www.nature.com/scientificreports SC丨ENTF| C REPORTS OPEN Characterization of the microbiome of nipple aspirate fluid of breast cancer survivors Received: 23 March 2016: Alfred A Chan'', Mina Bashir2, 3,. Magali N Rivas,', Karen Duvall+. Peter A Sieling, Accepted: 31 May 2016: Thomas R. Pieber, Parag A Vaishampayan, Susan M. Love&DelphineJ Lee Published: 21 June 2016 The microbiome impacts human health and disease Until recently, human breast tissue and milk were presumed to be sterile. Here, we investigated the presence of microbes in the nipple aspirate fluid(NAF) with a history of breast cancer(BC) and healthy control women(HO) using 165 rRNA gene amplicon nd their potential association with breast cancer. We compared the NAF microbiome between wom sequencing. The NAF microbiome from BC and HC showed significant differences in community composition. Two Operational Taxonomic Units(OTUs)showed differences in relative abundances between NAF collected from BC and HC. In NAF collected from BC, there was relatively higher incidence of the genus Alistipes. By contrast, an unclassified genus from the Sphingomonadaceae family was relatively more abundant in NAF from HC. These findings reflect the ductal source DNa since there were no differences between areolar skin samples collected from BC and HC. Furthermore the microbes associated with BC share an enzymatic activity, Beta-Glucuronidase, which may promote breast cancer This is the first report of bacterial DNA in human breast ductal fluid and the differences between NAF from HC and BC. Further investigation of the ductal microbiome and its potential role in breast cancer The human microbiome is the term applied to the universe of microbes that inhabit our skin and mucosal surfaces. Epidemiologic studies suggest that the human microflora contributes to 16% or more of worldwide malignancies-3. Increased cancer risk is associated with the presence of chronic, persistent, and dysregulated inflammation.. Many of the studies on microbes in relation to cancer have focused on the gut microflora. For instance, infection with Helicobacter pylori, a gram-negative bacterium that selectively colonizes the gastric pithelium and induces gastric inflammation, is correlated with a higher incidence of gastric cancer. Further, merging evidence suggests that infection with Fusobacterium nucleatum, a common member of the oropharyn geal flora and a pathogenic agent involved in gingival and periodontal disease, is associated with the development of human colorectal cancer.. Bacteria present near the tumor site are part of the tumor microenvironment On one hand, the microbiome might promote malignancy by inducing chronic inflammation, by altering the balance of host cell proliferation and death, and by triggering uncontrolled innate and adaptive immune responses. On the other hand, certain microbes might play a preventative role in breast carcinogenesis by affecting levels of strogen or by promoting antitumor immunity and immune surveillance The National Institutes of Health(NIH) Human Microbiome Project(HMP)has a reference collection of bac- terial genomes associated with multiple body sites from healthy human adults. Although an important number of anatomical sites were sampled for their microbiome, the human breast was omitted, presumably because most would claim that, at steady state, the breast tissue is sterile and devoid of any bacterial presence. However, the presence of six to eight ductal openings at the surface of the human nipple allows microbes from the environment, kin, and mouth( through sexual activity and breast feeding) to access the breast ductal system. In fact, microbes Dirks/ Dougherty Laboratory for Cancer Research, Department of Translational Immunology, John Wayne Cancer Institute at Providence Saint Johns Health Center, Santa Monica, CA, USA. Biotechnology and Planetary Protection Group, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA. Division of Endocrinology and Metabolism, Medical University of Graz, Graz, Austria. "Breast Center at the University of California Los Angeles (UCLA), Westwood, Los Angeles, CA, USA. Dr Susan Love Research Foundation, Encino, CA, USA. These authors contributed equally to this work. Correspondence and requests for materials should be addressed to D.J. L(email g) SCIENTIFIC REPORTS 6: 28061 DO1: 10.1038/srep28061
Scientific Reports | 6:28061 | DOI: 10.1038/srep28061 1 www.nature.com/scientificreports Characterization of the microbiome of nipple aspirate fluid of breast cancer survivors AlfredA. Chan1,* , Mina Bashir2,3,* , Magali N. Rivas1,* , Karen Duvall4,5, PeterA. Sieling1, Thomas R. Pieber3, ParagA.Vaishampayan2, Susan M. Love5 & Delphine J. Lee1 The microbiome impacts human health and disease. Until recently, human breast tissue and milk were presumed to be sterile. Here, we investigated the presence of microbes in the nipple aspirate fluid (NAF) and their potential association with breast cancer. We compared the NAF microbiome between women with a history of breast cancer (BC) and healthy control women (HC) using 16S rRNA gene amplicon sequencing. The NAF microbiome from BC and HC showed significant differences in community composition. Two Operational Taxonomic Units (OTUs) showed differences in relative abundances between NAF collected from BC and HC. In NAF collected from BC, there was relatively higher incidence of the genus Alistipes. By contrast, an unclassified genus from the Sphingomonadaceae family was relatively more abundant in NAF from HC. These findings reflect the ductal source DNA since there were no differences between areolar skin samples collected from BC and HC. Furthermore, the microbes associated with BC share an enzymatic activity, Beta-Glucuronidase, which may promote breast cancer. This is the first report of bacterial DNA in human breast ductal fluid and the differences between NAF from HC and BC. Further investigation of the ductal microbiome and its potential role in breast cancer are warranted. The human microbiome is the term applied to the universe of microbes that inhabit our skin and mucosal surfaces. Epidemiologic studies suggest that the human microflora contributes to 16% or more of worldwide malignancies1–3. Increased cancer risk is associated with the presence of chronic, persistent, and dysregulated inflammation1,3,4. Many of the studies on microbes in relation to cancer have focused on the gut microflora. For instance, infection with Helicobacter pylori, a gram-negative bacterium that selectively colonizes the gastric epithelium and induces gastric inflammation, is correlated with a higher incidence of gastric cancer5 . Further, emerging evidence suggests that infection with Fusobacterium nucleatum, a common member of the oropharyngeal flora and a pathogenic agent involved in gingival and periodontal disease, is associated with the development of human colorectal cancer6,7. Bacteria present near the tumor site are part of the tumor microenvironment. On one hand, the microbiome might promote malignancy by inducing chronic inflammation, by altering the balance of host cell proliferation and death, and by triggering uncontrolled innate and adaptive immune responses8 . On the other hand, certain microbes might play a preventative role in breast carcinogenesis by affecting levels of estrogen or by promoting antitumor immunity and immune surveillance9 . The National Institutes of Health (NIH) Human Microbiome Project (HMP) has a reference collection of bacterial genomes associated with multiple body sites from healthy human adults. Although an important number of anatomical sites were sampled for their microbiome, the human breast was omitted10, presumably because most would claim that, at steady state, the breast tissue is sterile and devoid of any bacterial presence. However, the presence of six to eight ductal openings at the surface of the human nipple allows microbes from the environment, skin, and mouth (through sexual activity and breast feeding) to access the breast ductal system11. In fact, microbes 1 Dirks/Dougherty Laboratory for Cancer Research, Department of Translational Immunology, John Wayne Cancer Institute at Providence Saint John’s Health Center, Santa Monica, CA, USA. 2 Biotechnology and Planetary Protection Group, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA. 3 Division of Endocrinology and Metabolism, Medical University of Graz, Graz, Austria. 4 Breast Center at the University of California Los Angeles (UCLA), Westwood, Los Angeles, CA, USA. 5 Dr. Susan Love Research Foundation, Encino, CA, USA. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to D.J.L. (email: LeeDJ@JWCI.org) received: 23 March 2016 accepted: 31 May 2016 Published: 21 June 2016 OPEN
ww.nature.com/scientificreports/ A Nipple Skin from HC(n=B) b 亡 Nipple Skin from HO(m=8) Phylum Distribution Adonis test, p-value=0. 945 nonparametric t-test, p-value=0. 929 ge 凸春 ■ Pc1(225%) Sequencing Depth mposition.(a)PCoA plot using Bray-Curtis dissimilarity based on genus- level OTUs from nipple skin swab samples collected from either HC or BC. (b)Number of bacterial OTUs observed on the nipple skin swabs as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs depicts the average phylum-level percentages of the nipple skin microbiome measured from HC andBC randomly sampled ten times at 918 sequencing depth. Error bars represent standard deviation. (c) Bar ch present in human breast milk and breast tissue have recently been characterized using next-generation sequenc ing and pan-pathogen array technologies Until now, the potential role of the local breast ductal microbiome with breast cancer has not been explored. this study, we used 16S rRNA gene sequencing to characterize the microorganisms present in nipple aspirate fluid (NAF). NAF is constantly secreted and absorbed by the epithelial cells lining the breast ducts and can be obtained non-invasively from at least one duct in a majority of women by applying negative pressu syringe attached to a suction cup. NAF collected from breast cancer patients has been shown to have a cantly different proteomic profile compared to NAF collected from healthy volunteers. Here, we collected NAF from healthy control ith a history of breast cancer(BC)(all were ductal carcinoma to investigate the breast ductal microbiome. Results Nipple skin microbiome from HC vS. BC. Nipple/areola skin was sampled with a sterile cotton swab as a control to compare to NAF. DNA extracted from the nipple skin samples was sequenced, the reads were clustered into unique Operation Taxonomic Units(OTUs), and then the OTUs classified to the genus taxonomic level. The relative differences amongst the nipple skin communities were calculated using the Bray-Curtis index and graphically visualized using Principal Coordinates Analysis(PCoA), whereby a shorter distance between points indicates increasing similarity in microbial composition. Though the skin varies according to body site and is expected to randomly vary across individuals, we hypothesized that the nipple skin microbiome would be inde pendent of breast cancer history. The nipple skin microbiome from HC(n=8)and BC(n=5)did not cluster parately and did not have a significant difference in bacterial composition(Adonis, p-value =0.945)(Fig. la) The rarefaction curves, which plot the number of unique species as a function of the number of reads sampled, reached a plateau for the nipple skin samples as well as for the other sample types(post-Betadine skin and NAF), indicating that our sampling depth provided sufficient coverage to capture most members of the bacterial com- munities(Figs 1b, 2b and 3b). The bacterial diversity was not significantly different between the nipple skin sam pled from HC vS BC (nonparametric t-test, p-value=0.929, Fig. 1b) In nipple skin samples from both HC and BC, bacterial composition at the phyla level was predominantly Proteobacteria(average 36.5%6), Firmicutes(average 33. 8%), and Bacteroidetes(average 19.5%)(Fig. 1c). Of all he OTUs in the nipple skin samples, Alistipes(recently reclassified Bacteroides putredinis)was the most abun dant(average 11.8%), followed by an unclassified genus from the Sphingomonadaceae family(average 11.3%), Rhizobium(average 6.7%), and an unclassified family from Acidobacteria Gp4(average 4.9%) None of the OTUs rom the skin swabs were significantly different when comparing their relative abundance between the HC and C groups(Kruskal-Wallis test, Supplemental Table S1). In summary, the nipple skin microbiome from HC and BC were not significantly distinguishable by their community composition, their diversity, or their individual OTUs, indicating that the nipple skin microbiome is independent of breast cancer history (Fig. 1) Microbiome in post-Betadine treated skin swab from HC vS BC. Betadine is an iodine-based broad trum disinfection agent used to clean skin before surgeries and other procedures. The nipple skin was treate ith Betadine to avoid contamination of NAF samples with normal nipple skin flora. We found that Betadine leaves a small residual flora that is detected by the extremely sensitive methodology used in this study. As a base line control, we collected post-Betadine samples of the nipple skin to characterize the residual microflora Reads from the post-Betadine skin swabs were clustered into OTUs and classified to the genera level. We calculated the Bray-Curtis dissimilarity index and performed PCoA to visualize community-wide differences amongst the ost-Betadine skin swabs. The post-Betadine skin microbiome from HC (n=5)and BC (n=7)did not separat SCIENTIFIC REPORTS 6: 28061 DO1: 10.1038/srep28061
www.nature.com/scientificreports/ Scientific Reports | 6:28061 | DOI: 10.1038/srep28061 2 present in human breast milk and breast tissue have recently been characterized using next-generation sequencing and pan-pathogen array technologies9,12–15. Until now, the potential role of the local breast ductal microbiome with breast cancer has not been explored. In this study, we used 16S rRNA gene sequencing to characterize the microorganisms present in nipple aspirate fluid (NAF). NAF is constantly secreted and absorbed by the epithelial cells lining the breast ducts and can be obtained non-invasively from at least one duct in a majority of women by applying negative pressure with a syringe attached to a suction cup16. NAF collected from breast cancer patients has been shown to have a significantly different proteomic profile compared to NAF collected from healthy volunteers17. Here, we collected NAF from healthy control women (HC) and women with a history of breast cancer (BC) (all were ductal carcinomas) to investigate the breast ductal microbiome. Results Nipple skin microbiome from HC vs. BC. Nipple/areola skin was sampled with a sterile cotton swab as a control to compare to NAF. DNA extracted from the nipple skin samples was sequenced, the reads were clustered into unique Operation Taxonomic Units (OTUs), and then the OTUs classified to the genus taxonomic level. The relative differences amongst the nipple skin communities were calculated using the Bray-Curtis index and graphically visualized using Principal Coordinates Analysis (PCoA), whereby a shorter distance between points indicates increasing similarity in microbial composition. Though the skin varies according to body site18 and is expected to randomly vary across individuals, we hypothesized that the nipple skin microbiome would be independent of breast cancer history. The nipple skin microbiome from HC (n= 8) and BC (n= 5) did not cluster separately and did not have a significant difference in bacterial composition (Adonis, p-value= 0.945) (Fig. 1a). The rarefaction curves, which plot the number of unique species as a function of the number of reads sampled, reached a plateau for the nipple skin samples as well as for the other sample types (post-Betadine skin and NAF), indicating that our sampling depth provided sufficient coverage to capture most members of the bacterial communities (Figs 1b, 2b and 3b). The bacterial diversity was not significantly different between the nipple skin sampled from HC vs. BC (nonparametric t-test, p-value=0.929, Fig. 1b). In nipple skin samples from both HC and BC, bacterial composition at the phyla level was predominantly Proteobacteria (average 36.5%), Firmicutes (average 33.8%), and Bacteroidetes (average 19.5%) (Fig. 1c). Of all the OTUs in the nipple skin samples, Alistipes (recently reclassified Bacteroides putredinis19) was the most abundant (average 11.8%), followed by an unclassified genus from the Sphingomonadaceae family (average 11.3%), Rhizobium (average 6.7%), and an unclassified family from Acidobacteria Gp4 (average 4.9%). None of the OTUs from the skin swabs were significantly different when comparing their relative abundance between the HC and BC groups (Kruskal-Wallis test, Supplemental Table S1). In summary, the nipple skin microbiome from HC and BC were not significantly distinguishable by their community composition, their diversity, or their individual OTUs, indicating that the nipple skin microbiome is independent of breast cancer history (Fig. 1). Microbiome in post-Betadine treated skin swab from HC vs. BC. Betadine is an iodine-based broad spectrum disinfection agent used to clean skin before surgeries and other procedures. The nipple skin was treated with Betadine to avoid contamination of NAF samples with normal nipple skin flora. We found that Betadine leaves a small residual flora that is detected by the extremely sensitive methodology used in this study. As a baseline control, we collected post-Betadine samples of the nipple skin to characterize the residual microflora. Reads from the post-Betadine skin swabs were clustered into OTUs and classified to the genera level. We calculated the Bray-Curtis dissimilarity index and performed PCoA to visualize community-wide differences amongst the post-Betadine skin swabs. The post-Betadine skin microbiome from HC (n=5) and BC (n=7) did not separate Figure 1. Nipple skin microbial composition. (a) PCoA plot using Bray-Curtis dissimilarity based on genuslevel OTUs from nipple skin swab samples collected from either HC or BC. (b) Number of bacterial OTUs observed on the nipple skin swabs as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs randomly sampled ten times at 918 sequencing depth. Error bars represent standard deviation. (c) Bar chart depicts the average phylum-level percentages of the nipple skin microbiome measured from HC and BC
ww.nature.com/scientificreports/ 合F Post-Betadine skin from HC (n=5) b Post-Betadine skin from HC (n=5) Phylum Distribution nonparametric t-test, p-value=0. 15 seeN吧宁 口 □ Addobaster Pc1(21.7%) Sequencing Depth skin from Hc skin from bc Figure 2. Post-Betadine skin microbial composition (a) PCoa plot using Bray-Curtis dissimilarity based on genus-level OTUs from post-Betadine skin samples from either HC (n=5)or BC(n=7).(b)Number of bacterial OTUs observed on the post-Betadine skin as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs randomly sampled ten times at 1164 sequencing depth. Error bars represent standard deviation. (c). Bar chart depicts the average phylum-level percentages of the post- Betadine skin swab microbiome asured from HC and bc into distinct clusters(Adonis, p-value =0.478)(Fig 2a), and the bacterial diversity between the two groups wer not significantly different(nonparametric t-test, p-value=0. 151)(Fig 2b). The post-Betadine swab microbiom is primarily composed of bacteria belonging to the phylum Proteobacteria(average 49.4%), Firmicutes(averag 23.8%6, and Bacteroidetes(average 13.0%)(Fig. 2c). None of the OTUs was significantly different between the ost-Betadine skin samples collected from HC and BC( Kruskal-Wallis test, Supplemental Table S2). As expected, he post-Betadine nipple skin microbiome from HC vS BC was not significantly different by comparing commu nity composition, by comparing species diversity, or by comparing relative OTU abundances, indicating that the NAF microbial composition from HC vS. o assess whether there is a difference in microbial compo- sition between the NAF from HC and BC, we calculated sample-to-sample variations by Bray-Curtis dissimilarity and then visualized the compositional differences by PCoA(Fig 3a). We found that the NAF microbiome from HC and bc clustered separately with a significant difference(Adonis, p-value =0.002). The Adonis test indi- cates that having had a history of breast cancer significantly affects the NAF microbial composition and explains approximately 13. 5% of the variation among the samples. The bacterial diversity of naf was not significantly different between those collected from HC vs those from BC(nonparametric t-test, p-value=0.65)(Fig. 3b) The most abundant bacteria in NAF samples were those belonging to the phylum Firmicutes(averaging 329% Bacteroidetes differences in relative abundance between NAF from HC and from BC by Kruskal-Wallis test( Supplemental Table $3). The genus Alistipes(Otu00009)was only present in the NAF from BC(p-value=0.0068)(Fig. 3d In contrast, an unclassified genus from the family Sphingomonadaceae(Otu00007)was present in NAF from both HC and BC, but was relatively more abundant in the NAF from HC compared to BC (p-value=0.02846) demonstrated significant clustering betwe nipple skin and the post-Betadine controls, the microbiome from NAF Fig 3e). Unlike the microbiome from health states Microbial composition from nipple skin vs NAF. To address whether the microbes in NAF correspond to those on the overlying nipple and areolar skin, we compared the microbiome between the nipple skin and NAF Because of the variability of microbiome across individuals, the analysis was limited to paired NAF and corresponding nipple skin samples collected from the same breast. We calculated the sample-to-sample variation by the Bray-Curtis dissimilarity and visualized it using PCoA(Fig. 4a). There was no compositional difference between the microbes present on nipple skin and in NAF when combining all HC and BC samples(Adonis, p-value=0.2734)(Fig 4a). However, a history of breast cancer is a confounding variable, so the samples from HC group and BC group were analyzed separately and the Bray-Curtis dissimilarity recalculated for each group ( Fig. 4b, c) Within the HC sample pairs(n=6), the nipple skin and NAF microbiome were significantly different by paired Adonis test using the strata parameter(p-value=0.0313). However, there was no differe the nipple skin diversity (paired t-test, p-value=0.62), and none of the OTUs were significantly different be d NAF groups( Paired Wilcoxon signed-rank test, Supplemental Table S4). Within the BC samples(n=3), the paired comparison between the nipple skin and NAF microbiome was not significantly different(Adonis, p-value=1.000). Comparing the nipple skin and NAF sample types from BC, SCIENTIFIC REPORTS 6: 28061 DO1: 10.1038/srep28061
www.nature.com/scientificreports/ Scientific Reports | 6:28061 | DOI: 10.1038/srep28061 3 into distinct clusters (Adonis, p-value=0.478) (Fig. 2a), and the bacterial diversity between the two groups were not significantly different (nonparametric t-test, p-value=0.151) (Fig. 2b). The post-Betadine swab microbiome is primarily composed of bacteria belonging to the phylum Proteobacteria (average 49.4%), Firmicutes (average 23.8%), and Bacteroidetes (average 13.0%) (Fig. 2c). None of the OTUs was significantly different between the post-Betadine skin samples collected from HC and BC (Kruskal-Wallis test, Supplemental Table S2). As expected, the post-Betadine nipple skin microbiome from HC vs. BC was not significantly different by comparing community composition, by comparing species diversity, or by comparing relative OTU abundances, indicating that the post-Betadine controls are independent of breast cancer history (Fig. 2). NAF microbial composition from HC vs. BC. To assess whether there is a difference in microbial composition between the NAF from HC and BC, we calculated sample-to-sample variations by Bray-Curtis dissimilarity and then visualized the compositional differences by PCoA (Fig. 3a). We found that the NAF microbiome from HC and BC clustered separately with a significant difference (Adonis, p-value= 0.002). The Adonis test indicates that having had a history of breast cancer significantly affects the NAF microbial composition and explains approximately 13.5% of the variation among the samples. The bacterial diversity of NAF was not significantly different between those collected from HC vs. those from BC (nonparametric t-test, p-value=0.65) (Fig. 3b). The most abundant bacteria in NAF samples were those belonging to the phylum Firmicutes (averaging 42.1%), Proteobacteria (averaging 32.9%), and Bacteroidetes (averaging 14.5%) (Fig. 3c). Two OTUs showed differences in relative abundance between NAF from HC and from BC by Kruskal-Wallis test (Supplemental Table S3). The genus Alistipes (Otu00009) was only present in the NAF from BC (p-value= 0.0068) (Fig. 3d). In contrast, an unclassified genus from the family Sphingomonadaceae (Otu00007) was present in NAF from both HC and BC, but was relatively more abundant in the NAF from HC compared to BC (p-value= 0.02846) (Fig. 3e). Unlike the microbiome from the nipple skin and the post-Betadine controls, the microbiome from NAF demonstrated significant clustering between HC and BC with two differentially abundant OTUs between the two health states. Microbial composition from nipple skin vs. NAF. To address whether the microbes in NAF correspond to those on the overlying nipple and areolar skin, we compared the microbiome between the nipple skin and NAF. Because of the variability of microbiome across individuals10, the analysis was limited to paired NAF and corresponding nipple skin samples collected from the same breast. We calculated the sample-to-sample variation by the Bray-Curtis dissimilarity and visualized it using PCoA (Fig. 4a). There was no compositional difference between the microbes present on nipple skin and in NAF when combining all HC and BC samples (Adonis, p-value= 0.2734) (Fig. 4a). However, a history of breast cancer is a confounding variable, so the samples from HC group and BC group were analyzed separately and the Bray-Curtis dissimilarity recalculated for each group (Fig. 4b,c). Within the HC sample pairs (n= 6), the nipple skin and NAF microbiome were significantly different by paired Adonis test using the strata parameter (p-value= 0.0313). However, there was no difference in bacterial diversity (paired t-test, p-value= 0.62), and none of the OTUs were significantly different between the nipple skin and NAF groups (Paired Wilcoxon signed-rank test, Supplemental Table S4). Within the BC samples (n= 3), the paired comparison between the nipple skin and NAF microbiome was not significantly different (Adonis, p-value=1.000). Comparing the nipple skin and NAF sample types from BC, Figure 2. Post-Betadine skin microbial composition. (a) PCoA plot using Bray-Curtis dissimilarity based on genus-level OTUs from post-Betadine skin samples from either HC (n=5) or BC (n=7). (b) Number of bacterial OTUs observed on the post-Betadine skin as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs randomly sampled ten times at 1164 sequencing depth. Error bars represent standard deviation. (c). Bar chart depicts the average phylum-level percentages of the post-Betadine skin swab microbiome measured from HC and BC
ww.nature.com/scientificreports/ b Adonis test, p-value=. D02 nonparametric t-test, p-value=D,65 NAF from Hc NAF from BC PC1(18% Sequencing Depth F. Rikenellaceae: G. Alistipes (otu00009) e, F. sphingomonadaceae: G unclassified (otu00007) NAF from Hc NAF from B NAF from Hc NAF from BC Figure 3. NAF microbial composition (a) PCoA plot of Bray-Curtis dissimilarity based on genus-level OTUs of NAF collected from either HC (n=9)or BC(n=6).(b)Number of bacterial OTUs observed in the NAF as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs randomly sampled ten times at 934 sequencing depth Error bars represent standard deviation. (c) Bar chart depicts the average phylum-level percentages of the NAF microbiome measured from HC and BC. (d) Strip-plot comparing relative abundances of the genus Alistipes(Otu00009)between HC and BC group. (e)Strip-plot comparing relative abundances of the unclassified genus from the Sphingomonadaceae family(otu00007)between HC and BC group there was no difference in bacterial diversity(paired t-test, p-value =0.52), and none of the OTUs were signifi antly different(Paired wilcoxon signed-rank test, Supplemental Table S5) Functional prediction based on NAF microbiome comparing HC vS BC. To investigate whether the microbes present in the breast ducts have a metabolic activity related to malignancy or health, we examined the gene sequences associated with 18 selected KEGG metabolic pathways that have been implicated in color cancer20-27 using the PICRUSt package 2. None of the 18 selected KEGG pathways were statistically significant at 5% alpha level after correcting for multiple hypothesis testing. The pathway"flavone and flavonol biosynthesis (ko00944)produced the lowest p-value(Kruskal-Wallis test, adjusted p-value=0.066)(Fig 5a) Of the 12 KEGG orthologs that comprise the flavone and flavonol biosynthesis pathway, Beta-Glucuronidase(k01195)"was th only KEGG bacterial ortholog predicted by the NAF microbiome. Furthermore, the gene "Beta-Glucuronidase" is inferred from a combination of eight OTUs(Table 1)and the combined relative abundances of these eight OTUs were statistically higher in NAF from BC than HC(Kruskal-Wallis test, p-value =0.0036)(Fig 5b) Di IscUSSIon This study establishes the existence of a microbiome in the breast ductal system, by detecting the presence of bacterial DNA in NAF. Analysis of the beta-diversity demonstrates that the NAF microbial cor position is different in women having had a history of breast cancer. Previous microbiome studies of breast tis- sue have described bacteria belonging mainly to the phylum Proteobacteria, Firmicutes, Actinobacteria, and Bacteroidetes. 25. Our studies on NAF show a similar microbial composition, with Firmicutes and Bacteroidetes. While previous studies on NAF have attempted to define biomarkers to help etection and prevention of breast cancer using cellular morphology, protein expression, growth factors, and hormones in the intraductal microenvironment, this is the first study to identify bacteria in the ductal syster SCIENTIFIC REPORTS 6: 28061 DO1: 10.1038/srep28061
www.nature.com/scientificreports/ Scientific Reports | 6:28061 | DOI: 10.1038/srep28061 4 there was no difference in bacterial diversity (paired t-test, p-value= 0.52), and none of the OTUs were significantly different (Paired Wilcoxon signed-rank test, Supplemental Table S5). Functional prediction based on NAF microbiome comparing HC vs. BC. To investigate whether the microbes present in the breast ducts have a metabolic activity related to malignancy or health, we examined the gene sequences associated with 18 selected KEGG metabolic pathways that have been implicated in colon cancer20–27 using the PICRUSt package28. None of the 18 selected KEGG pathways were statistically significant at 5% alpha level after correcting for multiple hypothesis testing. The pathway “flavone and flavonol biosynthesis (ko00944)” produced the lowest p-value (Kruskal-Wallis test, adjusted p-value= 0.066) (Fig. 5a). Of the 12 KEGG orthologs that comprise the “flavone and flavonol biosynthesis” pathway, “Beta-Glucuronidase (k01195)” was the only KEGG bacterial ortholog predicted by the NAF microbiome. Furthermore, the gene “Beta-Glucuronidase” is inferred from a combination of eight OTUs (Table 1) and the combined relative abundances of these eight OTUs were statistically higher in NAF from BC than HC (Kruskal-Wallis test, p-value=0.0036) (Fig. 5b). Discussion This study establishes the existence of a microbiome in the breast ductal system, by detecting the presence of bacterial DNA in NAF. Analysis of the beta-diversity demonstrates that the NAF microbial community composition is different in women having had a history of breast cancer. Previous microbiome studies of breast tissue have described bacteria belonging mainly to the phylum Proteobacteria, Firmicutes, Actinobacteria, and Bacteroidetes9,12,15. Our studies on NAF show a similar microbial composition, with Firmicutes, Proteobacteria, and Bacteroidetes. While previous studies on NAF have attempted to define biomarkers to help early detection and prevention of breast cancer using cellular morphology, protein expression, growth factors, and hormones in the intraductal microenvironment29, this is the first study to identify bacteria in the ductal system. Figure 3. NAF microbial composition. (a) PCoA plot of Bray-Curtis dissimilarity based on genus-level OTUs of NAF collected from either HC (n=9) or BC (n=6). (b) Number of bacterial OTUs observed in the NAF as a function of sequencing depth as assessed by diversity rarefaction curves. The difference in diversity was compared by a non-parametric t-test using the average of observed OTUs randomly sampled ten times at 934 sequencing depth. Error bars represent standard deviation. (c) Bar chart depicts the average phylum-level percentages of the NAF microbiome measured from HC and BC. (d) Strip-plot comparing relative abundances of the genus Alistipes (Otu00009) between HC and BC group. (e) Strip-plot comparing relative abundances of the unclassified genus from the Sphingomonadaceae family (Otu00007) between HC and BC group
ww.nature.com/scientificreports/ PCoA: Paired NAF V5. Skin from BC PCoA: Paired NAF vs Skin from HC PCoA: Paired NAF vS Skin from Bc Adonis, p-value=0. 2734 Adonis, p-value=0.0313 Adonis, p-value= 1.0000 AF from Bc 爸am■4△ 口 NAF from Hc △●△●△●△●△ ●△●△●△ △ sain 5 Left 0.6△skmn5Rk Pc1(20.17%) Pc121.74%) Pc1(33.97%) gure唾 le skin and Naf microbial arison.(a)PCoA plot of Bray-Curtis dissin between paired skin and NAF samples collected from both HC and BC(n=9). (b)PCoa plot of Bray-Curtis dissimilarity between paired skin and NAF samples collected from HC(n=6).(c)PCoA plot of Bray-Curtis dissimilarity between paired skin and NaF samples collected from BC(n=3). All subjects with sequences obtained from both skin and NAF are shown. The paired community wide testing was performed ing Adonis test with the strata parameter to allow permutation only within patients and not across While microbiome studies of Naf have never been reported, breast milk has been studied for its commu- nity composition 430. Both NAF and milk are secreted by the epithelial cells lining the breast ducts. Breast milk contains bacteria that are orally transmitted to the breastfed infant with a key function to populate and establish the neonate's gastrointestinal flora. The bacteria present in milk belong mostly to the phylum Firmicutes( Staphylococcus, Clostridium, Lactobacillus), Actinobacteria(Propionibacterium, Corynebacterium Proteobacteria( Ralstonia, Sphingomonas, Pseudomonas, Bradyrhizobiaccea), and Bacteroidetes(Prevotella)3. 14.30. Similarly, in NAF samples from both HC and BC, we found Clostridium and Prevotella, as well as an unclassified genus from the Sphingomonadaceae family. Thus, the NAF and milk microbiomes exhibit some overlap Though the existence of bacteria in breast tissue has previously been described, the current study establishes the presence of bacterial DNA in the breast ductal system and thereby raises the question where do breast ductal bac- teria originate?", Ductal spaces may be populated by bacteria from oral sources, resulting from sexual activity or breast-feeding. An entero-mammary pathway by which some bacteria from the gut could reach the mammary gland via an endogenous route has been postulated by others. Alternatively, it is tempting to speculate that skin bacteria from the nipple surface might reach the breast ducts through the ductal orifices via retrograde migra tion since the nipple skin and NAF microbiome had no significant OTUs by paired wilcoxon signed-rank test However, amongst the samples from HC, the nipple skin and the naf had significantly different community omposition. Additional studies with larger sample sizes are required to investigate this preliminary trend. Two individual OTUs showed differences in relative abundances between NaF from HC and BC. First the enus Alistipes(Otu00009)was present only in the NAF from BC, and was absent in the NAF collected from HC. listipes has been included among other bacteria to have an increased relative abundance in colorectal cancer2. Another orgaundant in NAF from HC than from BC. The consensus sequence from this OTU was queried against sm of interest, an unclassifed genus from the Sphingomonadaceae family(otu00007), was rela he NCBi refseq_rna using the default blaSt algorithm(Basic Local Alignment Search Tool) 3. The OtU matched two strains of Sphingobium yanoikuyae(NR_1137301 and NR_115241)with 100%identity. Notably, in our previous study, four out of the eight OTUs enriched in paired normal compared to estrogen-receptor positive (ER+) breast tumor tissue belonged to the Sphingomonadaceae family with Sphingobium yanoikuyae having the greatest statistical significance Members of the Sphingomonadaceae family are known for their ability to degrade aromatic hydrocarbons Polycyclic aromatic hydrocarbons have been associated with breast cancer. It is interesting that a microbe capa ble of degrading aromatic hydrocarbons, such as estrogen, would be comparatively higher in the healthy state and lower in the estrogen-dependent breast tumor 6. This may be especially relevant because the sex hormones have far higher concentration in the nipple aspirate fluid than in the serum. In the present study, the metabolism of aromatic hydrocarbons by Sphingomonadaceae may have a protective role and account for the greater abundance of these organisms in the NAF from healthy women, a possibility that warrants further investigation. If bacterial metabolism were related to breast cancer, then certain metabolic pathways would be discern le between the naf collected from HC and bC. PicruSt analysis infers the microbial gene contents and their functional profiles using a database of annotated sequences. None of the 18 selected KEGG pathwa were statistically significant at 5% alpha level after correcting for multiple hypothesis testing. The pati flavone and flavonol biosynthesis (ko00944)"produced the lowest p-value(Kruskal-Wallis test, adjusted p-value=0.066)(Fig 5a) Of the 12 KEGG orthologs that comprise the "flavone and flavonol biosynthesis path way,"Beta-Glucuronidase(k01195)"was the only KEGG bacterial ortholog predicted by the NAF microbiome The predicted Beta-Glucuronidase levels were higher in NAF from BC than from HC Beta-Glucuronidase is known to be a procarcinogenic enzyme in gastrointestinal cancer, and has been used as a marker to evaluate dietary effects on colon carcinogenesis.3. In addition, bacterial Beta-Glucuronidase activi has also been reported to be higher in patients with colon cancer than healthy controls. Notably, the NAF OTUs SCIENTIFIC REPORTS 6: 28061 DO1: 10.1038/srep28061 5
www.nature.com/scientificreports/ Scientific Reports | 6:28061 | DOI: 10.1038/srep28061 5 While microbiome studies of NAF have never been reported, breast milk has been studied for its community composition13,14,30. Both NAF and milk are secreted by the epithelial cells lining the breast ducts. Breast milk contains bacteria that are orally transmitted to the breastfed infant with a key function to populate and establish the neonate’s gastrointestinal flora14. The bacteria present in milk belong mostly to the phylum Firmicutes (Staphylococcus, Clostridium, Lactobacillus), Actinobacteria (Propionibacterium, Corynebacterium), Proteobacteria (Ralstonia, Sphingomonas, Pseudomonas, Bradyrhizobiaccea), and Bacteroidetes (Prevotella)13,14,30. Similarly, in NAF samples from both HC and BC, we found Clostridium and Prevotella, as well as an unclassified genus from the Sphingomonadaceae family. Thus, the NAF and milk microbiomes exhibit some overlap. Though the existence of bacteria in breast tissue has previously been described, the current study establishes the presence of bacterial DNA in the breast ductal system and thereby raises the question ‘where do breast ductal bacteria originate?’9,12. Ductal spaces may be populated by bacteria from oral sources, resulting from sexual activity or breast-feeding. An entero-mammary pathway by which some bacteria from the gut could reach the mammary gland via an endogenous route has been postulated by others31. Alternatively, it is tempting to speculate that skin bacteria from the nipple surface might reach the breast ducts through the ductal orifices via retrograde migration since the nipple skin and NAF microbiome had no significant OTUs by paired Wilcoxon signed-rank test. However, amongst the samples from HC, the nipple skin and the NAF had significantly different community composition. Additional studies with larger sample sizes are required to investigate this preliminary trend. Two individual OTUs showed differences in relative abundances between NAF from HC and BC. First, the genus Alistipes (Otu00009) was present only in the NAF from BC, and was absent in the NAF collected from HC. Alistipes has been included among other bacteria to have an increased relative abundance in colorectal cancer32. Another organism of interest, an unclassified genus from the Sphingomonadaceae family (Otu00007), was relatively more abundant in NAF from HC than from BC. The consensus sequence from this OTU was queried against the NCBI “refseq_rna” using the default BLAST algorithm (Basic Local Alignment Search Tool)33. The OTU matched two strains of Sphingobium yanoikuyae (NR_113730.1 and NR_11524.1) with 100% identity. Notably, in our previous study, four out of the eight OTUs enriched in paired normal compared to estrogen-receptor positive (ER+) breast tumor tissue belonged to the Sphingomonadaceae family with Sphingobium yanoikuyae having the greatest statistical significance9 . Members of the Sphingomonadaceae family are known for their ability to degrade aromatic hydrocarbons34. Polycyclic aromatic hydrocarbons have been associated with breast cancer35. It is interesting that a microbe capable of degrading aromatic hydrocarbons, such as estrogen, would be comparatively higher in the healthy state and lower in the estrogen-dependent breast tumor36. This may be especially relevant because the sex hormones have far higher concentration in the nipple aspirate fluid than in the serum16. In the present study, the metabolism of aromatic hydrocarbons by Sphingomonadaceae may have a protective role and account for the greater abundance of these organisms in the NAF from healthy women, a possibility that warrants further investigation. If bacterial metabolism were related to breast cancer, then certain metabolic pathways would be discernable between the NAF collected from HC and BC. PICRUSt analysis infers the microbial gene contents and their functional profiles using a database of annotated sequences. None of the 18 selected KEGG pathways were statistically significant at 5% alpha level after correcting for multiple hypothesis testing. The pathway “flavone and flavonol biosynthesis (ko00944)” produced the lowest p-value (Kruskal-Wallis test, adjusted p-value=0.066) (Fig. 5a). Of the 12 KEGG orthologs that comprise the “flavone and flavonol biosynthesis” pathway, “Beta-Glucuronidase (k01195)” was the only KEGG bacterial ortholog predicted by the NAF microbiome. The predicted Beta-Glucuronidase levels were higher in NAF from BC than from HC. Beta-Glucuronidase is known to be a procarcinogenic enzyme in gastrointestinal cancer, and has been used as a marker to evaluate dietary effects on colon carcinogenesis37,38. In addition, bacterial Beta-Glucuronidase activity has also been reported to be higher in patients with colon cancer than healthy controls39. Notably, the NAF OTUs Figure 4. Nipple skin and NAF microbial composition comparison. (a) PCoA plot of Bray-Curtis dissimilarity between paired skin and NAF samples collected from both HC and BC (n=9). (b) PCoA plot of Bray-Curtis dissimilarity between paired skin and NAF samples collected from HC (n=6). (c) PCoA plot of Bray-Curtis dissimilarity between paired skin and NAF samples collected from BC (n=3). All subjects with sequences obtained from both skin and NAF are shown. The paired community wide testing was performed using Adonis test with the strata parameter to allow permutation only within patients and not across