Prognostic Implications of Codon-Specific KRAS Mutations in Localized and Advanced Stages of Pancreatic Cancer.

abstract cc by

PURPOSEAlthough KRAS mutations represent the primary oncogenic driver in pancreatic ductal adenocarcinoma (PDAC), the association between codon-specific alterations and patient outcomes remains poorly elucidated, largely because of a lack of data sets coupling genomic profiling with rich clinical annotations across disease stages.MATERIALS AND METHODSWe used American Association for Cancer Research's GENIE Biopharma Consortium Pancreas v1.2 data set to test the association of codon-specific KRAS mutations with clinicogenomic features and patient outcomes in patients with PDAC diagnosed with localized (stages I to III) and advanced disease (stage IV). Overall survival (OS) was compared using Kaplan-Meier and multivariable Cox proportional hazards methods.RESULTSAmong 1,032 eligible patients, 949 (92%) exhibited mutant KRAS. These mutations were predominantly observed at G12D (n = 390, 41%), G12V (n = 305, 32%), and G12R (n = 149, 16%). In the group of patients who presented with localized disease, those with G12V mutation had notably longer survival compared with G12D mutation (P = .03). By contrast, patients with G12V mutation who presented with metastatic disease experienced shorter OS compared with those with G12R (P = .04) and G12D mutations (P = .04). Furthermore, no significant differences were observed in the frequencies of coaltered driver genes, including TP53, CDKN2A, and SMAD4, across the different KRAS mutations.CONCLUSIONThese findings demonstrated that codon-specific KRAS mutations affect PDAC outcomes differently based on disease stage at diagnosis. As studies testing KRAS inhibitors continue to emerge and mature, the prognostic variability of individual KRAS mutations must be carefully considered to avoid confounding and ensure accurate evaluation of therapeutic efficacy in early-phase studies.

introduction cc by

Pancreatic ductal adenocarcinoma (PDAC) has a 5-year survival rate of approximately 10%, mainly because of the lack of effective methods for early detection and treatment.1,2 Over 80% of patients present with advanced disease at the time of diagnosis, and outside of clinical trials, standard treatment is cytotoxic chemotherapy that provides limited benefit.3,4 In patients diagnosed with localized disease, the combination of surgical resection and systemic chemotherapy can improve survival in select patients; however, most patients treated with curative intent develop disease recurrence.5 These dismal outcomes highlight the need for improved understanding of the biologic underpinnings of PDAC. KRAS is mutated in over 90% of all patients with PDAC, serving as the primary oncogenic driver in this cancer. These KRAS mutations predominantly occur at codon 12, with less frequent mutations at other codons. By far, the most common codon-specific mutations are G12D (approximately 40%), G12V (approximately 30%), G12R (approximately 15%), and Q61 (H, K, L, R; approximately 6%).6,7 In preclinical studies, the various codon-specific KRAS mutations have demonstrated unique biochemical properties, suggesting that they may similarly promote differential clinicopathologic features in patients.8,9 Decades of research on the prognostic role of specific KRAS mutations in various cancers have significantly deepened our understanding of their importance in predicting clinical outcomes.10-13 However, results remain conflicting regarding their prognostic role in pancreatic cancer, largely because of a lack of large data sets that couple genomic profiling with rich clinical annotations.6,7,14-19 Additionally, previous studies have typically focused on either early or late stages of the disease, rather than across all stages. The American Association for Cancer Research (AACR's) Project GENIE (Genomics Evidence Neoplasia Information Exchange) is a publicly accessible cancer registry that includes robust clinicogenomic data from multiple cancer centers. In this study, we conducted a retrospective analysis to elucidate the impact of codon-specific mutations on PDAC patient outcomes across disease stages, leveraging the detailed demographic-, clinical-, and treatment-level annotations which have not previously been available in other genomic profiling efforts.

context cc by

Key Objective To evaluate the prognostic significance of codon-specific KRAS mutations in patients with pancreatic ductal adenocarcinoma (PDAC) diagnosed at localized and advanced stages. Knowledge Generated Using the American Association for Cancer Research GENIE Biopharma Consortium Pancreas data set, which includes genomic and clinical data from over 1,000 patients, we analyzed survival outcomes by KRAS mutations. In localized disease, the G12V mutation was associated with significantly longer survival compared with G12D. However, in metastatic cases, G12V was linked to worse outcomes than G12D or G12R, revealing a reversal in its prognostic role depending on disease stage. Relevance As KRAS-targeted therapies enter clinical trials, understanding the distinct behavior of codon-specific mutations becomes increasingly important. In particular, the paradoxical prognostic impact of G12V, found in more than one third of PDAC patients, may confound treatment efficacy assessments and should be carefully considered in trial design, data interpretation, and outcome analysis.

patient selection cc by

We used the Pancreas v1.2-consortium multi-institutional and multilayer database from the AACR's GENIE Biopharma Consortium (BPC).20 The GENIE BPC PANC cohort comprises 1,130 samples obtained from 1,109 patients as previously reported.21,22 Samples with cancer types other than pancreatic adenocarcinoma, missing KRAS mutation data, those with the co-occurrence of KRAS mutations in the different categories were excluded (Data Supplement, Fig S1).

clinical and genomic data cc by

Clinical data were curated using the previously described PRISSMM framework.23 Patients were categorized on the basis of stage at diagnosis: localized (stages I to III) and metastatic (stage IV). The mutation data were extracted from four centers through Synapse using the cBioPortal interface. Sequencing details can be found in the GENIE data guide (GENIE 15.1 public release). This cohort included genomic data on 39 genes that were assessed in all oncopanels. Copy number variation (CNVs) data for these genes were available for 975 patients. The term altered was used for the tumor suppressor genes CDKN2A, SMAD4, and TP53 if any copy number loss or mutations were observed. We calculated the genomic alterations based on changes in CNVs or mutations in 35 nondriver genes, excluding KRAS, CDKN2A, SMAD4, and TP53. The Data Supplement (Table S1) lists all gene panels.

statistical analyses cc by

Descriptive statistics were used to summarize baseline patient and clinical characteristics. Comparisons between groups were performed using Chi-square or Fisher exact tests for categorical variables and the Wilcoxon rank-sum test for nonparametric continuous variables. False discovery rate correction was applied for multiple comparisons where appropriate. Survival probabilities were estimated using the Kaplan-Meier method, and differences in survival were compared using the log-rank test. Overall survival (OS) was defined as the time from initial diagnosis to either the date of death or last known follow-up, and patients alive at the last follow-up were considered censored in the analysis. The Cox proportional hazards model was used to assess prognostic factors and estimate hazard ratios (HRs) across codon-specific KRAS mutations while controlling for clinical features. All statistical analyses were two-sided, with a P value <.05 considered statistically significant. Analyses were performed using IBM SPSS software, v28.0.1.1, GraphPad Prism, v10.2.2, and Python.

patient characteristics cc by

A total of 1,032 patients were included, consisting of 464 female (45%) and 568 male (55%) patients, with a median age of 65 (IQR, 57-72) years. Patients data were derived from four centers: Memorial Sloan Kettering (MSK, 500, 48%), Dana Farber Cancer Institute (DFCI, 421, 40%), Vanderbilt Ingram Cancer Center (VICC, 63, 6.1%), and University Health Network (UHN, 57, 5.5%). Patients were relatively evenly split between those diagnosed with localized disease (565, 55%) and those diagnosed with metastatic disease (467, 45%). Complete baseline patient characteristics stratified by stage at diagnosis are listed in Table 1. KRAS mutations were present in 949 patients (92%) and absent in 83 patients (8%). Among the KRAS mutations, the most common was G12D, found in 390 cases (37.8%), followed by G12V in 305 cases (29.6%), G12R in 149 cases (14.4%), Q61 (H, R, K, and L) in 74 cases (7.2%), and other less frequent mutations in 31 cases (3%). A detailed list of KRAS mutations is provided in the Data Supplement (Table S2). Baseline patient characteristics according to codon-specific KRAS mutations are shown in Table 2.

genomic landscape of pdac in the genie data set cc by

To investigate the genomic alterations, we focused on a panel of 39 genes that had complete somatic mutation and CNV data across all four centers. Other than KRAS, the most commonly altered genes were TP53 in 759 (73.5%), CDKN2A in 402 (39%), and SMAD4 in 379 (36.7%). The genomic landscape, stratified by stage at diagnosis and codon-specific KRAS mutations, are summarized in Figure 1. Given that PDAC tumors often exhibit co-occurring alterations in these driver genes, we next assessed the frequency of genomic alterations in TP53, CDKN2A, and SMAD4 in the context of wild-type and mutant KRAS (Data Supplement, Figs S2A and S2B), as well as the most common codon-specific KRAS mutations (Data Supplement, Fig S2C-S2F). Compared with patients with wild-type KRAS, patients with KRAS mutations were more likely to harbor co-occurring alterations in TP53 (80.3% v 45.7%; P < .001), TP53 and CDKN2A (36.6% v 14.2%; P < .001), TP53 and SMAD4 (33% v 12.8%; P = .003), as well as alterations in TP53, SMAD4, and CDKN2A (20% v 7.1%; P = .004). The absence of any of these coalterations was less common in patients with mutant KRAS compared with wild-type KRAS (10.4% v 27.4%; P = .003). Notably, no significant differences in coalterations were observed between the codon-specific KRAS mutations (all P > .05). Moreover, alterations in TP53, CDKN2A, and SMAD4 genes, independent of KRAS mutation status, had a negative prognostic impact on survival (Data Supplement, Figs S3A-S3D). In patients with KRAS-mutant tumors, we observed that the absence of alteration in TP53, CDKN2A, and SMAD4 was associated with the best survival (28.3 months), whereas patients with tumors containing just one of these altered tumor suppressors had significantly reduced OS. Notably, the worst survival outcomes were observed in patients with tumors that harbored altered CDKN2A, with similar survival whether CDKN2A was altered alone (15.55 months), concurrently with SMAD4 (14.4 months), or concurrently with TP53 (14 months; Data Supplement, Fig S3E and Table S3). Thus, in patients with KRAS-mutant PDAC, loss of CDKN2A appears to confer a particularly dismal prognosis.

association of codon-specific cc by

Next, we investigated whether different KRAS mutations have distinct impacts on OS. In the main cohort (not stratified by stage), there were no statistically significant differences in OS based on codon-specific KRAS mutations (P = .2; Figs 2A and 2B), nor were the frequencies of coalterations in the common tumor suppressor genes significantly different among the codon-specific KRAS mutation groups (Fig 2C and Data Supplement, Table S4). Interestingly, however, codon-specific KRAS mutations were associated with distinct survival outcomes in patients diagnosed with localized (Figs 2D and 2E) compared with metastatic disease (Figs 2G and 2H), suggesting a prognostic role for these mutations based on disease stage. In patients diagnosed with stage I to III disease, the G12V mutation was associated with significantly longer survival compared with the G12D mutation (37.3 months v 27 months; HR, 0.69 [95% CI, 0.55 to 0.87]; P = .03; Fig 2E). We investigated whether this difference in survival could be explained by differential frequencies of TP53, CDKN2A, and SMAD4 coalterations among codon-specific KRAS mutations. However, these coalterations were not significantly different among the codon-specific KRAS mutation groups (Fig 2F and Data Supplement, Table S5). Conversely, in patients diagnosed with stage IV disease, the G12V mutation was associated with shorter survival (10.6 months) compared with G12D (12.6 months; HR, 1.37 [95% CI, 1.07 to 1.75]; P = .04) and G12R (15.5 months; HR, 1.54 [95% CI, 1.13 to 2.10]; P = .04) mutations (Fig 2H). Similar to all stages and stages I to III, no significant differences were observed for the frequencies of tumor suppresser alternations among codon-specific KRAS mutations (Fig 2I and Data Supplement, Table S6).

paradoxical impact of cc by

The G12D and G12V mutations are the most frequent KRAS mutations observed in PDAC, occurring in over 70% of cases. Our survival analyses revealed a paradoxical impact of G12V compared with G12D mutation in patients diagnosed with localized versus metastatic disease. We aimed to determine whether this observation was independently associated with KRAS codon status or influenced by other factors, such as genomic coalterations in TP53, CDKN2A, and SMAD4 genes, patterns of distant metastasis, or treatment regimens. To test the independent associations between G12V and G12D status and survival outcomes, we applied univariable and multivariable Cox regression models (Fig 3 and Data Supplement, Fig S4). In univariable analysis of patients diagnosed with stage I to III disease, the G12V mutation was associated with longer OS compared with G12D (HR, 0.69 [95% CI, 0.55 to 0.87]; P = .002). This association was conserved in the TP53 (HR, 0.69 [95% CI, 0.52 to 0.90]; P = .007) and the CDKN2A altered subgroups (HR, 0.53 [95% CI, 0.35 to 0.79]; P = .002), as well as across different treatment regimens. In patients with SMAD4 coalterations, those who eventually developed liver and thoracic metastasis, and in patients who underwent surgery, the G12V mutation was again associated with improved OS, but these differences did not reach statistical significance (all P > .05). Conversely, in univariable analysis of patients diagnosed with metastatic disease, G12V mutation was associated with worse survival compared with the G12D mutation (HR, 1.37 [95% CI, 1.07 to 1.75]; P = .009). This trend remained consistent in the presence of alterations in TP53 (HR, 1.39 [95% CI, 1.06 to 1.84]; P = .01), CDKN2A (HR, 1.57 [95% CI, 1.11 to 2.21]; P = .005), and SMAD4 (HR, 1.92 [95% CI, 1.31 to 2.80]; P = .001), as well as in patients diagnosed with liver metastases (HR, 1.48 [95% CI, 1.13 to 1.93]; P = .002) and patients who received any first-line systemic therapy (HR, 1.42 [95% CI, 1.09 to 1.83]; P = .005). No statistically significant difference in OS between G12D and G12V mutations were detected within stage IV groups that received specific systemic regimens, possibly because of the limited size of each of these cohorts. These findings were further confirmed with multivariable analysis using Cox proportional hazards models adjusted for age at diagnosis, sex, race, and treatment center (Data Supplement, Fig S4). The treatment profiles of G12D or G12V mutations are detailed in the Data Supplement (Tables S7-S9).

paradoxical impact of cc by

Considering the paradoxical impact of G12V mutation status on PDAC prognosis based on stage at diagnosis, we investigated whether the frequency of CNVs or mutations (referred here as genomic alterations) in the 35 nondriver genes evaluated across all gene panels correlated with different clinical outcomes. Interestingly, the frequency of genomic alterations in 15 of the 35 genes was significantly lower in G12V patients diagnosed with stage I to III disease compared with patients diagnosed with stage IV disease (Fig 4A). However, no significant differences in genomic alterations of nondriver genes were observed between patients with G12D mutation in the stage I to III subgroup versus those in the stage IV subgroup (Fig 4B). Next, we analyzed the overall genomic alterations across these 35 nondriver genes in the G12D and G12V subgroups by calculating the overall genomic alterations frequencies per group (Fig 4C). Independent of stage, there were no statistically significant differences in the genomic alteration frequency of nondriver genes between patients whose tumors harbored G12D or G12V mutations. We asked whether the frequency of alterations in nondriver genes was different when stratified by stage at diagnosis. Intriguingly, patients with G12V mutations diagnosed with localized disease had the lowest overall genomic alterations, whereas those with G12V diagnosed at advanced-stage had the highest number of genomic alterations (P < .001). By contrast, patients with G12D mutation diagnosed with stages I to III versus IV disease did not demonstrate significant differences in nondriver genomic alterations (P = .1). Thus, these results suggest that the accumulation of genomic alterations in patients with G12V mutations diagnosed at a later disease stage may explain the paradoxical impact of G12V mutation across localized and advanced stages and might have contributed to the poor prognosis of G12V patients diagnosed with stage IV disease.

other genomic predictor of survival cc by

To further elucidate genomic predictors of outcomes in PDAC, we performed univariable and multivariable analyses to assess associations between genomic factors and OS based on stage at diagnosis (Data Supplement, Tables S10 and S11). In patients diagnosed at stage I to III, univariable Cox regression demonstrated significant associations between OS and KRAS G12D, G12V, TP53, and CDKN2A alterations. However, no significant associations were found for KRAS G12R, Q61, other KRAS mutations, or SMAD4 alterations. In multivariable Cox analysis, KRAS G12D, G12V, and CDKN2A alterations remained significant for OS (Data Supplement, Table S10). In patients diagnosed at stage IV, univariable analysis found KRAS G12V, TP53, and CDKN2A alterations were significant predictors of OS. In multivariable analysis, all three genomic alterations remained significantly associated with worse survival (Data Supplement, Table S11). These findings suggest that in patient diagnosed with localized PDAC, KRAS G12D and altered CDKN2A may serve as predictive biomarkers for reduced OS, while G12V may indicate a more favorable prognosis. By contrast, in patients diagnosed with metastatic disease, KRAS G12V and alterations in TP53 and CDKN2A appear to confer the worst survival outcomes.

discussion cc by

We used data from AACR's Project GENIE to analyze 1,032 patients with PDAC and specifically test how codon-specific KRAS mutations are associated with co-occurring genomic alterations and PDAC outcomes across disease stages. We found that the KRAS G12V mutation is associated with significantly longer survival in patients diagnosed with localized PDAC, yet is associated with shorter survival in patients with metastatic disease at the time of diagnosis. Multivariable analyses confirm the paradoxical impact of the G12V mutation across different disease stages, regardless of metastatic patterns, use of different treatment modalities, or alterations in the common tumor suppressors TP53, CDKN2A, and SMAD4. Detailed analyses of additional genomic alterations indicate that changes in frequencies of nondriver genes may explain the paradoxical impact of the G12V mutation on patient outcomes in localized and advanced PDAC. Taken together, our findings highlight the importance of considering codon-specific KRAS mutations as key factors in PDAC prognosis based on stage at diagnosis. To date, studies evaluating the impact of codon-specific KRAS mutations in PDAC have produced conflicting results, in part due to the fact that these studies often evaluated patient populations that different based on stage at diagnosis or lacked important clinical annotations.6,7,14-16,19,24,25 More recently, studies have emerged that provide detailed depictions of more uniform patient populations. Consistent with our findings, several studies evaluating patients with localized disease have now reported that G12D mutation is associated with significantly worse outcomes.6,14,15 Most recently, McIntyre et al compared OS in 397 patients with PDAC diagnosed at stage I to III and found that those diagnosed at stage I with G12D mutation had significantly worse survival compared with patients with G12R and G12V mutations, consistent with the favorable prognosis of G12V in localized PDAC, but not G12R.16 This discrepancy for G12R mutation may be due to the limited patient numbers in their cohort and their focus on stage I patients, which had a higher frequency of KRAS G12R mutation. In patients diagnosed with advanced metastatic disease, we found that G12V was associated with worse OS. In alignment with our study, Cheng et al and Norton et al reported that the KRAS G12V mutation was associated with poor prognosis in advanced pancreatic cancer.26,27 By contrast, Pan et al highlighted the adverse impact of the Q61 mutations on survival in patients with stage IV PDAC,19 while another cohort linked G12D to poorer OS in stage IV disease.7 Additionally, several studies reported longer OS in patients with metastatic PDAC with the G12R mutation,27,28 which does not align with our findings. These discrepancies may reflect the fact that these patient cohorts were sourced from a single institute, as well as limited number of patients in the cohorts. The G12V mutation has shown variable prognostic significance depending on disease stage in KRAS-driven malignancies. In early-stage lung adenocarcinoma, G12V has been associated with better outcomes,29 consistent with our finding in localized PDAC. In advanced lung cancer, this mutation has been linked to worse prognosis compared with other KRAS-mutant subtypes.30-32 Similarly, the G12V has been associated with poor outcomes in metastatic colorectal cancer.11,33 These findings in advanced colorectal and lung cancers, along with our data, suggest that G12V may confer unfavorable outcomes in metastatic KRAS-driven cancers. Each KRAS mutation has distinct biochemical features that affect guanosine triphosphate hydrolysis, nucleotide exchange, and effector interactions, resulting in varied neoplastic phenotypes.34-39 Grimont et al recently reported that G12D more strongly promotes acinar-to-ductal metaplasia and PDAC progression than G12V or G12R. Although transcriptomic profiles were broadly similar, G12D triggered more sustained and pronounced changes over time, aligning with its more aggressive behavior in localized PDAC.40 The rapid development and adoption of targeted KRAS inhibition through various approaches, including G12C inhibitors,41,42 G12D inhibitor,43 pan-KRAS inhibitor,44 pan-RAS inhibitors,45,46 and KRAS degraders47 will almost certainly alter how PDAC is treated across stages. Individual KRAS mutations exhibit distinct prognostic implications across different cancer types and stages, which can confound outcome interpretation in early-phase clinical trials. If these differences are not accounted for, they may obscure the true therapeutic effects being evaluated. Therefore, stratifying patients or adjusting analyses based on specific KRAS codon variants is critical to accurately assess treatment efficacy. Notably, the frequency of genomic alterations reported in this study for KRAS, TP53, CDKN2A, and SMAD4 genes is consistent with other large data sets.14,19,28,48 Alterations in each of these tumor suppressor genes were individually associated with worse survival outcomes.15,16,19,27,49,50 Although we did not find any association between the codon-specific KRAS mutations and alterations in TP53, CDKN2A, and SMAD4 as the main driver genes in PDAC, we interestingly uncovered an association between the prognostic impact of KRAS G12V in localized and advanced disease stages and differential genomic alterations, primarily driven by CNVs or mutations in 35 nondriver genes in PDAC. However, future studies are required to find any association between codon-specific KRAS mutations and whole genomic alterations in PDAC. Our study's strengths include a large patient cohort from multiple institutions and a comprehensive analysis of diverse clinical variables across all disease stages. However, despite the large sample size, limitations remain. This is a retrospective study and detailed imaging and surgical data, and dynamic variables such as smoking status were not captured. Additionally, the heterogeneity of treatment data restricts us to assess the impact of codon-specific KRAS mutations on different chemotherapy, radiation, and investigational treatment responses. Despite these limitations, this study represents the largest effort to evaluate codon-specific KRAS mutations across stages of PDAC, leveraging robust clinical annotations to help adjust for confoundings.