Preoperative detection of KRAS mutated circulating tumor DNA is an independent risk factor for recurrence in colorectal cancer

Preoperative ctDNA status in relation to recurrence in cases of CRC remains unclear. We examined preoperative ctDNA detection by targeting KRAS gene mutations as a predictive marker for recurrence after CRC surgery. We measured the preoperative KRAS mutated ctDNA status and analyzed the correlation with clinicopathologic features of 180 patients that underwent surgery for CRC. We studied the association between preoperative KRAS mutated ctDNA and postoperative recurrence in patients (n = 150) that underwent radical surgery. KRAS mutated ctDNA was detected in 59 patients (32.8%). Median mutant allele frequency of KRAS in ctDNA was 0.20%. KRAS status in ctDNA and lymph node metastasis and distant metastasis were not significantly different. Among patients that underwent radical resection, recurrence occurred in 21 (14.0%, median follow-up 24 months). In Kaplan–Meier analysis, preoperative detection of KRAS mutated ctDNA was associated with inferior recurrence-free interval (RFI) (p = 0.002) and recurrence-free survival (RFS) (p = 0.025). In a multivariate Cox proportional hazards model, preoperative detection of KRAS mutated ctDNA was an independent factor related to both RFI (HR = 3.08; p = 0.012) and RFS (HR = 2.18; p = 0.044). Preoperative measurement of KRAS mutated ctDNA could be useful to decide postoperative treatment.

KRAS mutation in ctDNA and lymph node and distant metastasis. Among 180 patients, 59 were positive for KRAS mutation in ctDNA (32.8%). The median MAF of KRAS in ctDNA was 0.20% (range 0.04-68.99%). Patient demographic data and KRAS status in ctDNA by clinicopathologic factors are shown in Table 1. The frequency of KRAS mutation positive in ctDNA was higher in cases with histological types other than well-differentiated, in cases with invasion depth of T3 or T4, cases with lymphatic invasion, cases with venous invasion, and stage IV cases. However, multivariate analysis did not reveal significant difference in these factors. Subsequently, we investigated the relationship between KRAS status in ctDNA and lymph node metastasis and with distant metastasis (Table 2). Factors such as invasion depth of T3 or T4, lymphatic invasion, venous invasion and positivity for preoperative serum CEA had correlation to lymph node and distant metastasis. Regarding KRAS status in ctDNA, cases with preoperative KRAS mutation positive for ctDNA tended to have lymph node metastasis and distant metastasis. Multivariate analysis showed venous invasion was significantly related to lymph node metastasis (p = 0.011), but there were no significant differences between KRAS status in ctDNA and lymph node metastasis or distant metastasis (Table 3).  Among the 150 patients with stage III or lower stage CRC that underwent radical resection, recurrence occurred in 21 patients (14.0%), including 12 of the 45 patients with detectable KRAS mutations in ctDNA (26.7%) and 9 of the 105 patients without such mutations (8.6%). Fifteen of the 150 patients (10.0%) died before median follow-up of 24 months (range 12-37 months). Patients with preoperative detectable KRAS mutations in ctDNA had an increased risk of recurrence relative to those without them (p = 0.003). Kaplan-Meier analysis showed that preoperative detection of KRAS mutant ctDNA was associated with inferior recurrence-free interval (RFI) (p = 0.002) (Fig. 2a) and recurrence-free survival (RFS) (p = 0.025) (Fig. 2b). In a multivariate Cox proportional hazards model, preoperative detection of KRAS mutated ctDNA was the significant factor correlated to both RFI (HR = 3.08; p = 0.012) and RFS (HR = 2.18; p = 0.044) ( Tables 4 and 5).

Discussion
KRAS mutation in CRC primary lesions were reported to be associated with a high risk of postoperative recurrence [18][19][20] . We demonstrated that preoperative KRAS mutation in ctDNA is an independent risk factor for recurrence in patients with CRC. Although KRAS mutation in ctDNA may be detected in plasma without malignancy, it is extremely rare. Hence, preoperative measurement of KRAS mutation in ctDNA could reflect the primary lesion 23,24 . The concordance rate of KRAS status between cancer tissue and ctDNA has been reported as 75-96% [25][26][27] , because the measurement of gene mutations in tissue has a problem of heterogeneity. Meanwhile a previous study of pancreatic cancer showed that mutant KRAS in plasma was significantly associated with recurrence and prognosis, but not in tumor tissue samples 27 . The measurement of KRAS mutation in ctDNA or in primary tumors may reflect different aspects of CRC. www.nature.com/scientificreports/ In recent years, several reports have demonstrated "postoperative" detectable ctDNA as a predictive factor for recurrence [11][12][13] . However, the detection rate of ctDNA has been reported to decrease after surgery. Reinert et al. reported that in stage I-III CRC, the detection rate of ctDNA decreased from 88.5% preoperatively to 10.6% postoperatively 13 . If ctDNA mutations in "postoperative" plasma samples after radical resection can be detected, it may indicate remnant cancer cells resulting in recurrence and poor prognosis. However, if patient has a small amount of cancer cells after radical surgery, ctDNA mutations could not be detected. Preoperative KRAS mutations in ctDNA reflecting mutations of primary lesion may present malignant phenotype of primary tumor. Therefore, we analyzed "preoperative" ctDNA mutations, but not postoperative mutations in the current study. A comparison between preoperative and postoperative ctDNA may provide more information regarding recurrence and prognosis. Further investigation is needed to address this issue. In addition, preoperative ctDNA positive cases were reported to have a high risk of recurrence in localized pancreatic cancer, even if postoperative  www.nature.com/scientificreports/  www.nature.com/scientificreports/ ctDNA becomes negative 15 . We showed that preoperative KRAS mutation in ctDNA is associated with recurrence of CRC. Preoperative detection of KRAS mutated ctDNA may provide adequate postoperative screening and appropriate postoperative adjuvant chemotherapy. In the current study, the significance of preoperative ctDNA measurement to determine the indication for adjuvant chemotherapy was not clarified due to a small number of patients. Clinical trial is needed to address this issue. Circulating tumor DNA is a novel means of detecting early phase CRC recurrence. Postoperative detection of ctDNA in stage II or III CRC reflected minimal residual disease and predicted recurrence 11,12 . It was measured in these reports by NGS, but clinical application was very costly. ddPCR has been reported to measurable at a lower cost and in a shorter time, yet with higher sensitivity than NGS 26 . In the current study, KRAS mutated ctDNA measured by ddPCR was significantly correlated with recurrence of CRC and was an independent risk factor for recurrence of CRC.
Our study has several important limitations; only a comparatively small number of patients that had recurrent CRC were included, and the study was of explorative design and there was no validation cohort. Further investigations are required to address these issues.
In conclusion, the presence of KRAS mutated ctDNA before surgery was significantly associated with recurrence after radical resection in cases of CRC. Preoperative KRAS mutated ctDNA measurement was suggested to be a potentially useful biomarker to predict postoperative recurrence. Recurrence may be reduced by administering adjuvant chemotherapy to ctDNA positive patients.

Methods
Patients. In this study, we investigated the relationship between KRAS status in ctDNA, lymph node metastasis, distant metastasis and clinicopathologic factors including age, gender, tumor site, differentiation, tumor depth, lymphatic invasion, venous invasion, preoperative serum CEA value, co-morbidity and adjuvant chemotherapy. In order to consider the effect of co-morbidity on recurrence and prognosis, we evaluated by using the Charlson Comorbidity Index (CCI) 28,29 .
Enrolled in this study were 183 patients with CRC that underwent surgery at the Second Department of Surgery, Wakayama Medical University, between April 2017 and December 2018. We excluded patients that received preoperative treatment, such as chemotherapy, radiotherapy or endoscopic resection. We also excluded cases diagnosed with other primary cancers and cases with other tumors found by preoperative imaging examinations. In addition, in three cases, surgery was performed for preoperative clinical diagnosis of CRC, and the patients were not diagnosed with adenocarcinoma by postoperative pathological diagnosis, and these were also excluded from statistical analyses. All research was performed in accordance with relevant guidelines/regulations. This study was approved by the Wakayama Medical University Human Ethics Review Committee (Approval Number 1949) and informed consent was obtained from all included patients.
Blood sample collection and extraction of cfDNA. Just before the start of surgery, 5 mL blood samples were obtained in EDTA tubes from each patient and centrifuged at 1900g for 10 min within 2 h after collection. Plasma was collected and stored at − 80 °C until use. After thawing plasma samples, they were centrifuged at 16,000g for 10 min. cfDNA was extracted from 2 mL of plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen) according to the manufacturer's instructions. Samples were eluted in 75 μL elution buffer and cfDNA was frozen at − 80 °C until analysis. Blood for CEA was collected at the first visit and measured within 2 h. Detection of KRAS mutated ctDNA. KRAS mutations in ctDNA was analyzed by ddPCR. QX200 Droplet Digital PCR system (Bio-Rad) and ddPCR KRAS multiplex assays including G12A, G12C, G12D, G12R, G12S, G12V, G13D mutations (Bio-Rad) were used according to the manufacturer's protocols. A reaction volume of 20 µL including 8 µL of cfDNA was used as a template for each PCR. Droplets were generated using the QX200 droplet generator (Bio-Rad) and PCR reaction was performed in a C1000 Touch Thermo Cycler (Bio-Rad) under the following conditions: 95 °C for 10 min, 40 cycles of 94 °C for 30 s and 55 °C for 1 min, and 98 °C for 10 min. Data analysis were performed using the Quantasoft software Ver1.7.4 (Bio-Rad).
Statistics. Statistical analysis was performed using JMP ver. 14.1.0 (SAS Institute). Differences between groups were determined using Pearson's chi-squared test to compare categorical variables as appropriate. Factors with p < 0.10 on univariate analysis were analyzed by multivariate logistic regression, and an odds ratio with a 95% confidence interval was calculated for each factor. The Kaplan-Meier method was used to estimate recurrence-free interval (RFI) and recurrence-free survival (RFS), and the log-rank test was used to determine the statistical significance. Cox proportional hazards model was used to assess the risk ratio under simultaneous contributions from several covariates. Final statistical results were considered significant at p < 0.05.

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.