Association of Maternal Secretor Status and Human Milk Oligosaccharides With Milk Microbiota: An Observational Pilot Study

Abstract

Background and Objectives: 

Breast milk contains several bioactive factors including human milk oligosaccharides (HMOs) and microbes that shape the infant gut microbiota. HMO profile is determined by secretor status; however, their influence on milk microbiota is still uncovered. This study is aimed to determine the impact of the FUT2 genotype on the milk microbiota during the first month of lactation and the association with HMO.

Methods: 

Milk microbiota from 25 healthy lactating women was determined by quantitative polymerase chain reaction and 16S gene pyrosequencing. Secretor genotype was obtained by polymerase chain reaction-random fragment length polymorphisms and by HMO identification and quantification.

Results: 

The most abundant bacteria were Staphylococcus and Streptococcus, followed by Enterobacteriaceae-related bacteria. The predominant HMO in secretor milk samples were 2’FL and lacto-N-fucopentaose I, whereas non-secretor milk was characterized by lacto-N-fucopentaose II and lacto-N-difucohexaose II. Differences in microbiota composition and quantity were found depending on secretor/non-secretor status. Lactobacillus spp, Enterococcus spp, and Streptococcus spp were lower in non-secretor than in secretor samples. Bifidobacterium genus and species were less prevalent in non-secretor samples. Despite no differences on diversity and richness, non-secretor samples had lower Actinobacteria and higher relative abundance of Enterobacteriaceae, Lactobacillaceae, and Staphylococcaceae.

Conclusions: 

Maternal secretor status is associated with the human milk microbiota composition and is maintained during the first 4 weeks. Specific associations between milk microbiota, HMO, and secretor status were observed, although the potential biological impact on the neonate remains elusive. Future studies are needed to reveal the early nutrition influence on the reduction of risk of disease.

What Is Known

• Human milk oligosaccharide differ between secretor and non-secretor mothers.
• Human milk oligosaccharide and milk microbes drive the infant gut microbial colonization.
• Maternal secretor status influences neonatal gut microbiota.
• Disruption of early microbiota colonization process has been related to risk of disease.

What Is New

• The maternal secretor status affects specific milk bacteria load, microbial composition.
• The impact of maternal secretor status in microbiota affects different bacterial groups during lactation.
• Interactions between human milk oligosaccharide and microbes are dependent on maternal genotype.

Human milk (HM) is a dynamic and complex fluid with compositional changes over lactation. Beyond its nutritive role, HM also contains a wide range of bioactive compounds such as HM oligosaccharides (HMO) and microbes that support the neonatal gut microbiota colonization, which affects mucosal and systemic immunity ^(1,2). It is known that breast-fed infant microbiota is different than formula-fed infant microbiota, which can be partially explained, by milk microbes and HMO ^(3,4). Staphylococcus and Streptococcus have been reported universally present in HM ⁽⁵⁾, although other bacteria as Bifidobacterium, Lactobacillus, and Acinetobacter are frequently detected ^(2,6). Those bacteria have been also reported to able to use HMO in in vitro models ^(7–12). It is known that bacterial strains are shared in mother-infant pair ^(13,14) and also, specific HMO are associated with Bifidobacterium microbiota ^(15,16).

Moreover, HMO profile is determined by host secretor status, encoded by fucosyltransferase 2 (FUT2) gene, based on the secretor/Lewis blood group ^(17,18). Shifts in gut microbiota depending on secretor status have been previously reported ^(19–21). Those studies were focused on the reduced Bifidobacterium diversity and abundance in non-secretor as compared with secretor individuals, although other studies showed no impact ^(22,23). Other studies highlighted the relevance of maternal secretor status in microbiota changes during pregnancy ⁽²¹⁾ and also, on the neonatal gut microbiota ^(16,24,25). In general, non-secretor microbiota harbors lower Actinobacteria, Bifidobacterium, and Bacteroides groups, and lower microbial diversity than secretor microbiota ^(19–21). In this scenario, it is, however, needed to determine the impact of maternal secretor status on HM microbes during lactation and also, to ascertain the potential milk microbes-HMO relationship which would have relevant biological consequences in infant gut development and health programming.

MATERIALS AND METHODS

Subjects and Design

Twenty-five healthy Spanish mothers with exclusive breast-feeding practices participated in this pilot study. Breast milk (BM) samples were collected longitudinally during first month of lactation (≤5 days: colostrum, <15 days: transitional milk, and 1 month: mature milk). Inclusion criteria included no metabolic or chronic diseases and also, no probiotics consumption during pregnancy and lactation, use of medications, or drugs.

Clinical data such as age, body mass index (BMI), gestational age, mode of delivery, and antibiotic use were recorded as well as neonatal weight at birth. Women received detailed study information and signed informed consent form. The protocol was approved by the Ethics Committee of the Hospital Clínico Universitario de Valencia (Spain), and the Bioethics CSIC Subcommittee (CSIC, Spain). This pilot study was carried out within the MAMI birth cohort (NCT03552939).

Breast Milk Samples

Mothers were given oral and written instructions for standardized milk collection in the morning at hospital. Breast skin was cleaned with 0.5% chlorhexidine solution to reduce bacterial concentration and first drops were discarded. Then, BM was collected by use of a sterile BM pumper (Medela Symphony, Barr, Switzerland) in sterile bottles. Finally, BM samples were aliquoted and stored at −80°C until further analysis.

DNA Extraction and FUT2 Genotyping

Samples were centrifuged at 10,000g for 10 minutes to pellet the cells and also, to concentrate the sample. Total DNA was isolated using QiAGEN Kit (QIAgen, Hilden, Germany) according to the manufacturer's instructions with some modifications, including a bead-beater step and enzyme incubation as described previously ⁽²⁶⁾. The secretor status was investigated by genotyping the FUT2 gene by polymerase chain reaction (PCR)-random fragment length polymorphisms (RFLPs) as described previously ⁽²⁷⁾.

Microbial Composition by Quantitative Polymerase Chain Reaction

PCR primers targeted to total bacteria and Bifidobacterium, Lactobacillus, Enterococcus, Staphylococcus, and Streptococcus groups were conducted as previously described ⁽²⁶⁾. qPCR were performed in LightCycler 480 real-time PCR System (Roche) by use of SYBR Green PCR Master Mix (Roche). A melting curve analysis was made and bacterial concentration was calculated by comparing the Ct values from standard curves.

Bifidobacterium species prevalence (B longum, B breve, B bifidum, B adolescentis, and B catenulatum) was also determined as described elsewhere ⁽²⁸⁾. For statistical analysis, SPSS 17.0 software was used. Microbiota data were Log transformed and parametric tests as T test were performed. Parametric repeated measures analysis of variance test was used to compare microbial groups along time (paired samples). The chi-square test was applied to establish differences in bacterial prevalence between the studied groups. A P < 0.05 was considered statistically significant.

Microbial Composition by 16S Gene Pyrosequencing

A specific 16S gene amplicon (500 pb size from V1-V3 region) were amplified as described previously ⁽²⁹⁾. Amplicons were subsequently cleaned using the Ampure purification system (Beckman and Coulter, Takeley, UK). Purified PCR products were pooled in equimolar amounts, as described by 454 Roche protocols, and the 16S-pyrosequencing was performed at FISABIO-GVA (Valencia, Spain) using the 454 FLX sequencer (Roche, Basel, Switzerland).

Raw sequences were filtered on the basis of quality and length. An end-trimming was performed by removing low-quality nucleotides at the 3′ end through windows of 20 nt of average quality values <20, and a second filtering was performed by removing those reads with an average quality value <20 and with <200 nt of length. These steps were performed through the galaxy server (http//getgalaxy.org/). These sequences were further filtered to remove chimeric sequences with the Usearch program and the gold database ⁽³⁰⁾. Alpha and beta-diversity was determined using QIIME pipeline (Version 1.9.0, default parameters) using clustering in operational taxonomic units with 97% identity to the RPD database. Relative abundances of bacteria, Alpha diversity indices (Chao1 and Shannon) were also obtained.

Human Milk Oligosaccharide Profile

HMO were extracted and identified following the procedure described elsewhere ⁽³¹⁾. Briefly, milk was centrifuged after the addition of pure water. Solid phase extraction with porous graphitic carbon cartridges (HyperSep-96 Wells, 25 mg; Thermo Scientific, Bellefonte, PA) was performed via a Hamilton Microlab Starlet liquid handling system (Hamilton Robotics, Reno, NV). After elution, OS were dried overnight in a vacuum centrifuge, and resuspended in ultrapure water. An HPAEC-PAD system, ICS-5000, equipped with a Carbo Pac PA-1 column (250 × 4 mm; 30°C) and a guard column, was operated by Chromeleon 6.80 software (all: Dionex, Idstein, Germany). For statistical analysis, SPSS 17.0 software was used. Parametric tests as T test and repeated measures analysis of variance test were used to compare HMO concentrations. A P < 0.05 was considered statistically significant.

Associations Between Milk Microbiota and Human Milk Oligosaccharide

To know the clustering of the different types of samples, together with the different HMO values, principal component analysis (PCA) was used to identify the largest sources of variation between the different groups. Spearman rank test allowed the study of the correlation between variables and significance was established at a coefficient of 0.5%. In addition, significant correspondences were estimated by Spearman correlation and adjusted P value based on the Benjamini-Hochberg procedure was carried out to correct for multiple testing for qPCR genus and sequencing microbiota results versus different oligosaccharides concentrations with the statistical package (http://www.t-project.org/). Statistical significance established at adjusted P < 0.05.

RESULTS

Maternal Characteristics, Milk Samples, and FUT2 Genotype

All mothers were healthy with a pregestational BMI <25 kg/m² (normal weight). C-section was present in 44% (11/25) (Table S1, Supplemental Digital Content, https://links.lww.com/MPG/B523). Maternal secretor status obtained by genotyping the FUT2 gene by PCR-RFLP was used to stratified samples as secretor and non-secretor. The proportion of non-secretor mothers was 44.0% (11/25) and secretors were 56% (14/25) according to FUT2 RFLPs-PCR genotyping. No significant differences were observed between FUT2 groups in terms of age, mode of delivery, gestational age, and antibiotic use. All mothers had the same socioeconomic status and demographics.

Microbial Composition by Quantitative Polymerase Chain Reaction

Maternal secretor status had an impact on the levels of specific bacteria in milk (Fig. 1 A–C). In colostrum, higher levels of Lactobacillus (P = 0.0004), Streptococcus (P = 0.0030), and Enterococcus (P = 0.014) were observed in secretor milk samples compared to non-secretor (Fig. 1A). In transitional milk, higher levels of Streptococcus (P = 0.006) and Enterococcus (P = 0.003) were observed in secretor milk (Fig. 1B). In mature milk, higher levels of Streptococcus (P = 0.043) were observed in secretor compared to non-secretors samples (Fig. 1C). Despite the significant lower levels of Bifidobacterium levels in secretor samples in colostrum (P = 0.040) compared to non-secretor samples, no differences were found in transitional and mature milk. The prevalence of different species of Bifidobacterium genus, as B longum, B bifidum, and B. breve, was, however, higher in secretor than in non-secretor (Table 1), which was dominated by B longum and B breve with lower prevalence or absence of B bifidum, B adolescentis, and B catenulatum.

FIGURE 1: Microbiota composition by quantitative polymerase chain reaction (qPCR) differs according to the stage of lactation and secretor status. Microbiota composition by qPCR in colostrum (A), transitional (B), and mature milk (C). Principal component analysis (PCA) analysis of microbiota profile in mature milk (light grey = secretor and dark grey = non-secretors). ^∗ P value <0.05. ^# Tendency P value ≥0.05 and ≤0.070. Unit variance scaling is applied to rows; singular value decomposition with imputation is used to calculate principal components. X- and Y-axis show principal component 1 and principal component 2 that explain 42.6% and 23.3% of the total variance, respectively. Prediction ellipses are such that with probability 0.90, a new observation from the same group will fall inside the ellipse. N = 25 data points.

TABLE 1: Prevalence of Bifidobacterium group and species in breast milk samples of secretor and non-secretor women during lactation

PCA showed 2 different clusters according to maternal secretor status on the qPCR profiles in mature milk (Fig. 1D). Scatter plot of the first 2 principal components (comprising 68% of total variance) suggested the potential different microbiota profile between secretor and non-secretor milk samples.

Microbial Composition by 16S Gene Pyrosequencing

We analyzed 25 mature milk samples and after quality checking, a total number of 22 samples were included (3 samples were discarded due failure in quality controls). After quality filtration and the length trimming, a total of 134,140 sequences were obtained, with an average of 7494 (±2193 SD) 16S rRNA high-quality sequences were generated per sample.

Higher interindividual variation in milk microbiota was observed. In general microbiota was dominated by Firmicutes and Proteobacteria. Among Firmicutes, the most predominant microbial genera were Streptococcus and Staphylococcus spp and among Proteobacteria, the main genera were Enterobacteria, Pseudomonas, and Acinetobacter. We found lactic acid bacterial families such as Streptococcaceae, Leuconostocaceae, and Lactobacillaceae but also typical inhabitants of the oral cavity such as Veillonellaceae and Flavobacteriaceae.

Different microbial profile at phylum and family level were found between BM samples depending on secretor status (Fig. 2A,B). Higher relative abundance of Actinobacteria phylum was significantly different in secretor milk samples compared with non-secretor (8.08% vs 1.00%, P = 0.050). At family level, we also found lower relative abundance of Enterobacteriaceae (12.65% vs 25.56%, P = 0.006), Veillonellaceae (1.18% vs 3.75%, P = 0.033), and Pseudomonaceae (13.10% vs 0.9%, P = 0.049) and higher relative abundance of Leuconostocaecae (5.52% vs 8.64%, P = 0.038) in secretors compared with non-secretors. At genus level, lower abundance of Veillonella (P = 0.030) and higher abundance Corynebacterium (P = 0.022) in secretor samples. Maternal secretor status did not affect the alpha diversity indexes, Chao 1 and Shannon, for diversity and richness, respectively.

FIGURE 2: Microbiota and human milk oligosaccharide (HMO) profile differs according to the stage of lactation and secretor status. Microbiota composition (n = 22 samples) by 16S gene sequencing in mature milk at phylum (A) and at family level (B) (internal circle = secretors and external circle: non-secretors); and HMO profile during lactation according to secretor status (C) (light grey = secretor and dark grey = non-secretors).

Human Milk Oligosaccharide Profile

Significant differences were found in total HMO in colostrum, transitional and mature milk according to maternal secretor status (Figs. 2C and 3A). In the present study, 14 mothers were identified by HPAEC-PAD as secretors having an HMO profile dominated by high concentrations of 2’FL and lacto-N-fucopentaose I (LNFP I). Both HMOs, 2’FL and LNFP I, were missing in non-secretors milk supporting the data obtained by PCR-RFLPs. Lacto-N-difucohexaose 1 (LNDFH 1), although secretor specific as well, could not be separated well from 3FL at the used HPAEC-PAD conditions. Then, the data were showed as 3FL+LNDFH 1 and therefore, not used for the statistical evaluations between single HMOs and microbiota.

FIGURE 3: Total HMO profile and microbiota during the first 4 postnatal weeks according to maternal genotype status. A, Total, neutral, and acidic HMO profile during lactation in g/L. B, Association between specific bacteria identified by quantitative polymerase chain reaction (qPCR) and total HMO at different lactation stages. C, Spearman heat map plot showing the impact of secretor status influence on HMO-microbes interaction during lactation. (light grey = secretor and dark grey = non-secretors). ^∗ P < 0.05. HMO = human milk oligosaccharide.

Association Between Milk Microbiota and Human Milk Oligosaccharide

Different associations between milk microbiota and HMO were found in milk of the first month. In general, Lactobacillus, Staphylococcus, and Streptococcus spp were the groups linked to HMO concentrations (Fig. 3B). Higher Lactobacillus levels were consistently associated to higher levels of total and neutral HMO while higher levels of Staphylococcus were significantly associated to lower concentration of total and neutral HMO in colostrum samples. Interestingly, Streptococcus levels were significantly linked to total and neutral HMO in transitional milk but not in colostrum and mature milk samples. Heat map plot (Fig. 3C) based on a hierarchical clustering of a real-valued similarity matrix, showed specific association between microbes and HMO depending on secretor status during lactation. Higher levels of Lactobacillus were linked to higher concentration of 2’FL (Rho = 0.542, P = 0.038 in colostrum and Rho = 0.700, P = 0.0013 in mature milk). Lactobacillus levels in mature milk were also associated with higher concentrations of LNFP I (Rho = 0.676, P = 0.0014) and lactodifucotetraose (Rho = 0.54, P = 0.0134) and lower levels of LNFP II (Rho = −0.650, P = 0.0020) and LNDFH II (Rho = −0.493, P = 0.0250).

We also found specific association between HMO and microbiota profile obtained by pyrosequencing. A higher concentration of total HMO was related to lower relative abundance of Enterobacteriaceae (Rho = −0.68, P = 0.050) and unclassified Enterobacteriaceae (Rho = −0.766, P = 0.021) and unclassified Clostridiales (Rho = −0.803, P = 0.009). Neutral HMOs were associated to lower unclassified Enterobacteriaceae (Rho = −0.768, P = 0.017), whereas acidic HMOs were linked to lower relative abundances of Leuconostoc (Rho = −0.700, P = 0.043).

Interestingly, higher abundance of 2’FL was associated to higher abundance of unclassified Enterococcaceae (Rho = 0.858, P = 0.0031). Higher concentration of LNnT in mature milk was associated to lower relative abundance of unclassified Clostridiales (Rho = −0.785, P = 0.012), Staphylococcaceae (Rho = −0.782, P = 0.013), and Flavobacteriaceae (Rho = −0.707, P = 0.0033).

DISCUSSION

In recent years, HMO and milk microbiome received much attention. Therefore, among others, questions of whether maternal secretor status affects the milk microbial profile need to be addressed. Our results reveal that maternal secretor status is associated to the milk microbiota composition. Our findings also provide supporting evidence that milk bacterial communities are associated with HMO profiles during the first 4 weeks of lactation.

It is well known that gut microbiota of exclusively breast-fed infant is different from those formula-fed ^(32,33). Different studies highlight the vertical transmission of specific bacterial species that are shared between maternal gut, milk, and infant feces ^(13,14). Recent study showed that milk microbes contributed in 27.7% and areola skin in 10.3% to the neonatal gut microbiota ⁽²⁾, demonstrating the relevance of milk microbes in the infant colonization process.

Despite the high microbial variability, we found higher relative abundance of Firmicutes and Proteobacteria, both being the major contributor to milk microbiome and also, Streptococcus and Staphylococcus spp being the most abundant bacteria in agreement with other studies ^{(2,5,9,29,33)}. In vitro studies have shown that most of the bacteria identified in milk as Bifidobacterium, Bacteroides, and Staphylococcus can use HMO to grow ^(8–11). HMOs are complex carbohydrates of which more than 200 different forms have now been described which presence is unique for humans. HMO may have many biological functions including protection against pathogenic bacteria and virus, improving intestinal health and contribute to brain development ^(34,35). Furthermore, HMO have been suggested to be associated to neonatal gut microbiota composition ^(15,36), on infant growth and body composition ^(16,37,38) and in the allergy risk ⁽³⁹⁾. HMO composition is determined by host secretor status, encoded by FUT2 gene ^(18,19). In agreement with previous reports ^(18,31), we found significant differences in total and specific HMO profiles during lactation according to maternal secretor status.

Furthermore, accumulating evidence is showing the FUT2 potential impact on gut microbiota composition and diversity ^(20,21), although other studies showed no influence ^(22,23). Higher levels of Bifidobacterium and Bacteroides were reported in breast-fed neonates during first month from secretor mothers compared to non-secretor ⁽¹⁶⁾. Moreover, the impact of maternal secretor status on BM microbes during lactation is still not known. Our data demonstrate the impact of maternal secretor status on the levels of specific bacteria determined by qPCR during the postnatal period and on the mature milk microbial profile by high-throughput sequencing.

By qPCR, higher levels of Lactobacillus, Streptococcus, and Enterococcus were reported in secretor samples compared to non-secretor milk during the first weeks of lactation. We found significant lower levels of Bifidobacterium in secretor samples in colostrum compared to non-secretor samples, although no differences were found in transitional and mature milk. Furthermore, PCA showed 2 differential qPCR milk microbiota clusters according to maternal secretor status in agreement with previously reported gut Bifidobacterium denaturing gradient gel electrophoresis-PCA plot between secretor and non-secretors ⁽¹⁹⁾. Furthermore, different studies have shown that secretor gut microbiota harbors a distinct Bifidobacterium species diversity ^(17,21). We found a reduced prevalence of Bifidobacterium spp in non-secretor compared to secretors milk samples. B bifidum, B adolescentis, and B catenulatum species are less prevalent in non-secretors compared to secretor milk samples. Our observation is in agreement with other data showing that non-secretor gut microbiota is lacked (or noncolonized) by B bifidum, B adolescentis, and B catenulatum/pseudocatenulatum⁽²¹⁾.

By sequencing, we also found distinct BM microbial profile at phylum and family level between secretor and non-secretor mothers, although no differences in microbial richness (Chao 1 index) and diversity (Shannon index) were detected. Higher relative abundance of Actinobacteria phylum (mostly represented by Bifidobacterium spp) and also, lower abundance of Enterobacteriaceae, Pseudomonaceae, and Veillonellaceae and higher abundance of Lactobacillaceae and Leuconostocaecae were found in secretors compared to non-secretors.

In addition, our results are support the limited data on the complex interaction between microbiota and HMO during lactation ^(40,41). We did not quantify the 3FL and LNDFH 1 as single components because of the overlapping coelution. Hence, the data were shown as 3FL+LNDFH 1. Therefore, 3FL+LNDFH 1 was not included in the microbiota and HMO associations. In general, Lactobacillus, Staphylococcus, and Streptococcus spp were the main bacterial groups associated to HMO. Higher Lactobacillus and lower Staphylococcus levels associated to lower total and neutral HMO in colostrum samples. Interestingly, Streptococcus levels were significantly linked to total and neutral HMO in transitional milk but not in colostrum and mature milk samples. We found that Lactobacillus levels were associated positively with 2’FL, LNFP I, and lactodifucotetraose and negatively with LNFP II and LNDFH II. Lactobacillus group was also linked to LNnT (P = 0.069 in mature milk) in agreement with other studies showing that L acidophilus was found to be most efficient at utilizing LNnT ⁽⁴²⁾. Our data, however, overlap partially with other data ^(40,41). Those different findings would be explained by methodological aspects (methodology qPCR vs sequencing; 16S gene region, sequencing platform, etc) and also, clinical (age, BMI, mode of delivery, lactation stage, gestational age) and environmental (diet, geographical location, ethnicity) factors and also, host genotype (FUT2). The relevance of geographical location in milk microbiota ^(6,43) and in the HMO ⁽⁴⁴⁾ has been already demonstrated.

Given the current challenges and limitations of this observational pilot study, that includes the number of volunteers and methodologies, our results reveal a substantial impact of maternal FUT2 genotype on the specific microbial groups in HM during lactation. Despite our small subsample cohort from a specific, homogeneous, and well-controlled population in Spain, it is needed to expand the study and also, to design futures studies with higher number of BM samples, with the use of new sequencing methodologies coupled with culture techniques and multilocation studies are needed and the information on infant health outcomes. Therefore, future studies are stringently needed to understand the complex links between host genotype-HMO-microbiome and its impact on health by microbial, metabolic, and immunological programming.