An Extensively Hydrolyzed Formula Supplemented with Two Human Milk Oligosaccharides Modifies the Fecal Microbiome and Metabolome in Infants with Cow's Milk Protein Allergy

Abstract

Cow's milk protein allergy (CMPA) is a prevalent food allergy among infants and young children. We conducted a randomized, multicenter intervention study involving 194 non-breastfed infants with CMPA until 12 months of age (clinical trial registration: NCT03085134). One exploratory objective was to assess the effects of a whey-based extensively hydrolyzed formula (EHF) supplemented with 2'-fucosyllactose (2'-FL) and lacto-N-neotetraose (LNnT) on the fecal microbiome and metabolome in this population. Thus, fecal samples were collected at baseline, 1 and 3 months from enrollment, as well as at 12 months of age. Human milk oligosaccharides (HMO) supplementation led to the enrichment of bifidobacteria in the gut microbiome and delayed the shift of the microbiome composition toward an adult-like pattern. We identified specific HMO-mediated changes in fecal amino acid degradation and bile acid conjugation, particularly in infants commencing the HMO-supplemented formula before the age of three months. Thus, HMO supplementation partially corrected the dysbiosis commonly observed in infants with CMPA. Further investigation is necessary to determine the clinical significance of these findings in terms of a reduced incidence of respiratory infections and other potential health benefits.

Keywords: amino acids; bile acids; fecal community type; human milk oligosaccharides; metabolomics; metagenomics; microbiome; short-chain fatty acids.

Figure 1

Description of study design: One hundred ninety-four infants were randomized to a test group fed EHF supplemented with 2′-FL and LnNT and a control group fed the same EHF without HMO. Due to the large age heterogeneity at enrollment (see Supplementary Figure S1), the population was divided into an early enrollment (EE: aged 0 to 3 months) and a late enrollment cohort (LE: aged 3 to 6 months). Fecal samples were collected at 1 month (V1) and 3 months from the start of the study formulas (V3), as well as at 12 months of age (V6). If sufficient sample volume was available, metagenomics and targeted metabolomics analysis were performed. In total, samples from 132 infants (EE: n = 60; LE: n = 72) were available for metagenomic analysis, and 84 samples for fecal metabolomics (EE: n = 41; LE: n = 43), respectively. More details about sample numbers in the control and test groups at baseline (V0), visit 1 (V1), visit 3 (V3), and visit 6 (V6) are available in Supplementary Figure S2.

Figure 2

Comparison of microbiota compositions and trajectories between feeding groups (test vs. control), stratified by early enrollment (0–3 months of age; EE) and late enrollment (3–6 months of age; LE). (A) Alpha diversity (Faith’s phylogenetic diversity [PD]) of the gut microbiomes of the infants in the two feeding groups at each timepoint (V0, V1, V3, V6) stratified for the EE and LE cohorts. Box plots show the median, 25th and 75th percentiles with Tukey whiskers. At each time point, the two treatment groups were compared, stratified by EE and LE, using a two-sided Mann–Whitney U test. Nominal p values are indicated above the groups. (B) Taxonomical overview of the 5 fecal community types (FCT) at the genus level. Bar plots display the mean abundance within each FCT of the 10 most abundant taxa; the remainder are grouped in “Other.” (C) The transition model analysis depicts the temporal progression from early to late FCT clusters based on all 481 samples. Node sizes represent the fraction of infants in a given cluster per age group (column). Line widths represent the fraction of transitions per age group (column). The line color indicates the transition category (grey: no change, blue: progression, red: regression). In analogy to the methodology used by Stewart et al., only transitions with frequencies >4% are shown [2]. (D,E) Distribution of FCTs per feeding group in the EE (D) and LE cohort (E). For each time point, the proportion of samples assigned to each of the five FCTs is shown, stratified by treatment group and enrollment cohort. (F,G) Kaplan–Meier plot illustrating the transitioning to FCT3 (or later FCT, i.e., FCT4 or FCT5) as a function of age for the EE cohort (F) and LE cohort (G). Time of event is defined as the age at the earliest visit where the infant is observed to have the FCT3 or later FCTs. The two treatment groups were compared using the log-rank test, and nominal p values are indicated above each plot. Vertical lines indicate censored data. The seven infants with only a V6 sample are excluded. The bottom panel shows the number of samples for a given age (bin width = 15 days) and illustrates the uneven age distribution. NB: The LE cohort has no observations during the first 90 days. A cross on the survival line is marked when data are no longer available for a given infant beyond that time point.

Figure 3

Bifidobacterium species enrichment in the test group at V1 and V3 in the EE cohort. Each circle indicates the effect size (Cliff’s delta) of a single metagenomic species (MGS) in comparison to the relative abundance between the control and HMO-treated infants at (A) V1 and (B) V3. The included MGS (y-axis) are sorted by their corresponding Cliff’s delta (x-axis). The following numbers of samples were included at V1: EE control, n = 27; EE test, n = 32; and at V3: EE control, n = 25; EE test, n = 33.

Figure 4

HMO supplementation impacts the fecal metabolic profile in the EE cohort. (A) PLS regression score plot modeling age on the first component and feeding groups on the second component at V3 and V6. (B) Volcano plot showing cumulative VIP score and correlation coefficient to treatment (PC2). Each dot corresponds to a metabolite. Most discriminating metabolites (VIP > 1 & |PLS coefficient (p(corr))| > 0.2) to feeding groups are highlighted in green. (C) log₂ scaled concentration (nmol/g) of 2′-FL, lactose, 3-hydroxyphenylacetic acid, phenylacetic acid, isobutyric acid, and isovaleric acid in the control (highlighted by pink box plots) and test group (highlighted by blue box plots) at each visit. The cross symbols show the outliers that are 1.5 times the interquartile range away from the bottom or top of the box. Significant differences between treatment groups were calculated using the Wilcoxon–Mann–Whitney test (p-value * <0.05, *** <0.001).

Figure 5

HMO supplementation maintains fecal unconjugated/conjugated bile acid ratios and fecal acetic acid levels at each study visit. (A) The trajectory of the total unconjugated/conjugated bile acid ratio (total unconjugated bile acids is the sum of cholic acid, chenodeoxycholic acid, lithocholic acid, and deoxycholic acid). The total conjugated bile acids are the sum of taurocholic acid, taurochenodeoxycholic acid, taurodeoxycholic acid, glycocholic acid, taurolithocholic acid, glycochenodeoxycholic acid, and glycodeoxycholic acid). (B) Unconjugated/conjugated cholic acid ratio (C) Acetic acid in the control group and test group at each visit. Control group at each timepoint is presented in pink and test group is presented in blue. Significant differences between visits or between treatment groups were calculated using the Wilcoxon–Mann–Whitney test (p-value # <0.1 * <0.05, ** <0.01).

Figure 6

The fecal metabolome reflects HMO-related changes in the bacterial proteolytic activity (A) Heatmap illustrating the correlation between KEGG Orthologs (KO) and stool metabolites. Association map of the analyses integrating FCT distribution, test formula effect, the gut microbiome, and the stool metabolome in the early enrollment (EE) cohort at V3 and V6. The main ‘heatmap’ panel shows Kendall’s correlation between a priori selected bacterial enzyme (KOs) and stool metabolites using all V3 and V6 samples with matched metagenomic and metabolomic information (samples, n = 70). The colors indicate the direction and magnitude of the correlation (Kendall’s Tau correlation coefficient), where red means a positive correlation between the KO and the stool metabolite, and blue means a negative correlation. Statistically significant correlations are indicated with filled grey (Kendall correlation p < 0.05) or black (Kendall correlation FDR < 10%) circles. The right and bottom ‘sidebar’ panels show associations between the same KOs and metabolites, respectively, and HMO-treatment groups (+HMO) or correlation with FCT distribution (FCT) at V3 and V6. For +HMO, the colors indicate the direction and magnitude of the association (Cliff’s delta). Brown means that the KO or metabolite is more abundant in HMO-treated infants, and green means that the KO or metabolite is more abundant in control infants. For FCT, the colors indicate the correlation (Kendall’s Tau correlation coefficient), where red means that the KO or metabolite is enriched in late FCT, and blue means it is enriched in early FCT. Statistically significant associations/correlations are indicated with open (MWU/Kendall correlation p < 0.05) or filled (MWU/Kendall correlation FDR < 10%) circles. For sidebars, the following number of samples were included: At V3/V6, associations with KOs, control, n = 25/24; test, n = 33/25; associations with metabolites, control, n = 17/18; test, n = 21/14 (detailed statistics in Supplementary Tables S8–S10). (B) Production of fusel acids and fusel alcohols from amino acids via the Ehrlich or the biogenic amine pathway.