Human milk oligosaccharides (HMOs) may provide health benefits to infants partly by shaping the development of the early-life intestinal microbiota. In a randomized double-blinded controlled multicentric clinical trial, healthy term infants received either infant formula (control) or the same formula with two HMOs (2=- fucosyllactose and lacto-N-neotetraose; test) from enrollment (0 to 14 days) to 6 months. Then, all infants received the same follow-up formula without HMOs until 12 months of age. Breastfed infants (BF) served as a reference group. Stool microbiota at 3 and 12 months, analyzed by 16S rRNA gene sequencing, clustered into seven fecal community types (FCTs) with marked differences in total microbial abundances. Three of the four 12-month FCTs were likely precursors of the adult enterotypes. At 3 months, microbiota composition in the test group (n 58) appeared closer to that of BF (n 35) than control (n 63) by microbiota alpha (within group) and beta (between groups) diversity analyses and distribution of FCTs. While bifidobacteriaceae dominated two FCTs, its abundance was significantly higher in one (FCT BiH for Bifidobacteriaceae at high abundance) than in the other (FCT Bi for Bifidobacteriaceae). HMO supplementation increased the number of infants with FCT BiH (predominant in BF) at the expense of FCT Bi (predominant in control). We explored the association of the FCTs with reported morbidities and medication use up to 12 months. Formula-fed infants with FCT BiH at 3 months were significantly less likely to require antibiotics during the first year than those with FCT Bi. Previously reported lower rates of infection-related medication use with HMOs may therefore be linked to gut microbiota community types. (This study has been registered at ClinicalTrials.gov under registration number NCT01715246.) IMPORTANCE Human milk is the sole and recommended nutrition for the newborn infant and contains one of the largest constituents of diverse oligosaccharides, dubbed human milk oligosaccharides (HMOs). Preclinical and clinical association studies indicate that HMOs have multiple physiological functions largely mediated through the establishment of the gut microbiome. Until recently, HMOs were not available to investigate their role in randomized controlled intervention trials. To our knowledge, this is the first report on the effects of 2 HMOs on establishing microbiota in newborn infants. We provide a detailed description of the microbiota changes observed upon feeding a formula with 2 HMOs in comparison to breastfed reference infants’ microbiota. Then, we associate the microbiota to long-term health as assessed by prescribed antibiotic use.
At delivery, a microbiologically essentially sterile infant ( 1 , 2) is exposed to a multitude of microbes from the mother and the environment ( 3 , 4). The infant’s gut is progressively colonized with a dense microbial population. Donor effects are important, as seen from gut microbiota differences between Cesarean and vaginal deliveries (3, 5–8). However, nutrition also has an important impact on the composition of the gut microbiota, as seen from differences between breastfed and bottle-fed infants ( 8 , 9) and from the cessation of breastfeeding ( 6). Neonatal gestational age ( 5 , 10 , 11), antibiotic therapy (12 , 13), and diarrhea (14 , 15) are additional factors affecting the development of the gut microbiome. Microbes play a key role in the development of the immune system (16–18) and host metabolism (19 , 20). They are therefore speculated to exert a key impact on neonate and infant health that may last until later in life (21–23). Milk is a rich biological fluid providing both protection and nutrition for the suckling newborns. Human milk contains nutrients and innate immune factors to support normal growth and development. Nondigestible and structurally diverse oligosaccharides, known collectively as human milk oligosaccharides (HMOs), form one of the major breastmilk components. They may support immune function through the modulation of the gut microbiome ecology, resulting in colonization resistance, and the establishment of an age-appropriate gut microbiota, educating the mucosal immune system in its development (16 , 24 , 25). Due to their structural similarity with mucosal glycans, HMOs may also function as soluble decoy receptors in the gut, protecting the neonate from enteric pathogens (26), and may directly interact with gut epithelial cells, yielding changes that may modulate host-microbial interactions (25). In human milk, the oligosaccharides are extensions of lactose by one or more of the following monosaccharides: glucose, galactose, N-acetylglucosamine (GlcNAc), fucose, and sialic acid ( N-acetylneuraminic acid) (25 , 27). Three classes of oligosaccharides coexist: neutral fucosylated, neutral nonfucosylated with N-acetylglucosamine, and acidic with sialic acid. In contrast, cow’s milk contains very low levels of oligosaccharides, which are primarily neutral nonfucosylated with galactose only and acidic with sialic acid (27–29). Consequently, cow’s milk-based infant formula contains only relatively low levels of oligosaccharides, expected to be less than 100 mg/liter of the reconstituted formula (27), which, moreover, do not match with the major oligosaccharide classes found in human milk. We previously reported on the primary outcome of a randomized double-blinded controlled multicentric safety clinical trial, in which a formula containing two major HMOs, namely, 2 =-fucosyllactose (2 =FL) and lacto-N-neotetraose (LNnT), was found to be safe and well tolerated, allowing for age-appropriate growth of the infants (30). As part of the secondary objectives, we observed associations between feeding of the twoHMO formula and reduced rates for reported illnesses (in the lower respiratory tract) and infection-related medication use (antibiotics and antipyretics). Here, we report on the impact of these HMOs (2 =FL and LNnT) on the establishment of the gut microbiota, and we further explore its relationship with the reported illnesses and infection-related medication use.
RESULTSClinical trial. The randomized, double-blinded, controlled, multicenter interventional clinical trial with two parallel formula-fed groups was registered at ClinicalTrials- .gov (registration number NCT01715246, 16 October 2012). Healthy term infants received either infant formula without HMOs (control group) or the same formula with two HMOs (1.0 g/liter 2 =FL and 0.5 g/liter LNnT; test group) from enrollment to 6 months. Then, all infants received the same follow-up formula without HMOs until 12 months of age. A group of 38 infants exclusively breastfed (BF) since birth and whose mothers intended to exclusively breastfeed at least to 4 months was enrolled as a
reference. From 4 months of age, complementary feeding (solid food) was allowed. The trial details and clinical findings related to the primary objective and supportive secondary objectives were recently published (30). The per-protocol (PP) infants who completed the 6-month treatment and for whom we had stool samples at 3 months of age represented 74% (control, 64/87) and 66% (test, 58/88) of the corresponding intention-to-treat population (Fig. 1). This well-controlled subpopulation was used to characterize the impact of the HMO supplementation on the stool microbiota. Taxonomic composition in the stool microbiota by 16S rRNA gene sequencing. Stool samples were collected at 3 months and 12 months of age. Microbiota composition was determined by multiplexed high-throughput sequencing of amplicons obtained from the V3 and V4 regions of the 16S rRNA gene. After quality filtering, 16,014,421 sequences described the microbiota of 282 samples with an average coverage of 47,430 (median) sequences per sample classified into 336 operational taxonomic units (OTUs). Four samples of the per-protocol (PP) set with fewer than 10,000 sequences were excluded. Finally, the 3-month samples described 72% (control, 63/87) and 66% (test, 58/88) of the ITT population, and the 12-month samples described 56% (control, 49/87) and 53% (test, 47/88) of the ITT population (Fig. 1). Working with a subpopulation may affect the bias elimination of the study randomization. Therefore, we tested that the ITT population and the PP population reported here showed no bias in the baseline characteristics between the formula groups (Table 1) and no major difference in the clinical data (Table 2). Noteworthy, we used an approach allowing accurate annotation of the 16S rRNA sequences belonging to the genera Bifidobacterium (the dominant taxon at 3 months) and Lactobacillus down to the species or subspecies level (15). The taxonomic composition of all samples is reported at genus level in Fig. S1 in the supplemental material. At 3 months, formula with two HMOs shifted stool microbiota composition and diversity toward that of BF infants. When comparing the phylogenic diversity (31) between feeding groups at 3 months, the lowest was in the BF group, and the test group was significantly lower than the control group (P 0.05) and therefore closer to the BF group (Fig. 2A). Stratification by delivery mode showed a similar trend (Fig. 3A).
The global difference in microbiota compositions between feeding groups was statistically assessed by random permutations of redundancy analysis (RDA). At 3 months, the three groups were significantly separated at the genus level (P 0.001) (Fig. 2B), with the test group closer to the BF group. Likewise, the test and control groups were significantly separated at the genus level (RDA1 component 1%; P 0.036). After stratification by mode of delivery, RDA at the genus level revealed a stronger contribution of the Caesarean-delivered infants to the overall difference between the formula-fed groups and the BF group (Fig. 3B). We calculated the Bray-Curtis distances between samples at the genus level and evaluated the separation between the test (T), the BF, and the control (Ct) groups (Fig. 3C and Table S1). As shown on Fig. 3C, the groups of Caesarean (C)-section delivered infants (the triangle T.C↔BF.C↔Ct.C) were more distantly related to each other than the groups of vaginally delivered infants (the triangle T.V↔BF.V↔Ct.V). Although the BF infants were always separated from the formula-fed infants, irrespective of delivery mode, the separation was clearer with the control group (Ct.C↔BF.C compared to T.C↔BF.C, and Ct.V↔BF.V compared to T.V↔BF.V). In the vaginally delivered infants, the separation between the test and BF groups did not reach significance (T.V↔BF.V). Noteworthy, in the test group, we could not distinguish between the delivery modes (T.C↔T.V), similarly to the