The Microbiome of Aseptically Collected Human Breast Tissue in Benign and Malignant Disease

Abstract

Globally breast cancer is the leading cause of cancer death among women. The breast consists of epithelium, stroma and a mucosal immune system that make up a complex microenvironment. Growing awareness of the role of microbes in the microenvironment recently has led to a series of findings important for human health. The microbiome has been implicated in cancer development and progression at a variety of body sites including stomach, colon, liver, lung, and skin. In this study, we assessed breast tissue microbial signatures in intraoperatively obtained samples using 16S rDNA hypervariable tag sequencing. Our results indicate a distinct breast tissue microbiome that is different from the microbiota of breast skin tissue, breast skin swabs, and buccal swabs. Furthermore, we identify distinct microbial communities in breast tissues from women with cancer as compared to women with benign breast disease. Malignancy correlated with enrichment in taxa of lower abundance including the genera Fusobacterium, Atopobium, Gluconacetobacter, Hydrogenophaga and Lactobacillus. This work confirms the existence of a distinct breast microbiome and differences between the breast tissue microbiome in benign and malignant disease. These data provide a foundation for future investigation on the role of the breast microbiome in breast carcinogenesis and breast cancer prevention.

Figure 1. Ordination plot of samples from 33 women shows a distinct clustering pattern of different sample types (breast tissue, breast skin tissue, skin swab and buccal swab).

The microbiota samples are embedded in the two-dimensional space based on the first two principal coordinates (PCs) from PCoA on the unweighted UniFrac distance. The percentage of explained variability of each PC is indicated on the axis. Each point represents a sample and is colored by sample types (blue diamonds - skin swab, green squares - buccal swab, purple triangles - breast skin tissue, red circles - breast tissue). The ellipse reflects the probability distribution of each sample type.

Figure 2. The microbiota of breast tissue is distinct from skin tissue in rare bacterial lineages.

(A) Barplots of the taxonomic profiles of the breast and skin tissue microbiota at phylum, family and genus level for taxa with a relative abundance >0.5%. (B,C) Rarefaction curves compare the two alpha-diversity measures (observed OTU number (B) and Shannon index (C)) between the two tissue types. (D) Heat map shows the OTU presence and absence of all the tissue samples (column: samples, row: OTUs). The hierarchical clustering (top) is built based on the Euclidean distance of the OTU presence/absence profiles with a complete linkage. (E,F) Ordination plots show the clustering pattern of the two tissue samples based on unweighted (E) and weighted (F) UniFrac distance. (G,H) Differential taxa between breast and skin tissue microbiota based on a permutation test. Taxa with a nominal p value < 0.05 at the family and genus level are shown with their mean abundances in each tissue type (G) and their significance (H). Error bars represent standard error of the mean. (I) Boxplot compares classification error between a microbiota-based predictor (Left: Microbiota) and a predictor solely based on the majority class in the training set (Right: Guess). Random forest based on genus-level abundance is used to build the predictive model and 100 bootstrap samples are used for assessing the classification error.

Figure 3. The microbiota of breast tissue adjacent to invasive cancer is distinguishable from that adjacent to benign disease (BBD-non-atypia).

(A) Ordination plot based on unweighted UniFrac distance shows the clustering pattern of the breast tissue microbiota between the two disease states. (B) Differential taxa between the breast tissue microbiota of malignant and benign states based on a permutation test. Taxa with a nominal p value < 0.05 are highlighted on the cladogram with red and blue indicating increase and decrease in invasive cancer respectively. (C–G) Barplots show the abundances of the five differential genera between the two disease states. Each bar represents a sample. (H) The differential KEGG pathways with a nominal p value < 0.05 between the microbiota of the two states as revealed by PIRCRUSt analysis.