Effect of breast milk with or without bacteria on infant gut microbiota

The participants in this observational study were 20–35 years old women who achieved singleton pregnancy spontaneously. Furthermore, women with pregnancy complications, infectious diseases or chronic diseases, as well as women who had used antibiotics or probiotics during pregnancy were excluded from the study. According to the above criteria, 23 pregnant women from the First Affiliated Hospital of Jinan University (Guangzhou, China) were enrolled in our study. All of the pregnant women delivered vaginally, and their infants’ Apgar scores were all normal. Meanwhile, their infants were also included in this study. All pregnant women who participated in the study signed an informed consent form.

Fecal samples were taken from pregnant women after full term of pregnancy. On the 14th postnatal day, breast milk and infant feces were collected. Fecal samples were collected by professionals using sterile spoons and sterile boxes. Before collecting the milk, the mothers cleaned the nipple and surrounding skin with soap and water. Women wore sterile gloves to squeeze out the milk and filled in sterile tubes. All samples were transported on ice and then stored in the refrigerator at -80℃ until DNA extraction. We also collected clinical information about pregnant women and infants, such as maternal age, height, weight, gestational weeks, neonatal length, weight, and gender.

The CTAB/SDS technique was used to extract total bacterial DNA from samples. DNA concentration and purity were monitored on a 1% agarose gel. According to the concentration, DNA was diluted to 1 ng/μL with sterile water. The V3-V4 hypervariable region of 16S rRNA gene was amplified using a specific primer (341F-806R) with the barcode. All PCR reactions were performed with Phusion® High-Fidelity PCR Master Mix (New England Biolabs). The PCR products were mixed with the same volume of 1X loading buffer (containing SYB Green) and electrophoresed on a 2 percent agarose gel. PCR products were mixed in equal density ratio. The mixed PCR products were then purified using Qiagen Gel Extraction Kit (Qiagen, Germany). Sequencing libraries were generated by using TruSeq® DNA PCR-Free Sample Preparation Kit. The library was sequenced on a NovaSeq6000 platform after assessing on the Qubit@ 2.0 Fluorometer.

Each read was stripped of the barcode and the primer, and then they were spliced by using FLASH (V1.2.7, http://​ccb.​jhu.​edu/​software/​FLASH/​) [26]. Next, effective tags were generated by intercepting, filtering, and removing chimeric sequences from the raw tags. The Uparse algorithm (Uparse v7.0.1001,http://​www.​drive5.​com/​uparse/​) was used to cluster all effective tags into sequences, and then all sequences were clustered into operational taxonomic units (OTUs) with 97% identity. Finally, representative sequences from each OTU were aligned to the SSUrRNA database (http://​www.​arb-silva.​de/​).

Alpha diversity was used to assess diversity within samples, whereas Beta diversity was used to compare the composition of microbial communities between samples. The Simpson index and Shannon index were calculated in QIIME software (Version 1.7.0) to estimate Alpha diversity. A higher Shannon and Simpson index implies a community with greater diversity. Comparison of Alpha diversity indices between groups using the Wilcoxon test with GraphPad Prism software (Version 8.0.2). Beta diversity was demonstrated using Principal Coordinates Analysis (PCoA) based on weighted Unifrac distances. To estimate the correlation between genera, we calculated the Spearman coefficients for the top 100 genera by relative abundance. Genera with absolute Spearman correlation coefficients greater than 0.6 and P values less than 0.05 were considered for construction of the network. The visualization of each network and calculation of topological properties were performed using Gephi software (Version 0.9.2). The relative abundance histogram and the Venn diagram were plotted using R software. To further investigate biomarkers between different groups, we conducted linear discriminant analysis effect size (LEfSe). The clinical data were compared between groups using independent-samples t test or Chi-square test. A P value less than 0.05 was considered statistically significant.

Seventeen of 23 mother-infant pairs were eventually enrolled in our study. Five pairs were eliminated due to lack of breast milk or personal reasons, and one pair was excluded because the mother suffered from mastitis. Six of the 17 milk samples could not be sequenced due to low bacterial loads or the absence of bacteria. The detection rate of microbe in milk was 64.7%. Breast milk that failed to be sequenced in this study was tentatively considered sterile milk. Based on the sequencing results of breast milk, the subjects were divided into four groups: pregnant women with bacterial milk (PBM, n = 11), pregnant women with sterile milk (PSM, n = 6), infants fed with bacterial milk (IBM, n = 11), infants fed with sterile milk (ISM, n = 6). The rarefaction curves of samples in each group tended to be flat as the sequence number increased, indicating that the sequencing depth was sufficient (Fig. 1a).

There were no significant differences in age, height, weight, BMI, and gestational week between pregnant women, and there were no significant differences in birth weight, length, and gender between infants (Table 1).

The microbial composition of maternal breast milk was shown in Fig. 1b-c and Table 2. Among the microbial communities of breast milk, Firmicutes predominated (73.1%), followed by Proteobacteria (9.2%), Actinobacteria (6.3%), and Bacteroidetes (2.2%), while the relative abundance of the rest of the phylum was less than 1%. At the genus level, eight genera had mean relative abundances greater than 1%. They were Streptococcus (30.1%), Staphylococcus (20.2%), Bacillus (7.4%), Gemella (5.7%), Rothia (5.4%), Acinetobacter (4.3%), Lactobacillus (2.5%), and Clostridium_sensu_stricto_1 (1.1%). Among the above genera, 73.1% of the bacteria were aerobic or facultative anaerobes, and 3.6% were anaerobes (Table 3).

It was noted that specific gut bacteria such as Bifidobacterium and Bacteroidetes were present in 10 out of 11 milk samples. The presence of Bifidobacterium and Bacteroides was 90.9%, but their relative abundances were less than 1.0% (Supplementary Table S1).

Infants fed bacterial milk and infants fed sterile milk had similar intestinal microbial compositions, but microbial diversity varied. Proteobacteria and Actinobacteria constituted the dominant phylum in infant feces (Fig. 1b-c, Table 2). The most abundant genera were Bifidobacterium, Escherichia-Shigella, and Bacteroides (Fig. 1c). However, the IBM group had significantly higher Shannon and Simpson indices than the ISM group (P = 0.020, P = 0.048), revealing that infants fed bacterial milk have a richer microbial community (Fig. 2a-b). Beta diversity was estimated by using PCoA to seek interindividual differences in microbial communities. PCoA indicated similar clustering of microbiota among the IBM and ISM groups (P = 0.264, Fig. 2c).

We then drew a Venn diagram to visualize the shared and unique OTUs of the IBM and ISM groups. There were 273 shared OTUs between groups. Meanwhile, the IBM group has 1063 unique OTUs, which far exceeds those in the ISM group (Fig. 2d). This suggested that infants fed bacterial milk were more enriched with microbes.

The LEfSe analysis was performed to find biomarkers between the two groups. As illustrated in Fig. 2e, we found that Eggerthellaceae, Lachnospiraceae, and Lachnospiraceae were enriched in the IBM group. The abundance comparison diagram showed that Eggerthellaceae are absent from the ISM group, and Lachnospiraceae are only found in one infant gut (Fig. 2f-g).

Network analysis can reveal the internal relationships of a microbial community. We constructed the co-occurrence network and calculated the topological properties. The number of nodes, edges, and average degree in microbial network were higher in the IBM group (Fig. 2h-i, Table 4), which indicated the microbial community in infants fed with bacterial milk was more connected and complex.

As there were no significant differences in clinical characteristics between pregnant women, the gut microbiota was further examined. The microbial composition of PBM and PSM groups was similar. The primary phyla were Firmicutes and Bacteroidetes, while the main genera were Faecalibacterium and Bacteroides. In addition, Bifidobacterium was more abundant in the PSM group (5.3%) than in the PBM group (1.5%) (Supplementary Table S2).

To evaluate variations in the community structure and diversity between pregnant women with or without bacterial milk, we performed Alpha and Beta diversity. Shannon and Simpson indices revealed no significant differences between the PBM and PSM groups (P = 0.591, P = 0.180) (Fig. 3a-b). Meanwhile, PCoA showed no separation of communities between the PBM and PSM groups, suggesting that the two bacterial communities had similar compositions and structures (Fig. 3c).

To further search for specific taxa between the two groups, we performed a LEfSe analysis. A total of five biomarkers were identified when LDA score = 4. The PSM group had significantly more Actinobacteria, Bifidobacteriales, Bifidobacteriaceae, Prevotellaceae, and Bifidobacterium than the PBM group (Fig. 3d). The evolutionary branch diagram demonstrated that Actinobacteria, Bifidobacteriales, Bifidobacteriaceae and Bifidobacterium belong to the same branch (Fig. 3e). Co-occurrence networks were constructed to exhibit the relationship between microbes of pregnant women, focusing on the relationship between Bifidobacterium and other genera. It is worth noting that Bifidobacterium negatively correlated with six genera in the PSM group, while it negatively correlated with only one genus in the PBM group (Fig. 3f-g).

In this study, microbes were detected in 11 of 17 breast milk samples, with a positive detection rate of 64.7%. This result was close to the detection rate of our previous study [6] and the study of Kordy et al. [7]. Many researchers have detected microorganisms in human breast milk. Firmicutes was reported to be the most abundant phylum in milk [27, 28]. The common genera in breast milk included Staphylococcus, Streptococcus, Pseudomonas, and Lactobacillus [20, 29‐32]. In this study, Firmicutes were abundant in breast milk. Meanwhile, Streptococcus and Staphylococcus had the highest abundance, which was generally consistent with other studies [20, 27, 31, 32]. Furthermore, the milk was also rich in Lactobacillus in our study. However, the abundance of Pseudomonas in breast milk was low, and we detected the commensal skin bacterium Propionibacterium in only one milk sample. A previous research [19] comparing the breast milk bacterial communities of mothers from Spain, Finland, South Africa, and China, has discovered higher levels of Pseudomonas and Propionibacterium in the breast milk of Spanish mothers who delivered by cesarean section than those from other countries. Furthermore, Pseudomonas was much lower in Chinese women who delivered vaginally compared to those who delivered by cesarean section [19]. Therefore, we attribute differences in results to geographic location and delivery mode.

The detection rate of Bifidobacterium and Bacteroides in milk was as high as 90.9%. These bacteria are strict anaerobes that often reside in the gut and are unlikely to originate from the skin. A study has found the existence of gut-associated bacteria such as Bifidobacterium in maternal and infant feces and breast milk [25]. Moreover, Jost et al. [33] found a strain of Bifidobacterium in all three ecosystems of a mother-neonate pair. These findings indicate that some gut microbes can enter milk and then colonize neonatal intestinal.

It has been reported that delivery modes, feeding modes, gestational age at birth and timing of solid food introduction may affect bacterial communities in infant feces [1, 14, 15, 34]. Our study manifested that the presence or absence of milk microbes has a minor effect on infant gut microbial composition but has a greater impact on microbial diversity and microbial network relationships. Proteobacteria and Actinobacteria constituted the dominant phylum in infant feces, which was in accordance with Pannaraj’s report [1]. After being fed with bacterial milk, Alpha diversity and microbial network associations increased. By contrast, infants fed with sterile milk had fewer microbial species and lower microbial network complexity. Alpha diversity has been regarded as a marker of microbiome health, and a reduction in Alpha diversity is seen as a manifestation of intestinal malnutrition [35]. Low diversity is thought to be related to metabolic disorders [36]. In addition, it has been reported that microbial network stability is related to its complexity, and the complexity of the network will contribute to its stability [37]. Lachnospiraceae, which was abundant in the IBM group, is described as a potential beneficial bacteria [38]. Lachnospiraceae can generate butyrate by hydrolyzing amylase and sugar. Butyrate provides energy to colon cells and maintains the integrity of the intestinal barrier. It appears that bacterial milk may help infants shape a healthy gut microbiome. However, long-term follow up is required to track infants’ health outcomes.

Notably, the microbial composition of the IBM and ISM groups was similar, although some breast milk was not detected for bacteria. Bifidobacterium and Bacteroides were also enriched in the ISM group, which indicated that they could be transmitted to newborns in multiple ways. The researchers identified DNA and cellular structure of intestinal bacteria such as Bifidobacterium in the placenta, amniotic fluid, and fetal membrane [39‐41], suggesting that the intestinal microbiota such as Bifidobacterium may have colonized the fetal intestine during pregnancy. In addition, prebiotics such as human milk oligosaccharides can promote the proliferation and growth of beneficial bacteria such as Bifidobacterium and Lactobacillus [16]. Therefore, we still recommend breastfeeding even if maternal milk contains extremely low bacteria or is even sterile. It is imperative to carefully consider whether microbiota interventions are necessary for infants fed with sterile milk.

Geographical location, delivery mode, maternal BMI, infant gender and feeding mode could affect milk microbiota composition [20, 22, 27, 42, 43]. To figure out why some milk had very low bacterial content or was sterile, the participants’ characteristics were compared. The pregnant women all live in Guangzhou, China. Meanwhile, they delivered vaginally and fed their newborns exclusively with breast milk. There were no significant differences between the two groups in geographical location, delivery mode and feeding mode. Other clinical characteristics, such as age, BMI, gestational week and infant gender, showed no significant differences.

Since some researchers have proposed the entero-mammary pathway theory, we analyzed the gut microbiota of the PBM group and the PSM group. Gut microbiota composition and diversity were similar between the two groups. The main difference was that Prevotellaceae and Bifidobacterium were abundant in the PSM group, and Bifidobacterium had more negative correlations. Several in vitro and in vivo studies validated that Bifidobacterium was able to affect the rearrangement of tight junction proteins and strengthen the intestinal barrier [44‐46]. We speculated that Bifidobacterium might inhibit bacterial growth and improve intestinal permeability, leading to a decrease in the migration of microbes entering milk via the entero-mammary pathway. In contrast, Simpson et al. investigated the effect of probiotic supplementation on maternal breast milk and reported that no significant changes occurred in milk microbial community despite maternal supplementation [10]. Of note, only a few samples could be sequenced in both the probiotic and placebo groups due to low bacterial DNA quantities [10]. It has been suggested that newborns’ oral cavity and breast skin may provide additional microbes for milk. Regrettably, we didn’t collect oral samples and skin samples. However, if there are multiple sources of milk microbes, there should be more bacteria in milk. Taken together, further research is needed to determine the cause of undetectable bacteria in milk.

Many factors affect milk and gut microbes, such as skin microbes, environmental factors, and diet. Although we attempted to control for differences in race, age, BMI, and delivery mode among participants, it cannot rule out the possibility that other factors influenced the study's results. Our sample size is small, and studies with a larger sample size are needed to reduce the bias. Nevertheless, this study found distinct differences in Alpha diversity and network associations among infants fed with bacterial milk and sterile milk. This result deserves to be of interest.