Genomic Encyclopedia of Type Strains of the Genus Bifidobacterium

Abstract

Bifidobacteria represent one of the dominant microbial groups that are present in the gut of various animals, being particularly prevalent during the suckling stage of life of humans and other mammals. However, the overall genome structure of this group of microorganisms remains largely unexplored. Here, we sequenced the genomes of 42 representative (sub)species across the Bifidobacterium genus and used this information to explore the overall genetic picture of this bacterial group. Furthermore, the genomic data described here were used to reconstruct the evolutionary development of the Bifidobacterium genus. This reconstruction suggests that its evolution was substantially influenced by genetic adaptations to obtain access to glycans, thereby representing a common and potent evolutionary force in shaping bifidobacterial genomes.

INTRODUCTION

Bifidobacteria represent one of the dominant microbial groups that occur in the gut of various animals, including warm-blooded mammals and social insects (1, 2). In these environments, bifidobacteria reach a particularly high relative abundance as part of the infant gut microbiota (3–5), and this early life prevalence supports their purported role as modulators of various metabolic and immune activities of their immature host (1). Various members of the genus Bifidobacterium have attracted substantial scientific and commercial interest due to various professed beneficial health effects that they exert on their human host (6–10). Currently, the genus Bifidobacterium includes 47 taxa, involving 38 species and 9 subspecies (2, 11–14). Genomics has been crucial in revealing the evolutionary development as well as the biology of any taxonomical group of bacteria and thus in understanding the genetic forces that sustain specific adaptations to an ecological niche (15). However, representatives of only 10 of the 47 currently recognized bifidobacterial sub/species have been genomically decoded (1). Here, we describe the genome analysis of representatives of all 47 (sub)species that are currently assigned to the Bifidobacterium genus. Based on the generated genome information, we hypothesize that the bifidobacterial genome coevolved with its animal host via gene loss and, in particular, genetic acquisition events, of which the latter events appear to be responsible for species-specific adaptations to a glycan-rich environment.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

All Bifidobacterium strains were cultivated in an anaerobic atmosphere (2.99% H₂, 17.01% CO₂, and 80% N₂) in a chamber (Concept 400, Ruskin) on De Man-Rogosa-Sharp (MRS) broth (Scharlau Chemie, Barcelona, Spain) supplemented with 0.05% (wt/vol) l-cysteine hydrochloride and were incubated at 37°C. Bacterial cultures were subjected to DNA extraction using a previously described protocol (4).

Genome sequencing and bioinformatics analyses.

The genome sequences of all studied Bifidobacterium species were determined by GenProbio srl (Parma, Italy) using an Ion Torrent PGM platform (Life Technologies, Carlsbad, CA). A genomic library was generated using 1 μg of genomic DNA and an Ion Xpress Plus fragment library kit and employing the Ion Shear chemistry according to the user guide. After a dilution to 2.66 × 10⁷ molecules/μl, 4.5 × 10⁸ molecules were used as the template for clonal amplification on Ion Sphere particles during the emulsion PCR according to an Ion Xpress Template 400 kit manual. The quality of the amplification was estimated, and the amplification product was loaded onto an Ion 316 chip and was subsequently sequenced using 125 sequencing cycles according to an Ion Sequencing 400 kit user guide. A total of 125 sequencing cycles resulted in an average read length of approximately 400 nucleotides. The MIRA program (version 3.4.0) was used for de novo assembly of each bifidobacterial genome sequence (16). The number of contigs generated by MIRA was further subjected to manual inspection and alignment using SeqMan (Lasergene) software in order to identify putative overlaps between contig ends. These overlaps were validated by PCR, thus reducing the number of gaps in each bacterial chromosome.

Sequence annotation.

The analyzed genomes consisted of five complete and publicly available bifidobacterial genome sequences plus, as part of this study, 42 newly sequenced genomes. In order to ensure that identical sequence quality standards were applied to all investigated genomes, the five publicly available nucleotide sequences that we used as part of this study were reanalyzed using common software and parameters (see below). Overall DNA analyses of the similarities between the bifidobacterial genomes were carried out using BLASTN (17) and Artemis (18). Protein-encoding open reading frames (ORFs) were predicted using a combination of Prodigal (19) and BLASTX (17) for comparative analysis. Results of the gene-finder program were combined manually with data from BLASTP (20) analysis of a nonredundant protein database provided by the National Center for Biotechnology Information. The combined results were inspected by Artemis, which was used for a manual editing effort to verify and, if necessary, to redefine the start of each predicted coding region or to remove or add coding regions.

Assignment of protein functions to predicted coding regions of the bifidobacterial genomes was performed manually. Moreover, the revised gene/protein set was searched using the Swiss-Prot (www.expasy.ch/sprot/)/TrEMBL, PRIAM (http://priam.prabi.fr/), protein family (Pfam, http://pfam.sanger.ac.uk/), TIGRFam (http://www.jcvi.org/cms/research/projects/tigrfams/overview/), Interpro (INTERPROSCAN; http://www.ebi.ac.uk/Tools/InterProScan/), Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/), and COG (http://www.ncbi.nlm.nih.gov/COG/) databases, in addition to BLASTP (17). Functional assignments were defined by manual processing of the combined results. Manual corrections of automated functional assignments were completed on an individual gene-by-gene basis as needed.

Additional bioinformatic analyses included the following: identification of tRNA genes using tRNAscan-SE (21) and detection of rRNA genes using RNAmmer (http://www.cbs.dtu.dk/services/RNAmmer/) followed by manual annotation on the basis of BLASTN searches and Enzyme Commission (EC)/Gene Onthology (GO) annotation of ORFs using annot8r (22).

Insertion sequence (IS) families were assigned using ISFinder (http://www-is.biotoul.fr/), restriction-modification systems were searched using the REBASE database (23), transporter classification was performed according to the Transporter Classification Database scheme (24), and ORF attribution to a specific COG family of clusters of orthologous genes (COGs) was made by searching the COG database (http://www.ncbi.nlm.nih.gov/COG/).

Pan-genome and extraction of shared and unique genes.

For all bifidobacterial genomes used in this study, a pan-genome calculation was performed using the PGAP pipeline (25); the ORF content of all genomes was organized in functional gene clusters using the GF (Gene Family) method involving comparison of each protein to all other proteins using BLAST analysis (cutoff E value of 1 × 10⁻⁴ and 50% identity over at least 50% of both protein sequences), followed by clustering into protein families, named Bifidobacterium-specific clusters of orthologous genes (BifCOGs), using MCL (graph-theory-based Markov clustering algorithm) (26). A pan-genome profile was built using an optimized algorithm incorporated in PGAP software, based on a presence/absence matrix that included all identified bifCOGs in the analyzed genomes. Following this, the unique protein families for each of the 47 bifidobacterial genomes were classified. Protein families shared between all genomes, named core BifCOGs, were defined by selecting the families that contained at least one single protein member for each genome.

The PGAP pipeline calculation was performed again with the inclusion of the remaining members of the family Bifidobacteriaceae in order to predict the pan-genome and core COGs of the entire family.

Each set of orthologous proteins constituting core COGs with one member per genome was aligned using MAFFT (27), and phylogenetic trees were constructed using the neighbor-joining method in Clustal W version 2.1 (28). The supertree was built using FigTree (http://tree.bio.ed.ac.uk/software/figtree/). PhyloPhlAn (29) was used to construct an additional phylogenetic tree based on >400 proteins optimized from among 3,737 bacterial genomes. This method measures the sequence diversity of all clades, classifies genomes from deep-branching candidate divisions through closely related subspecies, and improves the consistency of the phylogenetic and taxonomic groupings based on the 400 most conserved bacterial proteins.

Prediction of gene acquisition and loss.

Prediction and tree visualization of gene acquisition and loss were performed with BlastGraph (30). Data from BLASTP (17) comparisons of all the deduced proteins derived from the pan-genome to each other were used as the input, and the clustering cutoff was set at 50% identity over at least 50% of both protein sequences.

Prediction of the mobilome of bifidobacteria.

The identification of the so-called bifidobacterial “mobilome” (i.e., the genes that may have been acquired by horizontal gene transfer [HGT]) was achieved by merging results from DarkHorse v1.5 (31) and suite COLOMBO v3.8 implemented with the program SIGI-HMM (32). DarkHorse was run with default parameters, and only results with an E value of <1e-30 were retained, while COLOMBO was run with a sensitivity value of 0.4.

Identification of CRISPR.

Clustered regularly interspaced short palindromic repeats (CRISPR) were identified using the CRISPR Finder software (33). Once CRISPR were identified, flanking coding sequences were analyzed and mined for the presence of cas genes. Once cas genes were identified, the universal cas1 gene, in combination with the signature genes for type I, type II, and type III CRISPR-associated proteins (Cas) systems, namely, cas3, cas9, and cas10, respectively, were used for CRISPR type assignment. Furthermore, CRISPR locus orientations were determined using the widely applicable codirectional transcription pattern of cas genes with the CRISPR-spacer array. Once the orientation of CRISPR was determined and the corresponding sequence established, CRISPR within a locus were identified, and interspacing sequences were established as spacers.

Data deposition.

The sequences reported in this paper have been deposited in the GenBank database under the accession numbers indicated in Table 1.

TABLE 1.

General features of the Bifidobacterium genomes^a

Genome	Bifidobacterium strain	Fold coverage	Genome status (no. of contigs)b	Approximate genome size (nt)	GC content	No. of ORFs	No. of ORFs with an assigned function	No. of tRNAs	No. of complete CRISPR loci	No. of partial CRISPR locic	Avg length of ORFs (nt)	ORF region (%)	Isolation source	GenBank accession no.
1	B. actinocoloniiforme DSM 22766	88.61	Draft (4)	1,823,388	62.71	1,484	1,190	46	1		1,066.01	86.99	Bumblebee digestive tract	JGYK00000000
2	B. adolescentis ATCC 15703		Complete	2,089,645	59.18	1,649	1,419	54	1		1,093.98	86.33	Intestine of adult	AP009256.1
3	B. angulatum LMG 11039	84.47	Draft (6)	2,003,806	59.41	1,523	1,314	48	1		1,139.59	86.61	Human feces	JGYL00000000
4	B. animalis subsp. animalis LMG 10508	61.16	Draft (13)	1,915,007	60.47	1,527	1,254	52	1		1,081.93	86.27	Rat feces	JGYM00000000
5	B. animalis subsp. lactis DSM 10140		Complete	1,938,606	60.48	1,518	1,242	52	1		1,100.61	86.18	Fermented milk	CP001606.1
6	B. asteroides LMG 10735 (PRL2011)		Complete	2,167,304	60.05	1,653	1,363	44	1		1,138.82	86.86	Honeybee hindgut	CP003325.1
7	B. biavatii DSM 23969	45.11	Draft (56)	3,252,147	63.1	2,557	2,068	61			1,095.68	86.15	Feces of tamarin	JGYN00000000
8	B. bifidum LMG 11041	118.97	Draft (2)	2,208,468	62.67	1,704	1,270	53			1,088.44	83.98	Feces from breast-fed infant	JGYO00000000
9	B. bohemicum DSM 22767	140.24	Draft (5)	2,052,470	57.45	1,632	1,271	47	1	1	1,049.90	83.48	Bumblebee digestive tract	JGYP00000000
10	B. bombi DSM 19703	103.8	Draft (4)	1,895,239	56.08	1,454	1,121	48	1		1,081.75	82.76	Bumblebee digestive tract	ATLK00000000
11	B. boum LMG 10736	71.1	Draft (18)	2,171,356	59.31	1,726	1,412	49	1		1,075.57	85.50	Bovine rumen	JGYQ00000000
12	B. breve LMG 13208	21.14	Draft (31)	2,263,780	58.88	1,887	1,506	53	2		1,036.17	86.42	Infant intestine	JGYR00000000
13	B. callitrichos DSM 23973	65.86	Draft (33)	2,887,313	63.52	2,364	1,970	58	1		1,051.77	86.11	Feces of common marmoset	JGYS00000000
14	B. catenulatum LMG 11043	31.21	Draft (11)	2,082,756	56.11	1,664	1,396	55	1		1,072.31	85.67	Adult intestine	JGYT00000000
15	B. choerinum LMG 10510	107.33	Draft (20)	2,096,123	65.53	1,672	1,397	55	1		1,074.54	85.71	Piglet feces	JGYU00000000
16	B. coryneforme LMG 18911	182.57	Complete	1,755,151	60.51	1,364	1,133	56			1,130.49	87.85	Honeybee hindgut	CP007287
17	B. crudilactis LMG 23609	90.52	Draft (6)	2,362,816	57.72	1,883	1,606	45	1		1,089.40	86.82	Raw cow milk	JHAL00000000
18	B. cuniculi LMG 10738	120.38	Draft (41)	2,531,592	64.87	2,194	1,661	63	3		994.86	86.22	Rabbit feces	JGYV00000000
19	B. dentium LMG 11045 (Bd1)		Complete	2,636,367	58.54	2,129	1,625	55	2	2	1,067.06	86.17	Oral cavity	CP001750.1
20	B. gallicum LMG 11596	109.79	Draft (12)	2,004,594	57.61	1,507	1,293	58	2		1,116.18	83.91	Adult intestine	JGYW00000000
21	B. gallinarum LMG 11586	244.47	Draft (10)	2,160,836	64.22	1,654	1,384	53			1,131.86	86.64	Chicken cecum	JGYX00000000
22	B. indicum LMG 11587	280.81	Complete	1,734,546	60.49	1,352	1,141	47			1,129.67	88.05	Insect	CP006018
23	B. kashiwanohense DSM 21854	63.96	Draft (30)	2,307,960	56.2	1,948	1,618	53			1,023.36	86.37	Infant feces	JGYY00000000
24	B. longum subsp. infantis ATCC 15697		Complete	2,832,748	59.86	2,500	1,939	79			974.00	85.96	Intestine of infant	AP010889.1
25	B. longum subsp. longum LMG 13197	34.84	Draft (8)	2,384,703	60.33	1,899	1,556	71			1,083.59	86.29	Adult intestine	JGYZ00000000
26	B. longum subsp. suis LMG 21814	70.88	Draft (36)	2,335,832	59.96	1,955	1,675	55	1		1,027.46	86.04	Pig feces	JGZA00000000
27	B. magnum LMG 11591	80.25	Draft (13)	1,822,476	58.72	1,507	1,234	56	1		1,060.59	87.64	Rabbit feces	JGZB00000000
28	B. merycicum LMG 11341	78.12	Draft (16)	2,280,236	60.33	1,741	1,413	53	1	1	1,105.42	84.45	Bovine rumen	JGZC00000000
29	B. minimum LMG 11592	231.93	Draft (18)	1,892,860	62.73	1,590	1,356	53	1		1,032.77	86.75	Sewage	JGZD00000000
30	B. mongoliense DSM 21395	128.8	Draft (43)	2,170,490	62.78	1,798	1,514	47			1,040.02	86.15	Fermented mare's milk	JGZE00000000
31	B. pseudocatenulatum LMG 10505	52	Draft (10)	2,283,767	56.36	1,771	1,527	53	1		1,112.53	86.27	Infant feces	JGZF00000000
32	B. pseudolongum subsp. globosum LMG 11569	151.96	Draft (26)	1,935,255	63.39	1,574	1,367	52			1,091.36	88.76	Bovine rumen	JGZG00000000
33	B. pseudolongum subsp. pseudolongum LMG 11571	85.19	Draft (11)	1,898,684	63.06	1,495	1,310	52	2		1,111.81	87.54	Swine feces	JGZH00000000
34	B. psychraerophilum LMG 21775	72.17	Draft (11)	2,615,078	58.75	2,122	1,809	45			1,080.93	87.71	Pig cecum	JGZI00000000
35	B. pullorum LMG 21816	99.94	Draft (11)	2,153,559	64.22	1,691	1,466	53	1		1,097.52	86.18	Chicken feces	JGZJ00000000
36	B. reuteri DSM 23975	61.47	Draft (28)	2,847,572	60.45	2,149	1,747	53		2	1,127.08	85.06	Feces of common marmoset	JGZK00000000
37	B. ruminantium LMG 21811	102.57	Draft (23)	2,249,807	59.18	1,832	1,433	50	1	5	1,068.51	87.01	Bovine rumen	JGZL00000000
38	B. saeculare LMG 14934	68.91	Draft (14)	2,263,283	63.75	1,857	1,524	48			1,079.55	88.58	Rabbit feces	JGZM00000000
39	B. saguini DSM 23967	189.92	Draft (33)	2,787,036	56.35	2,321	1,853	59	1		1,055.87	87.93	Feces of tamarin	JGZN00000000
40	B. scardovii LMG 21589	68.58	Draft (34)	3,141,793	64.63	2,480	2,098	55	1	1	1,070.48	84.50	Blood	JGZO00000000
41	B. stellenboschense DSM 23968	108.97	Draft (40)	2,812,864	65.34	2,202	1,810	59	1		1,100.08	86.12	Feces of tamarin	JGZP00000000
42	B. stercoris DSM 24849	173.24	Draft (15)	2,304,613	59.38	1,891	1,548	54		1	1,070.70	87.85	Adult feces	JGZQ00000000
43	B. subtile LMG 11597	123.73	Draft (27)	2,790,088	60.92	2,260	1,881	47		3	1,027.75	83.25	Sewage	JGZR00000000
44	B. thermacidophilum subsp. porcinum LMG 21689	130.58	Draft (3)	2,079,368	60.2	1,738	1,229	40		1	1,009.07	84.34	Piglet feces	JGZS00000000
45	B. thermacidophilum subsp. thermacidophilum LMG 21395	75.19	Draft (8)	2,233,072	60.38	1,823	1,339	48	1		1,021.36	83.38	Anaerobic digester	JGZT00000000
46	B. thermophilum JCM 1207	84.48	Draft (12)	2,099,496	59.91	1,700	1,305	44	2		1,065.02	86.24	Swine feces	JGZV00000000
47	B. tsurumiense JCM 13495	89.67	Draft (25)	2,164,426	52.84	1,629	1,403	46	2		1,114.56	83.88	Hamster dental plaque	JGZU00000000

^ant, nucleotides.

^bThe genome sequences were fragmented into various numbers of contigs; the numbers are provided within parentheses.

^cCRISPR partial loci represent a CRISPR-Cas system without the full complement of cas genes.

RESULTS AND DISCUSSION

General features of Bifidobacterium genomes.

Genome sequences were determined for 42 distinct bifidobacterial strains, while an additional five bifidobacterial genome sequences were retrieved from the NCBI public database, together representing the neotype for each of the currently described 47 species and subspecies within the Bifidobacterium genus (34). The sequencing and assembly statistics of the 42 newly determined bifidobacterial genomes are summarized in Table 1. The approximate Bifidobacterium genome size ranged from 1.73 Mb (Bifidobacterium indicum) to 3.25 Mb (Bifidobacterium biavatii), corresponding to 1,352 and 2,557 predicted protein-encoding open reading frames, respectively (Table 1). Given the close phylogenetic relationship between bifidobacteria, such a substantial size difference suggests that bifidobacterial genomes have evolved as a result of many gene loss and/or acquisition events (35). Genome features of the sequenced bifidobacterial strains are presented in Table 1; functional annotations were assigned for 81.9% of the predicted ORFs identified in the analyzed members of the Bifidobacterium genus, representing the Bifidobacterium pan-genome (see below). The remaining 18.1% ORFs were assigned as proteins with an unknown function. Results from BLASTP searches of the NCBI database show that 17.7% of these ORFs of “unknown function” (corresponding to 3.2% of the total Bifidobacterium pan-genome) have homologs in other bacterial genera within the Bifidobacteriaceae family (i.e., members of the genera Scardovia, Parascardovia, Metascardovia, and Gardnerella). It is noteworthy that approximately 12.4% of the annotated ORFs were attributed to carbohydrate metabolism. These data are a genetic reflection of the metabolic commitment of bifidobacteria to a saccharolytic life style, a notion observed for other bacteria of the human gut microbiota (36).

The pan-genome, core genome, and variome of the Bifidobacterium genus.

Genome sequences from each of the 47 Bifidobacterium (sub)species were used to analyze the corresponding pan-genome, the core genome, and the variome (variable genome sequences), determined as described previously (37). A total of 18,181 BifCOGs (Bifidobacterium-specific clusters of orthologous genes), of which 6,464 had members present in at least two genomes, and which together represent the pan-genome of the Bifidobacterium genus, were identified in the 47 bifidobacterial genomes. The pan-genome size, when plotted versus the number of included genomes, clearly shows that the power trend line has yet to reach a plateau (Fig. 1). Nevertheless, the number of new genes discovered by sequential addition of genome sequences was reduced from 770 to 588 BifCOGs in the first three genome additions to 252 to 249 BifCOGs in the final three additions, indicating the existence of an open pan-genome within the Bifidobacterium genus. These findings suggest that additional sequencing efforts are needed in order to identify (essentially) all genes of members of this genus. Analysis of the set of predicted BifCOGs allowed the identification of 551 COGs shared by all 47 Bifidobacterium (sub)species, thereby representing the core of bifidobacterial genomic coding sequences (core BifCOGs). Plotting the identified number of core BifCOGs as a function of the included number of genomes shows that the core BifCOG set is not expected to be significantly reduced in number by the addition of further genomes since the exponential trendline essentially reached a plateau (Fig. 1). Inclusion of available genome sequences of other members of the family Bifidobacteriaceae (i.e., Gardnerella vaginalis 409-05, Metascardovia criceti DSM 17774, Parascardovia denticolens DSM 10105, Scardovia inopinata F0304, and Scardovia wiggsiae F0424) generated a core COG set of the family Bifidobacteriaceae consisting of 451 members. This relatively high number of members of the conserved genetic arsenal within the Bifidobacteriaceae is indicative of a close evolutionary relationship between members of this family (38). Examination of the functional annotation of the core BifCOGs, based on the updated COG database (39), suggests, as anticipated, that most of the conserved core genes specify housekeeping functions or functions related to adaptation to or interaction with a particular environment, such as carbohydrate metabolism, cell envelope biogenesis, amino acid biosynthesis and transport, or nucleotide biosynthesis and transport (see Fig. S1 in the supplemental material). Notably, only 5.5% of the core genome is involved in carbohydrate metabolism (see Fig. S1), whereas the carbohydrate metabolism functional family is the most highly represented COG family within the Bifidobacterium pan-genome (13.7%) (see Fig. S1). This indicates that a strong selective pressure exists with respect to the acquisition and retention of accessory (novel) genes for carbohydrate utilization by bifidobacteria in order for them to be competitive in the particular ecological niche in which they reside. The pan-genome analysis also allowed the identification of the variome, which includes truly unique genes (TUGs), i.e., genes present in just one of the examined bifidobacterial genomes. Predicted TUGs were validated by BLASTx searches in the analyzed genomes in order to avoid false positives imputable to the gene-calling algorithm. The numbers of TUGs range from 47 for B. indicum to 595 for Bifidobacterium cuniculi LMG10738 in the 47 bifidobacterial genomes analyzed (see Fig. S1). The mean number of TUGs found in the Bifidobacterium genome data set is 249. The large deviation from the mean is indicative of a high degree of genome diversity within members of the genus Bifidobacterium, which is typical for related species that have individually adapted to different environments (40). As expected, the majority (54.1%) of TUGs have no functional annotation (see Fig. S1). Nevertheless, 13.2% of TUGs can be attributed to a COG family representing proteins involved in carbohydrate metabolism, including glycosyl hydrolases (GH) and proteins involved in carbohydrate uptake. TUG identification in bifidobacteria may serve to identify targets for functional studies on adaptive abilities, in particular, studies on host interactions and metabolism of (saccharidic) host/diet-derived components (1).

FIG 1.

Pan-genome and core genome of the genus Bifidobacterium. The pan-genome (panel a) and core genome (panel b) are represented as variations of the sizes of their gene pools upon sequential addition of the 47 bifidobacterial genomes. The x axes represent the numbers of genomes, whereas the y axes represent the numbers of genes. Expon., exponential.

Phylogenomics of Bifidobacterium genus.

The availability of genome sequences for all members of the genus Bifidobacterium and for five members of the Bifidobacteriaceae family allows an in-depth analysis of the projected evolutionary development of this genus and family. A phylogenetic supertree was constructed based on the concatenated protein sequences of 404 identified Bifidobacteriaceae core COGs, excluding paralogs from the same genome (Fig. 2), an approach that increases the robustness of phylogenetic analyses (41). A consistent phylogeny was obtained using PhyloPhlAn (29), whereas certain discrepancies in the branching of the various bifidobacterial (sub)species were noticed in comparisons of the bifidobacterial core COG-based tree with the 16S rRNA gene-based tree. This observation reveals evolutionary development within the Bifidobacterium genus that is somewhat different from that previously reported, although it did confirm that bifidobacteria represent the deepest branch separating them from other genera within this family (34) (Fig. 2). Furthermore, the Bifidobacterium asteroides phylogenetic group is positioned close to the root in the core genome-based supertree, suggesting a close relationship of members of this group to the Bifidobacterium ancestor, as was previously noticed for the genome of B. asteroides PRL2011 (42).

FIG 2.

Phylogenomic overview of the family Bifidobacteriaceae. A supertree based on the alignment of 404 core COGs (with a single representative identified for each genome of members of the family Bifidobacteriaceae) was constructed in order to obtain a robust phylogenetic reconstruction. Phylogenetic clusters are highlighted with similarly colored branches, and nodes with bootstrap values higher than 70% are marked with a purple dot. The phylogenetic clusters close to the root of the tree may represent species that are most closely related to the ancestor of the Bifidobacterium genus. Circles surrounding the tree represent the approximate genome sizes (in blue), numbers of TUGs (in red), percentages of genes predicted to have undergone horizontal gene transfer (in green), and percentages of genes predicted to be subject to horizontal gene transfer and carbohydrate metabolism and transport (in orange). The outermost layer represents the numbers of the complete predicted degradation pathways. E. coli, Escherichia coli; met., metabolism.

Evolution of bifidobacterial genomes.

Evolution by gene acquisition and loss of the genus Bifidobacterium following speciation from a common ancestor of all Bifidobacteriaceae can be reconstructed through BlastGraph (30), thereby generating a tree based on information regarding the presence or absence of COGs in every taxa of this family and on the use of the maximum-parsimony algorithm (43) (Fig. 3). The observed difference in species clustering as revealed by this tree compared to that shown by the core COG-based supertree (Fig. 2) is highly informative with respect to possible horizontal gene transfer (HGT) events (44). Such analyses predict that the genome of the common ancestor of the genus Bifidobacterium consisted of approximately 1,048 COGs. This putative ancestor possessed just 179 fewer COGs than the number harbored by the B. indicum genome and as many as 1,091 fewer COGs than the B. biavatii chromosome, representing the smallest and largest genomes, respectively. Thus, the evolution of current bifidobacterial species appears to have involved a relatively limited number of ancestral gene loss incidences but an extensive number of gene acquisition events (Fig. 3). This contrasts with other bacteria, for example, the genomes of genera belonging to the lactic acid bacteria, which are believed to have undergone extensive simplification (45). Various changes identified at this stage of evolution may be linked to the transition to life in an environment characterized by high complexity and abundance of microbial communities. In this context, the acquisition of genes required for the utilization of diet/host-derived carbohydrates provided a clear competitive advantage in a complex microbial community such as the ecological niches of bifidobacteria. An example of this evolutionary trend is represented by the milk-adapted Streptococcus thermophilus and the closest phylogenetic neighbor Streptococcus salivarius. S. salivarius is an inhabitant of the oral cavity of mammals, and, despite the high-level phylogenetic relationship with S. thermophilus, the two species show extremely different carbohydrate utilization patterns, with only a few sugars utilized by the latter (46). The predicted Bifidobacterium ancestor would have been a microaerophile or facultative aerobe, which is reflected by the loss of the genes specifying the electron chain transport cytochrome bd subunits and particular enzymes (i.e., catalase and superoxide dismutase), which allow removal of toxic products that arise as a result of oxygen-mediated respiration (predicted to be present in members of the B. asteroides phylogenetic group). Gain of new gene families that originated either by lineage-specific gene duplication or by acquisition of paralogous genes through HGT seems to be a prevailing trend in the evolution of the genus Bifidobacterium (Fig. 3) and is different from what is observed in other bacterial lineages, e.g., the genus Lactobacillus (45). The evolution of the genome of lactobacilli is thought to have involved ancestral gene decay and metabolic simplification but also a substantial number of duplications and acquisition of unique genes, most of which are predicted to code for peptidases or proteases (45). Lineage-specific gene acquisition appears to have been extensive within the Bifidobacterium genus, as illustrated by the B. biavatii and Bifidobacterium longum subsp. infantis taxa (showing acquisition of 1,091 and 1,092 COGs compared to the presumed ancestral Bifidobacterium taxon, respectively), while these two species seem to have undergone relatively limited genome decay (Fig. 3). Gene acquisition events occurring in the course of evolution of microbial genomes are believed to support adaptation to a new ecological niche or acquisition of increased competitiveness in an existing ecological niche (40). Analysis of the gene families putatively involved in acquisition events indicates that adaptation to growth in environments rich in complex carbohydrates, such as the animal gut, has been the main driving force responsible for retention of gene duplications and HGT-acquired genes during the speciation of Bifidobacterium. An intriguing finding supporting this hypothesis is the presence of a large arsenal of genes encoding enzymes involved in carbohydrate metabolism, especially glycosyl hydrolases, many of which are predicted to have been duplicated or acquired at different times throughout the evolution of this genus (Fig. 4). GHs feed cell bioenergetics, i.e., ATP-producing pathways, which is known to be under high selection pressure during evolution (47), thus representing a strong driving force in genome shaping. Notably, we identified eight COGs predicted to encompass GH43 family members, which are GHs crucial for the degradation of plant polysaccharides (48) and appear to have been acquired early in the evolution of bifidobacteria, while seven COGs encompass members of the large GH13 family, representing α-amylases (48), and appear to have been acquired during the evolution of Bifidobacteriaceae and prior to the GH43 member acquisition (Fig. 4). Furthermore, several presumably acquired genes were identified that encode proteins with predicted carbohydrate uptake functions, including ATP-binding cassette (ABC) transporters, phosphoenolpyruvate-phosphotransferase system (PEP-PTS) transporters, and major facilitator superfamily (MFS) transporters. This supports the hypothesis that bifidobacteria selectively acquired new metabolic capabilities which allowed them access to a larger number of carbon and energy sources. While it seems clear that gene gain was and is the main driving force of bifidobacterial evolution (Fig. 3 and 4), gene decay and metabolic simplification may still be very important for niche-specific adaptation. Various gene loss events, in particular, loss of those encoding biosynthetic enzymes, were detected in the main phylogenetic groups of the genus Bifidobacterium, which presumably reflects analogous environmental pressures. Regarding GHs, it was observed that GH43 family members involved in the degradation of plant polysaccharides appear to have been largely lost in more recent times by a subgroup of 18 Bifidobacterium species (Fig. 3 and 4), while most GH13 family members encompassing α-amylases seem to have been deleted (with respect to the predicted Bifidobacteriaceae ancestor) from the genomes of the clade encompassing bifidobacteria isolated from honeybees and bumblebees (B. asteroides, Bifidobacterium actinocoloniiforme, B. indicum, Bifidobacterim coryneforme, Bifidobacterium bombi, and Bifidobacterium bohemicum), perhaps because these metabolic abilities became obsolete due to the particular diet of their arthropod hosts (Fig. 4).

FIG 3.

Gene gain and loss events in a reconstruction of data representing the family Bifidobacteriaceae. A tree was constructed using information related to the presence or absence of COGs for the whole Bifidobacteriaceae pan-genome. Each node is represented by a pie diagram showing the acquired COGs (in black) and the COGs derived from the previous node (in gray). Furthermore, additional information is displayed at each node as follows: number of acquired genes/number of lost genes/total number of COGs. The predicted Bifidobacterium ancestor is highlighted with thick black circle surrounding the pie diagram.

FIG 4.

Reconstruction of gene gain and loss events regarding genes encoding members of the GH3, GH13, and GH43 families in the family Bifidobacteriaceae. A tree was constructed using information related to the presence or absence of COGs for the whole Bifidobacteriaceae pan-genome. Each node is marked by a pie diagram showing the acquired COGs (in black) and the COGs derived from the previous node (in gray). Furthermore, the number of members of the glycosyl hydrolase families GH3, GH13, and GH43 that had been acquired (in black) or lost (in gray) is indicated close to each diagram.

Mobilome of bifidobacterial genomes.

The identification of genes that may have been acquired by HGT (the so-called mobilome) was performed using the software suite COLOMBO v3.8 implemented with the program SIGI-HMM (32) and DarkHorse software (31). The obtained results were merged, and the identified percentages of predicted alien genes, compared to the total number of ORFs, were shown to range from 6.1% in B. indicum to 26.5% in Bifidobacterium saguini (Table 2). Predicting the donors of these putative alien genes indicated a preferential origin from other members of the Actinobacteria class (28.5%), followed by Bacillus (11.7%), Gammaproteobacteria (8.7%), Clostridium (8.7%), and Alphaproteobacteria (5.9%) (Table 3). It is noteworthy that members of these donor classes are also widespread in the gut environment (49). These data are supportive of the idea that HGT events are the major driver for evolutionary development in members of the Bifidobacterium genus.

TABLE 2.

Predicted horizontal gene transfer in the Bifidobacterium genus

Bifidobacterium strain	No. of native genes	Putative no. of alien genes	Native genes (%)	Putative alien genes (%)
B. actinocoloniiforme DSM 22766	1,230	258	82.7	17.3
B. adolescentis ATCC 15703	1,475	174	89.4	10.6
B. angulatum LMG 11039	1,420	103	93.2	6.8
B. animalis subsp. animalis LMG 10508	1,366	161	89.5	10.5
B. animalis subsp. lactis DSM 10140	1,373	145	90.4	9.6
B. asteroides LMG 10735 (PRL2011)	1,227	426	74.2	25.8
B. biavatii DSM 23969	1,918	639	75.0	25.0
B. bifidum LMG 11041	1,499	205	88.0	12.0
B. bohemicum DSM 22767	1,388	244	85.0	15.0
B. bombi DSM 19703	1,278	176	87.9	12.1
B. boum LMG 10736	1,532	194	88.8	11.2
B. breve LMG 13208	1,563	324	82.8	17.2
B. callitrichos DSM 23973	1,921	443	81.3	18.7
B. catenulatum LMG 11043	1,540	124	92.5	7.5
B. choerinum LMG 10510	1,467	205	87.7	12.3
B. coryneforme LMG 18911	1,264	100	92.7	7.3
B. crudilactis LMG 23609	1,476	407	78.4	21.6
B. cuniculi LMG 10738	1,700	494	77.5	22.5
B. dentium LMG 11045 (Bd1)	1,831	298	86.0	14.0
B. gallicum LMG 11596	1,339	168	88.9	11.1
B. gallinarum LMG 11586	1,365	289	82.5	17.5
B. indicum LMG 11587	1,269	83	93.9	6.1
B. kashiwanohense DSM 21854	1,703	245	87.4	12.6
B. longum subsp. infantis ATCC 15697	1,845	655	73.8	26.2
B. longum subsp. longum LMG 13197	1,648	251	86.8	13.2
B. longum subsp. suis LMG 21814	1,635	321	83.6	16.4
B. magnum LMG 11591	1,346	161	89.3	10.7
B. merycicum LMG 11341	1,506	236	86.5	13.5
B. minimum LMG 11592	1,342	248	84.4	15.6
B. mongoliense DSM 21395	1,444	354	80.3	19.7
B. pseudocatenulatum LMG 10505	1,578	193	89.1	10.9
B. pseudolongum subsp. globosum LMG 11569	1,413	161	89.8	10.2
B. pseudolongum subsp. pseudolongum LMG 11571	1,360	135	91.0	9.0
B. psychraerophilum LMG 21775	1,574	548	74.2	25.8
B. pullorum LMG 21816	1,363	328	80.6	19.4
B. reuteri DSM 23975	1,791	358	83.3	16.7
B. ruminantium LMG 21811	1,608	224	87.8	12.2
B. saeculare LMG 14934	1,427	430	76.8	23.2
B. saguini DSM 23967	1,707	614	73.5	26.5
B. scardovii LMG 21589	1,858	622	74.9	25.1
B. stellenboschense DSM 23968	1,865	337	84.7	15.3
B. stercoris DSM 24849	1,716	175	90.7	9.3
B. subtile LMG 11597	1,692	568	74.9	25.1
B. thermacidophilum subsp. porcinum LMG 21689	1,572	166	90.4	9.6
B. thermacidophilum subsp. thermacidophilum LMG 21395	1,571	252	86.2	13.8
B. thermophilum JCM 1207	1,441	259	84.8	15.2
B. tsurumiense JCM 13495	1,416	213	86.9	13.1

TABLE 3.

HGT in the Bifidobacterium pan-genome

Putative donor	%a
Actinobacteria	28.5
Alphaproteobacteria	5.9
Bacilli	11.7
Bacteroides	0.2
Bacteroidia	0.6
Betaproteobacteria	3.4
Chlorobia	1.1
Chloroflexi	0.2
Clostridia	8.7
Deltaproteobacteria	0.9
Erysipelotrichia	0.4
Flavobacteria	2.0
Gammaproteobacteria	8.7
Halobacteria	0.7
Methanopyri	0.4
Negativicutes	0.7
Nitrospira	1.1

^aPercentage of the total HGT sequences identified in bifidobacteria corresponding to the various taxa.

The predicted bifidobacterial mobilome, with exclusion of prophage-associated and transposase-encoding genes or genes with no known function, was analyzed through COG assignment, revealing that the most highly represented (15.3%) functional class is that of carbohydrate metabolism and transport (Table 4). Notably, HGT events encompassing genes involved in carbohydrate metabolism and transport include genes encoding key enzymes such as GHs (representing 3.4% of the predicted mobilome) and genes predicted to specify glycosyl transferases (GTs) and carbohydrate transporters (ABC, MFS, and PTS classes), which constitute 2.6% and 4.0% of the predicted mobilome, respectively, while genes involved in exopolysaccharide (EPS) biosynthesis (with the partial inclusion of GTs) correspond to 3.7% of the predicted mobilome. Interestingly, the GH families that appear most affected by HGT events are GH43 and GH3, representing, respectively, 10.7% and 8.7% of the total pool of GHs involved in HGT. Members of GH43 and GH3 families have been shown to be involved in the breakdown of polysaccharides encompassing arabinose and xylose residues (50), thus supporting the hypothesis that the ability to utilize plant polysaccharides has been acquired by HGT in recent ancestors or actual members (e.g., Bifidobacterium reuteri, B. biavatii, and Bifidobacterium scardovii) of the genus Bifidobacterium rather than by vertical evolution.

TABLE 4.

COG function

Category of cluster of orthologous genes	%a
Translation, ribosomal structure, and biogenesis	3.8
RNA processing and modification	0.0
Transcription	10.7
Replication, recombination, and repair	5.7
Chromatin structure and dynamics	0.0
Cell cycle control, cell division, chromosome partitioning	1.2
Nuclear structure	0.0
Defense mechanisms	7.4
Signal transduction mechanisms	3.3
Cell wall/membrane/envelope biogenesis	7.5
Cell motility	0.1
Cytoskeleton	0.0
Extracellular structures	0.0
Intracellular trafficking, secretion, and vesicular transport	0.6
Posttranslational modification, protein turnover, chaperones	1.8
Energy production and conversion	3.6
Carbohydrate transport and metabolism	15.3
Amino acid transport and metabolism	9.4
Nucleotide transport and metabolism	2.3
Coenzyme transport and metabolism	2.2
Lipid transport and metabolism	2.2
Inorganic ion transport and metabolism	4.9
Secondary metabolites biosynthesis, transport, and catabolism	1.1
General function prediction only	12.8
Function unknown	4.0

^aPercentage of coding sequences for each category.

In silico analysis of the bifidobacterial pan-genome highlights an abundance of prophage-like elements (3.2% of the total pan-genome size, representing about 7.6% of the predicted bifidobacterial mobilome) and a rich arsenal of insertion sequences (IS), belonging to 16 IS families, with an abundance of IS3, IS21, IS256, ISL3, and IS200/IS605 family members, constituting approximately 3.2% of the total predicted mobilome.

Additional putative mobile elements identified in the Bifidobacterium pan-genome are represented by CRISPR loci. CRISPR and CRISPR-associated proteins (Cas) constitute the CRISPR-Cas system, which provides adaptive immunity against exogenous genetic elements in bacteria and archaea (51). Typically, DNA from invasive elements is captured in CRISPR loci and subsequently transcribed into small interfering RNAs that guide Cas nucleases for sequence-specific targeting and cleavage of cDNA (52). We identified the three main types of CRISPR-Cas systems, namely, type I, type II, and type III, in the genomes of bifidobacteria and observed 43, 6, and 7 systems, respectively (Table 1). Overall, we identified 56 distinct loci in 35 genomes, and the high level of occurrence of type I systems (43 loci with a type I CRISPR and 29 loci with a cas3 signature gene) is consistent with their prevalent distribution in bacteria (53). Interestingly, we observed 6 type II systems and identified 5 cas9 signature genes. Lastly, we observed remnants of 7 putative type III loci, including several cmr genes. Overall, a diversity of CRISPR-Cas systems occurs in bifidobacteria, at a frequency (35/47 genomes, 75%) much higher than that generally observed in the genomes of bacteria, of which just 46% contain CRISPR loci (54). Beyond diversity at the CRISPR-Cas system type level, we further observed diversity in terms of locus size with loci ranging from 4 to 172 CRISPR spacers, with an average of 60 spacers, which is also unusually high. It is noteworthy that we observed CRISPR loci in all the major phylogenetic groups of bifidobacteria, indicating that these systems are evolutionarily widespread throughout this genus. This is consistent with previous analyses reporting their occurrence in various Bifidobacterium species (55, 56), and matches between CRISPR spacer sequences and those of bacteriophages and plasmids suggest these widespread loci may also provide adaptive immunity against viruses and plasmids in most bifidobacteria.

In silico analyses of central metabolism.

In order to depict an overview of the metabolic capabilities of the entire genus Bifidobacterium, we conducted a prediction of complete metabolic pathways in every species through the use of Pathway tools software. Homologs of all enzymes necessary for the fermentation of glucose and fructose to lactic acid and acetate through the characteristic “fructose-6-phosphate shunt” (57), as well as a partial Embden-Meyerhoff pathway, were annotated in the Bifidobacterium core genome. These metabolic pathways are important for generation of pyruvate and oxidation of NADH, as well as for synthesis of an additional ATP molecule per glucose during the conversion of pyruvate to acetate, producing a higher energetic yield than lactic acid bacteria (58).

Genes encoding complete biosynthetic pathways for amino acids, purines, and pyrimidines from glutamine were variously present within the genus Bifidobacterium, with generally fewer of such pathways in the genomes of bifidobacteria isolated from insects (Fig. 5).

FIG 5.

Prediction of complete amino acid, vitamin, and cofactor biosynthesis pathways and non-carbohydrate degradation pathways. Panel a shows a heatmap of the amino acid, vitamin, and cofactor biosynthetic pathways present in the analyzed bifidobacterial genomes. Panel b displays a heatmap that shows all complete non-carbohydrate degradation pathways found in the genus Bifidobacterium. Panel c shows a heatmap illustrating glycodeoxycholate and taurodeoxycholate degradation capabilities, along with the presence of a bile salt hydrolase gene, in the analyzed bifidobacterial species. Each bifidobacterial genome analyzed is numbered according to the numbering of the species displayed in Table 1. Black and gray squares in panels a to c represent the absence and presence of genes.

Similarly, homologs for pathways to produce the vitamins riboflavin (B2), tetrahydrofolate (B9), thiamine (B10), and pyridoxal 5′-phosphate (B6) are also variously distributed in the genomes of this bacterial genus (Fig. 5). Interestingly, while tetrahydrofolate is not produced by mammals (59), it is predicted to be synthetized by all the analyzed species of bifidobacteria isolated from humans (with the sole exception of B. gallicum) or other primates (with the sole exception of B. biavatii). This suggests that tetrahydrofolate production by gut bacteria represents an important source of vitamin B11 for the host and a clear example of microbe-host coevolution (Fig. 5). Additionally, an intermediate in the riboflavin biosynthetic pathway has been shown to be involved in activation of mucosa-associated invariant T (MAIT) cells (60). Notably, four bifidobacterial species are predicted to possess a complete riboflavin biosynthesis pathway (Fig. 5), which may represent an additional mechanism for microbe-host interaction by stimulation of the host's immune system (Fig. 5).

Notably, a hierarchical representation of these biosynthetic pathways highlighted a closer coclustering of those species isolated from insects as well as from rabbit and poultry (Fig. 5), suggesting specialization with respect to these ecological niches following an adaptation to their host diet. Other metabolic capabilities of the genus Bifidobacterium predicted by our in silico analyses are displayed in Fig. 5. Interestingly, B. asteroides, B. indicum, B. coryneforme, B. actinoloniiforme, B. bohemicum, and B. bombi, isolated from various insect guts and with a small genome size compared to those of other members of the Bifidobacterium genus, possess narrow repertoires of biosynthetic pathways, while B. callitrichos and B. stellenboschense, possessing two of the largest genomes within the Bifidobacteriaceae, seem to have retained a much broader biosynthetic inventory.

Furthermore, none of the currently described taxa belonging to the Bifidobacterium genus possess a complete mevalonate pathway for isoprenoid biosynthesis, except for seven members of the most ancient branches of the core COG tree encompassing the B. actinocoloniiforme, B. bohemicum, B. bombi, B. crudilactis, B. mongoliense, Bifidobacterium psychraerophilum, and B. subtile taxa (Fig. 5). With the exception of B. psychraerophilum and B. crudilactis, this pathway was displaced by the alternative, non-mevalonate 2-C-methyl-d-erythritol 4-phosphate/1-deoxy-d-xylulose 5-phosphate pathway (MEP/DOXP pathway) for isoprenoid biosynthesis. Interestingly, the intermediate HMB-PP [(E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate] is an activator for human Vγ9/Vδ2 T cells, the major γδ T cell population in peripheral blood (61, 62), playing an important (even if not fully understood) role in the initial training and subsequent regulation of the mucosal immune system.

Pathways for degradation of alcohols (2,3-butanediol, ethanol, and glycerophosphodiester), amines and polyamines (4-aminobutyrate [GABA]), N-acetylglucosamine, and urea as well as allophanate, gluconate, phospholipids, 2-aminoethyl phosphonate, nucleotides, and xylitol are widely distributed among bifidobacterial species (Fig. 5). Notably, only B. actinocoloniiforme and B. bohemicum are predicted to possess a complete citrate degradation pathway and a complete d-glucarate degradation pathway. With respect to nitrogen metabolism, only the genome of B. callitrichos appears to encompass the nitrate reduction VI (assimilatory) pathway, which is predicted to be involved in nitrogen assimilation (63). Interestingly, the genetic locus encompassing the nitrite reductase also includes a gene encoding ferredoxin-NADP reductase, a flavodoxin-encoding gene, and a gene encoding an ABC-type nitrate/nitrite porter. Other intriguing bifidobacterial metabolic properties here identified involved the presence of a complete pathway for degradation of d-glucuronate, one of the main constituents of proteoglycans, which are present only in the genomes of B. asteroides, B. indicum, B. coryneforme, and B. biavatii. Proteoglycans have an important role in the physiology of insect gut since they constitute the peritrophic matrix, a physical barrier that plays a role analogous to that of mucous secretions of the vertebrate digestive tract (64). Thus, the presence of a degradation pathway for d-glucuronate in genomes of bifidobacterial species isolated from honey bees (B. asteroides, B. indicum, and B. coryneforme) and isolated from the gut of the insect-feeding tamarin monkey (B. biavatii) may represent a key example of strict genetic adaptation of bifidobacteria to the gut of insects. The genomes of the species isolated from insect gut such as B. asteroides, B. indicum, B. coryneforme, B. actinocoloniiforme, B. bohemicum, and B. bombi, in addition to B. mongoliense and B. subtile, highlighted the presence of a complete electron transfer chain consisting of four complexes (complex I, NADH dehydrogenase, flavin mononucleotide, and iron-sulfur cluster-containing protein; complex II, succinate dehydrogenase; complex III, cytochrome d oxidase; and complex IV, F₁F₀-ATPase), which suggests that these species have the option of operating a simplified respiratory metabolism (65).

The metabolic potential of Bifidobacterium is complemented by its predicted transport capabilities. In particular, ABC transporters that represent putative sugar uptake systems are present in greater numbers than those that represent predicted amino acid, peptide, and metal uptake systems. Among the detected carbohydrate uptake systems, those predicted to be specific for oligosaccharides and glycosides outnumber transporters for free sugars.

Conclusions.

This report represents an extensive comparative analysis of the genomes of all representative species belonging to the Bifidobacterium genus, revealing a distinct saccharolytic genotype. An extensive gene acquisition trend over the course of evolutionary development of bifidobacteria through HGT events seems to have allowed the enrichment of metabolic traits sustaining the utilization of a vast array of carbohydrates, in terms of both transport and degradation. In the ancestral bifidobacteria that are believed to closely resemble the current members of the B. asteroides phylogenetic group, carbohydrate metabolism is centered on the use of simple sugars commonly identified in plant cells. Furthermore, subsequent specialization of bifidobacterial taxa associated with the mammalian gut seems to have been subject to Darwinian selection that led the acquisition of genetic pathways indispensable for the metabolism of complex carbohydrates found in the mammalian diet.

Lastly, the results of the comparative genomic analyses provided here also indicate that a revision of the taxonomy of the currently distinguished Bifidobacterium species may be necessary, as these analyses revealed very close phylogenetic relatedness of bifidobacterial taxa that are currently considered separate species.