Although distinct dysbiotic microbial signatures related to disease states are being increasingly recognized using high throughput sequencing techniques, the microbiome of healthy breast tissue is under explored. Here, we characterized the microbiome of the normal breast by examining a large cohort (N = 403) of breast tissue biopsies donated by healthy women. We identified bacteria uniquely abundant in the normal tissue as compared with the uninvolved tissue adjacent to tumor (NAT) and tumor and, for the first time, correlated their level with BC risk factors and host transcriptomic changes. Our data also confirmed that, as previously reported [42], NAT, often used as a surrogate for healthy controls, displays bacterial dysbiosis as compared with the normal breast tissue from healthy donors, which is similar to what observed in tumor samples.
Recent findings demonstrated the existence of microbiota in internal organs once believed sterile, including the lung, pancreas, and breast [15]. Regarding the origin of the microbiota in the breast, several hypotheses have been proposed and investigated including the skin, via the nipple-areolar orifices, nipple-oral contact via lactation and/or sexual contact, and, more recently, translocation from the gastrointestinal tract [16, 43]. It is suggested that organ-specific microbiota plays a role in tissue homeostasis, tumor development and therapeutic resistance [17]. New discoveries in the cancer-related microbiome have been made possible with the use of next-generation sequencing technologies. However, considerable differences in the employed methodologies, with respect to specimen treatment after collection, DNA isolation, target hypervariable region selection for sequencing, and sequence analysis workflows, have resulted in considerable heterogeneity in results, delaying the assessment of the existence of a link between dysbiosis and BC. Inconsistency in choice of hypervariable region amplified to define the breast tissue microbiota remains a major concern, especially since specific hypervariable regions are more likely to identify certain taxa [44]. In our study of the breast tissue, five amplicons covering the nine 16S rRNA hypervariable regions were examined (V1V2, V2V3, V4V5, V5V7, and V7V9), and our findings confirmed the data from He et al., showing a difference between the amplified regions in both alpha diversity, with the V2V3 displaying the lowest diversity, and number of reads, with V1V2 generating overall a higher ASV count.
Investigation of low biomass specimens such as the breast tissue, where the microbiota abundance is relatively limited as compared to microbe-rich organs as the guts, is challenging. The analysis of such low biomass tissue needs to consider the impact of external contaminants and experimental artifacts [37]. We used multiple “no template controls” such as storage buffers, elution buffers, or water in our study as good approximates for contaminants introduced during sample collection and storage, extraction and library preparation steps. Hence, our analysis revealed not only the main contaminants (i.e., Burkholderiaceae and Propionibacteriaceae) often reported abundant in BC [45–47], but also that contaminants composition varied in relation to both negative control type and amplicon primers used for the 16s rRNA sequencing. This clearly indicates that appropriate negative control needs to be included in the workflow for microbiome analysis of the breast.
In breast tissue, Proteobacteria and Firmicutes have been reported to be the most abundant phyla, which is distinct from other tissues where these phyla, especially Proteobacteria, represent a small portion of the total bacterial load [10, 15, 17, 42]. Analysis of breast tumors and NAT from the same patient showed unique microbial communities associated with tumors, with the high abundance of Sphingomonas yanoikuyae in normal tissue and Methylobacterium radiotolerans in tumor tissue [17]. Moreover, Banerjee et al detected a distinct microbial signature associated with TNBC [18, 48]. Urbaniak et al reported that NAT from women with BC compared to tissue from healthy controls had higher relative abundance of Bacillus, Enterobacteriaceae, and Staphylococcus [10]. However, a study with a Mediterranean population found more similarities than differences between NAT and tumors [49]. A recent publication from Tzeng et al revealed that tumor tissues contained a much higher percentage of the families Pseudomonadaceae and Enterobacteriaceae and the genera Pseudomonas and Proteus [8]. Recently, Hoskinson et al reported the compositional shifts in bacterial abundance in NAT and tumor tissues as well as breast tissues prior to a clinical manifestation of cancer as compared with healthy breasts [21, 42]. Independent from the findings’ variability, these reports suggest a link between microbial dysbiosis and BC. The variability in the specific bacteria identified can be largely explained by differences in methodology, not only in the care to avoid contaminants, but importantly, in the choice of “normal” controls. In early microbiome investigations, breast tissue obtained from women undergoing reduction mammoplasty were used as poor substitutes for healthy controls [10, 15]. These tissues show hyperproliferative phenotype [50]. More recently, the histologically normal tissue surrounding the tumor lesions has been used to as the ‘healthy’ control in these experiments [8, 47]. However, multiple recent publications have documented a “field effect” of BC, with histologically normal tissue displaying both genetic and epigenetic aberrations [21–24, 51].
With the exception of the investigation by Hoskinson et al where 49 breast tissue cores from healthy women were examined, our study represents the first large scale analysis of the microbiota of the normal breast. We detected Lactobacillaceae (Firmicutes phylum), Acetobacterraceae, and Xanthomonadaceae (both Proteobacteria phylum) as more abundant families (> 2%) and Acetobacter and Liquorilactobacillus as the more abundant genera in the normal breast as compared to both the NAT and tumor samples. These findings further confirmed the biological difference between normal and NAT tissues. The report by Hoskinson et al revealed the similarity in microbial abundance between NAT and tumor tissue as compared with healthy breasts [42].
Four main bacterial species were identified as predominant in the normal breast as compared with the other tissue here analyzed: Lactobacillus paracasei, Lactobacillus vini, Acetobacter aceti, and Xanthomonas sp. In exploring the roles of these bacteria in the breast, the literature pointed to possible involvement in metabolic pathways. The genus Lactobacillus, including Lactobacillus paracasei, was previously reported to be more common in healthy breast tissues than in cancerous tissues and, because of its immunomodulatory effect, may have a role in BC prevention [19, 52]. Lactobacillus vini, generally isolated on organic matrices, was detected as member of human fecal microbiota for the first time by Rossi et al [53]. Acetobacter aceti is an aerobic bacterium widespread in sugary, acidic and alcoholic niches. The investigation of this bacterium in human tissues is limited. Aghazadeh et al showed that a strain of the same genus, Acetobacter syzygii, exhibited significant cytotoxicity towards a squamous cell carcinoma cell lines [54] suggesting that, similarly to Lactobacillus, this bacterium may also have beneficial properties in the breast. Interestingly, synergistic interaction between Acetobacter and Lactobacillus was reported in Drosophila melanogaster gut and led to nutriments availability modulation with reduction in hist triglyceride [41]. In the Aghazadeh study a direct correlation between the levels of these bacteria in the normal breast was also observed. The family Xanthobacteraceae was reported decreased in abundance in NAT and tumor tissues compared with normal breast from healthy women, thus confirming our data [42]. Also, Xanthomonas sp. was recently found abundant (6%) in normal breast of Chinese women [45], but the literature on the role of this microbial species is human is very limited. Although, because of the decontamination approach here employed, the ASV levels were lower than those previously described, consistently with previous reports [19], tumor samples displayed an abundance of Enterobacteriaceae (1.2%), Staphylococcus (2%), Corynebacteriaceae (1.7%), Corynebacterium (0.98%), Anoxybacillus (3%), Prevotella (0.98%), and Rothia (0.5%) as compared to the Normal group. Nevertheless, Ralstonia (Proteobacteria phylum), almost absent from the normal tissues, showed higher abundance in both NAT and tumor as compared with the Normal (p = 0.014, p = 0.018, respectively). This bacterium was previously detected in human milk [55] and silicone breast implant biofilms [56]. Our data confirm previous findings indicating Ralstonia as the most dominant bacterial genus in the breast tumor tissues [57–59], however, we also showed Ralstonia presence in the NAT. Interestingly, the fact that other studies report the abundance of Ralstonia in Normal breast is explained by their use of NAT as source of normal tissue.
Exogenous and endogenous factors can promote fluctuations in microbial abundance and functions. Many of the important life-style risk factors for cancer like obesity, smoking, diet, and alcohol can also cause perturbations in the microbial composition [40]. Here, we examined the correlation between the abundance of Ralstonia, Acetobacter aceti, Lactobacillus vini, Lactobacillus paracasei, and Xanthomonas sp. in a cohort of normal breast tissue cores (N = 190) and BC risk factors, such as age, parity, breastfeeding, smoking, alcohol consumption, age at menarche, and body mass index. Moreover, a cohort of 58 breast tissues from women either at genetic risk for BC or who developed BC post tissue donation [21, 42] was examined with respect to microbial enrichment. As opposed to the work by Tzeng et al, where subtle differences in microbial profiles between healthy control and high-risk tissues were detected [8], no significant difference was observed except for Acetobacter aceti being less abundant in genetically predisposed breast tissue. Ralstonia failed to show correlation with any of the examined risk factors, probably due to its low level in the normal tissue. The Normal-specific bacteria inversely correlated with age, the strongest risk factor for BC, and were enriched in nulliparous and parous women who did not breastfeed as compared with parous women who did breastfeed. Whether these bacteria, abundant in the normal breast and, except for Acetobacter aceti, previously detected in breast milk [60], are lost via breastfeeding requires further investigation. Racial background is another key determinant of BC. While BC incidence overall is higher in Caucasian women, African American women are at a higher risk of developing triple-negative BC, more aggressive disease. Unexpectedly, Acetobacter aceti, Lactobacillus vini and paracasei, and Xanthomonas sp. were enriched in normal breasts from African American women as compared with the Caucasian and Asian cohorts. Interestingly, Lactobacillus was previously reported highly abundant in TNBC from White non-Hispanic patients as compared with Black non-Hispanic tumor tissues [57], whereas the family Xanthomonadaceae was abundant in White non-Hispanic tumors [58]. Although previous reports evaluated Ralstonia abundance in breasts from Black non-Hispanic and White non-Hispanic women, their results were discordant with our current study because they used NAT as control tissue [57, 58].
Next, to test the hypothesis that microbial–host crosstalk may influence the tumor microenvironment, we examined the transcriptome changes in 190 normal breast tissues and their association with the abundance of Ralstonia, Acetobacter aceti, Lactobacillus vini, Lactobacillus paracasei, and Xanthomonas sp. Overall, our findings confirm the data from Tzeng et al, where the breast bacteria exhibit significant associations with immunomodulatory genes [8]. The gene set enrichment analysis of the DEGs linked with microbial level in normal breast revealed the involvement of immune pathways including the IL17 signaling (Acetobacter aceti), T cell receptor signaling (Lactobacillus paracasei), inflammatory response (Lactobacillus vini), and phagocytotic process [61](Lactobacillus vini and Xanthomoas sp). Moreover, although the influence of Acetobacter aceti on the tissue homeostasis appeared limited, its abundance inversely correlated with keratin 16 (KRT16), a structural protein recently shown to regulate innate immunity in response to epidermal barrier stress [62]. A recent investigation by Hoskinson et al, where Spearman’s rank correlation analysis between the host transcriptome and microbial taxa and genes in healthy breast tissues was performed, identified the gene CYP24A1, encoding for 24-hydroxylase, inversely associated with a number of bacterial nutrient transport and metabolic pathways [42]. Similarly, we found the relative abundance of certain microbes in the normal breast from healthy women to be linked with metabolic pathways such as lactose and galactose metabolism as well as fatty acid and cortisol synthesis. Specifically, in our study the abundance of both Lactobacillus vini and Xanthomonas sp resulted inversely correlated with SCD expression, which encodes a stearoyl-CoA desaturase involved in fatty acid biosynthesis and whose elevated expression in human BCs predicts poor survival [63]. Interestingly, tissues abundant in Ralsonia, in addition to upregulation of carbohydrate metabolism-related genes, presented a significant downregulation in DOK7, which was recently reported to inhibit proliferation, migration, and invasion of BC cells though the PI3K/PTEN/AKT pathway [64]. Further mechanistic investigation is required to link these bacterial species to any functional role of this microbial population in heathy breasts.
Overall, our data show a distinctive microbiota for the normal breast tissue, which seem to reduce and even disappear in NAT and, in a greater degree, in tumor. However, no correlation with Tyrer-Cuzick risk score was identified, thus suggesting that dysbiosis in the breast may occur late during the carcinogenesis process and might be a consequence rather than the cause of the changes in the microenvironmental milieu dictated by the already established tumor cells. An exception is given by the recently reported relationship between breast and gut microbiome. The breast microbiota composition can be directly [43] and indirectly [40, 65–67] influenced by alterations of the gut microbiota. Moreover, recently, Parida et al identified a gastrointestinal pro-oncogenic bacterium, Bacteroides fragilis, in breast tumor samples and suggested a potential role in BC initiation [43].
This study bears several limitations, including the lack of experimental validation of the data via immunostaining and mechanistic insights into the role of the identified microbial species in altering the breast tissue homeostasis and promoting BC. In vitro investigation shall follow to address these points. Moreover, we detected a lack of correlation between the bacteria and BMI, which was unexpected since bacteria can adapt to the fatty acid environment in the breast tissue. Hematoxylin eosin analysis may be a more appropriate approach to elucidating the association of the adipose tissue area in the analyzed biopsy with the microbial level. Finally, microbiota-derived metabolites are crucial mediators of host-microbial interactions [68], therefore metabolomic analysis would be critical for the defining the consequences of the dysbiosis.