Disclosing the Genomic Diversity among Members of the Bifidobacterium Genus of Canine and Feline Origin with Respect to Those from Human

ABSTRACT

In recent decades, much scientific attention has been paid to characterizing members of the genus Bifidobacterium due to their well-accepted ability to exert various beneficial effects upon their host. However, despite the well-accepted status of dogs and cats as principal companion animals of humans, the bifidobacterial communities that colonize their gut still represents a rather unexplored research area. To expand and further investigate the bifidobacterial ecosystem inhabiting the canine and feline intestine, strains belonging to this genus were isolated from fecal samples of dogs and cats and subjected to de novo sequencing. The obtained sequencing data, together with publicly available genomes of strains belonging to the same bifidobacterial species of our isolates, and of both human and animal origin, were employed for in-depth comparative genome analyses. These phylogenomic investigations highlighted a different degree of genetic variability between human- or pet-derived bifidobacteria depending on the considered species, with B. pseudocatenulatum strains of pet origin showing higher genetic variability than human-derived strains of the same bifidobacterial species. Furthermore, in silico evaluation of metabolic activities coupled with in vitro growth assays revealed the crucial role of diet in driving the genetic assembly of bifidobacteria as a result of their adaptation to the specific ecological niche they colonize.

IMPORTANCE Despite cats and dogs being well recognized as the most intimate companion animals to humans, current knowledge on canine and feline gut microbial consortia is still far from being fully dissected compared to the significant advances achieved for other microbial ecosystems, such as the human gut microbiota. In this context, a combination of in silico genome-based analysis and in vitro carbohydrate growth assay allowed us to further explore the canine and feline bifidobacterial community with respect to that inhabiting the human intestine. Specifically, these data revealed how strains of different bifidobacterial species seem to have evolved a different degree of host-specific adaptation. In detail, genotypic and phenotypic evidence of how diet can be considered the main factor of this host-specific adaptation is provided.

INTRODUCTION

Human and animal intestines harbor trillions of microbial cells collectively forming an extremely dense and complex community, known as the gut microbiota (1, 2). Millions of years of coevolution and coadaptation between microorganisms and their hosts have contributed to the establishment of interspecies relationships, resulting in multiple trophic interactions that impact host health (3). Among the multitude of microorganisms inhabiting the mammalian gastrointestinal tract, in recent years particular attention has been paid to the characterization of members of the genus Bifidobacterium (4).

Bifidobacteria are Gram-positive, anaerobic, nonsporulating, nonmotile, and saccharolytic microorganisms, encompassing 107 unique (sub)species and representing the deepest branching lineage within the phylum Actinobacteria (5–9). Although bifidobacteria have been isolated from various ecological niches, including human blood, oral cavity, sewage, raw and fermented milk, and hindgut of social insects and birds, the vast majority of the currently isolated and characterized bifidobacterial species originate from the gastrointestinal tract of humans and other mammals. At this site, it has been largely demonstrated that bifidobacteria play a crucial role in exerting multiple host health-promoting effects (4, 10). Specifically, bifidobacteria significantly contribute to host metabolism via saccharolytic fermentation of a wide variety of complex diet- and host-derived glycans, a feature that ensures their successful colonization of the mammalian intestine and provides accessible nutrients to the host as well as to other intestinal microbial taxa through cross-feeding strategies (2, 10). Furthermore, these commensal microorganisms promote intestinal barrier integrity, influence the development and activity of the host immune system, prevent pathogen proliferation, and produce a plethora of metabolites, such as vitamins, conjugated linoleic acids, polyphenols, and short-chain fatty acids, that elicit a beneficial impact on host enterocytes and gut bacteria (10–13). Due to their purported role in promoting host health, substantial scientific efforts have been made to characterize the composition of bifidobacterial communities colonizing the human intestine through both metagenomic and culture-dependent approaches (14–16). Current knowledge on the gut microbiota of dogs and cats, despite their special position as principal companion animals for humans, is still incomplete compared to the significant advances achieved for the human gut microbial ecosystem, especially concerning their bifidobacterial communities (17). Indeed, just two metagenomics-based studies have recently provided an accurate profiling of bifidobacterial species inhabiting the canine and feline intestine (17, 18). However, the latter did not take into account any culturomic approach or genomic comparative analysis to shed light on genetic and metabolic properties of bifidobacteria colonizing the canine and feline gut.

In the current study, a culturomics approach applied to canine and feline fecal samples was carried out to explore their intestinal bifidobacterial communities. This effort allowed the isolation of 12 novel strains belonging to Bifidobacterium longum, Bifidobacterium pseudocatenulatum, and Bifidobacterium pseudolongum (sub)species, whose genomes were analyzed by de novo sequencing, together with four B. pseudolongum strains isolated from human fecal samples in the context of a previous cultivation effort. The obtained sequencing data, together with genome sequences of all publicly available strains of the same species from human, canine, or feline origins, were subjected to in-depth comparative genome analysis. Combining this in silico data with in vitro carbohydrate growth assays, we revealed how strains of different bifidobacterial species apparently evolved host-specific adaptations.

RESULTS AND DISCUSSION

Isolation and genomic features of bifidobacterial strains from canine and feline origins.

Recently, an internal transcribed spacer (ITS) bifidobacterial profiling analysis, performed on fecal samples of different canine and feline breeds, revealed the ubiquitous presence of bifidobacterial species in the intestinal tract of these two companion animals (17, 18). However, data on the isolation and genomic characterization of bifidobacterial taxa inhabiting the gut of dogs and cats is rather limited, with only 15 genomes out of 1,641 currently publicly available bifidobacterial genomes belonging to strains isolated from canine or feline fecal samples (https://www.ncbi.nlm.nih.gov/genome/). To address this apparent paucity of information, a culture-dependent approach was applied to 20 fecal samples of dogs and cats, allowing the isolation of 32 strains belonging to B. animalis, B. longum, B. pseudocatenulatum, B. pseudolongum, and B. pullorum (sub)species (see Table S1 in the supplemental material). Notably, B. pullorum and B. animalis were excluded from downstream analyses since they are not typical colonizers of the human host, thereby preventing cross-host species comparisons (11, 19). Therefore, the genomes of 12 different bifidobacterial strains, belonging to B. longum, B. pseudocatenulatum, and B. pseudolongum (sub)species, were sequenced (Table 1). However, among the latter, only two strains were isolated from feline fecal samples, with most bifidobacterial isolates of feline origin belonging to the B. pullorum subsp. gallinarum taxon (Table S1). The difficulty in isolating other bifidobacterial species from feline feces may, as previously reported, be due not only to the high abundance of B. pullorum subsp. gallinarum typical of the feline gut microbiota but also to the dominant ecological behavior of this subspecies, which, when present, tends to exclude other bifidobacterial species (17). Furthermore, chromosomal DNA of four additional B. pseudolongum strains of human fecal origin, isolated in the framework of a previous cultivation effort (20), was subjected to sequencing to increase the limited number of publicly available human-derived strains belonging to this bifidobacterial species (Table 1).

TABLE 1.

General genome features of the 16 newly sequenced bifidobacterial strains

Newly sequenced genome	Species	Host	No. of bases	Avg coverage	No. of contigs	No. of ORFs	No. of rRNA loci	No. of tRNAs	GC%	Accession no.
93B	B. pseudolongum subsp. globosum	Homo sapiens	1,733,436	121	13	1618	3	53	63.46	JAIZHF000000000
243B	B. pseudolongum subsp. globosum	Homo sapiens	2,025,463	162	20	1623	3	52	63.09	JAIZHE000000000
495B	B. pseudolongum subsp. globosum	Homo sapiens	2,143,979	79	34	1808	4	58	63.03	JAIZHD000000000
598B	B. pseudolongum subsp. globosum	Homo sapiens	2,147,431	110	47	1809	3	58	63.04	JAIZHC000000000
2013B	B. pseudocatenulatum	Canis lupus familiaris	2,322,928	180	16	1954	3	58	56.48	JAIZHB000000000
2140B	B. pseudolongum subsp. globosum	Canis lupus familiaris	2,147,548	133	31	1814	3	58	63.13	JAIZHA000000000
2204B	B. longum subsp. suis	Felis silvestris catus	2,526,524	148	21	2115	3	56	60.36	JAIZGZ000000000
2211B	B. pseudolongum subsp. globosum	Canis lupus familiaris	2,109,557	200	20	1790	4	53	63.05	JAIZGY000000000
2226B	B. longum subsp. longum	Canis lupus familiaris	2,356,777	65	32	1944	1	55	60.26	JAIZGX000000000
2233B	B. longum subsp. suis	Canis lupus familiaris	2,413,224	286	14	2065	3	56	60.06	JAIZGW000000000
2234B	B. pseudocatenulatum	Canis lupus familiaris	2,255,332	101	11	1813	4	54	56.61	JAIZGV000000000
2235B	B. pseudocatenulatum	Canis lupus familiaris	2,175,837	116	9	1730	4	54	56.6	JAIZGU000000000
2241B	B. pseudocatenulatum	Felis silvestris catus	2,236,392	133	13	1819	5	54	56.86	JAIZGT000000000
2242B	B. pseudolongum subsp. globosum	Canis lupus familiaris	1,976,082	184	11	1623	4	52	63.41	JAIZGS000000000
2243B	B. pseudolongum subsp. globosum	Canis lupus familiaris	2,077,189	94	43	1737	4	57	63.1	JAIZGR000000000
2244B	B. pseudocatenulatum	Canis lupus familiaris	2,335,833	69	19	1904	4	54	56.86	JAIZGQ000000000

Accordingly, these 16 newly isolated bifidobacterial strains were decoded by means of a de novo genome sequencing approach. The obtained sequences displayed a coverage depth ranging from 65- to 286-fold, which, upon assembly, resulted in 9 to 47 contigs per genome, which were ordered and oriented based on each species’ type strain chromosome sequence (Table 1). Furthermore, the assembly allowed us to determine the genome length as well as the number of predicted open reading frames (ORFs) of each newly sequenced strain, resulting in a genome size that ranged from 1,733,436 bp of B. pseudolongum subsp. globosum 93B to 2,526,524 bp of B. longum subsp. suis 2204B and an ORF count ranging from 1,618 to 2,115 (Table 1).

Phylogenomic analyses of bifidobacterial strains of human, canine, or feline origin.

To assess the genetic relatedness between bifidobacterial strains isolated from dogs/cats and those retrieved from humans, the 16 newly isolated strains were subjected to comparative genome analysis by applying a previously described phylogenomic approach (21–23), including all publicly available genomes belonging to B. longum, B. pseudocatenulatum, and B. pseudolongum strains isolated from human, canine, or feline fecal samples. Furthermore, those genomes exhibiting an average nucleotide identity (ANI) of ≥99.5% were excluded from the comparative genomic analysis to reduce genetic redundancy, generating a final collection of 217 B. longum, 50 B. pseudocatenulatum, and 18 B. pseudolongum genomes (Table S2). The selected bifidobacterial genomes were further subjected to interspecies pangenome analyses, from which the core genomes, i.e., the set of genes shared by all considered strains per bifidobacterial species, were extrapolated. Subsequently, paralogs were excluded from each generated core genome through a pangenome analysis pipeline (PGAP) (see Materials and Methods), resulting in 539, 567, and 1,086 core genes shared by all assessed genomes belonging to the B. longum, B. pseudocatenulatum, and B. pseudolongum species, respectively. Finally, the concatenated amino acid sequences corresponding to the core genome sequences were aligned to reconstruct phylogenomic supertrees (Fig. 1 and Fig. S1).

FIG 1.

Phylogenomic supertrees of bifidobacterial taxa. Panels a, b, and c show the proteomic trees based on the concatenation of 539, 1,086, and 567 core genes identified in the pangenome analysis of the selected B. longum, B. pseduolongum, and B. pseudocatenulatum strains, respectively. Specifically, to obtain a clear phylogenetic graphical visualization, the B. longum-related supertree displays a selection of 58 representative genomes belonging to the B. longum species, selected to maximize the genetic diversity coverage based on ANI data. Trees were constructed by the use of the neighbor-joining method through the inclusion of genome sequences of Bifidobacterium breve LMG 13208, B. animalis subsp. lactis DSM 10140, and Bifidobacterium dentium LMG 11045 as selected outgroups for the B. longum, B. pseudolongum, and B. pseudocatenulatum trees, respectively. Bootstrap percentages above 50 are indicated at node points, based on 1,000 replicates. Circles surrounding the trees represent the ecological origin of each bifidobacterial strain: purple, humans; blue, dogs; orange, cats. The cluster containing B. pseudocatenulatum strains isolated from dogs and cats is highlighted with a lilac box in the relative supertree.

Despite the small number of B. longum taxa retrieved from dogs and cats, the B. longum-associated supertree highlighted that B. longum strains isolated from canine and feline fecal samples did not fit into a separate cluster occupied by either B. longum subsp. longum or B. longum subsp. suis strains but rather are distributed in various subclusters together with human-associated bifidobacterial taxa (Fig. 1). This finding indicates that none of the strains considered in the present study possesses highly divergent genetic signatures when considering their core genome sequences, suggestive of low genetic diversity among B. longum strains isolated from dogs, cats, or humans. Conversely, in-depth scrutiny of the B. pseudolongum subsp. globosum cluster revealed the presence of two subdivisions. Indeed, while one subcluster (Bpg1) only contained human-derived strains with the exception of two canine-associated strains, the second subcluster (Bpg2) encompassed only strains isolated from canine fecal samples (Fig. 1). Furthermore, the B. pseudocatenulatum-based phylogenomic supertree displayed a clear separation of bifidobacterial strains isolated from canine/feline fecal samples with respect to those retrieved from human stool. Indeed, the five novel B. pseudocatenulatum isolates form a separate cluster within the generated supertree together with a single human-derived strain (Fig. 1). These findings indicate a higher level of genetic variability among B. pseudolongum subsp. globosum or B. pseudocatenulatum strains inhabiting the human or canine/feline intestine compared to that observed for B. longum strains, suggesting that the core genome genetic sequences of strains belonging to B. pseudolongum and B. pseudocatenulatum undergo significant alterations as a sign of adaptation to the specific ecological niche they colonize.

Prediction of the enzymatic arsenal of bifidobacterial strains isolated from humans or cats/dogs.

To evaluate possible genetic differences related to the metabolism of bifidobacterial strains isolated from canine and feline fecal samples compared to those of human origin, the overall enzymatic repertoire of each strain was investigated by means of the METAnnotatorX2 pipeline (24). Although no variations common to the three considered bifidobacterial (sub)species were recorded, statistically significant differences were observed in the abundance of several enzyme subclasses between bifidobacterial strains of canine and feline origins compared to those isolated from humans (P < 0.05 by t test) (Table 2 and Table S3). This suggests that members of different bifidobacterial subspecies have evolved distinct metabolic mechanisms of adaptation to a specific ecological environment. In detail, only the genes coding for 18 enzymes appeared to be differentially abundant in B. pseudolongum subsp. globosum strains of canine and feline origin compared to those collected from human stool, while 29 and 110 enzyme-coding genes significantly differed in B. longum subsp. longum and B. pseudocatenulatum strains, respectively (Table S4). Specifically, B. pseudolongum subsp. globosum strains derived from canine and feline fecal samples were provided with a statistically significant higher number (P < 0.05 by t test compared to the human-related strains) of genes predicted to encode a mannosyl-glycoprotein endo-beta-N-acetylglucosaminidase (EC 3.2.1.96; corresponding to GH18, GH73, and GH85) (Table S4), an enzyme involved in endohydrolysis of high-mannose glycopeptides or glycoproteins. Similarly, although not statistically significant, pet-derived B. pseudocatenulatum strains were shown to possess an average number of genes predicted to encode α-mannosidase (EC 3.2.1.24, including GH31, GH38, and GH92) that was, on average, 54-fold higher than that obtained for human isolates. On the other side, B. longum subsp. longum taxa of canine and feline origin showed an 8-fold higher abundance of genes encoding putative mannosylglycoprotein endo-β-mannosidases (EC 3.2.1.152) (GH2) than strains retrieved from human stool (Table S4). These data support specific commitment of bifidobacterial strains of companion animal origin toward the breakdown of mannose subunits in complex carbohydrates. In this context, modern commercial food formulations for dogs and cats are typically enriched with prebiotics, including yeast extracts that are an important source of mannan-oligosaccharides and N-acetylglucosamine, which may have contributed to the selection of strains with a dedicated genetic repertoire toward degradation of mannose-containing carbohydrates (25–27).

TABLE 2.

Statistically significant EC subclasses between bifidobacterial strains isolated from feces of dogs/cats and humans

Species and EC no.	EC subclass	Avg EC for:		P value
		Cat/dog	Human
B. longum subsp. longum
1.1	Oxidoreductase, acting on the CH-OH group of donors	32.400	34.828	0.01062
1.2	Oxidoreductase, acting on the aldehyde or oxo group of donors	6.800	7.113	0.02199
3.1	Hydrolases, acting on ester bonds	45.600	44.005	0.04379
4.4	Lyases, carbon-sulfur lyases	3.400	4.335	0.00251
6.5	Ligases, forming phosphoric ester bonds	1.000	1.145	1.4E−08
7.3	Translocases, catalyzing the translocation of inorganic anions and their chelates	22.600	18.842	0.00206
B. pseudocatenulatum
1.14	Oxidoreductase, acting on paired donors, with incorporation or reduction of molecular oxygen	1.000	0.844	0.00668
1.17	Oxidoreductase, acting on CH or CH(2) groups	7.000	6.533	0.00871
1.3	Oxidoreductase, acting on the CH-CH group of donors	12.000	10.311	2.3E−14
2.6	Transferring, transferring nitrogenous groups	15.200	12.022	1.2E−05
2.8	Transferring, transferring sulfur-containing groups	4.000	3.578	0.00111
3.5	Hydrolases, acting on carbon-nitrogen bonds, other than peptide bonds	18.200	16.511	0.00859
3.6	Hydrolases, acting on acid anhydrides	21.200	19.133	0.01602
4.3	Lyases, carbon-nitrogen lyases	10.000	9.622	0.00964
4.4	Lyases, carbon-sulfur lyases	3.400	4.335	0.00251
5.3	Isomerases, intramolecular oxidoreductases	7.600	6.911	0.04652
6.1	Ligases, forming carbon-oxygen bonds	21.000	19.356	0.00854
6.3	Ligases, forming carbon-nitrogen bonds	34.000	32.622	0.04062
7.2	Translocases, catalyzing the translocation of inorganic cations	7.000	4.044	2.7E−10
7.4	Translocases, catalyzing the translocation amino acids and peptides	19.800	16.356	0.01214
B. pseudolongum subsp. globosum
1.16	Oxidoreductase, oxidizing metal ions	0.182	0.667	0.04855
2.1	Transferring, one-carbon groups	30.182	27.833	0.0025
4.1	Lyases, carbon-carbon lyases	9.909	9.000	0.01858
7.1	Translocases, catalyzing the translocation of hydrons	9.818	9.000	5.3E−05
7.2	Translocases, catalyzing the translocation of inorganic cations	4.455	5.500	0.01397
7.4	Translocases, catalyzing the translocation amino acids and peptides	4.364	5.500	0.00305

Furthermore, genomes of B. pseudocatenulatum isolates of canine/feline origin were shown to contain a significantly (P < 0.05 by t test) enhanced number of genes predicted to catalyze the metabolism of amino acids and peptides (EC 7.4), including transferases, lyases, and ligases (Table S4). In this context, the high-protein, carnivorous diet of companion animals may have influenced the genetic evolution of B. pseudocatenulatum strains inhabiting their intestine to maximize energy recovery from diet.

Conversely, human-derived B. longum subsp. longum and B. pseudocatenulatum strains were predicted to possess a higher abundance of coding sequences involved in metabolic pathways dedicated to the degradation and isomerization of complex plant-derived glycans (Table S4). This suggests that the human diet, typically enriched in vegetable-associated complex carbohydrates, modulated the genetic assembly of certain bifidobacterial strains toward a larger number of genes involved in metabolic pathways for glycan degradation.

Insights into the glycobiome of the canine/feline bifidobacterial strains.

Bifidobacteria are commensal gut inhabitants known for their ability to degrade a wide range of complex carbohydrates, including both host- and diet-derived glycans (28–31). In this context, to further investigate the enzyme repertoire of each considered bifidobacterial strain associated with polysaccharide degradation activities, the complete carbohydrate-active enzyme arsenal was assessed. Specifically, the total number of genes predicted to encode glycosyl hydrolases (GH) associated with B. longum strains retrieved from cats/dogs and those isolated from humans did not significantly differ from each other (P < 0.05 by t test) (Fig. 2). Conversely, B. pseudocatenulatum and B. pseudolongum strains originating from companion animals were shown to possess a statistically significantly higher GH total number compared to that of human-derived taxa (Fig. 2). This indicates that while the analyzed B. longum strains did not seem to have undergone any host-associated adaptive events in their carbohydrate metabolism-related genetic arsenal in terms of GH number, B. pseudocatenulatum and B. pseudolongum strains isolated from canine and feline fecal samples appear to have undergone a numerical expansion of GH-encoding genes, perhaps as a result of the particular diet of dogs and cats. Indeed, due to their carnivorous classification, the canine and feline diet is characterized by a reduced intake of carbohydrates compared to the average human diet (25). Therefore, it can be hypothesized that, to better compete and ensure their survival in an environment limited in the preferred bifidobacterial energy source, i.e., fermentable carbohydrates (10, 28), B. pseudocatenulatum and B. pseudolongum strains from the canine and feline gut may have acquired a larger number of GH to expand the diversity of carbohydrates they can metabolize.

FIG 2.

Predicted glycobiome of the selected bifidobacterial strains. Panel a shows a whisker plot reporting the average GH index calculated for bifidobacterial strains isolated from fecal samples of dogs and cats and those of human origin for each Bifidobacterium species. Panel b displays the average for the predicted glycobiome divided into canine/feline and human group for each considered bifidobacterial species.

Further scrutiny of each GH family revealed that B. pseudolongum genomes of animal origin encode a statistically significantly higher number of GH1 and GH130 (P < 0.05 by t test), which are two GH families encoding putative β-mannosidase and mannose-based polysaccharide phosphorylase (Fig. 2 and Table S5). Similarly, although not statistically significant (P = 0.083 by t test), GH130 abundance was shown to be 2-fold higher in pet-derived B. longum taxa than in those isolated from human fecal samples. Furthermore, although no variations were observed in the average abundance of GH130 in B. pseudocatenulatum strains of companion animal origin, the latter showed a higher number of GH38 as well as a significant increase of GH1 and GH2 (P < 0.05 by t test). Specifically, while GH38 corresponds to a carbohydrate-active enzyme family predicted to encode a putative metabolic activity similar to that of GH130, i.e., α-mannosidase, able to catalyze the hydrolysis of terminal α-d-mannose in mannose-containing polysaccharides, the other two GH families are both involved in β-mannosidase activities (Fig. 2 and Table S5). Similarly, GH85 members, which represent endo-β-N-acetylglucosaminidase activities involved in endohydrolysis of high-mannose glycopeptides and glycoproteins, were exclusively encoded by certain B. pseudocatenulatum strains derived from canine and feline fecal samples (Fig. 2 and Table S5). These observations reinforce the notion that a diet enriched in mannose-based complex polysaccharides may have caused the selection for specific bifidobacterial strains to suit the ecological niches representative of the canine and feline gut.

Conversely, B. longum strains of human origin displayed an increased number of genes predicted to encode GH27, GH31, GH55, and GH127 members (P < 0.05 by t test). In contrast, human-derived B. pseudolongum strains showed a higher abundance of GH4 (P = 0.043 by t test), i.e., GH families involved in enzymatic activities aimed at releasing glucose from complex plant-derived polysaccharides. Similarly, the genomes of B. pseudocatenulatum taxa retrieved from the human gut not only display a significantly higher number of GH51 members than those derived from dogs and cats but also possess genetic sequences encoding GH10 and GH30 that do not appear in the genomes of strains isolated from the two above-mentioned companion animals. Interestingly, the latter three GH families all perform enzymatic activities related to the degradation of xylans (Fig. 2 and Table S5). In this context, regardless of the species, human-derived bifidobacterial taxa seem to have evolved a glycobiome enriched in genes encoding enzymes involved in the degradation of complex plant-derived glycans, representing another genetic sign of adaptation to the high plant-derived carbohydrate content typical of the human diet (28). Altogether, these observations corroborate the generally accepted crucial role of diet in shaping the genetic assembly of the mammalian gut microbiota as a result of their adaptation to the particular ecological niche they colonize.

Growth profiles on a mannose-containing polysaccharide.

To validate the above-described in silico prediction, growth assays were performed involving the 16 newly sequenced bifidobacterial strains together with all those B. longum subsp. longum, B. longum subsp. suis, B. pseudocatenulatum, and B. pseudolongum subsp. globosum strains of pet and human origin isolated from our laboratory and employing mannan as the sole carbon source. As suggested by the in silico metabolic activity analyses, strains of canine/feline origins displayed statistically higher growth performances when cultivated for 48 h on mannan if compared to the human-derived strains (P < 0.01 by t test) (Fig. 3). Indeed, except for B. longum subsp. longum 108B and B. pseudolongum subsp. pseudolongum 243B, both of which seemed to be able to degrade mannan and exploit the released carbohydrates as carbon source, the other human bifidobacterial strains were unable to grow on this polysaccharide (Fig. 3). Conversely, canine- and feline-isolated strains displayed an appreciable ability to metabolize the tested mannose-containing polysaccharide (Fig. 3). Indeed, except for five bifidobacterial strains showing growth performances comparable to those obtained for strains isolated from human fecal samples, the other 15 strains of companion animal origins presented significantly higher optical density at 600 nm (OD₆₀₀) values than human-derived strains unable to grow on mannan (P < 0.05 by analysis of variance [ANOVA]) (Table S6).

FIG 3.

Evaluation of growth performances of feline/canine and human bifidobacterial strains in synthetic medium using mannan as the unique carbon source. Whisker plots report the average OD₆₀₀ measurement calculated for bifidobacterial strains of human or pet origin (left) and OD₆₀₀ measurement for each selected bifidobacterial strain (right). Below the latter, a heat map represents the growth performances of selected bifidobacterial strains on mannan. Cells were grown in biologically independent triplicates. In the whisker plots, the x axis reports bifidobacterial strains, while the y axis shows the optical density values obtained for each bifidobacterial strain. The boxes are determined by the 25th and 75th percentiles. The whiskers are determined by the maximum and minimum values and correspond to the box extreme values. The lines inside the boxes represent the average, while squares correspond to the medians. Boxes in pink and blue refer to bifidobacterial strains isolated from fecal samples of humans and pets, respectively.

Furthermore, in-depth analyses of growth performances of each considered bifidobacterial strain revealed that, among those strains isolated from canine and feline feces, B. pseudocatenulatum strains elicited, on average, the most appreciable growth performance on mannan. Thus, the in vitro growth profiles obtained by cultivating bifidobacterial strains on mannan substantiate the obtained in silico data, according to which the genomes of pet-derived B. pseudocatenulatum strains encompass a genetic arsenal enriched in genes involved in the degradation of mannose-containing polysaccharides.

Conclusions.

Despite their role as principal companion animals for humans, investigations of the taxonomic composition and metabolic arsenal of the gut microbiota of dogs and cats are still in their infancy. Specifically, in spite of the universally accepted ability of bifidobacteria to colonize the intestine of a wide range of mammals (including dogs and cats) (4, 17, 32, 33) and their well-known ability to exert multiple beneficial effects on the host (11, 13, 34), bifidobacterial communities colonizing the canine and feline gut are still far from being fully characterized. In this context, to further explore the bifidobacterial communities colonizing the intestine of these companion animals and to evaluate possible genomic differences between bifidobacterial strains isolated from cats/dogs and those retrieved from humans, a comparative genome analysis was performed, involving both newly isolated and publicly available genomes belonging to B. longum, B. pseudocatenulatum, and B. pseudolongum (sub)species. The phylogenomic analyses revealed a different intraspecies degree of genetic variability between strains isolated from humans or dogs/cats. Furthermore, in silico prediction of the glycobiome and metabolic pathways of the selected bifidobacterial strains highlighted the crucial role of diet in shaping the genetic assembly of bifidobacterial strains, in accordance with the specific ecological niche they colonize.

However, since fecal samples used in this study for the isolation of bifidobacterial species were collected from dogs and cats that live in close contact or even cohabit with their owners, the isolated strains may have been acquired through horizontal transmission events from cohabiting family members, as previously reported for members of different bacterial genera (35), a hypothesis that can be confirmed by the reduced abundance, if not total absence, of bifidobacteria in the gut microbiota of the feral counterparts of dogs and cats, i.e., the gray wolf and wild cat, respectively (18, 25, 33). However, to validate this hypothesis, the application of a Bifidobacterium-specific isolation protocol to fecal samples of gray wolves and wild cats as well as to animal companion owners is required.

MATERIALS AND METHODS

Ethical statement.

This study was carried out in compliance with the rules, regulations, and recommendations of the ethical committee of the University of Parma. Protocols were approved by the Comitato di Etica Università degli Studi di Parma, Italy. All animal procedures were performed in strict compliance with national guidelines (Decreto Legislativo 26/2014).

Bifidobacterial genome sequences.

Publicly available genomes (complete and draft genome sequences) belonging to strains of Bifidobacterium longum, Bifidobacterium pseudocatenulatum, and Bifidobacterium pseudolongum and isolated from human, canine, or feline feces were retrieved from the National Center for Biotechnology Information (NCBI) public database and subsequently subjected to genome quality assessment. In detail, only genome sequences with a genome coverage higher than 30-fold and containing less than 100 contigs were considered, resulting in a total of 388, 92, and 12 high-quality genomes of B. longum, B. pseudocatenulatum, and B. pseudolongum, respectively. Furthermore, the genome sequences of 16 newly isolated strains belonging to the abovementioned bifidobacterial species were also included in this study (Table 1).

Isolation and characterization of bifidobacterial strains and growth conditions.

Fecal samples were collected from dogs and cats that live in close contact or even cohabit with their owners. In detail, by using a dedicated sterile tube provided with a sampling spoon, fecal samples were collected immediately after defecation, shipped to the laboratory under anaerobic conditions, and immediately processed. Specifically, 1 g of a given fecal sample was mixed with 9 mL of phosphate-buffered saline (PBS; pH 6.5), serially diluted, and plated using de man-Rogosa-Sharpe (MRS) agar (Scharlau Chemie, Barcelona, Spain) supplemented with 0.05% (wt/vol) l-cysteine hydrochloride and 50 μg/mL mupirocin (Delchimica, Italy). The agar plates were subsequently incubated in an anaerobic chamber (2.99% H₂, 17.01% CO₂, and 80% N₂) (Concept 400; Ruskinn) at 37°C for 48 h. Colonies grown on MRS agar plates were randomly picked and restreaked to isolate purified bacterial strains. Subsequently, one colony per plate was inoculated into 10 mL of MRS broth supplemented with 0.05% (wt/vol) l-cysteine hydrochloride and incubated overnight at 37°C in an anaerobic atmosphere. Cells were then harvested by centrifugation at 6,000 rpm for 8 min. The obtained cell pellets were used for DNA extraction using the GenElute bacterial genomic DNA kit (Sigma-Aldrich, Germany) by following the manufacturer’s instructions. Subsequently, the internal transcribed spacer (ITS) sequences were amplified from the extracted DNA using primer pair Bif23S_ITS (5′-AGATGTTTCACTTCCCTGCG-3′) and Bif16S_ITS (5′-CCTTGTACACACCGCCCG-3′), as previously described (36, 37). Amplification was carried out according to the following protocol: one cycle of 95°C for 5 min, followed by 30 cycles of 95°C for 30s, 60°C for 1 min, 72°C for 1 min, and a final cycle of 72°C for 5 min. PCR amplicons were purified using the QIAquick PCR purification kit (Qiagen, United Kingdom) and subsequently sequenced. Nucleotide sequencing of the ITS region was performed by the Eurofins Mix2Seq kit service (Eurofins Genomics, Germany) using ITS_bif-SEQ1 (5′-CGTCAAGTCATGAAAGTGGG-3′). Finally, the obtained ITS sequences were compared to a public database composed of bifidobacterial ITS sequences using the basic local alignment search tool (BLAST). This procedure allowed for discarding suspected clonal or nearly identical strains within samples. Table 1 lists the isolated strains involved in this study.

Bifidobacterial genome sequencing and assemblies.

The chromosomal DNA of the newly isolated bifidobacterial strains was decoded by GenProbio srl (http://genprobio.com) with a MiSeq platform (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol by using the Nextera XT DNA Library Prep kit (Illumina), as previously described (20, 38). The library samples obtained were then pooled into a Flow Cell V3 600 cycle (Illumina). Fastq files of paired-end reads generated from each genome sequencing event were employed as input for the genome assembly by using the MEGAnnotator pipeline (39). The SPAdes program v3.14.0 was used for the de novo assembly of each bifidobacterial genome sequence with the pipeline option “–careful” and a list of k-mer sizes of 21, 33, 55, 77, 99, and 127 (40). Subsequently, MEGAnnotator employed contigs longer than 1,000 bp to predict protein-coding open reading frame (ORFs) using Prodigal (41). Predicted ORFs were then functionally annotated using RAPSearch2 (reduced alphabet-based protein similarity search) (cutoff E value of 1 × 10⁻⁵ and minimum alignment length of 20), employing the NCBI reference sequence (RefSeq) database (42) together with hidden Markov model profile (HMM) searches (http://hmmer.org/) performed against the manually curated Pfam-A database (cutoff E value of 1 × 10⁻¹⁰). tRNA genes then were detected through tRNAscan-SE v1.4 (43), while rRNA genes were identified by means of RNAmmer v1.2 (44).

Comparative genomic analysis.

The genome content of the 16 reconstructed bifidobacterial strains (Table 1) was subjected to a pangenome calculation by means of the pangenome analysis pipeline (PGAP) (45). The predicted proteome of a specific B. longum, B. pseudocatenulatum, or B. pseudolongum strain was screened for orthologues against the proteome derived from other available genomes of the same bifidobacterial species employing BLAST analysis (46) (cutoff E value of <1 × 10⁻⁵ and 50% identity over at least 80% of both protein sequences). The data obtained were then clustered into protein families, also named clusters of orthologous groups (COGs), employing MCL (graph theory-based Markov clustering algorithm) by means of the gene family (GF) method (47). Using the PGAP software, a pangenome profile for each of the three bifidobacterial species was built based on a presence/absence matrix including all identified COGs of the analyzed strains. The core genome of the B. longum, B. pseudocatenulatum, and B. pseudolongum species was then defined based on protein families shared among each analyzed genome.

Phylogenetic and phylogenomic analysis.

The concatenated sequences of amino acids belonging to the core genome of each strain for a single considered bifidobacterial species were aligned by means of the MAFFT software (48), and the resulting phylogenetic trees were built using the neighbor-joining method through the ClustalW v2.1 program (49). Subsequently, the graphical viewer of phylogenetic trees, i.e., FigTree v1.4 (http://tree.bio.ed.ac.uk/software/figtree/), was used to build a phylogenetic tree for each of the considered bifidobacterial species. A value of average nucleotide identity (ANI) was calculated through the program fastANI, using each genome pair as the input (50).

Genomic analyses.

Prediction of carbohydrate-active enzyme-encoding genes for each of the selected bifidobacterial taxa was assessed based on sequence similarity to genes collected in the Carbohydrate-Active enZyme (CAZy) database (51). For this purpose, GH data of 17,538 bacterial genomes available in the CAZy database were used to retrieve functional annotation using RAPSearch2 (cutoff E value, 1 × 10⁻³⁰). In addition, the prediction of metabolic pathways was performed by employing the METAnnotatorX2 pipeline (24).

Bifidobacterial carbohydrate growth assays.

Bifidobacterial strains were cultivated on semisynthetic MRS medium supplemented with 0.05% (wt/vol) l-cysteine hydrochloride at 37°C under anaerobic conditions. Subsequently, cells were washed with PBS, resuspended in MRS without glucose, and inoculated in MRS without glucose supplemented with 0.5% (wt/vol) mannan (Carbosynth, UK) in a 96-well microtiter plate to reach a final OD₆₀₀ of ∼0.1 and incubated in an anaerobic cabinet. Cell growth was evaluated by monitoring the optical density at 600 nm with the use of a plate reader (Biotek, USA). Plates were read in discontinuous mode, with absorbance readings performed at 3-min intervals three times after 48 h of growth, and each reading was ahead of 30 s of shaking at medium speed. Cultures were grown in triplicates, and the resulting growth data sets were expressed as the average of these replicates.

Statistical analyses.

Student's t test was performed with SPSS software (IBM, Italy).

Data availability.

The genome sequences of 16 newly isolated bifidobacterial strains were deposited at DDBJ/ENA/GenBank under the accession numbers reported in Table 1 (BioProject no. PRJNA768761).