Circulating Tumor DNA By Next Generation Sequencing As A Prognostic And Predictive Biomarker In Metastatic Colorectal Cancer

Background and aim:Colorectal cancer(CRC)is the fourth common malignancy in gastrointestinal tract in China.Only 12.5%of patients with metastases are alive 5 years after diagnosis.Chemotherapy is the usual first-line treatment for patients with metastatic colorectal cancer(mCRC).Biomarkers of treatment response and prognosis,which complement clinical examinations,imaging techniques,and biopsies,would be crucial for optimizing therapeutic strategies.One promising method is the analysis of circulating tumor DNA(ctDNA)in the plasma of mCRC patients,which is cell-free endogenous tumor DNA in peripheral blood.The aim of our study was to explore the applicability of ctDNA as a prognostic and predictive marker in mCRC.Methods:Sequential patients with mCRC receiving standard first-line chemotherapy were included.Both plasma ctDNA and serum CEA were assessed in samples obtained before treatment and after 4 cycles of chemotherapy(C4).Computed tomography(CT)scans were carried out at baseline and post-C4(8-10 weeks)and were assessed using RECIST v1.1 criteria.Target-capture deep sequencing with a panel covering 1021 genes was performed to detected somatic mutations in ctDNA.Clonal population structures were identified based on variations from ctDNA using Bayesian cluster with Pyclone.Molecular tumor burden index(mTBI)was calculated with the mean variant allele frequency of mutations in trunk clonal population.Results:A total of 20 patients were prospectively included and treated with FOLFOX(15/20)or FOLFIRI(5/20)from Sep 2015 to Aug 2016.Median follow-up was 6.9 months(range 1.6-26.6).Somatic mutations for baseline ctDNA analysis were identified in 850%(17/20)of the patients.The accordance of RAS and BRAF mutations identified in tumor tissue and pre-treatment ctDNA levels were 100%(6/6)and 100%(5/5),respectively.In a multivariable analysis,a high baseline mTBI was associated with the risk of disease progression(HR,1.091;95%CI:1.032-1.153;P=0.002).In a bivariate correlations analysis,baseline mTBI was related to progression-free survival(PFS)(Spearman correlations,P=0.0083,r=-0.5725).The optimal baseline mTBI for predicting PFS,as determined by the ROC curves(ROC area=0.83,P=0.0126),was 6.81%.Patients with baseline mTBI below 6.81%had longer PFS compared to those above(median 9.9 versus 4.35 months;HR,2.966;95%CI:1.075-8.184;P=0.0115).Mutation variations of ctDNA after chemotherapy were tested in 16/20(80%)of the patients.In a multivariable analysis,emergence of new detected mutations was related to shorter PFS(HR,5.965;95%CI:0.7019-50.69;P=0.0003).Patients with new detected mutations had shorter PFS compared to those without(median 3.038 versus 7.25 months;HR,2.966;95%CI:1.075-8.184;P=0.0115).Fold changes in mTBI from baseline to post-C4 were obtained in 16/20(80%)of the patients.In the Mann-Whitney U-test,there was a statistically significant difference in fold change in mTBI between groups of patients with tumor response and non-response(P=0.0085).In a bivariate correlations analysis,fold change in mTBI was related to both tumor response(Spearman correlations,P=0.005,r =-0.661)and PFS(Spearman correlations,P=0.0023,r=-0.7201).The optimal fold reduction in mTBI for predicting PFS,as determined by the ROC curves(ROC area=0.8438,P=0.0209),was 0.8-fold.Patients with fold reduction in mTBI above 0.8-fold had longer PFS compared to those below(median 9.29 versus 4.1 months;HR,4.506;95%CI:1.294-15.7;P=0.0008).In contrast,neither serum CEA level nor its variation did correlate with PFS or tumor response,respectively.Conclusions:High baseline mTBI and inferior PFS were correlated in mCRC.Fold reductions in mTBI were related to both tumor response and PFS.Therefore,ctDNA could be potentially a biomarker for prognosis and efficacy.Background and aim: Approximately 25% of the patients with colorectal cancer(CRC)present with raetastases at initial diagnosis,and nearly 50% of patients with CRC will develop metastases,thus contributing to the high mortality rate of CRC.If a patient' s non-response to one treatment could be reliably assessed earlier on in the treatment process,an earlier switch to an alternative therapy could help optimize the treatment of metastatic colorectal cancer(mCRC).While traditional serum tumor markers,such as carcinoembryonic antigen(CEA)and carbohydrate antigen 19-9(CA19-9),and imaging method have limitations,it is mandatory to find new markers.The results of our exploratory study in part 1 have suggested that changes in ctDNA after chemotherapy might be related to radiologic response in the patients with mCRC;monitoring changes in ctDNA by collecting blood samples at multiple timepoints could help to understand the tendency of ctDNA variations and find an optimal method to predict tumor response early.This prospective study was designed to explore the serial changes in plasma circulating tumor DNA(ctDNA)as an early marker of therapeutic response to systemic treatment in mCRC.Methods: Sequential patients with mCRC receiving standard first-line chemotherapy were included.Serial blood samples for ctDNA analysis were collected,and traditional serum tumor markers(CEA and Cal9-9)were measured at the following four defined time-points: pretreatment(within seven days before commencing cycle1 treatment,named CO)and prior to cycles 2,3,and 4 of treatment(within three days before commencing the next cycle of treatment,named Cl,C2,and C3,respectively).CT of the chest,abdomen,and pelvis was performed at baseline and after four cycles of therapy(usually 8-10 weeks after starting treatment)and assessed using RECIST vl.1 criteria.The DNA sequencing were generated with the Ion AmpliSeq Library Kit 2.0 using a panel to detect somatic mutations from plasma samples in 50 cancer-related genes.The mutation of maximal frequency in pretreatment plasma ctDNA was selected as each patient' s candidate mutation for analysis.Receiver operating characteristic(ROC)curves and the Z test for the area under curve(AUC)were used to determine the appropriate parameters(changes in ctDNA,CEA,and CA19-9 levels after treatment)and the optimal cutoff for differentiating patients with responses of PD and non-PD at the first restaging. A classification tree analysis based on chi-squared automatic interaction detection(CHAID)that considered all of the predictive variables with statistical significance was used to identify the ideal indicator to discriminate imaging PD and non-PD patients.Results: Forty-seven mCRC patients receiving first-line therapy were enrolled.Mutations in pretreatment ctDNA could be detected in 95.7% of patients(45/47).The RAS mutation status between the matched plasma and tissue from each patient could be analyzed in 30 patients,which was concordant in 29 of 30 cases(overall agreement: 96.7%).41 patients were monitored serially and treated with F0LF0X(36/41)or FOLFIRI(5/41).The median follow-up was 8.5 months.In the MannWhitney U-test,significant differences were observed in the changes in ctDNA levels between the non-progressive disease(PD)group and the PD group after the first,second,and third cycles(P = 0.000,P = 0.004,P = 0.001,respectively).However,when comparing the changes in CEA and CA199 levels after treatment between the two groups,a significant difference was only observed after the third cycle of treatment(P = 0.007,P = 0.006,respectively).ROC curve analyses indicated that all the changes in ctDNA from pretreatment to after cycles 1,2,or 3 had certain values in the prediction of radiologic disease progression(ROC area =0.987,0.954,0.992,respectively;the optimal cutoff log2 values =-0.126,-0.655,-0.471,respectively);for CEA and CA19-9,only changes from pretreatment to after cycle 3 had the potential to predict disease progression(ROC area =0.885,ROC area = 0.896,respectively).Classification tree analyses indicated that changes in ctDNA after cycle 1 treatment provided the best predictive capacity.The optimal log2 value of fold-change in ctDNA after cycle 1(log2(C1/C0))for predicting PD was-0.832.In the Kaplan-Meier survival analysis,patients with ctDNA log2(C1/C0)above-0.832 showed significantly worse progression-free survival than did those with log2(C1/C0)below-0.832(median 2.5 versus 9,0months;P =0.016).Conclusions: The present exploratory study suggests that early changes in ctDNA that are detected via targeted sequencing might potentially predict later radiologic responses in mCRC.More reduction in ctDNA after cycle 1 and superior PFS were correlated.Therefore,early changes in ctDNA could be potentially a bioraarker for both tumor response and prognosis.