Pig intestinal microbiota were investigated in terms of their functionalities in host health and growth promotion (Lee and Mazmanian, 2010; Brestoff and Artis, 2013). The development of pig intestinal microbiota, particularly the predominance of Lactobacillus spp., is considered a key event for health promotion (Konstantinov et al., 2006; Petri et al., 2010), and this lactobacillal predominance may relate to the significant reduction of Enterobacteriaceae and Clostridium perfringens (Inoue et al., 2005). Such development is affected by the weaning, dietary changes, and host-immune development (Katouli et al., 1997; Mackie et al., 1999; Inoue and Ushida, 2003; Ushida et al., 2008).

In addition to Lactobacillus spp., Bifidobacterium spp. are regarded as health-promoting bacteria in humans, because humans harbor Bifidobacterium as a predominant lactic acid bacteria (Mitsuoka and Kaneuchi, 1977; Matsuki et al., 2004). Especially, human breast-fed infants harbor high levels of Bifidobacterium spp., which may explain their low susceptibility to enteric diseases (Gibson and Wang, 1994; Roger et al., 2010; Guaraldi and Salvatori, 2012). Pigs do carry Bifidobacterium spp. as the primary component of intestinal microbiota, but to a lesser extent than Lactobacillus spp. as determined by both a culture-dependent method (Mikkelsen et al., 2003) and a culture-independent method (Hermann-Bank et al., 2013). We recently estimated the effect of domestication and modern pig feeding on the intestinal microbiota of Suidae by the metagenomic analysis of fecal DNA of various species of Suidae, such as Sus scrofa scrofa, S. s. domesticus, and Potamochoerus porcus (Ushida et al., 2016). In that study, we showed that Bifidobacterium was a predominant member (top 10 primary member) in the gut microbiota of wild boars and, to a lesser extent (top 20 member), in domestic pigs. Since Lactobacillus spp. were at the top 20 level of bacterial genera in wild boars, domestication and a modern feeding system may have induced the predominance of Lactobacillus spp. over Bifidobacterium spp. in pigs. Although the reasons for such a substitution are unrevealed, genomic analyses of Suidae-associated Bifidobacterium may detect the genetic changes associated with the adaptation to domesticated hosts and/or those grown in an industrialized pig-production system. For such study, we need to identify and isolate the Suidae-associated Bifidobacterium carried by both wild boars and domestic pigs.

Many strains of Lactobacillus spp. and Bifidobacterium spp. have been sequenced so far (Kant et al., 2011; Drissi et al., 2014; Lugli et al., 2014). But B. thermacidophilum intraspecies comparison has not been performed yet.

In this study, we analyzed the genomes of seven strains of B. thermacidophilum, which were isolated from pigs and wild boars. We found several fundamental differences in their draft genome sequences between the isolates from domestic pigs and those from wild boars. Although the number of strains tested here was limited, the difference in the draft genome sequences suggested the adaptation of Suidae-associated Bifidobacterium to the conditions in wild and artificial feeding.

Antimicrobial resistance (AMR) is now well-recognized as the primary threat to public health (O'Neill, 2014), and animal agriculture is considered an important industrial sector in which AMR continuously emerges due to antimicrobial use that promotes clinical and prophylactic growth (Woolhouse and Ward, 2013). Accordingly, the propagation of AMR of animal origin through the food chain has been routinely discussed. Probiotics have been regarded as a potential alternative for prophylactic drugs in animal agriculture (de Vrese and Schrezenmeir, 2008). However, since human-associated and probiotic strains of Bifidobacterium spp. are resistant to a range of antibiotics, including tetracycline (Moubareck et al., 2005), the transmission of AMR from probiotic strains to other bacterial members of intestinal microbiota could occur (Van Hoek et al., 2008). In this study, in addition to the biased distribution of genes responsible for carbohydrate and nitrogen metabolism in seven B. themacidophilum strains, resistance to antimicrobials was investigated.

We isolated seven strains of Bifidobacterium from pigs and wild boars. These isolates were identified as B. thermacidophilum according to their 16S rRNA sequences. Bifidobacterium is not a primary lactic acid bacteria in the pigs (Mitsuoka and Kaneuchi, 1977; Ushida et al., 2016). In the literature, B. globosum, B. pseudolongum, B. thermophilum, B. boum, and B. choerinum were detected from pig feces, and B. boum was a predominant Bifidobacterium in piglets (Mikkelsen et al., 2003). However, these species have a wide host range and are not limited to pigs (Scardovi et al., 1979; Yaeshima et al., 1992; Jans et al., 2013). B. thermacidophilum has also been detected in pigs (Mølbak et al., 2007). Marti et al. (2009) demonstrated that B. thermacidophilum is a pig-specific Bifidobacterium and can be used as a specific marker for pig-manure contamination. Accordingly, B. thermacidophilum was regarded as a Sus-specific Bifidobacterium species in this study, although the type strain of B. thermacidophilum subsp. thermacidophilum was isolated from waste water from a soybean curd factory in Beijing (Dong et al., 2000).

According to the 16S rRNA phylogenetic tree, pig isolates 45 and 53 were close to RBL67. These three strains and another pig isolate, 194, were close to both B. thermacidophilum subsp. porcinum and B. thermophilum JCM 1207^T.

Since B. thermacidophilum is a recognized species of the very close relative of B. thermophilum (Dong et al., 2000), their close relationship is not surprising.

However, at least the phylogenetic position of B. thermophilum RBL67, a human infant isolate, is confusing and should be reclassified. In fact, the low ANI value (88%) shown by the RBL67 genome with that of B. thermophilum JCM 1207^T clearly indicates the distant phylogenetic relationship between them. Kheadr et al. (2007) listed this particular strain under the D group of B. thermacidophilum subsp. porcinum. We constructed phylogenetic trees based on draft genome sequences in this study: a maximum-likelihood phylogenetic tree of the amino acid sequence of the single-copy core CDSs (Figure 1B), a maximum-parsimony tree based on all of the SNPs (Figure 1C), and a pan-genome dendrogram (Figure 1D). All of these trees showed the same topology and several inconsistencies with a tree constructed by 16S rRNA sequences due to the insufficient resolution of 16S rRNA sequence alignment at subspecies level as indicated by Lugli et al. (2014). In draft genome-based phylogenetic trees, pig isolate 194 and B. thermophilum RBL67 were always placed in close proximity, and their relatedness to B. thermacidophilum subsp. porcinum JCM 16945 was suggested. Pig isolates 45 and 53 were always close to B. thermacidophilum subsp. thermacidophilum JCM 11165 rather than to B. thermacidophilum subsp. porcinum JCM 16945. Four wild boar isolates, 195, 196, 197, and 198, were always in the independent clade in these trees. The ANI values of around 93%, which was smaller than the 95% criteria for species identification, were shown by strains 195, 196, 197, and 198 with both B. thermacidophilum subsp. porcinum JCM 16945 and B. thermacidophilum subsp. thermacidophilum JCM 11165. This may suggest that these wild boar isolates could be reclassified in the future at the species level. In this study, we constructed a kSNP tree in addition to a core CDSs based analysis and a pan-genomic presence/absence matrix of the core genome to confirm the phylogenetic relationship of seven strains. In fact, the kSNP tree accurately represented the phylogenetic relationship of 58 Streptococcus suis strains in a previous study (Okura et al., 2017). Since the kSNP tree is based on k-mer selected as the optimal value predicted by the kSNP-associated Kchooser script, it does not consider the effect of recombination. The involvement of multicopy genes in the phylogenetic analyses may suggest different phylogenetic relationships according to which multicopy genes are selected. Therefore, we selected the single-copy genes for the phylogenetic analyses. Pseudogenes were also removed from the present analyses because individual pseudogenes had been under a different level of selection pressure. The concordant feature of pan-genomic analyses with the kSNP tree can represent the robust phylogenetic relationship of seven strains. Therefore, concordant results shown by phylogenetic trees (Figures 1B,C) and a dendrogram (Figure 1D) are not weak evidence for the phylogenetic relationship of these seven strains.

The separation of pig isolates from wild boar isolates depends on the biased distribution of genes (Tables S2, S3). This biased distribution may represent the adaptation of B. thermacidophilum to the environmental condition of the host intestine.

Obviously, the present genomic analyses were based on draft sequences, in which we can discuss the genes actually present, not absent, in the draft genome sequences. It is possible that many of the essential genes presumably present in the genome were missing. Therefore, we still need to complete the genome sequencing to confirm the present findings and following discussion. In addition, a recent study indicated that a small number of COGs for the carbohydrate metabolism represented in the core genome of Bifidobacterium, although the carbohydrate metabolism is the major functional family of COGs (Milani et al., 2014). This suggests the importance of analyses on paralogous genes for complete understanding of the adaptation of Bifidobacterium to the host environment. Although such limitations, several particular differences in gene distributions between two groups of isolates (the domestic pig cluster and the wild boar cluster) may indicate the relatedness of gene distribution in the genome to the feeding condition of the host. For example, the biased presence of fiber-degrading enzymes, endoglucanase, chitinase, pullulanase, and beta-mannosidase, is interesting because wild boars are omnivores whose diet is predominantly plant materials (Ballari and Barrios-Garcia, 2014). Cellulose and xylan comprise the major part of plant carbohydrates of the above-ground part, which comprises 70% of the plant materials consumed by wild boars (Ballari and Barrios-Garcia, 2014). Adaptation of these wild boar isolates to a fibrous diet was, therefore, suggested in comparison with isolates from domestic pigs that rely more on cereal starch than plant fiber. In addition to fiber-degrading enzymes, the possible presence of chitinase may reflect the dietary habit of wild boars, because fungi are the one of their major food items (Fournier-Chambrillon et al., 1995; Hohmann and Huckschlag, 2005). The unique presence of chitinase in wild boar isolates again suggests adaptation of the host food habit. Glucosidases and genes for a nucleic acid-base metabolism, such as cytosine deaminase, xanthine, and other nucleic acid-base permeases, were found to be genes associated with wild boar isolates, which suggests that a nitrogen-scavenging capacity is essential for the survival of intestinal bacteria under a protein-limited intestinal environment. This may reflect the protein-limited feeding condition of wild boars normally observed in the winter season (Genov, 1981). These particular profiles in the carbohydrate metabolism of wild boar isolates were shared by the type strain of B. thermacidophilum subsp. thermacidophilum and not by the pig-derived type strain B. thermacidophilum subsp. porcinum, which supports the adaptation of this species to the dietary habit of the host. Genes for iron metabolism were also distributed to wild boar isolates. These genes were not shared by pig isolates and the two type species of B. thermacidophilum. Since some of these genes were detected in a pig-derived B. thermophilum JCM1207 and human infant–derived B. thermophilum RBL67, it is inconclusive whether these iron metabolisms were related to the wild boar feeding habit.

The abundance of Type VI pilin (T4P) is also distinctive for wild boar isolates. T4P is generally displayed on the surface of Gram-negative bacteria such as Pseudomonas aeruginosa (Bradley, 1972) but also has been identified in several Gram-positive bacteria (Melville and Craig, 2013). DNA uptake was one of the functions of T4P system (Claverys et al., 2006). This also suggests the adaptation of wild boar isolates to the nitrogen-limiting condition that accelerates the nucleic acid metabolism in the GIT.

The mobile genetic elements associated genes were found more abundant in the pig isolates than in the wild boar isolates. This tendency was also found in the pig-derived type strain B. thermacidophilum subsp. porcinum (20 mobile element proteins) as compared to B. thermacidophilum subsp. thermacidophilum (10 mobile element proteins). Genes for the site-specific recombinase, integrase, transposase, and TrsK-like protein were detected in pig isolates, which suggests the capacity of a horizontal gene transfer. As a a barrier to horizontal transfer and acquisition of mobile genetic elements, Type 1 restriction enzymes (EC 3. 1. 21. 3. and EC 2. 1. 1. 72.) were deployed in pig isolates, but these genes were not shared by the type strains.

The CARD and SEED suggested the variety of drug-resistance genes. The genes for virulence and defense, such as the tetracycline resistance gene, tetW, and mercuric ion reductase (EC 1. 16. 1. 1.), and the lincomycin resistance gene were found to be pig-associated genes (under the intensive feeding system). A tRNA synthetase gene, ileS, can be considered to be involved in the natural resistance of B. thermacidophilum (Serafini et al., 2011). The similar tRNA synthetase gene, alaS, which is involved in novobiocin resistance, was suggested for all isolated strains, although B. animalis subsp. lactis and B. longum were susceptible to this antibiotic (D'Aimmo et al., 2007).

Disk diffusion tests for presumably acquired resistance showed a low level of concordance with the CARD and SEED predictions. Several fluoroquinolone resistance genes were suggested for all strains, including three type strains, but any resistance to ofloxacin was not proven. Although the macrolide resistance gene, desR, was suggested for wild boar isolates, they are all susceptible to erythromycin. Colistin resistance was also conferred for the strains harboring PmrE, but colistin resistance apparently was not related to the presence of PmrE. Fosfomycin resistance was conferred by the presence of murA, but the fosfomycin resistance was not concordant to the presence of murA. Such low level of concordance may be related to the lower level of similarities (≤ 60%), which suggests the detection of non-functional genes. In case of gyrA gene, SNPs were also suggested by CARD, which indicated the mutation within the gene. The concordance between the prediction and the disk susceptibility test was the tetracycline resistance. Therefore, we focused on tetracycline resistance in this study, although the tetracycline resistance in Bifidobacterium is well-known (Moubareck et al., 2005). The biased distribution of tetW to strains 45 and 53, from a farm where tetracycline is heavily used, is noteworthy, because the isolates from a pig under a drug-free feeding condition and from wild boars were free of tetW and susceptible to tetracycline. While Bifidobacterium possesses a naturally occurring resistance to aminoglycoside antibiotics such as gentamycin (Moubareck et al., 2005; Ammor et al., 2008), acquired resistance was suggested for tetracycline and macrolide among the known antimicrobials (Moubareck et al., 2005; Van Hoek et al., 2008, 2011). Among the presently studied seven isolated strains, only two strains, which were isolated from pigs in a commercial pig farm with routine tetracycline use, were found to possess the tetW gene. In the literature, strains of B. adolescentis, B. animalis lactis, and B. pseudocatenulatum, mostly human-associated species, showed relatively high resistance (≥32 mg/L) to tetracycline due to the acquisition of tetW (Masco et al., 2006; Ammor et al., 2008). B. thermophilum strains of a pig or cattle origin also showed very high resistance to aminoglycoside but with varying degrees of resistance to tetracycline (Mayrhofer et al., 2011).

The tetW gene is now widely distributed to the genus Bifidobacterium in humans and livestock (Masco et al., 2006; Aires et al., 2007). The tetW gene product detected in the present study showed high similarity (>99%) to many tetW sequences (>50 sequences) deposited in the public database (BLAST search was executed on March 1, 2016, through the DDBJ website). Because of the high sequence similarly between the tetW genes identified so far, it has been suggested that tetW has recently appeared and rapidly spread among various bacteria (Scott et al., 2000). In fact, the tetW genes previously detected in B. longum and Bifidobacterium sp. ISO3519 are most closely related to the tetW found in other bacterial species, such as Megasphaera elsdenii, Roseburia hominis, Mitsuokella multocida, and Butyrivibrio fibrisolvens, inhabiting animals and humans (Aminov and Mackie, 2007).

A serine recombinase is often associated with integrative genetic elements (Bibb et al., 2005). Although attL/R-like direct repeat sequences could not be found around this tetW–serine recombinase gene cassette, this cassette may represent a mobile genetic element that mediates the horizontal transfer of tetW. The GC content of this cassette was 53.8%, which is lower than those in the flanking 4,000 bps region (59.4%). Such a difference in GC content also suggests the horizontal transfer of tetW.

The presence of this serine recombinase in Bifidobacterium was not known before, but this recombinase showed high similarity with those of T. succinifaciens DSM2489 and L. amylovorus 30SC, both of which were isolated from pigs (Cwyk and Canale-Parola, 1979; Oh et al., 2011). Although T. succinifaciens DSM 2489 also contains the tetW gene, it is 20 kb apart from the serine recombinase gene. Therefore, the relationship of these two genes in T. succinifaciens DSM 2489 is unknown.

A comparison of the tetW cassette–flanking regions of strains 45 and 53 with the analogous regions of tetW-negative strains revealed that the cassette has been inserted in a gene encoding an arabinose efflux permease family protein, and the 3′-region of this permease protein gene was predicted to be an MDR-permease by SEED (Figure 2, Figure S1).

The presently detected two tetW genes were identical. Moreover, the serine recombinase was also identical for both pig isolates, with a minor difference in an intergenic spacer sequence. The presence of a common origin for the tetW gene cassette in these two isolates was suggested. Since strains 45 and 53 are phylogenetically very close, it is, therefore, plausible that these two strains are progeny of a strain of B. thermacidophilum that horizontally received this tetW cassette from an unknown origin sometime in the past.

Bifidobacterium is the potential target of probiotics (Sanders, 2008; Amagase, 2011), since the transferable drug resistance associated with this bacterium is a serious concern for the safety of their probiotic use (Gueimonde et al., 2013). Tetracycline is still frequently used in animal agriculture, particularly in pig production in many countries, including Japan (Itoh, 2014; Van Rennings et al., 2015). Therefore, there is still a considerable risk of tetracycline resistance dissemination into human society through the food supply (Barbosa et al., 1999; Patterson et al., 2007; Kazimierczak et al., 2009). In this circumstance, B. thermophilum that is detected in both pigs and humans may become a more direct threat to our society than the presently analyzed B. thermacidophilum. In fact, B. thermophilum JCM 1207^T was proven to harbor a functional tetW. However, B. thermophilum RBL67, which was isolated from a human infant, was actually very close to our pig isolates, B. thermacidophilum strains 45 and 53, according to phylogenetic analyses on a draft genome basis. This suggests that a certain B. thermacidophilum can become a member of the human intestinal microbiota. The possible risk of tetracycline resistance dissemination into human society by pig-origin B. thermacidophilum cannot be ruled out.

One of the important differences between pig-derived isolates (under the intensive feeding system) and wild boar–derived isolates was that the former possessed a horizontally transferred tetW gene cassette with a putative serine recombinase, though the mobility of this tetW cassette is yet to be determined. Although B. thermacidophilum can be recognized as a pig-specific Bifidobacterium species, B. thermacidophilum subsp. porcinum and its close relative, B. thermophilum, have been recognized in both humans and pigs (Kheadr et al., 2007; Mayrhofer et al., 2007; Von Ah et al., 2007). This suggests a potential risk for the transmission of tetracycline resistance to human intestinal Bifidobacterium through the food chain. It is clear that the routine use of tetracycline in pig farms supports the continuous presence (or selection) of tetW-positive B. thermacidophilum, because a strain from a sow in our experimental farm, where no antimicrobials were fed, was free from tetW resistance gene, together with those derived from wild boars, which were likely to feed on antimicrobial-free natural diets.

ST contributed to the design of the study, isolation of strains, and physiological analyses on bacteria and interpreted the data and wrote the initial draft of the manuscript. FM contributed to the Bioinformatic analysis and interpretation of data, and assisted in the preparation of the manuscript. YO and AT contributed to the sequencing by Next-generation sequencer and interpretation of the data. TH and MO contributed the critical revision of the articles for important intellectual content respectively for genome analyses and for phylogenetic analyses. KU contributed to the Conception and design of the study, physiological analyses, and data interpretation. The final version of the manuscript was approved by all authors.