A genetic system for Akkermansia muciniphila reveals a role for mucin foraging in gut colonization and host sterol biosynthesis gene expression

Abstract

Akkermansia muciniphila, a mucophylic member of the gut microbiota, protects its host against metabolic disorders. Because it is genetically intractable, the mechanisms underlying mucin metabolism, gut colonization, and its impact on host physiology are not well understood. Here, we developed and applied transposon (Tn) mutagenesis to identify genes important for intestinal colonization and for the use of mucin. An analysis of Tn mutants indicated that de novo biosynthesis of amino acids was required for A. muciniphila growth on mucin medium and that many glycoside hydrolases are redundant. We observed that mucin degradation products accumulate in internal compartments within bacteria in a process that requires genes encoding pili and a periplasmic protein complex, that we term mucin utilization loci (MUL). We determined MUL genes were required for intestinal colonization in mice but only when competing with other microbes. In germ-free mice, MUL genes were required for A. muciniphila to repress genes important for cholesterol biosynthesis in the colon. Our genetic system for A. muciniphila provides an important tool to uncover molecular links between the metabolism of mucins, regulation of lipid homeostasis and potential probiotic activities.

Main

Akkermansia muciniphila is a Gram-negative bacterium that colonizes the human gastrointestinal (GI) tract and has been associated with various beneficial health effects, including protection from metabolic disease¹, neurological disorders², infection³, and an enhanced response to cancer immunotherapy⁴. Some factors that contribute to A. muciniphila’s metabolic and immunomodulatory activity have been identified^1,5,6, however most of the genes that contribute to colonization of the GI tract and regulate interactions with the host remain to be identified.

A. muciniphila uses mucins as its preferred nutrient source. Mucins are large, highly glycosylated proteins that make up the bulk of the intestinal mucus layer that separates the host epithelia from the microbiota⁷. The predominant mucins of the GI tract, Muc2 in the colon and Muc5AC in the gastric mucosa, consist of a protein backbone rich in serine, threonine, and proline residues, decorated with O-linked glycans. The complex linkages of mucin glycans and terminal sulfation make them a challenging substrate for microbes to degrade. Nonetheless, a range of intestinal microbes display enhanced growth in media supplemented with mucin⁸. A. muciniphila is particularly adept at using mucins, reaching a high optical density when cultured in liquid medium with gastric mucin as a sole carbon and nitrogen source (Extended Data Fig. 1a). A. muciniphila encodes multiple putative glycoside hydrolases (GH) and other mucolytic enzymes, some of which have been characterized in vitro^9–16. A. muciniphila’s repertoire of mucin degrading enzymes includes GHs belonging to the GH20^14,17, GH35^15,16, and GH16¹² enzyme families, an aspartic protease¹³, peptidases^9,10, and an O-glycopeptidase¹¹. While several A. muciniphila GHs have been characterized, the intractability of Akkermansia spp. to molecular genetic manipulation has limited our understanding of what role the metabolism of mucin by this microbe plays in the colonization.

Here, we apply Tn mutagenesis to identify genetic determinants of mucin utilization and colonization in A. muciniphila. We found that A. muciniphila accumulates mucin in intracellular structures and identified a mucin transport system required for mucin uptake. Using mutants unable to grow on mucin, we show that access to mucin is required for A. muciniphila to compete against other gut microbes and for stable engraftment in the GI trac. Finally, we demonstrate that mucin metabolism represses the expression of lipid biosynthesis genes in the colon, revealing a link between this key characteristic of A. muciniphila and host physiology.

Results

Akkermansia accumulates mucins in intracellular compartments

To visualize mucin uptake, we used fluorescein (FL)-labeled mucin glycans generated from porcine gastric mucin. Commercial porcine gastric mucin, which is pepsin-digested and partially purified, consists of 13 major O-linked glycans composed of the mucin sugars N-acetylglucosamine, N-acetylgalactosamine, galactose, and fucose¹⁸. Additional mucin modifications include sialylation and sulfation. We observed an accumulation of these labeled mucins or their degradation products in intracellular compartments using super resolution STED microscopy (Fig. 1a, Extended Video S1). The formation of these intracellular structures, which we operationally term “mucinosomes”, appear to be specific to Akkermansia spp., as we did not observe them in Bacteroides thetaiotaomicron, which also metabolizes mucin (Extended Data Fig. 1b).

Figure 1. A. muciniphila accumulates mucin glycans and requires amino acid biosynthesis for replication in mucin as a sole carbon and nitrogen source.

(a-c) Mucin accumulates within A. muciniphila intracellular compartments. (a) STED imaging of A. muciniphila showing mucin glycans or mucin degradation products within intracellular compartments (“mucinosomes”). Bacteria grown with fluorescein (FL) labeled mucin (purple) were stained with anti-Akkermansia antisera (anti-Akk, cyan). The lower panel shows orthogonal views of A. muciniphila (white lines denote the orthogonal plane) and the right panel shows a 3D reconstruction of a single cell. Images are representative of three independent experiments. (b) Flow cytometry of A. muciniphila incubated with FL-mucin over a 24 h. (c) Live cell imaging over 3 min shows accumulation of FL-mucin (green) at the A. muciniphila cell pole (Extended Video S2). (d-f) Amino acid biosynthesis is required for optimal growth of A. muciniphila on mucin. A complex Tn mutant pool was grown in synthetic medium (Input) or mucin medium for eight generations (Output). Each dot represents the pooled Tn insertions for each gene as determined by INSeq. Unadjusted P values were determined by Mann-Whitney U-test. (d). A comparison of the transcriptional response to growth on mucin (RNAseq) to the fitness of the corresponding mutants (IN-Seq) (e). Each dot represents one gene, and only genes that were detected in both the INSeq and RNAseq datasets are shown. RNAseq represents the Log₂ fold change in expression in mucin vs synthetic media, while INSeq represents the Log₂ fold change in the abundance between input Tn mutant pools versus the output. Dashed lines represent no change. (f) Overrepresentation analysis to identify differentially abundant KEGG pathways (Log₂ fold change > 1.5, p < 0.05) detected by INSeq and RNAseq following growth in mucin. P values were calculated by hypergeometric distribution. (Scale bar =1μm)

To examine the kinetics of mucin uptake, we pulsed A. muciniphila with FL-labeled mucin and analyzed fluorescence associated with bacteria over time by flow cytometry. The association with FL-mucin occurred rapidly and peaked by 6h post labeling, followed by a subsequent decrease between 6 and 24h (Fig. 1b). Next, we used live-cell imaging to monitor the uptake of FL-mucin in real-time (Fig. 1c). Mucinosome formation occurred within minutes and was a highly dynamic process (Extended Video S2). These structures did not form after incubation with FL-labeled dextran (Extended Data Fig. 1c) and were diminished in the presence of the ionophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) which dissipates the cell’s proton motive force (Extended Data Fig. 1d-e, Extended Video S2). CFDA, a cell-permeable dye that fluoresces green upon cleavage by esterases, confirmed that cells remained viable during CCCP treatment and continued to passively take up molecules (Extended Data Fig. 1f). Overall, these observations suggest that the acquisition of mucin glycans by A. muciniphila is a selective and energy-dependent process.

Development of a system for Tn mutagenesis

To define the genetic requirements for growth on mucin, we developed a system for Tn mutagenesis after identifying a suitable selectable marker (chloramphenicol (Cm) acetyl transferase), defining conditions for DNA conjugation with an E. coli donor, and codon-optimizing the Himar1c9 transposase¹⁹ for expression in Akkermansia (Extended Data Fig. 2a-b). The absence of plasmid sequences in A. muciniphila Cm^R colonies confirmed that transposition had occurred (Extended Data Fig. 2c) and a Southern blot analysis of total DNA extracted from these strains indicated that most mutants had a single Tn insertion (Extended Data Fig. 2d).

We assembled two mutant libraries, one consisting of a mixed pool of Tn mutants and one where individual Tn mutants were arrayed in 96-well plates. The Tn insertion sites in both pools were identified by DNA sequencing on an Illumina platform as previously described¹⁹. The pooled Tn mutant library consisted of 2680 unique insertions in 721 genes (33% of the total CDS), 198 intergenic regions, and 5 tRNAs. The arrayed Tn library consisted of 1063 unique insertions in 406 genes, 58 intergenic regions, and two tRNA genes. We then used a Cartesian pooling strategy²⁰ to map the locations of the majority of the Tn insertions to individual mutants in the arrayed collection.

Proteins of unknown function enable growth on mucin

To identify genes required for growth in mucins, Tn mutants from the mixed pool were grown for eight generations in a defined medium with gastric mucin as the sole source of carbon and nitrogen²¹. The impact of mutations in each gene on bacterial growth was expressed as the Log₂ fold change between the normalized abundance in the input inoculum pool and the output pool, where negative values are indicative of mutants with growth defects. We observed a significant decrease in the abundance of mutants with Tn insertions in genes required for amino acid biosynthesis, particularly branched chain amino acids and Arg (Fig. 1d, Extended Data Fig. 3a and Supplementary Data 1). This result was not surprising given that these amino acids are not abundant in the protein core of GI mucins²². Mutants with Tn insertions in genes required for assimilatory sulfate metabolism, which are required for Cys and Met biosynthesis²³, were also significantly depleted. The growth defect of these mutants could be rescued by the addition of protein hydrolysates (Extended Data Fig. 3b and Supplementary Data 1). These results indicate that A. muciniphila relies on de novo biosynthesis of amino acids during growth in mucin, presumably because some amino acids in the mucin protein backbone become rate limiting to support the bacterium’s anabolic needs.

We anticipated that GHs would be required for A. muciniphila growth on mucin. A. muciniphila strain Muc^T is predicted to encode 60 GHs, belonging to 24 different families. Our Tn mutant collection included insertions in genes encoding 38 GHs, representing 20 families²⁴. Following growth in mucin medium, only 8 GHs had a Log₂ fold change decrease in abundance > 2, indicating that most of GHs (30/38) were not essential under these conditions. Several mutants displayed a slight growth advantage, including mutants in genes encoding previously characterized enzymes such as a β-galactosidase¹⁵, a β-hexosaminidase^25,26, a β-N-acetyl hexosaminidase¹⁴, and a predicted mucin binding chitinase²⁷. Nonetheless, mutations in a subset of GH genes resulted in profound growth defects in mucin medium (Extended Data Fig. 3c-d and Supplementary Data 1), including genes predicted to encode a β-galactosidase (GH2), an α-N-acetylglucosaminidase (GH89), an α-amylase (GH13), a galactosidase (GH43), and a β-hexosaminidase (GH20). Mutants in genes encoding enzymes that catalyze the initial steps in mucin breakdown had particularly strong growth defects and included a sialidase (GH33)²⁵, a fucosidase (GH95)²⁵, and an outer membrane-associated endo O-glycanase which can break down mucin in vitro (GH16)¹².

The overlap between genes required for mucin utilization by A. muciniphila as determined by metabolic modeling²⁸ and INSeq analysis was low, suggesting that most of the enzymes predicted to have roles in mucin degradation are functionally redundant (Supplementary Data 1).

It is possible that the low representation of GH mutants among strains defective for the consumption of mucin is the result of trans complementation by neighboring cells. To test if cross-feeding in batch cultures led us to miss important factors needed for mucin utilization, we performed Droplet Tn-Seq²⁹ to analyze A. muciniphila Tn mutants grown in microdroplets. A comparison of the abundance of mutants grown in batch culture versus microdroplets indicated that the growth defects were largely independent of whether the bacteria were grown in isolation (Extended Data Fig. 3e and Supplementary Data 1), suggesting that any cross-feeding is limited. Exceptions to this observation include genes involved in capsule production and genes encoding a predicted sialidase, fucosidase, and sulfatase — all enzymes that can act on the terminal motifs of mucin glycans. These results are consistent with prior findings indicating fucosidase activity is associated with the Akkermansia outer membranes ³⁰.

We next performed an RNAseq analysis of A. muciniphila cultured in the same mucin medium used for INSeq selections and in the mucin-free synthetic medium used to propagate the Tn mutant libraries. In response to mucin, 103 genes were significantly upregulated (fold change > 4, adjusted P value < 0.05), and a gene set enrichment analysis showed that pathways involved in glycan degradation and galactose metabolism dominated this response (Fig. 1e-f) as previously observed²⁸. Genes that were highly expressed in response to mucin included an O-glycopeptidase⁹, putative sulfatases, and 24 GH enzymes. The correlation between the genes required for growth in mucin and those that were strongly upregulated by mucin was low (Fig. 1e-f). These observations indicate that, unlike the polysaccharide utilization (Pul) genes of Bacteroidetes³¹, the Akkermansia genes that are essential for growth in mucin are not preferentially induced by mucins.

Approximately 35% of the A. muciniphila genome encodes proteins with no predicted function²⁷. We mapped Tn insertions to 226 genes annotated as encoding either ‘conserved hypothetical’ or ‘hypothetical’ proteins, which accounts for 31.3% of all the mutated coding sequences in our pooled library. Mutations in 54/179 genes encoding hypothetical proteins that were either Akkermansia specific or had homologs restricted to the PVC superphylum led to decreases in abundance of > 2-fold during growth in mucin medium (Extended Data Fig. 4a, Supplementary Data 1). Manual curation of these proteins of unknown function for associated Pfam³² motifs and Conserved Domains³³ indicated that 19% contained either a tetratricopeptide repeat domain, which mediate protein-protein interactions³⁴, or were associated with pili assembly (Extended Data Fig. 4b-d, Supplementary Data 1). Proteins potentially involved in pili or type II secretion included Amuc_1100 and Amuc_1102, which have been characterized for their anti-obesity activity and whose structures have been solved^35,36. Overall, these findings suggest that mucin utilization by A. muciniphila requires genes encoding pili-like proteins and several poorly characterized hypothetical proteins whose expression is not regulated by mucin.

Mucin metabolism enables Akkermansia to compete in the gut

To identify A. muciniphila genes required to colonize the GI tract, we performed an INSeq analysis of the Tn arrayed mutant pool in four mouse models: germ-free (GF), mice stably colonized with the eight-member Altered Schadler Flora (ASF), conventionally raised (CONV) mice, and Muc2^−/− mutant mice lacking the most prominent intestinal mucin³⁷. Mice raised with a conventional microbiota had to be pre-treated with tetracycline to displace their endogenous Akkermansia, which prevents the engraftment of the Muc^T A. muciniphila strain³⁸. Mice were then orally gavaged with a pool of Tn mutants. The relative fitness of the Tn mutants in the cecum or after growth in mucin medium was assessed by INSeq (Fig. 2a).

Figure 2. The metabolic requirements for A. muciniphila to colonize the GI tract increase as the host microbiota becomes more complex.

(a) Chromosomal location of genes with Tn insertions that produced a significant change in abundance after growth in mucin medium (Mucin), or in the cecum of the following mouse models: germ-free (GF), Altered Schaedler Flora mice (ASF), conventional (CONV), and Muc2-deficient (Muc2^−/−). Lines represent the location of Tn inserts that result in growth defects (Log₂ fold change > 5, p < 0.05, unadjusted Mann-Whitney U-Test). Dots along the outer ring indicate locations of all Tn insertions and numbers represent chromosomal location (bp). (b) KEGG pathways overrepresented (q < 0.05) among mutants that displayed significant fitness defects. (c) Global analysis of the abundance of mutants with fitness defects in mucin medium as determined by INSeq and analyzed with Omics Dashboard⁷⁹. The larger nodes represent the mean Log₂ fold change for components of a metabolic pathway, and the smaller nodes represent individual genes within each pathway. The boxed figure shows the mean Log₂ fold change for individual amino acid biosynthesis pathways. (d) Venn diagram showing overlapping genes required for growth under various conditions. For each condition, genes with a Log₂ fold change in abundance >1 were compared. The number of shared genes and the percentage of the total mutant pool are shown for each condition. (e) Omics Dashboard analysis of genes encoding putative cell surface components. Genes with annotations corresponding to predicted functions in capsule and exopolysaccharide biosynthesis were manually curated (Exopolysacchar). Additional functions related to plasma membrane biogenesis (Plasma Mem), cell wall biogenesis (Cell Wall Gen), and transport of amino acids, carbohydrate, and ions across the cell wall (Transport) were predicted in BioCyc. (f) Detailed view of exopolysaccharide biosynthesis or capsule genes and the relative fitness of respective Tn mutants in each environment. (g) INSeq analysis of genes in the assimilatory sulfate reduction (ASR) pathway. Heatmaps represent the Log₂ fold change of input versus output for mutants in the pathway.

While growth in mucin medium and mice are not directly comparable due to contributions from the host environment, diet and the microbiota, our analysis indicated that approximately half of the genes whose disruption led to a > 2-fold decrease in abundance in the GI tract of mice were also required for growth in mucin (Fig. 2a). As the complexity of the microbiota increased from GF to CONV mice, additional genes became conditionally essential (Fig. 2a-e). As we observed in mucin medium, Tn insertions in genes required for amino acid homeostasis were significantly overrepresented among mutants that failed to colonize the GI tract (Fig. 2b, c). In addition, mutants defective for Ala, Asp, Glu, and Arg biosynthesis were depleted in ASF and CONV mice (Fig. 2c-e). Mutants in the glyoxylate pathway, particularly components of the Gly cleavage system, also had significant growth defects. The Gly cleavage system generates 5,10-methylene-tetrahydrofolate, a one carbon donor used in Ser biosynthesis, which may be particularly important for colonization of Muc2^−/− mice because they lack this Ser-rich mucin. Overall, these findings indicate that the synthesis of key amino acids, especially in the context of a complex microbiota, is required for A. muciniphila to successfully colonize the GI tract.

Metabolic and cellular pathways overrepresented in A. muciniphila mutants with fitness defects in mice indicated that Tn insertions in central processes such as transcription, protein metabolism, and RNA metabolism led to similar growth defects in the GI tract as in mucin medium (Fig. 2c). In contrast, several genes were required for survival in the GI tract, but not for growth in mucin medium. These included putative exopolysaccharide and capsule biosynthesis genes and a locus harboring glycosyl transferases (Fig. 2f), which are orthologous to the Bacillus biofilm genes epsH and epsJ ³⁹. These findings are consistent with A. muciniphila cell surface modifications playing a prominent role in intestinal colonization.

We identified eight genes that were selected against in mucin medium and GF mice but rescued in mice with a microbiota (Fig. 2d). Three of these genes encoded components of the assimilatory sulfate reduction (ASR) pathway (Fig. 2g), which is required for the reduction of sulfate to hydrogen sulfide and the biosynthesis of cysteine, methionine, and sulfur containing metabolites. Mutants in a fourth gene in the ASR pathway, Amuc_1298, had strong growth defects under all conditions. The growth defect of these ASR^- mutants was partially complemented in GF mice, indicating that A. muciniphila can harvest hydrogen sulfide or cysteine from the host or its diet. In both ASF and CONV colonized mice, ASR mutants had a slight growth advantage (Fig. 2g), but this was context dependent as ASR mutants displayed significant growth defects in Muc2^−/− mice (Fig 2g).

Tn insertions in hypothetical genes accounted for 23.6% (48/203) of the mutants with strong growth defects (Log₂ fold change > 5) in GF mice and 27.3% (67/245) in CONV mice. Many of the hypothetical proteins required for growth on mucin medium were also required to colonize the GI tract, with the greatest growth defects occurring in genes encoding proteins with predicted N-methyl domains, a hallmark of pili proteins, or TPR domains (Extended Data Fig. 4c-d).

Two genetic loci define a mucin transport system

We identified a cluster of Tn mutants with strong mucin growth defects in and around Amuc_0544, which encodes a predicted secreted protein of unknown function found only in Akkermansia and other Verrucomicrobia (Fig. 3a). Amuc_0544 has multiple TPR domains, which is reminiscent of the polysaccharide transport protein SusD in the Bacteroides Starch Utilization System (SUS)⁴⁰. Amuc_0543 and Amuc_0544 are part of a larger gene cluster that includes two genes annotated as putative ExbD biopolymer transporters, and an ExbB-like proton channel. ExbBD are components of TonB dependent transporters (TBDT), also found in SUSs, which transduce the proton motive force to transport substrates across the outer membrane⁴¹. Reads mapped by RNAseq indicate that the region is likely expressed as at least two mRNA transcripts (Extended Data Fig. 5a). We retrieved A. muciniphila strains with Tn insertions in Amuc_0543 and Amuc_0544 and confirmed that they were impaired for growth in mucin medium (Fig. 3b-e, Extended Data Fig. 5b), leading us to name the genomic region spanning Amuc_0543 to Amuc_0550 as Mucin Utilization Locus I (MUL1) (Fig. 3a).

Figure 3. MUL loci encode for a mucin transport complex in A. muciniphila.

(a) Map of the Mucin Utilization Loci, mul1 (Amuc_0543 – Amuc_0550) and mul2 (Amuc_1098 – Amuc_1102). Blue arrows represent genes with Tn inserts and their Log₂ fold change in abundance after culturing in mucin. Genes without Tn inserts are represented as grey arrows. Bars represent the mean normalized reads for each Tn insert in the input pool and the output after growth in mucin. (b-g) Characterization of A. muciniphila mul mutants defective for mucin utilization display different patterns of association with FL-mucin. Growth curves of mutants grown in synthetic or mucin medium (b-e), where mucin is the sole carbon and nitrogen source, and (f) flow cytometric analysis of fluorescein-mucin acquisition by mul mutants. Mean fluorescein intensity was quantified for bacteria detected with anti-Akkermansia antibodies. (g) Mutants lacking mul1A and mul2A, but not a galactose epimerase (Amuc_0029), failed to accumulate mucin or mucin degradation products in intracellular compartments. Cultures were grown in synthetic medium supplemented with FL-mucin prior to staining with anti-Akkermansia antibodies (anti-Akk). The insets show the corresponding image with enhanced brightness to visualize FL-mucin in mul1A and mul2A mutants. Images are representative of three independent experiments. The scale bar is 1 μm. (h) Multiple proteins associate with the TPR domain protein Mul1A. Mass spectrometric (LC-MS/MS) analysis of proteins that co-immunoprecipitated with Mul1A. Numbers refer to the gene ID (amuc). Node size reflects the Log₂ fold change in normalized spectral counts over immunoprecipitations performed with mul1A mutants and edge thickness is scaled to the -Log₁₀(P value). Nodes are color coded based on Pfam. Statistical analysis of peptides used two-tailed heteroscedastic t-test on Log2-transformed data. (i) Proposed model of the MUL transporter that imports mucin glycans or mucin degradation products. Mul1A and Mul1B form a complex with accessory proteins that include sulfatases, GHs, and potential inner and outer membrane transporters (sodium solute symporter (SSS) and TonB dependent transporter (TBDT)).

A second major group of genes required for mucin utilization were annotated as encoding pili, including type IV-like pili proteins encoded by the Amuc_1098 to Amuc_1102 locus^1,42. These proteins are among the most abundant produced by A. muciniphila⁴² which includes Amuc_1100, a TLR2 agonist^1,42 and modulator of host metabolism¹. The structure of Amuc_1102³⁵ indicates that it is related to archaeal type IV pili. Amuc_1101 is annotated as the cell division protein FtsA, but it belongs to the PilM Pfam, suggesting a more likely role in pili biogenesis. We renamed this locus, spanning amuc_1098 to amuc_1102, as MUL2 (Fig. 3a, Extended Data Fig. 5a) given that Tn insertions in MUL2 genes exhibited strong growth defects in mucin medium (Fig. 3c, Extended Data Fig. 5b, Supplementary Data 1). Additional potential mul loci are scattered throughout the genome (Fig. 2a), including genes encoding additional potential pili proteins with N-methylphenylalanine domains typical of type IV pilins, further suggesting a broader role for pili in mucin acquisition (Extended Data Fig. 4b-d).

The largest gene in the MUL1 locus, mul1A (Amuc_0544), is predicted to encode a protein with a type I signal peptide. Given its relatively large size and multiple predicted TPR domains, we hypothesized that Mul1A serves a structural role in the assembly of protein complexes important for the transport of mucin. We generated antisera specific for Mul1A and performed native immunoprecipitations from lysates of wild type A. muciniphila or mul1A::Tn mutants, coupled to quantitative mass spectrometry (Extended Data Fig. 5c). Mul1A showed cleavage of the N-terminal signal sequence confirming that Mul1A is secreted across the cytoplasmic membrane. The top interacting protein was Amuc_0543, designated Mul1B, which efficiently co-precipitated with Mul1A as a complex (Fig. 3h, Supplementary Data 1), indicating that both proteins function in the same step in the transport of mucin.

Peptides belonging to the pili subunit protein Amuc_1102 (mul2A), were also significantly enriched with Mul1A complexes suggesting that pili may play a key role in the acquisition of mucin (Fig. 3h). Similar observations have been made in the related bacteria Planctomycetes limnophila and Gemmata obscuriglobus, which bind dextran glycans with pili-like fibers⁴³. These findings also suggest that components of the MUL1 and MUL2 loci may cooperate to transport mucin fragments. We next tested if these mul mutants would be impaired for the import of mucins across the outer membrane. Indeed, Tn mutants in either mul1A or mul1B were unable to internalize FL-mucin (Fig. 3f-g). The loss of FL-mucin labeling in the mul mutants is not an indirect consequence of their inability to metabolize mucins as a Tn mutant in amuc_0029, encoding a UDP-glucose 4-epimerase in the Leloir pathway for galactose degradation⁴⁴, could not grow in mucin medium (Fig. 3e, Supplementary Data 1), yet displayed high intracellular staining with FL-mucin (Fig. 3f-g).

Mul1A interacted with several GH enzymes, two sulfatases, and two disulfide reductases. We also detected peptides from a putative sodium-solute symporter transporter (SSS), Amuc_0970, which in other microbes can import galactose⁴⁵ across the inner membrane. The SSS transporter is a potential route for mucin components to cross the inner membrane, a function performed by major facilitator superfamily (MFS) transporters in other PUL systems. A second putative transporter with a conserved outer membrane protein beta-barrel domain, Amuc_1687, was also enriched, suggesting a potential route for the transport of mucin glycan fragments (Fig. 3h, Supplementary Data 1), although the precise mechanism remains to be determined. Comparison of the mass spectrometry data to the transcriptional analysis in response to mucin indicates that core components of the MUL system are constitutively expressed, while MUL-accessory proteins such as GHs are inducible (Fig. 1e, 3h). Finally, we detected peptides belonging to the Von Willebrand type D domain and the C-terminal cysteine knot domain of Muc5AC specifically co-precipitating with anti-Mul1A antisera (Extended Data Fig. 5d), indicating that Mul1A or associated proteins can engage the non-glycosylated portion of the gastric mucin backbone. Based on these observations, we propose that the MUL system functions as a complex with associated pili (Fig. 3i), perhaps in conjunction with additional mul loci not yet identified, to capture extracellular mucins or mucin fragments and import them into the cell, where they ultimately accumulate in mucinosomes.

Mucin utilization modulates host transcriptional responses

We tested the requirement for the MUL system to colonize the GI tract in GF mice. Unexpectedly, when administered alone mul1A mutants colonized GF mice as efficiently as wild type A. muciniphila (Fig. 4a), although wild type A. muciniphila readily outcompeted mul1A and mul2A mutants when the strains were co-administered (Fig. 4b). We next tested the ability of mul mutant to colonize the GI tract in the presence of an intact microbiota by using a mouse colony that was naturally devoid of Akkermansia (Akk-free). These mice have an intestinal microbiota like that of conventional mice (Fig. 6a-d) and do not require pre-treatment with Tet to enable the engraftment of Akkermansia. Both mul mutants exhibited more pronounced defects in their ability to colonize mice with complex microbiotas (Fig. 4c). Similar results were obtained in antibiotic pretreated CONV mice (Extended Data Fig. 6e). We also performed competitions between wild type A. muciniphila and a mutant in amuc_0029, which binds to and acquires mucin, but cannot grow in it (Fig. 3e-g). Like the mul mutants, the amuc_0029 mutant was rapidly outcompeted by wild type A. muciniphila (Extended Data Fig. 6e), suggesting that the colonization defects of mul mutants are not simply because they can’t bind to mucins in the GI. Taken together, these findings indicate that the MUL system and mucin metabolism are required for A. muciniphila to establish residence in the mouse GI tract but only in the context of other microbes.

Figure. 4. Mucin utilization enables A. muciniphila to compete against members of the microbiota and leads to repression of genes in cholesterol biosynthesis.

(a-c) Mucin utilization gives A. muciniphila a competitive advantage. In GF mice, growth on mucin is not required for mono-colonization (a), but mul mutants are outcompeted by wild type A. muciniphila (WT) (b). Similarly, mul mutants were outcompeted in mice with a conventional microbiota that lacks A. muciniphila (Akk-free) (c). Each point represents the average A. muciniphila per gram of feces per cage (GF, n = 5; Akk-free, n = 6). (d) SCFA content of cecal contents from GF mice colonized with wild type A. muciniphila (WT, n=15), mul1A Tn mutants (n=13), or vehicle controls (PBS, n=12). Acetate and propionate were analyzed by One Way ANOVA (F statistic). The ratio of propionate to acetate was analyzed by a two-tailed student’s t test. Data are presented as mean values +/− SD. (e-h) Comparison of the transcriptional profiles of colonic tissue from GF mice colonized with mul1A mutants or wild type A. muciniphila. (e) RNAseq of colon tissue. Each point represents a gene (grey), and colors indicate a Log₂ fold change > 1 (green, dashed line), a Benjamini-Hochberg adjusted P value < 0.05 (blue, dashed line), or both (red). Significance was determined using the Wald test. (f) MetaScape enrichment plot and (g) network visualization showing pathways enriched in mice colonized with the mul1A mutants. Node colors correspond to cluster annotation and edge thickness denotes relatedness of the pathways. BH adjusted P values were determined in MetaScape with a hypergeometric test. (h) Differential expression of genes along the mevalonate and cholesterol biosynthesis super pathways visualized with BioCyc. Enzyme names are color coded to indicate relative expression levels between mice colonized with mul1A mutants versus wild type A. muciniphila. (i) Expression of selected genes in response to mul1A mutants (Wald test with BH adjusted P values). (LoD = limit of detection)

Colonization by mul1A mutants in GF mice indicates that A. muciniphila can access other nutrient sources and is consistent with observations that Muc2^−/− mice are colonized by an endogenous Akkermansia strain. We found that both wild type and mul1A or mul1B mutants colonized the GI tract of Tet treated Muc2^−/− mice (Extended Data Fig. 6f). Thus, the requirement for mucin to colonize the GI tract is context dependent and Akkermansia can use other nutrients from the diet, host, or the microbiota. However, mul mutants were still outcompeted by wild type A. muciniphila in Muc2^−/− mice (Extended Data Fig. 6g), suggesting that the MUL system is required to scavenge other related glycoproteins.

We reasoned that mul1A mutants would produce different fermentation products than wild type A. muciniphila since it produce varying ratios of acetate and propionate in response to different carbon sources²⁸. We found that GF animals colonized with the mul1A mutant or wild type A. muciniphila differed in their cecal short chain fatty acids (SCFA) levels. GF mice colonized with wild type A. muciniphila had higher levels of total acetate and propionate compared to mice colonized with the mulA mutant (Fig. 4d) but a lower ratio of acetate to propionate (Fig. 4d). It is unknown if these differences reflect SCFA production by A. muciniphila or adsorption by the host. Thus, mucin utilization could impact host physiology either as a direct result of mucin consumption or through metabolites produced by A. muciniphila during mucin metabolism.

Differential expression of host genes in response to various components of A. muciniphila, including live and pasteurized cultures⁴⁶, purified surface proteins¹, and conditioned supernatants⁴⁷ have been reported. Because wild type and mul1A mutants colonized the GI tract of female GF mice at similar levels (Extended Data Fig. 7a), we chose to test the host’s transcriptional responses to mucin foraging by A. muciniphila in these mice. An RNAseq analysis of colonic tissue from GF mice revealed differentially expressed genes in the category of immunoglobulin genes, indicating a generalized defense response to bacteria (Fig. 4e-g). However, the most statistically significant difference in the colonic transcriptional response between GF mice colonized with wild type A. muciniphila versus mul1A mutants occurred in female mice and was in pathways associated with lipid metabolism (Fig. 4e-g, Supplementary Data 1). Expression of the cholesterol biosynthetic genes Sqle, Hmgcs, Hmgcr, as well as most genes along the cholesterol/mevalonate synthesis pathway were repressed in mice colonized with wild type A. muciniphila, as were genes that modulate cholesterol uptake, including Ldlr and Pcsk9 (Fig. 4d-i, Extended Data Fig.7b-c). Single cell RNAseq datasets suggest that cholesterol/mevalonate biosynthetic genes are expressed in colonic intestinal epithelial cells and goblet cells⁴⁸ (Extended Data Fig. 7d), where cholesterol biosynthesis contributes to stem cell proliferation⁴⁹ and is required to maintain the integrity of the intestinal epithelium⁵⁰. Colonic epithelial cells are the most likely to respond directly to A. muciniphila and its metabolites. These observations are consistent with reports that A. muciniphila metabolites, including SCFAs, repress lipid metabolism in intestinal organoids⁴⁷, and that treatment with A. muciniphila lowers serum cholesterol in mice^1,51 and humans⁵². While GF mice showed intermediate levels of expression of sterol biosynthesis genes (Extended Data Fig. 7b), these pathways were significantly elevated in GF mice mono-colonized with the mul1A mutant (Fig. 4d-i, Extended Data Fig. 7b). This observation suggests that the colon’s transcriptional response to A. muciniphila is directly influenced by the bacterium’s ability to take up or metabolize mucins and related glycoproteins.

Discussion

A. muciniphila is a prevalent member of the intestinal microbiota and a potential next generation probiotic. However, we have an incomplete understanding of A. muciniphila biology and how it interacts with its host due to a lack of genetic tools and because Akkermansia proteins share limited homology with other prominent gut microbes. Here, we established methods for Tn mutagenesis and applied INSeq⁵³ to identify A. muciniphila genes required for mucin utilization and colonization. This led to the discovery of genes encoding hypothetical proteins, MULs, that were essential for growth on mucin. Unlike the substrate inducible polysaccharide utilization systems (PULs) that are common among Bacteroides and other intestinal microbes⁵⁴, many of the mul genes were constitutively expressed, possibly reflecting A. muciniphila’s specialized niche using mucin as its preferred nutrient source. The MUL complex was required for active transport of mucin across the outer membrane and appears to be a specialized system tailored to mucin acquisition, and possibly other glycans, in A. muciniphila.

Our findings indicate that A. muciniphila can access multiple nutrient sources in the GI tract, yet the ability to use mucin via the MUL system provides a competitive advantage against other members of the intestinal microbiota. Furthermore, we were able to determine that A. muciniphila mucin consumption repressed the expression of cholesterol biosynthesis genes in the colons of GF mice. Therefore, in addition to the beneficial immunomodulatory activities that have been assigned to cellular components of A. muciniphila^1,5, the active catabolism of mucin by Akkermansia may provide additional health benefits by regulating the expression of genes involved in lipid biosynthesis.

Methods

Animal studies

Animal care and study protocols were approved by Duke University’s Institutional Animal Care and Use Committee. All experiments with mice were performed under protocols A118–20-05.

Media and strains

Bacteria were grown in an anaerobic chamber (Coy Laboratory) with the following gaseous characteristics: 5% hydrogen_, 5% carbon dioxide, 90% nitrogen. A. muciniphila was grown in mucin medium based on previous work²¹ (3 mM KH₂PO₄, 3 mM Na₂PO₄, 5.6 mM NH₄Cl, 1 mM MgCl₂, 1 mM Na₂S·9H₂O, 47 mM NaHCO₃, 1 mM CaCl₂ and 40 mM HCl, trace elements and vitamins, and 0.25% porcine gastric mucin (Type III, Sigma-Aldrich)). Additional media used in this study included synthetic medium, where porcine gastric mucin was replaced with 0.2% GlcNAc, 0.2% glucose, 16g/L of soy peptone (Amresco) or Phytone (BD) and 4g of threonine/L¹, and Brain Heart Infusion (BHI, BD) supplemented with 2% tryptone and 1x hemin and vitamin K (Hardy Diagnostics). A. muciniphila BAA-835²¹, Bacteroides vulgatus, and B. thetaiotaomicron were obtained from ATCC and additional strains, including A. glycaniphila, Bifidobacterium longum, Bifidobacterium bifidum, Ruminococcus gnavus, Ruminococcus torques, and Peptostreptococcus russellii were obtained from DSMZ. E. coli S17 harboring the pSAM_Bt plasmid was provided by Andrew Goodman at Yale University. When required, antibiotics were used at the following concentrations for Akkermansia: chloramphenicol 7 mg/ml, gentamicin 10 mg/ml, kanamycin 12 mg/ml. E. coli was cultured in LB medium, and antibiotics were added as required at the following concentrations: ampicillin 100 μg/ml, chloramphenicol 35 μg/ml, kanamycin 30 μg/ml.

Labeling of mucin glycans

Porcine gastric mucin was labelled with 6-aminofluorescein by adapting a previously described protocol to label polysaccharides ⁵⁵. Prior to labelling, commercial pepsin digested gastric mucin (Type III mucin, Sigma) was filtered through a 0.45 μM filter and transferred to a 5 ml tube. In a fume hood, 1 ml of filtered mucin was combined with 30 mg of CNBr diluted in 350 μl of water. The reaction was carried out for 25 min while monitoring the pH using paper strips. Aliquots of 0.25 M NaOH were added as needed to maintain the pH > 10 for the duration of the reaction. Following activation, excess CNBr was removed using a Bio-Rad 10DG desalting column equilibrated with 0.2M sodium borate buffer, pH 8. Large mucin glycoproteins were eluted in the void volume into a tube containing 2 mg 6-aminofluorescein and reacted overnight in the dark. The labelled mucin glycans were then dialyzed against 16 L of 20 mM sodium phosphate buffer, pH 8, using an 8 kDa MWCO membrane and quantified using the phenol sulfuric acid method.

Microscopy

To test Akkermansia’s interaction with labeled glycans by microscopy, cultures were grown overnight to saturation and sub-cultured 1:1 in fresh mucin medium for 5 hours. The actively growing cultures (600 ml) were combined with 1.3 ml fresh synthetic media and either 20 mg carbohydrate/ml fluorescein-mucin or 20 mg/ml fixable Fluoro-Emerald (fluorescein) dextran (10,000 MW, Thermo D1820). For experiments involving carbonyl cyanide m-chlorophenylhydrazone (CCCP), the cultures were pretreated with 50 μM CCCP for 30 min prior to the addition on FL-mucin. Cultures were then incubated at 37^oC for 3h and subsequently harvested by centrifugation at 14 000 xg for 5 min, washed once with PBS, and fixed with 4% formaldehyde solution for 30 min on ice. The cells were washed with PBS to remove formaldehyde, resuspended in GTE buffer (50 mM glucose, 25 mM Tris, 10 mM EDTA), and applied to poly-lysine coated coverslips. Unbound cells were removed by washing with PBS. The cells were then treated with lysozyme (1 μg/ml) for 2 min, washed with PBS, and blocked with 2% BSA in PBS for 10 min at room temperature. Finally, the cells were stained with anti-Akkermansia antisera overnight at 4°C. After staining with fluorescently labeled secondary antibodies (goat anti-rabbit IgG Alexa Fluor 647, Invitrogen A21244). The cells were imaged on an inverted confocal laser scanning microscope (LSM 880; Zeiss) equipped with a motorized stage, diode (405, 561 nm), argon ion (488 nm), and helium-neon (633 nm) lasers, and an Airyscan detector module with Airyscan Fast capability. Images were acquired using a Plan Apo 63x/1.4 NA objective (Zeiss) and the Zen software (Zeiss) with Airyscan deconvolution.

To test fluorescent mucin uptake by B. thetaiotaomicron, starter cultures were grown to an OD₆₀₀ of 0.4–0.5 in chopped meat media supplemented with hemin and vitamin K, or for A. muciniphila, cultures were grown in synthetic medium. The cells were washed once and suspended in PBS. The cultures (600 μl) were combined with 1.3 ml of a modified version of synthetic media prepared with 0.25% mucin instead of glucose/GlcNAc and supplemented with 100 μl FL-mucin (final concentration 20 μg carbohydrate/ml). Labeled mucin accounts for approximately 0.8% of the total carbohydrate content of the media. The cultures were incubated for 3h at 37°C. Membranes were stained with 10 μg/ml fixable FM4–64fx (Invitrogen, F34653) and the cells were fixed with 4% formaldehyde and prepared for microscopy as described above.

For super-resolution microscopy, A. muciniphila was grown with FL-labeled mucin and prepared for imaging as described above, except that lysozyme was omitted and the cells were stained with anti-Akkermansia antisera (1:100 dilution) followed by STED compatible fluorescently labeled secondary antibodies (goat anti-rabbit IgG Alexa Fluor 594, Invitrogen A32740). Super resolution microscopy was carried out using a Leica TCS SP8 STED (STimulated Emission Depletion) microscope equipped with a Leica DMi8 inverted motorized stage, a pulsed white light laser, high-sensitivity GaAsP HyD detectors, and STED depletion lasers (592nm, 660nm, 775nm). For each image, 40 stacks in the Z plane were acquired with a 100x/1.4 objective HCX PL APO OIL DIC WD 90 μm (Leica) and LAS X software (Leica), followed by deconvolution with Huygens Professional (SVI) and 3D rendering with Imaris (v9.5, Oxford Instruments).

Live cell imaging of mucin acquisition

Cells were imaged under 1.5% agarose pads. The pads were prepared by melting 0.3 g agarose (Apex General Purpose LE Agarose) in 20 ml synthetic media and spotting ~200 ml of molten agarose onto glass bottom dishes. The pads were solidified and equilibrated for 2 h under anaerobic conditions prior to use. To prepare for live cell imaging, A. muciniphila cultures were grown overnight in mucin medium (OD₆₀₀ ~0.05) and subsequently diluted into a 1:1 mix of fresh synthetic and mucin medium. The cultures were incubated for an additional 3 h at 37^oC to ensure that the cells were actively growing. For CCCP treated cells, 50 μM CCCP was added during the last 15 min of incubation. Cells from 1.5 ml of culture were then harvested by centrifugation at 10 000 xg for 3 min and resuspended in 50 μl synthetic media supplemented with 15 ml of FL-labeled mucin (1.4 μg). Aliquots of the cells (10 μl) were immediately plated on 35 mm glass-bottom dishes (Cellvis), overlaid with an agarose pad, and sealed with parafilm under anaerobic conditions. The cells were imaged on a Nikon Ti2 inverted microscope equipped with a motorized stage, LED light source (Sola Light Engine), an ORCA-Flash4.0 V3 sCMOS digital camera (Hamamatsu), and an environmental chamber (Okolab) set to 37°C. Images were acquired every 30 sec for 30 min using a Plan Apo 100x/1.4 NA oil objective (Nikon) with 1.5x zoom in the NIS-Elements software. The images (resolution = 43.33 nm/pixel) were deconvolved in the NIS-Elements using the automatic 2D deconvolution function and bleach corrected with histogram matching in ImageJ (NIH), converted to 8-bit TIFFs and reconstructed in Adobe Photoshop where only linear adjustments were made to fluorescence intensity. Annotations were added using ImageJ.

Kinetics of mucin uptake

A. muciniphila was grown in mucin medium to an OD₆₀₀ of 0.5 and mixed 1:1 with fresh synthetic media and 96 mg/ml fluorescein labelled mucin. Cultures were incubated anaerobically at 37°C. At each timepoint, the OD₆₀₀ was measured and 1.5 ml of culture was collected and the bacteria were harvested by centrifugation at 14 000 xg, 3 min. The resulting cell pellets were either fixed with 4% formaldehyde and used in flow cytometry as described below. To test CCCP inhibition of mucin transport, cultures were treated with 50 μM of CCCP for 15 min prior to the addition of mucin. As a control for active transport, the cell permanent esterase 5-Carboxyfluorescein Diacetate (CFDA, Thermo Fisher Scientific, C1354) was used in place of mucin. CFDA staining and killed controls were prepared as described previously⁵⁶. CFDA fluoresces at 488 nm in viable cells.

Flow cytometry

Flow cytometry was used to measure the fluorescent intensity of bacteria grown in the presence of fluorescein-mucin. Cultures were grown in mucin medium overnight, subcultured 1:1 into fresh mucin media and grown for an additional 5h. The actively growing cultures (600 ml) were combined with 1.3 ml synthetic media and 20 mg carbohydrate/ml fluorescein-mucin. For mutants with mucin growth defects, synthetic media was used throughout. Cultures were grown in triplicate and fixed as described above. The cells were then incubated with rabbit anti-Akkermansia antisera in PBS/2% BSA (1:100 dilution) overnight at 4°C, washed with PBS, followed by incubation with anti-rabbit IgG Alexa 647 antibodies (1:1000, Invitrogen A-21244). Cells were analyzed on a BD Canto II flow cytometer and 50 000 events were collected for each sample. Flow cytometry data was acquired using FACSDiva software (v9, BD) and additional analysis was run using FlowJo (v9 and 10) software. The samples were gated on the population of single cells that stained positive with anti-Akkermansia/Alexa-647, and the mean intensity of fluorescein-mucin staining was recorded.

Growth curve assays

Growth kinetics of wild type A. muciniphila and selected Tn mutants was measured in both mucin and synthetic media. Starter cultures were prepared by growing bacteria in synthetic medium, supplemented with chloramphenicol (7 μg/ml) for mutants and diluted 1:1 into fresh medium and grown for an additional 5 h. The resulting cultures were then diluted 1:25 into fresh media (OD₆₀₀ 0.01 – 0.05) and 150 ml aliquots were dispensed into 96-well microplates. Each well was covered with 100 ml of paraffin oil and incubated at 37°C in a BMG SpectroStar Nano plate reader under anaerobic conditions. The optical density (OD₆₀₀) was measured at 1 h intervals for 72 h. Results were obtained from three biological replicates per strain. Assays with intestinal microbes other than Akkermansia were carried out as described above, except that the starter cultures were prepared in Brain Heart Infusion medium (BHI, BD) with 1x hemin vitamin K and 2% tryptone. The BHI media for A. muciniphila was also supplemented with 0.25% mucin. Growth curves assessed in mucin medium²¹ supplemented with 1x hemin vitamin k, where mucin is the sole source of carbon and nitrogen, or in BHI with hemin vitamin K and tryptone. Results reflect two biological replicates per strain.

Generation of anti-Akkermansia anti-sera

Polyclonal anti-serum was raised against whole A. muciniphila cells. To prepare the antigen, overnight cultures of A. muciniphila were washed three times with PBS and incubated at 4°C aerobically for 24h. The cells were then fixed with 4% formaldehyde, washed, and suspended in PBS. Antibodies were raised in New Zealand white rabbits. The antigen was mixed 1:1 with Freund’s complete adjuvant and 450 μl was used for subcutaneous inoculation, followed by three additional injections using Freund’s incomplete adjuvant.

Generation of anti-Amuc_0544 anti-sera

Polyclonal anti-sera was raised against a 47 kDa fragment of Amuc_0544 protein that spans the TPR domains (from 1216 bp to 2457 bp). The fragment was amplified with primers #127 and #129 (Supplementary Data 1) using Phusion polymerase and 5X HF Buffer (Thermo Scientific) and cleaned with a QIAquick PCR purification kit (Qiagen). The resulting PCR product and the expression vector pET28a (EMD Biosciences) were digested with EcoRI-HF and HindIII-HF (NEB #R3101, #R3104), gel extracted using a QIAquick Gel Extraction kit, and ligated overnight with T4 DNA ligase (NEB, #M0202). The ligated DNA was cloned into competent E. coli XL-1 Blue (Agilent), and after the construct was confirmed by sequencing, the final plasmid was transformed into E. coli BL21(DE3) (NEB, #C2527I) for protein expression. The Amuc_0544 fragment was produced as a soluble protein with an N-terminal hexahistidine tag. E. coli BL21(DE3) was grown in LB medium with 30 μg/ml kanamycin to mid-exponential phase and induced with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 3 h. The cells were harvested by centrifugation (5000 xg, 10 min) and the pellet was suspended in buffer (50 mM sodium phosphate, 300 mM NaCl, 20 mM imidazole; pH 7.4) and lysed by sonication with 30 cycles, 30 s intervals alternating with 30s on ice (Branson Sonifier 450). Cell debris was removed by centrifugation (15 000 xg, 45 min, 4°C), and the supernatant was applied to a Ni-NTA agarose bead column (Prometheus Protein Biology Products). The column was washed with 25 ml of buffer containing 40 mM imidazole, and protein was eluted in 1 ml fractions with 300 mM imidazole. Antibodies were raised in rabbits as described above using 2.5 mg of purified protein per injection.

RNAseq analysis of A. muciniphila grown in mucin and synthetic media

RNAseq was used to assess bacterial transcriptional responses to growth in mucin and synthetic media. For each medium to be tested, 25 ml cultures were prepared in triplicate and grown to mid-log (OD₆₀₀ ranging from 0.35 to 0.5). Cells were harvested by centrifugation at 10 000 xg, 5 min, 4°C. RNA was isolated using a Qiagen RNeasy kit following the enzymatic digestion protocol (proteinase K + lysozyme). The resulting RNA was concentrated by precipitation with 2 volumes of isopropyl alcohol and 0.3M sodium acetate, pH 5. rRNA was depleted using a Ribominus Bacteria kit (Thermo). RNA quality was assessed by running on an Agilent Bioanalyzer. Finally, 100 ng per sample was used as input RNA to prepare libraries with a NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB #E7760S) and NEBNext Multiplex Oligos for Illumina. Libraries were sequenced on a Hiseq 4000 in a 50 bp SR run. Following sequencing, adaptor sequences were trimmed and filtered using the program fastqmcf¹. The reads were then mapped to the A. muciniphila genome (accession: CP001071.1) using STAR⁵⁸ with the parameter --alignIntronMax set to 1. Differential expression analysis was carried out with DESeq2⁵⁹ and pathway analysis was run using clusterProfiler⁶⁰.

pSAM_Akk plasmid construction

Transposon mutagenesis was carried out with a modified version of pSAM_Bt¹⁹ adapted for use in Akkermansia. First, the ermG gene was replaced with a cat gene to confer chloramphenicol resistance after observing that A. muciniphila could acquire spontaneous resistance to erythromycin. The cat gene and its promoter was amplified from pRL1342⁶¹ with the primers #80 and #81 (Supplementary Data 1) and introduced into pSAM_Bt using the MfeI and XbaI cut sites. Cloning was carried out in the pir+ strains E. coli Pir2 (Invitrogen) and E. coli S17. Next, the himar1C9 transposase was codon optimized for expression in Akkermansia. To generate a codon table for Akkermansia, we used a selection of representative genes (Amuc_R0036, rpoD, secA, eftu, DNA pol I, DNA pol III alpha, ftsZ, ftsY, yidC, GAPDH, GroEL, enolase) and the tools provided in Sequence Manipulation Suite⁶². Comparison between A. muciniphila codon usage and himar1C9 revealed rare codons for arginine and serine. These were substituted for codons frequently used in Akkermansia, and the resulting gene was synthesized as a gBlock (IDT). The codon optimized himar1C9 was then ligated into pSAM_Bt using the NdeI and NotI cut sites to produce pSAM_Akk. Replacement of the original himar1C9 gene was confirmed by sequencing.

Transposon mutagenesis

Basic protocol.

Initial attempts at mutagenesis using pSAM_Akk were unsuccessful, and we observed that the plasmid can transiently remain in A. muciniphila, even in the absence of the pir genes required for the R6K origin of replication. To facilitate transposition, cure the donor plasmid, and remove potential residual donor E. coli contamination, several outgrowths in the absence of selection were required. To carry out conjugations, Akkermansia and E. coli were combined at a 10:1 ratio. A. muciniphila starter cultures were grown to saturation in 5 ml synthetic medium and subsequently diluted 1:10 into 20 ml synthetic medium. The cultures were grown to an OD₆₀₀ = 1 and harvested by centrifugation at 15 000 xg, 5 min. The E. coli donor was prepared by growing in LB medium (Fisher Bio-reagents, #BP1426–2) with ampicillin (100 μg/ml). Overnight cultures were subcultured 1:50 and grown to an OD₆₀₀ = 0.4 to 0.6. Cells were harvested by centrifugation at 1 500 xg and washed with LB media to remove residual antibiotics. The pellets were combined to create a slurry of cells that was distributed in 100 μl puddles on pre-reduced synthetic media plates. Conjugations were incubated aerobically at 37°C for 14 to 16h. Following conjugation, the puddles were scraped into a 1:1 mixture of PBS and 50% glycerol and resuspended. The cell suspension was diluted 1:10 into fresh synthetic media supplemented with gentamicin (10 μg/ml) and kanamycin (12 μg/ml) to counter select against the E. coli donor. Cultures were grown anaerobically for an additional 48 h to allow Akkermansia to recover, and subsequently sub-cultured 1:10 into fresh media two more times, at 24 h intervals. After the final subculture, 150 μl of the suspension was plated on synthetic media containing gentamicin, kanamycin, and chloramphenicol (10 μg/ml, 12 μg/ml, 7 μg/ml). Colonies appeared after 4 to 6 days incubation at 37°C. Colonies were routinely tested for contamination from residual E. coli donor by growing aerobically at 37°C.

To confirm that transposition had occurred, initial quality control tests were performed by PCR to screen for the presence of plasmid backbone by amplification of the bla gene (primers #11/12), Akkermansia specific 16S (primers #75/76), and the cat gene located on the transposon (primers #80/81). Once transposition was confirmed, arbitrary PCR was used for small-scale mapping of Tn insert locations⁶³, using the following sequential PCR settings: Reaction 1, run with primers #22/88: 94°C for 5 min; 6 cycles of 94°C × 30 s, 30°C × 30 s, 72°C for 1 min; 30 cycles of 94°C × 30 s, 45°C × 30 s, 72°C for 1 min; and 72°C for an additional 5 min. Reaction 2 was run using 2 ul of Reaction 1 as template and primers #23/89: 95 °C for 1 min followed by 30 cycles of 95°C × 30 s, 45°C × 30 s, 72°C for 2 min; 72°C for 5 min. The resulting PCR products were sequenced with primer #89. Finally, to confirm that the Tn inserted randomly into the genome, a set of representative mutants were analyzed by Southern blot. DIG labelled probes were generated to detect the cat gene and the blot was run using a DIG High-Prime Labeling and Detection II kit (Sigma). Southern blots, as well as DNA gels, were imaged with a Li-Cor Odyssey Imager and images were labeled using Adobe Photoshop v.24.1.1.

Construction of Tn libraries

Tn libraries were prepared in two formats for INSeq analysis. 1. Pooled Tn library: The first library consisted of approximately 25 000 mutant colonies which were prepared by scraping agar plates and pooling the resulting cell suspensions. This library was used for in vitro INSeq experiments. 2. Arrayed Tn library: The second approach used a smaller, but well-defined library constructed with equal inputs of know mutants from a library arrayed in 96-well plates. This library was used for all in vivo INSeq experiments. By generating pools of defined mutants to be used for INSeq analysis, we were able to filter the data against known inputs and remove spurious reads generated by sequencing errors. A subset of the arrayed library was used for a Cartesian mapping strategy, described below, to facilitate retrieval of mutants of interest.

i. Pooled Tn mutant Library.

Because of the low transposition efficiency, we ran 15 conjugation reactions and then subdivided each reaction into 16 independent outgrowths using 96 deep-well plates. The outgrowths were carried out as described above by culturing conjugation mixes for an initial 48h followed and two additional subculture steps at 24h intervals in the synthetic medium with gentamicin and kanamycin to inhibit E. coli growth. To select for transconjugants, the reactions were plated on synthetic media with gentamicin, kanamycin, and chloramphenicol and incubated anaerobically at 37°C for 6 days. Colonies were scraped into PBS with 20% glycerol and the number of colonies and volumes used were recorded for each reaction. This information was used to determine volumes required to mix equal numbers of colonies from each reaction. In total, the large pool was estimated to represent approximately 25 137 colonies (2680 insertions in 198 intergenic regions and 725 genes). The pooled cell suspensions were aliquoted and stored at −80°C and used as inoculum for in vitro INSeq experiments.

ii. Arrayed Tn mutant library.

To generate an arrayed library of Akkermansia Tn mutants, mutagenesis was carried out as described above, except that individual colonies were picked and arrayed into a total of 54 × 96-well plates. Colonies were grown in synthetic medium with antibiotics until all wells were turbid. Glycerol was added to a final concentration of 20% for storage at −80C. To identify the Tn insertion locations for each mutant, we used the INSeq library and mapping protocol. For each plate, 60 ul per well was pooled and DNA was extracted with Qiagen’s DNeasy Bloods and Tissue kit. The resulting DNA was used to prepare sequencing libraries, enabling the identification of each Tn insertion on a plate. Cultures from the arrayed plates were then combined to create a single pooled library comprised of all of the arrayed mutants in approximately equal amounts. The pooled cell suspensions were aliquoted and stored at −80°C and used as inoculum for in vitro and in vivo INSeq experiments.

Because the collection of 54 individual 96-well plates was derived from multiple conjugations and the extensive outgrowths required to obtain A. muciniphila Tn-mutants, plates were first sequenced separately. Three plates displayed extreme levels of clonality and were not used further. We chose to use the Arrayed Tn mutant library for the majority of experiments, because this ensured that any Tn-mutants identified from INSeq experiments could potentially be retrieved for follow-up experiments.

Cartesian Mapping strategy to map Tn insertion sites in the A. muciniphila arrayed collection

Bacteria from individual 96-well plates were pooled, assigned a barcode, and the Tn insertion sites sequenced as previously described⁵³ This approach provided very high sequencing depth for each 96-well plate and provided confidence in distinguishing true sequence reads derived from Tn-insertions on the 96-well plate from PCR amplification and sequencing artifacts. As the signal-to-noise varied between individual samples and sequencing runs, a Matlab script (INseq_read_filter_v3.m) was used for viewing data and setting thresholds. This process was used to build a table of Tn-insertion sites present on each plate, and collectively in the library. A high degree of clonal redundancy was observed in the Arrayed Collection (mean of 3.7 replicates of each clone), likely as a result of the extensive outgrowths that were required to effectively remove the E. coli donor strain prior to selection of transconjugants.

Clonal redundancy creates a challenge for mapping Tn-insertions to specific plate-well locations using methods that are based on orthogonal pooling and high throughput sequencing because the presence of multiple representatives of a clone in a pool either eliminates information on location (binary code¹⁹) or makes that information ambiguous (Cartesian pooling and Sudoku²⁰). Thus, we faced a trade-off between capturing the full diversity of Tn insertions in the collection, and fewer plates to derive more useful location mapping information. Simulations were used to model this trade-off, and to optimize the choice of 96-well plates taken for location mapping, with the aim of capturing the highest number of unique clones and the lowest level of redundancy.

i. Simulations:

An optimization routine was implemented in Matlab (sudoku_plate_compare_v1.m) to find the subset of 96-well plates that would yield the highest diversity and lowest redundancy when combined. Intra-plate redundancy was first simulated for each plate to generate an in silico set of 96-well plates. The number of clonal replicates on a plate was taken as 96 minus the number of unique Tn-insertions identified on that plate by deep sequencing. Replicates were then assigned to specific Tn-insertions identified on a plate by uniform random sampling with replacement. A full set of in silico 96-well plates generated by this process was then used for inter-plate comparisons designed to capture the series of plate-sets that minimizes clonal redundancy. The first in this series is simply the plate with the most unique clones (lowest intra-plate redundancy). In subsequent steps, all remaining plates are tested one at a time in combination with the set of previously chosen plates (1..i-1), to find the plate (i) that creates a new set (1..i) that has the lowest overall redundancy.

ii. Evaluating the trade-off between coverage-redundancy when choosing a subset of plates suitable for Cartesian mapping:

Simulation data for each optimized set of plates was then used to evaluate the trade-off between coverage and redundancy. The number of unique Tn-insertions and the mean redundancy increase for optimized plates sets were determined for a size of 1 through 40 (Extended Data Fig. 2e). The plate set that yields the largest diversity-for-redundancy is found where the distance between these two curves is greatest (set size 16: 873 unique Tn-insertions, mean redundancy = 1.76).

The degree of redundancy per clone changes with increasing plate-set size in the simulated data (Extended Data Fig 2f). This plot was used to further inform our choice and establish a cut off for the plate-set series used for Cartesian mapping. Cartesian mapping is precise when a clone is present in the library in one copy, allowing unambiguous identification of the Plate-Row-Column location for each Tn-insertion directly from the sequencing data, without any need for retrieval and secondary evaluation. This is particularly useful for high-throughput screening (forward genetics), where the genetic nature of hits in the screen can be immediately identified from the Cartesian mapping table. Clones present in two and three copies in the library still provide useful location information for Cartesian mapping, but this information is not precise and manual retrieval and secondary confirmation by PCR are necessary to match Tn-insertion information with a Plate-Row-Col identifier. The search space for redundant clones scales as the <number of copies> to the power of the <number of pooling dimensions>, so a clone present three times in the collection would be mapped to 27 potential Plate-Row-Col locations (3^3). In most cases, only one of the three copies needs to be retrieved for further work, effectively eliminating one pooling dimension and reducing the search space to nine Plate-Row-Col locations. Thus, hypothesis-directed retrieval of clones from the library (reverse genetics) also benefits from reducing the degree of clonal redundancy. Striking a balance between genome coverage and useful location mapping, we chose the set of 19 optimized plates (Akk SudSet-19) for location mapping. This set of plates captures 950 of the 1277 Tn-insertions in the entire collection, 86% of disrupted genes (insertion occurred with the 90th percentile of the CDS pos=0.9), and reduces the mean redundancy from 3.7 to 1.9 copies per clone

IN-seq analysis of conditionally essential genes in mucin medium

The mixed pooled mutant library was inoculated at 1:20 dilution in Synthetic Medium containing 10 ug/ml gentamycin, 12 ug/ml kanamycin, 5 ug/ml chloramphenicol and grown to saturation. Cells were washed twice with media and used to inoculate 5 ml test media at a 1:100 dilution. Growth in mucin was tested in mucin medium with a final concentration of 0.5% porcine gastric mucin. Growth in mucin with added nitrogen was tested in mucin medium supplemented with 0.5% mucin, 3 g/L threonine and 16 g/L soy-peptone. As growth rates varied significantly in the different test media, cultures were monitored regularly for OD₆₀₀ by withdrawing a small volume of culture. To obtain samples that had gone through a similar number of generations after inoculation with the Input pool, cultures were collected as they reached saturation, rather than after a fixed length of time.

INSeq analysis of conditionally essential genes in mice

INSeq analysis was carried out using four C57BL6/J mouse models of colonization: germ-free (GF, n = 4), Altered Schaedler Flora colonized (ASF, n = 7), conventional (CONV, n = 11), and Muc2^−/− mice³⁷ (n = 4). Colonization of CONV and Muc2−/− mice required pre-treatment with tetracycline to eliminate endogenous Akkermansia and allow the mutant library to engraft (3 g/L tetracycline suspended in distilled water with 1% sucrose for two weeks). DNA was then extracted from fecal pellets using a Qiagen Fast Stool Mini kit and tested for residual Akkermansia by PCR with Akkermansia specific 16S primers (#75/76). Once the Akkermansia was cleared, the antibiotics were withdrawn 48 h prior to the delivery of Tn mutant libraries by oral gavage. GF and ASF mice did not receive antibiotics.

Tn mutant libraries (Arrayed Collection) were prepared for gavage by diluting frozen library stocks 1:10 into synthetic media and incubating at 37°C for 36 h. Cells were harvested by centrifugation at 10 000 xg, 5 min, 4C, washed once with reduced PBS before being resuspended in 1/10^th of the initial volume. Mice were inoculated by intragastric gavage with 150 μl of the Tn mutant pool (~10⁹ bacteria), and the mutant library was allowed to colonize for 14 days. Mice were euthanized and cecal contents were collected and stored at −80°C until they were processed for DNA extraction. While most samples could be used directly for DNA extraction, the ASF mice had lower yields of Akkermansia DNA, and an enrichment step was added to increase recovery of A. muciniphila. Cecal contents from ASF mice (~200 mg) were combined with 10 ml of synthetic medium containing chloramphenicol, kanamycin, and gentamicin and grown for 8 generations (19 h). DNA was extracted from 9 ml culture per mouse and subsequent steps of INSeq library preparation were carried out using the same approach for all samples.

Droplet INSeq

Starter cultures were prepared by diluting frozen Tn library stocks (Arrayed library) 1:10 into synthetic media and grown for 36 h. To determine the dilution factor required for droplet single cell loading, serial dilutions of cultures were made on mucin-supplemented agar plates for a CFU assay. Given the OD₆₀₀ of the starter culture and the plate CFU counts, the CFU/mL concentration could be estimated from an O₆₀₀ measurement prior to droplet loading⁶⁴. Droplets were generated using syringe pumps and a T- junction microfluidic chip with six droplet generators (Dolomite Microfluidics) with media-containing cell suspension as the aqueous phase and 2% Pico-Surf surfactant (Sphere Fluidics) in NOVEC 7500 (3M) as the oil phase. Mucin media was passed through a 40 μM filter to prevent clogging of the microfluidic chip. All droplet generation and culturing were performed in an anaerobic chamber outfitted with a portable microscope with LCD screen for viewing droplet generation on the microfluidic chip (Celestron). Following droplet encapsulation, cultures were grown at 37°C in anaerobic conditions for 72h. qPCR with Akkermansia-specific primers (#75/76, Supplementary Data 1) was used to assess cell abundance between droplet cultures. To retrieve cells, the droplet emulsions were broken with an equal volume of PFO, vortexed and spun down briefly at 300 g. The aqueous phase was placed in a new tube, cells were pelleted at 10000 g for 1 minute, the supernatant was removed, and the cell pellet was stored at −20 C for DNA extraction. Nine replicate cultures for each condition were pooled to generate sufficient DNA for INSeq library preparation.

INSeq library preparation

Sequencing libraries were prepared as described in Goodman et al.¹⁹ with minor modifications. Libraries prepared from in vitro samples used 2 μg of input DNA, while libraries prepared from mice used 0.5 μg of A. muciniphila input DNA as determined by qPCR. Linear PCR was performed with Pfx polymerase (Invitrogen) and the primer Tru-BioSamA (Supplementary Data 1). Two 50 μl reactions were run per sample using the following cycling conditions: 94°C for 2 min followed by 50 cycles of 94°C for 15 sec and 68°C 1 min. The resulting PCR products were cleaned using a Qiagen QiaQuick PCR cleanup kit and eluted in 50 μl elution buffer. The resulting biotinylated PCR products were then bound to Dynabeads M-280 Streptavidin magnetic resin (Thermo). Washing, second strand synthesis, MmeI digestion, and adaptor ligation were performed as described in Goodman et al.¹⁹, except that six-base pair adaptor sequences⁶⁵ were used. Finally, the samples were amplified with Pfx polymerase using the buffer at 2x concentration, additional Mg2⁺ (4 mM final), and the primer pair TruP7-PCR_5 and TruP5-PCR_3. PCR was performed with the following conditions: 94°C for 2 min followed by 18 cycles of: 94°C 15 sec, 60°C 1 min, 68°C 2 min, followed by 68°C 4 min. The resulting PCR products were size selected on 2% E-Gel SizeSelect II agarose gels (Thermo). The DNA was then quantified using a Qubit dsDNA HS assay kit (Thermo) and sequenced on an Illumina Hiseq 4000 50 bp SR flow cell, samples from ASF and Muc2^−/− mice were sequenced on a NovaSeq 6000 using an S1 50 bp lane. Primer sequences were modified to produce libraries that were compatible with sequencing on Illumina Hiseq 4000/Novaseq 6000 sequencers. All sequencing was performed at the Duke Sequencing and Genomic Technologies Shared Resource.

INSeq mapping and quantification of transposon insertion junctions and normalization Pre-processing:

Raw sequencing data was processed using a python script (repair_barcodes.py) to format raw sequence data files into sample-specific fasta files. These served as the input for Perl scripts (INseq_pipeline_v3.pl, minor modifications from Goodman et al.¹⁹ that identify sequence reads containing the Mariner-Tn, and extract the bounding genomic sequence, which are grouped and counted. Bowtie v1.2.2 was used for mapping 16 bp reads to the Akkermansia MucT genome. Reads that failed to map to a unique genomic locus were discarded. Output read count tables used for subsequent analysis steps contain the Tn insertion coordinate and the number of instances a sequence read was mapped to this site.

The NCBI Akkermansia Muc^T genome sequence contains an erroneous G nucleotide at position 1704819 that results in a frame shift in the highly conserved gene recG (Amuc_1422). Corrected sequence and annotation files that restored the Amuc_1422 reading frame were used for mapping and annotating INSeq sequence reads.

Quality control:

A MATLAB script (INseq_QCsummary_v3.m) was used to extract and plot quality control metrics for all samples from a sequencing run. We encountered problems with broadly distributed noise as well as small numbers of spurious sequences that dominated read depth. Quality control metrics were used to establish criteria for sample inclusion in the data set. [Depth] QC metric 1: Samples were excluded when more than 50% of the total sequencing data represented sequence reads not found in the deeply sequenced Input Sample.

[Noise] QC metric 2: The INSeq library preparation should generate genomic DNA fragments from both sides (L and R) of the Mariner-Tn insertion site. High quality data showed very little variation in the relative abundance of reads mapping to the L and R sides of individual Tn insertion sites. However, some samples had a high degree of L-to-R read variability. This was quantified by calculating the variance of log10(ratLR), as follows: 1) a psuedocount of 1 was added when either L or R read count was zero; 2) for each insertion site, ‘ratLR’ was calculated using the larger of L or R, so that all ratios of side-bias are greater than one; 3) calculate the sample-wide variance in log10(ratLR). Samples were excluded when this metric exceeded 0.2.

Filtering:

For the Arrayed Library, Tn-insertion sites identified by sequencing individual 96-well plates provided a high-confidence set of sites for filtering data sets acquired using this collection. An analogous set of high-confidence Tn-sites was identified in the Pooled Library, by performing multiple deep sequencing runs of the Input library and combining the data for additional depth. Dynamic filtering scripts (INseq_read_filter_v3.m, INseq_read_filter_v4.m) allow users to plot the data and view quality control metrics, apply thresholds, re-plot after thresholding, and ultimately output a high-confidence set of Tn-insertion sites that could be used for filtering sequencing data from lower quality samples.

Cartesian mapping of Tn-insertions to Plate-Row-Column locations provided information about which clones had more than one Tn-insertion. Mutants with a single Tn insertion accounted for 77.5% of the arrayed library, while 17.3% were predicted to have two insertions, and the remaining 5.3% had ≥ 3 insertions. Tn-insertion sites present in multi-Tn clones in the arrayed pool were excluded from analysis, as the phenotype attributed to disruption of a gene by one Tn, could be driven by Tn disruption of another gene in the same clone.

Normalization:

We applied the normalization method used by the ARTIST package to account for changes in library complexity that occur in different test conditions⁶⁶. We made minor changes to the ARTIST method to account for our low-coverage library (ARTIST assumes high coverage and uses all TA sites in the genome for complexity calculations), and to improve computational efficiency (Artist_setup.m, Artist_run_v2.m). ARTIST is designed for single Sample-Input pairs, so an additional multi-sample normalization step (Artist_multiSample_normalization_exact.m) was required when using TRANSIT for analysis of multiple Samples (e.g. biological replicates) to account for variation in sequencing depth between Samples. Normalization features in TRANSIT were disabled.

Analysis of transposon mutant populations

The data analysis package TRANSIT⁶⁷ was used to identify genes with altered relative abundance in input and output populations. Normalized counts were analyzed using the TRANSIT Mann-Whitney U-test in command line mode with the flag -n nonorm and insertions were mapped to a protein table generated from the A. muciniphila ATCC BAA-835 complete genome (GenBank: CP001071.1, corrected as noted above). The output from TRANSIT was used for all downstream analyses, and p-values were adjusted by the Benjamini-Hochberg method.

Pathway analysis was carried out with the clusterProfiler⁶⁰ program enrichKEGG to analyze genes with Tn inserts that generated an absolute Log₂ fold change > 5 in input vs output samples and an uncorrected P value < 0.05. Enriched pathways with a q-value < 0.05 were considered significantly enriched. BioCyc⁶⁸ was used to analyze global pathways in the INSeq data. For this analysis, data was analyzed using the Log₂ fold change values for all genes and mapped to the A. muciniphila genome. Analysis of the assimilatory reduction pathway was carried out using a combination of BioCyc and KEGG. The KEGG pathway for assimilatory sulfate reduction was not included in the BioCyc database, thus the genes from the KEGG pathway amu_M00176 were retrieved in BioCyc and the Log₂ fold change for each gene was displayed using the Pathway Collage option. The Circos plot was created with Circa (omgenomics.com/circa) using INSeq data generated from mouse experiments and from INSeq data generated from growth in mucin medium using the same Tn arrayed library. Genes with a Log₂ fold change > 5 in input vs output samples and an uncorrected P value < 0.05 were mapped onto the Circos plot. To assess genes that were required for growth in vivo and in vitro, a Venn diagram was generated using the R package ggvenn. For this analysis, a less stringent cut-off of a Log₂ fold change > 1 was used to capture all genes with modest growth defects.

To identify putative glycoside hydrolases (GH) in the INSeq datasets, the list of predicted GH enzymes for A. muciniphila BAA-835/Muc^T was retrieved from the CAZY database²⁴. Genes in additional pathways, including amino acid biosynthesis genes, were obtained from KEGG. INSeq data was plotted using the R package ggPlot2 and specific genes of interest were highlighted using the R package gghighlight.

Colonization of mice with wild type A. muciniphila and Tn mutants

All mouse experiments were approved by Duke’s Institutional Animal Care and Use Committee. In vivo colonization experiments were carried out in GF C57BL/6J mice obtained from Duke’s Microbiome Core Facility, C57BL/6J mice obtained from Jackson Laboratories, Muc2^−/− mice³⁷ (kind gift from Leonard Augenlicht, Albert Einstein College of Medicine), and Akkermansia-free mice bred in Duke’s Division of Laboratory Animal Resources Breeding Core Facility. To generate an Akkermansia-free mouse colony, 5-week-old conventional C57BL/6J mice obtained from Jackson Laboratories were treated with tetracycline for two weeks to eliminate endogenous A. muciniphila. We then reconstituted the microbiota with a fecal slurry prepared using pellets from a previously described line of Akkermansia-free mice⁶⁹. The mice were fasted overnight and gavaged with 200 μl of fecal slurry (0.03g/ml). These mice were then used to start a breeding colony that we have maintained in sterile cages since January 2020. The colony is routinely tested for contamination using Akkermansia-specific 16S rRNA primers⁷⁰, and all mice are tested for contamination immediately before experiments. Akk-free mice were maintained on autoclaved water and autoclaved LabDiet 5K67, and CONV and Muc2−/− mice received reverse osmosis water and LabDiet 5353. Mice were between 8 and 12 weeks old, and were a mix of males and females, except for the CONV mice which were all females. CONV and Muc2^−/− mice were pre-treated with 3 g/L tetracycline suspended in distilled water with 1% sucrose for two weeks. Following antibiotic treatment, clearance of residual endogenous mouse Akkermansia was confirmed by PCR using Akkermansia-specific 16S rRNA primers. Antibiotics were replaced with regular water for two days before gavage. No antibiotic treatment was used for GF or Akkermansia-free mice.

Mice were colonized with mutants in mul1A (amuc_0544::Tn, insert at 644982 bp), mul1B (amuc_0543::Tn, insert at 642725 bp), mul2A (amuc_1102::Tn, insert at 1317866 bp) or a mutant with a Tn insertion immediately upstream of the amuc0029 start codon (insert at 38253 bp). To prepare the inoculum for colonization, A. muciniphila cultures were grown for 36h, concentrated by centrifugation, and suspended in PBS containing 20% glycerol. Cultures were standardized by optical density and plate counting. The inoculum was stored at −80°C and thawed immediately before oral gavage. All mice were inoculated by intragastric gavage with 150 – 200 μl containing ~10⁹ CFU. For the competition assay, wild type A. muciniphila was mixed with the mutants at a 1:1 ratio and mice were gavaged with 150 μl of a mixture containing approximately 0.5 × 10⁸ CFU of each strain. To partially reconstitute the microbiota in antibiotic-treated mice, the gavage material for CONV mice included a fecal slurry prepared from fecal pellets from Akkermansia-free mice. The slurry was prepared under anaerobic conditions by homogenizing fecal pellets in sterile PBS (0.1g/ml). Large particles were removed by centrifugation (300 xg, 1 min), and the resulting supernatant was used to prepare the gavage material (100 μl per mouse). All mice received a single gavage after an overnight fast. To measure A. muciniphila abundance, fecal pellets were collected at regular intervals and DNA was extracted using a Qiagen Fast Stool DNA kit or Qiagen PowerSoil Plus.

qPCR to assess A. muciniphila colonization levels in mice

A standard curve was generated using A. muciniphila genomic DNA isolated from a pure culture with a DNeasy Blood and Tissue kit (Qiagen), quantified with Qubit, and used to make a five-point standard curves with 10-fold dilutions of a 10 ng/ml starting concentration. A. muciniphila specific 16S rRNA primers were used to measure total Akkermansia loads, while strain specific primers designed to flank the Tn insertion site to measure total loads for each mutant were used for the competition experiments. To validate the primers for the competition experiments and ensure that they were specific to their intended target, each primer pair was tested using serial dilutions of the target strain diluted in a 10 μg/ ml solution of the competing strain. Primers are listed in Table 1. PCR was run with PowerUp SYBR Master Mix (Thermo) reagent on a QuantStudio 3 real time PCR system (Applied Biosystems) using fast cycling mode. The abundance of A. muciniphila was calculated as copy per gram fecal material or cecal content.

RNAseq of mouse tissues

To investigate host responses to A. muciniphila colonization and mucin grazing, RNAseq was performed on gastrointestinal tissues from GF mice colonized for 14 days with either wild type A. muciniphila or the mul1A::Tn mutant. The intestinal contents were collected to quantify A. muciniphila levels along the GI tract and tissue was collected to prepare RNA. The entire intestinal tract was removed, and the contents of the colon, cecum, ileum, jejunum and duodenum were collected and stored at −80°C. Colonic tissue was rinsed in cold PBS and immediately transferred to RNALater (Invitrogen). To prepare RNA, the tissue samples (~25 mg per mouse) were placed in 0.6 ml TRIzol in soft tissue homogenizing tubes (Precelly), and the samples were lysed using a bead beater set to 5000 rmp, 15 s, x 3 cycles. RNA was then isolated using Zymo’s Direct-zol RNA miniprep kit (Zymo, R2050) according to the manufacturer’s instructions. RNA was quantified with Qubit RNA Br kit and the RIN score was tested on an Agilent Bioanlyzer. Poly-A purification of mRNA was carried out using a NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB) and used to prepare libraries with the NebNext Ultra II directional RNAseq kit. Libraries were sequenced on a NovaSeq 6000 S-Prime 50bp PE flow cell. Sequencing data was trimmed and filtered as described for the bacterial samples, except that the reads were mapped to Mus musculus genome (assembly GRCm38) using STAR with the flag --sjdbOverhang 49. Differential expression was analyzed with DESeq2⁵⁹ and pathway analysis was run using MetaScape⁷¹. P-values were adjusted by the Benjamini-Hochberg method. To adjust for multiple testing, differentially expressed genes were plotted with the R package Enhanced Volcano.

Co-immunoprecipitation and mass spectrometry

Native immunoprecipitation coupled with LC-MS/MS was used to identify proteins that interact with Mul1A (Amuc_0544). Bacterial cultures were prepared with either wild type or the mul1A::Tn mutant as a negative control. Samples were prepared with the wild type strain grown in both mucin and synthetic media, and the mutant was grown in synthetic media only, and each sample was prepared in triplicate. Cultures were grown in 30 ml media to an OD₆₀₀ of 0.6 to 0.8 and the cells were harvested by centrifugation at 4000 xg, 20 min. The supernatant was discarded, and the cell pellets were stored at −80°C overnight, and subsequently thawed in 10 ml of a lysis buffer (150 mM NaCl, 50 mM Tris, pH 8, 1% dodecyl-beta-maltoside, 1 mM PMSF). The cells were lysed by sonication at 30 s intervals on ice and remaining cell debris was pelleted by centrifugation at 15 000 xg, 25 min, 4°C. The resulting supernatants were used for immunoprecipitation of Mul1A. Affinity purified anti-Mul1A was bound to protein A magnetic beads following the manufacturers protocol (Bio-Rad), and 50 μl of the resin was then added to cell lysates and incubated for 48h at 4°C with rotation. The beads were collected and washed four times with 50 mM NaCl, 25 mM Tris pH 8 buffer. Proteins were eluted with 25 μl 62.5mM Tris 2% SDS at 60°C for 10 min and submitted to the Duke Proteomics Core Facility for quantitative LC/MS/MS analysis.

Samples were reduced with 10 mM dithiothreitol for 30 min at 80°C and alkylated with 20 mM iodoacetamide for 30 min at room temperature. Next, they were supplemented with a final concentration of 1.2% phosphoric acid and 273 μL of S-Trap (Protifi) binding buffer (90% MeOH/100mM TEAB). Proteins were trapped on the S-Trap, digested using 20 ng/μl sequencing grade trypsin (Promega) for 1 hr at 47C, and eluted using 50 mM TEAB, followed by 0.2% FA, and lastly using 50% ACN/0.2% FA. All samples were then lyophilized to dryness and resuspended in 12 μL 1%TFA/2% acetonitrile containing 12.5 fmol/μL yeast alcohol dehydrogenase (ADH_YEAST). From each sample, 3 μL was removed to create a QC Pool sample which was run periodically throughout the acquisition period.

Quantitative LC/MS/MS was performed on 3 μL of each sample, using a nano Acquity UPLC system (Waters Corp) coupled to a Thermo Orbitrap Fusion Lumos high resolution accurate mass tandem mass spectrometer (Thermo) via a nanoelectrospray ionization source. Briefly, the sample was first trapped on a Symmetry C18 20 mm × 180 μm trapping column (5 μl/min at 99.9/0.1 v/v water/acetonitrile), after which the analytical separation was performed using a 1.8 μm Acquity HSS T3 C18 75 μm × 250 mm column (Waters Corp.) with a 90-min linear gradient of 5 to 30% acetonitrile with 0.1% formic acid at a flow rate of 400 nanoliters/minute (nL/min) with a column temperature of 55C. Data collection on the Fusion Lumos mass spectrometer was performed in a data-dependent acquisition (DDA) mode of acquisition with a r=120,000 (@ m/z 200) full MS scan from m/z 375 – 1500 with a target AGC value of 2e5 ions. MS/MS scans were acquired at Rapid scan rate (Ion Trap) with an AGC target of 5e3 ions and a max injection time of 200 ms. The total cycle time for MS and MS/MS scans was 2 sec. A 20s dynamic exclusion was employed to increase depth of coverage. The total analysis cycle time for each sample injection was approximately 2 hours. Following 13 total UPLC-MS/MS analyses (excluding conditioning runs but including 4 replicate QC injections; Supplementary Data 1), data was imported into Proteome Discoverer 2.2 (Thermo Scientific Inc.), and analyses were aligned based on the accurate mass and retention time of detected ions (“features”) using Minora Feature Detector algorithm in Proteome Discoverer. Relative peptide abundance was calculated based on area-under-the-curve (AUC) of the selected ion chromatograms of the aligned features across all runs. The MS/MS data was searched against the TrEMBL A. muciniphila database (downloaded in Apr 2017) and an equal number of reversed-sequence “decoys” for false discovery rate. The database was customized to include the Sus scrofa MUC5AC protein sequence (XP_020938242.1), the main component of pig gastric mucin. Mascot Distiller and Mascot Server (v 2.5, Matrix Sciences) were used to produce fragment ion spectra and to perform the database searches. Database search parameters included fixed modification on Cys (carbamidomethyl) and variable modifications on Meth (oxidation) and Asn and Gln (deamidation). Peptide Validator and Protein FDR Validator nodes in Proteome Discoverer were used to annotate the data at a maximum 1% protein false discovery rate.

The analysis identified 448 Mul1A interacting proteins. To further reduce this list to the most significant candidate proteins, we implemented several filtering steps. First, this list of proteins was filtered to only include proteins that were significantly enriched in the wild type strain vs the mul1A mutant negative control (P < 0.01) and to have > 1 peptide hit. The resulting 81 proteins were then screened for the presence of type I and type II signal peptides using the programs SignalP v3 and v4 and LipoP⁷². The resulting list of 32 proteins that were predicted to be secreted, and thus have a greater likelihood of co-localizing with Mul1A, were selected as probable Mul1A interactors. To visualize the interacting proteins, genes were plotted in Cytoscape⁷³. The size of each node was set to represent the Log₂ fold change normalized peptide abundance in wild type vs the mul1A mutant negative control and edge thickness represents the -Log₁₀(P value), the color of the nodes was used to represent their respective Pfam³² associations.

16S rRNA gene sequencing

Genomic DNA was extracted from fecal pellets collected from mice housed in separate cages using a Qiagen DNeasy PowerSoil DNA extraction kit (Qiagen). Amplicon sequencing was performed with primers targeting the v4 region of the 16S rRNA gene as previously described⁷⁴ and sequenced at Duke’s Sequencing and Genomic Technologies core facility using an Illumina MiSeq with paired end 250bp reads.

Taxonomy assignment and bioinformatic analysis

Demultiplexed sequences were obtained from the sequencing core. We used the DADA2 (v1.25.2)⁷⁵ R package to filter and trim the reads, keeping bases between positions 30 and 240. DADA2 was then used to denoise, dereplicate and merge the forward and reverse reads, and remove bimeras. Taxonomy was assigned using the Silva database (v138.1). Additional analysis was performed using the R packages phyloseq⁷⁶, ape, vegan, and microshades. Comparisons of microbial community composition were run using weighted UniFrac distances at the genus level Linear discriminant analysis Effect Size (LEfSe)⁷⁷ was run using default settings and accessed through the Huttenhower Galaxy website: https://huttenhower.sph.harvard.edu/galaxy.

Analysis of short chain fatty acids in the mouse caecum

Cecal contents were used to prepare 20% (w/v) slurries in 0.5 ml PBS. The slurries were acidified to a pH of <3 with 25 μl of 6N HCL and thoroughly mixed. Debris was removed by centrifugation at 14 000 x g for 5 min at 4°C, and the resulting supernatant was filtered through a 0.22 μm spin column. The filtrate was then transferred to an autosampler vial and analyzed on an Agilent 7890 gas chromatograph (GC) equipped with a flame-ionization detector (FID) and an Agilent HP-FFAP free fatty-acid column. The concentrations of acetate and propionate in the samples were determined using an 8-point standard curve (0.1 mM to 16 mM). Ratio of propionate to acetate was analyzed by a two-tailed student’s t test after removing two outliers identified by the Robust regression and Outlier removal (ROUT) method in GraphPad Prism with a Q coefficient value of 1%. Differences in acetate and propionate levels were analyzed by One Way ANOVA (F statistic) and Tukey’s post hoc test.

Statistical analysis

Unless otherwise noted, statistical analysis and plots were generated with GraphPad Prism version 9.0.0. Data distribution was assumed to be normal but this was not formally tested. Data collection and analysis were not performed blind to the conditions of the experiments. No statistical methods were used to pre-determine sample sizes but our sample sizes are similar to those reported in previous publications^53,78.

Extended Data

Extended Data Fig. 1. Akkermansia sp. are mucin specialists and the acquisition of mucin by A. muciniphila is selective and energy dependent.

(a) Growth curves, as assessed by optical density (OD₆₀₀) of a range of Gram-positive and Gram-negative mucin-degrading intestinal microbes, including A. muciniphila and A. glycaniphila, in the indicated medium. (b) A. muciniphila and Bacteroides thetaiotaomicron grown with fluorescein-mucin. The cells were grown with fluorescein mucin in a modified version of synthetic media with 0.25% mucin as the sole carbon source. Membranes were labeled with FM4–64. Experiments were three twice. (c-d) Mucin uptake is a specific and active process. A. muciniphila grown in the presence of either fluorescein-mucin or fluorescein-dextran (green) for 3 h and stained with anti-Akkermansia anti-sera (anti-Akk). All microscopy was performed at least three times (c). Flow cytometric analysis of cells grown in the presence of fluorescein-mucin for 3h, with or without pre-treatment with CCCP. Cells for flow cytometry were gated for the anti-Akkermansia positive population and the numbers under each curve represent the mean fluorescent intensity of fluorescein-mucin (d). A. muciniphila grown with fluorescein-mucin for 3h without CCCP or with CCCP treatment (e). Flow cytometry analyses of A. muciniphila grown with the cell permanent esterase carboxyfluorescein diacetate (CFDA) in the presence and absence of CCCP, and after heat inactivation (f). Scale bar = 1 μm. Error bars represent the standard error of the mean.

Extended Data Fig. 2. Transposon mutagenesis in A. muciniphila.

(a) Map of the A. muciniphila optimized INSeq plasmid. (b) Overview of the A. muciniphila conjugation protocol. (c) PCR analysis confirming transposition. DNA from representative Tn mutants was amplified with primers for A. muciniphila specific 16S rRNA, the bla gene located on the delivery plasmid backbone, and the cat gene located with the transposon. This analysis was performed for every transposition experiment. (d) Southern blot analysis of Tn mutant DNA digested with HindIII and probed with DIG-labeled probes that recognize the cat gene in the Tn insert. Data is representative of two experiments with similar results. (e-f) A Cartesian mapping strategy to identify Tn insertions. (e) Trade-off between genome coverage and clonal redundancy, using simulated subsets of the arrayed collection optimized for low redundancy. A series of 96-well plates drawn from the arrayed collection that minimizes clonal redundancy was identified by simulation. The tradeoff between increasing genome coverage (orange) and increasing clonal redundancy (blue) as the number of plates included from the optimized series grows (X-axis). (f) Estimating location mapping accuracy for different sizes of the optimized library. The distribution of clonal replicates for increasing sizes of the optimized library was drawn from the simulation. The estimated number of clones present in one, two, three, and four replicates are shown as a function of increasing collection size. For orthogonal pooling and Cartesian location mapping the search space for an individual clone scales as the number of replicates to the power of the number of pooling dimensions. A clone present in only one well has a unique plate-well address (1^3), while a clone with present three times in the collection would be mapped to 27 potential Plate-Row-Col locations (3^3).

Extended Data Fig. 3. INSeq analysis of relative nutritional requirements for A. muciniphila to grow in mucin medium and the role of putative glycan hydrolases.

Plot of INSeq data from Tn mutant pools grown for eight generations in mucin medium where each dot represents all inserts in a specific gene. Genes that belong to KEGG amino acid biosynthesis pathways are highlighted for cultures grown in (a) mucin medium and (b) mucin medium supplemented with Phytone. Predicted glycosyl hydrolases for A. muciniphila BAA-835 were identified using the CAZy database and highlighted on the INSeq plot for cultures grown in (c) mucin and in (d) mucin medium with Phytone. Statistical analysis on INSeq data was performed with a Mann-Whitney Utest. (e) Dropletseq analysis of A. muciniphila grown in mucin medium microdroplets. Tn mutants (Arrayed Pool) were injected into a microfluidic device at a low density to generate on average less than one bacterium per droplet. The graph displays the INSeq analysis and Log₂ fold change for cultures grown in mucin in batch culture (8 generations) versus single cell growth in droplets (72h). Selected genes that were depleted in one condition relative to the other are highlighted on the plot. GH, glycosyl hydrolase.

Extended Data Fig. 4. A significant proportion of A. muciniphila genes required for growth in mucin medium are specific to Akkermansia/Verru comicrobia.

(a) Number of genes required for optimal A. muciniphila growth in mucin medium that lack functional annotations. Genes corresponding to Tn mutants with a Log₂ > 2 fold decrease in abundance in mucin medium were used as the query for a BLAST search to identify potential homologs. The plot represents the number of genes encoding hypothetical proteins that were unique to Akkermansia spp. (Akk), homologs in other members of the PVC super phylum (PVC), homologs in other bacteria (other), and genes annotated as conserved hypothetical proteins (conserved). (b) Distribution of genes with Pfam designations belonging to pili or type II secretion families (Pili/T2SS), or TPR families in the INSeq analysis of genes required for growth in mucin medium in vitro, (c) in the cecum of germ-free mice, and (d) in the cecum of conventional mice.

Extended Data Fig. 5. Evidence for the presence of a stable Mul1A-Mul1B protein complex.

Transcriptional analysis of Mul1 operons. (a) View of RNAseq reads generated from wild type A. muciniphila grown in mucin medium mapped to genes in the mul1A and mul1B loci. (b) Growth curves for wild type A. muciniphila and mutants in mul1A and mul1B grown in triplicate in synthetic medium or with mucin as the sole carbon and nitrogen source and corresponding microscopy with FL-mucin (green). Cells are stained with anti-Akkermansia antisera (white). The scale bar is 1 μm. (c) Coomassie blue stained SDS-PAGE gel showing eluted proteins following immunoprecipitation with anti-Mul1 antibodies. Immunoprecipitations were performed with cell lysates from wild type A. muciniphila and in mul1A mutants. (d) Depiction of Conserved Domains (colors) in Muc5AC and locations of peptides identified as co-precipitating with Mul1A (vertical bars). The experiment was performed in triplicate.

Extended Data Fig. 6. Mucin utilization is required for A. muciniphila to compete in CONV mice and in Muc2^−/− mice.

A breeding colony of Akkermansia-free mice (Akk-free) was generated to facilitate mouse colonization without antibiotic pre-treatment. (a-c) Comparison of the microbiota of Akkermansia colonized (Akk-colonized) and Akk-free mice by 16S rRNA gene sequencing. (a) Relative abundances of fecal bacteria at the genus level in Akk-colonized and Akk-free mice. Each sample was obtained from separately housed mice (n = 3 per group). The center line is the mean and the whiskers show the minimum and maximum. (b) Principal Coordinates Analysis (PCoA) performed on weighted UniFrac distances. Statistical significance was determined by Permutational Multivariate Analysis of Variance (PERMANOVA). (c) Linear discriminant analysis Effect Size (LEfSe) analysis of Akk-colonized and Akk-free mice. The Kruskal-Wallis test was used to detect features with a significant differential abundance (p < 0.05). (d) Relative abundances of potential mucin-degrading taxa at the family level. (e) CONV mice were pre-treated with antibiotics and gavaged with a 1:1 mix of WT and mutant A. muciniphila prepared with a fecal slurry from Akk-free mice to partially reconstitute the microbiota (n = 6 per group). Bacterial loads in fecal pellets were quantified by qPCR, each point represents one cage. (f) Colonization of mucin deficient Muc2^−/− mice with A. muciniphila. Each point represents the average A. muciniphila per gram of feces (WT, n = 4; mul1A::Tn, n = 4; mul2A::Tn, n = 6). (g) Competition between wild type A. muciniphila and the mul1A::Tn mutant in Muc2^−/− mice. Mice were gavaged with a 1:1 mix of wild type and mutant and abundance was monitored over time using strain specific primers. Each point represents the average amount of A. muciniphila (n = 4), error bars represent the standard error.

Extended Data Fig. 7. The impact of mucin utilization by A. muciniphila in colonization along the GI tract, SCFA production and transcriptional responses.

(a) Abundance of A. muciniphila wild type and mul1A mutants along the GI tract of female GF mice (n =3). Intestinal contents were scraped from sections along the GI tract and A. muciniphila levels were quantified by qPCR. Data are presented as mean values +/− SEM. The analysis was carried out with the same female mice that were used for RNAseq. (b) Expression of cholesterol biosynthesis genes in male and female mice, and control mice gavaged with sterile PBS. (c) Normalized expression of genes that are pivotal to cholesterol biosynthesis (Hmgcr) and uptake (Ldlr) in relation to cecal acetate and propionate levels. (d) Representative single cell RNAseq expression data from the Tabula Muris⁴⁸. Violin plots show the expression of Ldlr and Hmgcr in mouse colonic epithelial and goblet cells.

Supplementary Material

MovS2

Extended Video S2 Live imaging of A. muciniphila grown with fluorescein labeled mucin under an agarose pad, with and without the addition of 50 μM CCCP. Images were captured every 30s for 20 min. Arrows indicate cells with active mucinosome formation (untreated), and locations where mucin accumulates at the cell surface but fails to form mucinosomes in the presence of CCCP.

MpvS1

Extended Video S1 View through an orthogonal section of a 3D-reconstructed STED image of a mucinosome inside A. muciniphila. Fluorescein mucin is shown in magenta and the A. muciniphila cell surface (anti-Akk) is cyan.

Supplementary Data 1

Supplementary Data 1. Data tables with INSeq, RNAseq, mass spectrometry outputs, and primer and adaptor sequences.

Data availability

Analyzed data is available in the ReportingTools data file. Sequencing data used for this study can be found at NCBI BioProject PRJNA955715. Unprocessed data from mass spectrometry and additional supporting data are available upon reasonable request from the corresponding authors.