KRAS Mutations in Cholangiocarcinoma: Prevalence, Prognostic Value, and KRAS G12/G13 Detection in Cell-Free DNA

Abstract

Background/Aim

Cholangiocarcinoma (CCA) is an aggressive hepatobiliary malignancy characterized by genomic heterogeneity. KRAS mutations play a significant role in influencing patient prognosis and guiding therapeutic decision-making. This study aimed to determine the prevalence and prognostic significance of KRAS mutations in CCA, asses the detection of KRAS G12/G13 mutations in plasma cell-free DNA (cfDNA), and evaluate the prognostic value of KRAS G12/G13 mutant allele frequency (MAF) in cfDNA in relation to clinicopathological data and patient survival.

Materials and Methods

A retrospective analysis of 937 CCA patients was performed using data from cBioPortal to examine KRAS mutation profiles and their association with survival. Plasma from 101 CCA patients was analyzed for KRAS G12/G13 mutations in the cfDNA using droplet digital PCR, and the results were compared with tissue-based sequencing from 78 matched samples.

Results

KRAS driver mutations were found in 15.6% of patients, with common variants being G12D (37.0%), G12V (24.0%) and Q61H (8.2%). Patients harboring KRAS mutations exhibited decreased overall and recurrence-free survival. KRAS G12/G13 mutations were detected in 14.9% of cfDNA samples, showing moderate concordance with tissue sequencing, and achieving 80% sensitivity and 93% specificity. Elevated KRAS G12/G13 MAF in cfDNA, combined with high CA19-9 levels, correlated with poorer survival outcomes.

Conclusion

The presence of KRAS mutations was associated with poor survival in CCA, underscoring the importance of KRAS mutations as prognostic markers. The detection of KRAS mutations in cfDNA demonstrated potential as a promising non-invasive alternative for mutation detection and, when combined with CA19-9 levels, may improve prognostic efficacy in CCA.

Cholangiocarcinoma (CCA), a malignancy of epithelial cells in the biliary tree, is the second most prevalent primary liver cancer. The prevalence of CCA is notably high in specific regions, including the northeast of Thailand (1). Several risk factors contribute to the development of CCA, including liver fluke infection and chronic biliary and liver diseases that induce chronic inflammation and cholestasis in the liver (1). Due to a typically late-stage diagnosis, the effectiveness of standard chemotherapy is limited, and surgical treatment is generally reserved for patients with early-stage disease. As a result, CCA has a poor prognosis with a 5-year overall survival (OS) of only 3-24% (2-4). There is an urgent need for effective biomarkers to predict patient prognosis and to guide the development of improved therapeutic strategies. Recent studies have highlighted the potential of neoadjuvant therapies in CCA, especially given the high recurrence rates post-surgery (5). Additionally, the integration of immuno-therapies, such as immune checkpoint inhibitors, is under investigation, and may offer new treatment options (6,7). These advances are particularly relevant in the context of molecular profiling, which enables tailored and targeted therapy strategies for CCA. Additionally, precise prognostic risk stratification will support clinical decision-making, ranging from selecting initial treatment regimens to identifying candidates for new clinical trials.

Several studies have reported the molecular characteristics of CCA, highlighting the importance of identifying genetic alterations that improve prognostic and treatment strategies (8-10). Among the genes targeted for therapy, KRAS mutations are the most frequent and are present in 7-40% of all CCA subtypes (9,11-14). A majority of KRAS mutations occur in codons 12 or 13, but mutations in codon 12 are the most common, occurring in >90% of KRAS mutation-driven cancer cases (15). The distribution of KRAS variants varies across different cancer types, with non-small cell lung cancer (NSCLC) commonly harboring G12C and colorectal cancer (CRC) and pancreatic ductal adenocarcinoma (PDAC) most commonly exhibiting G12D and G12V mutations (16-18). In CCA, KRAS mutations play a role in pathogenesis by affecting cell signaling pathways (19). A studies using mouse models have shown that KRAS activation combined with PTEN loss can activate the ERK/MAPK pathway, leading to CCA development (20). Moreover, the coexistence of isocitrate dehydrogenase and KRAS mutations contributes to the expansion of liver progenitor cells, subsequently promoting the development of intrahepatic CCA (21). Notably, the existence of KRAS mutations progressively increases from non-neoplastic large bile duct to peribiliary gland lesions and biliary intraepithelial neoplasia (BilIN) to intrahepatic CCA, with ~33% of BilIN lesions exhibiting KRAS mutations, indicating their importance as an early molecular event in cholangiocarcinogenesis (22). The presence of KRAS mutations in biliary tract cancer (BTC) is associated with poor OS, with specific allelic variants correlating with different survival outcomes (23,24). More notably, several KRAS inhibitor clinical trials are also ongoing in patients with BTC, including adagrasib, a KRAS G12C inhibitor that has revealed a favorable objective response rate in patients with KRAS G12C-mutant gastrointestinal cancer (including 8 patients with BTC) in the KRYSTAL-1 study (25). Therefore, determining the KRAS mutation status could help stratify patients for individualized treatment strategies, given its predictive value for prognosis.

However, the highly desmoplastic nature of CCA limits the accuracy of cytological and pathological approaches during diagnosis (26). Moreover, obtaining suitable tissue samples is often challenging and invasive, further complicating the diagnostic process. Therefore, there is a growing interest in liquid biopsy as a non-invasive alternative for CCA diagnosis and prognosis assessments, which involves analyzing tumor-derived biomarkers in the bloodstream. Studies have revealed the presence of genetic alterations in cell-free DNA (cfDNA) and demonstrated the diagnostic and prognostic values of plasma-based genotyping assays, with a number of assays relying on gene panel next-generation sequencing (NGS) (27-32). However, NGS frequently encounters sensitivity issues, particularly when detecting low-abundance mutations in cfDNA due to the low amount of cfDNA present (33,34). Patients with CCA, particularly those in the early stages of disease, exhibit low levels of cfDNA and mutant allele frequency (MAF) (29). Droplet digital PCR (ddPCR) has the potential to overcome these challenges by sensitively detecting and quantifying low-abundance mutations (35). Consequently, quantifying the number of KRAS mutations in cfDNA has been demonstrated to predict survival and recurrence rates in patients with cancer including in the gastrointestinal tract (36,37). Of note, the recent approval of sotorasib by the U.S. Food and Drug Administration (FDA), the first KRAS G12C inhibitor for KRAS G12C mutant-NSCLC, was based on cfDNA testing (38,39). Additionally, using ddPCR to identify KRAS mutations in cfDNA is sensitive, specific and aligns with the standard clinical mutation testing of archival tissue samples from advanced cancer (40). Despite these advances, the clinical implications and the concordance of KRAS mutations in cfDNA with tissue mutations in patients with CCA are yet to be fully explored.

At present, serum levels of CA19-9 are commonly used in the diagnosis and prognosis of CCA. However, the prognostic and diagnostic roles of CA19-9 remain controversial due to its low sensitivity and specificity (41). In addition, CA19-9 levels can also be elevated in other conditions, such as pancreatitis, cholangitis and cirrhosis (42), and the detection of CA19-9 is not applicable in patients lacking plasma Lewis antigens (43). There is growing evidence suggesting that the combination of CA19-9 with other genetic biomarkers might improve its diagnostic and prognostic efficacy. Studies, including the study by Watanabe et al. (44,45) have shown a correlation between KRAS-mutated cfDNA and CA19-9 levels in PDAC. Furthermore, an optimal CA19-9 cut-off value for prognostic prediction in PDAC was identified based on the presence of KRAS-mutated cfDNA. Additionally, a synergic prognostic value for the combination of KRAS cfDNA-based liquid biopsy with the conventional CRC biomarker, CEA, was found in CRC (35).

Thus, the aims of the present study were as follows: i) to determine the prevalence and distribution of KRAS mutations in CCA tumors by analyzing multiple CCA cohorts from the cBioPortal database; ii) to detect the 7 most clinically significant KRAS hotspot mutations (G12A, G12C, G12D, G12V, G12R, G12S and G13D) in plasma cfDNA from treatment-naive CCA patients using ddPCR, and compare their concordance with the KRAS G12/G13 mutations identified in CCA tissues via Sanger sequencing; and iii) to evaluate the association between the KRAS G12/G13 MAF in cfDNA with clinicopathological data, and evaluate its prognostic value in predicting patient survival outcomes.

Materials and Methods

Analysis of KRAS mutations in datasets from public cancer genomics databases. Somatic KRAS mutation data and clinical information on CCA tumors were retrieved from the cBioPortal database (46), including 6 independent datasets (Figure 1A). Specifically, data originated from the International Cancer Genome Consortium (ICGC) (12), 3 studies from the Memorial Sloan Kettering Cancer Center (MSK) (47-49), The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/) and a study by Zou et al. (50). The sample inclusion criteria were as follows: i) Genomic data from hepatobiliary cancer and CCA; and ii) data profiled for KRAS mutations. Exclusion of overlapping cases was conducted. In total, 937 patients with CCA were included in the present study, comprising 417 cases from the ICGC dataset, 382 cases from the MSK datasets, 35 cases from TCGA dataset and 103 cases from the study by Zou et al. Data on the liver fluke-related status were also obtained from the ICGC and MSK cohorts.

Figure 1.

Included patients and the study design. (A) Overview of sample inclusion from the cBioPortal database. (B) Sample inclusion for KRAS G12/G13 detection in cfDNA and CCA tissues. CCA, Cholangiocarcinoma; ICGC, International Cancer Genome Consortium; TCGA, The Cancer Genome Atlas.

Sample collection for KRAS G12/G13 detection. At diagnosis, a total of 101 whole blood samples and 78 formalin-fixed paraffin-embedded (FFPE) tissues were obtained from treatment-naive patients who underwent surgery at the Srinagarind Hospital, Khon Kaen University (Khon Kaen, Thailand), between January, 2019 and February, 2022 (Figure 1B). Biopsies were performed in all cases and were pathologically diagnosed by pathologists. Pathological TNM staging was conducted according to the 8th American Joint Committee on Cancer staging system (51). 101 plasma samples, 78 FFPE samples, clinical data and information on the CA19-9 levels of the patients in the present study were obtained from the Cholangiocarcinoma Research Institute, Khon Kaen University (Khon Kaen, Thailand) between February, 2022 and May, 2022. The OS was defined as the interval from the date of surgery to the time of death or the last follow-up date, and RFS was defined as the interval from the date of surgery to the occurrence of local, regional, or distant tumor recurrence, or to the last follow-up date for patients without recurrence (52). All patients provided informed written consent. The present study was approved by the Human Ethics Committee of Khon Kaen University (approval no. HE651040).

Plasma cfDNA extraction and quantification. Blood samples were collected in 10 ml K2EDTA tubes (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). The whole blood was centrifuged at 2,300 × g for 10 min at 4˚C, within 4 h of blood collection. The plasma was again centrifuged at 4,100 × g for 15 min at 4˚C to remove additional cellular debris. The plasma samples were then stored at –80˚C until DNA extraction. Total cfDNA was extracted from 2-4 ml plasma from each sample using the MagMAX cell-free DNA isolation kit (Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to the manufacturer’s protocol. cfDNA was eluted in 15 μl DNase/RNase-free distilled water (Invitrogen; Thermo Fisher Scientific, Inc.) and quantified using the Qubit dsDNA HS Assay kit (Invitrogen; Thermo Fisher Scientific, Inc.) and a Qubit 4.0 fluorometer (Invitrogen; Thermo Fisher Scientific, Inc.). Isolated cfDNA was then stored at –20˚C until analysis.

DNA extraction from FFPE tissues. Hematoxylin and eosin-stained sections were examined by pathologists and confirmed to consist of ≥80% tumor cells. FFPE DNA was extracted from 10-μm-thick sections using QIAamp DNA FFPE advanced kits (Qiagen, Inc., Hilden, Germany) according to the manufacturer’s protocol. DNA was quantified using the Qubit dsDNA HS Assay kit and a Qubit 4.0 fluorometer.

Mutation detection by ddPCR assay. The KRAS G12/G13 Screening Multiplex Kit (cat. no. 1863506; Bio-Rad Laboratories, Inc., Hercules, CA, USA) and a QX200 Droplet Digital PCR System (Bio-Rad Laboratories, Inc.) were used to perform ddPCR. This multiplex kit was used to screen for 7 common KRAS exon 2 mutations (G12A, G12C, G12D, G12R, G12S, G12V and G13D). Each reaction mixture contained 1X KRAS G12/G13 multiplex screening primers/probes reagent, which contained wild-type probe labeled with HEX dye and mutant probe labeled with FAM dye (Bio-Rad Laboratories, Inc.), 1X ddPCR Supermix for probes (no UTP; Bio-Rad Laboratories, Inc.) and 2-5 μl cfDNA template, adjusted to a final volume of 20 μl with DNase/RNase-free distilled water. A droplet generator cartridge (Bio-Rad Laboratories, Inc.) was loaded with the reaction mixture and 70 μl droplet generating oil (Bio-Rad Laboratories, Inc.) was loaded into the oil well. Next, the cartridge generated a droplet using a QX200 Droplet Generator (Bio-Rad Laboratories, Inc.), which was transferred to a PCR plate. PCR was then conducted using the following protocol: 95˚C for 10 min, followed by 40 cycles of 94˚C for 30 s and 55˚C for 1 min. These cycles were followed by a final step at 98˚C for 10 min. The processed droplets were analyzed using a QX200 Droplet Reader (Bio-Rad Laboratories, Inc.) to determine the level of fluorescence in each droplet. Each experiment also included a negative template control (NTC) and a positive template control (PTC). The PTC and NTC were KRAS G12/G13 positive and KRAS wild-type reference standards (Horizon Discovery, Ltd., Cambridge, UK), respectively. QuantaSoft Analysis Pro Software (version 1.0.596; Bio-Rad Laboratories, Inc.) was used to identify positive and negative droplets and convert results to copies/μl of reaction. Wells with <10,000 accepted droplets were excluded from the analyses. The KRAS G12/G13 MAF was calculated using the mutant allele concentration (copies/μl reaction, MT), the wild-type allele concentration (copies/μl reaction, WT) and the following equation: [MT/(MT + WT)] ×100.

Determination of the detection limit of the ddPCR-based KRAS G12/G13 multiplex assay using plasma cfDNA. For determination of the limit of blank (LoB) and the limit of detection (LoD) of the KRAS G12/G13 multiplex assay when using plasma cfDNA, the Multiplex I cfDNA Reference Standard Set (cat. no. HD780; Horizon Discovery, Ltd.) was used as a PTC. The PTC mutant template was diluted with wild-type DNA (multiplex cfDNA, 0%) to obtain a series of standard samples with the desired MAF (including 5.0, 2.5, 1.0, 0.5, and 0.1%). NTC wells (wild-type only control wells) were used to determine the LoB, and a serial dilution of mutant template was used to determine the LoD. The LoB was determined by calculating the mean of the measured MAF in 7 replicate NTC wells. The LoD was defined as the MAF value in which the 95% confidence interval (CI) of all replicate detections presented values above the LoB. In the present study, the LoB of the ddPCR analysis was 0.21% and the LoD was 0.5%, for 2.5 ng cfDNA template. Thus, a cut-off value of 0.5% MAF was used to assign cases, where the KRAS mutational status in plasma cfDNA was assigned “mutant” for cases with a MAF >0.5% and as “wild-type” for cases with a MAF ≤0.5%, as determined by ddPCR.

Sanger sequencing. Sanger sequencing mutation screening for KRAS G12/G13 was performed using FFPE DNA. The primer set used was as follows: KRAS forward, 5’-GTACTGGTGGAGTATTTGAT-3’ and KRAS reverse, 5’-GTCCTGCACCAGTAATATGC-3’, resulting in a 245 bp PCR amplicon. The PCR conditions used were as follows: 1X PCR buffer (no Mg), 3.75 mM MgCl₂, 0.5 mM dNTP mix, 1 μM primer, 2.5 units of Platinum Taq DNA Polymerase and at least 100 ng FFPE DNA, which was adjusted to a final volume of 25 μl with DNase/RNase-free distilled water (all Invitrogen; Thermo Fisher Scientific, Inc.). The PCR amplification conditions were as follows: 94˚C for 10 min, followed by 40 cycles of 94˚C for 30 s, 54˚C for 1 min and 72˚C for 1 min, and 1 cycle of 72˚C for 10 min. The PCR products were electrophoresed in a 1.5% agarose gel, stained with SafeView DNA Stain (Applied Biological Materials, Inc., Richmond, BC, Canada) and visualized using UV light. The PCR products were then treated with ExoSAP-IT PCR Product Cleanup Reagent (Applied Biosystems; Thermo Fisher Scientific, Inc.), according to the manufacturer’s protocol. Finally, the purified PCR products were Sanger-sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit and a 3730xl DNA Analyzer (Applied Biosystems; Thermo Fisher Scientific, Inc.). The nucleotide sequence data were analyzed using SnackVar free software (version 2.4.3) (53).

Statistical analysis. Data were analyzed using R statistical software (version 4.3.1; https://www.r-project.org) and Rstudio (version 2023.12; http://www.rstudio.com). Normality was assessed using the Shapiro-Wilk or Kolmogorov-Smirnov tests. The sensitivity, specificity and concordance rate were calculated using the “DTComPair” package (version 1.2.2). The Mann-Whitney U (for two groups) or Kruskal-Wallis (for more than two groups) tests were used to compare differences in the KRAS MAF between patient groups. Correlation analysis was performed using the Spearman’s rank-order test. Survival analysis was conducted using Kaplan-Meier OS curves and assessed using the log-rank test or Renyi test (54) with the “survminer” package (version 0.4.9). The 5% trimmed mean value of KRAS G12/G13 MAF cfDNA and the median CA19-9 level were used as the cut-off in the OS analysis, as described previously (40,55). Univariate analysis was performed by the Cox regression model to test the independent significance of covariates using the “survival” package (version 3.5-5), and data are presented as hazard ratios (HRs) and 95% CIs. p<0.05 was considered to indicate a statistically significant difference.

Results

KRAS mutation frequency in CCA tumors and its association with patient survival across public cancer genomics databases. A total of 937 patients were included in the analysis, of which 146 (15.6%) had KRAS driver mutations (including G12D/V/C/A/R/S, G13, Q61, A146, A18 and Q22). Among these 146 patients, the most prevalent KRAS allelic variants were G12D, G12V and Q61H (37.0, 24.0 and 8.2%, respectively; Figure 2A and B). The distribution of specific KRAS mutations in relation to liver fluke-associated CCA was analyzed using 77 patients of the ICGC cohort and 50 patients of the MSK studies with KRAS-mutant CCA. Notably, KRAS G12D mutations were the most frequent in fluke-negative CCA, occurring in 40.7% of cases, whereas in fluke-positive CCA, the highest frequencies were observed for KRAS G12D (15.8%), G12V (15.8%), G13D (15.8%), and G12C (15.8%) (Figure 2C).

Figure 2.

Prevalence and distribution of KRAS driver mutations in CCA tumors. (A) The lollipop plot shows the identified KRAS variants relative to a schematic representation of the KRAS gene. Any position with a mutation contains a circle and the length of the line depends on the number of mutations detected at that codon. (B) The frequencies of the individual KRAS mutations among all the identified KRAS driver mutations in the study cohort. (C) Distribution of the KRAS mutations identified in patients with CCA stratified based on liver fluke-related status. Data was retrieved from the cBioPortal database. *p<0.05. CCA, Cholangiocarcinoma.

Next, it was determined whether there was a difference in survival outcome according to the presence of the KRAS driver mutations. The median OS time of the 100 patients with a KRAS driver mutation (17.23 months; 95%CI=12.16-20.98 months) was significantly shorter than that of the 640 patients with KRAS wild-type (30.72 months; 95%CI=25.59-35.55 months) (p<0.0001; Figure 3A). Similarly, the RFS time of KRAS mutant cases (10.94 months; 95%CI=4.76-19.25 months) was significantly shorter than that of KRAS wild-type cases (20.17 months; 95%CI=14.92-31.31 months) (p=0.0031; Figure 3B). Further subgroup analyses based on the KRAS mutation subtype (G12/G13 vs. non-G12/G13) were performed. KRAS G12/G13 mutations showed a significant association with decreased OS and RFS (p<0.0001 and p=0.026, respectively; Figure 3C and D). Conversely, non-G12/G13 KRAS mutations did not show a significant association with OS, but a significant association was observed with RFS (p=0.0084; Figure 3E and F). These findings highlighted the prognostic relevance of KRAS G12/G13 mutations in CCA.

Figure 3.

Kaplan-Meier survival analysis by KRAS mutation status in patients with cholangiocarcinoma. Kaplan-Meier curves for (A) OS and (B) RFS stratified by KRAS mutation status. Kaplan-Meier curves for (C) OS and (D) RFS comparing patients with KRAS G12/G13 mutations to those with wild-type KRAS. Kaplan-Meier curves for (E) OS and (F) RFS comparing patients with non-G12/G13 KRAS mutations to those with wild-type KRAS. OS, Overall survival; RFS, recurrence-free survival.

KRAS G12/G13 mutations in plasma cfDNA. A total of 101 patient blood samples were analyzed for KRAS G12/G13 mutations. Overall, the median cfDNA level was 16.75 ng/ml of plasma, ranging 1.10-373.78 ng/ml of plasma (Table I). The ddPCR assay was used to determine and quantitatively evaluate the KRAS G12/G13 mutations in the plasma cfDNA. In the selected cohort, 14.9% (15/101) of patients harbored KRAS G12/G13 mutations (Table II). The median KRAS MAF in plasma cfDNA was 0.02%, ranging 0-8.46% (Table I). A significant association between the KRAS G12/G13 MAF and clinicopathological parameters was not observed (Table I). Moreover, no correlation was observed between the KRAS G12/G13 MAF and either cfDNA levels or tumor size in the study cohort.

Table I. Correlation of the KRAS G12/G13 MAF in cfDNA with the clinical status of patients with cholangiocarcinoma.

^aMann-Whitney U-test and ^bKruskal-Wallis test.

Table II. Comparison of KRAS G12/G13 mutation detection in cfDNA by droplet digital PCR and with detection in tissues by Sanger sequencing.

cfDNA, Cell-free DNA; CI, confidence interval; SE, standard error.

To assess the concordance between ddPCR-based detection in plasma and Sanger sequencing in tissue, matched plasma and tissue samples from 78 cases were compared. KRAS G12/G13 mutations were identified in 5 tissue samples via Sanger sequencing. Of these 5 mutation-positive cases (G13D, n=3; G12D, n=1; G12V, n=1), 4 cases (80%) showed concordant mutations in plasma cfDNA detected by ddPCR, with a MAF range of 0.52-2.7%. For the remaining 73 wild-type tissues, 68 (93%) showed agreement with the wild-type KRAS status in plasma. The overall concordance rate was 92%. Collectively, the cfDNA test showed a sensitivity of 80% (95%CI=0.45-1.00) and a specificity of 93% (95%CI=0.87-0.99) (Table II). The Cohen’s ĸ coefficient was 0.533 (standard error=0.165; 95%CI=0.209-0.857; p=0.001), suggesting a moderate agreement between the two methods.

Increased levels of both KRAS G12/G13 MAF and CA19-9 are associated with OS in CCA. It was next determined whether the baseline MAF of KRAS G12/G13 in cfDNA was associated with OS and RFS. Patients were divided into two groups according to the KRAS G12/G13 MAF (MAF ≤0.174% vs. MAF >0.174%). However, no significant difference in OS was observed between these two groups (p=0.52; Figure 4A). Similarly, no significant difference in OS was observed between patients classified based on the median CA19-9 value (p=0.21; Figure 4B). Notably, compared to KRAS MAF or CA19-9 alone, patients with elevated levels of both KRAS G12/G13 MAF and CA19-9 (MAF >0.174% and CA19-9 >49.99 U/ml) had a worse OS time compared to those with low levels of the combined biomarkers (15.8 vs. 39.0 months; p=0.046; Figure 4C). An elevation of both KRAS G12/G13 MAF and CA19-9 levels was the only independent negative factor for OS (HR=3.19; 95%CI=1.03-9.91; Table III). In contrast, no significant difference in RFS was observed between patients classified based on median CA19-9, KRAS MAF alone, or their combination (Figure 4D-F).

Figure 4.

Analysis of OS and RFS based on the KRAS G12/G13 MAF and CA19-9 levels in patients with cholangiocarcinoma. (A) OS Kaplan-Meier curve stratified by KRAS G12/G13 MAF in cell-free DNA. (B) OS Kaplan-Meier curve stratified by CA19-9 levels. (C) Kaplan-Meier curve demonstrating the combined impact of the KRAS G12/G13 MAF and CA19-9 levels on OS. (D) RFS Kaplan-Meier curve stratified by KRAS G12/G13 MAF in cell-free DNA. (E) RFS Kaplan-Meier curve stratified by CA19-9 levels. (F) Kaplan-Meier curve demonstrating the combined impact of KRAS G12/G13 MAF and CA19-9 levels on RFS. cfDNA, Cell-free DNA; OS, overall survival; RFS, recurrence-free survival; MAF, mutant allele frequency.

Table III. Cox univariate regression analysis assessing the prognostic performance of the KRAS G12/G13 MAF in plasma, CA19-9 levels and the clinicopathological variables of patients with CCA.

^aHazard ratio (HR), estimated from Cox proportional hazard regression model. ^bConfidence interval of the estimated HR. Statistical significance was defined as *p<0.05. CCA, Cholangiocarcinoma; MAF, mutant allele frequency; OS, overall survival; HR, hazard ratio.

Discussion

In the present study, the prevalence and prognostic impact of KRAS driver mutations in CCA tumors and their detection in plasma cfDNA were examined. Among the included 937 CCA tumors from multiple genomic datasets, KRAS driver mutations were identified in 15.6%, with G12D emerging as the most prevalent allelic variant. Further analysis demonstrated that KRAS G12/13 mutations were significant prognostic indicators in CCA and associated with decreased OS and RFS times. In the present study, 14.9% of patients harbored KRAS G12/G13 mutations in cfDNA, as identified by a clinically validated ddPCR assay, which showed moderate agreement with tissue Sanger sequencing. In this cohort, the KRAS G12/G13 MAF was not associated with clinicopathological parameters. However, further analysis revealed that elevated levels of both KRAS G12/G13 MAF and CA19-9 were significantly associated with a shorter OS time. This underscores the prognostic value of integrating multiple biomarkers. A key novelty of this research is its thorough examination of KRAS mutations in cfDNA, which has not been extensively explored in CCA. By employing ddPCR, this study enhances the sensitivity and specificity of detecting low-abundance KRAS mutations in cfDNA and correlates these findings with patient prognostic outcomes. This approach offers a less invasive diagnostic alternative to traditional biopsy methods and provides critical insights into the potential of cfDNA analysis to act as a predictive tool for patient survival.

KRAS mutations are recognized as crucial targets for targeted therapy in both tissue and cfDNA (25,56). In the present study, KRAS mutations were identified in 15.6% of CCA tissues, with G12D, G12V and Q61H being the most prevalent variants. The prevalence of KRAS mutations in CCA varies, ranging 7-40% across different published studies (9,11-14). Specifically, Xu et al. (57) reported a high frequency (38.2%) of KRAS mutations in Chinese patients with CCA. A broader study by Wang et al. (58) reported KRAS mutations in 8.7% of patients, with G12D and G12V being the most common mutations, which aligns with the findings of the present study. Zhou et al. (23) observed that G12 variants were associated with a worse survival and increased recurrence risk, paralleling the observations in the present study where KRAS mutations were associated with a poor prognosis in CCA. It was also found in the present study that the Q61H variant was the third most common subtype in CCA tumors, accounting for 8.2% of KRAS mutations. While targeted therapies, such as sotorasib for KRAS G12C mutations, have gained FDA approval, the development of inhibitors for other KRAS mutation subtypes remains in the early stages. A recent study demonstrated that casein kinase 2 inhibition affects macropinocytosis in KRAS mutant CCA, suggesting a novel angle for targeting metabolic pathways in KRAS-driven tumors (59). This insight into CK2’s role in metabolic adaptation further supports the exploration of targeted treatments that disrupt KRAS-driven metabolic processes, offering new therapeutic avenues in CCA management.

In the present study, to detect and quantify the seven most common hotspot mutations in KRAS G12/G13 in cfDNA, a sensitive ddPCR-based technology was used that enabled the detection of these mutations in cfDNA at a low abundance, with a median MAF of 0.02%. KRAS G12/G13 mutations were identified in 14.9% of plasma samples from patients with CCA, consistent with the KRAS mutation rates of 13.9% found in cfDNA of advanced BTC (28). The concordance in KRAS G12/G13 mutation status as determined using ddPCR for plasma and using Sanger sequencing for FFPE samples was also examined in the present study. The concordance between plasma ddPCR and tumor tissue Sanger sequencing was 92%. This concordance result for KRAS G12/G13 mutations compares favorably to previous studies. In CRC and NSCLC, the inter-assay concordance between plasma ddPCR and tissue sequencing in the KRAS mutation detection rates was 24-85% (40,60-63). In the present study, the ddPCR method showed a sensitivity of 80% and a specificity of 93% for detecting KRAS G12/G13 in plasma when compared with tissue Sanger sequencing. Similarly, Takase et al. demonstrated that KRAS mutations in cfDNA from pancreatic cancer patients, derived from endoscopic ultrasound-guided fine-needle aspiration specimens, achieved a sensitivity of 63.2% and a specificity of 100%, highlighting the potential of cfDNA analyses as ancillary diagnostic tools (64). Furthermore, in the present study, the Cohen’s ĸ coefficient of 0.533 indicated a moderate agreement between cfDNA and tissue-based KRAS mutation detection. This highlights an important consideration in the use of cfDNA as a diagnostic tool. The moderate agreement may be attributed to several factors. For instance, 1 tissue mutant case was undetected in plasma, potentially due to limitations in detecting low MAFs and the short half-life of ctDNA in the blood (65). Additionally, 5 cases that were KRAS wild-type in tumor tissue but KRAS mutant in cfDNA were also found. Possible reasons for this discordance may involve intra-tumoral heterogeneity (62,65-67). Additionally, tissue sequencing techniques have inherent detection limits, missing subclonal alterations if they fall below a certain fractional abundance threshold in sampled sections (68,69). In the present study, matched tumor analysis by ddPCR was not possible due to limited specimens. It should also be acknowledged that KRAS mutations may occur in healthy individuals due to clonal hematopoietic cells (70,71). In the present study, Sanger sequencing was performed to confirm that all patients with KRAS mutations identified in the study had wild-type KRAS in their blood cells.

In the present study, the baseline KRAS G12/G13 MAF in cfDNA was not associated with patient clinicopathological data, OS or RFS. The prognostic role of the KRAS MAF in cfDNA remains controversial according to previous reports. Kim et al. (63) reported that the serum KRAS G12/G13 MAF was not related to clinicopathological parameters in CRC, while it was reported to be associated with tumor staging and distant metastasis in PDAC (72,73). In CCA, this limited prognostic utility may be attributed to the low mutational burden. In the present study, KRAS G12/G13 mutations were detected in cfDNA at a low abundance, with a median MAF of 0.02% (range=0-8.46%). This contrasts with the higher median and range of MAF reported in other gastrointestinal and epithelial malignancies associated with a poor prognosis, including a median MAF of 0.34% in pancreatic cancer (range=0-34.71%), a median MAF of 0.5% in advanced CRC, NSCLC and melanoma, and a MAF of 0.53-10% in metastatic CRC (40,63,72). Specifically, Janku et al. (40) reported that patients with a low KRAS G12/G13 MAF exhibited a significantly longer OS time than those with a high KRAS G12/G13 MAF in the cfDNA of advanced cancer including CRC, NSCLC and melanoma. Moreover, the KRAS MAF in the preoperative plasma of patients with CRC and liver metastases was inversely associated with OS (35). This suggests that in CCA, KRAS mutations predominantly exist as subclonal variants within the circulating DNA, often constituting <1% of the total cfDNA. Such a prevalence of subclonal populations may therefore diminish the prognostic value of the KRAS MAF. However, further studies with larger cohorts are warranted to elucidate whether distinct clinical outcomes become more apparent with higher MAF thresholds. Additionally, the significance of very low-level mutant fractions needs careful consideration when correlating ctDNA profiles with clinical prognosis.

While the prognostic roles of CA19-9 remain controversial, in the present study, the combined elevation of CA19-9 and KRAS G12/G13 MAF was associated with worse survival outcomes. There is evidence that combination of the KRAS MAF in cfDNA with routine biomarkers, such as CA19-9 and CEA, could be a preferable circulating biomarker detection method for prognosis and diagnosis (35,72). These findings highlight the potential of combining multiple biomarkers for a more accurate prognostic assessment in cancer and could thus potentially guide more tailored therapeutic strategies.

Study limitations. First, the sample size for mutation detection in plasma was small, which restricted the extent of the conclusions that could be drawn from the findings. Therefore, further studies with larger sample sizes are required to validate the results of the present study. Second, the multiplex ddPCR assay used in the present study was limited to detecting only a narrow range of KRAS mutations that have been shown to have clinical relevance, and thus may not fully reflect the prognostic efficacy of KRAS mutations in plasma cfDNA. Other notable aspects include assessing KRAS MAF alterations in the cfDNA of patients with CCA treated with standard chemotherapy, targeted treatments or for predicting metastasis and disease relapse. Thus, the dynamic change in KRAS MAF in the cfDNA of patients with CCA during treatment should be continuously detected in future work. Third, while the present study provided valuable insights into the prognostic relevance of KRAS G12/G13 mutations in CCA tumors, a significant constraint is the lack of Cox or univariate/multivariate analyses in the present study. This was due to the diversity and varied origin of the data obtained from cBioPortal, which compiles information from multiple studies. For robust validation of the association between KRAS mutations and clinical outcomes in CCA, further studies should focus on uniformly collected and standardized data.

Conclusion

The present study, based on more than 900 genomic profiles from CCA patients, identifies KRAS mutations as significant prognostic factors. We found KRAS mutations in 15.6% of CCA tissues, of which the most prevalent variants included G12D, G12V, and Q61H. Notably, KRAS G12/G13 mutations in cfDNA were found in 14.9% of patients with moderate concordance with tissue sequencing and showing 80% sensitivity and 93% specificity. Moreover, the integrated KRAS MAF in cfDNA with CA19-9 levels correlated with decreased survival times, highlighting the prognostic significance of multi-marker strategies in the clinical setting. These findings suggest that KRAS may serve as a prognostic biomarker and offer an opportunity for a non-invasive method to monitor prognosis and therapy response in routine surveillance.

Declaration of Generative AI and AI-assisted Technologies in the Writing Process

During the preparation of this work the Authors used ChatGPT and QuillBot to enhance the language in certain sections of the text. After using this tool, the Authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Conflicts of Interest

The Authors declare that they have no competing interests.

Authors’ Contributions

AJ conceptualized the study, designed the methodology, performed the investigation, supervised the project, provided funding, and wrote and edited the manuscript. PT conceptualized the study, designed the methodology, performed analysis, and wrote the original manuscript draft. JC confirmed the clinical data. CA, PI, and WK performed the histological examination. SK, AT, WL, NN, and AT.T reviewed and edited the manuscript. AJ and PT confirm the authenticity of all the raw data. All Authors read and approved the final manuscript.

Acknowledgements

The Authors express gratitude to all patients involved in this study. We acknowledge the Centre for Research and Development of Medical Diagnostic Laboratories (CDML). We would like to express our sincere gratitude to Professor Narong Khuntikeo for his invaluable support and guidance throughout this project.

Funding

This research was supported by Office of the Permanent Secretary, Ministry of Higher Education, Science, Research and Innovation (OPS MHESI), Thailand Science Research and Innovation (TSRI) and Khon Kaen University (Grant No. RGNS 64-051), NSRF under the Basic Research Fund of Khon Kaen University through Cholangiocarcinoma Research Institute (CARI-BRF64-22), and a grant from Khon Kaen University to A.J. (KKU62 Grant No. 6200010002).