Efficacy and Imaging-Enabled Pharmacodynamic Profiling of KRAS G12C Inhibitors in Xenograft and Genetically Engineered Mouse Models of Cancer

KRAS encodes a small membrane-bound GTPase, which switches between inactive GDP-bound and active GTP-bound states to regulate signal transduction from cell surface growth factor receptors to downstream effectors. The transition between states is tightly controlled by guanine nucleotide-exchange factors (GEF), which load GTP to activate KRAS and by GTPase-activating proteins (GAP) that hydrolyze GTP to GDP to inactivate it. Once activated, KRAS promotes signaling to effectors such as the RAF–MEK–ERK and PI3K–AKT pathways, which in turn control cellular processes including proliferation, differentiation, survival, migration, and glucose metabolism (1).

Although important for normal cell function, KRAS is one of the most commonly mutated oncogenes in human cancers, and 25% to 30% of all tumors harbor KRAS-activating point mutations (2, 3). These alterations are most prevalent in pancreatic, lung, and colorectal cancers, but the frequency of KRAS mutation and mutation subtypes differ across indications. For example, 90% of pancreatic adenocarcinomas carry KRAS mutations, of which the G12D variant makes up 40% of cases whereas G12C only accounts for 1% (3). Non–small cell lung cancer (NSCLC), in contrast, is 30% KRAS mutant, but G12C is the dominant mutation and accounts for approximately 40% of cases (2).

Therapeutic targeting of KRAS-driven tumors has been a subject of research for decades, but direct targeting of KRAS was challenging due to the increased GTP binding affinity and lack of suitable small molecule binding pockets in the mutant proteins. However, recent breakthroughs identified compounds capable of specifically targeting the G12C protein by covalently binding to the mutant cysteine in a pocket only found in its GDP-bound form (4–6). Small-molecule KRAS G12C inhibitors, such as sotorasib (AMG510) and adagrasib (MRTX849), thereby irreversibly bind and trap KRAS G12C in its inactive form, shutting down oncogenic KRAS signaling and aberrant tumor cell proliferation (4, 7). Although early clinical trials for sotorasib and adagrasib have shown promising response rates in KRAS G12C NSCLC, a significant fraction of patients have innate resistance and fail to respond; those who respond initially also inevitably develop resistance and relapse (8). Thus, understanding the mechanisms of resistance to G12C inhibitors and identifying strategies to overcome them remains necessary for improving patient outcomes.

Although preclinical studies rely heavily on cell line-derived xenografts (CDX) to demonstrate in vivo proof-of-concept, these models are limited by their subcutaneous implant sites and lack of host immune systems. Genetically engineered mouse models (GEMM), in contrast, enable de novo tumor development that can recapitulate human cancer progression in the appropriate immune-competent microenvironment. Moreover, GEMMs have become valuable translational tools for studying relapse and both intrinsic and acquired drug resistance (9, 10). Knock-in models where mutations are engineered into the endogenous gene locus and only expressed in target tissues of interest are especially relevant as they allow for physiologic levels of gene expression and native regulatory elements that may be important for elucidating mechanisms of resistance. To address the lack of G12C-specific GEMMs in cancer research, we engineered and characterized a knock-in mouse model of NSCLC driven by Cre-induced Kras G12C activation and p53 deletion (Kras^LSL-G12C/wtp53^fl/fl), hereafter referred to as the G12C KP GEMM.

Here, using the MiaPaCa2 CDX and the G12C KP GEMM, we provide a framework consisting of multimodality preclinical imaging—microCT and ¹⁸F-FDG PET, a readout of glucose metabolism—for pharmacodynamic (PD) and efficacy evaluations of KRAS inhibition by an in-house developed KRAS G12C inhibitor (hereafter referred to as Compound A) and sotorasib. Conventional endpoints such as tumor growth measurement, survival, and IHC were also included to cross-validate with imaging endpoints. Our results establish new preclinical and translational tools to aid ongoing investigation of targeting KRAS G12C and overcoming resistance to inhibitors.

For the following in vitro and ex vivo assays, see Supplementary Materials and Methods for details: Cell Titer Glo (CTG) assay, ELISA RBD assay, target occupancy assay, and IHC, slide scanning and digital image analysis, histology-based lung tumor burden measurement, and pharmacokinetics (PK).

All animal studies were performed as per the animal use protocols approved by Pfizer Inc.’s Institutional Animal Care and Use Committee in an AAALAC-accredited facility. For the human CDX model, MiaPaCa2 cells were resuspended in serum-free RMPI media and mixed 1:1 with Matrigel. A total of 5 × 10⁶ cells were injected subcutaneously in 200 μL into the right flank of female nu/nu mice (Charles River Laboratories). Bodyweight and tumor volume (width² × length × 0.5) were monitored twice a week. The efficacy study was initiated when average tumor volume reached ∼200 mm³. Animals were randomized into treatment groups (n = 10/group) and dosed once daily for 2 weeks with 10 mL/kg of vehicle or Compound A at 1, 5, and 30 mg/kg. Tumor growth inhibition (TGI) was calculated as [1 − (T_t − T₀)/(C_t − C₀)] × 100, where T₀ and T_t are the mean tumor volume of treated group on day 0 and the last day of treatment, respectively, and C₀ and C_t are the mean tumor volume of the control group on those respective timepoints. End-of-study plasma samples were collected at 1, 7, and 24 hours after the last dose to assess exposure of Compound A. In a separate PK/PD study, mice with tumors averaging ∼400 mm³ were treated daily for 3 days at 1, 5, and 30 mg/kg. Plasma and tumors were collected at 1, 3, 5, 8, and 24 hours after the third dose for analysis of PK, target occupancy by LC/MS, and biomarker modulation by IHC.

Mutations in Kras^LSL-G12C/wt p53^fl/fl and Kras^LSL-G12D/wt p53^fl/fl genetically engineered mouse models were induced in 8 to 12 weeks old mice via 50 μL inhalations of 1 × 10⁶ IFU of Adenoviral Cre (Viral Vector Core, University of Iowa). Mice were monitored for tumor burden scoring using microCT pre- and post-induction at 2- to 3-week intervals and randomized into treatment groups. Animals were dosed once daily with vehicle, 30 mg/kg Compound A, or 100 mg/kg sotorasib for 5 weeks or until animals reached euthanasia criteria. In addition to microCT imaging, mice were monitored weekly for clinical symptoms and body weight loss. Plasma for PK was collected at 1, 7, and 24 hours post-dose on day 28 of dosing. Lungs were inflated with 10% neutral-buffered formalin for paraffin embedding and IHC.

Mouse microCT images were acquired on Skyscan 1278 (Bruker) using a 0.5 cm Al filter at 918 μA, 50 kV, 100 μm resolution, and a step size of 0.5° per projection for 360°. Animals were anesthetized with 2% isoflurane/98% oxygen inhalation during the scan. Tomography projection images were reconstructed into cross-section images using NRecon (version 1.7.1.0; Bruker) with ring artifact correction of 7, a 12% beam hardening correction, and Gaussian smoothing with a kernel size of 2. The thoracic region of interest was defined by the tracheal bifurcation (the cranial limit) and the descending distal curvature of the diaphragm (the caudal limit). Reconstructed images were viewed in Amira (version 6.3; Thermo Fisher Scientific). A visual score on a 0 to +++ subjective scale was then assigned based on manual examination of tumor burden. Independently, reconstructed images were imported into the MLAST program and analyzed quantitatively (11). The MLAST score represents the percentage of the thoracic cavity occupied by healthy, non-tumorous, lung tissue. Tumor progression is therefore reflected by a reduction of the MLAST score. As a reference, healthy lung tissues usually constitute 37% to 45% of the thoracic cavity in a tumor-free C57Bl/6 mouse.

Fluorine 18-fluorodeoxyglucose (¹⁸F-FDG) was procured from PETNET Solutions, Inc. For each PET imaging session, mice were fasted for at least 4 hours prior and approximately 80 μCi ¹⁸F-FDG was injected intravenously. Mice were then anesthetized under 2% isoflurane inhalation and remained anesthetized and kept warm at 37°C during a 1-hour uptake period and throughout the imaging session. MicroPET/CT images were acquired using the G8 benchtop scanner from Sofie Biosciences with the following parameters: ¹⁸F peak-calibrated 7-minute static scan with a vendor provided 3D ML-EM reconstruction at 1.4 mm resolution, immediately followed by a 2-minute co-registered CT scan [50 kVp, 200 μA, and 75-μm resolution on a step-and-shoot (5 degrees/step)]. Animals were euthanized after the final timepoint imaging. Tumors were collected, weighed, placed in scintillation vials, submerged in 10% neutral-buffered formalin, and subject to ex vivo gamma counting using the Wizard 2480 Gamma Counter (PerkinElmer). Tails were collected and gamma-counted to correct for the injected dose (ID). PET/CT images were processed and analyzed using the VivoQuant software (Invicro). Regions of interest (ROI) were manually drawn based on the co-registered CT images of the tumor. ¹⁸F-FDG uptake was calculated as (the amount of radioactivity detected in the tumor) / (total ID) × 100% / (tumor volume/weight) and reported in the unit of either %ID/cm³ or %ID/g, assuming tissue density of 1 g/cm³. Tumor samples fixed in formalin were then paraffin embedded, made into blocks, and evaluated by IHC (see Supplementary Materials and Methods for details).

The data generated in this study are available upon request from the corresponding author.

We developed Compound A, (1-(4-(8-((5,6-dichloro-1H-indazol-4-yl)oxy)-2-(((3R,4R)-4-methoxy-1-methylpyrrolidin-3-yl)oxy)pyrido[3,4-d]pyrimidin-4-yl)piperazin-1-yl)prop-2-en-1-one), as a tool covalent inhibitor of the KRAS G12C mutant (structure shown in Supplementary Fig. S1, patent; ref. 12). By binding in the switch-II pocket similarly to other KRAS G12C inhibitors (12), it enabled facile yet selective modification of the KRAS protein at the target mutant cysteine. Biochemical activity of Compound A was measured via ELISA quantification of the key protein–protein interaction between KRAS and the RAS-binding domain (RBD) of RAF-1 that affects downstream MAPK signaling and is dependent on KRAS being in the GTP-bound active form (12). Compound A potently inhibited KRAS G12C in MiaPaCa2 and H358 cell lines, with IC₅₀ values of 4 and 6 nmol/L, respectively (Supplementary Table S1). No inhibition was observed (IC₅₀ > 10 μmol/L) in a KRAS G12S control cell line A549. In a 7-day cell proliferation assessment, Compound A was shown to have 8 nmol/L potency in H358. The in vitro biochemical and anti-proliferative activity of Compound A was comparable with the clinically approved G12C covalent inhibitor, sotorasib (Supplementary Table S1).

To evaluate target engagement and downstream PD modulation in vivo, we administered Compound A at 1, 5, and 30 mg/kg once daily for 3 days to mice bearing subcutaneous MiaPaCa2 pancreatic cancer xenografts. Tumors were collected at different timepoints after 3 days of treatment, and total KRAS proteins were immunoprecipitated from the tumor lysates and subjected to LC/MS to measure unbound KRAS G12C peptides (LVVGAC^[CAM]GVGK) and determine the % occupied by Compound A. Although greater variability was observed at the 1 mg/kg dose, consistently high levels of KRAS G12C occupancy were observed in the 5 and 30 mg/kg groups. Peak target occupancy occurred around 3 hours and averaged 91% and 98% at 5 and 30 mg/kg of Compound A, respectively (Fig. 1A). Overall, average KRAS G12C engagement remained above 75% and 88% for the 5 and 30 mg/kg doses, respectively, during the 24-hour dosing interval. Matching tumor samples were analyzed by IHC staining for targets downstream of KRAS. Digital image analysis using pixelwise H-score (13) indicated pERK levels were decreased in a dose-dependent manner (Fig. 1B). Over time, the lowest dose of Compound A only resulted in a moderate decrease in pERK at 1 and 3 hours, which afterwards recovered to baseline levels matching that of the vehicle-treated group. Sustained pERK inhibition was achieved at 30 mg/kg (Fig. 1B). These data confirm that Compound A specifically targets KRAS G12C and inhibits its catalytic activity in vivo.

To assess the effect of KRAS G12C inhibition on glucose uptake, we performed ¹⁸F-FDG PET imaging with the MiaPaCa2 xenograft model. ¹⁸F-FDG is a glucose analog and has been widely accepted in preclinical and clinical settings to be a suitable pharmacodynamic (PD) biomarker to reflect glucose uptake and metabolism (14–16). As demonstrated by both image analysis (Fig. 1C and D) and ex vivo gamma counting quantitation (Supplementary Fig. S2A), while all animals showed a moderate level of baseline ¹⁸F-FDG uptake consistent with the glucose avidity of this tumor model (17), the Compound A-treated groups demonstrated a significant decrease in their tumor uptake following 3 days of treatment. The mean tumor uptake of ¹⁸F-FDG was 2.50%ID/g and 2.56%ID/g for the 5 and 30 mg/kg groups, respectively, compared to 4.56%ID/g in the vehicle group. Both dose levels showed similar and significant inhibition, suggesting that the 5 mg/kg dose was effective at inhibiting KRAS-driven glucose uptake. Notably, consistent reduction of pERK signaling was also detected in this cohort of animals by IHC (Supplementary Fig. S2B).

To evaluate efficacy of Compound A in vivo, we treated animals bearing MiaPaCa2 tumors with vehicle or Compound A at 1, 5, and 30 mg/kg for 2 weeks. Serial blood (plasma) sampling at 1, 7, and 24 hours post-final dose showed dose-dependent exposures, where 30 mg/kg maintained unbound plasma concentrations above the unbound cellular IC₅₀ for the entire 24-hour dosing interval (Fig. 1E). All dose levels were well-tolerated without significant weight changes (Supplementary Fig. S3A). We observed dose-dependent tumor growth inhibition (TGI), where 1, 5, and 30 mg/kg achieved 80.8% (P = 0.15), 123.5% (P = 0.0002), and 135.9% (P < 0.0001) TGI, respectively (Fig. 1F; Supplementary Fig. S3B). On the last day of the study (day 26), the average tumor volumes of groups treated with Compound A were significantly smaller (P = 0.004 for 1 mg/kg, P < 0.0001 for 5 and 30 mg/kg) compared to vehicle (Supplementary Fig. S3C). Furthermore, the endpoint tumor volumes of the 5 and 30 mg/kg groups were substantially smaller than their initial volumes, indicating tumor regression (Supplementary Fig. S3D). These results are consistent with Fig. 1A and B showing sustained target binding and pERK modulation in these higher doses. Together, these studies demonstrate that Compound A inhibits KRAS G12C, downregulates MAPK signaling and glucose metabolism, and has significant anti-tumor efficacy in MiaPaCa2 xenografts.

Although KRAS is mutated in ∼90% of pancreatic cancers, the MiaPaCa2 model represents a rare population of pancreatic ductal adenocarcinoma (PDAC), as G12C mutations only account for ∼1% of KRAS mutant PDAC (3). In contrast, KRAS G12C is the most prevalent KRAS mutation in NSCLC. Further, compared to xenografted human tumors in immune-deficient mice, GEMMs enable the study of primary tumor formation and progression in an immune-competent setting to better model the complex tumor microenvironment. However, genetic engineering of the murine Kras locus to model lung cancer in mice has to date been focused on the Kras G12D mutation. To generate a clinically relevant model with the targeted Kras G12C driver mutation, we performed genetic editing of the Kras^LSL-G12D/wtp53^fl/fl knock-in mouse line (18, 19) to mutate Kras G12D to Kras G12C. Tumor development in the resulting animals was induced by administering Cre-expressing Adenovirus via intranasal inhalation, and subsequent tumor growth in the lungs was monitored by longitudinal microCT imaging. Tumor burden was evaluated by both manual image assessment (Fig. 2) and an automated quantitation algorithm, Mouse Lung Automated Segmentation Tool (MLAST; ref. 11). On the basis of the visual assessment, animals were assigned a score of 0 (no tumor), + (small, isolated tumor nodules), ++ (large, distinct tumor nodules), or +++ (significant tumor burden with coalescing nodules) as shown in Fig. 2A. Manual assessment and scoring of the tumor-bearing animals show that at week 19 post-induction ∼75% of animals on study had no detectable tumors and <25% of animals were presenting score + tumors. As expected, the number of animals presenting tumors and tumor sizes increased at later timepoints (Fig. 2B).

We compared the novel G12C-driven lung model with its G12D counterpart and observed a marked difference in tumor latency. Small G12D-driven tumor nodules were visible in CT images around 12 weeks post-induction, whereas G12C-driven tumor nodules were first observed around 18 weeks post-induction. Average MLAST scores began to steadily decrease at 18 to 21 weeks in the G12D KP mice and 25 to 27 weeks in the G12C KP mice (Fig. 3A; Supplementary Fig. S4A). The G12C KP animals survived significantly longer (P = 0.01, log-rank Mantel–Cox), where the median survival was 36.6 weeks post-induction, compared to 27.7 weeks in the G12D KP animals (Fig. 3B). Although the time to tumor onset and overall survival was longer in the G12C mice, once tumors were detected, the rate of progression towards death was not noticeably different. Morphologically, both models developed individual nodules which grew and coalesced into large grade 3 to 4 adenocarcinomas. Histologically, both models were comprised of a mixture of acinar and papillary subtypes, but G12C-driven tumors had a higher frequency of poorly differentiated areas and little-to-no mucinous subtype (Fig. 3C). Intercellular bridging, an early indication of differentiation to squamous cell carcinoma, was also rarely observed. IHC for pERK, pAKT, and pS6 showed significantly lower pAKT in the G12C KP animals but also modest decreases in pERK and pS6, suggesting an overall trend in reduced RAF–MEK–ERK and PI3K–AKT–mTOR signaling in the G12C-driven tumors compared to their G12D-driven counterparts (Supplementary Fig. S4B). Lymphocyte proliferation was observed surrounding some larger tumor foci in 2 of 10 G12D tumors and in none of the G12C tumors. Given that this observation was based on histologic sections which represent a 2D sampling of a 3D specimen, this does not preclude the possibility that the G12C tumors may also exhibit lymphocytic infiltration.

To evaluate the glucose avidity of this model, ¹⁸F-FDG PET imaging was performed on animals with typical score ++ tumors. ROIs were manually drawn around discrete tumor nodules, and, as control, in a healthy lung area of the same animals. As shown in Fig. 4, while normal lung tissues demonstrated an average of 2.74%ID/g of ¹⁸F-FDG uptake, nodules of the G12C KP tumors showed an average of 4.37%ID/g uptake (P < 0.01). Similar results were seen in the G12D KP tumors.

To assess the PD effect of Kras G12C inhibition in the G12C KP GEMM, mice with distinct lung tumor nodules were treated with 30 mg/kg of Compound A or vehicle for 7 consecutive days. ¹⁸F-FDG PET imaging was performed at baseline and 1 hour post-final dose (“post-treatment”). As shown in Fig. 5 and Supplementary Fig. S5, all animals demonstrated a moderate baseline level of tumor ¹⁸F-FDG uptake. At the post-treatment timepoint, visual examination of the accompanying CT images indicated varied degrees of tumor progression in the vehicle-treated animals (Supplementary Fig. S5). Consistently, these mice showed slight increases in their tumor ¹⁸F-FDG uptake compared to baseline (average post-to-baseline fold change: 1.18; Fig. 5B and C). In contrast, CT images showed most of the animals (5 of 7) receiving Compound A had either stable disease or tumor reduction (Supplementary Fig. S5), and their tumors showed either no change or a decrease in ¹⁸F-FDG PET signal at the posttreatment timepoint (average post-to-baseline fold change: 0.87; Fig. 5B and C). Notably, two animals in the inhibitor-treated group appeared to be progressors/non-responders, exhibiting evident tumor growth and increased ¹⁸F-FDG uptake (Supplementary Fig. S5). These two mice were categorized in a separate subgroup for the analyses and are examples of potential intrinsic resistance and tumor heterogeneity in the G12C KP GEMM. IHC examination of tumor-bearing lung tissues revealed a slight trend of decreased pERK in Compound A-treated animals (Fig. 5D; did not reach statistical significance), but no changes were detected in pAKT or pS6. Interestingly, the two non-responders also showed reduced pERK staining. The exact mechanism warrants additional assessment.

To examine whether treatment with KRAS G12C inhibitors could achieve anti-tumor efficacy in endogenous primary lung tumors, a cohort of G12C KP animals with distinct tumor nodules was randomized into three groups and treated daily with vehicle, 30 mg/kg of Compound A, or 100 mg/kg of sotorasib (Fig. 6). Animals were imaged every 2 to 3 weeks by microCT to monitor disease progression. After 5 weeks of treatment, examination of lung sections revealed a marked reduction in both tumor incidence and tumor burden (Fig. 6A–C). 86% (12/14 mice) of the vehicle group had detectable tumors, and only 43% (6/14 mice) and 31% (5/16 mice) of the Compound A and sotorasib groups had detectable tumors, respectively (Fig. 6B). Furthermore, visual analysis of microCT images pre- and post-treatment showed that the vehicle group had progressed towards scores ++ and +++, whereas more G12C inhibitor-treated mice had shifted towards scores 0 and + (Fig. 6D). These results were likewise supported by MLAST quantification where the percentage of healthy lung tissues decreased in the vehicle group but was stable in the inhibitor-treated groups (Fig. 6E; Supplementary Fig. S6A). IHC did not show statistically significant changes in pERK or pAKT but did reveal a significant change in pS6 in the G12C inhibitor-treated samples compared to vehicle-treated samples (Supplementary Fig. S6B). The discrepancy in pERK and pS6 modulation between these tumors and those from the PET study described earlier (Fig. 5D) is likely due to differences in the timing of sample collection. Tumors from the PET study were collected 1 hour after the final dose of a 7-day regimen, whereas tumors from the efficacy study were collected 24 hours after the final dose of a 34-day regimen. It is possible that modulation of pERK levels in the GEMM had recovered by 24 hours or that the ERK pathway was no longer as sensitive to KRAS G12C inhibition after 5 weeks of treatment.

The duration of response to G12C inhibition was tested in a follow-up study where animals were treated with vehicle or Compound A for approximately 3 months or until a survival endpoint was reached (Supplementary Fig. S7; Supplementary Table S2). Median survival of the vehicle group was 61 days post-treatment initiation, whereas the median survival remained undefined for the inhibitor-treated group as most of these animals exhibited reduced tumor burden in the lungs and did not meet disease-related euthanasia criteria. Because of the advanced age of the animals in this study, several in both the vehicle and inhibitor-treated groups had to be euthanized due to age-related, non-tumor health complications (deaths were censored). Still, a significant difference in survival was observed between the two groups (P < 0.0001), where only 3 of 17 animals progressed while being treated with Compound A compared to 11 of 17 that were treated with vehicle.

With sotorasib and adagrasib progressing through clinical trials, KRAS is no longer the elusive target it was once considered. Identification of the allosteric switch-II binding pocket in the RAS GDP-bound form and subsequent discovery of small molecules with the ability to covalently bind the cysteine 12 residue (4–6, 20) have greatly improved the outlook for patients with KRAS G12C-driven cancers. Specifically, for KRAS G12C NSCLC, the sotorasib phase II trial (CodeBreaK100) reported a 37.1% objective response rate with median progression-free survival of 6.8 months and median duration of response of 11.1 months (21). The adagrasib phase I to II trial (KRYSTAL-1) reported an objective response rate of 45% with median duration of response of 8.5 months (8). Although the emerging data show clinical benefit for many patients, it is also clear that responses are not permanent, and more than half of patients have intrinsic resistance or develop resistance to the inhibitors (8).

As is common with preclinical drug discovery, we first utilized a traditional xenograft model (MiaPaCa2) to interrogate in vivo efficacy and PK/PD of Compound A, but such models are often easier to treat and represent best-case scenarios for the clinic. Indeed, we show that MiaPaCa2 tumors are sensitive to KRAS G12C inhibitor treatment, and Compound A achieved efficacy that was comparable to the published preclinical MiaPaCa2 data with sotorasib and adagrasib (4, 5, 20). Analysis of MiaPaCa2 tumor growth in individual animals showed uniform responses (Supplementary Fig. S3B), which was less obvious when tested in endogenous lung tumors arising in the Kras G12C-driven GEMM (Supplementary Fig. S6). Given the growing need to understand tumor heterogeneity and drug resistance, and the role of the immune microenvironment in the context of inhibitor treatment (7, 22), GEMMs offer advantages that may bridge preclinical and clinical observations.

The Kras-driven GEMMs that were first pioneered two decades ago by Jacks and Tuveson (18, 23) are widely used in the literature, but these models utilize the G12D mutation, which is only the third most prevalent KRAS mutation subtype in NSCLC (2, 3). To our knowledge, the G12C KP (Kras^LSL-G12C/wtp53^fl/fl) mouse model we described here is the first reported instance of a G12C knock-in mouse model, where one allele of the endogenous wild-type Kras has been replaced with Kras G12C. This contrasts with another recently reported KRAS G12C mouse model that was generated by inserting one copy of human KRAS G12C in the Col1a1 locus while still maintaining two wild-type alleles of mouse Kras (1). Targeting of Kras in its endogenous locus allows for physiologic levels of gene expression and retention of native gene regulatory elements that may be important for recapitulating signaling and potential development of resistance. Although our mouse model does not express human KRAS G12C, the protein is highly conserved between mouse and human.

We characterized the novel Kras G12C GEMM by comparing its tumor growth and morphology to that of its Kras G12D counterpart. Both models use identical gene targeting strategies and are in the same strain of C57BL6 mice, allowing a direct comparison of the tumorigenic potential. We observed a significant difference in overall survival and tumor latency indicating Kras G12C signaling was slower to initiate tumors than G12D. Staining for biomarkers in the RAF–MEK–ERK and PI3K–AKT–mTOR pathways revealed that Kras G12C induced lower levels of signaling in both pathways, although only the decrease in pAKT was statistically significant. Interestingly, the reduced signaling through ERK and AKT that we observed in our Kras G12C knock-in compared to the Kras G12D knock-in was not corroborated by the findings of Li and colleagues, which reported similar levels of pAKT and higher levels of pERK in their KRAS G12C model when compared to a Kras G12D (1). Because their KRAS G12C model was generated with a different targeting strategy, it is plausible that regulation of gene expression is altered in their model. Notably, the latency and signaling differences we observed in our models is consistent with inherent differences in biochemical activity and signaling that have been previously reported for different KRAS mutants (24). Studies comparing human NSCLC cell lines driven by KRAS G12C or KRAS G12D have shown that cell lines driven by G12D have higher activation of PI3K–AKT and MEK signaling, whereas cells driven by G12C have decreased levels of AKT signaling and preferentially signal through the RAL pathway (25, 26).

In addition to conventional endpoints such as TGI and survival, we took advantage of multiple in vivo imaging technologies to examine tumor growth and response to therapy. Longitudinal microCT imaging and accompanied manual and automated analyses provided valuable tools for rapid and quantitative in-study assessment of lung tumor burden, without the need of interim takedowns. We further employed ¹⁸F-FDG PET functional imaging to evaluate glucose uptake as one of the PD effects of Kras inhibition. This biomarker was selected based on reports illustrating the KRAS–ERK–glucose metabolism axis (27–30). Both MiaPaca2 xenograft and G12C KP GEM models showed a decrease in tumor ¹⁸F-FDG uptake when treated with Compound A, although this response was more ubiquitous and evident in MiaPaca2. The spontaneously formed lung tumors may have presented a more challenging model to treat, possibly due to a more pro-tumor parenchymal environment. Indeed, the 3-day treatment regimen that was effective in inducing PD response in MiaPaca2 tumors was ineffective in the GEMM, for which a prolonged, 7-day regimen was needed to enable PD analysis. The G12C KP tumors also exhibited greater inter-individual heterogeneity than the xenograft model, and two of the lung tumor-bearing mice were clearly nonresponders based on both ¹⁸F-FDG uptake and volumetric assessment of tumor burden. Nonetheless, the trend of reduced glucose utilization upon Kras inhibition was consistent between the two models. Of note, we did not observe differential expression of GLUT1 in the tumor following Kras inhibition. Therefore, the mechanism of reduced glucose uptake is yet to be elucidated, and the dependence of these Kras-driven tumors on glutamine metabolism and the animals’ overall glucose homeostasis after prolonged treatment deserve future investigation.

¹⁸F-FDG is a widely accepted probe for assessing tumor burden in patients in the clinic (14–16). Thus, preclinical ¹⁸F-FDG PET imaging is readily translatable and bears good clinical relevance. Of note, this probe is often used in the clinic to quantify tumor burden systemically and identify (occult) cancer metastases either at baseline or post-therapy. This is based on the assumption that unlike the surrounding normal tissues, most cancerous cells are glucose/FDG avid. Here, we extended the utility of ¹⁸F-FDG beyond tumor burden assessment and validated its capability in monitoring the metabolic changes following Kras inhibition. Importantly, all imaging results presented here are normalized to tumor volume/weight, which ensures that any reduction in ¹⁸F-FDG uptake value reflects a true decrease in glucose metabolism, rather than a mere reduction in tumor volume.

In summary, we presented a comprehensive in vivo framework to evaluate the pharmacology of KRAS G12C inhibitors using a multitude of in vivo imaging modalities in addition to PK, efficacy/survival, and histopathology. We validated an internal KRAS G12C-specific inhibitor (Compound A) in a conventional human xenograft model as proof-of-concept and then applied the workflow to a novel genetically engineered Kras G12C knock-in mouse model of NSCLC. We characterized tumor development of the G12C KP GEMM and benchmarked the anti-tumor effects of Compound A against sotorasib in the primary mouse lung tumors. This work showcased the relevance of a novel G12C-driven mouse model and the applicability of translational imaging technologies to understand the pharmacology and mechanisms of response in drug development. The G12C KP GEMM will be valuable for elucidating potential tumor intrinsic and extrinsic resistance mechanisms to new or existing clinical KRAS G12C inhibitors, as well as immunomodulatory drugs. The workflow and tools described will likewise facilitate testing of combination therapies to overcome resistance or improve patient responses to treatment.

All authors are or were employees of Pfizer, Inc. when the work was conducted. No other financial support was received by the authors for this work. All authors report employment with Pfizer, Inc. No disclosures were reported by the other authors.

C. Lee: Conceptualization, investigation, methodology, writing–original draft, writing–review and editing. Z.K. Jiang: Conceptualization, data curation, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Planken: Conceptualization, investigation, methodology, writing–original draft. L.K. Manzuk: Data curation, investigation, methodology. R. Ortiz: Data curation, investigation, methodology. M. Hall: Data curation, investigation, methodology. K. Noorbehesht: Investigation, methodology. S. Ram: Investigation, methodology, writing–original draft. T. Affolter: Conceptualization, investigation. G.E. Troche: Investigation, methodology. N.T. Ihle: Conceptualization, investigation, methodology. T. Johnson: Investigation, methodology. Y. Ahn: Investigation, methodology. M. Kraus: Conceptualization, resources, supervision. A. Giddabasappa: Conceptualization, resources, supervision, investigation, methodology, writing–original draft, writing–review and editing.

We thank the Pfizer - Comparative Medicine (La Jolla) technical staff for their assistance with experiments. The funding for this research was provided by Pfizer Inc.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.