The Akkermansia spp. strains isolated in this study originated from fecal samples of five healthy human subjects. Strain isolation refers to the method of Mushtaq, with some modifications (Mushtaq et al., 2021). Briefly, 1 g of fecal material was vortex-mixed with 9 mL of sterile physiological saline in an anaerobic chamber to prepare a 10⁻¹ dilution. Serial ten-fold dilutions were then performed up to 10⁻⁶ dilution. A 100 μL aliquot of the 10⁻⁶ dilution was spread onto blood agar plates (Jiangmen Kailin P0903). After uniform spreading, plates were incubated in an inverted position at 37 °C under anaerobic conditions for 3–5 days to obtain isolated colonies. Single colonies were selected, and the near-full-length 16S rRNA gene was amplified using universal primers 27F and 1492R (Heuer et al., 1997). The resulting sequences were submitted to the NCBI Nucleotide BLAST database for analysis. Sequence comparisons confirmed that the isolates belonged to the genus Akkermansia. A. muciniphila AKM Lab-01 was isolated from a fecal sample donated by a healthy 56-year-old male volunteer in Dali, Yunnan, China, and has been deposited at Guangdong Microbial Culture Collection Center (GDMCC) under accession number GDMCC 63782. They were cultured in self-optimized MM01 liquid medium (peptone, 15 g; glucose, 20 g; yeast extract, 15 g; cysteine, 1 g; sodium acetate, 5 g; sodium citrate, 4 g; dipotassium phosphate, 2 g; magnesium sulfate, 0.1 g; manganese sulfate, 0.05 g; Tween 80, 1 g contained per liter; pH 6.3–6.5) and grown anaerobically under an atmosphere of 10% CO₂, 10% H₂ and 80% N₂ at 37 °C for 48 h. All the reagents were vented in an anaerobic atmosphere for at least 24 h prior to use. Cultures were concentrated in deoxygenated PBS (containing 0.05% Cys-HCL) with 25% (v/v) glycerol under strictly controlled anaerobic conditions for use as a live bacterium. The A. muciniphila culture was pasteurized at 70 °C for 30 min, followed by immediate refrigeration at −80 °C until further use. In addition, pasteurized A. muciniphila was centrifuged and re-suspended in sterile pure water and then freeze-dried for 48 h in a vacuum freeze-dryer to obtain A. muciniphila drug substance.

Bacterial genomic DNA was extracted from bacterial pellet samples using the ALFA-SEQ Fast Magnetic Bacteria DNA Kit (DC313-03, Findrop, China), following the manufacturer's instructions. DNA libraries were constructed using the VAHTS Universal Plus DNA Library Prep Kit for Illumina V4 (ND801, Vazyme, China) in accordance with the manufacturer's protocol. The whole genome shotgun sequencing was performed on the NextSeq 2000 platform. Reads were filtered with fastp (version 0.20.0), assembled with SPAdes (version 3.14.0), and annotated with PROKKA (version 1.14.5). The genome assembly accession number is JBQKMJ000000000, and the public repository URL address is https://www.ncbi.nlm.nih.gov/nuccore/JBQKMJ000000000. Orthofinder (version 2.5.5) was used for comparative analysis of 123 genomes, including strains from NCBI. The Clusters of Orthologous Groups (COGs) databases identified functional categories of gene families in A. muciniphila and A. massiliensis. A phylogenetic tree was constructed using IQ-TREE (version 2.2.2.7) with single-copy core genes aligned by MAFFT (version 7.520) and refined by Gblocks (version 0.91b; Minh et al., 2020). Given that Amuc_1409, Amuc_1100, and P9 proteins are well-recognized as core functional effectors mediating Akkermansia's probiotic activities in metabolic regulation (e.g., intestinal barrier maintenance, inflammation modulation, and GLP-1 induction), we focused on these three proteins for homology analysis. Homology analyses were performed on Amuc_1409, Amuc_1100, and P9 proteins using blastp (version 2.14.0+) against the corresponding proteins in A. muciniphila ATCC BAA−835. Virulence factors were identified by Virulence Factor Database (VFDB).

Micrographs of A. muciniphila AKM Lab-01 were obtained using scanning electron microscopy (SEM). Bacterial cells were prefixed in 3 % (v/v) glutaraldehyde at 4 °C for 4–24 h, rinsed six times (20 min each) with 0.1 M phosphate-buffered saline (PBS, pH 7.0), and then dehydrated through an ethanol series: 30 % and 50 % (two changes, 10 min each), 70 % (overnight at 4 °C), 90 % (one change, 15 min) and 100 % (three changes, 15 min). Samples were critical-point-dried, sputter-coated with gold, and examined by SEM (Apreo 2S, Thermo Scientific).

The methods for assessing tolerance to pH and bile salts were referred to in a previous publication (Hassanzadazar et al., 2012). A. muciniphila AKM Lab-01 was initially inoculated into MM01 medium at 37 °C for 72 h. Subsequently, the activated A. muciniphila AKM Lab-01 was inoculated into MM01 medium adjusted to different pH levels (3 to 10) or containing various concentrations of bile salts (0 to 0.4%) and incubated for 48 h at 37 °C. After culture, the optical density at 600 nm was measured.

Auto-aggregation assessment was referred to the previous publication (Cozzolino et al., 2020), A. muciniphila AKM Lab-01 was cultured in MM01 medium and incubated at 37 °C for 48 h. Cell precipitates were obtained by centrifugation (4 °C; 6,000 rpm, 10 min), washed twice, and resuspended in phosphate-buffered saline (PBS, pH 7). The optical density was adjusted to 0.5 at 600 nm. The bacterial suspension was then incubated at 37 °C, and the optical density at 600 nm was measured every 30 min for up to 6 h. The auto-aggregation of AKM Lab-01 was calculated using the following formula (Cozzolino et al., 2020):

where OD₀ is the initial optical density at time 0, and OD_final is the initial optical density measured after 0.5, 1, 1.5, 2.5, 3.5, 4.5, 5, 5.5, and 6 h.

The method was referred to the previous publication (Francisco et al., 2023). A. muciniphila AKM Lab-01 was cultured in MM01 medium at 37 °C for 48 h. After incubation, cell precipitates were collected by centrifugation at 4 °C and 6,000 rpm for 10 min, washed twice, and resuspended in phosphate-buffered saline (PBS, pH 7). The optical density was adjusted to an OD₆₀₀ of 1.0. A 100 μL aliquot of the bacterial suspension was then spread onto an MM01 agar plate. An antibiotic test strip (Liofilchem, Italy) was placed on the surface of the agar. The antibiotic diffuses radially into the agar, creating a concentration gradient that inhibits the growth of susceptible bacteria, forming a zone of inhibition. The size of this zone correlates with the antibiotic concentration. By measuring the intersection points of the inhibition zone with the test strip, the minimum inhibitory concentration (MIC) of the antibiotic against the specific bacterium was accurately determined (Table 1).

Quantitative detection of Amuc_1100 was performed by mass spectrometry with Multiple Reaction Monitoring (MRM) was referred to the previous publication with modifications (Zhang et al., 2011). Briefly, bacterial total protein was extracted using urea lysis buffer (24.0240 g urea, 0.6057 g Tris base in 40 mL), and the protein concentration was measured using the BCA assay (Bicinchoninic Acid Assay). Proteins were precipitated with acetone and redissolved in 8 M Urea. The proteins were reduced with DTT (Dithiothreitol, 10 mM) and alkylated with IAA (Iodoacetamide, 50 mM), then diluted to 1 M urea in the sample, enzymatically digested at 37 °C overnight at a 1:50 mass ratio (trypsin: protein) and finally desalted and lyophilized for mass spectrometry. The characteristic peptide of the Amuc_1100 protein was synthesized and subsequently used for analysis via triple quadrupole LC-MS/MS (Agilent 1290-6470B) to establish a quantitative standard curve for Amuc_1100. Based on this standard curve, the signal intensity of the characteristic peptide in the samples was quantified by mass spectrometry, allowing the final determination of the Amuc_1100 protein content in the samples.

Five to six weeks old male C57BL/6J mice (GemPharmatech Inc.) were maintained in a specific pathogen free (SPF) environment. All animal experiments were carried out according to protocols approved by the Institutional Animal Care and Use Committee of Moon Biotech Co., Ltd, in compliance with the Guide for the Care and Use of Laboratory Animals. All mice were cared for under controlled temperature and humidity (24 ± 2 °C, 50% ± 10%) with a 12 h light/dark cycle. Animals exhibiting abnormal behavior or physiological conditions were excluded prior to group assignment. All mice were randomly assigned to experimental groups by body weight to ensure balanced distribution across groups. Experimental groups were identified exclusively by alphanumeric codes, and the experimenters were unaware of treatment group identities.

For (i) Experiment 1 (IACUC Number: IACUC202407-4O, a prevention model): Forty-two male C57BL/6J mice were randomly divided into 7 groups after a 1-week acclimation period: one normal chow diet (NCD) group (normal diet: 3.8 kcal/g, with 20% of energy derived from protein, 10% from fat, and 70% from carbohydrate; Diet D12450B; Research Diets, Inc., New Brunswick, New Jersey, USA) and six high-fat diet (HFD)-fed groups [HFD: 5.2 kcal/g, with 20% of energy derived from protein, 60% from fat, and 20% from carbohydrate (Diet D12492; Research Diets, Inc., New Brunswick, New Jersey, USA)] (n = 6 mice per group). All diets were administered for 8 weeks. Mice in the A. muciniphila S1–S5 groups (AKK groups) received daily oral gavage of the corresponding A. muciniphila strains [10⁹ colony-forming units (CFU), used for live bacteria counting, in 200 μL PBS], while those in the NCD and HFD groups received 200 μL PBS via oral gavage from Day 1. According to the experimental protocol, body weight and food intake was measured twice weekly. On Day 56, mice were euthanized by cervical dislocation following anesthesia induction with 4–5% isoflurane (R510-22-10, Jinan Ante Biochemical Pharmaceutical Co., Ltd.) in oxygen (200–400 mL/min) delivered via an inhalation anesthesia system (CL-1000-S4, Shanghai Yuyan Instruments Co., Ltd.) within an induction chamber.

For (ii) Experiment 2 (IACUC Number: IACUC202407-14O, a treatment model): After 1 week of acclimatization, 20 mice were fed HFD (D12492; Research Diets, inc., New Brunswick, New jersey) for 10 weeks to build an obesity model. Subsequently, 12 mice weighing 38–46 g were selected and randomly divided into two groups: HFD-Control group and viable AKM Lab-01 group. Obese mice in these groups received daily oral gavage (0.2 mL per animal) for 28 days: PBS for the HFD-Control group (n = 6) and viable AKM Lab-01 (10⁹ CFU) for the treatment group (n = 6) from week 10. According to experimental procedure, body weight and 24 h food intake was measured weekly or twice weekly. One week before the experiment's endpoint, glucose tolerance tests were conducted as follows: After a 12-h fasting period, the mice were weighed and administered a glucose solution at a dose of 2 g/kg body weight via oral gavage. Blood samples were collected from the tail tip at 0, 15, 30, 60, 90, and 120 min to measure serum glucose levels. Glucose concentrations were determined using a glucometer with glucose test strips. On Day 28, all mice underwent an overnight fast prior to terminal procedures. An inhalation anesthesia apparatus (CL-1000-S4, Shanghai Yuyan Instruments Co., Ltd.) was utilized for mouse sedation. Within the induction chamber, animals received 4–5% isoflurane (R510-22-10, Jinan Ante Biochemical Pharmaceutical Co., Ltd.) in oxygen (flow rate: 200–400 mL/min) until loss of righting reflex was observed. The mice were then positioned in a nose cone for maintenance of surgical anesthesia using 1.5–2.5% isoflurane. Orbital venous plexus blood collection was carried out under deep terminal anesthesia, immediately followed by euthanasia via cervical dislocation. Adipose depots (epididymal fat, perirenal fat, mesenteric fat and Inguinal fat) and liver were precisely dissected and weighed. Epididymal fat, liver and colon tissues were fixed in paraformaldehyde or Karnoy's solution and subjected to histological staining and analysis. Serum chemistry parameters (TCHO, TG, LDL, HDL, ALT, AST, BUN, and CRE) were tested by Wuhan Servicebio Technology Co., Ltd. using automatic biochemical analyzer.

For (iii) Experiment 3 (IACUC Number: IACUC202408-7O, a treatment model): After 1 week of acclimatization, 40 C57BL/6J male mice were randomly divided into 5 groups according to body weight (8 mice/group, 4 mice per cage). The NCD group were fed a normal control diet (D12450B; Research Diets, inc., New Brunswick, New jersey). The remaining four groups were fed HFD (D12492; Research Diets, inc., New Brunswick, New jersey). After 4 weeks on HFD, the mice had reached mild obesity (body weight 28–34 g) and were stratified by weight into four groups (n = 8 each): HFD-Control, viable AKM Lab-01 (AKM Lab-01-L, 10¹⁰ CFU per animal), pasteurized AKM Lab-01 (AKM Lab-01-P, 10¹⁰ CFU per animal), and pasteurized AKM Lab-01 drug substance (AKM Lab-01-D, 10¹⁰ CFU per animal). From week 5 onward the NCD-Control and HFD-Control groups received 0.2 mL PBS daily by oral gavage for 56 days, while the three treatment groups received 0.2 mL of the corresponding bacterial preparation daily for the same period. According to the experimental procedure, body weight and food intake was measured twice weekly. On Day 56, after an overnight fast, mice were anesthetized with isoflurane (4–5% induction, 1.5–2.5% maintenance). Orbital venous plexus blood was collected under deep anesthesia before euthanasia by cervical dislocation. Tissues and fecal samples were then harvested from all animals for subsequent analyses.

For (iii) Experiment 4 (IACUC Number: IACUC202507-4O, a treatment model): After a 1-week acclimatization, thirty male C57BL/6J mice (6 weeks old) were fed a high-fat diet (HFD; D12492, Research Diets, New Brunswick, NJ, USA) for 10 weeks to induce obesity. Eighteen animals that reached 36–44 g were then selected and randomly assigned to three groups (n = 6 each): The HFD-Control group received 0.2 mL PBS daily, while the AKM Lab-01 and BAA-835 groups received 1 × 10⁹ CFU of the respective live strain in 0.2 mL PBS daily; all gavages were given once daily for 28 consecutive days from week 10. Body weight and food intake was measured twice weekly. On Day 28, after overnight fasting, mice were anesthetized with isoflurane (4–5% induction, 1.5–2.5% maintenance) for orbital blood collection under deep anesthesia, followed by cervical dislocation euthanasia and tissue harvesting.

Liver and adipose tissue samples were fixed in paraformaldehyde, stained with hematoxylin and eosin (H&E), and examined for pathology by Wuhan Servicebio Technology Co., Ltd. The tissue was photographed at 200 × magnification. The severity of non-alcoholic fatty liver disease (NAFLD) was assessed using inflammation, steatosis, and ballooning scores, as well as the NAFLD activity score (NAS).

The thickness and integrity of the colonic mucus layer were evaluated using Alcian Blue staining. Colon tissues were fixed in Karnoy's solution and stained with Alcian Blue, which selectively binds to acidic mucins, resulting in blue staining. The tissues were photographed at 200 × magnification. The mucus layer thickness was measured under a light microscope to assess the structural integrity of the colonic mucus.

Two hundred fifty microliter blood was quickly mixed with 0.1 nmol/L Dipeptidylpeptidase IV Inhibitor I (Sigma, 416200), 7.5 μL 0.5 M EDTA, and 2.5 μL Protease Inhibitor Cocktail (Sigma, P8340)). Plasma was then isolated by centrifugation and stored at −80 °C until subsequent hormone analysis. Total GLP-1 was measured using Glucagon-Like Peptide-1 (Active) ELISA kit (Merck, EGLP-35K) according to the provided protocols.

Bacterial genomic DNA was extracted by Guangdong Magigene Biotechnology Co., Ltd. (Guangzhou, China) using commercial kits. DNA integrity and purity were monitored on 1% agarose gels. DNA concentration and purity were measured using Qubit 3.0 (Thermo Fisher Scientific, Waltham, USA) and Nanodrop One (Thermo Fisher Scientific, Waltham, USA). Sequencing libraries were generated using ALFA-SEQ DNA Library Prep Kit. The library quality was assessed on the Qubit 4.0 Fluorometer (Life Technologies, Grand Island, NY, USA) and Qsep400 High-Throughput Nucleic Acid Protein Analysis system (Houze Biological Technology Co., Hangzhou, China). At last, the library was sequenced on an Illumina NovaSeq 6000 platform and 150 bp paired-end reads were generated. Raw sequencing reads were quality-filtered using fastp (version 0.20.0). Kraken2 (version 2.1.3) and Bracken (version 2.9) were used to remove host reads and to generate the relative abundance in each sample. Subsequent alpha diversity analysis, PCoA and non-parametric test were performed on R scripts.

Serum samples collected from the mouse study (Experiment 3) were used to quantify the concentrations of lipopolysaccharides (LPS), tumor necrosis factor α (TNF-α), and interleukin-6 (IL-6) via enzyme-linked immunosorbent assay (ELISA), following the manufacturers' instructions with specific kits: LPS was detected using the Mouse Lipopolysaccharides (LPS) ELISA Kit (Cusabio, Catalog No.: CSB-E13066m), TNF-α using the Mouse Tumor Necrosis Factor α (TNF-α) ELISA Kit (Cusabio, Catalog No.: CSB-E04741m), and IL-6 using the Mouse IL-6 ELISA Kit (MULTI SCIENCES, Catalog No.: EK206). Serum samples were diluted to the recommended concentrations, absorbance was measured at the specified wavelengths, and target molecule concentrations were calculated via standard curve fitting; each sample was assayed in triplicate to ensure data reliability.

To evaluate the potential systemic toxicity of AKM Lab-01 drug substance (freeze-dried pasteurized A. muciniphila Lab-01), a 90-day subchronic oral toxicity study was conducted in vivo according to the OECD guidelines (IACUC Number: IACUC202408-5AT). Forty 6- to 7-week-old BALB/c mice (20 males and 20 females) were obtained from GemPharmatech Inc. and housed under controlled conditions (temperature 24 ± 2 °C, humidity 50% ± 10%, 12-h light/dark cycle). After a 1-week acclimatization period, the mice were randomly divided into four groups: Group 1 received saline only; Groups 2, 3, and 4 received AKM Lab-01 drug substance at dosages of 8.8 × 10¹⁰, 8.8 × 10⁹, and 8.8 × 10⁸ TFU/mouse per day, respectively. The suspensions were administered orally and daily for 90 days. All the mice were fed standard maintenance chow diet (3.8 kcal/g, with 21.5% of energy derived from protein, 11.1% from fat and 67.4% from carbohydrate; Diet XTI01WC-009, Jiangsu Province Collaborative Pharmaceutical Biotechnology Engineering Co., Ltd., China) during the whole study. Clinical signs were recorded daily, and comprehensive clinical observations were performed weekly. Body weight, food intake, and water consumption were monitored weekly. After 90 days, following an overnight fast, mice received isoflurane anesthesia (induction: 4–5%; maintenance: 1.5–2.5%). Orbital venous plexus blood was collected under deep terminal anesthesia before euthanasia by cervical dislocation. Tissue specimens were subsequently harvested from all animals. Blood samples were analyzed for hematology, coagulation, and clinical chemistry, while tissues were examined for macroscopic changes, organ weights, and histopathology.

Statistical analysis was performed using GraphPad Prism (version 10.2.3), and data are presented as the mean ± standard deviation (Mean ± SD) unless otherwise indicated. Raw data from in-vitro assays were pre-processed with Cytation5 software (version 3.11.19) prior to analysis. Normality was assessed with the Shapiro–Wilk test, and homogeneity of variances with Levene's test; outliers were identified and excluded using the ROUT method (Q = 1%). For multiple-group comparisons among more than three groups, one-way ANOVA with Dunnett's multiple comparison test was applied. For all pairwise comparisons following ANOVA, Tukey's multiple comparisons test was applied. When analyzing two independent variables, two-way ANOVA with Dunnett's post hoc test was used. For two-group comparisons, a two-tailed unpaired t-test was used when data were normally distributed and variances were equal; otherwise, the Mann–Whitney U test was applied. All graphical representations were generated using GraphPad Prism (version 10.2.3). Significance levels are indicated as: ns, not significant; ^*p < 0.05; ^**p < 0.01; ^**p < 0.001; ^****p < 0.0001.

In our study, we isolated 222 strains of Akkermansia from healthy human feces. To minimize strain redundancy, we retained only one strain isolated from each donor's fecal sample and selected 26 strains for whole-genome sequencing. These 26 strains, together with 97 additional strains obtained from public databases, were used to form a comprehensive genomic dataset of Akkermansia strains, designated as Supplementary Dataset S1. Firstly, a phylogenetic tree was constructed using IQ-TREE based on 338 single-copy core genes (Supplementary Dataset S4). Verrucomicrobium spinosum DSM 4136, Rubritalea marina DSM 17716, Terrimicrobium sacchariphilum NM-5, and Rubritalea squalenifaciens DSM 18772 were selected as out-groups. To ensure the accuracy of our pan-genome analysis, strains were identified based on the following criteria: they must be in the same branch of the phylogenetic tree, exhibit ≥95% Average Nucleotide Identity (ANI) with the type strain of the species (if available), show ≥95% ANI with other strains within the same species, and have < 95% ANI with strains of other species. Based on these criteria, we labeled strains in the same branch as the type strains of A. muciniphila (n = 88) and A. massiliensis (n = 14) for further analysis (Figure 1). We found that strains such as Akkermansia sp. CAG:344 and Akkermansia muciniphila BSH01 were in the same branch as the type strain of Akkermansia massiliensis, suggesting potential inaccuracies in previous species identification (Guo et al., 2017; Xing et al., 2019).

Additionally, two potential novel species were identified. The first potential novel species, including Akkermansia sp. MNH40091, Akkermansia sp. UBA7059, Akkermansia sp. UBA3271, and Akkermansia sp. UBA7090, was closely related to A. muciniphila. The second potential novel species, including Akkermansia muciniphila GP22 and Akkermansia muciniphila GP24, was in an outer branch of the large clade containing A. muciniphila, A. massiliensis, and the first potential novel species (Figure 1). Moreover, strains previously identified as Candidatus Akkermansia timonensis (including Candidatus Akkermansia timonensis Akk2633, Candidatus Akkermansia timonensis Akk0496a, Candidatus Akkermansia timonensis Akk0490, Candidatus Akkermansia timonensis Akk0496b) and Akkermansia sp. (including Akkermansia sp. KLE1605, Akkermansia sp. KLE1797, Akkermansia sp. KLE1798) were reclassified as Akkermansia biwaensis (Figure 1). In total, eight Akkermansia species were included in our study: A. muciniphila, Akkermansia massiliensis, the potential novel species closely related to A. muciniphila, the potential novel species in the outer branch, Akkermansia biwaensis, Candidatus Akkermansia intestinavium, Candidatus Akkermansia intestinigallinarum, and Akkermansia glycaniphila.

Subsequently, we conducted a pan-genome analysis. For A. muciniphila, the analysis revealed a total of 4,167 gene families, with 1,448 core gene families shared among the 88 strains (Supplementary Dataset S2). The rarefaction curve showed that the increase in both total and core gene families slowed down as new strains were added, indicating that our sample size for A. muciniphila is representative (Supplementary Figure S1A). In contrast, A. massiliensis had a total of 2,952 gene families, with 1,817 core gene families among the 14 strains (Supplementary Dataset S2). Given that A. massiliensis is a newly discovered species, its strain number is significantly lower than that of A. muciniphila. The instability in the rarefaction curves for both total and core gene families indicates that the sample size for A. massiliensis is insufficient (Supplementary Figure S1B).

After obtaining core and strain-specific gene sequences, we compared their functional categories using the Clusters of Orthologous Groups (COGs) database (Tatusov et al., 2003). In A. muciniphila, COG annotation covered 75.7% (1096/1448) of core gene families, 27.4% (575/2099) of accessory gene families, and 9.4% (58/620) of strain-specific gene families. In contrast, in A. massiliensis, COG annotation covered 71.1% (1292/1817) of core gene families, 26.3% (256/975) of accessory gene families, and 5.6% (9/160) of strain-specific gene families. When comparing species-unique core gene families, A. muciniphila had 64.7% (121/187) annotated, while A. massiliensis had only 57.4% (319/556) annotated (Figure 2A, Supplementary Dataset S2). The lower annotation rates, especially for accessory and strain-specific gene families, suggest that the functions of many strains in these two species remain poorly studied. The COG annotation results for A. muciniphila core gene families were consistent with previous research at the functional class level (Xing et al., 2019), while A. massiliensis core gene families showed a similar functional class composition. The species-unique core gene families between A. muciniphila and A. massiliensis were primarily annotated as metabolism, cellular processing and signaling, and information storage and processing (Figure 2A). Detailed annotations showed that the three most abundant functions among core gene families annotated as metabolism were amino acid transport and metabolism, carbohydrate transport and metabolism, and coenzyme transport and metabolism (Figures 2B, C). Notably, A. massiliensis had more core gene families annotated in these categories than A. muciniphila.

Homology analyses of key proteins—Amuc_1409, Amuc_1100, and P9—revealed that Amuc_1409 is conserved in A. muciniphila and closely related new species. Meanwhile, Amuc_1100 and P9 are present in most strains of the Akkermansia genus, except for strains such as Candidatus Akkermansia intestinavium ChiGjej6B6-8097, Candidatus Akkermansia intestinigallinarum 14975, and Akkermansia glycaniphila APytT (Figure 1). Previous studies have shown that Amuc_1409 promotes gut health by enhancing intestinal stem cell proliferation and regeneration (Kang et al., 2024). Therefore, A. muciniphila and closely related new species are likely to share similar probiotic functions, such as promoting intestinal epithelial development and improving barrier integrity. Amuc_1100 strengthens epithelial tight junctions and reduces body weight and fat mass through TLR2 activation (Plovier et al., 2017; Wang et al., 2020, 2021). Additionally, this protein triggers a range of anti- and pro-inflammatory cytokines, thereby preventing obesity, insulin resistance, and inflammation in visceral adipose tissue (Si et al., 2022). Moreover, P9 induces the secretion of glucagon-like peptide-1 and activates brown adipose tissue thermogenesis (Yoon et al., 2021). These findings suggest that most strains of the Akkermansia genus possess probiotic functions associated with weight loss.

To identify the optimal Akkermansia strain for preventing obesity and improving related metabolic disorders, we conducted an exploratory, product-oriented screening of 26 isolates in our strain library. First, genome-based prediction of virulence factors and core-genome COG annotation were performed; all isolates carried similarly low burdens of putative virulence genes, and functional profiles for amino-acid, carbohydrate and coenzyme transport/metabolism were highly similar across the set (Supplementary Dataset S3). Consequently, these genomic criteria served as qualitative references rather than exclusion filters. Next, cultivation performance was quantified by OD600 after 72 h in MM01 medium (Supplementary Figure S2); the five isolates exhibiting the highest biomass yield (S1 > S2 ≈ S3 ≈ S4 ≈ S5) were selected for in-vivo testing. These five isolates consist of four A. muciniphila strains—representing the most prevalent species in healthy human gut samples—and one A. massiliensis strain, which is the second-most abundant species. COG analysis revealed that A. massiliensis harbors a greater number of species-specific core gene families across several functional sub-categories than A. muciniphila, indicating a potentially broader substrate-utilization capacity (Figure 2A, Supplementary Dataset S2 and Supplementary Dataset S3). Consequently, these five Akkermansia strains were selected for efficacy evaluation in HFD-induced obese mice model with supplementation concurrently with the high-fat diet (Figure 3A). After an 8-week intervention, A. muciniphila S1 exhibited a substantial inhibitory effect on weight gain, significantly outperforming the HFD-Control group (Figures 3B, C). A. muciniphila S2, S3, S4, and S5 also demonstrated trends toward weight gain prevention, although these effects did not reach statistical significance (Figures 3B, C). Given its superior efficacy, A. muciniphila S1 was selected for further development and was renamed AKM Lab-01.

To assess the effects of AKM Lab-01 on diet-induced obesity and associated metabolic disorders, we established an obesity model in mice by feeding them HFD for 10 weeks, followed by a 4-week intervention with AKM Lab-01 (Figure 3D). After the intervention, we evaluated several key indicators, including body weight, food intake, blood glucose levels, liver and kidney function, and gut barrier integrity. The results demonstrated that AKM Lab-01 significantly reduced body weight gain (Figures 3E, F) in HFD-induced obese mice compared to the HFD-Control group. Additionally, AKM Lab-01 treatment led to a significant decrease in food intake (Figure 3G), fasting blood glucose levels (Figure 3H), and area under the curve (AUC) during an oral glucose tolerance test (OGTT; Figures 3I, J). Furthermore, AKM Lab-01 significantly improved liver function, as evidenced by reduced levels of alanine aminotransferase (ALT, Figure 3K) and aspartate aminotransferase (AST, Figure 3L). It also enhanced kidney function, with significant reductions in blood urea nitrogen (BUN, Figure 3M) and creatinine (CRE, Figure 3N). Moreover, AKM Lab-01 increased the thickness of the colonic mucus layer (Figure 3O), indicating improved gut barrier function. Collectively, these findings suggest that AKM Lab-01 can effectively mitigate diet-induced obesity and associated metabolic disorders by modulating multiple physiological pathways, including energy balance, glucose metabolism, liver and kidney function, and gut barrier integrity.

To further elucidate the effects of AKM Lab-01 on lipid metabolism in HFD-induced obese mice, we examined its impact on serum lipid profiles, including total cholesterol (TCHO), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) during the 4-week intervention period (Figure 4A). AKM Lab-01 significantly reduced serum TCHO levels compared to the HFD-Control group (Figure 4A).

In addition to its effects on serum lipids, AKM Lab-01 also exerted significant effects on adipose tissue. It markedly decreased the weights of epididymal fat, perirenal fat, mesenteric fat, and inguinal fat, as well as the proportion of visceral fat to body weight (Figure 4A). Histopathological analysis of adipose tissue revealed that AKM Lab-01 reduced adipocyte size, indicating a reduction in adipose tissue hypertrophy (Figure 4B). Regarding liver health, AKM Lab-01 significantly decreased liver weight and the proportion of liver weight to body weight (Figure 4A). Histopathological analysis of liver tissue showed that AKM Lab-01 reduced hepatic lipid accumulation, lobular inflammation, hepatocellular ballooning, and the non-alcoholic fatty liver disease (NAFLD) activity score compared to the HFD-Control group (Figure 4C). Collectively, these findings demonstrate that AKM Lab-01 can significantly improve lipid metabolism and ameliorate adipose tissue and liver pathology in HFD-induced obese mice, suggesting its potential as a therapeutic agent for obesity-related metabolic disorders. To benchmark the anti-obesity efficacy of AKM Lab-01, we compared it with the widely used reference strain, live A. muciniphila ATCC BAA-835. After 10 weeks of HFD induction, obese mice were treated with 1 × 10⁹ CFU of AKM Lab-01 or BAA-835 daily for 28 days. Both strains significantly reduced cumulative weight gain compared to HFD-Control, with no significant difference between them, indicating that AKM Lab-01 is at least as effective as the standard strain in this obesity model (Supplementary Figure S5).

To clarify A. muciniphila AKM Lab-01′s taxonomic status and genomic distinctiveness with reference Akkermansia strains, we performed the following analyses: First, we aligned the 16S rRNA sequence of AKM Lab-01 with that of reference Akkermansia strains. Phylogenetic analysis of the 16S rRNA gene confirmed that AKM Lab-01 clusters within the same clade as other validated A. muciniphila strains (Supplementary Figure S3A). Furthermore, AKM Lab-01 shared >99% ANI with A. muciniphila BAA-835 (Supplementary Figure S3B), confirming AKM Lab-01′s assignment to the species A. muciniphila. Subsequently, we performed comparative genomic analysis through COG annotation of AKM Lab-01, BAA-835, and three additional A. muciniphila strains examined in this study. This analysis revealed a relative enrichment of genes assigned to COG category P (Inorganic ion transport and metabolism) in AKM Lab-01 (Supplementary Figure S3C). This distinctive genomic signature may indicate unique functional attributes potentially contributing to its probiotic characteristics, though the underlying mechanism requires further study.

A. muciniphila AKM Lab-01 is a Gram-negative, strictly anaerobic, non-motile, and non-spore-forming bacterium that resides in the human gut (Figure 5A). Individual cells typically measure between 0.5 and 1.0 μm in length and can be observed singly, in pairs, in short chains, or in aggregates (Figures 5A, B). This strain exhibits a broad pH tolerance, with growth observed at pH levels ranging from 6 to 10 (Figure 5C). For the bile salt tolerance assay, AKM Lab-01 exhibited effective growth in media containing bile salt concentrations up to 0.25% (p < 0.05), indicating its favorable bile salts tolerance (Figure 5D). To effectively exert beneficial effects, probiotics must achieve adequate mass through aggregation, a trait highly desirable due to its impact on adhesion to intestinal epithelial cells (Gómez et al., 2016). AKM Lab-01 exhibits an auto-aggregation capacity of 54% within 6 h (Figure 5E). Collectively, these findings suggest that AKM Lab-01 possesses potential probiotic characteristics and is capable of effectively colonizing the gut.

To further evaluate the effects of different AKM Lab-01 formulations, we established a mild obesity model in mice by feeding them HFD for 4 weeks, followed by an 8-week intervention with three distinct formulations: viable (live) cells, pasteurized cells, and the drug substance form (Figure 6A). Our results revealed that all these three formulations-viable, pasteurized, and drug substance-effectively reduced weight gain in HFD-induced obese mice (Figures 6B, C). Previous studies have demonstrated that A. muciniphila combats obesity by regulating host energy metabolism through several mechanisms. Specifically, it thickens the intestinal mucus layer, thereby enhancing gut barrier function via the action of Amuc_1100 and reducing endotoxin entry into the bloodstream. This, in turn, lessens systemic inflammation (Mo et al., 2024). Additionally, A. muciniphila boosts the secretion of gut hormones like GLP-1 by producing SCFAs, particularly acetate and propionate. It also acts through the secreted protein P9 or the membrane protein Amuc_1100, thereby improving glucose tolerance and insulin sensitivity (Yoon et al., 2021). In the present study, we measured GLP-1 levels in blood samples from mice. The results showed that GLP-1 levels were significantly higher in the AKM Lab-01 treatment groups compared to the HFD-Control group (Figure 6D). Furthermore, we used mass spectrometry to quantify the content of Amuc_1100 protein in the different formulations of AKM Lab-01. The results demonstrated that the concentration of Amuc_1100 protein was 350 ng/mg in viable bacterial preparations and 450 ng/mg in the drug substance form (Figure 6E). Additionally, we assessed inflammation-related markers in mouse serum following AKM Lab-01 treatment. LPS levels were significantly reduced by viable AKM Lab-01 treatment compared to HFD controls, while pasteurized AKM Lab-01 showed a milder effect (Figure 6F). Similar reductions were observed for TNF-α and IL-6 (Figures 6G, H). These results indicate that AKM Lab-01 reduces endotoxin translocation and attenuates systemic inflammation.

To further assess the microbiome-mediated effects of AKM Lab-01, we conducted metagenomic sequencing comparing the HFD group and AKM Lab-01-treated group. While alpha diversity analysis revealed no significant changes in species diversity following AKM Lab-01 treatment (Supplementary Figure S4A), PCoA demonstrated significant alterations in gut microbiota composition (Supplementary Figure S4B). Further analysis of shifts in gut microbiota composition between the HFD group and the AKM Lab-01 treatment group is presented in Supplementary Dataset S5 and Supplementary Figures S4C, D. At the genus level, the abundance of Anaerotruncus and Lawsonibacter—genera associated with metabolic improvement (Lv et al., 2019; Liu et al., 2025)—was significantly increased after AKM Lab-01 treatment (p < 0.05). In contrast, the abundance of Romboutsia, Allobaculum, and Eggerthella—genera linked to obesity progression (Therdtatha et al., 2021; Zheng et al., 2021; Samudio-Cruz et al., 2025)—was markedly reduced following AKM Lab-01 supplementation. At the species level, Lawsonibacter asaccharolyticus (belonging to the genus Lawsonibacter) was notably increased, while Romboutsia ilealis (belonging to the genus Romboutsia) was significantly decreased in the AKM Lab-01 group (p < 0.05). We next conducted gene set enrichment analysis (GSEA) based on the KEGG pathway database to assess the functional impact of gut microbiota following AKM Lab-01 treatment. Notably, GSEA revealed a significant down-regulation in most disease-related pathways after treatment with AKM Lab-01, including Pathways of neurodegeneration – multiple diseases, arkinson disease, Alcoholic liver disease, and Alzheimer's disease (p < 0.05, FDR < 0.2, Supplementary Dataset S6 and Supplementary Figure S4E). These pathways have previously been linked to obesity (Neto et al., 2023; Zeng et al., 2025; Chiang and McCullough, 2014). Collectively, these results suggest that the amelioration of obesity-related phenotypes by AKM Lab-01 is associated with a remodeling of the gut microbiota in HFD-induced obese mice.

To evaluate the safety of A. muciniphila AKM Lab-01 drug substance (freeze-dried pasteurized AKM Lab-01), we conducted a 90-day dietary exposure study in mice. AKM Lab-01 drug substance was administered at daily dosages of 8.8 × 10¹⁰, 8.8 × 10⁹, and 8.8 × 108 TFU per mouse (Figure 7A). Throughout the study, no mortality or clinical signs were observed at any dosage level. There were no significant differences in body weight (Figures 7B, C), daily weight gain (Figures 7D, E), final body weight (Figures 7F, G), food consumption (Figures 7H, I), or food efficiency ratio between the treated and control groups (Figures 7J, K). Clinical chemistry analysis revealed that AST levels decreased in the low-dose group but increased in the high-dose group of female mice (Supplementary Table S1). Hematology and coagulation results were generally normal. However, a few non-dose-related changes were observed: a decrease in lymphocyte percentage at the high dose and in MCHC at the low dose, and an increase in neutrophil number at the low dose in male mice (Supplementary Table S2); an increase in basophil number, MCV, and MCH at the low dose in female mice (Supplementary Table S3). Organ weights were comparable between treated and control groups, except for a slight decrease in the weight of the liver, kidney, prostate glands, and epididymis at the medium dose. These changes were non-dose-related and attributed to individual variability rather than consumption of AKM Lab-01 (Supplementary Table S4). Histopathological analysis confirmed that AKM Lab-01 did not cause any abnormalities in the architecture of visceral organs, including the lung, liver, kidney, spleen, and heart (Figure 7L).

Therefore, under the study conditions and based on the toxicological parameters evaluated, no adverse effects were observed that could be attributed to the ingestion of AKM Lab-01 drug substance. Mice were able to safely consume AKM Lab-01 drug substance at concentrations up to 8 × 10¹⁰ TFU/day, corresponding to a daily dietary intake of 3.2 × 10¹² TFU/kg body weight per day. Drawing on literature data (EFSA Panel on Nutrition et al., 2021) and the guidance of EFSA Scientific Committee (Committee, 2012), we applied an uncertainty factor of 200 [10 (interspecies variability) × 10 (intraspecies variability) × 2 (sub-chronic to chronic study duration)] to the NOAEL derived from the 90-day repeated-dose oral toxicity study in mice. Based on this analysis, we concluded that a daily intake of AKM Lab-01 drug substance at 1.12 × 10¹² TFU is safe for an average 70 kg human, provided that the viable cell counts in AKM Lab- 01 drug substance remains below 10 TFU/g.