Gut Microbiome Predicts Neoadjuvant Chemoradiotherapy Response in Locally Advanced Rectal Cancer Patients

Abstract

Several studies suggest that the gut microbiome modulates tumor response to checkpoint blockade immunotherapy; however, it has not been well-characterized in rectal cancer patients who underwent neoadjuvant chemoradiation. This research aims to evaluate the feasibility of microbiome as a marker for rectal cancer and its ability to predict tumor response to neoadjuvant chemoradiation. This study involved a total of 85 locally advanced rectal cancer (LARC) patients who received neoadjuvant chemoradiation in our institution. We collected 85 pre-treatment fecal samples, 84 post-treatment fecal samples from these patients, as well as 31 fecal samples from normal healthy individuals (N) and had all of these samples complete 16S ribosomal RNA sequencing. We characterized gut microbiome, identified microbial markers and constructed treatment response classifier in these patients. Treatment response was determined by the tumor regression grade (TRG) score in which TRG 0-1 as good response (responders, R) and 2-3 as poor response (non-responders, NR). We constructed a classifier tool in 42 patients with random forest model, and validated it in another 43 patients. Community diversity analysis revealed a significant decrease of Chao1 index in LARC versus healthy individuals, and in post-treatment samples versus pre-treatment samples. Clustering and ordination analysis showed a discrepancy in microbiome composition between R and NR, and results were consistent in different sampling timepoints. With linear discriminant analysis, we further confirmed the discrepancy and found that Bifidobacteriaceae and some common “good” genera in Firmicutes were significantly increased in R group, including Roseburia, Anaerostipes and Doreo. To transform this compositional variation into a predicting tool, we constructed a classifying model whose area under curve (AUC) reached 83.18% (CI: 70.48%-95.88%) in the training set, and 73.48% (CI: 58.24%-88.72) in validation set. Reviewing the top 10 genera in the classifier, 6 genera were significant discriminator between R and NR, and 2 between LARC and N. This study revealed the diverse microbiome characteristics in LARC with different response after neoadjuvant chemoradiation. Gut microbiota-targeted biomarkers represent potential noninvasive tools for predicting treatment response in LARC. Further validation is needed in a larger sample of patients.