Integrated analyses of metagenomics, metabolomics and culture-based assays reveal functional roles of gut microbiota in Felidae

Abstract

The functional roles of gut microbiota in carnivores remain poorly understood. Here, we integrated metagenomics, metabolomics, proteomics and culture-based functional assays to characterize metabolic potential of gut microbiota across 14 captive Felidae species. Comparative metagenomics analysis revealed that the Felidae gut microbiome is distinct from that of non-Felidae and reflects carnivorous dietary patterns. Genus-level core microbiota were dominated by Clostridium, Collinsella and Bacteroides, with functional enrichment in carbohydrate and amino acid metabolism. Of 219 reconstructed metagenome-assembled genomes (MAGs), 27 were identified as core MAGs containing proteases- and lipases- encoding genes, with ATP-dependent Clp proteases predominating and enriched KEGG orthologs mainly associated with amino acid metabolism. Fecal metabolomics identified 1316 metabolites shared among Felidae species, with KEGG analysis showing they were involved in amino acid and lipid metabolism and significantly enriched in protein digestion and absorption pathway. The amino acid- and lipid-related metabolites were correlated with the relative abundance of core MAGs. Culture-based assays revealed proteolytic and lipolytic activities across isolates, supported by proteomics evidence of predominant ATP-dependent proteases. In vitro fermentation with representative isolates generated fatty-acid-dominated metabolites consistent with fecal metabolomic profiles. Together, our findings demonstrate that Felidae gut microbiota play a critical role in amino acid metabolism for carnivory.

Introduction

The mammalian gut represents one of the largest reservoirs of microbes on Earth¹, containing trillions of microorganisms with cell counts of 10¹² cells/ml². The composition of these complex gut microbiota is largely shaped by three important factors: host phylogeny, diet, and environment³. Evolutionary theory posits that gut microbial community composition in mammals is primarily affected by host phylogeny, reflecting long-term host-microbe co-evolution^4–6. In contrast, ecological theory emphasizes the role of diet and environmental factors in shaping the gut microbiota, highlighting the rapid adaptation and plasticity of the microbial community in response to these external conditions⁷. The different dietary strategies, such as herbivory and carnivory, drives the formation of specialized microbial communities adapted to these nutritional niches⁵. Lifestyle (captive versus wild) can alter microbial richness and diversity, although these effects often vary across different mammalian families⁸.

Previous studies in mammals showed that host evolutionary history and diet exert distinct influences on gut microbiome diversity, with each factor shaping different aspects of community structure⁹; in some cases, dietary effects can outweigh those of host phylogeny or environment^5,6. Compared to herbivores and omnivores, carnivores exhibit lower gut microbiota diversity, with functions significantly enriched in amino acid degradation^5,6. Besides, different carnivore lineages exhibit additional gut microbiome functions that are closely associated with long-term dietary patterns^10,11. Felidae, as obligate carnivores with persistently high-protein and high-fat diets, represent an extreme dietary model; however, most existing studies on Felidae gut microbiota have focused on the community structure affected by external factors such as captivity, diet, or medical interventions^12–17, rather than on the microbial functions associated with obligate carnivory. This represents a critical gap in our understanding of carnivore gut microbiomes.

In recent decades or so, the study of gut microbiota using omics approaches has developed rapidly¹⁷. Metagenomics, involving the shotgun sequencing of extracted microbial DNA, has been pivotal in this field. For instance, a metagenomic study showed that the enterotypes of Felidae are characterized by a predominance of Bacteroides and Clostridium¹⁶; in Ursidae, the dominant gut microbial phyla, Firmicutes and Proteobacteria, display seasonal shifts and are involved in supplying energy to the host through the metabolic pathway of short-chain fatty acids¹⁸. Advances in understanding gut microbiota extend beyond mere description, with more and more experiments validating predicted functions related to specific microbes^19,20. To bridge the gap between in silico predictions and functional confirmation, culturomics has emerged as a powerful complementary approach. By expanding the culturable fraction of gut microbes, culturomics enables the correction of microbial community composition and the acquisition of bacterial strains²¹. Concurrently, enhanced bioinformatic analysis pipelines have been developed to construct metagenome-assembled genomes (MAGs), which are crucial for species-level functional analysis^22–24. MAG reconstruction combined with microbial culture could provide a chance to characterize the microbial functions hypothesized from metagenomics at species level. Although comparative microbiome studies are becoming more common, integrated studies that combine species-level microbial resolution with in vitro functional validation have remained relatively scarce.

Given the diverse dietary strategies across Carnivora, Felidae species were selected as a model for studying obligate carnivory, exhibiting a strict meat-based diet in contrast to the more varied diets of other families (e.g., Ursidae, Musteloidea). Although host-microbe associations across mammals have been widely studied, most investigations have focused on community-level patterns. In contrast, relatively few studies have resolved the species-level functional capacities of gut microbes, particularly through integrative, species-level functional analyses in Felidae. Due to the elusive and rare nature of wild Felidae, obtaining fresh fecal samples has been challenging; therefore, we focused on captive individuals maintained in Chinese zoos. We generated a comprehensive gut metagenomic dataset from 14 captive Felidae species and compared these data with publicly available gut metagenomes from mammals spanning diverse taxonomic orders, dietary categories, and lifestyles to identify the core microbes driven by a carnivorous diet. Then, we characterized gut microbial community structure and reconstructed high-quality MAGs to resolve species-level functional potentials of Felidae core microbial taxa. To couple microbial composition with functional metabolism, we integrated metabolomics, proteomics, and culture-based enzyme activity assay. This multi-omics framework enabled deep functional resolution, particularly through species-level MAG analyses and experimental validation of metagenomic predictions, moving the study beyond descriptive community profiling and providing a comprehensive view of diet-associated functional specialization of the gut microbiota in Felidae.

Results

Comparative metagenomics identified the core gut microbiota in Felidae

We first analyzed 70 gut metagenomes from captive Felidae animals. Sequencing statistics, including raw reads and raw base (bp) for each sample, are provided in Suppl. Table 1. Microbial profiling identified Firmicutes (56.22 ± 19.48%), Actinobacteria (16.48 ± 13.14%), Bacteroidetes (13.14 ± 14.00%) and Proteobacteria (6.41 ± 9.84%) as dominant phyla (Fig. 1A). Comparative analysis across mammalian metagenomes showed that these phyla were significantly more abundant in Felidae than in non-Felidae (Analysis of Molecular Variance, AMOVA, p < 0.001, Fig. 1B). A similar pattern was observed between carnivores and herbivores (p < 0.001, Fig. 1C) but not omnivores (p = 0.06, Fig. 1C). Specifically, significant differences were observed in the relative abundances of Proteobacteria, Actinobacteria, and Firmicutes between Felidae and non-Felidae (Wilcoxon rank-sum test, Suppl. Fig. 1A); whereas for dietary groups, a significant difference was found only for Proteobacteria (Kruskal-Wallis H test, Suppl. Fig. 1B). At the genus level, captive Felidae gut microbiota were dominated by Clostridium (18.28 ± 18.88%)), Collinsella (15.55 ± 15.07%), and Bacteroides (7.06 ± 8.36%) (Fig. 1A), all significantly enriched compared with non-Felidae (AMOVA, p < 0.001, Fig. 1D). Subsequent Wilcoxon rank-sum test on individual genera revealed that 10 of them were the primary contributors to this significant distinction (Suppl. Fig. 1C). Carnivores also showed higher relative abundance of these genera than herbivores and omnivores (AMOVA, p < 0.001, Fig. 1E). The Kruskal-Wallis H test revealed that seven genera were the primary contributors to differences in gut microbiota composition across dietary habits (Suppl. Fig. 1D). Random forest analysis identified Proteobacteria, Firmicutes, Bacteroidetes, and Actinobacteria, as well as Blautia, Peptacetobacter, Fusobacterium, Clostridium, and Mediterraneibacter, as key discriminative taxa (Suppl. Fig. 1E, F). The genus-level model achieved a cross-validated accuracy of 80%.

Fig. 1. Patterns of gut microbial communities in 45 mammalian species.

(A) showed the relative abundance of the dominant phyla and genera across mammalian species with different diets and habitats. Species highlighted in purple represent Felidae. Analysis of Molecular Variance (AMOVA) analysis at the phylum level revealed differences in microbial community composition between Felidae and non-Felidae (B), and among different dietary groups (C). AMOVA analysis at the genus level revealed differences in microbial community composition between Felidae and non-Felidae (D), and among different dietary groups (E). The y-axes of panels B-E represent relative abundance.

Thirty genera were defined as the Felidae core microbiota (Felidae core genera) based on prevalence ( > 90%) across samples, all of which had a mean relative abundance greater than 0.5% (Suppl. Table 2). Among them, Clostridium, Collinsella, Bacteroides, Unclassified_f_Lachnospiraceae (6.19 ± 3.92%), and Blautia (5.86 ± 3.29%) were the most abundant (Suppl. Table 2). Stability analysis using the coefficient of variation (CV) indicated relatively consistent abundance across samples (average: 1.32, median: 1.15). Functional annotation of Felidae core genera showed enrichment in carbohydrate and amino acid metabolism (Kyoto Encyclopedia of Genes and Genomes, KEGG) (Suppl. Fig. 1G). Within carbohydrate metabolism, amino sugar and nucleotide sugar metabolism, glycolysis/gluconeogenesis, and starch and sucrose metabolism were dominant pathways (Suppl. Fig. 1H). Since amino sugars and nucleotide sugars are primarily synthesized by the host, this pathway metabolizes host-derived carbohydrates rather than dietary carbohydrates^25,26. Glycolysis and gluconeogenesis represent opposing pathways of glucose metabolism²⁷. In Felidae, which consume high-protein, high-fat diets with minimal exogenous carbohydrates, glucose homeostasis is likely largely maintained through gluconeogenesis, with amino acids derived from dietary protein serving as major gluconeogenic substrates²⁸. Taken together, these results suggest that microbial amino acid metabolism constitutes the primary mode of nutritional metabolism in the gut of captive Felidae.

Metabolic and enzymatic potential of core microbes at species level

To investigate species-level members of gut microbiome, metagenomic reads from captive Felidae and publicly available datasets were assembled into contigs. A total of 219 non-redundant near-complete MAGs (completeness ≥ 90%, contamination ≤ 5%) were recovered based on dereplication using dRep and quality assessment with CheckM (Suppl. Table 3). The relative abundance of these MAGs in each sample was calculated using CoverM. AMOVA analysis revealed that the relative abundances of these MAGs were significantly higher in Felidae compared to non-Felidae (R² = 0.215, p < 0.001). Similarly, carnivores showed significantly higher abundance levels than herbivores (R² = 0.231, p < 0.001) and omnivores (R² = 0.222, p = 0.001). A one-way permutational multivariate analysis of variance (PERMANOVA) revealed that MAG composition differed significantly among dietary groups (F = 3.89, R² = 0.217, p = 0.001). Wilcoxon rank-sum test identified 45 MAGs that differed significantly (corrected p < 0.05) between Felidae and non-Felidae (Fig. 2A). Among these, 44 MAGs were significantly enriched in Felidae. Notably, 27 of these 44 MAGs belonged to Felidae core genera (Felidae core MAG). Among them, Clostridium (6 MAGs) represented the most abundant genus, followed by Blautia (5 MAGs) and Collinsella (3 MAGs). The Felidae core MAG encoded 80 proteases, with ATP-dependent Clp protease (24.67%) and ATP-dependent Lon protease (15.08%) being the most abundant (Fig. 2B). Additionally, 37 lipases were identified, among which lysophospholipase (13.54%) and phospholipase (12.1%) were the most prevalent (Fig. 2C).

Fig. 2. Felidae-specific MAGs and their functional enrichment.

Wilcoxon rank-sum test identified 44 MAGs enriched in Felidae (A). Protease (B) and lipase (C) composition encoded by the Felidae core MAGs. KEGG functional annotation of Felidae-enriched KEGG Orthologs (KOs) (D).

KEGG functional annotation was performed on all 219 non-redundant MAGs to generate a MAG×KEGG Ortholog (KO) matrix. The relative abundance of each KO in different host species was calculated by combining MAG-level KO abundances with the relative abundance of each MAG across species. Wilcoxon rank-sum test comparing Felidae and non-Felidae identified 13,931 KOs significantly enriched in Felidae (fold change ≥ 2, corrected p < 0.01, Suppl. Table 3). These Felidae-enriched KOs were mapped to KEGG pathways, revealing that most were associated with Metabolism (62.39%), with amino acid metabolism being the most prominent (Fig. 2D). Although a large number of KOs were identified as significantly enriched, these KOs were mainly concentrated in metabolism-related pathways rather than being randomly distributed. Collectively, these results demonstrate that Felidae gut microbiota exhibit functional specialization for amino acid metabolism, characterized by enrichment of ATP-dependent Clp protease.

Fecal metabolomics reveals abundant amino acids and fatty acids

To test this hypothesis, fecal metabolites were first profiled, capturing metabolites derived from both host and microbial metabolic processes. Fifty-six samples from captive Felidae species were analyzed using liquid chromatography tandem mass spectrometry (LC-MS/MS), revealing 2874 and 1843 metabolites in positive and negative ion modes, respectively (Suppl. Table 4). The identified metabolites were mainly involved in amino acid metabolism, biosynthesis of secondary metabolites, and lipid metabolism (Fig. 3A). Venn analysis demonstrated overlap in metabolite profiles across Felidae species, with 867 metabolites shared in the positive ion mode and 449 metabolites shared in the negative ion mode (Fig. 3B, C). These shared metabolites were dominated by amino acids and fatty acids, including 4-hydroxy-2-nonenal-glutathione conjugate, pantothenic acid, and oleoyl leucine (Suppl. Fig. 2). Notably, shared metabolites were significantly enriched in the protein digestion and absorption pathways (Fig. 3D, E). Five processes related to protein digestion and absorption were identified in the feline gut, including stomach acid-assisted breakdown, small intestinal metabolism, transport to blood, degradation into branched-chain fatty acids absorbed by colonic epithelial cells, and microbial fermentation in the colon leading to amines, phenols, and indoles (Supplementary Figs. 3 and 4). Further analysis focused on metabolites involved in amino acid and lipid metabolism, retaining 85 metabolites (Suppl. Table 4). Spearman correlation analysis between the abundances of these metabolites and the relative abundances of Felidae core MAGs across species showed that 24 of the 27 Felidae core MAGs were significantly associated with one or more of these metabolites (Fig. 4).

Fig. 3. Fecal metabolomic features and functional enrichment across Felidae species.

A Functional classification of identified metabolites showing predominant involvement in amino acid metabolism, biosynthesis of secondary metabolites, and lipid metabolism. Venn diagram shows the shared metabolites on positive (B) and negative ion modes (C). KEGG enrichment analysis of shared metabolites (D: positive, E: negative) highlighting significant enrichment in protein digestion and absorption pathways.

Fig. 4. Spearman correlation analysis showing associations between metabolite abundance and the relative abundance of Felidae core MAGs.

Significant correlations are highlighted in red.

Microbial screening reveals Bacillus, Enterococcus, and Escherichia as key protease and lipase producers

Next, live gut bacteria were isolated and cultured to assess their protease and lipase production. Except for fecal samples from Caracal caracal and Pardofelis temminckii used for metagenomics and metabolomics, 14 fecal samples from the remaining Felidae species were analyzed. Samples were serially diluted, plated on 12 different types of media, and incubated to obtain distinct colonies, which were then identified by 16S rRNA sequencing following DNA extraction and PCR amplification. A total of 494 isolates were obtained, among which 244 non-repetitive strains were identified (Supplementary Table 5). Screening for enzyme activity identified 166 protease-producing strains exhibiting clear hydrolytic zones (Suppl. Fig. 5). The ratio of the diameter of the clear zone to that of the colony was calculated (Suppl. Table 5), with six strains showing ratios > 6.0, including Bacillus sp. (belonging to Firmicutes) strain Bh10, Bacillus velezensis strain 411, Enterococcus faecalis strain UIS0927, Bacillus mycoides GBPI_Hb312, Enterococcus faecalis strain IMAU98469, and Enterococcus sp. CSQRZN1.1.4. Additionally, 33 isolates were identified as lipase-producing strains based on precipitation zones (Suppl. Fig. 6). Quantitative assays confirmed that selected strains exhibited both proteolytic and adipolytic activities (Suppl. Table 5). 79.52% of protease-producing strains and 63.64% of lipase-producing strains were affiliated with Bacillus, Enterococcus, and Escherichia, consistent with Felidae core genera identified by metagenomics. Six strains with the highest proteolytic activity (NN-2-3: Enterococcus hirae, NN-8-3: Bacillus altitudinis, OM-5-7: Bacillus subtilis, PB-11-1: Bacillus subtilis, PL-8-4: Enterococcus faecalis, FB-4-4: Bacterium strain) and six strains with the highest adipolytic activity (PP-12-2: Escherichia coli, PP-5-1: Escherichia coli, LL-5-2: Shigella flexneri, PTI-12-2: Citrobacter freundii, PP-11-2: Kurthia gibsonii, PP-7-2: Enterococcus faecalis) were selected for further analyses (Fig. 5 and Suppl. Table 5). With the exception of Bacterium strain, the genera of all other strains were Felidae core genera. Furthermore, Enterococcus faecalis and Escherichia coli were included in the Felidae core MAGs (MAG105, MAG148, Suppl. Table 3). Therefore, except for the Bacterium strain, all other strains were subjected to proteomic analysis. Proteomics identified nine major protease families, dominated by the ATP-dependent Clp protease (30.14%), and two major lipase families, with phospholipases being predominant (Suppl. Table 5). These findings were consistent with the metagenomic results.

Fig. 5. Identification of bacterial strains with proteolytic and adipolytic activities.

A Screening results of strains showing the highest proteolytic activity. B Screening results of strains showing the highest adipolytic activity.

Fermentation products of protease-producing strains inoculated with protein

The enrichment of protein digestion and absorption pathway together with high protease activity suggests that microbe-assisted protein degradation represents a key digestive mechanism in Felidae. To further investigate the protein-derived products generated by these enzyme-producing microbial species, the three highest-activity protease-producing strains (PL-8-4: Enterococcus faecalis, OM-5-7: Bacillus subtilis, NN-2-3: Enterococcus hirae) with significant proteolytic activity were selected for bacterial metabolomic analysis. All three belonged to Felidae core genera. Fermentation products were analyzed by LC-MS/MS. Each strain was individually cultured by liquid fermentation in medium containing casein as a substrate (protein group), with ISP2 liquid medium (ISP2 group) serving as a control. The presence of the protein substrate resulted in metabolic profiles distinct from those observed in the control medium (Suppl. Fig. 7). Metabolomic profiling identified 10 types of compounds (Suppl. Table 6). Lipids and lipid-like molecules, particularly fatty acids and derivatives, were the most abundant metabolites in the protein group, accounting for 46.3% ± 4.17%, compared to 27.25% ± 1.17% in the ISP2 group. Amines, indoles, and phenols constituted 1.49% ± 0.69%, 3.31% ± 2.77%, and 4.43% ± 1.77% of total metabolites in the protein group, respectively, similar to the ISP2 group (amines: 1.66% ± 0.82%, indoles: 1.73% ± 0.77%, phenols: 7.57% ± 2.74%). These metabolic patterns were consistent with the fecal metabolomics.

Discussion

Our study characterizes the gut microbiota composition of captive Felidae and highlights microbial features associated with a carnivorous dietary pattern. The gut microbiota of Felidae was dominated by Clostridium, Collinsella, Bacteroides at the genus level, which contributed substantially to distinguishing Felidae from other mammals. Random forest analysis further identified several genera, including Blautia, Peptacetobacter, Fusobacterium, Clostridium, and Mediterraneibacter, as potential microbial biomarkers associated with carnivorous hosts. Thirty Felidae core genera were identified, among which Clostridium, Collinsella, Bacteroides, Unclassified_f_Lachnospiraceae, and Blautia being the most abundant. Clostridium, a member of Firmicutes, could be accumulated by a high-fat and high-protein diet in humans²⁹. A high-fat diet is associated with a decreased abundance of Collinsella and high proportions of Bacteroides in humans^29,30. Blautia is inversely correlated with obesity in humans, whereby the process is linked to unique amino acid and carbohydrate metabolisms, resulting in simultaneous modification of the gut bacterial composition³¹. Notably, Collinsella was detected at relatively high abundance in Felidae in the present study, suggesting that diet-microbiota associations described in humans may not be fully applicable to Felidae. Although these patterns may provide useful context, these observations are derived primarily from human studies and should be interpreted cautiously when applied to Felidae. Functional annotation also confirmed that the Felidae core genera mainly participated in carbohydrate metabolism and amino acid metabolism. The similar metabolic roles have been described in humans and mice. Amino acid metabolism was annotated as plausible because dietary proteins are hydrolyzed into amino acids by hydrolase during the digestive process that can be utilized by certain gut microbes to participate in metabolism and physiological function of host (e.g., humans)³². However, carbohydrates are not present in the diet of Felidae, but carbohydrate metabolism was a major enriched microbial function. We found that amino sugar and nucleotide sugar metabolism, glycolysis/gluconeogenesis, and starch and sucrose metabolism were the major enriched pathways within carbohydrate metabolism. Given the low carbohydrate content of the Felidae diet, these pathways likely do not represent direct utilization of dietary carbohydrates. Instead, they are most likely associated with microbial metabolism of host-derived glycans (e.g., intestinal mucus) and carbon metabolism contributing to host energy and glucose balance^25,26,33. In addition, possible causes included functional annotation bias and the presence of multifunctional enzymes involved in multiple metabolic processes. The community level analysis suggests that the gut microbiota enriched during fermentation may play an important role in amino acid metabolism in Felidae.

Although several studies have reconstructed the MAGs of some mammals^22–24, little is known about MAGs in Felidae. One preliminary study annotated the MAGs of domestic cats (Felis catus)³⁴. In this study, metagenomic reads from captive Felidae together with publicly available datasets from non-Felidae mammals were jointly assembled, resulting in 219 non-redundant MAGs. These MAGs showed significantly higher relative abundances in Felidae than in non-Felidae, and in carnivores than in herbivores and omnivores, with PERMANOVA revealing clear compositional separation among dietary groups. Of the 44 MAGs enriched in Felidae, 27 belonged to the Felidae core genera were considered Felidae core MAG, including Clostridium, Blautia, and Collinsella. Our results were supported by previous studies. Blautia and Clostridium are considered beneficial genera and are commonly detected in the feces and intestines of mammals^35,36. As expected, Felidae core MAGs encoded multiple proteases and lipases, with ATP-dependent Clp protease representing the most abundant protease. At the pathway level, KEGG annotation showed that KOs enriched in Felidae were predominantly associated with Metabolism, with amino acid metabolism emerging as the most prominent category. Our MAG data reveal species-level functional specializations in Felidae gut microbiomes that are consistent with community-level functional profiles.

To validate the metagenomic predictions, we profiled fecal metabolites from captive Felidae, revealing a substantial presence of metabolites, although we cannot discriminate gut microbe-derived metabolites from host-derived metabolites. KEGG annotation indicated that metabolites shared across Felidae species were primarily associated with amino acid metabolism, lipid metabolism, and biosynthesis of secondary metabolites, consistent with the predicted functional capacities of the gut microbiota. Notably, KEGG enrichment analysis revealed that these shared metabolites were significantly enriched in protein digestion and absorption pathways, including degradation into branched-chain fatty acids (absorbed colonically) and colonic fermentation yielding amines, phenols, and indoles. Previous studies have confirmed that once endogenous and exogenous nitrogenous compounds move through the ileocaecal junction into the colon after the protein digestion in the small intestine, these compounds are not absorbed by the large intestinal mucosa but undergo intense proteolysis, leading to the release of numerous bacterial metabolites, such as ammonia, amines, phenols, fatty acids, etc^37,38. These metabolites influence various signaling pathways in epithelial cells, regulate the mucosal immune system in the host, and modulate the gene expression of bacteria, which results in the synthesis of enzymes associated with amino acid metabolism³⁷. Additionally, a high dietary protein intake could result in potentially harmful metabolites in the gut from proteolysis³⁸. Sometimes, the metabolism of protein by gut bacteria is potentially detrimental due to the production of toxic metabolites such as ammonia, amines, p-cresol, and indole^32,38. Although toxic compounds of amines, phenols, and indoles were found in our results, their quantities were much smaller than those of fatty acids. To further reveal the relationship between microbial functional potential and metabolite profiles, we performed co-variation patterns between metabolites and Felidae core MAGs. Of the 27 Felidae core MAGs, 24 showed significant associations with one or more of the metabolites related to amino acid and lipid metabolism. These patterns suggest that the Felidae gut microbiome is closely associated with metabolic features characteristic of protein- and fat-rich diets, although contributions from host digestion and dietary inputs cannot be excluded.

To further evaluate microbial functional potential, we isolated and cultured live gut bacteria from Felidae fecal samples and assessed their capacity for protease and lipase production. Most isolates belonged to Felidae core genera, including Bacillus, Enterococcus, and Escherichia. A large proportion of isolates exhibited proteolytic and lipolytic activity, and extracellular enzyme assays confirmed active protein and lipid degradation by these strains. Proteomic analysis further revealed that ATP-dependent Clp protease was the dominant protease, consistent with metagenomic predictions. The enrichment of protein digestion and absorption pathways, together with widespread protease production among cultured isolates, indicates that microbial protein degradation represents a major functional process in the Felidae gut. To further characterize metabolic outputs associated with microbial proteolysis, representative high-activity protease-producing strains were subjected to metabolomic analysis under protein-rich conditions. These strains produced a diverse set of metabolites, with fatty acids and their derivatives being the most abundant. Such metabolic patterns may be explained by previous finding showing that amino acids can contribute intermediates to lipid metabolism³⁹. In contrast, potentially harmful metabolites such as amines, indoles, and phenols were detected at relatively low levels. This metabolic profile was consistent with fecal metabolomics. Collectively, culture-based functional assays, proteomics, and metabolite profiling support the role of Felidae gut microbiota in protein and lipid processing. These findings provide experimental evidence consistent with metagenomics and metabolomics analyses. Together, these results establish a functional linkage from core microbial taxa to key enzymes, metabolic pathways, and downstream metabolites, providing a systematic framework connecting microbioal composition with functional and metabolic outputs in the captive Felidae gut.

However, this study has several limitations. Our findings are based on captive animals, and standardized diets, limited environmental exposure, and possible antibiotic history may influence the observed microbial patterns⁸; thus, conclusions are framed under the context of captivity. Causal links between diet and microbial function cannot be directly established and require validation through controlled feeding experiments with metagenomics. In addition, cultured isolates may not reflect in vivo abundance and are affected by cultivation conditions. So the culture-based results are interpreted mainly for functional validation. The microbial culture experiment was conducted under aerobic conditions, as protease- and lipase-producing bacteria usually thrive in aerobic environments^40,41. However, some anaerobic bacteria in the gut might have been overlooked; in the future, an anaerobic culture would be required to obtain comprehensive understanding of gut microbiota. Finally, the in vitro fermentation system represents a simplified model and does not fully capture the complexity of the feline gut ecosystem.

Methods

Feline sample and metagenomic data collection

Sample collection was conducted between April and June, 2023. Fecal samples were obtained from captive Felidae individuals housed in five zoos in China: Beijing Wildlife Park, Beijing Zoo, the Breeding Center of Beijing Zoo, Chongqing Zoo, and Xining Wildlife Park. Fresh fecal samples were collected immediately after defecation to ensure minimal environmental contamination. Animals were housed individually in zoo enclosures, and fecal samples were collected from the corresponding enclosures, ensuring that each sample could be assigned to a single individual at the time of collection. One fecal sample was collected per individual. All sampled individuals were adults, clinically healthy, and had not received antibiotic treatment prior to sampling. 73 fecal samples were collected from 14 Felidae species, including two subspecies and two coat color variants (Suppl. Table 7). These included Acinonyx jubatus (AJ), Caracal caracal (CC), Felis bieti (FB), Leptailurus serval (LS), Lynx lynx (LL), Neofelis nebulosa (NN), Otocolobus manul (OM), Panthera leo (white) (PLW), Panthera leo (PL), Panthera onca (PO), Panthera pardus (black) (PPB), Panthera pardus (PP), Panthera tigris altaica (PAL), Panthera tigris amoyensis (PAM), Panthera tigris tigris (PTI), Panthera uncia (PU), Pardofelis temminckii (PT), and Prionailurus bengalensis (PB). Animals were fed meat-based diets consisting primarily of chicken, mutton, beef, and pork, following standard zoo husbandry practices, one to three times per day. The number of individuals per species, along with sex and dietary information, are provided in Suppl. Table 7. All fecal samples were immediately frozen in dry ice after collection and transported back to the laboratory, where they were stored at -80 °C until further processing.

To enable comparative analyses, we downloaded publicly available gut metagenomic datasets (raw sequencing data) comprising 321 samples from 31 mammalian species representing diverse dietary strategies and lifestyle. The number of samples per species ranged from 2 to 48, with all species having more than two samples (Suppl. Table 7). All raw reads will be reprocessed using the same bioinformatic pipeline.

Metagenomic sequencing and MAGs construction

Except for three captive Felidae samples that failed DNA extraction (highlighted in Suppl. Table 7), the remaining 70 samples were subjected to metagenomic sequencing and analysis. Total DNA was extracted using the EZNA Soil DNA Kit (Omega Bio-Tek, USA) according to the protocol. DNA concentration, purity, and integrity were assessed using a TBS-380, NanoDrop 2000, and 1% agarose gels. DNA libraries were constructed using the Nexflex Rapid DNA-Seq (Bioo Scientific, USA) and sequenced on the Illumina Novaseq platform. Raw reads were processed using Trimmomatic⁴² and FastQC⁴³ for quality control. Host-derived reads were removed using Bowtie2⁴⁴ by mapping against reference genomes (Suppl. Table 8), and high-quality reads were assembled using MEGAHIT⁴⁵. Taxonomic classification was performed using Kraken2⁴⁶ with the standard database, and relative abundance estimation was refined using Bracken⁴⁷. Gene prediction from metagenomic contigs was performed using Prokka to obtain predicted protein sequences⁴⁸. Functional annotation was conducted using eggNOG-mapper⁴⁹, and KOs assigned by eggNOG were used for downstream pathway and functional analyses.

Host reference genomes and corresponding host removal rates are provided in Suppl. Table 8. To minimize potential batch effects, ComBat was applied to the normalized abundance matrix prior to downstream analyses⁵⁰. Sequencing platform was used as the batch variable, and samples were grouped into four batches accordingly (Suppl. Table 9). To avoid removing biological signals of interest, correction was performed separately using biologically relevant grouping factors (Felidae and non-Felidae and carnivore, herbivore, and omnivore). Beta diversity differences were assessed using AMOVA based on Bray-Curtis distances. Group comparisons were conducted using the Wilcoxon rank-sum test and Kruskal–Wallis test. Dunn’s post hoc test was applied for pairwise comparisons. Multiple testing correction was performed using the false discovery rate (FDR) with a significance threshold of adjusted p < 0.05. Random forest analysis was performed using ten-fold cross-validation with 500 decision trees. Model performance was evaluated based on cross-validated accuracy, and feature importance was assessed using mean decrease in accuracy. Optimal model parameters were selected using default settings of the implemented package (Majorbio pipeline). Thresholds for defining core microbiota vary across studies⁵¹; in this study, a strict criterion of > 90% prevalence across samples was applied⁵². To evaluate stability, the CV (standard deviation/mean) was calculated for each core genus.

MAGs were reconstructed using MetaBAT2⁵³. Genome quality (completeness and contamination) was assessed using CheckM⁵⁴. Redundant bins were removed using the dRep at a 95% average nucleotide identity (ANI) threshold to obtain species-level representative genomes⁵⁵. Taxonomic classification of MAGs was performed using GTDB-Tk (v2.1.0). Protein-coding genes were predicted using Prokka. KEGG functional annotation of MAGs was performed using KOfam based on profile hidden Markov models (HMMs)⁵⁶. Relative abundances of non-redundant MAGs were calculated by mapping host-filtered reads to dereplicated genomes using CoverM⁵⁷, and averaged across all samples of each host species. A one-way PERMANOVA was performed based on Bray-Curtis dissimilarities to investigate differences in MAG composition between host group. Felidae-enriched KOs were identified using Wilcoxon rank-sum test, retaining only KOs with a fold change ≥ 2 and corrected p < 0.01. The host’s phylogenetic tree was constructed using Timetree (https://timetree.org) and iTOL (https://itol.embl.de/). The complete bioinformatic workflow scripts are publicly available at: https://github.com/dinihhh/Mammal_Metagenome.

Fecal metabolomic analysis

Samples from captive Felidae species were selected for metabolomic profiling as follows: for species with fewer than five individuals, all available samples were included; for species with more than five individuals, five samples were randomly selected to ensure balanced representation. In total, 56 samples were analyzed. For metabolite extraction, 50 mg of feces was suspended in 400 µL methanol:water (4:1, v/v) containing 0.02 mg/mL L-2-chlorophenylalanine as an internal standard. Samples were homogenized using a Wonbio-96c frozen tissue grinder (Shanghai Wanbo Biotechnology Co., Ltd.) at −10 °C and 50 Hz for 6 min, followed by ultrasonic extraction at 5 °C and 40 kHz for 30 min. After incubation at −20 °C for 30 min and centrifugation 13,000 × g for 15 min at 4 °C, the supernatant was transferred to LC-MS/MS vials for analysis. A pooled quality control (QC) sample was prepared by mixing equal volumes of all samples.

Metabolomic analysis was performed using a SCIEX UPLC-Triple TOF 5600 system equipped with an ACQUITY HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm; Waters, USA) at Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). The mobile phases consisted of (A) 0.1% formic acid in water:acetonitrile (95:5, v/v) and (B) 0.1% formic acid in acetonitrile:isopropanol:water (47.5:47.5, v/v). Mass spectrometry was performed in both positive and negative electrospray ionization modes. In positive ion mode, solvent B was programmed as follows: 0–3 min, 0–20%; 3–4.5 min, 20–35%; 4.5–5 min, 35–100%; 5–6.3 min, held at 100%; 6.3–6.4 min, 100–0%; 6.4–8 min, held at 0%. In negative ion mode, solvent B was programmed as follows: 0–1.5 min, 0-5%; 1.5-2 min, 5–10%; 2–4.5 min, 10–30%; 4.5–5 min, 30–100%; 5–6.3 min, held at 100%; 6.3–6.4 min, 100–0%; 6.4–8 min, held at 0%. The flow rate was set at 0.4 mL/min and the column temperature at 40 °C.

MS/MS data were acquired using a quadrupole-time-of-flight mass spectrometer (Triple TOFTM5600 + , Sciex, USA). Instrument parameters were set as follows: source temperature 550 °C; curtain gas 30 psi; ion source gases 1 and 2 at 50 psi; ion spray voltage −4000 V (negative mode) and + 5000 V (positive mode); declustering potential 80 V; collision energy 20–60 eV. Data were acquired in information-dependent acquisition (IDA) mode over a mass range of m/z 50–1000.

Raw data were processed using Progenesis QI (Waters Corporation, Milford, USA). Metabolite identification was performed by matching against the KEGG (http://www.genome.jp/kegg/), HMDB (http://www.hmdb.ca/), and Majorbio databases. Metabolite abundances were normalized using total-sum normalization. All metabolomics data were generated in a single analytical batch. QC samples were used to assess analytical stability, and metabolites with a relative standard deviation (RSD) ≤ 30% in QC samples were retained for downstream analysis. Prior to statistical analysis, data were log10-transformed to improve normality and reduce heteroscedasticity. Metabolic enrichment and pathway analyses were performed based on the KEGG database, and statistical enrichment analysis was carried out using the Python package scipy.stats. Multiple testing correction was performed using the Benjamini–Hochberg (BH) method, and pathways with an adjusted p-value < 0.05 were regarded as significantly enriched. Spearman correlation analysis was used to calculate correlation coefficients (R) and corresponding p-values between MAGs and metabolites, and hierarchical clustering was independently performed for both variables using the average linkage method. All analyses were performed on the Majorbio Cloud Platform (cloud.majorbio.com).

Bacterial isolation and identification

A total of 14 samples from each Felidae species were selected for bacterial isolation, except for the feces of Caracal caracal and Pardofelis temminckii, which had been fully consumed in prior metagenomic and metabolomic analyses. For each sample, 1 g feces was suspended in 10 mL of sterile water and homogenized thoroughly. The suspension was serially diluted to 10⁻² and 10⁻⁴. Aliquots (100 μL) of each dilution were spread onto 12 different mediums (Suppl. Table 10). Plates were incubated at 28 °C for 10 days. After incubation, morphological distinct colonies were selected and streaked onto ISP2 media for purification. The procedures for DNA extraction and PCR amplification followed those in Hu et al. ⁵⁸. Genomic DNA was extracted using a bacterial DNA extraction kit (Thermo Fisher Scientific, USA). PCR amplification of the 16S rRNA gene was performed using primers of 27 F/1492R⁵⁹. The PCR cycling conditions were as follows: initial denaturation at 95 °C for 5 min; followed by 35 cycles of denaturation at 94 °C for 1 min, annealing at 55 °C for 1 min, and extension at 72 °C for 2 min; with a final extension at 72 °C for 10 min. PCR products were sequenced using the Sanger method. The obtained compared against the NCBI database using BLASTn to identify the closest phylogenetic relatives⁶⁰ (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Protease- and lipase-producing strain screening and bacterium proteomic analysis

All identified colonies were inoculated onto two screening mediums containing casein⁶¹ or tween-80⁶² as the sole substrate for detecting protease and lipase activities, respectively. Plates containing casein were incubated overnight at 37 °C, whereas plates containing tween 80 were incubated for 2 days at 28 °C. Colonies exhibiting clear hydrolysis zones on casein-containing media or precipitation zones on tween 80-containing media were considered potential protease- or lipase-producing strains, respectively. For proteases-producing strains, the ratio of colony to clear zone was also measured. Enzymatic activities of the screened strains were further quantified using commercial protease and lipase activity assay kits (MlBio, China) according to the manufacturer’s instructions. Bacterial strains exhibiting high enzymatic activity were selected for proteomic analysis (n = 11). Based on enzyme activity assays, the five strains with the highest protease activity (NN-2-3: Enterococcus hirae strain Sha 28, NN-8-3: Bacillus altitudinis strain SE14, OM-5-7: Bacillus subtilis strain SA 91, PB-11-1: Bacillus subtilis strain Z33-33, PL-8-4: Enterococcus faecalis strain IMAU98469) and the six strains with the highest lipase activity (PP-12-2: Escherichia coli strain DL1501, PP-5-1: Escherichia coli strain YS, LL-5-2: Shigella flexneri strain 2439, PTI-12-2: Citrobacter freundii strain CitB, PP-11-2: Kurthia gibsonii strain KH2, PP-7-2: Enterococcus faecalis strain TFL1) were selected.

Selected strains were cultured in ISP2 liquid medium. Cells were harvested by centrifugation and subjected to trypsin digestion. Peptide mixtures were analyzed using an EASY nLC-1200 system (Thermo, USA) coupled with a timsTOF Pro2 mass spectrometer (Bruker, Germany) at Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). Peptides were separated on a C18-reversed phase column (75 μm × 25 cm, Ionopticks, USA) using solvent A (A: 2% acetonitrile with 0.1% formic acid) and solvent B (B: 80% acetonitrile with 0.1% formic acid) at a flow rate of 250 nL/min. The gradient was as follows: 0–45 min, 3–28% B; 45–50 min, 28–44% B; 50–55 min, 44–90% B; 55–60 min, 90–90% B. MS/MS spectra were searched against the corresponding reference protein database (Bacillus, Citrobacter, Enterococcus, Escherichia, Kurthia, and Shigella). Label-free quantification was applied for relative protein abundance analysis.

Bacterium metabolomic analysis

The three highest protease-producing strains with significant proteolytic activity (PL-8-4: Enterococcus faecalis strain IMAU98469, OM-5-7: Bacillus subtilis strain SA 91, NN-2-3: Enterococcus hirae strain Sha 28) were selected for bacterium metabolomic analysis. Fermentation broths from each strain cultured in casein-containing medium or ISP2 medium were processed individually. An aliquot (100 μL) of the culture supernatant was mixed with 400 µL of methanol: water (4:1, v/v) solution. The mixture was vortexed thoroughly and centrifuged at 15,000 × g and 4 °C for 15 min. The supernatant was collected for LC-MS/MS analysis. QC samples were prepared by mixing equal volumes of all samples. LC–MS/MS analysis was performed using a Vanquish UHPLC system coupled to a Q Exactive™ HF mass spectrometer (Thermo Fisher, Germany). Chromatographic separation was achieved on a Hypesil Gold column (100 × 2.1 mm, 1.9 μm, Thermo Fisher, USA). The column temperature was maintained at 40 °C, and the flow rate was set to 0.2 mL/min. For positive ion mode, the mobile phase consisted of solvent A (0.1% formic acid in water) and solvent B (methanol). For negative ion mode, solvent A consisted of 5 mM ammonium acetate (pH 9.0) in water and solvent B was methanol. The gradient program was as follows: 0-1.5 min, 98% A; 1.5–3 min, 98%–15% A; 3–10 min, 15%–0% A; 10–10.1 min, 0–98% A; 10.1–12 min, 98% A. Mass spectrometry was performed in electrospray ionization (ESI) mode with a scan range of m/z 100–1500. The spray voltage was set at 3.5 kV. The sheath gas flow rate was 35 psi, and the auxiliary gas flow rate was 10 L/min. The capillary temperature was set at 320 °C and the auxiliary gas heater temperature at 350 °C. Both positive and negative ion modes were applied. MS/MS acquisition was performed in data-dependent acquisition (DDA) mode. Raw data were processed using Compound Discoverer 3.1 (Thermo Fisher Scientific) for peak detection, alignment, and quantification. Retention time tolerance was set to 0.2 min and mass tolerance to 5 ppm. Metabolite identification was performed by matching MS/MS spectra against HMDB and KEGG databases. Blank samples were used to remove background ions. Peak areas were normalized using the formula: original peak area / (total metabolite peak area in sample/total metabolite peak area in QC sample), followed by relative quantification. Metabolite features with a CV > 30% in QC samples were removed prior to downstream analysis. All samples were analyzed in a single analytical batch.

Ethics approval

All experimental protocols were approved by the Animal Care and Use Committee of the Institute of Zoology, the Chinese Academy of Sciences (No. IOZ-IACUC-2022). All fecal samples were collected non-invasively.

Author contributions

Dini Hu, Yixin Liu, Tianchun Pu, Shunfu He, Yanqiang Yin collected the fecal samples. Dini Hu, Chenghang Sun, and Tong Xin finished the microbial experiments. Dini Hu and Yibo Hu wrote the manuscript and prepared the figures. All the authors approved the manuscript.

Data availability

The raw sequence data reported in this paper have been deposited into the Genome Sequence Archive in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences under accession number of CRA014688. The generated sequences of isolates were deposited in the NCBI database with the accession number of PRJNA1069337.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41522-026-01066-9.