RAS(Q61R/K) cells are insensitive to combined SHP2i+MEKi treatment due to feedback activation of RAS(Q61R/K) and induction of a SHP2 conformation with reduced binding to SHP2i

KRAS(G12X)-expressing cells exhibit differential sensitivity to panRAS- GEF(OFF)i

Abstract Therapeutically silencing the RAS/MAPK signaling cascade, an oncogenic driver in more than one-third of human cancers, is constrained by a fundamental trade-off: potent pathway inhibition in tumors versus dose-limiting toxicities in normal tissues. BRAF-mutant (BRAF-MUT) cancers are a notable exception, where current clinical RAF inhibitors (RAFis) (Type 1.5 - αC-OUT/DFG-IN) selectively inhibit monomeric BRAF(V600X), while paradoxically activating MAPK signaling pathway in settings where RAF signals as a dimer, including wild-type and RAS-Mutant (RAS-MUT) contexts. While this paradoxical activation limits the broader applicability to BRAF-MUT tumors, it has been therapeutically exploited in vertical MAPK-targeting combinations with MEK or EGFR inhibitors, enhancing antitumor efficacy while restoring physiological MAPK signaling in normal tissues, achieving an improved therapeutic window and enhancing tolerability.To target dimeric RAF-driven tumors, including RAS-MUT tumors, Type 2 (αC-IN/DFG-OUT) RAFis were developed to engage both RAF monomers and dimers. However, as single agents, Type 2 RAFis showed only modest activity. Combining them with MEK inhibitors improved efficacy but also exacerbated toxicities due to MAPK pathway suppression in normal tissues, limiting dosing and ultimately constraining therapeutic benefit.Here, we characterized ELV-3111, a next-generation, highly potent and selective Type 1 RAFi with broad activity across BRAF class I/II/III, CRAF, and RAS-MUT models, including contexts resistant to current MAPK-targeted therapies. Unlike Type 2 RAFis, ELV-3111 induces robust paradoxical MAPK hyperactivation selectively in normal tissues - a phenomenon we successfully modeled in cells. Using complementary biochemical and live-cell assays, alongside molecular dynamics simulations, we demonstrate that this MAPK hyperactivation occurs via a RAS-dependent allosteric mechanism distinct from the paradoxical activation described for Type 1.5 RAFis. This unique property can be therapeutically exploited. Combining ELV-3111 with a MEK inhibitor overcomes the therapeutic ceiling of MAPK pathway targeting by creating a pharmacologically complementary interaction: additive suppression in tumors, where both agents inhibit MAPK signaling, and opposing effects in normal tissues, where MEK inhibition counteracts RAFi-driven hyperactivation. This configuration produced profound and durable regressions across RAS- and BRAF-MUT models, including a RAS-MUT model refractory to current therapies, while maintaining favorable tolerability. This tumor-selective mechanism, previously exploited in BRAF-MUT cancers, can now be extended to RAS-MUT and other dimeric RAF-driven tumors, offering a renewed therapeutic opportunity and the potential to reshape combination strategies across a broader spectrum of MAPK-driven cancers. Citation Format: Mathieu Desaunay, Tara L. Peters, Evangelia Matenoglou, Beau Baars, Bijaya Gaire, Ana Orive-Ramos, Li Ren, Joseph P. Lyssikatos, Michael R. Burkard, Dalton Dacus, Matthew J. Sale, Stuart A. Aaronson, Frank McCormick, Evripidis Gavathiotis, Stefan D. Gross, Poulikos I. Poulikakos. Tumor-selective dimeric and monomeric RAF targeting with a next-generation Type 1 RAF inhibitor [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2940.

Abstract The therapeutic benefits of MAPK pathway inhibitors depend on achieving more potent pathway inhibition in tumors over normal tissues, maintaining a broad therapeutic index. In BRAF-mutant cancers, selective BRAF inhibitors (BRAFi) achieve this balance, and their combination with MEK inhibitors (MEKi) at full doses resulted in significant clinical benefit. In contrast, next-generation Type 2 dimeric RAF, also known as pan-RAF inhibitors (pan-RAFi), which target wild-type BRAF and CRAF and demonstrate single-agent activity in RAS-mutant (RAS-MUT) tumors, have failed to achieve comparable therapeutic impact in combination with MEKi, owing to toxicity-driven dose reductions that constrain clinical efficacy. In preclinical and clinical studies, these combinations have reached a therapeutic ceiling, yielding predominantly stable disease and only rare tumor regressions. We show that dose-limiting toxicity in pan-RAFi + MEKi combinations stems from relief of negative feedback by MEK inhibition, which amplifies RAF activation and pan-RAFi engagement in normal tissues, oversuppressing MAPK signaling and limiting the therapeutic index. RAF/MEK glues constitute a distinct category of MEK inhibitors that stabilize RAF-MEK complexes and functionally suppress RAF. Although structural studies have captured these glues bound to both active and inactive RAF conformations, we show that their inhibitory activity stems from a spatial-trapping mechanism: we found that MEK is constitutively cytosolic, RAF/MEK glues sequester RAF in the cytosol, blocking its membrane recruitment and dimerization, steps essential for activation. In line with this mechanism, the RAF/MEK glue avutometinib, when combined with a pan-RAFi, was well tolerated at full dose and, critically, drove tumor regressions across multiple RAS-MUT models, achieving a 90% ORR compared with 0% using a conventional pan-RAFi + MEKi regimen in an insensitive RAS-MUT model. Tumor regressions corresponded with deeper MAPK pathway suppression by both pharmacodynamic and transcriptional metrics, without increased toxicity. Together, these findings reveal an unrecognized mechanism of RAF inactivation by RAF/MEK glues and show that substituting the MEKi with a RAF/MEK glue can overcome the therapeutic ceiling of current MAPK-targeting regimens, shifting outcomes from mostly stable disease to frequent tumor regressions. More broadly, this work establishes drug-induced proximity as a means to increase tumor selectivity by reprogramming the spatial and biochemical state of wild-type signaling effectors, thereby widening the therapeutic window for oncogenic pathway inhibition and providing a new paradigm for precision oncology. Citation Format: Bijaya Gaire, Ana Orive-Ramos, Christos Adamopoulos, Beau Baars, Mathieu Desaunay, Evangelia Matenoglou, Silvia Coma, Nayeli Gutierrez-Trejo, Kevin Mohammed, Stuart A. Aaronson, Jian Jin, Tiphaine Martin, Ernesto Guccione, Evripidis Gavathiotis, Jonathan A. Pachter, Poulikos I. Poulikakos. A spatial-trapping mechanism of RAF by RAF/MEK glue enables full-dose combination with a pan-RAF inhibitor and drives potent, RAS-mutant tumor-selective MAPK and growth inhibition [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2948.

Sensitivity to any single RAS-targeted inhibitor is restricted to the subset of RAS(MUT)-specific inhibitor-sensitive tumors

KRAS(G13D) cells are insensitive to next generation SHP2i and SOS1i

Chemical structures of compounds used in this study

Abstract The landscape of RAS-directed therapies has rapidly advanced following the advent of mutant-selective KRAS(G12C) inhibitors, driving the development of additional RAS-targeting agents, including mutant-selective (e.g. KRAS(G12C), KRAS(G12D)), as well as paralog- and state-selective compounds. Non-mutant- specific RAS inhibition can currently be achieved through three strategies: (i) guanine nucleotide exchange-OFF inhibitors (panRAS-GEF(OFF)i) that indirectly inactivate RAS by targeting SHP2 or SOS1, (ii) KRAS-OFF inhibitors (panKRAS(OFF)i) that spare NRAS and HRAS, and (iii) active-state RAS(ON) inhibitors (panRAS(ON)i) that directly block binding of effector RAF. Although these therapeutic modalities have shown promise, their clinical effectiveness and tolerability ultimately depend on achieving a high therapeutic index, defined as potent inhibition of oncogenic signaling in tumor cells with minimal effects on normal cells. To more robustly quantify tumor selectivity in preclinical models, we introduce the signaling inhibition index (SII), which measures the differential suppression of oncogenic signaling between RAS(MUT) and RAS(WT) cells, providing a more structured metric of tumor selectivity that has previously been poorly defined. Here, we evaluated the SII for state- and paralog-selective RAS inhibitors across diverse RAS(MUT) and RAS(WT) models. PanRAS-GEF(OFF)i exhibited neutral or negative SII, reflecting reduced MAPK suppression in KRAS(G12X) cells compared to wild-type cells. KRAS(G13D) models, especially with NF1 loss, showed low sensitivity. Combining SHP2 and MEK inhibition resulted in low tumor-selectivity, while RAS(Q61X) models were resistant due to MEK inhibitor-induced NRAS reactivation and altered SHP2 conformations. Consistent with these findings, analysis of DepMap SHP2-inhibitor sensitivity and dependency datasets showed that RAS(MUT) cell lines are not more sensitive than RAS(WT) cells to SHP2 inhibition, further underscoring the limited tumor selectivity of panRAS-GEF(OFF)-based approaches. In parallel, we assessed panKRAS(OFF)i and panRAS(ON)i potency/selectivity across a panel of RAS(MUT) and RAS(WT) cell line models. KRAS(OFF) inhibitors demonstrated higher selectivity, whereas active-state RAS(ON) inhibitors showed broader activity but narrow selectivity. Comparative analyses of published datasets revealed correlated sensitivity patterns across RAS inhibitor classes, indicating that therapeutic activity is largely restricted to the same subset of RAS(MUT) cancers. These findings highlight the importance of systemic SII quantification for therapeutic selectivity and for guiding the rational design and clinical implementation of next-generation RAS-targeted therapies. Citation Format: Beau Baars, Ana Orive-Ramos, Matthew Emmett, Bijaya Gaire, Mathieu Desaunay, Ziyue Kou, Guangyan Li, Christos Adamopoulos, Stuart A. Aaronson, Shaomeng Wang, William R. Sellers, Tiphaine Martin, Evripidis Gavathiotis, Poulikos I. Poulikakos. Profiling tumor selectivity of state- and paralog-selective RAS inhibitors through a signaling inhibition index (SII) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3898.

Non-small cell lung cancers (NSCLCs) treated with tyrosine kinase inhibitors (TKIs) of the epidermal growth factor receptor (EGFR) almost invariably relapse in the long term, due to the emergence of subpopulations of resistant cells. Through a DNA barcoding approach, we show that the clinically approved drug sorafenib specifically abolishes the selective advantage of EGFR-TKI-resistant cells, while preserving the response of EGFR-TKI-sensitive cells. Sorafenib is active against multiple mechanisms of resistance/tolerance to EGFR-TKIs and its effects depend on early inhibition of MAPK-interacting kinase (MKNK) activity and signal transducer and activator of transcription 3 (STAT3) phosphorylation, and later down-regulation of MCL1 and EGFR. Using different xenograft and allograft models, we show that the sorafenib-EGFR-TKI combination can delay tumor growth and promote the recruitment of inflammatory cells. Together, our findings indicate that sorafenib can prolong the response to EGFR-TKIs by targeting NSCLC capacity to adapt to treatment through the emergence of resistant cells.

A high therapeutic index (TI), balancing potent oncogenic signaling inhibition in tumor cells with minimal effects on normal cells, is critical for effective cancer therapies. Recent advances have introduced diverse RAS-targeting inhibitors, including mutant-specific inhibitors (e.g., KRAS(G12C) and KRAS(G12D)), as well as paralog- and state-selective inhibitors. Non-mutant-specific RAS inhibition can be accomplished by 1) panRAS-GEF(OFF) inhibitors which inactivate RAS indirectly by inhibiting SHP2 or SOS1, thereby blocking the nucleotide exchange step of RAS activation, 2) direct KRAS(OFF)-selective inhibitors sparing NRAS and HRAS, and 3) panRAS(ON) inhibitors that directly target active RAS, by occluding binding of its effector RAF. However, the signaling inhibition index (SII) - the differential inhibition of oncogenic signaling between RAS-mutant (RAS(MUT)) and normal cells - remains poorly defined for these approaches. In this study, we evaluated the SII of state- and paralog-selective RAS inhibitors across diverse RAS-mutant (RAS(MUT)) and RAS-wild-type (RAS(WT)) models. PanRAS-GEF(OFF) inhibitors exhibited neutral or negative SII, with comparable or reduced MAPK suppression in KRAS(G12X) cells relative to RAS(WT) cells. KRAS(G13D) models showed low sensitivity (negative SII) to panRAS-GEF(OFF) inhibitors, particularly in the context of NF1 loss. Combination treatments with SHP2 and MEK inhibitors resulted in low SII, as pathway suppression was similar in RAS(MUT) and RAS(WT) cells. Furthermore, RAS(Q61X) models were resistant to combined SHP2 inhibitor+MEK inhibitor due to dual mechanisms: MEK inhibitor-induced NRAS(Q61X) reactivation and RAS(MUT)-induced SHP2 conformations impairing inhibitor binding. Overall, panRAS-GEF(OFF) inhibitors exhibited the lowest SII. PanKRAS(OFF) inhibitors demonstrated a higher SII, while panRAS(ON) inhibitors displayed broader activity but relatively narrow SII. We observed that tumors that were sensitive to RAS(MUT)-specific inhibitors, were also sensitive to the state-selective RAS inhibitors (OFF, or ON). In fact, all RAS inhibitors (mutant-specific and state- or paralog-selective) were active in the same portion of RAS(MUT) models, while the majority of RAS(MUT) cell lines were insensitive to all of them. These findings reveal significant SII variability among RAS-targeted inhibitors, depending on the specific RAS driver mutation and cell context and underscore the importance of incorporating SII considerations into the design and clinical application of RAS-targeted therapies to improve therapeutic outcomes. Main points:PanRAS-GEF(OFF) inhibitors have limited SII and effectiveness: The Signaling Inhibition Index (SII) - i.e. the differential inhibition of oncogenic signaling between tumor and normal cells - was neutral or negative for panRAS-GEF(OFF) inhibitors, with comparable or reduced MAPK suppression in KRAS(G12X) mutant versus RAS(WT) cells. KRAS(G13D) models showed reduced sensitivity, particularly with NF1 loss. SHP2+MEK inhibitor combinations also had low SII, with RAS(Q61X) models demonstrating resistance due to NRAS(Q61X) reactivation and impaired SHP2 inhibitor binding.PanKRAS(OFF) selective inhibitors have higher SII than panRAS-GEF(OFF) inhibitors: panKRAS(OFF)-selective inhibitors have a higher SII compared to panRAS-GEF(OFF) inhibitors, offering better tumor-versus-normal cell selectivity.PanRAS(ON) inhibitors have broad but modest SII: While panRAS(ON) inhibitors displayed a broader activity profile, their ability to selectively inhibit mutant RAS signaling over normal cells remained relatively narrow (low SII).Most KRAS-mutant tumors will be insensitive to any single RAS-targeted inhibitor: State- and paralog-selective inhibitors have enhanced activity in the same RAS-MUT cancer models that are also sensitive to RAS-MUT-specific inhibitors, suggesting that most KRAS-MUT tumors will not respond uniformly to any one RAS-targeting inhibitor.SII varies across RAS inhibitors, necessitating tailored therapeutic strategies: The effectiveness of paralog- and state-selective inhibitors depends on specific RAS mutations and cell context, highlighting the need to integrate SII considerations into the development and clinical application of RAS-targeted therapies.

Fig. S7. Aggregate tSNE plots generated from aggregate analysis of different clusters of WTC and control patients. A, Circled populations in t-SNE plots showed significant differences between WTC and control patients. B, List of markers used to define the clusters identified in tSNE plots (WTC, n=8 and non-WTC, n=12). C, Meta clusters showing the difference in cell density values occurring in IMC tissue samplings (n=2) for each patient sample (n=10).

Fig. S2. Volcano plots showing gene expression profiles for C57BL/6 mouse organs at 7d and 21d after WTC dust exposure normalized to PBS control mice (n=3).

Fig. S6. Cell densities of different cell populations in WTC and non-WTC patient tissues samples. A, Immunoreactive cells for individual antibodies tested in WTC and non-WTC IMC analysis (WTC, n=8 patient samples and non-WTC, n=12 patient samples). Graphs were generated based on the analysis of tSNE plots shown in Fig. S7. B, Cell numbers used for IMC analysis for non-WTC and WTC analysis

List of genes amplified at a frequency of 0.5% or greater in all cancer types in TCGA (used in GSEA). Acronyms for cancer types are expanded in the Materials and Methods.

Fig. S3. A-B, Expression of P-AKT-S473 in the prostatic acini from WTC dust treated Pb-Cre+PtenL/Wt mice shown at low and high magnifications (A = early progression, B = later progression). C, P-AKT-S473 expression in Pb-Cre+PtenL/L GEM mice (HE bar = 250 μM, IHC bar = 100 μm). D, IF expression of PTEN and P-AKTS473 in Pb-Cre+PtenL/Wt GEM mice treated with WTC dust.

Supplementary Figure 1 Legend from Clonal Selection in Malignant Transformation of Human Fibroblasts Transduced with Defined Cellular Oncogenes

ABSTRACT Non-small cell lung cancers (NSCLCs) treated with tyrosine kinase inhibitors (TKIs) of the epidermal growth factor receptor (EGFR) almost invariably relapse in the long term, due to the emergence of subpopulations of resistant cells. Here we show that the lack of sensitivity of these cells to EGFR-TKIs constitutes a vulnerability that can be potentially targeted. Through a DNA barcoding approach, we demonstrate that the clinically approved drug sorafenib specifically abolishes the selective advantage of EGFR-TKI-resistant cells, while preserving the response of EGFR-TKI-sensitive cells, thus resulting in overall inhibition of clonal evolution within the tumor cell mass population. Sorafenib is active against multiple mechanisms of resistance/tolerance to EGFR-TKIs and its effects depend on early inhibition of MAPK interacting kinase (MNK) activity and signal transducer and activator of transcription 3 (STAT3) phosphorylation, and later down-regulation of MCL1 and EGFR. Using several xenograft and allograft models to recapitulate different mechanisms and kinetics of acquired resistance, we show that the sorafenib-EGFR-TKI combination can substantially delay tumor growth and promote the recruitment of inflammatory cells. Together, our findings indicate that sorafenib can substantially prolong the response to EGFR-TKIs by targeting NSCLC capacity to adapt to treatment through the emergence of resistant cells.