Tumor microenvironment-adjusted prognostic implications of the KRAS mutation subtype in patients with stage III colorectal cancer treated with adjuvant FOLFOX

Several studies have reported that the prognostic effect of KRAS mutations on colorectal cancers (CRCs) varies depending on the type of mutation. Considering the effect of KRAS mutations on tumor microenvironment, we analyzed the prognostic significance of KRAS mutation types after adjusting for the tumor-infiltrating lymphocytes (TIL) and tumor-stromal percentage (TSP) statuses. In two independent cohorts, KRAS mutations were analyzed by Sanger sequencing and/or next-generation sequencing. TIL density and the TSP were quantified from whole-slide immunohistochemical images. KRAS-mutant CRCs were divided into three subgroups (G12D/V, other codon 12 mutations and codon 13 mutations) to examine their differential effect on TIL density, the TSP and recurrence-free survival (RFS). Among the KRAS mutations, only the G12D/V subgroups showed significantly less TIL infiltration than the wild-type CRCs. According to survival analysis, G12D/V mutations were associated with short RFS; codon 13 mutations showed discordant trends in the two cohorts, and other codon 12 mutations showed no significant association. Multivariate analysis further supported the prognostic value of G12D/V mutations. This result is not only consistent with a recent study suggesting the immunosuppressive effect of mutant KRAS but also provides insight into the type-specific prognostic effect of KRAS mutations.


Results
KRAS mutations were identified in 30.6% (n = 86) of patients in the discovery cohort and 37.3% (n = 79) of patients in the validation cohort. Among the CRCs with KRAS mutations, codon 12 mutations accounted for 68.6% (n = 59) of those in the discovery cohort and 83.5% (n = 66) of those in the validation cohort. The majority of codon 12 mutations were G12D and G12V mutations (n = 44 in the discovery cohort; n = 50 in the validation cohort), followed by G12S, G12C, G12A, and G12R, while codon 13 mutations were represented only by G13D (Fig. 1). www.nature.com/scientificreports/ Association of the KRAS mutation type with TIL density and the TSP. When comparing clinicopathologic parameters according to the KRAS mutation status, no significant difference was found between CRCs with wild-type KRAS and CRCs with mutant KRAS except for the density of TILs (Table 1 and Fig. 2). The association between KRAS mutations and TILs was identified by comparisons of CD3(totTILs), CD3(iTILs), CD3(sTILs), CD8(totTILs), CD8(iTILs), and CD8(sTILs) according to specific KRAS mutation types. KRASmutant CRCs were divided into three subgroups: G12D/V, other codon 12 mutation types, and the codon 13 mutation (G13D). The subgroup analysis showed that the TIL density of CRCs with G12D/V was significantly lower than that of wild-type CRCs, and a significant decrease in CD3(sTILs) was observed in both the discovery and validation cohorts. The lower density of CD8(totTILs) and CD8(sTILs) in the G12D/V group was statistically significant but only in the discovery cohort, not in the validation cohort. CD3(totTILs) and CD3(iTILs) tended to decrease in the G12D/V group; although the difference was not statistically significant in the discovery cohort, the difference was statistically significant in the validation cohort. In the discovery cohort, the TSP was www.nature.com/scientificreports/ significantly different between CRCs with wild-type KRAS and CRCs with the KRAS G12D/V mutant (Fig. 3A). However, the validation cohort did not show such a relationship between the TSP and KRAS mutation type (Fig. 3B).

Survival analysis of patients with stage III CRC according to the KRAS mutation type. In the
Kaplan-Meier survival analysis, G12D/V mutations were associated with poor RFS in both cohorts, while there was no difference in survival between the other G12 mutations and wild-type CRCs (P = 0.035 in the discovery cohort; P = 0.006 in the validation cohort, Fig. 4). The G13D mutation was associated with inferior outcomes in the discovery cohort (P = 0.056) but not in the validation cohort, showing inconsistent results. To further demonstrate the value of the KRAS mutation subtype as an independent prognostic factor, we performed multivariate Cox proportional hazards analysis in the discovery and validation cohorts (Tables 2, 3). Additionally, T category, N category, tumor differentiation, lymphovascular emboli, perineural invasion, CD8(iTILs) and    www.nature.com/scientificreports/ the TSP were included in the multivariate analysis because CD8(iTILs) and the TSP were found to be significant prognostic factors in the univariate survival analysis. Of the four TIL parameters, CD8(iTILs) showed the highest prognostic significance in the univariate analysis; thus, the CD8(iTILs) parameter was included in the multivariate analysis (Supplementary Table 1). On the multivariate analysis, the G12D/V mutation was associated with a poor prognosis in both cohorts, with marginal significance (HR = 2.015, 95% CI = 1.064-3.817, P = 0.032 in the discovery cohort; HR = 2.366, 95% CI = 1.109-5.046, P = 0.026 in the validation cohort). The G13D mutation demonstrated a significant effect on survival in the discovery cohort but not in the validation cohort (HR = 3.041, 95% CI = 1.411-6.554, P = 0.005 in the discovery cohort; HR = 0.000, 95% CI = 0.000-Inf, P = 0.997 in the validation cohort).

Discussion
KRAS mutations are common mutations in CRC and are well known as predictive markers for cetuximab therapy, but previous studies on their prognostic value have shown inconsistent results. In the present study, the KRAS G12D/V mutation was associated with a low TIL density and a poor prognosis in stage III CRC patients treated with adjuvant FOLFOX chemotherapy. To the best of our knowledge, this is the first study to validate the association between specific KRAS mutations and TIL density, which was quantified through whole-slide images and computer-based methods 21 . Although the association between a low TIL density and the KRAS G12D/V mutation raised concern over whether the association between a poor prognosis and the KRAS G12D/V mutation might be related to the low density of TILs, the multivariate analysis showed that both the KRAS G12D/V mutation and low TIL density were independent prognostic markers for a poor prognosis. www.nature.com/scientificreports/ The frequency of KRAS mutations was 30-40% in the present study (30.6% in the discovery cohort and 37.3% in the validation cohort). The majority of KRAS mutations consisted of G12D, G12V, and G13D mutations, followed by G12A, G12C, G12S, and G12R. The Cancer Genome Atlas (TCGA) dataset also showed that KRAS mutations were identified in 40.8% of CRC patients (218 out of 534), with G12D (n = 58) being the most common mutation subtype, followed by G12V (n = 49) and G13D (n = 37) 22 . Several studies have reported differences in the survival of CRC patients according to specific subtypes of KRAS mutations. In a large-scale study with stage III CRC patients, KRAS mutations were associated with poor outcome in MSS/MSI-L CRC but not in MSI-H CRC 9 . In MSS/MSI-L CRCs, all codon 12 mutations and G13D mutations were associated with a shorter time to recurrence and overall survival. By contrast, Jones et al. reported that codon 12 mutations were an independent prognostic factor, but codon 13 mutations were not associated with poor overall survival (OS). In particular, both G12V and G12C mutations were associated with poor OS 23 . Similar to these results, Margonis et al. demonstrated that G12V and G12S mutations were independent prognostic factors of poor OS. Additionally, G12V, G12C and G12S mutations were associated with a poor prognosis in patients who experienced tumor recurrence after the resection of CRC liver metastasis 11 . Bai et al. suggested that codon 12 mutations, in particular, G12D and G12V mutations, were associated with a poor prognosis in Chinese patients with metastatic CRC 24 . In the present study, we divided KRAS-mutant CRCs into three subgroups, taking into account previous studies on the mutation rate, prognosis, and relative affinity for RAF kinase of individual KRAS mutation types (G12D/V, other codon 12 mutations, and the codon 13 mutation (G13D)) 25,26 . G12D/V mutations were consistently associated with a poor prognosis in both the discovery and validation cohorts, while the results regarding the effect of the codon 13 mutation on RFS were inconsistent. The subgroup of other codon 12 mutations showed no significant association with survival compared to wild-type CRCs.
Several studies have suggested that oncogenic potential or aggressiveness may differ according to specific KRAS mutations. A previous study demonstrated that codon 12 mutations were associated with resistance to apoptosis and aggressiveness (compared to codon 13 mutations) 27 . Al-Mulla et al. reported the possibility that G12V generates more persistent and potentially oncogenic signals than G12D due to the differences in GTPase activity and affinity for GTP 28 . We focused on the association with TIL density to elucidate the prognostic value of specific KRAS mutation types. Recently, it was recognized that mutant KRAS regulates tumor-associated immune responses and induces the protumorigenic properties of immune cells 29  In the present study, we demonstrated the association between specific KRAS mutations and the density of TILs by whole-slide image analysis in a homogeneous cohort of stage III CRC patients who received adjuvant FOLFOX chemotherapy. G12D/V mutations were consistently associated with a low TIL density and poor outcomes. Considering the effect of MSI on the tumor microenvironment, we analyzed the TIL density according to the KRAS mutations in MSS/MSI-L or MSI-H CRCs. The TIL density of CRCs with G12D/V mutations was still significantly lower than that of the wild-type CRCs in the MSS/MSI-L CRCs of the discovery cohort, and a similar tendency was observed in the validation cohort with no statistical significance ( Supplementary Fig. 1). The MSI-H group failed to show a significant difference due to the limitation of the number of patients in both cohorts (n = 16 in the discovery cohort; n = 18 in the validation cohort).
Galon et al. designed a scoring system, "Immunoscore, " based on the quantification of cytotoxic and memory T lymphocytes or CD3 and CD8-positive T lymphocytes in the center and at the invasive margin of primary tumors 30,31 . Immunoscore was shown to be a powerful prognostic factor in CRCs 32,33 , and superior to MSI in predicting recurrence and survival 34 . Both KRAS mutations and Immunoscore were demonstrated as independent predictors of survival in the studies exploring the prognostic value of Immunoscore in CRCs 35,36 , which is consistent with the results of the present study: both CD8(iTILs) and KRAS G12D/V mutations were found to be independent parameters in multivariate survival analysis. However, to the best of our knowledge, there is no study analyzing the difference of Immunoscore in CRCs according to KRAS mutation subtypes.
The strengths of this study include a well-defined cohort with stage III CRC treated with curative surgery and adjuvant FOLFOX. It also provides accurate KRAS mutation statuses confirmed by both direct sequencing and targeted NGS and computer-based quantification of TIL density by whole-slide image analysis. However, there are some limitations. First, rare mutation types such as mutations in codon 61 or 146 were excluded from this study because direct sequencing was performed only in KRAS exon 2 and not full-length KRAS. Second, the number of patients with each individual mutation type was very small, making it difficult to establish statistical significance. It is necessary to verify these results in larger cohorts and elucidate the mechanism how the KRAS mutation affects the infiltration of various immune cells in the tumor microenvironment. Third, the lack of assessment of BRAF and NRAS mutations which may pollute the KRAS wild-type group, is likely to lead to an insufficient power for the multivariate analysis. The effects of NRAS/BRAF mutation on the tumor microenvironment need to be evaluated in future studies using larger cohorts.
In conclusion, the findings of the present study indicate that among three types of KRAS mutations, G12D/V mutations were consistently associated with less TIL infiltration and shorter RFS in two independent cohorts of stage III CRC patients treated with adjuvant FOLFOX. This finding is not only consistent with a recent study suggesting the immunosuppressive effect of mutant KRAS but also provides insight into the type-specific prognostic effect of KRAS mutations.

Materials and methods
Patients and samples. The discovery and validation cohorts consisted of stage III CRC patients who had undergone curative surgery and received oxaliplatin-based adjuvant chemotherapy; the inclusion and exclusion criteria for these patients were described in detail previously 21 KRAS mutation analysis. Targeted NGS of 40 genes, including KRAS, was performed as described previously 38 . For Sanger sequencing of KRAS, representative tumor portions were marked histologically, and the corresponding areas on unstained tissue slides were then subjected to manual microdissection. The dissected tissues were collected into microtubes containing lysis buffer and proteinase K and incubated at 55 °C for up to 2 days. The direct sequencing of KRAS exon 2 was performed to confirm the NGS results. A total of 281 samples that had consistent results between NGS and Sanger sequencing were included in the discovery cohort. In the validation cohort, mutations in KRAS exon 2 were analyzed by direct sequencing only.
Quantification of TIL density and the TSP from whole-slide immunohistochemical images. For each case, immunohistochemistry for CD3 and CD8 was performed on a representative tumor section, and the stains were subjected to computational quantification of TIL density and the TSP as described previously 21 .
Analysis of microsatellite instability and the CpG island methylator phenotype. The microsatellite instability (MSI) status was determined through the evaluation of five microsatellite markers (BAT25, BAT26, D2S123, D5S346 and D17S250) as standardized by the National Cancer Institute. An MSI-high (MSI-H) status was defined as when tumor DNA had altered alleles in two or more markers compared to normal DNA. An MSI-low (MSI-L) status was defined as when tumor DNA had altered alleles in one marker compared to normal DNA. Microsatellite-stable (MSS) was defined as when no altered allele was present in tumor DNA. The status of the CpG island methylator phenotype (CIMP) was evaluated by a real-time methylation-specific qPCR method (MethyLight) and eight CIMP-specific markers (CACNA1G, CDKN2A, CRABP1, IGF2, MLH1, NEUROG1, RUNX3, and SOCS1). Tumors were classified as CIMP-negative, CIMP-P1, or CIMP-P2 when ≤ 4, 5-6, and ≥ 7 markers were methylated, respectively, as described previously 37 .
Statistical analysis. In this study, statistical analysis was performed using SPSS version 25 (IBM, Armonk, NY, USA). Comparisons between categorical variables were conducted with the chi-square test or Fisher's exact test. To determine whether TIL densities and the TSP were normally distributed, a normality test was performed with Shapiro-Wilk's W test. TIL densities were not normally distributed, while the TSP was normally distributed. Because of these findings, both ANOVA and the Kruskal-Wallis test were performed to identify any difference in the means of parametric and nonparametric tests, respectively, between three or more groups. Student's t test and the Mann-Whitney test were used for the comparison of the means of parametric and nonparameteric tests between two groups, respectively. Survival analysis was performed using the Kaplan-Meier method with the log-rank test. Hazard ratios (HRs) were calculated using the Cox proportional hazards model. All variables that were associated with RFS with P < 0.10 were entered into the model. These variables were reduced by backward elimination. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.
Ethics approval and consent to participate. All patients gave informed consent prior to specimen collection according to our institutional guidelines. The institutional review board of Seoul National University Hospital and Seoul National University Bundang Hospital approved this study. This study was performed in accordance with the Declaration of Helsinki.

Data availability
The data sets used and/or analyzed during this study are available from the corresponding author on reasonable request.