Mechanism of 2′-fucosyllactose degradation by human-associated Akkermansia

ABSTRACT Among the first microorganisms to colonize the human gut of breastfed infants are bacteria capable of fermenting human milk oligosaccharides (HMOs). One of the most abundant HMOs, 2′-fucosyllactose (2′-FL), may specifically drive bacterial colonization of the intestine. Recently, differential growth has been observed across multiple species of Akkermansia on various HMOs including 2′-FL. In culture, we found growth of two species, A. muciniphila MucT and A. biwaensis CSUN-19,on HMOs corresponded to a decrease in the levels of 2′-FL and an increase in lactose, indicating that the first step in 2′-FL catabolism is the cleavage of fucose. Using phylogenetic analysis and transcriptional profiling, we found that the number and expression of fucosidase genes from two glycoside hydrolase (GH) families, GH29 and GH95, vary between these two species. During the mid-log phase of growth, the expression of several GH29 genes was increased by 2′-FL in both species, whereas the GH95 genes were induced only in A. muciniphila. We further show that one putative fucosidase and a β-galactosidase from A. biwaensis are involved in the breakdown of 2′-FL. Our findings indicate that the plasticity of GHs of human-associated Akkermansia sp. enables access to additional growth substrates present in HMOs, including 2′-FL. Our work highlights the potential for Akkermansia to influence the development of the gut microbiota early in life and expands the known metabolic capabilities of this important human symbiont. IMPORTANCE Akkermansia are mucin-degrading specialists widely distributed in the human population. Akkermansia biwaensis has recently been observed to have enhanced growth relative to other human-associated Akkermansia on multiple human milk oligosaccharides (HMOs). However, the mechanisms for enhanced growth are not understood. Here, we characterized the phylogenetic diversity and function of select genes involved in the growth of A. biwaensis on 2′-fucosyllactose (2′-FL), a dominant HMO. Specifically, we demonstrate that two genes in a genomic locus, a putative β-galactosidase and α-fucosidase, are likely responsible for the enhanced growth on 2′-FL. The functional characterization of A. biwaensis growth on 2′-FL delineates the significance of a single genomic locus that may facilitate enhanced colonization and functional activity of select Akkermansia early in life.

intact where they act as prebiotics for early bacterial colonizers of the human gut (8).HMOs are composed of only five monosaccharide units (glucose, galac tose, N-acetylglucosamine, fucose, and N-acetylneuraminic acid) that form structurally complex linear and branched polysaccharides (9)(10)(11).The presence and concentration of these HMOs vary across individuals and are influenced by maternal genetics (12,13).For example, one of the HMOs, 2′-FL, is highly produced by mothers with an active FUT2 secretor gene, accounting for ~30% of the total HMOs in these individuals (5).Higher levels of 2′-FL correlate with lower levels of allergies, and lower incidence of eczema and diarrhea in infants (14).In addition to regulating the host immune system, bacterial fermentation of HMOs produces short-chain fatty acids (SCFA) that help maintain the intestinal epithelium and regulate appetite (15).
While only a few microorganisms in the infant's gut possess the ability to use the entire suite of HMOs, many are capable of fermenting components of HMOs for growth (16)(17)(18).For instance, Bifidobacteria encode both glycoside hydrolases (GH) and transport systems (19,20) to catabolize HMOs.However, there are species and strain-to-strain variability in the mechanisms for HMO degradation across Bifidobacteria.For example, Bifidobacterium longum subsp.infantis uses intracellular enzymes, whereas Bifidobacterium bifidum uses extracellular enzymes, pointing to differences in compet itive strategies between these organisms (19)(20)(21)(22).In addition, other intestinal bacte ria such as Bacteroides thetaiotaomicron and Bacteroides fragilis repurpose machinery primarily responsible for the breakdown of mucins to degrade HMOs (23).
The genus Akkermansia is comprised of commensal mucin-degrading bacteria that colonize the gastrointestinal tract from early life to adulthood (24) and can also use HMOs for growth (25)(26)(27).During growth on HMOs, Akkermansia produces acetate, succinate, and propionate (26).The most abundantly produced SCFA, acetate, is involved in fueling gastrointestinal epithelial cells, leading to mucin secretion, and helping to maintain gut barrier integrity (28).In infant guts, there is a positive correlation between the abundance of Akkermansia and fucosylated HMOs in the mother's breast milk (29,30), suggesting a role for this organism in HMO metabolism in vivo.Given that Akkerman sia are largely considered beneficial members of the gut microbiome (31), understanding their ability to degrade HMOs may lead to therapeutic applications.
Although Akkermansia are predominantly mucin-degrading specialists, given the structural and compositional similarities between HMOs and mucin oligosaccharides, species may use similar enzymes to break down HMOs (26,32).Most of the efforts in developing a mechanistic understanding of Akkermansia utilization of host glycans thus far have centered around a single isolate of A. muciniphila (Muc T ).The Muc T strain has been shown to use a diversity of GHs including fucosidases, β-hexosaminidases, and β-galactosidases for degrading a variety of human glycans including HMOs and lactose (26,32,33).Recently, we and others have identified and isolated at least four species-level phylogroups (AmI-AmIV) of human-associated Akkermansia.Thus far, three phylogroups have been named, A. muciniphila (AmI), A. massiliensis [AmII (34)], and A. biwaensis [AmIV (35)].The fourth phylogroup, AmIII, is represented by our CSUN-56 isolate and two isolates from Guo and colleagues (36) but is not yet formally named.
We recently demonstrated that all four of these human-associated Akkermansia species have the genetic potential and functional ability to metabolize and grow using a variety of HMOs (27).Moreover, some species were more efficient and grew better on individual HMOs.In particular, A. biwaensis strain CSUN-19 (AmIV), grew to higher final optical densities (OD) on a suite of HMOs and encodes for a larger complement of putative GHs associated with HMO degradation.Interestingly, despite high growth yields when grown on 2′-FL, A. biwaensis consumed the least amount of the HMO across Akkermansia species.We hypothesized that strains with a greater ability to grow on 2′-FL will have greater expression of GHs or additional GHs required for growth on this substrate.Through phylogenetic analysis, transcriptional profiling, and heterologous expression systems, we identified genes encoding putative fucosidases (GH29 and GH95) and a β-galactosidase (GH2) from A. biwaensis involved in 2′-FL degradation.

Phylogenetic analysis of putative fucosidase encoding genes
To determine the phylogenetic relationships of select glycoside hydrolase (GH) genes across Akkermansia species, we analyzed the genomes of 17 publicly available isolates spanning the known phylogenetic diversity of human-associated Akkermansia (36)(37)(38)(39).Genome accession numbers of isolates used in the analysis are provided in Table S1 (Table S1).Assembled genomes were submitted to the dbCAN meta server for anno tation of GH encoding genes (40,41).dbCAN uses several tools for annotation and we considered only annotations that agreed using HMMER (42) and DIAMOND (43).All genes annotated as GH2, GH29, and GH95 were identified in the dbCAN output, extracted from the genomes of each strain, imported into MEGA11 (44)(45)(46), and aligned using MUSCLE (47,48).Maximum likelihood trees for each GH family were generated in MEGA11 using default parameters with 100 bootstrap iterations.After the identifi cation of putative fucosidases involved in 2′-FL catabolism, GH29 gene annotations from A. muciniphila Muc T , and A. biwaensis CSUN-19 were phylogenetically compared to functionally characterized GH29 fucosidases from other intestinal bacteria (49)(50)(51)(52)(53)(54).Accession numbers for proteins and the associated organisms used in the analysis are provided in Table S2 (Table S2).For these analyses, amino acid sequences were aligned using MUSCLE, and a maximum likelihood tree was generated in MEGA11 using default parameters with 50 bootstrap iterations.
All cultures were incubated at 37°C in a Vinyl Anaerobic Chamber (Coy Laboratory Products, Incorporated, Grass Lake, MI).To monitor growth, 500 μL of each culture was used to determine OD 600nm on a Biophotometer plus spectrophotometer (Eppendorf, Biophotometer Plus).Cultures with an OD 600nm > 1.0 were diluted with sterile BTTM and re-read.All experimental cultures were grown in quadruplicate.
For follow-up growth experiments with the Akk-EH114 (MucT Tn-Amuc_2072::HHJ01_10880-see below) strain, the following sugar concentrations were used: 4 mM 2′-FL (Glycom, Hørsholm, Denmark) with 0.4% porcine mucin.Akkermansia was grown overnight in 5 mL basal mucin medium (27) with 0.4% porcine mucin, sub-cultured in 1 mL BTTM containing media with the appropriate sugars, and 200 μl was transferred to a flat clear-bottom 96-well plate and read hourly on a Spectrostar Nano plate reader (BMG Labtech, USA) for 46 ± 2 hours with 37°C incubation and 10 seconds of orbital shaking prior to each read.Optical density was converted to OD 600nm using a linear equation and the initial (t = 0) read was subtracted from all time points as a media blank.All cultures were grown in triplicate and each experiment was repeated three times, except where indicated.

Targeted metabolomics
Culture supernatants were collected to measure 2′-FL and Glc degradation during RNA harvest time points.In addition to each parent sugar (2′-FL, Glc, and GlcNAc), their sugars (fucose, lactose, glucose, and galactose) were also quantitatively measured using high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) (57).Culture supernatants were transported frozen (dry ice) where they were thawed in a water bath, vortexed to homogenize, and centrifuged at 7,000 × g for 5 minutes at 10°C.Then, 0.5 µL of the culture media was injected into the HPAEC-PAD for the detection of the expected analytes described above.Carbohydrate analyses were done as described previously (27).

RNA extraction and purification
RNASeq was performed on cultures grown in BTTM supplemented with GlcNAc and either Glc or 2′-FL (described above) to identify potential genes involved in 2′-FL degradation.Samples were collected at approximately mid-log growth for each condition/species combination.For A. muciniphila Muc T , samples were collected after 12 hours of growth on Glc and 49 hours for growth on 2′-FL.For A. biwaensis CSUN-19, samples were collected after 14 hours of growth on Glc and 34 hours of growth on 2′-FL.At the indicated time points, 1 mL of each culture was centrifuged for 7 minutes at 4°C and 10,000 × g.Supernatants from each pelleted sample were removed and placed in the −80°C freezer for metabolomics analysis and pellets were flash frozen in liquid nitrogen before storage at −80°C.RNA was extracted from cell pellets using the phenol/chloroform extraction method outlined in the protocol in RiboPure Bacteria Kit (Life Technologies Corporation, Carlsbad, CA).Molecular grade Chloroform (MP Biomedicals, LLC, Solon, OH) and Absolute Ethanol (200 proof ) (Fisher Bioreagents, Fisher Scientific, Fair Lawn, NJ) were used for the chloroform/ethanol steps.The elution solution was replaced by water (for RNA work) that is nuclease-free and DEPC-treated (Fisher Bioreagents, Fisher Scientific, Fair Lawn, NJ).A total of 50 µL was eluted in two steps (25 µL each step) with 10-minute incubations for each elution step.Two DNase treatments were completed of 1-hour duration each: DNase I treatment from the RiboPure Bacteria Kit followed by Turbo DNase treatment (Life Technologies Corporation, Carlsbad, CA).During Turbo DNase treatment, a master mix containing 5 µL turbo DNase buffer and 1 µL turbo DNase per sample was made prior to adding to the samples.The reagent used for the Turbo DNase inactivation was the same as the RiboPure Bacteria Kit.
RNA was purified and concentrated according to the protocol for RNA Clean and Concentrator Kit -5 (Zymo Research Corporation, Irvine, CA) with an eluted volume between 30 and 35µL.Final RNA concentrations were measured using the Qubit 2.0 Fluorometer with RNA High Sensitivity (HS) Assay Kit (Life Technologies Corporation, Carlsbad, CA) after the purification.All samples were stored at −20°C until submission to Novogene USA, Inc. for Total RNA Sequencing.

RNA sequencing and analysis
RNA samples were sequenced by Novogene USA, Inc. using the Illumina Sequencing HiSeq platform (Illumina, CA, USA).Quality control of RNA samples was assessed using High Sensitivity RNA ScreenTape on TapeStation (Agilent Technologies Inc., CA, USA) and quantified using the Qubit 2.0 RNA High Sensitivity assay (Thermofisher, MA, USA).Ribosomal RNA depletion was performed with Ribo-Zero Plus rRNA Removal Kit (Illumina Inc., CA, USA).Samples were primed and fragmented based on the recommendation of the manufacturer.First-strand synthesis was performed with Protoscript II Reverse Transcriptase using a longer extension step of ~30 minutes at 42°C.NEBNext UltraTM II Non-Directional RNA Library Prep Kit for Illumina (New England BioLabs Inc., MA, USA) was used for the remaining steps of library construction.The final library size was around 350 bp with an insert size of 200 bp.The final quantity of the library was quantified using Qubit 2.0 and the quality was assessed by the TapeStation D1000 ScreenTape (Agilent Technologies Inc., CA, USA).For the Illumina HiSeq, 8-nt dual-indices were used and equimolar pooling of libraries was done based on the values from the QC before sequencing with a read length configuration of 150 PE for 20 M PE reads per sample (20M in each direction).
RNA sequencing analysis was performed using the DIY.transcriptomics pipeline (https://diytranscriptomics.com)(58).Raw reads from transcriptomic data were mapped to coding sequences (CDs, no rRNA, or tRNA) of each strain using Kallisto (59).The remaining analyses were completed in Rstudio (https://www.rstudio.com)(File S1 and S2).Kallisto transcript abundance measurements were imported into RStudio using the R package tximport (60) for filtering and normalization.The Bioconductor/R package edgeR (61) cpm function was used to create a list of counts per million (cpm) per transcript.Genes with no cpm were filtered out of the data set using base R's subsetting method.The edgeR function calcNormFactors was used to normalize the filtered data using the TMM method (Trimmed Mean of M-values) to the data set (62).
For analyses of differentially expressed genes (DEGs), a design matrix targeting treatment (Glc versus 2′-FL) was set up to compare the cpm of genes in the Glc control versus the 2′-FL treatment.To create a model mean-variance trend and fit the linear model to the data, the voom function from the Bioconductor/R package limma (63) was used to model a mean-variance relationship, the function lmfit was used to fit a linear model to the data, and makeContrasts was used to make contrasts of the design matrix between the 2′-FL group versus the Glc group.Bayesian statistics were computed for the linear model fit using the eBayes function from limma which resulted in moderated statistical values of differential expression (Adjusted P-value, Log fold change, Average Expression, T statistics).To extract DEGs from the linear model fit and Bayesian analysis, the function decideTests from limma was used to generate a list of significantly regulated DEGs with strict cutoff values (corrected P value ≤ 0.01, log2 fold change ≥ 2) (64).
The R package EnhancedVolcano was used to generate volcano plots highlighting the GH29 and GH95 genes for each Akkermansia strain with strict cutoff values (P adj ≤ 0.05, log2 fold change ≥2) (65).

Cloning and purification of putative β-galactosidase in E. coli
To characterize one of the β-galactosidases that may be involved in 2′-FL metabolism, the gene fragment of the HHJ01_10865 protein, which encodes the mature GH2 protein lacking the signal peptide (as predicted by SignalP 5.0) (66) was amplified from A. biwaensis CSUN-19 genomic DNA using the primers as shown in Table S3 (Table S3), cut with HindIII and EcoRI and inserted into the expression vector pET28a(+) (Novagen, Inc.).The resulting recombinant plasmids were transformed into E. coli Tuner BL21 for protein overexpression.Bacterial cultures were grown in 50 mL LB medium with kanamycin (50 μg/mL) at 37°C to OD 600nm ~ 0.5, followed by induction with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 2 hours at 37°C.Cells were harvested by centrifugation (5,000 × g for 5 minutes at 4°C), and the cell pellets were flashfrozen in liquid nitrogen for 5 minutes and stored at −20°C until purification.
Thawed cell pellets were resuspended in 5 mL of purification buffer (50 mM Tris pH 8.0, 200 mM NaCl) and lysed by 4 rounds of sonication (2 minutes of 1 second on/2 seconds off) with 40% amplitude (Qsonica, USA).Lysates were centrifuged for 30 minutes at 38,000 × g and 4°C and cleared supernatants loaded onto Ni-NTA columns with 0.5 mL bed volume (1018244 Qiagen, Germany).The columns were washed sequentially with three column volumes (CV) each of the purification buffer, a low-imidazole wash, and a mid-imidazole wash.Bound proteins were eluted with 4 CV of 100 mM imidazole in a purification buffer.Fractions containing recombinant protein, as assessed by SDS-PAGE analysis, were combined and dialyzed against the purification buffer, and the final protein concentration was determined using the Qubit protein assay kit (#Q33211 Life Technologies Corporation, USA).

Activity of purified proteins
All enzyme activity and inhibition reactions were performed in 50 mM Tris, 200 mM NaCl, pH 8.0, unless otherwise indicated.Reactions involving 2.5 mM ONPG were carried out for 1 hour at 37°C.The enzymatic product of the analog, orthonitrophenol (ONP), was determined by measuring A 405 nm using a 96-well plate reader Spectromax M5 (Molecular Devices, USA) and monitored every 5 minutes for 1 hour following the addition of substrate.

Expression of A. biwaensis HHJ01_10880 in A. muciniphila Muc T
The gene encoding the putative A. biwaensis GH29 fucosidase HHJ01_10880 (clade 4) was introduced into A. muciniphila Muc T using a modified mariner transposon (Tn) delivery system (pSAM_Akk) (67).The plasmid was re-engineered to insert genes within the inverted repeats of the Tn system and enable the expression of genes of interest in A. muciniphila under the control of the promoter of Amuc_1505 (RNA pol) promoter.The resulting plasmid (pE_Akk) contains a cat resistance cassette and the Amuc_1505 promoter flanked by Mariner transposase recognition sites.

Evolutionary relationship of GHs involved in 2′-FL metabolism
Enzymes that target 2′-FL include fucosidases that cleave the α1,2-linked fucose from the parent lactose, which can be further metabolized by a β-galactosidase.To identify mechanisms of 2′-FL degradation by human-associated Akkermansia, we first charac terized the evolutionary relationship of predicted fucosidases belonging to the GH29 and GH95 families and β-galactosidases belonging to the GH2 family across species.A phylogenetic analysis revealed seven well-supported clades for GH29 and three for GH95 (Fig. 1).Generally, when each Akkermansia species was represented within a clade, strains from the same species clustered into sub-clades showing phylogenetic congruence for each species.An exception to this pattern is for the GH95 tree, where sequences of A. biwaensis were quite divergent within clade 3.
Representation of the strains across the clades was not uniform, where some Akkermansia species had greater numbers and diversity of fucosidases.Specifically, across the GH29 genes, A. biwaensis strains including strain CSUN-19 used in this study were present in 6 of the 7 clades including one clade (clade 4) found exclusively in this species (Table 1).By contrast, A. muciniphila strains, including the type strain Muc T , were found in only 4 of 7 clades with none being exclusive to this species.Interestingly, however, clade 6 lacked sequences from either of these two representative species, but contained GH29 genes present in only phylogroup AmIb, a closely related sub-species of A. muciniphila previously defined by Becken and colleagues (38).Similarly, in the GH95 tree, A. massiliensis, including strain CSUN-17, Akkermansia sp.CSUN-56 from phylogroup AmIII, and A. biwaensis including strain CSUN-19 had genes present in all three clades, whereas A. muciniphila including the type strain Muc T , only had genes in two clades (Fig. 1; Table 1).
A similar phylogenetic analysis of GH2 genes in diverse Akkermansia revealed eight clades (Fig. 1).In line with previous findings, the representative species diverged, in which A. muciniphila Muc T had fewer copies of GH2 (5/8 clades) (Table 1), and A. biwaensis CSUN-19 had the highest copy number of GH2s in the genome (7/8 clades).Each species also contained unique GH2 sequences; clade 5 was unique to A. biwaensis strains while clade 7 was unique to a subset of A. muciniphila strains in the AmIb subgroup (Fig. 1).

Growth of A. muciniphila Muc T and A. biwaensis CSUN-19 on 2'-fucosyllactose and glucose
A. muciniphila and A. biwaensis, the two species with the most divergent number of predicted fucosidases, were used to determine the relative growth and generation of sugar metabolites in a minimal media with 2′-FL.Interestingly, although both strains grew using 2′-FL and GlcNAc, the final OD 600nm was different across species.By 36 hours, the OD 600nm of A. biwaensis CSUN-19 was nearly double that of A. muciniphila Muc T (Fig. 2).Although we are unable to calculate a formal growth rate because of sparse sampling, A. biwaensis CSUN-19 reached a final OD ≥1.5 by 36 hours, whereas by 120 hours, A. muciniphila Muc T had yet to reach this level of growth.This contrasts with the growth on Glc and GlcNAc, where both species reached similar densities across all time points (Fig. S2).
We tested whether these species were degrading and consuming the available sugars through similar pathways by measuring the concentrations of the various oligosacchar ides (e.g., 2′-FL, lactose) and monosaccharides (e.g., Glc, galactose, GlcNAc).Both strains degraded 2′-FL into fucose and lactose and consumed nearly all GlcNAc by the end of the experiment.Although lactose and fucose were found in the culture medium during growth on 2′-FL, fucose was quickly consumed, whereas lactose accumulated in cultures from both organisms (Fig. 2).In addition, small but appreciable amounts of glucose a Clade designations are based on maximum likelihood trees (Fig. 1).
and galactose accumulated in the culture medium of A. biwaensis CSUN-19 but not A. muciniphila Muc T .As anticipated, when grown on glucose and GlcNAc, both strains consumed these sugars within 34 hours of growth (Fig. S2) (25).

Gene regulation of A. muciniphila Muc T and A. biwaensis on 2′-fucosyllactose and glucose
We performed an RNASeq analysis of both A. muciniphila Muc T and A. biwaensis CSUN-19 during mid-log growth on GlcNAc and either Glc or 2′-FL.Because the strains grew at different rates in the different media backgrounds, RNA was extracted from cells harvested during mid-log growth (shading, Fig. 2; Fig. S2).Importantly, 2′-FL was present in the culture media at the time of sampling, indicating that consumption of this HMO was ongoing.The quality and quantity of all RNA extracts varied but sequencing yields were consistent across samples averaging approximately 15 million reads per sample (Table S4).On average, across all samples, 38% of the total reads were aligned to coding sequences (CDs) of each genome.Compared to growth on glucose, 146 of 2,115 (~6.9%) genes were differentially expressed by A. muciniphila Muc T when grown on 2′-FL, of which 51 were downregulated and 95 were upregulated (Fig. S3).Analysis of A. biwaensis CSUN-19 identified 507 out of 2,557 (~21%) differentially expressed genes when grown on 2′-FL versus glucose, of which 302 were downregulated and 225 were upregulated.Initial analysis of the top 20 most differentially expressed genes for each strain revealed limited overlap (highlighted rows, Tables S5 and S6).Within the top 20 most differentially expressed genes of A. muciniphila Muc T , multiple GHs were identified including a GH2 (putative beta-galactosidase), GH16 (active on β−1,4 or β−1,3 glycosidic bonds), and a GH88 (putative glucuronyl hydrolase), as well as multiple hypothetical proteins, a putative surface layer protein, and two NADPH-dependent oxidoreductases.By contrast, for A. biwaensis CSUN-19, none of the top 20 most differentially expressed genes encoded for GHs, instead constituting mostly genes related to iron transport and annotated as hypothetical proteins.
An analysis of the predicted fucosidases, GH29 and GH95 families, revealed that in A. muciniphila Muc T only Amuc_0846 was significantly upregulated when grown on 2′-FL (log2 fold change >|2|, P adj <0.05), and none of the GH95 genes were significantly regulated (Table 2).For A. biwaensis CSUN-19, three GH29 genes were significantly upregulated (log2 fold change >|2|, P adj <0.05), and one GH29 and one GH95 gene were significantly downregulated.The β-galactosidase family GH2 is also likely to be involved in lactose metabolism once fucose is liberated from 2′-FL.Our analysis indicated that two genes, Amuc_1666 and Amuc_0539 were significantly upregulated in A. muciniphila Muc T during growth on 2′-FL, whereas GH2 genes in A. biwaensis CSUN-19 were not significantly upregulated.

Genomic loci of GH2 and GH29/GH95 genes across human-associated Akkermansia
The differences in 2′-FL degradation across species may also be due to genetic variation between the strains; therefore, we compared the genomic locus of A. biwaensis CSUN-19 containing the most differentially upregulated GH29 (HHJ01_10880) across all four known human-associated Akkermansia species (Fig. 3).Surprisingly, A. biwaensis CSUN-19 was the only species containing this GH29 (HHJ01_10880), agreeing with our phyloge netic analysis.However, all four species have a GH2 β-galactosidase (A.muciniphila Muc T = Amuc_0824; A. biwaensis CSUN-19 = HHJ01_10865) within the same genomic locus that may be involved in the cleavage of the lactose after removal of the terminal fucose.In addition to the GH2, A. biwaensis CSUN-19, A. massiliensis CSUN-17, and Akkermansia sp.CSUN-56 have genes in this locus encoding a hypothetical protein that is likely a putative sialidase upon further BLASTp search (blue, Fig. 3).These same three species  that retain the putative sialidase also encode a putative α-galactosidase approximately 7 kb upstream (yellow, Fig. 3).Interestingly, A. muciniphila Muc T is missing all three of these putative glycoside hydrolases.

Full-Length Text
Journal of Bacteriology of HHJ01_10880 with related GH29 enzymes from other intestinal bacteria suggest that this enzyme is a GH29B enzyme (Fig. S5) and may require an additional binding of a second sugar.To determine the role of HHJ01_10880 in 2′-FL catabolism, we used a complementary approach of expressing the A. biwaensis putative fucosidase in A. muciniphila Muc T .A transposon-based HHJ01_10880 expression construct was delivered by conjugation into A. muciniphila Muc T .The insertion sites of the Tn insertions from four clones were confirmed, and all four clones were used to assess growth in BTTM + 0.4% mucin supplemented with or without 2′-FL across a 48-hour period (Fig. S6).
One clone (Akk-EH114) had comparable growth to A. muciniphila Muc T in background conditions but grew significantly better on 2′-FL (Fig. 4), similar to the growth of A. biwaensis CSUN-19 on 2′-FL.Together, these results suggest this enzyme is involved in 2′-FL catabolism.

DISCUSSION
One important function of HMOs is to facilitate the expansion of beneficial bacteria in the gut microbiota early in life.Recently, different species of human-associated Akkermansia have been shown to grow using HMOs (26,27), expanding their ecological niche and potential importance in the development of the human gut.The mechanisms by which Akkermansia uses HMOs, specifically the highly abundant 2′-FL, may affect how they interact in and shape both the composition of the infant gut microbiota and the development of the infant.Here, we used comparative genomics, phenotypic and transcriptomic profiling, and enzyme activity assays to explore the mechanisms of 2′-FL catabolism in human-associated Akkermansia species.We found that strains of A. biwaensis possess a higher diversity of both putative fucosidase (GH29 and GH95) and β-galactosidase (GH2) encoding genes than other Akkermansia species.Despite this genomic difference, both A. muciniphila Muc T and A. biwaensis CSUN-19 grew in 2′-FL in media lacking mucin.Furthermore, both strains significantly induced the expression of homologs of a gene encoding a GH29 fucosidase (clade 1), for which the purified protein from A. muciniphila has previously been shown to have weak activity against para-nitro phenyl-α-L-fucoside, an analog of 2′-FL (32).Interestingly, a gene unique to A. biwaensis strains encoding a different GH29 was also significantly upregulated on 2′-FL.Impor tantly, expression of this GH29 in A. muciniphila was sufficient to enhance its growth to levels similar to A. biwaensis in 2′-FL.An additional gene encoding a putative GH2 gene in the same genomic region as the differentially expressed GH29 was confirmed to be a β-galactosidase which is needed to fully deconstruct the 2′-FL trisaccharide.Surprisingly, A. biwaensis had a significantly downregulated GH95 fucosidase despite its importance in 2′-FL catabolism.Overall, these findings highlight the different strategies used by Akkermansia species to deconstruct fucose-containing HMOs and point to functional differences that may influence the colonization success and health impacts of Akkerman sia at different host life stages.

GH29 and GH95 family proteins are involved in 2′-FL catabolism
Analysis of putative fucosidase and β-galactosidase genes across Akkermansia species revealed differences in the number and diversity of GH29, GH95, and GH2 genes that were phylogroup specific.Recently, six fucosidases (four GH29 and two GH95 enzymes) from A. muciniphila Muc T have been characterized, demonstrating each enzyme has specific activity on different fucosylated substrates that are diet-or host-derived (32).These corroborate previous work demonstrating that members of the GH95 family preferentially hydrolyze 1,2-α-L-fucosylated linkages (70).By contrast, the other dominant fucosidase family, GH29, is divided into two subfamilies, GH29A and GH29B (49,50), in which Group A has little substrate specificity and GH29 Group B prefer entially hydrolyzes α−1,3/1,4-L-fucosidase linkages (71,72).Typically, GH29A enzymes show broader linkage specificity than the GH29B enzymes that show preference for α−1,3/4 fucosylated linkages.However, these linkage preferences are not absolute and many GH29B enzymes have broad substrate activities.Furthermore, we find that A.
massiliensis CSUN-17, Akkermansia sp.CSUN-56, and A. biwaensis CSUN-19 contain more copies of both GH29 and GH95 genes than A. muciniphila Muc T , suggesting the capacity to metabolize a broader set of fucosylated substrates or under a broader range of conditions, such as pH or temperature.Several studies have demonstrated the optimal performance of fucosidases, sialidases, neuraminidases, β-galactosidases, and β-acetyl hexosaminidases from multiple organisms, including Akkermansia, at specific pH values (26,72,73) and fucosidases with different activity depending on the temperature (74).
Given that the local pH will vary along the length of the infant intestine, it is therefore necessary to understand the conditions under which these enzymes are expressed to model colonization dynamics (75).Fucosidase expression patterns observed in A. muciniphila Muc T and A. biwaensis CSUN-19 under 2′-FL growth suggest that GH29 are the primary enzymes responsible for the cleavage of the terminal fucose residue.In our phylogenetic analysis, genes from clade 1 were significantly upregulated in both species.Interestingly, this protein in A. muciniphila Muc T , Amuc_0846, has weak activity on 2′-FL but strong mucin binding activity (32).Also in agreement, both strains downregulated (significantly only for A. biwaensis CSUN-19) another GH29 belonging to clade 7. Previous reports of this purified protein from A. muciniphila Muc T , Amuc_0010, demonstrated very weak activity of this purified enzyme against 2′-FL (26, 32).Our results further support these findings suggesting that this fucosidase may not be the primary enzyme for 2′-FL catabolism.
While there were similarities in fucosidase expression patterns across species, there were also notable differences.Most notable was the significant upregulation of HHJ01_10880 in A. biwaensis CSUN-19, a predicted GH29B fucosidase absent in other Akkermansia species.The purified HHJ01_10880 did not display fucosidase activity using both a substrate analog and the synthetic 2′-FL used in growth experiments.However, other purified GH29B enzymes from intestinal bacteria are not active on chromogenic substrates and often have a second binding pocket that must be occupied by a second sugar for efficient fucosidase activity (70).For example, the GH29B enzymes that share sequence similarity to HHJ01_10880 (Fig. S5) from Bacteroides thetaiotaomi cron (BT_2192) and Ruminococcus gnavus (E1_10125) bind β-D-galactose and sialic acid, respectively, during catabolism of fucosylated substrates (51,54).Thus, this inability to detect the activity of purified HHJ01_10880 was not unexpected, and further work is needed to characterize fucosidase activity with HMO sugar combinations.

Genomic linkage of GH genes involved in HMO metabolism
For Akkermansia species CSUN-19, the most differentially expressed GH29 when grown on 2′-FL was the predicted fucosidase HHJ01_10880.Interestingly, in the same genomic region as HHJ01_10880, there is also a β-galactosidase 3 genes downstream (Fig. 3), albeit on the opposite DNA strand.While this β-galactosidase was not differentially expressed under 2′-FL growth conditions, it was among the most highly expressed genes for both organisms (Table 2).In other Gram-negative HMO-degrading bacteria including Bacteroides, HMO-active GH genes are genomically co-localized and co-regu lated in polysaccharide utilization loci (PUL) to efficiently respond to available carbohy drates (23).PULs also contain substrate-binding proteins and transport proteins for the target polysaccharides.Akkermansia do not have this typical PUL organization despite several GHs mapping to this region.However, given the regional proximity and high abundance of both transcripts when grown on 2′-FL, we propose a model in which the HHJ01_10880 cleaves the fucose by breaking the α1-2 bond, leaving the residual lactose (β1-4) to be cleaved by the adjacent β-galactosidase (HHJ01_10865).Based on the presence of signal peptide sequences, we predict that both the GH29 and the GH2 operate in the periplasmic or extracellular space (26,67).With this proposed model, after extracellular cleavage of fucose, the lactose would have to be further processed by the β-galactosidase before cellular uptake.Supporting this model, the genome of A. biwaensis CSUN-19 lacks an ortholog of the LacY permease, suggesting lactose is not likely imported into the cytoplasm.While this is the first report of introducing and expressing genes in A. muciniphila based on the recently developed genetic modification system (67), expansion of this method to include targeted gene knockouts in A. biwaensis is eagerly anticipated to confirm this model.Alternatively, understanding the movement of sugar in these systems would be greatly improved by labeled substrates that can be tracked in real time through the cell.

Conclusions
The presence of Akkermansia in infants may be correlated with its ability to utilize HMOs.In this study, Akkermansia isolates from two divergent species, A. muciniphila and A. biwaensis, were shown to have different growth dynamics on 2′-FL, an abundant HMO in breastmilk.Phylogenetic analysis of putative fucosidases (GH29 and GH95) and β-galactosidases (GH2) involved with the deconstruction of 2′-FL coupled with transcriptional profiling pointed to two genes in a genomic locus that may be responsi ble for enhanced growth of A. biwaensis.We provide evidence that a GH29 present in A. biwaensis but absent in A. muciniphila is most likely responsible for this phenotype and that both this GH29 and an associated β-galactosidase perform their activities extracellularly (26).In addition to serving as a substrate for Akkermansia GH2 enzymes, in a complex ecosystem, the released lactose could also serve as a substrate for other members of the infant microbiota.A healthy adult gut microbiome begins in early life where it is shaped by human milk sugars through glycan-degrading microbes.The difference in Akkermansia species to break down HMOs suggests a speciesspecific role in the development of a healthy microbiome.This is significant because it points to the strainspecific mechanisms by which Akkermansia may colonize and shape the infant gut, ultimately leading to the rational selection of probiotic bacteria and prebiotic therapies.

FIG 1 Full
FIG 1 Maximum likelihood trees reveal seven clades of GH29 (A), three clades of GH95 (B), and eight clades of GH2 (C) genes across human-associated Akkermansia species.Sequences marked by circles represent genes from A. muciniphila Muc T (AmI) and sequences marked by diamonds are genes from A. biwaensis CSUN-19 (AmIV), the representative strains used throughout this study.All trees were generated using a maximum likelihood heuristic method by executing Neighbor-Join and BioNJ (NJ/BioNJ) algorithms and applied to a matrix of pairwise distances with 100 bootstrap iterations.Trees are drawn to scale, with branch lengths measured in the number of substitutions per site.The trees in (A) involved 78 nucleotide sequences and a total of 3,258 positions in the final data set.For the tree in (B), 44 nucleotide sequences and 2,735 sites were analyzed, while in (C), 106 nucleotide sequences and 4,919 positions were included.Evolutionary analyses were conducted using MEGA11(44)(45)(46).

FIG 2 A
FIG 2 A. muciniphila Muc T and A. biwaensis CSUN-19 use 2′-FL and GlcNAc for growth.Fucose, lactose, glucose, and galactose, breakdown products of 2′-FL degradation, are detectable in culture media during different stages of growth.Note the differences in time along the y-axis.HighlightingAsterisks denotes time points where cells were collected for RNAseq.Error bars represent the standard deviation of triplicate cultures.

TABLE 1
Presence of putative fucosidase (GH29 and GH95) and beta-galactosidase (GH2) genes in representative strains of the four species-level phylogroups reveals a greater suite of predicted fucosidases and beta-galactosidases in phylogroup IV (A. biwaensis) of human-associated Akkermansia a

TABLE 2 A
subset of glycoside hydrolases are regulated in the presence of 2′-FL a RNAseq analysis of the expression of genes encoding glycoside hydrolases potentially involved in 2′-FL metabolism including fucosidases, such as GH29 and GH95, and β-galactosidases, such as GH2.Fold change (log2FC) in gene expression compared to mid-log growth on glucose for both A. muciniphila Muc T and A. biwaensis CSUN-19.Average Expression for these genes is based on growth across both conditions.Shaded rows indicate genes that are significantly regulated by 2′-FL.