Gut Check: Sugar, Fiber, and Infant Microbiota

A comprehensive understanding of what drives differences in the composition of infant gut microbiota, the microbial communities in the intestinal tract, remains elusive despite growing evidence of its importance to health [1]. Some aspects have been appropriately described, such as the effects of breastmilk or formula on gut microbiota composition [2]. However, the effects of other dietary factors, such as complementary foods, are less clear. The gap in understanding the effects of dietary sugars and fiber on infant gut microbiota is particularly relevant, because intake of these nutrients, which is suboptimal for most United States infants, is likely to play a role in the childhood obesity epidemic [3]. Addressing these gaps and identifying affected microbes could lead to interventions that facilitate obesity prevention in childhood.

In this issue of the Journal of Nutrition, Mokhtari et al. conducted a cross-sectional study to examine how intake of sugars (other than lactose) and fiber is associated with gut microbiota composition among a study population comprising exclusively breastfeeding, mixed feeding, and nonbreastfeeding Latino infants aged 6 mo enrolled in the Southern California Mother’s Milk Study. They characterized gut microbiota of 105 infants using 16S rRNA sequencing of stool samples and quantified intake of sugars (total, free, and added) and fibers (soluble and insoluble) by averaging three 24-h diet recalls. The authors examined intake of sugars and fibers in relation to gut microbiota composition, as estimated by log-transformed genera relative abundances, using Spearman’s correlation for unadjusted analyses and Poisson regression models for adjusted analyses. Associations with microbiota α-(within-sample) or β-(between-sample) diversities are unknown because they were not reported.

Several genera were differentially abundant with respect to sugar and/or fiber intake. Before adjustment, 4 and 7 genera had positive associations with at least one fiber and one sugar type, respectively, whereas 3 genera had inverse associations with sugars and/or fibers. Of note, Bifidobacterium, an early colonizer of the pediatric gut, was positively associated with total sugar, but inversely associated with free or added sugars. After adjustment, 5 genera were positively associated with free sugar intake: Blautia, Clostridium (Lachnospiracea family), Dorea, Oscillospira, and Parabacteroides. Blautia, Clostridium (Erysipelotrichaceae family), and Dorea were also associated with higher added sugar intake. Lastly, Blautia, Dorea, and Oscillospira had higher relative abundances among infants consuming more fiber. Several of these genera have been associated with early introduction to solid foods (e.g., Blautia) [4], and with obesity (e.g., Blautia and Dorea) [5], consistent with the postulate that microbiota imbalance induced by dietary sugars cause metabolic dysfunction [6].

To the best of our knowledge, no prior human studies have investigated the effects of dietary sugars on infant gut microbiota [7], but experiments have been conducted in mice [8,9]. One study found that mice fed high glucose or fructose diets exhibited decreased relative abundance of Bacteroidetes (phylum), including Muribaculum instestinale, and increased Proteobacteria, including Desulfovibrio vulgaris [8]. Another murine model found that administration of fructose increased proportions of Parasutterella and decreased Intestinimonas [9]. Alas, it is difficult to generalize taxa-specific findings from mice to humans. Fiber-microbiota studies in infants are also lacking, but these effects have been characterized in adults. A systematic review of adult randomized trials highlights the effects of dietary fiber on the gut microbiota, noting that fiber-rich diets typically increase Bifidobacterium and Lactobacillus among other short-chain fatty acid-producing bacterial taxa [10]. Although the design, aims, and exposures/outcomes of prior studies were distinct from those of Mokhtari et al., they corroborate the general hypothesis that sugar and fiber intake affect gut microbiota composition.

Mokhtari et al. should be commended for their study’s strengths, including the research question’s novelty, using three 24-h diet recalls, and conducting their investigation in an underrepresented population at high risk of obesity. However, a few caveats warrant consideration when interpreting their findings. First, although the authors motivate their study by discussing the lack of data on the effects of complementary foods on gut microbiota, their findings are only partially driven by intake of complementary foods because almost half of their participants (49 of 105) were exclusively breastfed and thus not consuming complementary foods. Lactose, the primary sugar in breastmilk, was not included in free or added sugars, so intake of these sugars would be zero for exclusively breastfed infants. Exclusively breastfed infants’ higher exposure to human milk oligosaccharides, which are not classified as nutrients by the Nutrition Data System for Research (NDSR), could explain why Bifidobacterium, which is under strong selective pressure by human milk oligosaccharides [11], was inversely associated with free and added sugars. In general, this design feature (i.e., inclusion of breastfed infants) may have biased their findings, because exclusively breastfed infants do not meet the “positivity” assumption [12], since by definition, their probability of exposure to (nonbreastmilk) sugars and fibers is zero, and thus they should not be included when estimating associations between these nutrients and microbiome outcomes. Future analyses aiming to isolate the effects of sugars or fibers from complementary foods might consider restricting to infants exposed to complementary foods and with comparable exposure to breastmilk and formula.

Another caveat regards the investigators estimation of effects of individual nutrients without accounting for differential intake of other nutrients. In nutritional epidemiology studies, similar to theirs, dietary intake is usually replacement-like, meaning increased intake of a food/nutrient typically results in compensatory reduced intake of other foods/nutrients [13]. As such, it is difficult to know whether the associations observed by Mokhtari et al. were because of greater intake of sugars or fibers, or because of lower intake of other nutrients. Investigating the effects of individual nutrients may have greater validity when accounting for the differential intake of other nutrients. Future observational studies hoping to estimate the isocaloric effects of nutrients on the microbiome might consider using a leave-one-out model, with adjustment for total energy intake and nutrient sources of energy other than those being substituted with the exposure [13].

Other aspects of the study should also be considered when interpreting results. The unique features of microbiome data (e.g., zero-inflated, compositional) can result in spurious findings if the correct analytic approach is not used. This is particularly true for differential abundance (DA) testing, as evidenced by a recent study reporting high discordancy when identifying exposure-associated taxa through different testing methods, and thus suggested prioritizing agreement between at least two DA regression approaches [14]. It would be interesting to see if the findings by Mokhtari et al. are confirmed using DA regression approaches more commonly employed for analyses of zero-inflated, compositional data [14]. Analyzing microbial β-diversity would also be informative to investigate overall community structure. Furthermore, their study was cross-sectional and thus not able to establish temporality, and despite adjustment for confounders, one cannot rule out the influence of unmeasured or residual confounding.

These caveats withstanding, Mokhtari et al. offer an important original insight into potential effects of dietary sugar and fibers on the gut microbiome of Latino/a infants. Although further investigation is warranted, their findings suggest that intake of these nutrients is associated with differential relative abundance of bacterial genera that may be implicated in metabolic health outcomes. Future studies should consider using experimental or longitudinal designs, employing whole-genome shotgun sequencing of the microbiome, integrating other -omics, and connecting changes in microbiota to clinically meaningful metabolic phenotypes.

Author contributions

Both authors prepared, read, and approved the final manuscript.

Conflict of interest

The authors report no conflicts of interest.

Funding

NTM is supported by the National Heart, Lung, and Blood Institute (K01HL141589 and R01HL166473).