Sugar ABC transporter repertoires predict ecological dynamics in gut microbiome communities

TL;DR

This paper shows that bacteria with sugar ABC transporters dominate gut communities when sugar is present in the diet.

Contribution

The study identifies sugar ABC transporter genes as a genomic predictor of microbiome responses to dietary sugar.

Findings

Bacteria encoding sugar ABC transporters outcompete others in monocultures and complex consortia.

Deleting sugar transporter genes in E. coli reduces its ability to compete in gut microbial communities.

Dietary sugar intake correlates with expansion of sugar ABC transporter-positive genera in human microbiomes.

Abstract

The gut microbiome plays a central role in human health, but modern diets and lifestyles alter its composition. Increased sugar consumption is a hallmark of modern diets, yet its impact on the microbiome remains poorly understood. Here, we combine comparative genomics, experiments, and longitudinal human diet-microbiome records to show that the response of the microbiome to dietary sugars is explained by the carriage of sugar ABC (ATP-binding cassette) transporters. Bacteria encoding these transporters exhibit enhanced growth and consistently outcompete others in both monocultures and complex consortia across contexts. Targeted deletion of sugar transporter genes in Escherichia coli, a model gut pathobiont of the expanded Oligo-Mouse-Microbiota (OMM15) consortium, reveals that a specific sugar ABC transporter gene is required to compete, and invade this community. In gnotobiotic mice colonized with the OMM15 consortium, dietary sugar supplementation selectively increases the expansion of sugar ABC transporter-positive bacteria, including E. coli. Paired human diet-microbiome data reveal that intake of dietary sugars significantly correlates with the expansion of sugar ABC transporter-positive genera. Taken together, our work identifies a genomic predictor of microbiome responses to dietary sugars and suggests ways to anticipate major shifts in the abundances of important gut bacteria.

Figures

f](#F1)). As even minor differences in pH may selectively inhibit strains^[26](#R26),[27](#R27)^, we adjusted the pH of 1% GLU medium from 6.30 to 6.80 (matching 0.25% GLU), which significantly rescued *B. longum* growth (*p* = 0.0008, Supplementary Fig. 4), moderately improved *A. muciniphila*, which had shown a significant growth reduction ([Fig. 1f](#F1)), but failed to significantly improve growth in Bacteroidota (Supplementary Fig. 4). Analysis of transporter combinations after pH adjustments revealed that bacteria harboring sugar ABC transporters consistently benefitted from higher gluco

d](#F2)). These additional strains encoded genes for PTS and MFS transporters and at least one complete glycolytic pathway, but, as predicted, grew worse than their sugar ABC transporter-possessing counterparts from the same phylum ([Fig. 2e](#F2), Extended data Fig. 5). Moreover, these ABC-lacking strains did not benefit from elevated glucose concentrations ([Fig. 2f](#F2)), in contrast to their phylum-matched counterparts. Together, these results support a model whereby transporter repertoire, rather than taxonomy, explains growth on simple sugars.

c](#F3), Supplementary Fig. 16). To compare mutant fitness in a community, we separately inoculated WT *E. coli* and its Δ*mglB* and Δ*ptsG1* mutants together with the other OMM^14^ community members. While both Δ*ptsG1* and Δ*mglB* mutants showed reduced expansion relative to WT initially, the Δ*ptsG1* mutant eventually caught up, showing no significant expansion differences to the WT after 24 hours ([Fig. 3d](#F3)), in contrast to Δ*mglB*, which expanded less than both WT and Δ*ptsG1*. We next assayed community invasion fitness by first establishing a OMM^14^ community lacking *E. coli* and

b](#F4)). These findings demonstrate that dietary sugar intake drives substantial restructuring of this model consortium *in vivo*. A known limitation of 16S-based community profiling is that it provides compositional data, obscuring the biological interpretation of observed changes due to negative covariance biases^[34](#R34),[35](#R35)^. To address this, we implemented complementary qPCR to measure changes in density of bacterial strains between days 0 and 14, providing a direct measurement of population dynamics following sugar supplementation. This analysis revealed significant differences

a](#F5)). The largest positive coefficient was obtained for the genus *Blautia*, with an average computed sugar ABC transporter count of 46 genes. Conversely, of typical gut commensals, we found *Prevotella*^[2](#R2),[36](#R36),[37](#R37)^ to be most negatively associated with dietary sugar intake, and it lacks sugar ABC transporters ([Fig. 5b](#F5)). This suggested that sugar ABC transporters were specifically predictive of genus abundance associations with dietary sugar consumption. Thus, we next correlated the average transporter count per genus of all sugar transporter genes combined, as w

Full text

Introduction

Gut microbiome composition is associated with regulation of metabolism, immunity, and overall health^1,2^. Modern lifestyle factors, such as environmental changes, widespread antibiotic use, and changes in diet, including an increase in sugar consumption, are disrupting the human gut microbiome symbiosis, causing lasting multi-generational shifts in microbial composition^1,3–5^. Altered microbiome composition is associated with increasing rates of chronic diseases, including colitis, metabolic syndrome, obesity, type 2 diabetes, and inflammatory bowel disease^4,6,7^. Emerging evidence suggests that Western diets high in sugars, fats and low in fiber can reduce microbial diversity and alter bacterial gene repertoires^8,9^, and dietary sugars can also directly silence gene expression of gut colonization factors in bacterial symbionts^10^. Despite the growing evidence linking sugar consumption to microbiome disruption, the genetic and ecological mechanisms underpinning bacterial responses to dietary sugars remain unclear.

Predicting complex microbial ecosystem responses to changing diets is challenging as they may be nonlinear, context-dependent, and hysteretic, meaning they depend on prior ecological interactions and environmental conditions^11,12^. Advances in sequencing and computational modeling have identified correlations between high-sugar diets and microbiome shifts. In murine models, high sugar intake increased Pseudomonadota abundance while decreasing Bacteroidota^13^, liquid fructose-rich elevated the Bacillota-to-Bacteroidota ratio^14^, whereas a high-fat, high-sucrose diet reduced Bacteroidota and increased Bacillota abundance, which partially depended on host genetic background^15^. Human studies comparing rural African children who consumed high-fiber diets with European children on sugar-rich Western diets showed that fiber increased Bacteroidota diversity, while Western diets raised Bacillota abundance and lowered overall microbiome diversity^16^. Clinically, patients undergoing hematopoietic cell transplantation who consumed high-sugar diets experienced exacerbated antibiotic-induced loss of microbial diversity and expansion of antibiotic-resistant Enterococcus species^3^. While these studies reveal strong associations between sugar intake and microbiome changes, the altered microbial components vary from study to study, and the genetic determinants driving microbial fitness in sugar-rich diets—which could therefore consolidate these observations—remain elusive.

In this study, we linked sugar ABC transporter genes to ecological competitiveness of gut bacteria in vitro, in in vivo models of sugar enriched diets, and in human data. Analysis of thousands of bacterial genomes suggested that bacterial sugar metabolism and transport systems are under independent selection. Leveraging this independent variability in in vitro screens of bacterial growth under varying sugar conditions identified sugar ABC transporters as strongly associated with monoculture growth and community fitness, more so than glycolytic and other transporter genes. Supporting a crucial role in competition, deletion of a single sugar ABC transporter gene in E. coli, unlike other transporter deletions, significantly impaired its fitness in in vitro communities, suggesting that sugar ABC transporters are crucial for gut microbes when hosts consume high sugar diets. Consistent with this, feeding gnotobiotic mice various sugar-enriched diets showed that sugar ABC transporter gene carriage reliably predicted strain expansion. To assess whether this pattern holds in humans, we analyzed paired human diet-microbiome time-series data and confirmed that genera containing sugar ABC transporter-positive strains consistently had the strongest positive associations with sugar intake. These findings establish an outsized importance of sugar ABC transporters as predictive genomic markers for microbial ecosystem responses to dietary sugars, and establishes their crucial role in microbial competition.

Taken together, our work advances genetics-driven frameworks to forecast microbiome dynamics.

Results

Interphylum genomic characterization of sugar transporter repertoires and metabolic pathways in gut bacteria

Our goal is to understand the response of the gut microbiome to dietary sugars. A critical driver of microbial growth is the ability to access and metabolize sugars^17,18^. As core metabolic pathways in bacteria are highly conserved^19,20^, but nutrient transporter systems vary even between related strains^21,22^, we hypothesized that it is the sugar transporters that are most likely to explain the growth responses of gut bacteria to dietary sugars. To investigate this, we analyzed 1,147 bacterial genomes from five major phyla, broadly representing gut bacterial diversity^1^ (Supplementary Fig.1). Our analysis focused on six core genomic features involved in the utilization of simple sugars^22,23^: three sugar transport systems: ATP binding cassette (ABC), Major Facilitator Superfamily (MFS), and phosphotransferase system (PTS); and three central carbon metabolic pathways: Embden-Meyerhof-Parnas (EMP), pentose phosphate pathway (PPP), and Entner-Doudoroff (ED) pathways). Visualization of gene presence on a tree constructed from the analyzed genomes and pairwise mutual information (PMI) analysis of gene presence across phyla revealed only partial overlap between transporter and metabolic pathway modules (Fig. 1a). Among species of the same phylum, we observed considerable variation in transporter and glycolytic pathway presence (except Verrucomicrobiota, represented solely by Akkermansia muciniphila; Extended Data Fig. 1), indicating that sugar metabolism traits are evolutionarily labile.

We next asked whether this genomic variability would translate into functional differences that impact microbial competition. We selected 24 genetically diverse gut bacterial species from the five major phyla for experimental testing (Supplementary Table 1). Among these strains, sugar transporter profiles showed both phylum-level patterns and intra-phylum variability (Fig. 1b, Supplementary Fig. 2), reflecting the full genome data set (Fig. 1a). Pseudomonadota encoded comparable numbers of sugar ABC and PTS transporters. Bacillota showed heterogeneity, with sugar ABC and PTS transporters variably present across species, while MFS transporters were largely absent. Bacteroidota primarily encoded MFS transporters, with some strains also possessing PTS transporters, and consistently lacked sugar ABC transporters. Actinomycetota possessed both sugar ABC and MFS transporters but lacked PTS systems, whereas Akkermansia muciniphila (Verrucomicrobiota) encoded only PTS and MFS transporters. Glycolytic metabolism repertoires also varied by phylum and species (Fig.1c). The EMP pathway was nearly ubiquitous, absent only in Lactococcus lactis (Bacillota) and Actinomycetota species. The PPP pathway was broadly present but absent in A. muciniphila. The ED pathway only occurred in Pseudomonadota. To evaluate functional completeness, three essential enzymes were examined for each pathway (Supplementary Table 2). No genome encoded the full set of representative enzymes for all three pathways. Gene presence patterns varied across species, with some pathways missing specific enzymes and others sharing enzymes between pathways, reflecting the modular and overlapping organization of glycolytic pathways^24^ (Supplementary Fig. 3). This heterogeneity in sugar metabolism genes and transporters did not align with phylogeny, reinforcing that sugar utilization traits are at least partially decoupled from taxonomy and may have been shaped by strain-specific selective pressures. Taken together, our selected strains encode independent variation in sugar metabolism and transport systems.

Growth on simple sugars correlates with sugar ABC transporter repertoires

We next tested the importance of these genomic features for growth on sugars. We cultured each of the 24 bacterial species individually in a modified Brain Heart Infusion (BHI) medium lacking glucose (BHI-NG), and supplemented it with either 0.25% or 1% (w/v) glucose (GLU), reflecting the nutrient variability in gut environments^4,25^. After 24 hours of growth, colony-forming units (CFUs) were quantified and correlated with genes involved in sugar transport and processing (Fig. 1d). Growth correlated most strongly with sugar ABC transporter gene counts (Spearman’s rs = 0.79, p < 0.0001), while MFS transporter gene counts were negatively correlated with growth (rs = −0.56, p = 0.0003), with weaker associations between CFUs and central metabolism gene counts; our results were consistent in both glucose concentrations (Fig. 1d, e and Extended Data Fig. 2).

Next, to assess if increased sugar availability altered growth in a genetic feature-specific manner, we compared 24 h CFU ratios between 1% and 0.25% GLU. We found that some species exhibited reduced growth in 1% GLU (Fig. 1f). As even minor differences in pH may selectively inhibit strains^26,27^, we adjusted the pH of 1% GLU medium from 6.30 to 6.80 (matching 0.25% GLU), which significantly rescued B. longum growth (p = 0.0008, Supplementary Fig. 4), moderately improved A. muciniphila, which had shown a significant growth reduction (Fig. 1f), but failed to significantly improve growth in Bacteroidota (Supplementary Fig. 4). Analysis of transporter combinations after pH adjustments revealed that bacteria harboring sugar ABC transporters consistently benefitted from higher glucose concentration (Fig. 1g).

This observation was also robust to a binary classification of the strains by presence or absence of transporter types; strains encoding sugar ABC transporters exhibited significantly higher growth in both low and high glucose conditions compared to those lacking them, and this was not predicted by glycolytic enzymes (Extended data Fig. 3 and 4). Moreover, multivariable analysis of CFU counts with gene presence/absence at 12, 24, and 48 h showed that presence of ABC sugar transporters had the strongest and most significant association with CFU among all genetic features, except at 48 h in 0.25% GLU (Fig. 2a, Supplementary Fig. 5). We extended our experiments to other simple dietary sugars (Fig. 2b, Supplementary Fig. 6–9); using area under the curve (AUC) as a reliable proxy for bacterial growth among our strains (Fig. 2c), we found that species with sugar ABC transporters grew better on fructose (FRU), sucrose (SUC), and galactose (GAL).

As sugar ABC transporters seemed to accelerate growth, we next tested strains that, unlike other members of their phylum in our species set, lacked sugar ABC transporters, expecting slower growth compared to their relatives from the same phylum (Fig. 2d). These additional strains encoded genes for PTS and MFS transporters and at least one complete glycolytic pathway, but, as predicted, grew worse than their sugar ABC transporter-possessing counterparts from the same phylum (Fig. 2e, Extended data Fig. 5). Moreover, these ABC-lacking strains did not benefit from elevated glucose concentrations (Fig. 2f), in contrast to their phylum-matched counterparts. Together, these results support a model whereby transporter repertoire, rather than taxonomy, explains growth on simple sugars.

Sugar ABC transporters are associated with growth in in vitro communities

We thus hypothesized that sugar ABC transporters drive ecological success in microbial communities under sugar supplementation. To test this, we leveraged an established 15 species model community, the Oligo-Mouse-Microbiota (OMM^15^), a minimal consortium representing five dominant murine gut phyla^28–30^, whose genomes varied in transporter repertoires (Supplementary Fig. 10, Supplementary Table 3). While two strains (Acutalibacter muris and Streptococcus danieliae) did not grow on BHI-NG because they require blood-based and Tryptic Soy Broth media, respectively^31^, among the remaining strains, those harboring sugar ABC transporters showed enhanced monoculture growth across GLU concentrations (Supplementary Fig. 11) and other simple sugars (Fig. 3a, Supplementary Figs. 12–14), consistent with the human strain library.

To evaluate the role of sugar ABC transporter in microbial community competition, we first inoculated approximately 10^3^ ± 0.3 CFUs/ml of each of the 15 strains simultaneously into BHI-NG medium supplemented with either 0.25% GLU, FRU, or SUC. We monitored species’ density changes over a 72-hour period using quantitative PCR (qPCR) with primers specific to each strain’s 16S rRNA gene, defining a fold change from 0 to 72 h of greater than 1 as growth within the community (Fig. 3b). All sugar ABC transporter carriers expanded: most notably, E. coli and Enterococcus faecalis grew rapidly during the earlier time points (8 h) with other following around 12 h (Supplementary Fig. 15). In contrast, bacteria lacking sugar ABC transporters did not expand in the community despite their ability to expand in monocultures. An exception was Limosilactobacillus reuteri, which possesses PTS transporters and showed modest growth in the community settings. A. muris and S. danieliae, which failed to grow in monoculture, showed weak expansion in the community (Fig. 3b and Supplementary Fig. 15). Taken together, sugar ABC transporters were predictive of growth in different simple sugar conditions in a defined in vitro community of diverse murine gut bacterial strains.

A specific sugar ABC transporter is required for community invasion by E. coli

Next, to directly test the necessity of sugar ABC transporters for competitiveness, we systematically modified the transporter repertoire of E. coli MT1B1 of the OMM^15^ community (Supplementary Table 4). In 0.25% glucose, deletions of mglB (a sugar ABC transporter subunit) and ptsG1 (a PTS transporter subunit) resulted in the strongest growth defects relative to the wild type (WT, Fig. 3c, Supplementary Fig. 16). To compare mutant fitness in a community, we separately inoculated WT E. coli and its ΔmglB and ΔptsG1 mutants together with the other OMM^14^ community members. While both ΔptsG1 and ΔmglB mutants showed reduced expansion relative to WT initially, the ΔptsG1 mutant eventually caught up, showing no significant expansion differences to the WT after 24 hours (Fig. 3d), in contrast to ΔmglB, which expanded less than both WT and ΔptsG1. We next assayed community invasion fitness by first establishing a OMM^14^ community lacking E. coli and introducing the E. coli WT or the mutant strains after 12 h, i.e. when glucose concentrations reached ≈50% of the initial concentrations (Fig. 3e, Supplementary Fig. 17). The ΔmglB mutant showed significantly lower competitiveness in this assay than both WT and ΔptsG1 strains at 36 hours and 60 hours (Fig. 3f), highlighting the importance of a single sugar ABC transporter for community invasion.

Sugar ABC transporters are predictive of sugar-induced expansion in vivo

Our in vitro data suggest that sugar ABC transporters can be used to forecast microbiome dynamics under sugar enriched diets. To test this, we colonized germ-free mice with the OMM^15^ community and supplemented their drinking water with 10% GLU, FRU, or SUC for 14 days, modeling typical sugar concentrations in hospital or commercial beverages^6,32^. Both the control and sugar-fed groups were maintained under otherwise identical conditions in the same room, housed in HEPA-filtered ISOcages in a gnotobiotic facility, with ad libitum access to their drinking water. As reported previously, sugar-fed mice consumed more water (data not shown), gained weight, and exhibited reduced colon length and cecal weight compared to controls (Supplementary Figs. 18–20)^6,33^. To assess the impact of dietary sugars on the gut microbiota, we collected fecal samples at days 0 and 14 and profiled community compositions with 16S rRNA gene sequencing. Principal coordinates analysis (PCoA) showed that by day 14, the microbiota of GLU- and FRU-treated mice had diverged significantly from that of control mice as well as from their own baseline microbiota (Fig. 4a and Supplementary Fig. 21).

Analysis of relative abundances at day 14 between sugar treated mice and control revealed a significant relative abundance increase of A. muris, Bifidobacterium animalis, Turicibacter muris, and A. muciniphila in sugar-treated groups, while Blautia coccoides showed a significant decrease (Fig. 4b). These findings demonstrate that dietary sugar intake drives substantial restructuring of this model consortium in vivo. A known limitation of 16S-based community profiling is that it provides compositional data, obscuring the biological interpretation of observed changes due to negative covariance biases^34,35^. To address this, we implemented complementary qPCR to measure changes in density of bacterial strains between days 0 and 14, providing a direct measurement of population dynamics following sugar supplementation. This analysis revealed significant differences between control and sugar-treated mice (Supplementary Fig. 22), including increases of sugar ABC transporter-positive B. animalis and A. muris in all sugar-supplemented diets. Consistent with this, individual strain dynamics from day 0 to day 14 within each mouse (Fig. 4c) revealed significantly greater expansion of most sugar ABC transporter-positive strains in sugar-fed mice relative to mice on control diets, but none of the strains without sugar ABC transporters showed sugar-induced increased growth. Combining strains based on their transporter combinations confirmed that bacteria with sugar ABC transporters, exclusively or with other transporters, expanded more in vivo upon sugar supplementation (Fig. 4d, Extended data Fig. 6). We thus identified sugar ABC transporters as crucial drivers of bacterial competitiveness in high sugar diets.

Taxa associated with dietary sugar intake in HCT patients harbor abundant sugar ABC transporters

We next asked if sugar ABC transporters could be used to forecast human microbiome responses to dietary sugars. To test this, we analyzed data from fecal samples of patients undergoing HCT, in a cohort with high-resolution dietary intake data^3^. This data set comprises per meal macronutrient intake, including sugars, as well as food-item level resolution revealing the frequent consumption of sugar-sweetened beverages (Supplementary Table 5). For each microbiome sample, we modeled the association between total sugar intake in the prior two days^3,5^ and the centered log ratio (CLR)-transformed relative abundances of 91 bacterial genera using Bayesian regression models (Extended data Fig. 7; n = 1,009 samples from 158 patients). As this amplicon-based microbiome data set does not directly reveal transporter gene abundances, we estimated them by computing the average count of each sugar transporter type from up to 100 randomly sampled complete genomes per genus from the NCBI genome database. This revealed a consistent pattern: genera with confidently positive sugar intake associations exhibited a higher total number of sugar transporter genes compared to genera negatively associated with sugar intake (Fig. 5a). The largest positive coefficient was obtained for the genus Blautia, with an average computed sugar ABC transporter count of 46 genes. Conversely, of typical gut commensals, we found Prevotella^2,36,37^ to be most negatively associated with dietary sugar intake, and it lacks sugar ABC transporters (Fig. 5b). This suggested that sugar ABC transporters were specifically predictive of genus abundance associations with dietary sugar consumption. Thus, we next correlated the average transporter count per genus of all sugar transporter genes combined, as well as each transporter type individually, with the dietary sugar association coefficients for the corresponding genera. Both Spearman rank correlations and Bayesian linear regressions showed that total sugar transporter counts and specifically the sugar ABC transporter counts (Fig. 5c) were significantly and positively correlated with the dietary sugar association coefficients. In contrast, PTS and MFS sugar transporter counts did not show this relationship (Fig. 5c). These findings establish sugar ABC transporter repertoires as powerful predictors of human microbiome responses to dietary sugars.

Discussion

Our study shows that transport systems for simple sugars are key determinants in microbial competition, and that they are more informative of growth in complex gut communities than central metabolism gene repertoires. Moreover, we find that a single genomic feature—possession of sugar ABC transporters—has outsized predictive power of complex gut microbiome dynamics, including the ability to invade a resident community and forecast human microbiome dynamics. This predictive capacity spans in vitro assays, in vivo supplementation with dietary sugars, and human microbiome data, establishing a single genomic trait as a key determinant of dietary sugar-mediated community dynamics. Because our findings were derived from diverse bacterial strains with variable central metabolism gene repertoires, we were able to rank sugar ABC transporters as the dominant fitness-driving feature in various models of gut microbial competition. This is consistent with orthogonal prior work showing that deletion of ABC transporters in uropathogenic E. coli reduces growth rates in infection models^21^. More generally, the usage of ABC and PTS transport systems varies between species and can be regulated by substrate availability^21,38–42^. For example, in some E. coli strains, sugar ABC transporters are highly expressed under low-glucose conditions^38,39,43^, whereas in Pseudomonas stutzeri, they are used in both low and high sugar environments^44^.

How is the competitive advantage conferred by sugar ABC transporters during sugar supplementation? One possibility is that increased sugar availability concentrates competition within the microbial community, accentuating the fitness effects of specific transport systems^4,18^. As such, the high substrate affinity of sugar ABC transporters, combined with their capacity for efficient nutrient acquisition^45,46^ across sugar concentrations^38,43,47^ may explain why we find them singularly predictive of community dynamics. Additionally, different transporter systems may also contribute at distinct stages of community dynamics in sugar-supplemented environments^47,48^. Our knockout experiments show that while both PTS and ABC transporters support early bacterial growth, only the ABC transporter is essential for sustained expansion and successful invasion within the community context (Fig. 3d and 3f). This suggests that although multiple nutrient uptake systems influence microbial fitness^38,39,47,48^, sugar ABC transporters provide a persistent advantage.

Moreover, in HCT patients, whose microbiomes experience extreme antibiotic- and treatment-related microbiome disruption, which may sensitize them to effects of dietary sugars^3,49,50^, we found that genera with higher sugar ABC transporter counts had positive association with dietary sugar intake, consistent with sugar-induced expansion of such taxa. Interestingly, while not all genera with low transporter gene content declined, we observed a particularly strong negative association of Prevotella with dietary sugar intake, a genus that is typically lost in industrialized populations as a result of low-fiber diets and whose members lack sugar ABC transporters^2,16,36,37,51^. These findings demonstrate that sugar ABC transporter content is a genomic correlate of microbiome dynamics in response dietary sugars that extends to humans.

Taken together, the key outcome of this study is that sugar ABC transporters—a single genomic trait—consistently predict bacterial growth in sugar defined environments, across monoculture conditions, community invasion, in vivo expansion, and human microbiome responses to dietary sugars. As such, our work supports prioritization of functional traits over taxonomy in explaining microbiome dynamics^52–54^. The outsized predictive power of a single genomic feature described here is particularly notable given the complexity of microbial ecosystems, which are influenced by factors such as osmolarity, pH levels^26,27^, secondary metabolite production^28,55^ and cross-feeding^56^. As nutrient blocking is the prevailing ecological explanation for colonization resistance against pathogens^29,57^, we speculate that in diets defined by simple sugar intake, sugar ABC transporters are crucial mediators of this fundamental microbiome function. Moreover, our results contribute to explaining the increase of Pseudomonadota, Bacillota, and pathobionts observed in dysbiosis and as a consequence of modern dietary shifts^3,13–16,36,37,51,58^, which often involve elevated consumption of refined sugars alongside reduced fiber intake. As global sugar intake increases, these findings demonstrate that specific gene content can be used to anticipate microbiome shifts, offering a predictive link between microbial gene content and microbiome responses to diet-driven ecological restructuring.

Methods

Strains and media

All bacterial strains used in this study are listed in Supplementary Tables 1 and 4.

Starter cultures of human isolates were grown in Brain Heart Infusion broth without glucose (BHI-NG) (Bioworld), supplemented with 0.25% glucose (w/v) (Sigma-Aldrich). Starter cultures of the OMM^15^ community were also cultured in BHI medium under the same conditions, except A. muris and S. danieliae, which were cultured in Columbia agar medium (Remel) supplemented with 5% defibrinated sheep blood (Fisher Scientific). A. muciniphila and B. caecimuris were cultured in BHI medium supplemented with 0.25% GLU and 0.2% mucin (w/v) (Sigma-Aldrich). All bacteria were cultivated anaerobically at 37°C inside a Coy Model 2000 incubator in a type B anaerobic chamber (N2/CO2/H2, 95%/2.5%/5%) (Coy Laboratory Products, Inc., Grass Lake, MI, USA). Selective media for cloning purposes were prepared using LB broth or LB-agar, supplemented with antibiotics at the following final concentrations: 25 μg/mL chloramphenicol (Sigma-Aldrich) and 100 μg/mL ampicillin (Sigma-Aldrich).

Cross-Phylum Genome Analysis of Sugar Transporters, Glycolytic Enzymes, and Pathways

Gut microbiome species were obtained from the BacDive database. Species were ranked by the number of completed genomes available in the NCBI database as of May 12, 2025. For each phylum, the top 20 species were selected for analysis, with A. muciniphila (the sole representative of Verrucomicrobiota) included. For each species, all available complete genomes were downloaded; if more than 20 genomes were available for a species, a random subset of 20 genomes was selected. To determine the counts of sugar transporters and glycolysis enzymes in bacterial genomes, genome annotation files were downloaded from NCBI. The genomes of R. intestinalis, K michiganensis AD9, and E. coli AD24 were annotated using PGAP^59,60^ provided by NCBI and these annotation files were used for genome searches. For the screening of sugar transporters, genome annotations were searched using the keywords “sugar” and “transporter”. Identified sugar transporters were then categorized into subgroups, including sugar ABC transporters, PTS transporters, and MFS transporters; transporter categories were not resolved by sugar specificity. Key glycolysis enzymes were selected from three primary glycolytic pathways: Embden-Meyerhof-Parnas (EMP), Pentose Phosphate Pathway (PPP), and Entner-Doudoroff (ED). For the screening of glycolytic enzymes, genome annotations were searched using the specified gene names (Supplementary Table 2). For hexokinase, alternative search terms such as “sugar kinase,” “glucokinase,” and “ROK family kinase” were employed. These selections provided comprehensive coverage of the different glycolytic pathways. For KEGG pathway analysis, KOALA^61^ was used to assign KEGG Orthology (KO) terms to proteins in deduplicated bacterial protein files. The annotated KOs were then used to reconstruct KEGG pathways in each genome via KEGG Mapper. Only complete glycolysis pathways were retained and considered functional for this study.

CFU analysis

For CFU experiments, 5 mL of BHI-NG medium supplemented with 0.25% and 1% glucose (w/v) was inoculated with each bacterium to a final concentration of 10^5^ ± 0.3 CFU/mL, and cells were grown at 37 °C under anaerobic conditions without agitation. At various time points, cells were transferred to a 96-well flat bottom plate (Corning Costar), and serial dilutions ranging from 10^−1^ to 10^−7^ were performed in deoxygenated PBS. From these dilutions, 20 μL of sample was transferred to a BHI agar plate. Plates were incubated at 37 °C under anaerobic conditions, and CFUs were counted. Statistical analysis of gene presence/absence as predictor of CFUs was conducted with a linear mixed effects model using varying intercepts for biological repeats; the model was implemented in Python using the statsmodels library, with the following model equation for the regression of gene presence as predictors of CFUs at a given sugar concentration and time point: (CFU | time point, sugar concentration) ~ ABC + PTS + MFS + RPI + PGM + KDGP + (1 | uniqex), where ‘uniqex’ represents the index of the biological repeat. Transporter gene presence was represented using binary indicators for ABC, PTS, and MFS. Glycolytic pathway genes were selected based on prevalence, retaining only those present in more than 25% but fewer than 75% of species. Binary indicators RPI, PGM, and KDGP correspond to ribose-5-phosphate isomerase, phosphoglucomutase, and 2-keto-3-deoxy-6-phosphogluconate aldolase, respectively.

pH Experiments

For pH adjustment experiments (Supplementary Fig. 4), pH was measured using a glass double-junction electrode (Orion^™^ Star A111, Thermo Fisher Scientific), and media pH was titrated using 1 M NaOH to match across sugar concentrations.

Growth measurements

For growth assays, BHI-NG medium was supplemented with 0.25% or 1% (w/v) glucose, fructose, sucrose, or galactose and inoculated with each bacterium to a final concentration of 10^5^ ± 0.3 CFU/mL in 250 μL of medium. Cultures were grown in a 96-well flat-bottom plate (Corning Costar) at 37°C under anaerobic conditions without agitation, using a Tecan Infinite Mplex plate reader. Prior to each optical density measurement, a 10-second double-orbital shaking step was performed. Optical density at 600 nm (OD_600_) was recorded every 30 minutes for 72 hours.

E. coli MT1B1 mutants’ construction

All mutants generated in this study are reported in Supplementary table 5. Deletion mutants were generated as described previously in^62^. Briefly, plasmid pKD46 was introduced into wild-type E. coli MT1B1 via electroporation. A chloramphenicol resistance cassette (cat) was PCR-amplified from plasmid pKD3 using primers containing ~100 bp homology arms flanking the target gene. The resulting PCR product was electroporated into E. coli MT1B1 harboring pKD46. Successful recombinants were selected on chloramphenicol-containing media, and gene deletions were confirmed by colony PCR. Plasmids and primers used are listed in Supplementary Tables 6 and 7.

In vitro community experiments

OMM^15^ starter cultures were grown in their respective media as described previously. For community experiments, 500 mL of BHI-NG medium supplemented with 0.25% glucose (w/v) was inoculated with each bacterium to a final concentration of 10^3^ ± 0.3 CFU/mL. The cultures were grown at 37°C under anaerobic conditions without agitation. At defined time points, 20 mL aliquots were collected, centrifuged, and the resulting pellets were stored at −80 °C for subsequent processing and quantitative PCR analysis.

Glucose measurements

To quantify free glucose concentrations in OMM^14^ (excluding E. coli) and OMM^15^ communities over time, cultures were set up as described above. At defined time points 1 ml samples were collected, centrifuged at 8,000 × g for 5 minutes, and the supernatants were filtered (0.22 μm). Glucose levels in the cleared supernatants were measured using a glucose colorimetric detection kit (Invitrogen, EIAGLUC), following the manufacturer’s protocol. Absorbance was read at 560 nm using a plate reader (Tecan Infinite Mplex), and glucose concentrations were calculated based on a standard curve generated in parallel.

Invasion assay

The OMM^14^ community excluding E. coli was cultured as described above in BHI-NG medium supplemented with 0.25% GLU under anaerobic conditions at 37°C. After 12 hours, when glucose concentrations began to decline, wild-type E. coli MT1B1 or its respective deletion mutants were inoculated separately into established OMM^14^ cultures at a final concentration of approximately 10^3^ ± 0.3 CFU/mL. Cultures were maintained under the same anaerobic conditions, and 20 mL samples were collected at designated intervals, immediately stored at −80°C, and subsequently analyzed by quantitative PCR to determine bacterial population dynamics.

DNA extraction and Quantitative PCR of bacterial 16S rRNA genes

DNA was extracted from in vitro community and invasion assay using the Fecal/soil microbe MiniPrep Kit (Zymo Research, D6010) following the manufacturer’s protocol. Quantitative PCR was performed as described previously^30^, with 1 μL of the extracted DNA used for qPCR reactions. Strain-specific 16S rRNA primers (Supplementary Table 8) at 10 μM concentration were employed, with SYBR FAST Roche LightCycler 480 2X qPCR Master Mix. The qPCR cycling conditions included a pre-incubation step at 95°C for 3 minutes, followed by amplification with 3 s at 95°C, 10 s at 60°C, and 20 s at 72°C, and a elting curve of 5 s at 95°C and 1 min at 65°C for a total of 40 cycles. Fold changes in the density of each bacterial strain across time points or experimental conditions were calculated directly from normalized qPCR Ct values. Absolute 16S rRNA gene copy numbers were not determined, as our analysis focuses on relative changes within the same strain. Ct values were normalized to the volume of culture (in vitro) or fecal weight (in vivo) for fold change calculations. The final fold change was determined using the equation

$$[eqn]$$ $$[eqn]$$

In vivo experiments

Mice.

Mice were kept on a 12h light/dark cycle with ad libido access to autoclaved rodent chow and drinking water. Offspring of germ-free C57BL/6Tac mice housed in sterile isolators were used for all experiments.

Animal care and experimentation were consistent with NIH guidelines and approved by the Institutional Animal Care and Use Committee (IACUC) at New York University School of Medicine

OMM15 colonization.

T. muris and M. intestinale, strains were purchased from DSMZ, while the remaining strains were generously provided by Xiaomin Xiao and Ken Cadwell (NYU School of Medicine/University of Pennsylvania). Strains were cultured under anaerobic conditions as described above, combined at equal CFU ratios 10^8^ ± 0.3, and the resulting mixture was stored in 20% glycerol at −80°C for further use.

Eight-week-old male and female germ-free C57BL/6Tac mice, housed in sterile isolators, were used for colonization. On the day of colonization, the bacterial mixture was thawed and kept on ice until inoculation. Mice were inoculated via oral gavage and enema using a 1 mL syringe fitted with an 20ga, 38mm flexible plastic rodent feeding tube (Instech, Cat# 50-475-764). A second dose of the bacterial mixture was administered on day 2 post-initial inoculation.

To allow for recovery and minimize stress, breeders were set up on day 7 following the secondary inoculation. Vertically colonized mice were used for all experiments.

In vivo high sugar challenges.

Mice were randomly assigned to treatment groups and provided with either 10% glucose (Sigma-Aldrich), 10% fructose (Sigma-Aldrich), or 10% sucrose (Sigma-Aldrich) dissolved in filtered drinking water. Control mice received filtered drinking water without supplementation. All sugar solutions and control water were sterilized using a SteriCup 0.22 μm filter (Millipore) before administration.

Body weight and fluid intake were monitored throughout the experiment, and fluids were replaced every three days to maintain freshness and cleanliness. Fresh fecal samples were collected at baseline (day 0) and on day 14, immediately stored on dry ice, and subsequently kept at −80°C for long-term storage. Final body weight measurements were recorded on day of sacrifice (day 14).

Murine fecal DNA Extractions.

Fecal weights were recorded prior to DNA extraction. DNA was extracted from fecal samples using the Quick-DNA Fecal/Soil Microbe MiniPrep Kit (Zymo Research, D6010) following the manufacturer’s protocol.

Mouse intestine dissections.

Mice were sacrificed by cervical dislocation, and the entire intestinal tract, spanning from the gastroduodenal junction to the rectum, was carefully removed as an intact piece. The intestine was divided into segments as follows: Small intestine: Defined from the gastroduodenal junction to the ileocecal junction; Cecum: Isolated as a distinct segment; Colon: Defined as the section extending from 1 cm distal to the cecal-colonic junction to 1 cm proximal to the anus. The lengths of each segment were measured by aligning them along a ruler on a flat surface to ensure accuracy.

Bayesian models of sugar-genus associations and genome sugar transporter counts

The sequencing protocol for the 16S rRNA gene sequencing for this dataset has been reported previously^3^. Assigned taxonomy were aggregated at the genus level, discarding amplicon sequence variants without genus identification. A relative abundance and prevalence threshold was then set to keep genera for analysis that were present in at least 10% of samples at relative abundance 0.01%, or that were present in any sample at 10% relative abundance or more. Of these 117 genera present after filtering, 91 had complete genomes on NCBI. To count sugar transporters of each type, we downloaded up to 100 randomly sampled genomes from each genus and counted the occurrence of “ABC”, “MFS”, and “PTS” in the feature products where “sugar” and “transporter” were already present. We reported means of these counts to approximate the count of sugar transporters of each type for each genus. “All sugar transporters” is a sum of these means.

Sugar-genus models. The associations of the 91 genera with sugar intake in the prior 2 days were quantified with individual Bayesian univariate models with the following definition,

$$[eqn]$$ $$[eqn]$$ $$[eqn]$$ $$[eqn]$$ $$[eqn]$$

for sample i =1,009 for each of the 91 genera. The relative abundance matrix of samples and genera used in the analysis was centered-log ratio (CLR)-transformed^63^ the predictors and outcome were standardized, and a Hamiltonian Monte Carlo sampler was used to extract 4000 samples (across 4 chains) from the posterior distributions (R v4.3.2, rstan v2.32.6, Stan v2.32.2). Posterior distributions for the β parameter associated with the sugar predictor were summarized by plotting the mean value and the 66% and 95% credible intervals.

The relationship between sugar transporter counts and each genus’s association with sugar was investigated in two ways. Firstly, for each sugar transporter type (and all sugar transporters), a Spearman rank correlation was performed between the mean count of each sugar transporter type and the mean values from the sugar parameter of each Bayesian sugar-genus model. Secondly, a Bayesian linear regression was visualized following the model definition,

$$[eqn]$$ $$[eqn]$$ $$[eqn]$$ $$[eqn]$$ $$[eqn]$$

for j = 91 genera. The sugar parameters existed on the z-score scale already, and sugar transporter count was normalized by its maximum before sampling. For each genus, μj was simulated over the range of sugar transporter values using 4000 posterior samples (1000 × 4 chains), and the mean of μj was plotted with a 95% credible interval.

HCT patient dataset

The previously published dataset^3^ comprised 158 patients undergoing HCT, 1009 longitudinally collected fecal samples on which microbiome profiling was conducted with 16S rRNA gene sequencing, and 9419 meals with dry weight in grams of each food item consumed (as indicated in quarter-portions by the patient and verified by a nurse). We used the per-food-item data to sum up the diet (grams of each food item) of each patient every day, and matched this dietary information to microbiome samples by taking the mean of the day sums from the 2 days prior to each microbiome sample (e.g., a sample collected on day 3 is associated with the mean of the diet sums of day 1 and day 2). Macronutrient composition of each food item was collected from the USDA FNDDS database, where each of our food items represented the highest-resolution 8-digit food code. We could then calculate mean sugar intake in grams over the prior 2 days for every microbiome sample.

Data analysis and figures

Data was analyzed using Graphpad Prism 9, python (v.3.8.9) and R (v.4.1.2). Figures were generated using Adobe Illustrator CC (Adobe Inc.). The statistical analysis varied for different datasets and details on the statistical methods are reported in the figure legends.

Supplementary Material

Supplementary Files

This is a list of supplementary files associated with this preprint. Click to download.

• Suppinfo.pdf
• Extended gures.pdf