Study On Biomarkers And Prediction Model For Pathological Response To Neoadjuvant Chemoradiation In Rectal Cancer Based On Dynamic Metabolomics

BACKGROUND Rectal cancer is one of the most common gastrointestinal cancer.Due to lack of awareness of early screening,most patients are diagnosed at advanced stages or with distant metastasis.Preoperative neoadjuvant chemoradiation therapy(NCRT)followed by total mesorectal excision and additional chemotherapy was a recommended treatment therapy for locally advanced rectal cancer.However,patients show great heterogeneity in pathological response of neoadjuvant chemoradiation therapy.Only 45%patients show tumor regression after preoperative neoadjuvant chemoradiation therapy.At present,pathological tumor response was mainly evaluated according to the 2010 American Joint Committee on Cancer(AJCC)tumor regression grade(TRG)system.However,tumor regression grade system cannot evaluate the pathological response to NCRT before total mesorectal excision.Using biomarker to evaluate the pathological response to NCRT at early stage of treatment and adjusting NCRT according to NCRT sensitivity will improve prognosis of rectal cancer,and has important significance on personalized medicine of rectal cancer.Based on dynamic metabolomics research design,this study aimed to obtain metabolic information from serum sample by ultra-performance liquid chromatography to quadrupole time-of-flight mass spectrometry(UPLC-QTOF-MS),clarify differential metabolic profile,identify potential metabolic biomarkers for NCRT sensitivity,explore the relationship between altered metabolic pathways and NCRT sensitivity and established the optimal NCRT sensitivity prediction model for rectal cancer.METHODS Based on dynamic metabolomics design,this study included 106 patients treated with NCRT,diagnosed with T3-4 and/or N+rectal cancer and without proliferation of cancer.Biological samples and clinical information of the enrolled patients were collected in five consecutive time-points.UPLC-QTOF-MS-based untargeted metabolomics was used to profile human serum samples from included LARC patients.After data pretreatment and data quality assessment,451 peaks and 478 serum samples were included in subsequent analysis.Multilevel simultaneous component analysis(MSCA)and multilevel partial least-squares discriminant analysis(MLPLSDA)were used to explore differential metabolic profile between NCRT-sensitivity patients and NCRT-resistant patients.Then,we used fuzzy c-means clustering(FCM)to explore temporal change patterns in metabolites cluster and identify monotonously changing metabolites during NCRT treatment.Repeated measures analysis of variance(RMANOVA)and MLPLSDA was performed to select dynamic metabolic biomarkers.Meanwhile,we calculated dynamic metabolic biomarkers variabilities,including standard deviation,linear slope,difference between the first follow-up panel measurement and last follow-up panel measurement and ratio of the first follow-up panel measurement to last follow-up panel measurement.Using he first and last follow-up panel metabolic measurement information,we built cross-lagged path analysis model to explore temporal relationship between intercorrelated metabolites,identify metabolite pairs with significant path coefficient.Sensitivity analysis was used to detect differential metabolites pairs between NCRT-sensitivity patients and NCRT-resistant patients.Based on Kyoto Encyclopedia of Genes and Genomes(KEGG)and The Small Molecule Pathway Database(SMPDB),pathway analysis and enrichment analysis were completed by Metaboanalyst platform.A panel of dynamic metabolic biomarker was used to build logistic regression prediction models.To evaluate the classification performance,the area under the receiver operating characteristic curve was computed based on leave-one-out cross validation.Prognostic model was built by cox model to validate association between dynamic metabolic biomarker and rectal cancer.RESULTS A total of 106 patients treated with NCRT and total mesorectal excision were included in this study.In the present study,15605 metabolic peaks were detected using the untargeted metabolomics approach,475 metabolic peaks were detected with identification information.Only 4%Metabolic peaks whose RSDs larger than 30%in QC samples were removed from the dataset,indicating good analytical reproducibility of this metabolomics study.MSCA score plot and MLPLSDA three-dimensional score plot both show tendency of separation in different follow-up panels.We found two cluster,155 serum metabolites with monotonously changing patterns(Clusterl and Cluster 6).Cluster 1 remains decreasing trend during NCRT,while cluster 6 shows increasing trend over time.Using RMANOVA,MLPLSDA,and FCM,eight metabolites with p-value smaller than 0.05 and VIP larger than 1 were identified as dynamic differential metabolites to discriminate NCRT-sensitive and NCRT-resistant patients.Forty-two metabolite pairs with significant path coefficient were selected by cross-lagged path analysis.Sensitivity analysis identify 16 differential metabolite pairs(18 differential one-way metabolic pathway).Compared to logistic models composed of the first follow-up panel metabolic measurement(AUC=0.54,95%CI=0.43～0.65),logistic models composed of differential metabolite pairs(AUC=0.57,95%CI=0.46～0.68)and dynamic metabolic biomarkers variabilities(Standard deviation AUC=0.56,95%CI=0.45～0.67,Linear slope AUC=0.64,95%CI=0.53～0.75,Difference between the first follow-up panel measurement and last follow-up panel measurement AUC=0.67,95%CI=0.57～0.77,and ratio of the first follow-up panel measurement to last follow-up panel measurement AUC=0.60,95%CI=0.50～0.71)show better performance in predicting NCRT sensitivity.Prognostic analysis further validated the association between dynamic metabolic biomarkers variabilities and rectal cancer.L-Norleucine,betaine and acetylcholine show good prognostic ability for rectal cancer patients.Enrichment analysis shows eight differential metabolites were mainly enriched in valine,leucine and isoleucine degradation,phospholipid biosynthesis,purine metabolism,retinol metabolism,glycine and serine metabolism,methionine metabolism,betaine metabolism.Pathway analysis revealed difference in valine,leucine and isoleucine biosynthesis,valine,leucine and isoleucine degradation,pantothenate and CoA biosynthesis,glycerophospholipid metabolism,purine metabolism between NCRT-sensitive and NCRT-resistant patients.CONCLUSIONS This study found eight metabolites as dynamic differential metabolic markers to discriminate NCRT-sensitive and NCRT-resistant patients(L-Norleucine,Betaine,Hypoxanthine,Acetylcholine,1-Hexadecanoyl-sn-glycero-3-phosphocholine,Glycerophosphocholine,Alpha-ketoisovaleric acid,N-Acetyl-L-alanine).Pathway analysis revealed difference in branched chain amino acid,retinol,betaine metabolism and other metabolism pathways between NCRT-sensitive and NCRT-resistant patients.Prediction models composed of dynamic metabolic biomarkers variabilities show better performance in predicting NCRT sensitivity than models composed of the first follow-up panel metabolic measurement.Prognostic analysis further validated the association between dynamic metabolic biomarkers variabilities and rectal cancer.