The Prognostic Utility of KRAS Mutations in Tissue and Circulating Tumour DNA in Colorectal Cancer Patients

: This study aims to investigate the long-term prognostic utility of circulating tumour DNA (ctDNA) KRAS mutations in colorectal cancer (CRC) patients and compare this with KRAS mutations in matched tissue samples. Tumour tissue ( n = 107) and ctDNA ( n = 80) were obtained from patients undergoing CRC resection and were analysed for KRAS mutations. The associations between KRAS mutation and overall survival (OS), cancer-specific survival (CSS), and recurrence-free survival (RFS) were analysed. All outcomes were measured in years (y). A total of 28.8% of patients had KRAS mutations in ctDNA and 72.9% in tumour tissue DNA. The high frequency of KRAS mutations in tissue samples was due to 51.4% of these being a detectable low mutation allele frequency (<10% MAF). Comparing KRAS mutant ( KRAS mut) to KRAS wild-type ( KRAS wt) in ctDNA, there was no association found with OS (mean 4.67 y vs. 4.34 y, p = 0.832), CSS (mean 4.72 y vs. 4.49 y, p = 0.747), or RFS (mean 3.89 y vs. 4.26 y, p = 0.616). Similarly, comparing KRAS mut to KRAS wt in tissue DNA there was no association found with OS (mean 4.23 y vs. 4.61 y, p = 0.193), CSS (mean 4.41 y vs. 4.71 y, p = 0.312), or RFS (mean 4.16 y vs. 4.41 y, p = 0.443). There was no significant association found between KRAS mutations in either tissue or ctDNA and OS, CSS, or RFS.


Introduction
Colorectal cancer (CRC) was responsible for an estimated 862,000 deaths globally in 2018 which makes it the second leading cause of cancer-related death [1].Advances in treatment and early detection strategies over the last decade have significantly changed the medical management of this disease and improved overall survival.However, the outcomes for CRC patients remain closely related to the stage of cancer at diagnosis.The 5 year survival rate of patients is inversely related to the stage of the disease.At stage I, there is a 99% survival rate, at stage II it is 89%, at stage III it is 71%, and at stage IV it drops precipitously to 13% [2].Whilst most patients with stage II or III disease have good outcomes, there is a proportion of patients who are affected by disease recurrence.It has been shown that there is a survival benefit to offering adjuvant chemotherapy to patients with stage III disease; however, this benefit has not been seen in patients with stage II disease [3].Despite the absence of benefit in patients with stage II disease, there may be a subgroup who would benefit if they were able to be identified by appropriate biomarkers after surgical resection with curative intent.To date, it has not been possible to identify which patients with stage II or III disease will develop recurrence.However, recent evidence from Tie et al., who used sequencing to identify multiple mutations from primary tumour tissue and plasma ctDNA that subsequently guides therapeutic choices in stage II colon cancer patients, has shown non-inferiority to standard management [4].By using a ctDNA-guided approach there was a reduced use of adjuvant chemotherapy without a significant compromise in recurrence-free survival (RFS).
The potential of using KRAS mutations in primary tumour samples as a prognostic or predictive marker has been proposed by many researchers.Early studies suggested that KRAS mutations could be used as a prognostic marker; however, there have been conflicting results from more recent studies .Almost half the research in this field has found no significant association between KRAS mutation status and prognostic outcomes of OS, CSS, or RFS.During the last few years, the research in this area has become even more complicated since more recent studies have replaced tissue tumour samples with ctDNA analysis and the concordance of mutational status between tumour tissue samples and ctDNA varies widely between studies.Most of these studies on tissue and ctDNA have been small, with around 100 patients.KRAS mutations found in these different types of samples could have diverse prognostic or predictive utility and caution should be used when comparing results from these studies.This study aims to add to the evidence regarding possible utility of the long-term prognostic ability of ctDNA KRAS mutation detection in colorectal cancer patients and compare this with KRAS mutations in matched tissue samples.

Sample Collection and Ethics
Specimens were collected from consecutive patients undergoing a CRC resection between July 2011 and December 2013 at either the John Hunter Hospital or Newcastle Private Hospital.The tissue samples were collected from the primary tumour at the time of resection prior to fixation and snap frozen and stored at −80 • C. The blood samples were collected in either K2-EDTA tubes or lithium heparin tubes pre-operatively and were processed within 4 h of phlebotomy.Two centrifugation steps were used to prepare plasma samples, which were then stored at −80 • C (Supplementary S1).For the purposes of this study, a total of 121 patients had suitable specimens that were able to be utilised for either plasma (n = 80) or tissue (n = 107) analysis.A total of 66 patients had both tissue and plasma available for analysis.Histopathological examination and status of the tumour was confirmed by a certified pathologist and staged using the TNM system defined by the Union for International Cancer Control (UICC) [31].Collected patient characteristics included age, gender, body mass index (BMI), smoking status, comorbidities graded as the Charlson Co-morbidity Index (CCI), site of tumour, histopathology, tumour staging, use of adjuvant therapy, disease recurrence, and mortality.This study was conducted in accordance with the Helsinki Declaration and was approved by the Hunter New England Human Research and Ethics Committee (2019/ETH01147, 2019/ETH10205).Patient consent for specimen collection and analysis was obtained prior to their procedures in all cases.

DNA Extraction
A standard ethanol and salt extraction method was used to isolate genomic DNA from the fresh frozen tissue samples (Supplementary S1).The genomic tissue DNA was then prepared for droplet digital polymerase chain reaction (ddPCR) using Zymo DNA Clean and Concentrator kits (Zymo Research, Irvine, CA, USA).Purification of DNA from plasma was performed using Zymo Quick-cfDNA Serum and Plasma kits.The total amount of genomic DNA purified from the plasma samples was quantified using Qubit 2.0, dsDNA high-sensitivity assay (Life Technologies, Carlsbad, CA, USA).All methods were performed according to the manufacturer's instructions.

KRAS-Mutation Testing Using ddPCR
Genomic DNA extracted from plasma and tissue were analysed for 7 KRAS mutations (G12A, G12C, G12D, G12R, G12S, G12V, and G13D) by ddPCR using the KRAS G12/G13 Screening kit (Bio-rad, Hercules, CA, USA).KRAS multiplex analysis was performed using 1-8 µL volume of sample DNA.KRAS master mixes were made for all different volumes used in each run.For each reaction, the master mix contained 1.1 µL of KRAS primer/probe mix, 11 µL of ddPCR Supermix (Bio-rad, Hercules, CA, USA), and autoclaved Millipore water in variable volumes relative to the sample input volume.The sample and master mix were combined to achieve a total end volume of each PCR reaction of 22 µL.The 96-well plate was then sealed, centrifuged at 300 rpm for 5 s, gently vortexed, and recentrifuged at 300 rpm.The plate seal was removed, and the plate was then run on the QX200 AutoDG Droplet Digital PCR system, immediately foil heat sealed using the PX1 PCR Plate Sealer, and run on the C1000 Touch Thermocycler.The PCR cycling conditions were as per the manufacturer's instructions.The plate was then placed into the QX200 Droplet Reader for analysis, and the data were analysed using QuantaSoft software v1.2 (Bio-rad, Hercules, CA, USA).For each PCR plate, there were control samples for both mutation and wild-type KRAS that were made using Horizon reference standards.All assays included a no template control (NTC).

Calculation of the LoD and LoB
The limit of detection (LoD) and limit of blank (LoB) for the ddPCR analysis were measured according to the methods described by Armbruster et al. [32].The LoB was determined by 22 replicates of PCR wells that contained only KRAS wild-type controls.The LoD was determined by the following equation: The LoD using this method was 4.73 droplets.The cut-off for a positive KRAS mutation call was, therefore, conservatively placed at 6 or more positive droplets.Alternate cut-offs of more than 1%, 5%, and 10% mutation allele frequency (MAF) were also used for the tissue cohort in separate analyses.

Statistical Analysis
Overall survival (OS) was defined as the time from the cancer operation until death from any cause.Cancer-specific survival (CSS) was defined as the time from the cancer operation until death from colorectal cancer.Patients were excluded from OS and CSS analysis if they had metastatic disease that did not undergo an R0 resection plus metastasectomy.Recurrence-free survival (RFS) was defined as the time from the cancer operation until the first objective evidence of disease progression or death from colorectal cancer.Patients were excluded from RFS analysis if they survived < 30 days from the initial operation or had metastatic disease that did not undergo an R0 resection plus metastasectomy.The survival and recurrence of patients in each KRAS mutation group were examined using Kaplan-Meier curves and stratified by two potential prognostic factors: KRAS mutation in tissue and cell-free KRAS DNA.The analysis was performed using both total available survival data for each individual as well as survival data adjusted to an endpoint of 5 years.
Associations between the presence of a KRAS mutation and patient survival were examined using Cox regression.For all three survival outcomes, the non-adjusted (crude) effect of the presence of KRAS (in tissue as well as cell-free DNA) was examined for association.For OS and RFS, the effect of the KRAS mutations after adjustment for potential confounders was examined for association.The CCI had the most impact on results during multivariate analysis; hence, the adjusted variables presented utilise this confounder.Cox regression model estimates are presented as estimate hazard ratios (HRs) with 95% confidence intervals (CIs).
The proportional hazard assumptions were assessed by visual inspection of Kaplan-Meier curves assessing time-dependent covariates and plotting simulated cumulative Martingale residuals of each covariate.The assumption of proportional hazards was deemed appropriate.Statistical analyses were programmed using SPSS v28.A priori, p < 0.05 (two-tailed) was used to indicate statistical significance.

KRAS Mutation in ctDNA and Prognosis
Among the 121 patients in this study, there was insufficient plasma for analysis of 41 patients.Insufficient plasma occurred due to the use of plasma samples in another prior study.One further patient had sufficient ctDNA but had incomplete follow-up data so was also excluded.Among the remaining 80 patients, 23 were positive for KRAS mutation in ctDNA (28.8%).Patient demographic data and clinical characteristics by ctDNA KRAS status are shown in Table 1.The frequency of KRAS mutation was found to be significantly higher in non-smokers and those who received adjuvant chemotherapy.All other characteristics were not significantly associated with the KRAS mutation status.Interestingly, well and moderately differentiated tumours were the most predominant tumour grade in both KRAS negative and KRAS positive cases; however, the frequency of poorly differentiated tumours was three times higher in KRAS positive cases.Despite these differences, our analysis did not reveal a significant association between KRAS status and tumour grade.There was no significant association found between OS, CSS, or RFS in patients who tested positive for ctDNA KRASmut compared to those who were KRASwt (Tables 2 and 3, Figures 1-3).Adjustment for potential confounders produced no statistically significant changes.

KRAS Mutation in Tumour Tissue and Prognosis
Tumour tissue was available for analysis from 107 patients.Amongst these patients, 78 were positive for KRAS mutation in tumour tissue DNA when using the LoD cut-off (72.9%).This rate decreased sequentially with cut-off values of 1%, 5%, or 10% MAF, which resulted in frequencies of 24.3%, 22.4%, and 21.5%, respectively.Patient demographic data and clinical characteristics by tissue KRAS status are shown in Table 4.The frequency of KRAS mutation was not significantly different in any of these factors.There was no significant association found between OS, CSS, or RFS in patients who tested positive for tumour tissue DNA KRASmut compared to those who were KRASwt (Table 5, Figures 4-6).However, there was a trend towards decreased overall, cancer-specific, and recurrence-free survival for tumour tissue KRASmut positive cases.The same non-statistically significant trend was seen when analysis was performed using a > 10MAF cut-off point (Table 6).Similarly, although it is not consistent across every model, there was a trend towards an increased HR for all these outcomes (Table 7).Adjustment for confounders produced no statistically significant changes.

KRAS Mutation in Tumour Tissue and Prognosis
Tumour tissue was available for analysis from 107 patients.Amongst these patients, 78 were positive for KRAS mutation in tumour tissue DNA when using the LoD cut-off (72.9%).This rate decreased sequentially with cut-off values of 1%, 5%, or 10% MAF, which resulted in frequencies of 24.3%, 22.4%, and 21.5%, respectively.Patient demographic data and clinical characteristics by tissue KRAS status are shown in Table 4.The frequency of KRAS mutation was not significantly different in any of these factors.There was no significant association found between OS, CSS, or RFS in patients who tested positive for tumour tissue DNA KRASmut compared to those who were KRASwt (Table 5, Figures 4-6).However, there was a trend towards decreased overall, cancer-specific, and recurrence-free survival for tumour tissue KRASmut positive cases.The same non-statistically significant trend was seen when analysis was performed using a >10 MAF cut-off point (Table 6).Similarly, although it is not consistent across every model, there was a trend towards an increased HR for all these outcomes (Table 7).Adjustment for confounders produced no statistically significant changes.+ Adjustment for CCI only displayed in this example, * unadjusted follow-up survival data, ** survival adjusted to 5-year endpoints, MAF = mutation allele frequency, LOD = limit of detection.

Discussion
The results show a trend towards increased risk of disease recurrence and decreased OS and CSS for patients with KRAS mutation found in their tumour tissue.However, these trends were not statistically significant in univariate or multivariate analyses.It is possible that this study was underpowered; however, as previously demonstrated, the number of participants with tissue samples in this study (n = 107) was above the median number of participants (n = 97) found after a review of the literature when searching for articles analysing the prognostic utility of KRAS mutations in CRC tumour tissue or plasma (Chapter 1: Background-Table 1).Regarding features such as the TNM stages included the following: receipt of adjuvant chemotherapy, type of chemotherapy received, type of PCR method used, and methods of survival analysis; the heterogeneity of the research found illustrates the difficulty in comparing the current evidence in this field.
Our results show a non-significant trend towards an increased risk of disease recurrence, a decreased overall survival, and a decreased cancer-specific survival for patients with KRAS mutation-positive tissue at the time of diagnosis and initial treatment.These results were not reiterated in ctDNA, which failed to show any trend or significant association between KRAS mutation status and overall survival, cancer-specific survival, or recurrence-free survival.
The number of patients found to have KRASmut-positive tissue samples was well above the expected range of 20% to 40% [33].However, if the MAF cut-offs of 1%, 5%, and 10% are used, then the frequency falls within the expected range of 21-25%.Despite the higher KRASmut clonal population of cells present in the tumours with these cut-offs, this did not translate into any significant changes in survival or recurrence (Tables 6 and 7).
The limitations of this study are its relatively small sample size, retrospective nature, and lack of complete standardization of patient care and plasma collection methods.The patients were managed clinically at the discretion of the treating surgeon and oncologist, and CEA was not routinely performed for all patients, hence there were insufficient numbers to include this marker in the analysis.The slight differences in the collection of the plasma samples potentially could have affected the overall frequency of ctDNA KRAS mutations.However, since the classification of KRAS mutation status was based on the LoD, this should be unaffected by potential variances in cell-free DNA release at the time of collection due to different sample types.Furthermore, ddPCR has been found to have a high resistance to the effect of potential PCR inhibitors, and Sefrioui et al. found that heparinase treatment of samples did not quantitatively or qualitatively alter the ctDNA detection [34][35][36].Finally, although the sample size is small, it is comparable to other studies in this area and there is a significantly longer follow-up time for this cohort.The sample size was insufficient to perform subgroup analyses for the KRAS status stratified by either adjuvant therapy, site of cancer, or resection margins, which are all known to have effects on oncological outcomes.
Similarly, the sample size was insufficient to perform subgroup analyses based on the specific KRAS mutation type.This is in addition to the fact that the specific assay kit used was not able to sufficiently differentiate the mutation types to allow for certainty.There is mixed evidence regarding the effect that specific codon mutations have on prognoses.Jones et al. and Imamura et al. found that the G12C and G12V mutations conferred a worse prognosis compared to other mutations in codons 12, 13, or 61 [37,38].However, a pooled analysis of patients from five trials found conflicting results that suggested there was a similarly poorer prognosis with G12C mutations but that G12D and G12V mutations had no obvious impacts on OS in univariate and multivariate analyses [39].These studies were all performed on tissue samples, which highlights that there is still ongoing debate about the prognostic utility of KRAS mutations for CRC even when the mutational analysis is performed on the primary cancer itself.Studies that have found an association between KRAS mutations and OS are more likely to have recruited patients with metastatic colorectal cancer rather than early-stage cancer.For instance, Mendoza-Moreno et al. found that at 36 months more patients with peritoneal metastases and KRASwt tumours were alive compared to KRASmut (31% vs. 15%; p < 0.001) [40].Alkader et al. found a similar association of shorter overall survival in KRASmut when compared to KRASwt patients (21 months vs. 17 months) [41].
The poor concordance (44%) between ctDNA and tumour tissue KRAS mutation status in matched patient samples suggests the limited feasibility of ctDNA as a useful biomarker (Table 8).However, the situation is complex and the reason for this discordance is potentially multifactorial.Firstly, the accuracy and sensitivity of ctDNA seem to increase with the overall burden of disease.Several small and large studies have found that the diagnostic accuracy of ctDNA is related to the stage of disease, and, therefore, whilst earlystage tumours could harbor KRAS mutations, these are less likely to be simultaneously found in ctDNA when compared to stage III or IV cancers [20,[42][43][44].This pattern explains why many of the studies utilising ctDNA focused only on metastatic CRC [22,[25][26][27][28][29].Furthermore, the heterogeneity of genetic mutations within CRC means that in many of the tissue DNA samples, there is likely to be a small sub-clonal population of cells with KRAS mutations.This fractional volume of KRAS mutated cancer cells may not shed enough DNA into the circulation to produce a positive result in the plasma whilst still being sufficient to be detected at low volumes in the tumour tissue due to the high sensitivity of ddPCR.This is supported by the fact that the frequency of KRAS mutations drops from 72.9% to 21.5% if a cut-off of an MAF > 10% is used rather than the LoD methodology.Similarly, the concordance between ctDNA and tumour tissue also increases from 44% to 73% with this change in cut-off value.Although this study is not designed to answer any questions regarding the clinical significance of the detection of low-volume KRAS mutations, it is likely that this is related to the emergence of acquired resistance to targeted epidermal growth factor receptor (EGFR) blockade therapy [45][46][47][48].The use of anti-EGFR therapy in CRC is limited to patients with a KRAS mutation-negative status (generally a MAF < 5% on tumour tissue) since it has been shown that this treatment has a reduced effect on OS, CSS, or RFS in patients harbouring a KRAS mutation [49].However, in this cohort, around 50% of the cases were found to harbour low-frequency KRAS mutations that would lead to the development of treatment resistance.There is evidence that the emergence of resistance to anti-EGFR treatment can be overcome if combined with other therapies simultaneously, such as MEK pathway inhibition [50,51].Therefore, the identification of low-frequency KRAS mutations in patients who qualify for anti-EGFR therapy could be an indication for dual first-line therapy with treatment such as MEK inhibitors and is an area worth investigating but is beyond the scope of this study.

Conclusions
Our results show a non-significant trend towards an increased risk of disease recurrence, decreased overall survival, and decreased cancer-specific survival for patients with KRAS mutation-positive tissue at the time of diagnosis and initial treatment.These results were not reiterated in ctDNA, which failed to show any significant association between KRAS mutation status and overall survival, cancer-specific survival, or recurrence-free survival.

Table 1 .
Demographic and clinical characteristics of patients by ctDNA KRAS status.

Table 3 .
OS, CSS, and RFS hazard ratio by ctDNA KRAS status.
+ Adjustment for CCI only displayed in this example, * unadjusted follow-up survival data, ** survival adjusted to 5-year endpoints.Gastroenterol.Insights 2024, 15, FOR PEER REVIEW 6

Table 4 .
Demographic and clinical characteristics of patients by tissue DNA KRAS status.

Table 5 .
OS, CSS, and RFS by tissue DNA KRAS status.

Table 7 .
OS, CSS, and RFS hazard ratio by tissue DNA KRAS status.

Table 8 .
Concordance between ctDNA and tissue KRAS status.