Gut microbiota composition and outcomes following neoadjuvant therapy in patients with localized pancreatic cancer: A prospective biomarker study.

Abstract

4143 Background: The prognostic role of the gut microbiome, which modulates cancer development and therapeutic response, is unknown in patients with pancreatic ductal adenocarcinoma (PDAC). With increasing utilization of neoadjuvant therapy (NAT) prior to pancreatic cancer surgery, identification of patient-specific biologic signatures associated with NAT response a priori could help inform PDAC precision medicine. Thus, we assessed the influence of the baseline gut microbiome on clinical outcomes in patients with PDAC receiving NAT. Methods: Stool samples were collected at diagnosis from 42 consenting patients with localized PDAC intending to undergo NAT and surgical resection between 2018 and 2020. 16S rRNA sequencing was completed on pre-NAT stool and resected tumor samples. Microbiota alpha diversity and log transformed taxonomic classifications were utilized in multiple regression analyses to predict oncologic outcomes. Results: Five patients progressed during NAT and eight of the 37 operated patients were deemed NAT responders, demonstrating pathologic near complete or partial response (CAP 0-2), T1N0 AJCC 8th staging, and ≥ 80% CA 19-9 reduction. Compared to the gut, the tumor microbiota demonstrated remarkably reduced Shannon diversity (2.4±0.5 vs 1.5±0.4, p< 0.0001) and distinct taxa (75.2±19.7 vs 8.0±4.5, p< 0.0001) and, accordingly, was not associated with clinical outcomes. Overall, several taxa, including Bacteroides and Akkermansia, were enriched in the gut relative to the tumor ( p< 0.001). The gut microbiota of NAT non-responders had an increased proportion of the gemcitabine metabolizing Enterobacteriaceae (additive log ratio (ALR) -3.4±2.0 vs -8.7±0.8, p= 0.0004). NAT responders, by contrast, had an increased proportion of the adaptive immune cell activating Akkermansia (ALR -0.9±1.2 vs -6.2±3.3, p= 0.0004). The incidence of jaundice or biliary stent placement was associated with relative increases in Enterobacteriaceae (+3.3 ALR, p< 0.01 and +4.9 ALR, p< 0.0001) and decreases in Akkermansia (-3.2 ALR, p= 0.02 and -2.4 ALR, p= 0.057), without altering total gut microbiota diversity. Age, sex, BMI, comorbidities, receipt of pre-NAT antibiotics, diagnostic EUS tumor size, resectability, and pre-NAT CA 19-9 were not predictive of NAT response or survival in a univariate regression model. However, inclusion of gut microbiota data improved the ability of the model to predict NAT response ( p= 0.016, adjusted R2 0.76) and survival ( p= 0.001, adjusted R2 0.89), with Enterobacteriaceae (FDR p= 0.012) and Akkermansia (FDR p= 0.015) being significant negative and positive predictors of NAT response, respectively. Conclusions: The baseline gut microbiota could be leveraged as a biomarker of NAT response and prognosis in patients with pancreatic cancer and warrants further investigation.