A human milk oligosaccharide alters the microbiome, circulating hormones, and metabolites in a randomized controlled trial of older adults

Abstract

Aging-related immune dysfunction is linked to cancer, atherosclerosis, and neurodegenerative diseases. This 6-week randomized controlled trial evaluated whether 2'-fucosyllactose (2'-FL), a human breast milk oligosaccharide with established benefits in infants and animal models, could improve gut microbiota and immune function in 89 healthy older adults (mean age 67.3 years). While the primary endpoint of cytokine response change was not met, 2'-FL supplementation increased gut Bifidobacterium levels and elevated serum insulin, high-density lipoprotein (HDL) cholesterol, and FGF21 hormone. Bifidobacterium "responders" experienced additional metabolic and proteomic changes and also performed better on a cognitive test of visual memory. Nonresponders were more likely to lack Bifidobacterium in their gut microbiota at the start of the intervention. Multi-omics analysis indicated a systemic response to 2'-FL, which could be detected in blood and urine, showcasing the potential of this prebiotic to provide diverse benefits for healthy aging. This trial was registered at ClinicalTrials.gov (NCT03690999).

Keywords: 2'-fucosyllactose; Bifidobacterium; Human milk oligosaccharide; aging; metabolomics; prebiotics.

Graphical abstract

Figure 1

Design of a randomized controlled trial in healthy, older human subjects (A) The chemical structure of 2′-FL. 2′-FL is composed of lactose (purple) covalently linked to a fucose (orange) via a glycosidic linkage (green). (B) A Consolidated Standards of Reporting Trials (CONSORT) flow diagram of the individuals recruited and screened for this study. (C) The treatment arm allocation of the 89 individuals who were randomized to a treatment arm and received the product. The three arms were placebo, Low Dose 2′-FL (consuming 1 gram of 2′-FL per day; 0.5 g morning and evening), and High Dose 2′-FL (consuming 5 grams of 2′-FL per day; 2.5 g morning and evening). (D) Trial timeline (top) and sample collection schema (bottom). Each subject was tracked for a 2 weeks of baseline period; supplementation began on week 0 and continued for 6 weeks; participants were observed for a 4 week washout period after cessation of supplementation. Blood and stool were collected at each time point; urine was collected at the week 6 time point. Subjects also took cognitive tests at each time point.

Figure 2

2′-FL is detectable in plasma and urine of the High Dose 2′-FL group and results in significant increases in fasting insulin and HDL cholesterol, although it does not affect cytokine response score (A) 2′-FL is detected at significantly greater levels in plasma at the week 6 time point in the High Dose 2′-FL group compared to both the Low Dose 2′-FL group (p = 3.62 × 10⁻⁵, Wilcoxon rank-sum test) and the placebo group (p = 5.37 × 10⁻⁷, Wilcoxon rank-sum test). Lower limit of quantification is 0.02 μg/mL. p < 0.05 is indicated by ∗, p < 0.01 by ∗∗, and p < 0.001 by ∗∗∗. (B) 2′-FL is detected at significantly greater levels in urine at the week 6 time point in the High Dose 2′-FL group compared to both the Low Dose 2′-FL group (p = 0.0002, Wilcoxon rank-sum) and the placebo group (p = 2.56 × 10⁻⁶, Wilcoxon rank-sum). The Low Dose 2′-FL group trended toward higher concentrations of 2′-FL compared to the placebo group (p = 0.076, Wilcoxon rank-sum). Lower limit of quantification is 0.2 μg/mL. (C) A scatterplot showing the correlation between 2′-FL concentrations measured in plasma and urine, for participants that contributed both urine and plasma samples but excluding participants in the placebo arm. Pearson correlation coefficient is 0.729, and the p value of the association is 1.9 × 10⁻⁸. (D) Cytokine response score does not change significantly from baseline (week 0) to end of intervention (week 6) for any of the three treatment groups. (E) Percent change from baseline (week 0) to end of intervention (week 6) for metabolic markers. Fasting insulin and HDL cholesterol increased significantly from week 0 to week 6 in the High Dose 2′-FL group but not the Low Dose 2′-FL group or the placebo group. Dots are the mean change from baseline (week 0) to end of intervention (week 6), and lines are 95% confidence intervals of the means. p < 0.05 is indicated by ∗.

Figure 3

2′-FL induces a bloom of Bifidobacterium that is most prominent in week 3 and subsides to baseline levels during the washout period (A) Shannon (alpha) diversity of each treatment group at each time point during the intervention. (B) Beta diversity of each treatment group at baseline (week 0), midpoint (week 3), and end of intervention (week 6), depicted via principal coordinate analyses (PCoA). Treatment group explained a significant portion of variation at the midpoint (week 3) time point only (p = 0.03, Adonis test). Ellipses represent 95% confidence intervals around the centroids of each group. (C) Relative abundance of the genus Bifidobacterium for each treatment group at each time point during the intervention. The High Dose 2′-FL group had higher Bifidobacterium relative abundance at midpoint (week 3) compared to the placebo group (p = 0.0016, Wilcoxon rank-sum test) and also compared to the High Dose 2′-FL group at the pre-baseline week −2 (p = 0.01, Wilcoxon rank-sum test), baseline week 0 (p = 0.037, Wilcoxon rank-sum test), and washout week 10 (p = 0.037, Wilcoxon rank-sum test) time points. p < 0.05 is indicated by ∗, and p < 0.01 by ∗∗.

Figure 4

Serum metabolites and proteins that vary significantly by treatment arm during the intervention (A) Boxplots showing the log2 abundance of octanoylcarnitine (top), glutamate (middle), and taurine (bottom) from untargeted metabolomics analysis of serum samples. Three panels from left to right show metabolite abundance for the placebo group, Low Dose 2′-FL group, and High Dose 2′-FL group, respectively, in the baseline and intervention time points. P_int values next to each metabolite name are the p values of the linear mixed effect modeling, adjusted across all metabolites using the Benjamini-Hochberg method. p values within each facet are the results of Wilcoxon rank-sum post hoc tests between baseline and intervention time points. The baseline time points are weeks −2 and 0, and the intervention time points are weeks 3 and 6. (B) Boxplots showing the normalized protein expression values of FGF21 (top) and CD40 (bottom) from targeted proteomics analysis of serum samples. Three panels from left to right show metabolite abundance for the placebo group, Low Dose 2′-FL group, and High Dose 2′-FL group, respectively, in the baseline and intervention time points. P_int values next to each protein name are the p values of the linear mixed effect modeling, adjusted across all metabolites using the Benjamini-Hochberg method. p values within each facet are the results of Wilcoxon rank-sum post hoc tests between baseline and intervention time points. The baseline time points are weeks −2 and 0, and the intervention time points are weeks 3 and 6.

Figure 5

Subsetting subjects into responders and nonresponders based on change in Bifidobacterium abundance reveals widespread changes to levels of circulating cytokines (A) Subsetting of subjects into responders and nonresponders based on change in Bifidobacterium relative abundance between baseline (week 0) and midpoint (week 3). Participants were classified as responders if they were in the top tertile of Bifidobacterium increase from week 0 to week 3. (B) Bifidobacterium relative abundance at baseline (week 0) for responders and nonresponders is not significantly different (p = 0.055, Wilcoxon rank-sum test). (C) Plasma concentrations of 2′-FL at end of intervention (week 6) for responders and nonresponders. Responders have significantly greater plasma concentrations of 2′-FL (p = 0.038, Wilcoxon rank-sum test). The dashed line represents the lower limit of quantification (0.02 micrograms/mL). (D) Scatterplot showing the relationship between week 0 to week 3 change in Bifidobacterium abundance and plasma concentration of 2′-FL. Points are colored according to responder status; gray segments indicate standard error of linear trend. Linear regression analysis revealed that change in Bifidobacterium abundance from week 0 to week 3 was associated with higher levels of 2′-FL in plasma (p = 0.016, linear regression). Horizontal dashed line indicates limit of 2′-FL detection. (E) Scatterplot showing the relationship between week 0 to week 3 change in Bifidobacterium abundance and urine concentration of 2′-FL. Points are colored according to responder status; gray segments indicate standard error of linear trend. Linear regression analysis revealed that change in Bifidobacterium abundance from week 0 to week 3 was not associated with higher levels of urine 2′-FL concentration (p = 0.60, linear regression). Horizontal dashed line indicates limit of 2′-FL detection. (F) Volcano plot of untargeted serum metabolomics showing the metabolites enriched in responders and nonresponders. p values were false discovery rate corrected with the Benjamini-Hochberg method. Horizontal dashed line represents a q value threshold of 0.1. ACC, 1-aminocyclopropane-1-carboxylic acid. (G) Volcano plot of targeted serum proteomics showing the proteins enriched in responders and nonresponders. p values were false discovery rates corrected with the Benjamini-Hochberg method. Horizontal dashed line represents a q value threshold of 0.1.

Figure 6

Multi-omic associations between serum metabolomics and serum cytokine proteomics (A) A correlation network between serum metabolomics (green nodes), serum cytokine proteomics (orange nodes), and clinical markers (blue nodes). All-versus-all correlation network was based on centered and scaled baseline-to-end changes in individual metabolite/protein levels across all participants. Blue edges indicate positive correlation, and red edges indicate negative correlation. ACC = 1-Aminocyclopropane-1-carboxylic acid. (B) Significant pairwise correlations between serum metabolites and serum cytokines using Pearson correlation (p value correction with Benjamini-Hochberg method, corrected p values ≤ 0.05). Larger circles indicate more significant corrected p values. Color map indicates the strength of the Pearson correlation with blue indicating positive correlation and red indicating negative correlation. Metabolites and cytokines with asterisks were significantly associated with responder status.