Supplemental Table 1: Published Cellular IC50 values of PI3K inhibitors

KRASG12R-mutant cell lines exhibit reduced migration and collagen deposition. A, Representative images of migration assays from a panel of KRASG12D and KRASG12R PDAC cell lines. Scale bar, 50 μm. B, Box-and-whisker plot showing quantification of migration assays from A. Mean values are plotted, with each data point representing the percentage of area migrated from at least three technical replicates from three separate experiments for each cell line. C, Number of metastatic sites in each patient across indicated KRAS mutations. D, Schematic of the rapid autopsy program and how total liver involvement is examined for invasion. E, Liver involvement of metastatic lesions. For C and E, mean values are plotted, with each data point represented by a circle. KRASG12D, n = 37; KRASG12R, n = 17. F, Significant Molecular Signatures Database Gene Ontology pathway collagen containing extracellular matrix genes and all collagen genes from GSEA comparing PDX KRASG12D tumors with KRASG12R tumors. NES and P values are listed by corresponding pathways. G, Representative brightfield and polarized light images of collagen structures of KRASG12D/R using PSR. Polarized light images are representative and at 20× magnification. Scale bar, 300 μm. H, Box-and-whisker plot showing quantification of PSR in KRASG12D/R resected tumor samples (n = 3). Mean values are plotted, with each data point representing the percentage of the area of collagen normalized to the total tumor size. All P values were calculated using the Student t test comparing G12D and G12R. *, P < 0.05; ****, P < 0.00001; ns, nonsignificant, P > 0.05. Error bars, mean ± SEM.

The PDAC KRAS transcriptional signature is weakly activated in KRASG12R-mutant model systems. A, Volcano plot with changes in the 200 KRAS-dependent upregulated (red) and downregulated (blue) genes in the PKS in KrasG12D/+;Pdx1-Cre 4-week-old murine pancreas compared with KrasWT;Pdx1-Cre. All nonsignificant changes are denoted in gray (Padj. > 0.05), and lighter colors indicate nonsignificant <0.5 log2 fold change. The dotted lines represent the cutoff of 0.05 adjusted P value and ±0.5 log2 fold change. B, Volcano plot with PKS changes in KrasLSL-G12R/+;Pdx1-Cre 4-week-old murine pancreas compared with KrasWT;Pdx1-Cre control pancreas using methods outlined in A. C, Volcano plot with changes in PKS in ductal pancreatic organoids KrasLSL-G12D/+:Cas9-P2A-Cre & Trp53 sgRNA (KP-G12D) compared with KP-WT using methods outlined in A. D, Volcano plot with changes in PKS in KP-G12R ductal pancreatic organoids compared with KP-WT using methods outlined in A. E, Volcano plot showing PKS changes in the KRASG12D-expressing hTERT-HPNE E6/E7 cell line in comparison with the HPNE-EV control using methods outlined in A. F, Volcano plot showing PKS changes in the KRASG12R-expressing hTERT-HPNE E6/E7 cell line in comparison with the HPNE-EV control using methods outlined in A. G, Quantification of BRET measurements between CRAF and KRASG12D, KRASG12R, or KRASS17N. BRET signal serves as a proxy for binding affinity. Data represent the average of three independent experiments, each with eight technical replicates. H, Representative immunoblot of HA-KRASWT/G12D/R-expressing HPNE cells that were fractionated using a subcellular protein fractionation kit. Normalized portions of each extract were loaded for Western blot analysis (n = 3). C, cytoplasmic fraction; M, membrane fraction; N, nuclear fraction. I, Box-and-whisker plot quantifying nuclear pERK1/2 levels from HPNE subcellular fractionation. pERK1/2 was normalized to the loading control in the nuclear fraction. Murine RNA data were extracted from the Sequence Read Archive under accession PRJNA578549. GSEA was used to compare PKS between KRASG12D and KRASG12R. All comparisons were significant (P < 0.01) except for HPNE PKS UP. P values were calculated using one-way ANOVA. *, P < 0.05; ***, P < 0.0001. Error bars, mean ± SEM. Statistical tests were performed using R version 4.2.2.

Whole-body KrasLSL-G12R expression via Rosa26Cre-ERT2 drives heterogeneous hyperplasia in multiple tissues and increases endocrine pancreatic vacuole formation. A, Schematic of the KRRC. B, KRRC mice and littermate controls between 24 and 28 days received tamoxifen at 1 mg/10 g of body weight for 5 consecutive days via i.p. injections. C, Kaplan–Meier survival curves for KDRC mice: n = 11 (seven males and four females), Rosa26Cre-ER/LSL-EYFP control: n = 6 (three males and three females), KRRC mice: n = 5 (two males and three females), and KRRC controls: n = 2 (one male and one female). D, H&E staining of KRRC mouse tissues from the lung, spleen, liver, small intestine, and pancreas. A small tumor was detected in a single lobe of the lung in one KRRC mouse (black arrow). No major differences were observed in the spleen and liver between KRRC and control mice. Hyperplasia of the small intestine was observed in 2/5 KRRC mice (black arrow). KRRC mice also exhibited increased vacuole presence within exocrine cells of the pancreas (black arrow). Scale bar, 50 μm. E, Lung sections were immunostained for pErk1/2 and Ki-67 and false-colored. Images are representative of two Rosa26Cre-ERT2 control mice (one male and one female) and five KRRC mice (three males and two females). Scale bar, 100 μm. F, Box plot showing the average intensity of pErk1/2 in Rosa26CreERT2/+ and KrasLSL-G12R/+;Rosa26Cre-ERT2/+ mice. Five fields of view were taken from each lung. P values were calculated using a Student t test comparing control mice with G12R. *, P < 0.05. Error bars, mean ± SEM. G, Box plot showing the average intensity of Ki-67 in the pancreas of Rosa26Cre-ERT2/+ and KrasLSL-G12R/+;Rosa26Cre-ERT2/+ mice. Five FOV from each mouse (two control and three KRRC). P values were calculated using a Student t test comparing control mice with KRRC. Error bars, mean ± SEM. All statistical tests were run using R 4.2.2. ns, nonsignificant. Created in BioRender. Burge, R. (2025) https://BioRender.com/jv6j4ns.

Full body KrasLSL-G12D;Rosa26Cre-ER/LSL-EYFP increases endocrine pancreatic vacuoles and drives changes in the lung and spleen.

KRASG12R-mutant PDAC exhibits a distinct TME. A, Schematic of the PDX model. B, Heatmap of NESs from GSEA hallmark analysis using the Molecular Signatures Database (MSigDB) of PDX KRASG12D/R, compared with transcriptomics from three healthy pancreas samples. C, Significant MSigDB hallmark pathways from GSEA comparing PDX KRASG12D tumors with KRASG12R tumors. D, Volcano plot showing PKS changes in KRASG12D PDX PDAC compared with normal pancreas using methods outlined in Fig. 4A. E, Volcano plot of RNA from KRASG12R PDX PDAC tumors compared with normal pancreas, as outlined in Fig. 4A. F, Significant hallmark pathways in the TME of PDX-PDAC KRASG12D compared with PDX PDAC KRASG12R using GSEA hallmark analysis. NES and adjusted P values are listed for all corresponding pathways. G, Significant hallmark pathways in the TME of PDX PDAC KRASG12R compared with PDX PDAC KRASG12D using GSEA hallmark analysis. NES and adjusted P values are listed for all corresponding pathways. ROS, reactive oxygen species.

Engineering a KrasLSL-G12R/+ murine model.

The frequency of co-mutations in tumor suppressor genes is not dependent on KRAS mutation status in human pancreatic cancer patients in the AACR Project GENIE dataset.

Ubiquitous expression of all four PI3K isoforms in PDAC

The TGIF1 transcription factor gene is present on chromosome 18, which is subject to whole chromosome copy number reduction in colon cancer. Despite this, TGIF1 expression is significantly higher in cancer than in normal. In mice complete deletion of Tgif1 reduced tumor burden in an Apc mutant model of intestinal cancer. Here we show that reducing TGIF1 expression in a human colon cancer cell line slows proliferation and reduces growth of orthotopic xenografts. To ask if additional genes with copy number loss are more highly expressed in tumors we identified chromosomal regions subject to copy number reductions from ten TCGA cancer datasets. Within these regions a small proportion of genes, generally less than 10%, are expressed at higher levels in the tumor than in corresponding normal samples. Enrichment analysis using a set of 435 genes that have copy number reduction and increased expression identified mitosis as the most enriched gene set and FOXM1 and E2F family transcription factors as potential regulators. For mitotic genes, the average expression increase in tumor compared to normal is independent of copy number. In contrast, while DepMap common essential genes are generally more highly expressed in cancer than normal tissue, the relative increase in expression tracks well with copy number. Similarly, expression differences for gene sets such as S-phase, rRNA processing and DNA repair show increased expression in cancer versus normal, but changes also track with copy number. Thus, genes with increased expression despite copy number reduction may represent the output of key pro-tumorigenic transcriptional programs and could be potential therapeutic targets.

KRASG12D–mutant cell lines have increased proliferation rates compared to KRASG12R.

Glucose and glutamine restriction enhances macropinocytosis, PI3K signaling, and PTEN oxidation in PDAC cell lines

KPC cell lines utilize direct KRAS-PI3K–mediated signaling. A, Immunoblot of KRAS mutants expressed in hTERT-HPNE E6/E7 cell lines for 48 hours with doxycycline-induced KRAS expression. The image is representative of three independent experiments. B, Immunoblot of NIH “RASless” MEFs expressing BRAFV600E and KRASWT/G12D/R. The image is representative of three independent experiments. C, Immunoblot analysis from murine KPC cell lines after inhibition using the KRASG12D-specific inhibitor MRTX1133 at 20 nmol/L for 4 hours. Quantification of pAKTT308 after inhibition is shown underneath the blot. Blots are representative of two independent experiments. D, Immunoblot analysis of human KRASG12D PDAC cell lines after inhibition with the KRASG12D-specific inhibitor MRTX1133 at 20 nmol/L for 4 hours. Quantification of pAKTT308 after inhibition is shown underneath the blot. Data are representative of three independent experiments. E, Immunoblots with human KRASG12D PDX VMP PDAC cell lines after inhibition with the KRASG12D-specific inhibitor MRTX1133 at 20 nmol/L for 4 hours. Quantification of pAKTT308 after inhibition is shown underneath the blot. Data are representative of two independent experiments. F, Box-and-whisker plot with quantification of pAKTT308 in murine PDAC from C and human PDAC from D and E. P values were calculated using a Student t test comparing MRTX1133-treated cells with vehicle-treated cells. ***, P < 0.0002; ns, nonsignificant, P > 0.05. Error bars, mean ± SEM. Statistical tests were run using R 4.2.2.

Animals used for KRAS mouse models

KrasLSL-G12R/+ induced pancreatic hyperplasia and vacuole formation with limited tumor development after 1 year. A, Graph showing the percentage of KP48 mice that developed pancreatic tumors. The number of mice (n) is indicated above each corresponding bar. B, Box plot of terminal body weights of male mice that were sacrificed at the determined endpoint and had tissues harvested. KDP48 mice (n = 6), KRP48 mice (n = 6), and control mice (n = 6). Female mice followed similar trends. Data are presented as mean weight at endpoint with no statistically significant differences in body weight. C, Box plot of pancreas weights normalized to total mouse weight at the time of sacrifice. KDP48 mice had significantly larger pancreata in comparison with control and KRP48 mice. P values were determined using one-way ANOVA with Tukey multiple comparison test. ****, P < 0.0001. D, Representative gross images of the pancreas. KRP48 (no tumor) has an increase in islet cells (H&E 1) and proliferation of ductal epithelial cells (H&E 2, black arrow). Both KDP48 (tumor-bearing) and KRP48 (tumor-bearing) mice show increased periductal fibrosis. Representative immunostains for CK19 (ductal marker), αSMA (fibroblast), Ki-67 (proliferation), and Alcian blue–periodic acid-Schiff (AB-PAS) stain (glycogen and zymogen granules). H&E images were analyzed by a veterinary pathologist. Scale bar, 50 μm.

Pancreas-specific coexpression of KrasLSL-G12D/+;Trp53LSL-R172H/+, but not KrasLSL-G12R/+;Trp53LSL-R172H/+, drives invasive PDAC. A, Schematic of the generation of the KrasLSL-G12R allele. NEO, neomycin resistance cassette and was used as a selectable marker. B, Schematic of the KrasLSL-G12R/+;Trp53LSL-R172H/+;p48Cre-ERTM mouse model (KP48). C, Mice and littermate controls (24–28 days old) received tamoxifen (1 mg/10 g body weight) via i.p. injection for five consecutive days. D, Kaplan–Meier survival curves for controls (n = 26), KDP48 (n = 26), and KRP48 (n = 20) mice. E, Gross images of the pancreas (panc) and spleen (spl). Pancreas sections were additionally stained with H&E, immunostained for pErk1/2 and Kras, and stained with Masson’s trichrome. Images are representative of six mice from each cohort (three male; three female). Immunostained markers were false-colored brown for visualization. H&E images were analyzed by a veterinary pathologist. Scale bar, 100 μm. Created in BioRender. Burge, R. (2025) https://BioRender.com/jv6j4ns.

Pancreatic ductal adenocarcinoma (PDAC) is defined by the near universal occurrence of KRAS mutations. The KRASG12R mutation is detected in approximately 15% to 20% of patients with PDAC and rare in other KRAS-mutant cancers. KRASG12R is unable to activate the lipid kinase PIK3CA, suggesting that alternative mechanisms might be employed to activate KRASG12R-independent PI3K signaling in PDAC. In this study, we detected elevated expression of all four PI3K isoforms in PDAC cell lines, with the PIK3CG isoform showing higher overall expression in KRASG12R-mutant PDAC. All four PI3K isoforms contributed to global PI3K signaling, and inhibition of any single isoform was insufficient to limit PDAC proliferation. The combined inhibition of all PI3K isoforms was required to limit proliferation, providing a potential explanation for the limited efficacy of PI3K inhibitors in the clinic. Additionally, PTEN, a negative regulator of PI3K signaling, was inactivated in PDAC by the formation of an intramolecular disulfide, which elevated overall PI3K signaling and reduced the dependency of PI3K signaling on KRAS. Oxidation of PTEN was independent of KRAS mutation status. Finally, nutrient-limiting conditions mimicking the PDAC tumor microenvironment further elevated PTEN oxidation and significantly increased macropinocytosis. Thus, this study uncovered a mechanism that supports elevated PI3K signaling in PDAC, thereby reducing the need for KRAS to directly activate the PI3K pathway. SIGNIFICANCE:PTEN inactivation by intramolecular disulfide bond formation and elevated expression of PI3K isoforms in pancreatic cancer leads to unchecked KRAS-independent PI3K signaling, highlighting the need for therapeutic approaches targeting constitutive PI3K signaling. See related commentary by Tiriac and Engle, p. 1817 See related article by Burge et al., p. 1868 See related article by Kamgar et al., p. 2042.

Background: Clinical staging of American Joint of Cancer Committee (AJCC) tumor-node-metastasis (TNM) system for intrahepatic cholangiocarcinoma (ICC) is under ongoing debate. Perineural invasion (PNI) represents an adverse prognostic factor for ICC. The aim of this study was to propose a new T category for early-staged ICC by incorporating PNI. Methods: Patients undergoing curative-intent surgical resection for ICC were identified from an international multi-institutional database. T category for early-staged ICC was modified based on combination of tumor size, vascular invasion and PNI status. Disease-free survival (DFS) was utilized to assess prognostic differences among redefined stage groups. Results: A total of 307 stage T1-2 ICC patients with solitary mass-forming tumor were included for analysis. Multivariable Cox regression analysis identified that large tumor size (>5 cm) and PNI were independently associated with worse DFS. A modified AJCC T category (mT) was proposed: mT1 consists of tumor sizing <= 5 cm; mT2a consists of tumor sizing >5 cm; mT2b consists of tumor with vascular invasion and/or PNI. Among the updated mT2b group, 19 patients had concomitant vascular invasion and PNI, while 35 and 20 patients had vascular invasion alone and PNI alone, respectively. Patients with mT2b showed significantly inferior DFS than mT1 (25.0 months vs. not reached, P=0.001), yet an almost identical DFS outcome as mT2a (25.0 vs. 18.0 months, P=0.94). Conclusions: The new mT category better stratifies prognostic groups for early-staged ICC in comparison to the current system. These findings should be validated and considered in the future update of staging criteria for ICC.

BACKGROUND:Recurrence is a major driver of poor long-term outcomes after curative-intent resection for intrahepatic cholangiocarcinoma (iCCA), yet the association between postoperative pathology, first recurrence patterns, and post-recurrence outcomes remains unclear. METHODS:Patients who underwent curative-intent resection for iCCA (2000-2023) were identified from an international multi-institutional database. First recurrence patterns were classified as intrahepatic-only, extrahepatic-only, or combined intrahepatic and extrahepatic recurrence. Multivariable analyses assessed associations between postoperative pathological features, first recurrence patterns, post-recurrence survival (PRS), and post-recurrence curative-intent treatment. RESULTS:Among 1328 patients, 763 patients (57.5%) developed a recurrence; 717 patients had a classifiable first recurrence pattern (381 patients developed intrahepatic-only recurrence, 171 patients developed extrahepatic-only recurrence, and 165 patients developed combined intrahepatic and extrahepatic recurrence). Three or more metastatic lymph nodes (adjusted OR (aOR) 3.47 (95% c.i. 1.56 to 7.72)) and microvascular invasion (aOR 2.49 (95% c.i. 1.61 to 3.85)) were associated with higher odds of combined recurrence compared with intrahepatic-only recurrence, whereas perineural invasion (aOR 2.24 (95% c.i. 1.35 to 3.71)) and the absence of pathological nodal evaluation (aOR 2.04 (95% c.i. 1.24 to 3.36)) were associated with extrahepatic-only recurrence. Compared with intrahepatic-only recurrence, combined recurrence was associated with worse PRS (adjusted HR 1.63 (95% c.i. 1.29 to 2.06)) and lower odds of receiving curative-intent treatment (aOR 0.14 (95% c.i. 0.06 to 0.29)). CONCLUSION:Pathological nodal burden and invasive tumour features were associated with distinct first recurrence patterns after iCCA resection. Combined recurrence, more common with ≥3 metastatic lymph nodes or microvascular invasion, was associated with worse PRS and lower receipt of curative-intent treatment. Among patients who develop recurrence, pathological prognostic factors available after surgery may help characterize first recurrence patterns and inform risk-adapted postoperative surveillance.

BACKGROUND:This cross-sectional study evaluated risk-adjusted cumulative sum (RA-CUSUM), a tool for real-time monitoring of perioperative outcomes, in liver surgery across an international multicenter cohort. METHODS:Patients undergoing curative-intent hepatectomy for hepatocellular carcinoma (31%), intrahepatic cholangiocarcinoma (29.7%), or colorectal liver metastases (39.3%) across 25 centers (2000-2023) were analyzed. RA-CUSUM curves were generated for operative time, intraoperative blood loss, major complications (Clavien-Dindo ≥ III), and length of stay (LOS). Mixed-effects models were used for overall curves and center-specific regressions for institutional analyses. Secondary outcomes included R1 resection and 1-year recurrence comparing the first 50 and subsequent cases at each center. RESULTS:Among 5880 cases 38.2% were major hepatectomies, while 15.1% were minimally invasive. Median operative time was 240 min (IQR 150-356), median blood loss 300 mL (IQR 100-600), and median LOS 9 days (IQR 6-14). Major complications occurred in 35.1% of patients. RA-CUSUM curves demonstrated a biphasic pattern, with deterioration coinciding with the spread of minimally invasive surgery, and subsequent improvement. Center-level curves revealed marked heterogeneity. Secondary outcomes demonstrated no uniform early-period disadvantage. CONCLUSIONS:RA-CUSUM offers a real-time visualization of surgical performance and can guide quality review and support targeted quality-improvement initiatives in liver surgery units.