stress and functionally overlap with DDR-targeting approaches; combinations with chemotherapy are an active area of preclinical and clinical investigation (see RASolute 303 trial in the frontline setting; NCT07491445). While this study presents the first comprehensive evaluation of clinical and preclinical genomic mechanisms of acquired resistance to monotherapy daraxonrasib in PDAC, we only describe genomic features detectable by paired ctDNA analyses in patients with sufficient tumor shedding into circulation. The identification of candidate genomic resistance drivers in ~60% of patients examined indicates that ~40% of PDAC tumors that relapse on daraxonrasib treatment potentially acquire resistance via nongenomic means. Moreover, multiple preclinical studies have demonstrated that genetic and nongenetic mechanisms can coevolve within the same tumors19,40. Additionally, the detection of alterations in 800 genes in the DNA sequencing panel selected for this study cannot rule out other resistance mutations, including in PPIA, the gene encoding cyclophilin A, required for RAS(ON) inhibitor activity42. Future studies with larger cohorts and paired tumor biopsies will be needed to validate these findings and define genomic and nongenomic resistance mechanisms. Furthermore, a rigorous evaluation of whether the genomic resistance mechanisms reported in this study are acquired de novo (or pre-existing at baseline) will require data from a much larger patient cohort such as that from RASolute 302 mentioned herein13. In closing, we report the first characterization of treatment-emergent genomic alterations following daraxonrasib response and relapse in patients with PDAC, demonstrating that acquired resistance mechanisms converge on RAS pathway signaling hyperactivation and recapitulate patterns of acquired resistance for inhibitors of BRAF, MEK and RTKs18. These results reinforce that RAS/MAPK pathway-dependent cancers are oncogene-addicted and that the evolution of therapeutic resistance often selects for reactivation of driver oncogene signaling. Our insights guided the evaluation of rational combinations to avert and/or overcome daraxonrasib resistance in a series of preclinical models laying a translational foundation for the evaluation of promising combinations in PDAC aimed at increased and durable therapeutic benefit following on recent encouraging monotherapy results in the randomized phase 3 clinical trial RASolute 302 (ref. 13).

ethics statement cc by

The trial was conducted in accordance with the Good Clinical Practice guidelines of the International Council for Harmonization and the principles of the Declaration of Helsinki. The protocol and amendments were approved by the institutional review board at each participating site and regulatory authorities of participating countries. The institutional review boards that approved this study were Advarra, Dana-Farber Cancer Institute, WCG, Salus, John Hopkins Medicine, Memorial Sloan Kettering, Mary Crowley Medical Research Center, UCLA Office of Human Research Protection Program, NYU Grossman School of Medicine and Columbia University Irving Medical Center. All patients provided written informed consent. The trial was designed by Revolution Medicines (study sponsor). The data were collected by investigators and analyzed by statisticians employed by Revolution Medicines. All animal studies were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee at each study site and were performed in compliance with all applicable institutional and national guidelines for the care and use of laboratory animals.

statistics and reproducibility cc by

The clinical study NCT05379985 was conducted as a single-arm phase 1/2 trial; therefore, participants were not randomized. Investigators were not blinded to allocation during experiments and outcome assessment unless otherwise indicated. For preclinical studies, power analyses were used to drive sample size determination as appropriate. Randomization was used in preclinical in vivo studies. Sample sizes for preclinical experiments were determined based on study objectives, previous experience with similar models and practical considerations and were not predetermined using formal statistical methods. Unless otherwise specified, no data were excluded from the analyses.

circulating tumor dna assessment in patients with acquired resistance cc by

Patients with RAS mutant metastatic PDAC treated at 160 mg to 300 mg daraxonrasib as second-line or later therapy and disease progression were assessed for acquired resistance. Acquired resistance was defined as patients with a PFS of more than 3 months (including SD or PR or CR) and subsequent progression, in line with previous criteria applied for patients with acquired resistance19. Paired pretreatment and EOT at time of progression ctDNA assessment using the Guardant Health Infinity next-generation sequencing panel with 800+ genes was performed in 44 patients meeting these criteria. Tissue sequencing was also performed in one patient with a paired pretreatment and EOT tissue biopsy. An acquired alteration in EOT ctDNA was defined as a known or likely oncogenic mutation annotated by OncoKB and not detected in pretreatment ctDNA. To account for false-negative acquired variants due to low shedding in pretreatment ctDNA, pretreatment samples were required to have at least one somatic variant detected to be considered for paired analysis. Acquired oncogenic variants were examined in RAS, MAPK, RTK, PI3K/AKT and MYC pathways based on definitions by Vega et al.43. CN alterations were defined as any CN <2 copies for gene deletions and >2 copies for gene amplifications; patients without reported CN alterations in a gene thus had two copies (normal diploid state) detected in total cell-free DNA. Supplementary Table 4 lists all known or likely oncogenic alterations detected in pretreatment ctDNA.

animal studies using xenograft and syngeneic tumor models cc by

Studies were conducted at Revolution Medicines, Washington University School of Medicine and at contract research organizations, including Pharmaron and WuXi AppTec. No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications33. All CDX/PDX mouse studies and procedures related to animal handling, care, and treatment complied with all applicable regulations and guidelines of the Institutional Animal Care and Use Committee at each facility with approvals. The maximum tumor burden permitted was 2,000–2,300 mm3. Tumor growth was monitored regularly, and animals were killed according to the study protocol. In a small number of cases, due to rapid tumor growth between scheduled measurements, tumor volumes somewhat exceeded 2,000–2,300 mm3 just before killing. For studies with PSN-1 and Capan-1 CDX tumors, female BALB/c nude mice 6–8 weeks old were used. The animals were purchased from ZheJiang Vital River Laboratory Animal Technology Co. and Shanghai Sino-British SIPPR/BK Laboratory Animal Co. For studies with HPAC CDX tumors, female athymic nude mice aged 6 weeks were used. For study with HPAF-II CDX tumors, female SCID Beige mice 6–8 weeks old from Beijing Vital River Laboratory Animal Technology Co. were used. For study with KP-4 CDX tumors, female BALB/c nude mice 6–8 weeks old purchased from Shanghai Sino-British SIPPR/BK Lab Animal Co. were used. For studies with KPC/Y GEMM-derived models, female C57BL/6J mice 6–8 weeks old were used obtained from The Jackson Laboratory. For study with MIA Paca-2 CDX tumors, 8–12-week-old female NSG mice were purchased from The Jackson Laboratory.

in vivo model development cc by

Capan-1 tumor cells (ATCC-HTB-79) were maintained in vitro as a monolayer culture in IMDM supplemented with 20% heat-inactivated fetal bovine serum, at 37 °C in an atmosphere of 5% CO2 in air and routinely subcultured twice weekly by trypsin–EDTA treatment. Cells in the exponential growth phase were collected and counted for tumor inoculation. Each mouse was inoculated subcutaneously on the right flank with 5 × 106 Capan-1 tumor cells in a 0.2-ml PBS/Matrigel mixture (1:1 ratio) to initiate tumor development. For efficacy studies, treatments were initiated when the average tumor volume reached 146 mm3 (Fig. 3e) or 167 mm3 (Fig. 4c) Animals were assigned to groups using Studylog-based randomization software which performed stratified randomization based on tumor volumes. For the pharmacodynamic (PD) study, treatments were initiated when the average tumor volume reached 400–600 mm3. The HPAF-II tumor cells (ATCC, CRL-1997) were maintained in vitro as a monolayer culture in EMEM supplemented with 10% fetal bovine serum and 1% antibiotic–antimycotic at 37 °C in an atmosphere of 5% CO2 in air and routinely subcultured twice weekly. Cells growing in an exponential growth phase were collected, counted for tumor inoculation, and inoculated subcutaneously at the right upper flank with 3 × 106 tumor cells in 0.2 ml of PBS with Matrigel (1:1) per animal. Treatments were initiated at the time of randomization when the mean tumor volume reached 155 mm3. The Capan-1 tumor model resistant to daraxonrasib (25 mg kg−1 po qd) was established through serial passaging with daraxonrasib treatment in vivo (Extended Data Fig. 3). For each round of tumor passaging, recipient mice were dosed with daraxonrasib at 25 mg kg−1 for 3 days before the implantation of tumor fragments, followed by continued daily dosing of daraxonrasib until tumor volumes >600 mm3 at which time these were collected and re-passaged. Daraxonrasib-resistant tumors from passage 3 (P3) and passage 5 (P5) were used for PD and efficacy studies, respectively. Each mouse was inoculated subcutaneously in the right upper flank with daraxonrasib-resistant Capan-1 tumor slices (~30 mm3) to initiate tumor development. For the PD study, treatments were initiated when the average tumor volume reached 400–600 mm3. For the efficacy study, treatments were initiated when the average tumor volume reached 163–164 mm3. The animals were assigned to groups (n = 6) using a Studylog-based randomization software as above. Similarly, the HPAF-II tumor model resistant to RMC-7977 (10 mg kg−1 po qd) was established through serial passaged with RMC-7977 treatment in vivo (Extended Data Fig. 3). HPAC tumor cells (ATCC CRL-2119) were maintained as a monolayer culture in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37 °C in a humidified atmosphere containing 5% CO2. The culture medium was refreshed every 2–3 days, and cells were routinely subcultured at 80–90% confluence using trypsin–EDTA, with passages limited to a maximum of 4–5. Cells in the exponential growth phase were collected, counted for tumor inoculation and each mouse was inoculated subcutaneously in the right flank with 1 × 107 cells suspended in 100 µl of a 1:1 mixture of RPMI 1640 and Matrigel to establish tumors. Treatment commenced upon randomization once the mean tumor volume reached 139 mm3. For in vivo RAS(ON) inhibitor-resistant model derivation, HPAC xenograft tumors that emerged after initial tumor regressions in response to RMC-9945 (the RAS(ON) G12D-selective inhibitor tool representative of the investigational agent zoldonrasib26), wherein we observed amplification of mutant KRASG12D in vivo (Fig. 3d and Extended Data Fig. 5b). We then established a HPAC tumor-derived cell line from these RMC-9945-resistant tumors, which was maintained under constant RAS(ON) inhibitor pressure. KPC/Y GEMM-derived tumor cells 6694c2, 2838c3, 6499c4 and 6419c5, were obtained from Kerafast and are referred to as KPCYc2, KPCYC3, KPCYc4 and KPCYc5 thereafter. Cells were maintained as a monolayer culture in DMEM supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37 °C in a humidified atmosphere containing 5% CO2. The culture medium was refreshed every 2–3 days, and cells were routinely subcultured at 80% confluence using trypsin–EDTA, with passages limited to a maximum of 3–4. Cells in the exponential growth phase were collected, counted for tumor inoculation and each mouse was inoculated subcutaneously in the right flank with 3 × 105 cells resuspended in serum-free DMEM. For efficacy studies, treatment commenced upon randomization once the mean tumor volume reached 100–120 mm3. For studies performed in MIA Paca-2 (ATCC CRL-1420) CDX tumors, WU-0002, WU-0022 PDX tumors, 2–5 million PDAC cells or 5 x 5 x 5-mm chunks of cryopreserved PDXs were mixed 1:1 (v/v) with Matrigel matrix (Corning) and inoculated into both flanks of 8–12-week-old female NSG mice (purchased from The Jackson Laboratory) by needle injection or small incision subcutaneously at the flanks of each mouse. Treatments were initiated when the tumors reached ~100 mm3 in volume. KP-4 tumor cells (RIKEN-RCB1005) were maintained in vitro as a monolayer culture in IMDM supplemented with 20% heat-inactivated fetal bovine serum at 37 °C in an atmosphere of 5% CO2 in air and routinely subcultured twice weekly by trypsin–EDTA treatment. The cells growing in an exponential growth phase were collected, counted for tumor inoculation and each mouse will be inoculated subcutaneously at the right upper flank with KP-4 tumor cells (5 × 106) in 0.2 ml of PBS with Matrigel (1:1) for tumor development. Treatments were initiated at the time of randomization when the mean tumor volume reached 141 mm3 (Fig. 4e) or 160 mm3 (Fig. 4f). The PSN-1 tumor cells (ATCC CRL-3211) were cultured as a monolayer culture in RPMI 1640 medium supplemented with 10% fetal bovine serum at 37 °C in an atmosphere of 5% CO2 in air. The medium was renewed every 2–3 days and tumor cells were routinely subcultured at a confluence of 80–90% by trypsin–EDTA, not exceeding 4–5 passages. Cells in the exponential growth phase were collected, counted for tumor inoculation and each mouse was inoculated subcutaneously on the right flank with 5 × 106 PSN-1 tumor cells in 100 µl of an RPMI 1640/Matrigel mixture (1:1 ratio) to initiate tumor development. Treatments were initiated at the time of randomization when the mean tumor volume reached 171 mm3.

in vivo treatment cc by

For tumor growth evaluation studies, tumor-bearing animals were randomized and grouped for treatments (n = 4–10 per group). The vehicle, daraxonrasib, zoldonrasib, RMC-7977, RMC-9945, azenosertib or IAG933 at indicated doses were administered daily or at indicated dosing schedule via oral gavage with a 10 ml kg−1 dosing volume. Daraxonrasib (25 mg kg−1 po qd) combined with zoldonrasib (100 mg kg−1 po qd) was well tolerated in syngeneic models using immunocompetent C57BL/6J mice; however, this same dosing regimen was not tolerated in xenograft models in immunodeficient mice and a reduced daraxonrasib dose (10 mg kg−1 po qd) was used in combination with zoldonrasib (100 mg kg−1 po qd) to ensure tolerability. T-DXd was purchased from Gruechem and dosed at 4 mg kg−1 via i.v. injection weekly for studies performed in Capan-1 and daraxonrasib-resistant Capan-1 tumors. DS-8201a used in studies of MIA Paca-2, WU-0002, WU-0022 and Pa01C tumors was purchased from the Washington University Siteman Cancer Center Pharmacy and approved for use under Institutional Biosafety Committee protocol 16701 v.3 and dosed at 10 mg kg−1 via i.v. injection weekly. Azenosertib and IAG933 were purchased from Gruechem. Animals were euthanized as previously described8. Snap-frozen and formalin-fixed samples were collected following the same protocols as previously described28,44. RMC-9945, zoldonrasib, daraxonrasib, azenosertib, IAG933 and the combination of daraxonrasib plus azenosertib or IAG933 were formulated in 10:20:10:60 (%v/v/v/v) (DMSO:PEG400:Solutol HS15:water). T-DXd was formulated in sterile saline. For studies performed in MIA Paca-2, WU-0002, WU-0022 and Pa01C tumors, daraxonrasib was formulated in vehicle consisting of 5% DMSO (CAS no. 67-68-5; Fisher Scientific).

in vivo study data analysis cc by

In tumor volume plots, the average tumor volume of each group in CDX models was plotted post-implantation, while the x axis started on the first day of treatment in daraxonrasib-resistant Capan-1, KPC, WU-0002, WU-0022, Pa01C and MIA Paca-2 tumor models. Control and treatment groups were compared by two-way repeated-measures ANOVA (Dunnett’s multiple comparisons test) on the last measurement day of the control group. Tumor diameter was measured in two dimensions using a digital caliper and mice in the study were weighed twice per week. The tumor volume in mm3 was calculated based on the formula: volume = (length × (width)2)/2. An individual tumor was deemed to have regressed (R) if it showed >10% reduction in tumor volume from the initial volume and deemed to have CR if it showed ≥80% reduction in tumor volume from the initial volume on a given day. The percentage change in body weight for each animal on a given day was determined as ((body weight on test article administration end date/body weight on test article administration start date) − 1) × 100.

in vitro resistant cell line establishment and characterization cc by

HPAC CDX tumors relapsed on RMC-9945 (RAS(ON) G12D-selective inhibitor) were collected and dissociated using the Tumor Dissociation kit (human, cat. no. 130-095-929) in conjunction with a gentleMACS Octo Dissociator to isolate tumor cells. These isolated tumor cells were cultured in vitro under inhibitor pressure until they demonstrated sustained growth in the presence of 100 nM RMC-9945 (Extended Data Fig. 3 provides a schematic illustration). The resistant cells were characterized to confirm cell identity, purity, and KRAS G12D/WT CN via digital PCR, along with whole-exome sequencing with ~100× coverage at Azenta.

droplet digital pcr and digital pcr cc by

Genomic DNA (gDNA) was extracted from at least 20 mg of tumor tissue using a QIAamp DNA Mini kit (QIAGEN, cat. no. 51304). Then, 20 ng of gDNA was included for digital PCR (dPCR) using the Naica system multiplex digital PCR (Stilla) or Thermo Fisher QuantStudio Absolute Q Digital PCR system following the manufacturer’s protocol. Probes and primers for the following genes were included in the multiplexed droplet dPCR of PSN-1 end-of-study tumors: KRASWT (dPCR Mutation Detection Assay KRAS WT for p.G12C, Human (Apexbio, AA100902-WT)), KRASG12R (dPCR Mutation Detection Assay KRAS Mutant for p.G12R, Human (Apexbio, AA100904-MU)) and RPP30 (RPP30 probe, 5′-Cy5-TTCTGACCTGAAGGCTCTGCGC-3′; RPP30 primer F, 5′-TCAGCATGGCGGTGTTT-3′; RPP30 primer R, 5′-TGGCATGAGGTTGGCCA-3′); MYC (MYC probe FAM-TGCTGCTGCTGCTGGTAGAAGTTC; MYC primer F, 5′-GGTGCAGCCGTATTTCTACTG -3′; MYC primer R, 5′-AGCAGCTCGAATTTCTTCCAGA-3′). The results were normalized to RPP30 for CN assessment. For the dPCR analysis of HPAC resistant cells, DNA was extracted from ~2 million cells using the DNeasy Blood & Tissue kit (QIAGEN, cat. no. 69504). Then, 9 ng of gDNA was included for dPCR using the Thermo Fisher QuantStudio Absolute Q Digital PCR system following the manufacturer’s protocol. The following kits were used for KRAS WT, KRAS G12D and MYC detection and CN evaluation: Absolute Q Liquid Biopsy dPCR Assays, (Thermo Fisher Scientific, A53732, assay ID DGPRJZK), RNaseP (Thermo Fisher Scientific, 4485714) and MYC (Thermo Fisher Scientific, assay ID Hs00292858_cn). The results were normalized to RNaseP for CN assessment.

immunohistochemistry cc by

Tumor fragments collected from mice after the indicated treatments were fixed in 10% neutral buffered formalin at room temperature for up to 24 h, stored at 70% ethanol and then processed into paraffin blocks. All staining was performed on 4-µm tissue sections. For IHC staining of HER2 in tumor tissues collected from Capan-1 and daraxonrasib-resistant Capan-1 model, formalin-fixed paraffin-embedded (FFPE) sections were stained on a BOND-III Automated IHC Stainer using the manufacturer’s recommended settings. Anti-EpCAM rabbit monoclonal antibody (Cell Signaling Technology, cat. no. 14452, clone: D9S3P) was used at 1:400 dilution with EDTA-based pH 9 epitope retrieval solution. Anti-HER2 rabbit monoclonal antibody (Gene Tech, cat. no. GT224507) was used with EDTA-based pH 9 epitope retrieval solution. A BOND IHC Polymer detection kit (Leica, DS9800) was used to detect rabbit primary antibodies. DAB-stained slides were scanned and digitized with a 3DHISTECH Panoramic MIDI whole-slide scanner at ×200 magnification. To stain tissues from KPCY allograft models, a Biocare IntelliPATH automation system was used. Anti-phospho-ERK1/2 rabbit monoclonal antibody (Cell Signaling Technology cat. no. 4370, clone D13.14.4E) was used at 1:200 dilution and anti-panCK rabbit monoclonal antibody (Abcam, cat. no. ab308262, clone EPR28285-45) was used at 1:5,000 dilution with citrate-based pH 6.2 Heat-Induced Epitope Retrieval (Biocare, cat. no. DV2004). Primary antibodies were detected with the MACH4-HRP-polymer Detection System (Biocare, MRH534). Stained slides were scanned and digitized with a TissueScope LE (Huron Digital Pathology) whole-slide scanner at ×20 magnification. Image analysis was performed using HALO software from Indica Labs. For IHC staining of RAS in HPAC control tumors and tumors relapsed on RMC-9945 treatment, FFPE sections were stained on the Biocare IntelliPATH automated staining platform using the manufacturer’s recommended settings. Anti-human Ras rabbit monoclonal antibody (Cell Signaling, cat. no. 67648, clone E8N8L, lot no. 2, 44 μg ml−1) was used at 1:400 dilution (0.11 μg ml−1) with citrate-based pH 6.2 Heat-Induced Epitope Retrieval (Biocare, cat. no. DV2004); an isotype control (rabbit IgG) was used under the same conditions. Stained slides were digitized with a Leica ScanScope AT whole-slide scanner at ×200 magnification.

whole-slide image analysis cc by

panCK staining was used to identify epithelial cells. First, the area to be analyzed was delineated, excluding necrotic regions. The random forest Tissue Classifier from the HALO Image Analysis package was used to identify the tumor compartment. Three classes were created: glass, tumor (panCK-positive) and stroma. The tumor class mark was then copied onto the serial sections stained with markers to perform the image analysis on only the tumor compartment. Quantification of pERK was performed with the HALO Image Analysis software from Indica Labs using the Area Quantification module. The analysis was performed only on the tumor compartment copied from the panCK slide. The software was tuned to detect positive DAB staining (brown color). The percentage of area positivity was chosen to represent the results (area positive for brown/total area) and subjected to statistical analysis using GraphPad Prism (Dunnett multiple comparisons test).

western blotting cc by

Tumor tissues were dissected, cut into 35–50-mg fragments, snap frozen in liquid nitrogen and stored at −80 °C. Frozen tumor fragments were homogenized in a tissue grinder (Scientz, Scientz-48) with NP-40 cell lysis buffer (Invitrogen, FNN0021). Cells cultured in vitro were washed twice with ice-cold PBS and lysed with MSD Tris Lysis Buffer (MSD, R60TX-2), and scraped and collected before centrifugation. All lysis buffers were supplemented with protease and phosphatase inhibitors (Thermo, 78444). Lysates were centrifuged at 14,000 × rpm for 10 min at 4 °C. The protein concentrations in supernatant were measured by PierceTM BCA Protein Assay kit (Thermo, 23225). Then all the lysates with equal quantities of protein were denatured at 95 °C for 10 min after adding LDS sample buffer (Thermo, NP0008) and a reducing agent (Thermo, NP0009). Samples were resolved on 4–12% Bis-Tris gels (Thermo, WG1403BOX) and western blotting was performed following the standard protocol. The following primary antibodies were used: anti-KRAS (Santa-Cruz, sc-30), anti-pan RAS(Abcam, ab108602), anti-Ras (G12D mutant) (Thermo Fisher Scientific, MA5-36256), anti-HER2 (Cell Signaling Technology, 2242), anti-phospho-p44/42 MAPK (ERK1/2) T202/Y204 (Cell Signaling Technology, 4370), anti-p44/42 (ERK1/2) (Cell Signaling Technology, 4696 or 4695), anti-phospho-MEK1/2 Ser217/221 (Cell Signaling Technology, 9154), anti-MEK1/2 (Cell Signaling Technology, 9126), anti-Flag (Abcam, ab205606), anti-Vinculin (Cell Signaling Technology, 13901), anti-β-actin (Cell Signaling Technology, 4967) and anti-β-tubulin (Cell Signaling Technology, 86298). IRDye 800CW goat anti-mouse IgG (H + L) (Licor, 926-32210) and goat anti-rabbit IgG (H + L) HRP-conjugated (Thermo Fisher Scientific, 31462) were used as secondary antibodies.

2d cellular viability assays cc by

Cells were plated in tissue culture-treated 384-well plates and incubated overnight. The following day, ten serial dilutions of compound were prepared in the growth medium. Cells were treated with compound or DMSO (0.1% v/v) for 5 days (T5). Luminescence was detected before treatment (T0) and after treatment (T5) using the Envision Plate Reader (PerkinElmer). Then, 25 μl of CellTiter-Glo 2.0 reagent (G9243, Promega) was added to each well. Plates were shaken at 450 rpm for 2 min at room temperature and then incubated for an additional 10 min without shaking at room temperature and protected from light. Luminescence signal from the CellTiter-Glo 2.0 reagent was normalized to DMSO treated wells (% DMSO = (lumtreated(T5)-lumDMSO (T0)/(lumDMSO (T5)-lumDMSO (T0) × 100). Data were plotted as a function of inhibitor concentration, and a four-parameter sigmoidal concentration response model was fitted.

cell line engineering cc by

KRASG12R overexpression cell line experiments were performed at Pharmaron. The plasmid encoding Tet-On 3G and TRE3GS-inducible Flag-tagged KRASG12R was synthesized and packaged into lentivirus at Pharmaron. Lentivirus transductions were performed with addition of Polybrene (10 µg ml−1). Lentivirus particles were added to HuP-T3 cells and incubated for 24 h after which lentivirus particles were added again to improve transduction efficiency. Then, puromycin (2 µg ml−1) was added in the culture medium to maintain selective pressure. Expression of the transgene was induced by adding doxycycline (0.1–1 µg ml−1) for at least 24 h. AP009 KRASG12D overexpression cell line: AP009 KrasG12D/wt;Trp53null/null cell line was generated from a LSL-KrasG12D/wt mouse and maintained in adherent culture as described previously45. AP009c2 clonal cell line was isolated from parental AP009 cells with single-cell sorted by FACS. Expanded clones were confirmed with KrasG12D/wt;Trp53null/null genotype by sequencing. To generate clonal cell lines with mouse Kras overexpression, AP009c2 cells were transduced with lentivirus expressing mutant Kras G12D cDNA at optimized multiplicity of infection. On the third day after lentiviral infection cells were single-cell sorted by FACS into four 96 well plates. Expanded clones were verified by PCR and sequenced for mutant KRASG12D transgenes. Positive clones with mutant KRASG12D overexpression were evaluated by quantitative PCR using forward primer 5′-GGACTCTGAAGATGTGCCTATGGT-3′, reverse primer 5′-GCCTGTTTCGTGTCTACTGTTC-3′ and probe 5′-CCTGGTAGGGAATAAGTGTGATTTGCC-3′. The open reading frame of mouse Kras cDNA (NM_021284.7) with KRASG12D mutation was cloned into the custom lentiviral cDNA expression plasmid pRR-mCMV-cDNA (Cellecta). Lentivirus was generated by packaging plasmids into VSV-G pseudotyped viral particles using Cellecta’s Ready-to-use Lentiviral Packaging Plasmid Mix (cat. no. CPCP-K2A). The titer of lentiviral packaged constructs provided by Cellecta is functionally determined by transduction of 293T cells.

flow cytometry cc by

To assess surface HER2 expression, cells were stained with 4 µg ml−1 Alexa Fluor 647-conjugated anti-human HER2 antibody (BioLegend, cat. no. 324412) and analyzed using FACScan 5. Data were processed and quantified using FlowJo software (BD Biosciences).

bioinformatics analysis (preclinical studies) cc by

Whole-exome sequencing analysis was performed to ascertain gene mutations. For parental cell line-derived xenograft models, gene mutation data were obtained from the 23Q4 release of the DepMap Data Portal. DNA mutation calling was accomplished with TNSeq using the hg38 version of the human genome.

deep mutational scanning screen cc by

DMS data were processed as previously indicated. In brief, reads were aligned to the KRASG12C reference sequence and collapsed to reads per amino acid at each position and for each amino acid substitution included in the library. Substitution not included in the library and likely PCR artifacts were removed and not considered for downstream analyses. The abundance of each allelic variant was calculated as the fraction of its read compared to the total library-wide reads in each sample independently. Variant abundances were defined as the log2 fold change (FC) between 6-day-treated time points (RMC-7977 or DMSO) and the initial variants representation at day 0 (ETP) and were computed by applying a moderated t-test as implemented in the R package limma (v.3.54.2). We defined a ‘RASi resistance’ and a fitness threshold as the average log2FC + 2 × s.d. in the RMC-7977-day 6 versus ETP and DMSO day 6 versus ETP comparisons, respectively. Bulk-RNA sequencing analyses were performed to ascertain gene expression. Transcript-level quantification was performed using Salmon (v.1.0.8) in quasi-mapping mode. Reference transcriptome used for quantification was derived from the mouse genome (GRCm38). The transcript-level counts generated by Salmon were imported into R, and genes with low expression across the samples were filtered out before downstream analysis. Normalization of gene-level counts was performed using the trimmed mean of M-values method, which was used to calculate the log-transformed counts per million values using the CPM function from the edgeR package (v.4.2.2).

mapping phenotypic data onto protein structures cc by

Functional data from the DMS screen was mapped onto the structure of KRASG12C bound to GNP and in complex with RMC-7977 and PPIA (PDB: 8BTK). Graphical display was carried out using the ‘define by attribute’ and ‘render attribute’ functions implemented in Chimera X. For each residue of KRASG12C, the average log2FC of all substitutions affecting the same residue.

fish cc by

Fluorescent in situ hybridization was performed on either methanol acetic acid fixed in vitro cells or FFPE tumor sections using dual labeled KRAS/centromere 12 probes from Empire Genomics (KRAS-CON12-REGR). Metaphase cells were collected from cell lines treated with 0.2 μg ml−1 Colcemid (GIBCO cat. no. 15210-040) for 4 h, trypsinized and treated with 0.075 M KCl for 20 min at 37 °C before being fixed three times in 75% methanol/25% acetic acid. Fixed cells were stored at −20 °C and then metaphase spreads were prepared on glass slides and air dried. For hybridization, slides were dehydrated in a series of ethanol (70%, 85% and 100%) for 2 min each and air dried. Probes were mixed with hybridization buffer and applied to fixed metaphase slides, covered with a 22 × 22 mm cover-glass and sealed with Cytobond (SciGene cat. no. 2020-00-1). Slides were denatured at 73 °C for 3 min and hybridized overnight at 37 °C in a Cytobrite Plus slide incubation system (SciGene) and then washed with 0.4× SSC (saline sodium citrate) + 0.3% IGEPAL, pH 7 buffer (SciGene) at 73 °C for 2 min followed by another washing with 2× SSC + 0.1% IGEPAL (SciGene) at room temperature for 2 min. Slides were quickly rinsed in water and then air dried. Cells were stained with 4,6-diamidino-2-phenylindole (DAPI) with antifade and imaged with a ×40 immersion objective on a Leica DMi8 microscope. FFPE tumor sections were dewaxed and then treated with the Abbott FFPE Pretreatment kit to prepare the fixed DNA for hybridization. Sections were incubated in 0.2 N HCl for 20 min at room temperature, followed by incubation in sodium thiocyanate for 30 min at 80 °C and digestion with protease for 30 min at 37 °C. In situ hybridizations for FFPE sections were as described for in vitro cells, except that sections were denatured at 78 °C for 7 min.

reporting summary cc by

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

online content cc by

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41591-026-04537-w.

supplementary information cc by

Reporting Summary Peer Review File Supplementary Tables 1–5Supplementary Table 1. Acquired genomic alterations detected in paired baseline and EOT ctDNA profiled from 44 patients treated with 160–300-mg doses of daily daraxonrasib. Genomic alterations include short variants and copy number variants and changes in ctDNA RAS VAF detected at pretreatment (Pre), early on treatment (either cycle 2 day 1 or cycle 3 day 1) (On), and EOT (End) with daraxonrasib. Supplementary Table 2. Acquired genomic alterations detected in paired baseline and EOT ctDNA profiled from 13 patients treated with 120 mg or lower doses of daily daraxonrasib. Genomic alterations include short variants and copy number variants and changes in ctDNA RAS VAF detected at pretreatment (Pre), early on treatment (either cycle 2 day 1 or cycle 3 day 1) (On), and EOT (End) with daraxonrasib. Supplementary Table 3. Summary of 16 patients with acquired KRAS amplification. *Mutant KRAS allele amplification expected to increase ratio of mutant KRAS VAF relative to other tumor mutation VAF. Method is exploratory, and allele amplification is not confirmed. Values above 1 are putative amplification of RAS mutant allele and below 1 are putative amplification of RAS WT. VAFs not adjusted for CN. #Amplification copy number is observed copies in total cell-free DNA and may underestimate tumor copy number. Supplementary Table 4. Baseline genomic alterations from 57 patients treated with daily dose (20–300 mg) of daraxonrasib. Genomic alterations (short variants and copy number variants) in pretreatment ctDNA from 57 patients profiled in this study. Supplementary Table 5. Summary of preclinical models in this study.

source data cc by

Source Data Fig. 1Statistical source data. Source Data Fig. 2Statistical source data and CT scan images. Source Data Fig. 3Statistical source data, unprocessed blots. Source Data Fig. 4Inhibitor response source data. Source Data Fig. 5Inhibitor response source data. Source Data Extended Data Fig. 1Statistical source data for changes in ctDNA. Source Data Extended Data Fig. 2CT scan images. Source Data Extended Data Fig. 3DMS heatmap, scatter-plot, residue Fisher’s test, structural mapping. Source Data Extended Data Fig. 5Statistical source data. Source Data Extended Data Fig. 6Source data of mice body weight. Source Data Extended Data Fig. 7Source data of mice body weight. Source Data Extended Data Fig. 8Statistical source data for matrix combo response, unprocessed blots. Source Data Extended Data Fig. 9Statistical source data. Statistical source data.