Analysis of the human breast milk microbiome and bacterial extracellular vesicles in healthy mothers

The infant gut harbors over 100 trillion microbes that interact with the host to aid immune development. Early-life gut microbiome acquisition influences health throughout life, and factors such as delivery mode, antibiotics, care environment, and nutrition shape early colonization. Human breast milk (HBM) provides nutrition and immune components (e.g., antibodies, immune cells, antimicrobial proteins, cytokines, and oligosaccharides) and is now recognized to contain commensal bacteria that can affect infant gut colonization. Extracellular vesicles (EVs) are nano-sized vesicles released by both Gram-positive and Gram-negative bacteria and carry proteins, nucleic acids, and lipids. Bacterial EVs are detectable in body fluids and can modulate host cells, but had not been studied in HBM. The objective of this study was to characterize the HBM microbiota of healthy lactating mothers in Korea and to quantify key bacteria likely affecting infant gut colonization by analyzing both the microbiota and bacterial EVs using culture-independent sequencing.

Earlier culture-dependent work established that HBM contains commensal bacteria, and subsequent next-generation sequencing (NGS) studies expanded knowledge of HBM microbial diversity. A core HBM microbiome has been proposed but remains controversial across studies. Systematic reviews indicate that most bacterial species in breast tissue and milk belong to Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes, with commonly reported genera including Staphylococcus, Streptococcus, Lactobacillus, Pseudomonas, Bifidobacterium, Corynebacterium, Enterococcus, Acinetobacter, and others. Research on bacteria-derived EVs in other human samples (blood, urine, stool) shows that EVs can disseminate systemically and influence immunity and disease, but EVs had not been analyzed in HBM prior to this study.

Study design and participants: IRB-approved study (Chung-Ang University Hospital, IRB No. 1810-004-309). Healthy lactating mothers of full-term infants were recruited between July 2017 and June 2018, within 1 week to 4 months postpartum. Exclusions: samples collected within 1 week postpartum (colostrum), maternal disease during lactation, or antibiotic use; samples not delivered to the lab within 24 h. Sample collection and handling: Breast milk from 22 mothers was collected in sterile bags via manual or pump expression after hand washing and nipple cleansing with sterile water; the first 10 drops were discarded. Samples were transported within 24 h and stored at −80°C. EV isolation and DNA extraction: EVs were separated by differential centrifugation at 10,000 g for 10 min at 4°C (pellet=bacteria; supernatant=EVs). Supernatants were filtered through 0.22-μm filters to remove bacteria and particles. Bacteria and EVs were boiled at 100°C for 40 min, then centrifuged at 13,000 rpm for 30 min at 4°C; supernatants were collected. DNA was extracted using the DNeasy PowerSoil Kit (QIAGEN) and quantified by QIAxpert. 16S rRNA gene amplicon sequencing: V3–V4 regions were amplified with primers 16S_V3_F and 16S_V4_R. Libraries were prepared per Illumina MiSeq guidelines, pooled at equimolar ratios, and sequenced on an Illumina MiSeq. Bioinformatics: Adapter trimming with cutadapt v1.1.6; paired-end merging with CASPER; quality filtering per Phred score criteria; merged read length filtered to 350–550 bp. Chimeras were removed using VSEARCH against the SILVA gold database. OTU clustering was performed de novo at 97% similarity using VSEARCH. Representative OTU sequences were taxonomically assigned with Greengenes via UCLUST (QIIME v1.9.1 default settings). Statistics: Alpha diversity (Chao1, ACE, Shannon, Simpson, Fisher’s index) and richness (rarefaction) were computed; group comparisons used t-tests. Beta diversity was assessed via PCA based on Euclidean distance. Correlation networks were constructed using Pearson correlation coefficients. Analyses were conducted in R v3.3.2.

- Participants: 22 lactating mothers (mean age 33.1 ± 3.4 years); 72.7% primiparous; 36.4% vaginal delivery, 63.6% cesarean section; samples collected on average 11.1 ± 2.4 days postpartum. - Sequencing depth and richness: Mean reads were higher in bacterial samples than EVs (14,966.6 ± 10,189.1 vs 9,267.8 ± 6,507.0). Mean OTUs were higher in bacteria (1,202.3 ± 789.1) vs EVs (501.0 ± 446.8). - Alpha diversity: Bacterial samples exhibited significantly higher richness and evenness than EVs by Chao1 (p < 0.001) and ACE (p = 0.005), and higher Fisher’s index (p < 0.001). No significant differences were observed in Shannon index (p = 0.51) or Simpson index (p > 0.05). - Beta diversity: PCA based on Euclidean distance showed significant differences between bacteria and EVs within the same individual sample. Across individuals, bacterial-level distances differed significantly, whereas EV-level distances did not. - Taxonomic composition (bacteria): 23 phyla detected. Firmicutes dominated (56.4% ± 19.4), followed by Proteobacteria (19.6% ± 19.1), Bacteroidetes (9.8% ± 10.5), and Actinobacteria (9.0% ± 7.9). At the genus level, Streptococcus (25.1% ± 20.9) and Staphylococcus (10.7% ± 12.3) predominated; numerous other genera averaged <5%. - Taxonomic composition (EVs): 22 phyla detected. Firmicutes (35.8% ± 12.9), Proteobacteria (24.5% ± 13.3), Bacteroidetes (15.5% ± 8.8), and Actinobacteria (13.1% ± 8.3) dominated. At the genus level, Bacteroides (9.1% ± 5.4), Acinetobacter (6.9% ± 8.2), and Lactobacillaceae(f) (5.5% ± 9.1) were most abundant; many others averaged <5%. - Correlations between bacteria and EVs (genus level): Significant positive correlations for Acinetobacter, Ruminococcaceae(f), Bifidobacterium, Comamonadaceae(f), Rothia, and Clostridiaceae(f) (p < 0.05). Notably, Comamonadaceae(f) r = 0.95; Comamonas and Diaphorobacter (each <1% abundance) r = 0.94. Several genera (e.g., Bacteroides, Prevotella, Lachnospiraceae(f), Collinsella, Burkholderia) showed nonsignificant negative trends. - Subgroup analyses: Compared with multiparous mothers, primiparous mothers had significantly lower Staphylococcus (bacterial samples; 0.31-fold) and lower Collinsella (EV samples; 0.18-fold), while multiparous mothers had higher Haemophilus in bacterial samples (2.86-fold) (p < 0.05). Delivery mode: Cesarean section samples showed higher Streptophyta(o) (6.24-fold) in bacteria, and lower Actinomycetales (0.19-fold) in EVs compared with vaginal delivery (p < 0.05). No significant differences with calorie intake.

This study addresses whether bacterial extracellular vesicles are present in human breast milk and whether they might contribute to infant gut colonization. The findings confirm that HBM contains diverse bacterial communities dominated by Firmicutes and that bacterial EVs are detectable and differ compositionally from the resident bacterial community. EV-associated taxa included gut-associated genera (e.g., Bacteroides, Lactobacillaceae), suggesting that EV-producing bacteria may be metabolically active and potentially involved in vertical transfer signals impacting infant gut colonization and immune development. Significant positive correlations between certain genera (notably Bifidobacterium, Comamonadaceae) in bacteria and EVs imply that specific taxa may actively release EVs and thus exert functional effects beyond their relative abundance in HBM. Subgroup differences by parity and delivery mode suggest perinatal factors can influence both the microbiota and EV profiles. Overall, the data support a model in which bacterial EVs in HBM represent functional outputs of key bacteria that could modulate infant gut colonization independent of the dominant milk bacterial populations.

HBM from healthy lactating mothers harbors diverse bacterial communities and measurable bacterial extracellular vesicles with distinct community profiles. Gut-associated genera comprise a higher proportion of the EV fraction than of the bulk bacterial fraction, supporting the presence of key metabolically active bacteria and suggesting a role for EVs in vertical transfer of microbiota-related signals from mother to infant. Future research should elucidate mechanisms by which HBM-derived bacterial EVs influence infant gut colonization and immunity and explore strategies to promote beneficial EV-mediated effects.

- Small sample size (n=22), limiting generalizability. - Lack of concurrent sampling from maternal/infant oral cavity, skin, and stool, preventing source attribution of HBM bacteria. - No direct demonstration of mechanisms by which HBM microbiota or EVs colonize the infant gut or affect health outcomes. - Cross-sectional design near early postpartum; limited assessment of temporal dynamics.