Comprehensive analysis of KRAS aberrations in pan-cancer, with a focus on prognostic and therapeutic implications

Abstract

Background:

KRAS is one of the most frequently mutated oncogenes in humans. KRAS aberrations play a significant role in various solid tumors, affecting patient prognosis and treatment outcomes.

Objectives:

We identified features of genetic alterations in KRAS, including single amino acid substitutions and amplifications, based on the results of next-generation sequencing tests in 1667 advanced solid tumor patients.

Design:

Retrospective cohort study.

Methods:

Among 1667 patients, 28.1% (N = 468) had KRAS aberrations (single-nucleotide variant (SNV): N = 438, 26.1%; amplification (copy number variation): N = 48, 2.9%) in metastatic solid tumors.

Results:

The incidence rate of SNVs was higher in pancreatic cancer (N = 89, 89.9%) than in other tumors, including colorectal cancer (N = 259, 47.3%) and small bowel cancer (N = 4, 40.0%). Most of the mutations were missense mutations (N = 434, 99.1%). In addition, we examined the specific location of SNVs; the mutational type G12D (N = 178, 40.6%) was the most frequent, followed by G12V (N = 95, 21.7%) and G13D (N = 47, 10.7%). In the survival analysis, the mutational types of G12V and G13D influenced the poor survival of patients (G12V, the mutant type; 245 vs the wild type; 531 days; G13D, the mutant type; 435 vs the wild type; 531 days).

Conclusion:

In patients with KRAS amplification, the copy number range varied among the tumor types. Bladder cancer (147.9), cholangiocarcinoma (16.2), and gastric cancer (11.0) had relatively high median copy numbers of KRAS. Overall, our data are expected to provide valuable information for patients with various metastatic solid tumors and KRAS aberrations.

Introduction

The small GTPase protein family functions as a binary switch in the signal transduction of most growth factor receptors including epidermal growth factor receptors, tyrosine kinase receptors for Hepatocyte Growth Factor (HGF) (MET), and tyrosine kinase receptors for Stem Cell Factor (SCF) (KIT).^{1 –3} KRAS is a small GTP-binding protein, ⁴ and KRAS is the one of the most frequently mutated oncogenes in humans: more than 80% of pancreatic cancers and more than 30% of colorectal cancers (CRCs). ⁵ The tremendous downstream effects of dysregulated RAS activity lead to unchecked cell division and additional oncogenic mutations. ⁶ KRAS mutations are associated with limited prognosis and poor response rates to tyrosine kinase inhibitors. ⁷ However, several new clinical trials for KRAS mutant tumors have been recently conducted. ⁸ For example, the KRYSTAL-1 phase I/IB study reported on the safety, tolerability, recommended phase II dose, and preliminary efficacy of adagrasib, a potent covalent KRAS G12C inhibitor, in advanced solid tumors harboring the KRAS G12C mutation.^9,10 In nonsmall cell lung cancer, KRAS mutations were considered nontargetable alterations or undruggable for a long time. ¹¹ Nonetheless, the results of early clinical trials recently demonstrated that pharmacological inhibition of KRAS G12C mutated protein is practicable, opening the possibility of a new targeted treatment for patients with advanced nonsmall cell lung cancer. ¹² In addition, attempts have been made to search for treatments for non-G12C-KRAS mutant cancers. A clinical trial of VS-6766, a dual RAF–MEK inhibitor, reported early single agent activity in non-G12C mutated KRAS driven cancers. ¹³ On the flip side, with the rapid implementation of next-generation sequencing (NGS) in the oncology practice worldwide, clinicians encounter various KRAS aberrations in clinics.^14,15 Here, we investigated the features of KRAS aberrations and the association between KRAS aberrations and survival outcomes in various types of solid tumors using NGS.

Materials and methods

Patient enrollment

NGS was conducted at the time of the diagnosis of metastatic disease using formalin-fixed paraffin-embedded (FFPE) tissue specimens. The collection of specimens and associated clinical data used in this study were approved by the Institutional Review Board of Samsung Medical Center (IRB# 2019-09-052). All participants provided written informed consent before enrollment and specimen collection. This study was performed in accordance with the principles of the Declaration of Helsinki and the Korean Good Clinical Practice guidelines. The reporting of this study conformed to the STROBE guidelines (Supplemental STROBE Checklist).

DNA extraction

The tumor regions were microdissected from most tumor tissues, except for the samples used for genomic DNA extraction. Genomic DNA was isolated from FFPE tissue fragments and purified using an AllPrep DNA/RNA FFPE Kit (Qiagen, Venlo, The Netherlands). DNA concentration was measured using a Qubit dsDNA HS assay kit (Thermo Fisher Scientific, Waltham, MA, USA), and 40 ng of DNA was used as the input for library preparation. The DNA integrity number, which is a measure of DNA fragment size and consequently, DNA quality, was determined using the Genomic DNA ScreenTape assay on an Agilent 2200 TapeStation system (Agilent Technologies, Santa Clara, CA, USA).

Library preparation and data analysis

DNA library was prepared using a hybrid capture-based TruSight Oncology 500 DNA/RNA NextSeq Kit, following the manufacturer’s protocol. During library preparation, enrichment chemistry was optimized to capture nucleic acid targets from FFPE tissues. Unique molecular identifiers were used in TruSight Oncology 500 (TSO500; Illumina, San Diego, CA, USA) analysis to determine the unique coverage at each position and reduce any background noise caused by sequencing and deamination artifacts in FFPE samples. DNA library preparation enables the detection of variants at low variant allele frequencies, while simultaneously suppressing errors, thereby providing high specificity.

Sequence data were analyzed for clinically relevant classes of genomic alterations, including single-nucleotide variants (SNVs), small insertions and deletions (indels), CNVs, and rearrangements/fusions. SNVs and small indels with a variant allele frequency <2% were excluded. Average copy number variations of more than four were considered gains and those less than one were considered losses. Only gain (amplification) was analyzed in the TSO 500-CNV analysis, and RNA translocation-supporting reads of more than 4–12 were considered translocations, depending on the quality of the sample. Data outputs exported from the TSO 500 pipeline (Illumina) were annotated using the Ensembl Variant Effect Predictor Annotation Engine, with information from databases, such as dbSNP, gnomAD genome and exome, 1000 genomes, ClinVar, COSMIC, RefSeq, and Ensembl. Processed genomic changes were categorized according to a 4-tier system proposed by the American Society of Clinical Oncology/College of American Pathologists and annotated with proper references. The TSO 500 pipeline was used to determine the tumor mutation burden (TMB) and microsatellite instability (MSI) status. The TMB was calculated by (1) excluding any variant with an observed allele count ⩾10 in any of the GnomAD exome, genome, and 1000 genomes databases, and including, (2) variants in the coding region (RefSeq Cds), (3) variant frequency ⩾5%, (4) coverage ⩾50×, (5) SNVs and indels, (6) nonsynonymous and synonymous variants, and exclusion of (7) nonsynonymous and synonymous variants. The effective panel size for the TMB was the total coding region with a coverage of >50×. The MSI status was calculated from the microsatellite sites for evidence of instability relative to a set of baseline normal samples based on information entropy metrics. The percentage of unstable MSI sites among the total number of assessed MSI sites was reported as a sample-level microsatellite score. The TSO500 assay demonstrated strong performance and reliable accuracy in detecting the evaluated markers for pan-cancer. ¹⁶

Statistical analysis

Data are presented as mean ± SD. All statistical analyses were performed using R (Ver.3.4), R Studio (https://www.rstudio.com/), and GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA; http://www.graphpad.com/). Statistical significance was set at p < 0.05. All statistical tests were two-sided. Hazard ratios (HRs) were estimated using a Cox regression model with the forestmodel package (ver. 0.6.2) in R. The survival (Ver. 3.5-5) and survminer (Ver. 0.4.9) packages in R were used to generate survival curves and calculate p-values. Fisher’s exact test was performed to evaluate statistical differences between cancer types. The trackViewer (ver. 1.30.0) and GenVisR (ver. 1.26.0) packages were used to generate Laplace plots displaying the location and number of SNVs.

Results

Patient characteristics

A total of 1667 cancer patients underwent NGS, including 523 cancer genes (TSO500; Illumina), as a routine clinical practice at the Samsung Medical Center between October 2019 and April 2021 (Supplemental Table S1). The most common tumor types were CRC (N = 547, 32.8%), gastric cancer (GC; N = 381, 22.9%), and sarcoma (N = 155, 9.3%; Figure 1(a)). In total, 468 patients (28.1%) had KRAS aberrations in their tumors. Among them, 420 (26.3%) had only KRAS mutations (SNV), 30 (1.8%) had only KRAS amplification, and 18 (1.1%) had concurrent KRAS amplification and mutation (Figure 1(a)). No KRAS fusions were observed. We investigated the prevalence of KRAS mutations in each cancer type (Figure 1(b)). The percentage of KRAS mutations was the highest in pancreatic adenocarcinoma patients (89.9%), followed by CRC (47.3%) and small bowel cancer (40.0%) patients (Figure 1(b)). In contrast, the incidence of bladder cancer (BLCA; 3.1%) and hepatocellular carcinoma (2.8%) was lower than that of other tumors. The percentage of KRAS amplification was the highest in neuroendocrine tumor (9.1%) and head and neck cancer (7.1%) patients, whereas it was the lowest in BLCA patients (1.5%; Figure 1(c)). To assess whether the relative incidence differed significantly across each pair of cancer types, we conducted a Fisher’s exact test. The results are summarized in Supplemental Tables S2 (SNV) and S3 (CNV).

Figure 1.

Overview of enrolled cancer patients and the proportions of KRAS genetic variants. (a) Pie chart indicating the percentage of each type of cancer in a total of 1667 patients and a Venn diagram showing the number and percentage of patients with KRAS SNVs and CNVs (amplification). (b) Chart showing the number of patients with KRAS SNV mutations in order of the most common tumor types and the percentage of tumor types with KRAS SNVs: PACA (89.9%), CRC (47.3%), and small bowel cancer (40.0%). (c) The most common tumor types with KRAS CNV mutations and the percentage of tumor types with KRAS CNVs: NET (9.1%) and HNC ( 7.1%).

CRC, colorectal cancer; HNC, head and neck cancer; NET, neuroendocrine tumor; PACA, pancreatic adenocarcinoma; SNV, single-nucleotide variant.

KRAS mutations (SNV)

Of the 468 patients with KRAS aberrations, 438 (93.6%) had KRAS mutations (SNV) in their tumor specimens. The most common tumor type was CRC (N = 259, 59.0%), followed by pancreatic cancer (N = 89, 20.3%), cholangiocarcinoma (CCC; N = 35, 8.0%), and GC (N = 26, 5.9%; Figure 2(a)). Regarding transition and transversion, the Ampulla of Vater cancer, BLCA, and neuroendocrine tumors showed a pattern of high transversion proportion. In contrast, CCC, GC, and sarcoma showed a high proportion of T to G changes (Figure 2(b)). The most common mutation type was missense mutation (N = 450, 99.1%), except in four cases. The four cases included one nonsense mutation, two in-frame insertions, and one in-frame deletion (Figure 2(c)). The nonsense mutation was observed in small bowel cancer. Additionally, we examined the specific location of SNVs and found that the G12D mutation (N = 178, 40.6%) was the most frequent, followed by the G12V (N = 95, 21.7%) and G13D mutations (N = 47, 10.7%; Figure 2(d)). Interestingly, many SNVs (N = 404, 87.4%) occurred between the 10th and 17th amino acid from the start of the coding region (Figure 2(d)).

Figure 2.

(a) Pie chart showing the cancer types of KRAS-SNV patients: CRC (n = 259, 59.0%), PACA (n = 89, 20.3%), CCC (n = 35, 8.0%), and GC (n = 26, 5.9%). (b) Bar graph representing the proportion of each nucleotide change in the various cancer types. (c) Pie chart showing mostly missense mutations (n = 450, 99.1%). Four cases included one nonsense mutation, two in-frame insertions, and one in-frame deletion. (d) The most frequent missense mutations were G12D (n = 178, 40.6%), G12V (n = 95, 21.7%), and G13D (n = 47, 10.7%). (e) No significant differences in OS between patients with KRAS mutations and those with wild-type KRAS (p = 0.18). (f) Significantly poor OS in patients with mutations in G12V (p = 0.036) and (g) G13D (p < 0.001) than in patients with wild-type KRAS and KRAS mutations at other sites.

CCC, cholangiocarcinoma; CRC, colorectal cancer; GC, gastric cancer; OS, overall survival; PACA, pancreatic adenocarcinoma; SNV, single-nucleotide variant.

In the survival analysis, overall survival (OS) did not differ between patients with KRAS mutations and those with wild-type KRAS (median: 502 vs 526 days, p = 0.18; Figure 2(e)). However, patients with mutations in G12V and G13D had shorter OS than those with wild-type KRAS and KRAS mutations at other sites (G12V: median: 245 vs 505 (KRAS SNVs in other sites), 531 days (wild-type KRAS), p = 0.036, Figure 2(f); G13V: 435 vs 495 (KRAS SNVs at other sites), 531 days (wild-type KRAS), p < 0.001, Figure 2(g)). We further investigated the impact of the G12V and G13D mutations on OS across different cancer types. In CCC (p = 0.005), CRC (p < 0.001), and sarcoma (p = 0.039) patients, patients with the G12V mutation exhibited significantly shorter OS than those with other KRAS mutations (Supplemental Figure 1(A)). Similarly, in CRC (p < 0.001, p < 0.001) and GC (p = 0.001, p = 0.038) patients, patients harboring the G13D mutation showed a significantly reduced OS than those harboring other KRAS mutations or wild-type KRAS (Supplemental Figure 1(B)). The number of patients with other cancer types was insufficient to categorize them into the three groups; thus, they were not included in the survival plots.

Multivariate Cox regression analyses of CCC, CRC, GC, and sarcoma are shown in Supplemental Figures 3–6. The KRAS G12V mutation was significantly associated with an increased HR for OS in CCC, CRC, and sarcoma patients. The KRAS G13D mutation was significantly associated with an increased HR for OS in CRC and GC patients.

KRAS amplification

Among the 51 patients with KRAS amplification, the most common tumor type was CRC (N = 18, 37.5%), followed by GC (N = 9, 18.8%), CCC (N = 4, 8.3%), and melanoma (N = 4, 8.3%; Figure 3(a)). KRAS copy number ranged from 3.3 to 181.1 (median: 5.6; Figure 3(b)). Interestingly, a few patients had an extremely high copy number of KRAS: 1 BLCA (147.9) and 2 GC patients (144.1 and 181.7); the median copy number was the highest in BLCA patients (147.9), followed by CCC (16.2) and GC patients (11.0; Figure 3(c)). Conversely, the median value was the lowest in sarcoma patients (4.4). Supplemental Table S2 presents the p-values obtained from pairwise comparisons among cancer types; no significant differences in KRAS CNV were noted between the two different types of cancer.

Figure 3.

(a) Pie chart showing the distribution of tumor types with KRAS amplification (n = 48). (b) Chart showing the number of patient incidences according to the KRAS copy number range and cancer type. (c) Range of copy number in each cancer type. The square bar represents the mean value of copy number. (d) The proportion of patients with KRAS-amplified tumor in each tumor types. (e) No significant difference in OS between patients with KRAS-amplified tumors, those with amplification of other genes, or those without any amplification (p = 0.13). (f) Landscape of the genomic characteristics of patients with KRAS amplification. The first top panel: copy number of KRAS; middle panel: cancer type, TMB, microsatellite instability, PD-L1 status, sex, and age; and bottom panel: OncoPrint showing concurrent SNV genes in KRAS-amplified patients. Left: top gene list that mutated the most frequently and percentage of mutations in KRAS-amplified patients.

OS, overall survival; PD-L1, programmed death-ligand 1; SNV, single-nucleotide variants; TMB, tumor mutation burden.

The OS was not significantly different among patients with KRAS-amplified tumors, those with amplification of other genes, and those without any amplification (p = 0.13; Figure 3(d)). To gain further insights, we assessed the OS stratified by cancer type (Figure S2). Our analysis revealed that KRAS copy number variations were significantly associated with reduced OS in patients with CCC, CRC, and pancreatic cancer. Finally, we portrayed the landscape for the characteristics of patients with KRAS amplification (Figure 3(e)). Among 48 patients, 9 (18.8%) had TMB-high tumors and 12 (25.0%) had programmed death-ligand 1 (PD-L1)-positive tumors with combined positive score ⩾1%. All KRAS-amplified tumors were microsatellite-stable. Of note, 33 (68.8%) patients had concurrent TP53 mutations in their tumor specimens. Following TP53, NOTCH3 (N = 15, 31.3%) and BRCA2 (N = 14, 29.2%) were the most frequently mutated genes in KRAS-amplified patients.

Correlation between KRAS aberrations and the TMB

We also analyzed the effects of the TMB in patients with KRAS aberrations. In total, 1667 patients (250 (15.0%)) had tumors with a high TMB. A TMB-high tumor was defined as the one with ⩾10 mutations per megabase. Among the 468 patients with KRAS aberrations, 80 (17.1%) had tumors with high TMB (Figure 4(a)). We divided patients with KRAS aberrations into two groups: KRAS-SNV and KRAS-CNV. In the KRAS-SNV group, a high proportion of the TMB was observed in gallbladder cancer (N = 1/2, 50.0%) and metastatic tumors of unknown origin (N = 1/2, 50.0%; Figure 4(b)). In the KRAS-CNV group, BLCA (N = 1/1, 100%), gallbladder cancer (N = 1/1, 100%), CCC (N = 3/4, 75.0%), and GC (N = 2/7, 22.2%) showed a high percentage of high TMB (Figure 4(c)). However, the correlation between the score of the TMB and KRAS copy number was relatively weak, with a Pearson’s correlation value of 0.26 (p = 0.06; Figure 4(d)). In the survival analysis, in the KRAS-SNV group, patients with high TMB had significantly longer OS than those with low TMB (median: 754 vs 382 days, p = 0.01; Figure 4(e)). In patients with KRAS amplification, OS was not significantly different between those with high and low TMB (median: NA vs 685 days, p = 0.51; Figure 4(f)).

Figure 4.

(a) Pie chart indicating the percentage of patients with high TMB tumors among 1667 patients and among the 468 patients with KRAS aberrations. (b) Chart showing the proportion of high TMB in each cancer type in KRAS-SNV patients and (c) KRAS-CNV patients. (d) Weak correlation between the score of TMB and copy number of KRAS (Pearson’s correlation value of 0.26 (p = 0.06)). (e) Significantly longer OS in patients with high TMB than in those with low TMB in the KRAS-SNV group. (f) No significant difference in OS between patients with high and low TMB in the KRAS-CNV group.

OS, overall survival; SNV, single-nucleotide variants; TMB, tumor mutation burden.

Discussion

This study reported KRAS aberrations in 1667 patients with various metastatic solid tumors. A total of 468 patients (28.1%) had KRAS aberrations (SNVs, N = 438, 26.3%; CNVs, N = 48, 2.9%) in their tumor specimens. The incidence rate of KRAS SNVs was the highest in pancreatic cancer (N = 89, 89.9%), followed by CRC (N = 259, 47.3%). According to the specific location of the SNV, the mutational type G12D (N = 178, 40.6%) was the most frequent, followed by G12V (N = 95, 21.7%) and G13D (N = 47, 10.7%). The mutational types G12V and G13D influenced poor patient survival (G12V, the mutant type; 245 vs the wild type; 531 days; G13D, the mutant type; 435 vs the wild type; 531 days). The KRAS copy number varied among the tumor types. BLCA (147.9), CCC (16.2), and GC patients (11.0) had relatively high median KRAS copy number. These comprehensive data on KRAS aberrations in various solid tumors will be useful in establishing a treatment strategy for KRAS aberrations.

Adagrasib (MRTX849) is drawing attention as an oral, highly selective, small-molecule, covalent inhibitor of KRAS^G12C. ⁹ However, results of a phase I clinical trial indicated resistance to adagrasib following the initial clinical response to treatment. Resistance to adagrasib has been reported as follows; a case of KRAS Y96D mutation in the switch II pocket of KRAS acquired by adagrasib administration was described as a factor for resistance. ¹⁷ These mutations (H95D/Q/R, R68S) also confer resistance to adagrasib, with the assistance of additional genomic alterations in the RTK-RAS signaling pathway. ¹⁸ Therefore, research on co-occurring mutations is required in the future.

In the present study, the G12V and G13D mutations were associated with poor patient survival. For these mutations, RNA interference has been used in the last two decades to silence the expression of oncogenes and oncogenic signaling effectors and is highly potent in cancer therapies. ¹⁹ Silenseed Ltd. (Jerusalem, Yerushalayim, Israel) provided a remarkable development in the field with their local drug luteR (LODER) system, in which a KRAS G12D-targeting siRNA was embedded in a biodegradable polymeric matrix (siG12D-LODER). In a phase I/II study (NCT01188785), siG12D-LODER was implanted directly into the tumor site of patients with locally advanced pancreatic cancer using a standard endoscopic ultrasound-guided biopsy needle in combination with chemotherapy (gemcitabine or FOLFIRINOX). ²⁰ Evaluation of the primary outcome revealed that the therapy improved progression-free survival in the study population in a 1-year time frame without causing dose-related toxicity. ²⁰ A phase II study (NCT01676259) evaluated this biodegradable siG12D-bearing miniature drug administered as a single dose at 12-week cycles in combination with gemcitabine plus nab-paclitaxel in 80 participants. ²¹ HRS-4642 is a potent, selective KRAS G12D inhibitor that shows strong antitumor activity in preclinical and early clinical studies, with enhanced efficacy when combined with the proteasome inhibitor carfilzomib, partly by promoting an immune-permissive tumor microenvironment. ²² Kim et al. ²³ used doxycycline-inducible CRISPR/Cas9 as a therapeutic tool to target KRAS G12V, G12D, and G13D in CRC cells both in vitro and in vivo, demonstrating that a 7.2-fold reduction in tumor volume was achieved by the knockdown of the G12V mutant by highly specific KRAS G12V single-guide RNA in a xenograft model without altering the wild type allele. On the contrary, other studies have reported that G13D is not different from other mutations. Moreover, patients who had CRC with the KRAS p.G13D mutation appeared to benefit more from cetuximab than those who had tumors with KRAS codon 12 mutations. ²⁴ This meta-analysis demonstrated no significant difference between KRAS G13D and other KRAS mutational tumors in terms of treatment benefit from antiepidermal growth factor receptor monoclonal antibodies for CRC. ²⁵ In this analysis, the use of cetuximab was associated with longer overall and progression-free survival among patients with chemotherapy-refractory CRC with p.G13D-mutated tumors than among those with other KRAS-mutated tumors. ²⁶ However, our study showed that G13D is an important mutation that influences the short OS of patients. Therefore, further studies on this mutation are required.

The copy number range for KRAS amplification varied among solid tumors. BLCA, CCC, and GC had high median copy numbers, particularly in the present study. In our study, KRAS amplification did not affect the OS of the overall population. However, KRAS copy number variations were significantly associated with reduced OS in patients with CCC, CRC, and pancreatic cancer. There was no correlation between KRAS copy number variation and the TMB. Additionally, in patients with KRAS amplification, OS was not significantly different between those with high and low TMB (median: NA vs 685 days, p = 0.51). To date, no studies have demonstrated a clear association between KRAS amplification and the TMB in specific cancer types. Further investigations are warranted to evaluate the effect of immune checkpoint inhibitors in patients with KRAS amplification, particularly in relation to their TMB status.

This study had several strengths. It was based on a large real-world cohort and captured diverse tumor types encountered in routine clinical practice. The use of NGS enabled comprehensive genomic profiling, including the detection of SNVs, CNVs, and TMB. This analysis provides tumor-specific insights into the prevalence and prognostic impact of KRAS variants and integrates clinically relevant biomarkers such as the TMB, PD-L1 expression, and co-occurring mutations. Additionally, KRAS amplification, a relatively underexplored aberration, was evaluated across different cancer types.

This study had some limitations. Its retrospective, single-center cohort study design may have introduced a selection bias and limited generalizability. Treatment and response data were not uniformly available, restricting the assessment of clinical outcomes in relation to therapy. Small sample sizes in rare tumor subtypes limited the statistical power of the subgroup analyses.

Conclusion

This real-world, large-scale genomic analysis provides a comprehensive overview of KRAS aberrations across diverse solid tumors and provides useful information for establishing a treatment strategy for patients with metastatic cancer with KRAS aberrations. These findings support the integration of NGS into routine oncology practice and highlight the need for mutation-specific therapeutic strategies, including optimized sequencing of KRAS-targeted agents and immunotherapies.