Intestinal persistence of Bifidobacterium infantis is determined by interaction of host genetics and antibiotic exposure

Abstract Probiotics have gained significant attention as a potential strategy to improve health by modulating host–microbe interactions, particularly in situations where the normal microbiota has been disrupted. However, evidence regarding their efficacy has been inconsistent, with considerable interindividual variability in response. We aimed to explore whether a common genetic variant that affects the production of mucosal α(1,2)-fucosylated glycans, present in around 20% of the population, could explain the observed interpersonal differences in the persistence of commonly used probiotics. Using a mouse model with varying α(1,2)-fucosylated glycans secretion (Fut2WT or Fut2KO), we examined the abundance and persistence of Bifidobacterium strains (infantis, breve, and bifidum). We observed significant differences in baseline gut microbiota characteristics between Fut2WT and Fut2KO littermates, with Fut2WT mice exhibiting enrichment of species able to utilize α(1,2)-fucosylated glycans. Following antibiotic exposure, only Fut2WT animals showed persistent engraftment of Bifidobacterium infantis, a strain able to internalize α(1,2)-fucosylated glycans, whereas B. breve and B. bifidum, which cannot internalize α(1,2)-fucosylated glycans, did not exhibit this difference. In mice with an intact commensal microbiota, the relationship between secretor status and B. infantis persistence was reversed, with Fut2KO animals showing greater persistence compared to Fut2WT. Our findings suggest that the interplay between a common genetic variation and antibiotic exposure plays a crucial role in determining the dynamics of B. infantis in the recipient gut, which could potentially contribute to the observed variation in response to this commonly used probiotic species.


Introduction
Host-microbiome interactions play a pivotal role in shaping human physiology.The intestinal microbiome in particular is an important regulator of innate and adaptive immunity [1], metabolic control [2], the central nervous system [3], as well as contributing to energy and nutrient harvest [4], and suppressing pathogen proliferation [5].Given the association between disruption of the commensal gut microbiota and adverse outcomes, there is significant interest in approaches that facilitate its restoration following perturbation.Among the most well-established of these approaches is the ingestion of viable commensal bacteria in the form of probiotics.
Probiotics can be defined as "live microorganisms, which when administered in adequate amounts, confer a health benefit on the host" [6].Most commonly, these take the form of individual strains or multistrain consortia, of well-characterized commensal bacteria, prepared either as liquid suspensions or in freeze-dried capsules.The principal concept underlying the use of probiotics is that the introduction of live bacteria can re-establish physiological homeostasis by modifying the composition or behaviour of the gut microbiota or by directly providing regulatory cues to the host.Despite substantial evidence supporting the efficacy of probiotics in principle, their use remains poorly supported by empirical data in many physiological or health contexts [7,8].Further, substantial interindividual variance in probiotic persistence has been noted, in part explained by variation in colonization resistance by the microbiome [9,10].Consequently, the global probiotics market, which is projected to reach USD 73.9 billion by 2030 [11], is dominated by direct-to-consumer sales, with little or no consideration is given to recipient traits that might substantially inf luence probiotic efficacy.
Various mechanisms enable the human gut to regulate commensal microbiota composition.One of the principal mechanisms involves the secretion of specific types of sugars that are utilized by beneficial microbial species.Many mucosal constituents and secreted factors are decorated with glycans (oligosaccharides), which are added by a diverse family of glycosyltransferase enzymes [12].Of these, the FUT2 gene encodes a galactoside α(1,2)fucosyltransferase, which adds an L-fucose monosaccharide to nonreducing end Gal residues to form Fucα1-2Gal-O-R glycans, termed the H antigens [13,14].Expressed by multiple mucosal epithelial cell types, this H antigen is a highly versatile structure that can be further modified to form many other important glycans, including the AB blood group glycans.Because FUT2 controls the nature of the various α(1,2)-fucosylated glycans secreted by mucosal surfaces, it is commonly referred to as the "secretor" gene [13].
Across the human population, multiple nonsense singlenucleotide polymorphisms (SNPs) are found within the FUT2 gene [15], leading to a "nonsecretor" phenotype.The nonsecretor phenotype, like the AB blood groups, is one of the more common functional mutations maintained in the population, with approximately one-fifth of people carrying homozygous loss-of-function FUT2 genes [15].This high carriage of loss-offunction mutations is likely a result of positive selection from altered susceptibility to infections by certain bacterial and viral pathogens [16].However, as fucosylated glycans are an important nutrient source for gut microbes, their absence in nonsecretors has been shown to inf luence the composition of commensal microorganisms [16,17].
The intact commensal microbiota of secretor individuals is likely to be enriched for glycan-utilizing bacteria.In contrast, depletion of commensal taxa, for example, through antibiotic exposure, can provide a selective advantage to exogenous glycan utilizers that are absent in nonsecretors [9,18].Probiotic preparations typically contain Bifidobacteria (Bifidobacterium adolescentis, animalis, bifidum, breve, and longum) and/or Lactobacilli (Lactobacillus acidophilus, casei, fermentum, gasseri, johnsonii, paracasei, plantarum, rhamnosus, and salivarius).Both genera include species that encode the specific glycoside hydrolases (GHs), GH29, GH95, and GH151, which can utilize the H antigen.However, both genera also include species without this glycoside hydrolase capacity.Therefore, the ability of a probiotic to colonize and persist in an individual may depend on the presence of secreted glycans and the ability of the introduced bacterial strain to utilize them.This is supported by studies identifying increased persistence of such glycan-utilizing species when supplemented with exogenous oligosaccharides [19,20].
We hypothesized that the interplay between secretor status, the glycan utilization ability of the probiotic strain, and the presence of a disrupted commensal microbiota due to antibiotic exposure would collectively inf luence the abundance and persistence of probiotic populations in the gut.To test this hypothesis, we introduced probiotic Bifidobacterium strains into a murine model of secretor/nonsecretor status, with or without prior antibiotic depletion of commensal microbiota.

Materials and methods
Details of reagent catalogue numbers and resource links are provided in Supplementary Table S1.

Establishment of a Fut2 knockout mice
A Fut2 KO mouse line was developed using CRISPR/Cas9 technology in C57BL/6 mice (IMSR_JAX:000664) by South Australian Genome Editing (SAGE).Brief ly, a 1230-bp region of the Fut2 exon region was excised using targeted CRISPR guide sequences.Gene knock out was confirmed by Sanger sequencing and phenotype confirmed by α(1,2)-fucosylated glycan staining of intestinal biopsies using Ulex Europaeus lectin 1 (Supplementary Fig. S1), as described previously [21].Littermate Fut2 WT and Fut2 KO mice (6 weeks of age, gender-matched) were obtained by mating heterozygous male and female mice originating from F1 heterozygotes.

Breeding and housing
All mice were bred and maintained under specific and opportunistic pathogen-free (SPF) conditions at 22 • C ± 2 • C, under a 12-h light-dark cycle, at the South Australian Health and Medical Research Institute (SAHMRI).All mice were housed in individually ventilated cages, fed an identical diet (Teklad Global 18% Rodent Protein Diet, Envigo, Huntingdon, UK), maintained under the Federation of European Laboratory Animal Science Associations (FELASA) standards, and routinely screened using an SNP genotyping panel.
Heterozygous × heterozygous breeding was performed to allow for Fut2 KO and Fut2 WT littermates, while also standardizing effects of Fut2 that occur through vertical transmission.Fut2 KO , Fut2 HET , and Fut2 WT littermates were cohoused from birth until weaning (∼3 weeks), where they were genotyped by PCR amplicon melt curve using primers targeting the outer and inner regions of the Fut2 gene.Fut2 KO and Fut2 WT mice separated into cages after weaning based on sex and Fut2 genotype (Fig. 1A).No experiments were performed on Fut2 HET mice.In all experiments, 6-week old, age-and sex-matched mice were used.Each experimental group consisted of at least four cages to control for cage effects.Given the heterogeneous nature of the gut microbiome, each mouse was considered as a biological replicate rather than a technical replicate, even within cohoused littermates.

Antibiotic treatment
A cocktail of ampicillin (1 g/L, Sigma-Aldrich) and neomycin (0.5 g/L, Sigma-Aldrich) was provided to mice via drinking water for 7 days.Water bottle volume and mouse weight were monitored to assess water intake.Antibiotic activity was confirmed by qPCR targeting 16S rRNA gene (Supplementary Table S2) of faecal samples at day 7.

Probiotic supplementation
Mice received 5 × 10 7 colony-forming units (CFU)/g of mouse of either B. infantis, B. bifidum, or B. breve daily for 5 days via oral gavage.A starting gavage concentration of 5 × 10 9 CFU/ml in phosphate-buffered saline (PBS) was prepared daily from fresh overnight cultures.The dose was selected based on previous reports of safety, persistence, and immune modulation capability [22,23].
with glycan transporters, and B. breve JCM 1192 encodes only the GH95 family.

Faecal and tissue collection
At least two faecal pellets were collected from separated mice at the beginning of the light phase, unless specified otherwise.For intestinal tissue collection, mice were sacrificed by CO 2 asphyxiation and laparotomy was immediately performed using a vertical midline incision.Once the digestive tract was exposed, separate dissection tools were used to dissect tissue into four parts: the proximal small intestine; distal small intestine; caecum; and large intestine.For small and large intestine tissue segments, the luminal content was collected by instilling sterile PBS using a syringe barrel and the f lushed mucosal tissue was collected into separate tubes.All collected faecal samples, organs, and luminal contents were immediately frozen on dry ice and stored at −80 • C until further processing.

DNA extraction of faecal samples
Faecal pellets were weighed, and 25-mg samples (±10 mg) were resuspended in 300 μl of cold PBS (pH 7.2) by vortexing and pelleted by centrifugation at 10 000 × g for 10 min at 4 • C. Microbial DNA was extracted from faecal samples using the PowerLyzer PowerSoil DNA Isolation Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions as described previously [28].

DNA extraction on mucosal tissue samples
Mucosal tissue from the proximal small intestine, distal small intestine, and large intestine were semi defrosted, and 3 cm was removed from the tissue centre using sterile scalpel.The dissected tissues were cut open longitudinally and mixed with 750-μl Pow-erSoil bead solution and 60 μl solution C1 in a PowerSoil bead tube.The bead tube was then incubated at 65 • C for 10 min prior to bead beating.The subsequent DNA isolation was performed using the PowerLyzer PowerSoil DNA Isolation Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions as described previously [28].

16S rRNA gene amplicon sequencing and bioinformatic processing
Amplicon libraries of the V4 hypervariable region for 16S rRNA gene were prepared from DNA extracts using modified universal bacterial primer pairs 515F and 806R [29].Amplicon libraries were indexed, cleaned, and sequenced according to the 16S Metagenomic Sequencing Library Preparation protocol.Paired-end sequencing was performed using MiSeq Reagent Kit v3 (600-cycle kit) (Illumina) on a MiSeq System (Illumina), at the South Australian Genomics Centre (SAGC).Paired-end 16S rRNA gene sequence reads were analysed using QIIME2 version 2021.11.0 [30].Brief ly, de-noising was performed on demultiplexed sequences using the DADA2 plugin [31], resulting in a mean read depth of 15 563 ± 2719 for stool and 6941 ± 4503 for tissue.Taxonomic classification of amplicon sequence variants (ASVs) was performed based on the V4 hypervariable region of the SILVA 16S rRNA gene reference database (version 138) at 99% similarity [32].Sufficient coverage at this depth is confirmed by the rarefaction curve, which reached an asymptote.Sequence data have been deposited in the National Center for Biotechnology Information Sequence Read Archive (NCBI SRA) under accession number PRJNA1011386.

Microbiota characterization
The taxonomic relative abundance at the genus level was used to generate alpha diversity (within-group) and beta diversity (between-group) measures.Alpha diversity metrics (observed ASVs, Pielou's evenness, Shannon diversity, and Faith's phylogenetic diversity) were obtained from QIIME2 at sampling depth of 9883 reads (faecal samples) and 662 reads (mucosal tissue samples).The Bray-Curtis dissimilarity index was calculated to compare microbiome similarity between groups (beta diversity), using square-root-transformed species relative abundance data using the "vegan" package in R. Nonmetric multidimensional scaling (nMDS) for all beta diversity measures were generated using the "vegan" package in R. Core taxa were defined as those present in >95% of samples, with a mean relative abundance of >0.01%.Identification of taxa with α-1,2-L-fucosidases capability was determined by comparing the genus-level taxonomic classification to genomes identified by CAZy [26] as carrying either the GH29, GH95, or GH151 enzyme families.

Quantification of Bifidobacterium species and total bacterial load
We investigated the extent to which 16S amplicon sequencing could discriminate between different Bifidobacterium species.As expected, level 7 resolution (species-level output) was unable to differentiate bifidobacterial strains, ref lecting a well-recognized limitation of this approach.Given that, quantification of total bacterial load, B. breve, B. infantis, and B. bifidum was performed by SYBR Green-based qPCR assays (Supplementary Table S2).For all qPCR assays, 1 μl of DNA template was combined with 0.7 μl of 10 μM forward primer, 0.7 μl of 10 μM reverse primer, 17.5 μl of 2 × SYBR Green (Applied Biosystems, Waltham, MA, USA), and 15.1 μl nuclease-free water.All samples were run in triplicate (10 μl each replicate).Gene copy quantification was performed using a standard curve generated from a known concentration of a pure colony control.Any sample with a cycle threshold (CT) ≥40 cycles was defined as 40 (limit of detection).

Statistical analysis
Experimental mice were randomly assigned to different treatment groups.The investigators were not blinded to the experimental groups.No outliers have been removed from any of the data presented.All data analyses were performed using either R (R Foundation for Statistical Computing; version 4.1.0)or GraphPad Prism software (GraphPad Software, Inc.; version 9.00).For parametric data, unpaired Student's t test was used to compare data between two unpaired groups; one-way ANOVA was used to compare data among three or more unpaired groups.For nonparametric data, the Mann-Whitney U test was used to compare data between two unpaired groups; the Kruskal-Wallis test was used to compared data among three or more unpaired groups.Differences in Bray-Curtis dissimilarity between groups was performed by permutational multivariate ANOVA (PERMANOVA) and pairwise PERMANOVA, using the "adonis" package in R, with 9999.Linear discriminant analysis effect size (LEfSe) was applied to identify the abundant taxa in each site, using default parameters [34].The area under the curve (AUC) was calculated for in vitro growth experiments (using OD 600 values) and bifidobacterial persistence in mice (using copies/ng faecal DNA).The log-rank test was employed to compare survival time differences based on bacterial qPCR detection.One-tailed tests were used where differences between groups were hypothesized to be in a single direction.Statistical outcomes with P value <.05 were considered statistically significant.The core taxa plot was generated using GraphPad Prism; other data were visualized using R.

Fut2 shapes the faecal microbiota
Assessment of the faecal microbiota between SPF Fut2 WT and Fut2 KO littermates was performed at 6 weeks of age in both male and female mice (Fig. 1A).Faecal microbiota composition (Fig. 1B) differed significantly between Fut2 WT and Fut2 KO littermates (PER-MANOVA: R 2 = 0.028; P = .028,Fig. 1C) when male and female mice were assessed together.However, stratification according to sex identified a greater divergence according to genotype in male mice (PERMANOVA: R 2 = 0.12; P = .021)compared to female mice (PER-MANOVA: R 2 = 0.037; P = .38,Fig. 1C).Exploration of this sex effect identified a significant interaction between sex and Fut2 genotype (PERMANOVA: R 2 = 0.11; P = .0068,Supplementary Table S3).These findings were unchanged after adjustment for cage effects (Supplementary Table S3).Given the interaction between sex and genotype, all subsequent experiments involved male mice only.

Fut2-microbiota relationships can be recapitulated in vitro through glycan exposure
To further investigate the relationship between α(1,2)-fucosylated glycans and intestinal microbiology, faecal homogenate from Fut2 WT mice was grown in a mBasal media with or without the α(1,2)-fucosylated glycan, 2'-FL.Microbiota assessment following in vitro culture (Fig. 3A) confirmed the findings of the in vivo faecal microbiota analysis, with a significant difference in Bray-Curtis similarity between faecal cultures with and without 2'-FL (PERMANOVA: R 2 = 0.90; P < .0001,Fig. 3B).This difference was marked by an enrichment of glycan-utilizing genera (Bacteroides, Enterococcus, Lactobacillus), with compensatory decreases in the relative abundance of other taxa (Fig. 3C).
We investigated whether differences in microbiota composition between Fut2 WT and Fut2 KO mice ref lected selection for bacterial populations able to utilize α(1,2)-fucosylated glycans for growth in Fut2 WT animals.Faecal homogenates from Fut2 WT and Fut2 KO mice were used to inoculate mBasal media with or without 2'-FL supplementation.The increase in bacterial density between 2'-FL supplemented media and media alone was significantly greater when faecal homogenates were derived from Fut2 WT compared to Fut2 KO mice (median AUC [WT] = 20.5 [IQR = 16.6, 23.5]; AUC [KO] = 16.7 [13.1, 19.7]; P = 0.0051, Fig. 3F), consistent with a greater abundance of glycan-utilizing bacteria.

Prior antibiotic exposure profoundly affects the Fut2-probiotic relationship
To test whether the relationship between Fut2 genotype and probiotic strain characteristics were independent of antibiotic exposure, we supplemented nonantibiotic-exposed mice with B. infantis (Fig. 5A).No significant difference in B. infantis persistence postgavage was observed between Fut2 WT and Fut2 KO mice (Fig. 5B), and B. infantis was not detectable in intestinal tissue from either Fut2 WT or Fut2 KO mice at day 5 postgavage (Supplementary Fig. S5A).However, analysis of B. infantis abundance following gavage (based on AUC) revealed an effect that was opposite to that observed in antibiotic exposed mice, with B. infantis significantly higher in Fut2 KO mice compared to Fut2 WT (mean AUC [WT] = 3.8 [SD = 0.6]; AUC [KO] = 5.6 [0.90]; P = .00046;Fig. 5C).
Faecal levels of Bifidobacterium probiotics were assessed during the instillation period (samples collected 2 h prior to gavage and 6 h after gavage).Levels in postgavage samples (6 h) did not differ between Fut2 WT and Fut2 KO mice, consistent with the instillation of equal probiotic loads.However, at 22 h postgavage (2 h prior to gavage), levels in Fut2 KO mice were significantly higher than in Fut2 WT mice (P < .0001;Supplementary Fig. S5B).No cumulative effect was observed with repeat gavage and the decline in probiotic levels after 22 h was comparable to that observed in the 24 h postcessation of installation.

Discussion
Although probiotics have shown great potential in modifying host-microbiome interactions [36], their actual performance has been disappointing in many clinical contexts [37].Previous studies have investigated the reasons for this underperformance, relating to cohort-level effects and variation in response between individuals.Mode of delivery and dose have both been shown to contribute to overall efficacy [38,39], whereas the habitual diet of the recipient, particularly fibre intake, is an important determinant of probiotic response [40].The inf luence of factors that shape the gut microbiome on the abundance and persistence of probiotics is unsurprising [41,42], given the ability of resident gut microbiota to competitively exclude introduced populations [10].Indeed, exposure to antibiotics, a factor that greatly impacts the gut microbiome, has been shown to considerably inf luence probiotic effects at a microbiological level [9,43].However, while common FUT2 genetic variants are known to help shape intestinal microbiology [16,17], the effect of secretor status on probiotics had not been described.
Our study highlights several important points in relation to interindividual variance in intestinal microbiology and probiotic efficacy.With 20% of the global population also homozygous for a nonfunctional FUT2 gene [15], our findings suggest that these "non-secretor" individuals will also experience different probiotic population dynamics compared to "secretor" individuals, if the probiotic taken contains one of the many bacterial species able to utilize α(1,2)-fucosylated glycans (H antigens).In our study, this was ref lected in the significantly greater transience of B. infantis in the faecal and intestinal microbiome of nonsecretor (Fut2 KO ) mice compared with secretor mice following antibiotic exposure.
Bifidobacterial species that are commonly used as probiotics are relatively close phylogenetically but differ in their ability to use glycans, even at a strain level [44].We showed that neither B. breve (JCM 1192) nor B. bifidum (JCM 1255) differed in their abundance or persistence between secretor and nonsecretor animals.In contrast, B. infantis (JCM 1222) persisted for significantly longer and showed a significantly higher abundance in secretor mice compared to nonsecretor mice.This finding likely ref lects differences in H antigen hydrolysis and catabolism capacities between species when administered as a probiotic.For example, B. infantis encodes GH29, GH95, and GH151 family intracellular α-1,2-L-fucosidases, along with fucose transporters to facilitate internalization [44].While independent hydrolysis and catabolism of mucin-bound H antigens by B. infantis are not hypothesized [25], cross-feeding by organisms with extracellular α-1,2-L-fucosidases is likely, even following antibiotic supplementation [45].In contrast, B. bifidum, while expressing extracellular GH29 and GH95 α-1,2-L-fucosidases, does not consume fucose to facilitate growth [27].Finally, B. breve encodes a separate GH95 intracellular α-1,2-L-fucosidase along with fucose transporters.While this species is capable of utilizing the H antigen with support from cross-feeding [46], these findings suggest reduced persistence compared with B. infantis, when administered as a probiotic.
We found that antibiotic exposure inf luenced the persistence of probiotics in a secretor status-dependent manner.In the absence of microbiota depletion through antibiotic exposure, it would be expected that other commensal bacteria would utilize available glycans within the secretor gut.Moreover, such strains are highly adapted to an individual's gut environment, making them likely to outcompete any exogenous glycan utilizers that are introduced.When we explored this directly, we found that in the absence of a prior period of antibiotic exposure, the higher levels and greater persistence of B. infantis in secretors was inverted, with these B. infantis being significantly higher in nonsecretor mice.These findings likely ref lect the competitive exclusion of H antigen-utilizing probiotics in the secretor gut and highlight the importance of considering the ecological context in relation to probiotic impact.
It should be noted that our antibiotic mix contained a cocktail of ampicillin and neomycin, designed to deplete a wide range of bacteria.While most Bifidobacterium strains are resistant to neomycin, the tested strains are sensitive to ampicillin [47].We designed the experiment so that gavage with Bifidobacterium was immediately after ceasing antibiotic depletion to maximize colonization without competition from other bacteria.It is possible that residual antibiotics in the intestine deplete Bifidobacterium over the first days of gavage.For this reason, we performed gavage for 5 days, a time period that extends beyond the activity spectrum of the administered antibiotics.Such an antibiotic combination is common for mouse models [48,49], as well as empiric for suspected sepsis in humans [50].
While this study was performed in mice, the effect of secretor status on bifidobacterium supplementation has important implications for probiotic strategies in humans.It is crucial to consider individual host traits and recent antibiotic exposure when designing a probiotic intervention [ 51].The findings here suggest that the 20% of the population who are nonsecretors may have poorer persistence of H antigen utilizing probiotics, such as B. infantis, compared to secretors following antibiotic exposure.Conversely, in the absence of recent antibiotic exposure, higher levels of microbial niche occupancy in secretors may hamper B. infantis persistence compared to nonsecretors.An individualised supplementation with prebiotics may have potential as a means to optimize probiotic uptake in nonsecretors.For example, previous studies have shown that supplementation with human milk oligosaccharides can enhance B. infantis engraftment [20], with successful supplementation shown to reduce intestinal inf lammation in infants [52].Investigating additional α(1,2)-fucosylated glycans, given as prebiotics, may lead to improved outcomes of B. infantis supplementation in nonsecretor individuals.
Determining the impact of secretor status on other species commonly considered beneficial and marketed as probiotics is challenging due to their broad range of carbohydrate utilization capabilities [53].For instance, Akkermansia muciniphila, a mucindegrading species, has been associated with a reduced risk of chronic inf lammatory diseases in humans and mice [54], and is a potential target for probiotic development [51].However, its utilization of mucin glycoproteins, including the α(1,2)fucosylated glycan, 2 -fucosyllactose [53], suggests that it may also be affected by secretor status.Although Akkermansia was not detected in the mice of this study, we found that Candidatus Arthromitus, another genus associated with immune modulation [55], was enriched in the distal small intestinal mucosal tissue of secretor mice.Genome annotation of Candidatus Arthromitus has indicated other fucose utilization capabilities [56], indicating that the functional FUT2 gene may promote colonization by this species.
Our experiments involved SPF mice that were obtained through heterozygous mating.Such breeding was essential to allow comparison of Fut2 WT and Fut2 KO littermates from a maternal secretor lineage.The findings from this study are therefore independent of vertical transmission effects, which are known to inf luence the microbiome of the offspring [57,58], and indicate that a change in gut microbiology occurred postweaning.This difference in baseline gut microbiota composition between secretors and nonsecretors was also only evident in male mice.The effect of sex on the relationship between secretor status and the gut microbiome is difficult to explain but may relate to variable intestinal expression of Fut2, which can be altered factors such as stress [59].In addition, independent interactions between sex hormones and the gut microbiome [60] may affect the relationship between Fut2 and the gut microbiome.
We acknowledge the importance of considering blood antigens/ABO phenotypes in interpreting the inf luence of FUT2 gene on the gut microbiome, as indicated by recent studies [61,62].Indeed, in humans, FUT2 is responsible for the generation of the H antigen, which can be further modified to give the OLewis b , ALewis b , or BLewis b antigens [63].Each of these glycans can modulate the competitive advantage of particular microbes capable of cleaving the oligosaccharide constituents.In the absence of FUT2, these Lewis b antigens are not displayed, leading to a Lewis a antigen.While our study did not address these blood type variations, it should be noted that even in humans, a secretor O blood group and a nonsecretor O blood group are not the same.The impact of this on the gut microbiome is evidenced by studies reporting an association between H antigen concentrations and gut microbiome characteristics [58].
Our study demonstrates a Fut2-dependent genetic determinant for interindividual response to probiotic supplementation, which is affected by antibiotic exposure and glycan utilization capabilities of the probiotic strain.With prior antibiotic exposure, Fut2 functionality was associated with increased persistence of B. infantis, consistent with its ability to utilize the H antigen.However, without antibiotic exposure, Fut2 functionality was associated with lower abundance of B. infantis, relating to difference in baseline microbiology and niche space occupation.

Figure 2 .
Figure 2. Fut2 affects the intestinal microbiota.(A) Taxa bar plot of the intestinal mucosal microbiota of Fut2 WT and Fut2 KO mice.(B) NMDS plots of Fut2 WT and Fut2 KO intestinal mucosal microbiota.Significance: Permutational multivariate ANOVA.(C) Taxa significantly different between Fut2 WT and Fut2 KO intestinal mucosal tissue.Significance: LEfSe.No taxa differed between Fut2 WT and Fut2 KO proximal small intestine.

Figure 3 . 2 -
Figure 3. 2 -Fucosyllactose modifies the faecal microbiota in vitro and enhances growth of α(1,2)-fucosylated glycan-utilizing bacteria.(A) Taxa bar plot of the faecal microbiota of Fut2 WT mice following anaerobic growth either with or without 2 -fucosyllactose (2 -FL).(B) NMDS plot of faecal microbiota following anaerobic growth with or without 2 -FL.significance: Permutational multivariate ANOVA.(C) Taxa significantly different between Fut2 WT faecal samples following anaerobic growth with or without 2 -FL.Significance: LEfSe.(D) Comparison of colonies relative abundance of identified colonies following anaerobic growth with or without 2 -FL.* E. coli or Shigella.(E) OD following Fut2 WT faecal bacteria cultured with or without 2 -FL.(F) OD of Fut2 WT or Fut2 KO faecal bacteria following growth with 2'-FL.OD normalized to growth in media without 2'-FL.

Figure 4 .
Figure 4. Bifidobacterium infantis, but not B. breve or B. bifidum, persists longer in Fut2 WT mice following antibiotic pre-exposure.(A) Experimental design.(B) Persistence of detectable Bifidobacterium species in stool following antibiotic pre-exposure.Significance: log-rank test.(C) Bacterial copies of Bifidobacterium species in stool following antibiotic pre-exposure.Significance: T-test of area under the curve.(D) Bacterial copies of B. infantis in intestinal tissue mucosa following antibiotic pre-exposure.Significance: Mann-Whitney U test.

Figure 5 .
Figure 5. Bifidobacterium infantis persists longer in Fut2 KO mice without antibiotic pretreatment.(A) Experimental design.(B) Persistence of gavaged B. infantis in stool without antibiotic pre-treatment.Significance: Log-rank test.(C) Bacterial copies of B. infantis in stool without antibiotic pre-treatment.Significance: T-test of area under the curve.