Single-cell sensor analyses reveal signaling programs enabling Ras-G12C drug resistance

Oncogenic mutations in Ras occur in roughly one-third of cancers and drive strong MAPK pathway signaling. Although covalent inhibitors targeting the inactive, GDP-bound KRas-G12C (for example, AMG-510/sotorasib) have shown clinical efficacy, most patients develop resistance characterized by Ras–MAPK pathway reactivation. The cellular processes perturbed by Ras-G12C inhibitors at the signaling level, and the determinants that distinguish drug-resistant from quiescent cells, remain unclear. KRas-G12C cancers also express wild-type HRas and NRas, which may compensate during KRas-G12C inhibition, and Ras isoforms differ in subcellular localization (WT H/NRas access ER/Golgi; KRas4A can localize to mitochondria). The authors hypothesized that KRas-G12C inhibitor treatment induces adaptive reorganization of Ras signaling across subcellular compartments, enabling drug resistance, and set out to resolve these dynamics at single-cell and subcellular resolution using genetically encoded Ras biosensors.

Prior work established: (1) clinical efficacy and resistance to covalent KRas-G12C GDP-state inhibitors (AMG-510) and reports of pathway reactivation (Ras–MAPK rebound); (2) rapid, nonuniform adaptations to KRas-G12C inhibition across cells; (3) distinct Ras isoform biology, particularly subcellular localization of H/NRas to ER/Golgi and KRas4A to mitochondria; (4) reports that EGFR/GEF signaling can drive WT Ras activation during KRas-G12C inhibition; and (5) metabolic rewiring upon KRas-G12C inhibitor treatment, including increased glycolytic enzymes/metabolites. MVP has been implicated in intracellular signaling scaffolding (interactions with Shp2, Erk, RTKs) and in chemoresistance, but its mechanistic role in Ras inhibitor resistance was unresolved.

• Biosensors: Used two-component genetically encoded Ras sensors. Ras-LOCKR-S (FRET-based YFP/CFP) reports Ras-GTP-dependent cage–key association to measure Ras activity. Ras-LOCKR-PL uses split TurboID reconstitution to biotinylate proteins near active Ras (Ras signalosome). Both sensors can be subcellularly targeted (PM, ER, Golgi, mitochondria) via N-terminal localization sequences.
• Cell lines: KRas-G12C cancer lines H358 (heterozygous), MIA PaCa-2 and SW1573 (homozygous). WT mouse embryonic fibroblasts served as controls.
• Treatments: AMG-510 (Ras-G12C–GDP inhibitor), RasONi RMC-6291 (Ras-G12C–GTP inhibitor), EGFR inhibitor cetuximab, SOS1 inhibitor BI-3406. Time courses up to 72 h; refresh experiments as indicated.
• Single-cell readouts: Live-cell epifluorescence FRET for Ras-LOCKR-S at compartments; immunostaining for pErk; PLA (proximity ligation assay) for protein–protein proximity or biotinylation detection; Cyto-PercevalHR sensor for cytosolic ATP:ADP ratio; mito-Ras-LOCKR-S for mitochondrial Ras-GTP.
• Proteomics: Golgi-targeted Ras-LOCKR-PL with biotin labeling under DMSO vs AMG-510 (4 h, 24 h). Streptavidin pulldown, on-bead digestion, LC-MS/MS (Orbitrap Fusion Lumos). Data processed with MaxQuant/Perseus; selection criteria included log2 fold-change and statistical cutoffs; GO/PANTHER filtering for signaling.
• Genetic perturbations: siRNA knockdown of MVP, HRas and NRas; expression of compartment-targeted dominant negative Ras (Ras-S17N) via lentiviral transduction; rescue experiments with ectopic MVP.
• Metabolism assays: Bulk glucose consumption and lactate release (Lactate-Glo, Glucose-Glo) normalized to protein; single-cell ATP:ADP via Cyto-PercevalHR.
• Biochemistry: Immunoblotting for pErk, pAkt, MVP, Ras; MVP immunoprecipitation and anti–phosphotyrosine probing; phosphatase (PAP) treatment prior to MVP IP to assess phosphorylation-dependent interactions.
• Imaging and analysis: Standardized epifluorescence setup; background subtraction; FRET ratio calculation; ImageJ Coloc 2 for colocalization; PLA puncta quantification; statistics via t-tests and ANOVA; CV reported to assess heterogeneity.

• Heterogeneous MAPK rebound after AMG-510: pErk levels became heterogeneous (H358 CV 30–42% at 24–72 h), indicating a signaling-adaptive subpopulation.
• Compartmental Ras activity remodeling with AMG-510: PM Ras activity decreased at 4 h then partially rebounded; endomembrane Ras activity increased, most prominently at the Golgi (H358 Golgi Ras-LOCKR-S FRET increased; CV at 72 h ~19–23%). WT MEFs showed no changes.
• WT H/NRas drives Golgi Ras activity: EGFRi and SOS1i reduced AMG-510–promoted Golgi Ras activation but had minimal effect at the PM; Ras-G12C–GTP inhibitor (RasONi) reduced PM rebound but not Golgi Ras activation. HRas/NRas siRNA abrogated Golgi Ras activation and reduced endomembrane Ras localization.
• Golgi Ras activity correlates with MAPK rebound and growth: Only Golgi Ras activity correlated with pErk at single-cell level (H358 R^2=0.78; SW1573 R^2=0.62). Golgi-targeted Ras-S17N reduced Golgi Ras activity, prevented pErk rebound and blocked AMG-510–treated cell proliferation.
• MVP identified as Golgi Ras signalosome mediator: Golgi Ras-LOCKR-PL proteomics after AMG-510 enriched MVP (and PAPOLG, VDAC1/2) at 24 h. PLA and IP confirmed increased MVP biotinylation after AMG-510. MVP knockdown eliminated AMG-510–promoted Golgi Ras activation and suppressed adaptive growth/colony formation.
• MVP is required for global adaptive signaling: MVP KD eliminated pErk rebound and decreased pAkt; MVP levels increased within 24 h of AMG-510. MVP colocalization with Shp2, Erk, EGFR, FGFR3 increased after AMG-510 and correlated with Golgi Ras activity (R^2=0.85). MVP KD reduced Shp2–Erk and Shp2–EGFR proximities. MVP tyrosine phosphorylation increased over 72 h; phosphatase treatment reduced MVP co-IP with Shp2, Ras and pErk, indicating phosphorylation-dependent scaffolding.
• Metabolic adaptations with AMG-510: Cytosolic ATP:ADP decreased heterogeneously (H358 CV 28% at 72 h) and bulk assays showed increased glucose consumption and lactate release, consistent with enhanced glycolysis. Mitochondrial Ras (mito-Ras-LOCKR-S) showed initial decrease then heterogeneous increase (CV 20% at 72 h), interpreted as adaptive KRas4A-GTP upregulation; RasONi addition after AMG-510 decreased mito-Ras and increased ATP:ADP, implicating mutant KRas4A.
• MVP links to mitochondrial adaptations: AMG-510 increased MVP mitochondrial localization and MVP–VDAC1 PLA; MVP–VDAC1 colocalization inversely correlated with ATP:ADP (combined R^2=0.81). MVP KD abolished AMG-510–induced ATP:ADP changes and blunted mito-Ras activation; MVP KD reduced Ras–VDAC1 colocalization which inversely correlated with ATP:ADP (R^2=0.79).
• Ras-G12C–GTP inhibitor (RasONi) also induces MVP-dependent Golgi WT Ras activation: RasONi caused sustained PM Ras suppression and heterogeneous Golgi Ras increase (CV 29% at 72 h), partial pErk reactivation by 48–72 h, and tripled MVP expression by 72 h. MVP KD abrogated RasONi-induced Golgi Ras activation. Unlike AMG-510, RasONi sustained mito-Ras inhibition and did not alter ATP:ADP or glycolytic hallmarks.

The study resolves how KRas-G12C inhibitor–treated cancer cells rapidly rewire Ras signaling at subcellular scales to evade therapy. Single-cell biosensors revealed that Golgi-localized WT H/NRas activation, not PM or ER pools, underpins MAPK pathway rebound and drug-resistant proliferation. Mechanistically, MVP emerges as a central scaffold that becomes tyrosine phosphorylated and brings together Shp2, Erk, RTKs (EGFR/FGFR3) and WT H/NRas at the Golgi, thereby enhancing Ras-GTP accumulation and downstream MAPK signaling in a subpopulation of cells. In parallel, AMG-510 triggers an MVP-dependent mitochondrial program involving KRas4A and VDAC1/2 that lowers cytosolic ATP:ADP and promotes glycolysis, delineating a metabolically adaptive subpopulation. While Ras-G12C–GTP inhibition (RasONi) differs by effectively suppressing mitochondrial KRas4A activity and metabolic rewiring, it still induces MVP expression and MVP-dependent Golgi WT Ras activation, indicating a shared escape mechanism across inhibitor classes. These insights emphasize the importance of spatially resolved, single-cell analysis to uncover heterogeneity and identify actionable nodes like MVP that coordinate compensatory signaling and metabolic adaptations driving drug resistance.

This work introduces subcellularly targeted, single-cell Ras biosensors to dissect adaptive signaling underpinning KRas-G12C inhibitor resistance. The authors identify MVP as a key mediator that scaffolds MAPK components and WT Ras at the Golgi to reactivate oncogenic signaling, and coordinates with KRas4A and VDAC to promote glycolytic adaptations at mitochondria under AMG-510. Both GDP- and GTP-state KRas-G12C inhibitors induce MVP-dependent Golgi WT Ras activation, while only AMG-510 promotes mitochondrial and metabolic adaptations. Targeting MVP-mediated scaffolding or its phosphorylation, interrupting MVP–Shp2/Erk/RTK interactions, and co-targeting RTK/GEF pathways may suppress the signaling adaptive subpopulation. Future work should clarify how KRas-G12C inhibition upregulates MVP, define whether MVP accumulation is compartment-specific, identify Golgi-resident Ras GEF contributions, and validate these mechanisms in physiologically relevant in vivo models and across other Ras-mutant contexts.

• Spatial resolution of the Ras-LOCKR-PL proximity labeling is limited by long labeling times and potential diffusion, reducing precise subcellular localization of labeled interactors.
• The study did not identify a specific Ras GEF in the Golgi Ras signalosome; mechanisms that increase Ras-GTP at the Golgi remain partly unresolved.
• It is unclear whether MVP specifically accumulates at the Golgi versus broadly increasing across compartments due to overall expression changes.
• The upstream mechanisms by which KRas-G12C inhibition induces MVP expression are unknown and could involve cytokine signaling or other stress responses.
• Most experiments were in vitro cell lines; in vivo validation is needed to confirm relevance and therapeutic tractability.
• Antibody limitations constrained direct isoform-specific Ras proximity analyses; some conclusions rely on proxies and correlations.