Assessment of the differential profile of breast intratumoral bacteria in treatment responders and nonresponders.

Abstract

580 Background: Despite improvements in breast cancer therapy, many patients do not respond to treatment: the response rate varies between 31 and 80% [Walks et al., 2019]. The tumor microenvironment contains a microbiome which is both cancer type-specific and different to surrounding normal tissue [Nejman et al., 2020]. It plays an important role in tumour microenvironment crosstalk with tumor cells and has been implicated in disease pathogenesis, being recognised in the latest hallmarks of cancer [Hanahan et Weinberg, 2011]. It has also successfully predicted treatment response [Chen et al., 2022; Hermida et al., 2022]. Here, we use conventional microbiome analysis alongside our proprietary mechanistically-powered technology platform to compare the intratumoral bacteria between responders and non-responders. We hypothesise that there are significant differences between those groups regarding composition and mechanistic function of the bacteria. Methods: The tumor microbiome can be deduced from existing sequencing data of tumor biopsy samples. Using 34 tumor DNA samples from 5 breast cancer studies, we compared the intratumoral bacterial profiles of responders to non-responders of various treatments. We performed microbiome analysis in R, including alpha diversity, compositional abundance profiles, and PERMANOVA (permutational analysis of variance). Additionally, BioCorteX CarbonMirror version dated 2023-02-14 was used to infer mechanistic links between bacterial species and the up- and downregulation of genes implicated in the hallmarks of cancer. Results: The results show significant differences in diversity and compositional profile between responders and non-responders: Non-responders have significantly higher alpha diversity ( p<0.05) and a higher proportion of Proteobacteria. PERMANOVA analysis revealed that while responders are more likely to have commensal bacteria, non-responder tumour microbiomes are more likely to harbor pathogenic species such as Mycoplasmopsis fermentans. Mechanistic analysis showed that for all responders the tumor microbiome is consistently promoting tumor suppressor genes and downregulating proto-oncogenes. In non-responders, however, the tumor microbiome is upregulating and downregulating both with unclear consistency. Conclusions: Our results show microbiome differences between responders and non-responders which were mechanistically implicated in tumor pathogenesis. Further analysis divided by the hallmarks of cancer is required to fully understand the non-responder microbiome links. The reported findings could make the tumor microbiome a useful biomarker aiding patient stratification for both prognosis and treatment choice. They also highlight a potentially meaningful mechanistic link between tumour microenvironment and treatment response that could be leveraged as a novel therapeutic avenue.