The phylum Bacteroidetes is a very diverse bacterial phylum, the name of which changed several times over the past years. It is also known as the Cytophaga–Flexibacter–Bacteroides (CFB) group, an appellation that reflects the diversity of organisms found in this phylogenetic group (Woese, 1987; Woese et al., 1990). According to the Bergey's Manual of Systematic Bacteriology (Bergey's, 2011), the Bacteroidetes phylum comprises four classes: Bacteroidia, Flavobacteria, Sphingobacteria, and Cytophagia, representing around 7000 different species (NCBI, October 2010). The largest class is the Flavobacteria, grouping together around four times more species than the three others (Table 1). These bacteria are all Gram negative, cover a mixture of physiological types, from strictly anaerobic Bacteroides to strictly aerobic Flavobacteria. They are non-motile, flagellated, or move by gliding.

Members of the phylum Bacteroidetes have colonized many different ecological niches, including soil, ocean, freshwater, and the gastrointestinal tract (GIT) of animals, where they display various biological functions. In particular, they are well known degraders of polymeric organic matter. This review describes current knowledge on the role and mechanisms of polysaccharide degradation by Bacteroidetes in their respective habitats. We emphasize the features shared by members of the phylum that allow this functional -specialization in various environments. We address the links between these different microbial communities through food consumption, which raise the question of the evolution of gut microbes.

As a phylum, and especially due to their versatility in habitats, Bacteroidetes have access to an amazing diversity of carbon sources. Indeed, the chemical diversity of polysaccharides largely outnumbers the possibility for protein folds – it has been calculated that there are 1.05 × 10¹² possible linear and branched forms of a single hexasaccharide (Laine, 1994). Moreover, these structural variations have been harnessed by living organisms to fulfill very different roles: e.g., structural, storage, specific signaling, -specific -recognition, host–pathogen interactions to name but a few (Carpita and Gibeaut, 1993; Graham et al., 2000; Stahl and Bishop, 2000). Consequently, carbohydrates account for around 75% of the biomass on Earth, a natural resource that was not lost on competing organisms that were developing their own strategies to utilize this chemical energy for their own survival.

But even more importantly several classes of polysaccharides are niche specific. While cell walls are a characteristic feature of all plants, they are not exclusive to plants, with most bacterial and algal cells as well as all fungal cells also being surrounded by extracellular, macromolecular barriers (extracellular matrix or ECM). The macromolecular composition, however, is characteristically different among the major evolutionary lineages of the living world, linking specific life style or nutritional habits to specifically encountered biopolymers. A vivid example is provided by the polysaccharides of the marine environment that are typically and to a large majority sulfated (carrageenans and fucans) or highly ionic (alginates) and unique to this particular habitat (Michel et al., 2010a,b; Popper et al., 2011). In contrast, the basic polysaccharide components of plant cell walls are cellulose and hemicellulose (pectins, xylans, mannans, xyloglucans, etc.), whereas fungal cell walls primarily consist of chitin (Niklas, 2004). In metazoa, the ECM will predominantly consist of chondroitin or dermatan, which are essentially made of sulfated polysaccharides referred to as glycosaminoglycans (GAGs; Sugahara and Kitagawa, 2002) that are interconnected by fibrillar proteins (collagens). Other sources of carbohydrates in animals are glycosylation sites, such as mucin that contain a high proportion of sialic acid in addition to GAGs (Raman et al., 2005).

Reflecting this chemical diversity of the substrate, glycosidases, the enzymes responsible for the breakdown of di-, oligo-, and polysaccharides, as well as glycoconjugates, are ubiquitous through all domains of life (Turnbaugh et al., 2010). Carbohydrate processing enzymes (CAZymes), including glycosidases and glycosyltransferases (the enzymes which transfer saccharides to other saccharide moieties, small molecules, lipids, or proteins), constitute between 1 and 3% of the genome of most organisms (Davies et al., 2005). Noteworthy, the genomes of Bacteroidetes species have revealed that they are champions with respect to the diversity and number of CAZymes they contain, reflecting the molecular strategies evolved by this microbial community to differentiate, capture, and degrade complex glycans. Consequently and as a result of this ability to degrade host and plant glycans, cultured (environmental or gut) species are often used to isolate specific enzymes for polysaccharide degradation (Berg et al., 1980; Tierny et al., 1994; Bernardet and Nakagawa, 2006; Reichenbach, 2006).

The first sequenced genome of a Bacteroidetes representative was published in 2002 for the human symbiont B. thetaiotaomicron (Xu et al., 2003). Since then, many sequencing projects have been conducted to increase the genomic knowledge on this phylum. To date, 33 Bacteroidetes genomes are complete, publicly available and published (Table 2). Many others are in a draft state or have not yet been published (total of 125 sequences censed on NCBI). During the last 10 years, sequencing efforts have indiscriminately concerned environmental, pathogen, and symbiotic/commensal species with the aim to better understand their biological -functions, including their capacity to interact with their habitats. This data allows comparing the different enzymatic capabilities of various genera, which sheds new light on the specialization of Bacteroidetes toward degradation of organic matter. A striking common feature revealed by this comparative genomic approach is the trend of Bacteroidetes genomes to encode many polymer-degrading enzymes, acting either on proteins or carbohydrates. The census of Carbohydrate-Active enzymes (CAZ Ymes) in the CAZY database¹ (Cantarel et al., 2009) eases the comparison of the number of glycosylhydrolases (GH) and polysaccharide lyases (PL) in an increasing number of sequenced species. In each of the four classes of the Bacteroidetes phylum, there are examples of CAZYme-enriched species. In Bacteroidia, the proteomes of B. thetaiotaomicron, B. fragilis, and P. ruminicola comprise 272, 137, and 130 GH and PL respectively (Xu et al., 2003; Kuwahara et al., 2004), much more than other members of the gut microbiota, or outside the Bacteroidetes phylum, such as Clostridium perfringens (57 GH and PL) and Bifidobacterium longum (49 GH). The same is true for Flavobacteria [e.g., G. forsetii with 48 GH/PL (Bauer et al., 2006), Z. profunda with 120 GH (Qin et al., 2010), F. johnsoniae with 152 GH/PL (McBride et al., 2009)], Cytophagia [e.g., S. linguale with 151 GH/PL (Lail et al., 2010), D. fermentans with 98 GH/PL (Lang et al., 2009), and Sphingobacteria, e.g., C. pinensis with 184 and P. heparinus with 163 GH/PL, respectively (Han et al., 2009; Del Rio et al., 2010)]. In some cases, the prediction of CAZYme-encoding genes in newly sequenced organisms can even unveil unexpected catabolic capabilities toward specific substrates (McBride et al., 2009). Moreover, the predicted enzymatic battery of a bacterial species will help characterize its natural habitat (i.e., the available substrates) and its ecological function in organic matter recycling. Recently, this feature has been used for genome and habitat comparison, linking the number and occurrence of specific CAZyme-families to the environmental niche (Pope et al., 2010; Purushe et al., 2010). With this respect, the presence of a vast majority of exo-acting enzymes in the genome of B. thetaiotaomicron, suggests that the organism is able to use the saccharide decorations appended to the backbone of structural polysaccharides and glycoproteins (Xu et al., 2003). Another recent study elegantly demonstrates that this particularly evident expansion in exo-GHs of family GH92 enzymes (23 members) is indeed related to the α-mannosides present in the N-glycans of host and dietary glycoproteins (Zhu et al., 2010).

However, one has to keep in mind that a rough analysis of the number of degradation enzymes in general is insufficient; one needs to go down to the enzymatic sub-family to infer a putative metabolism. Indeed, the CAZyme classification based on sequence -similarity¹ has the consequence that gene families group together enzymes with widely different substrate or product specificities (Henrissat, 1991). Therefore, to derive knowledge useful for subsequent functional predictions, phylogenetic analyses defining subgroups that contain biochemically characterized representatives are needed to perform unambiguous assignments (Turnbaugh et al., 2010).

In spite of the general trend of Bacteroidetes to possess numerous degradation enzymes, there are several noteworthy exceptions. The fish pathogen F. psychrophilum genome harbors only 13 proteases, i.e., 4.5 proteases per megabase (Mb) and 3 GH per Mb (Duchaud et al., 2007). This is far less than its closely related soil-associated cousin F. johnsoniae, which possesses 20.5 proteases and 23 GH per Mb and has a genome more than twice as large. The reduced number of degradation enzymes in the fish pathogen can be explained by its dedication to the infection of animal tissues. For such a species, the relative restricted diversity of substrates (compared to a soil-associated species) would diminish the need of multiple families of hydrolases. A second, similar example is P. gingivalis that has a relatively lower number of hydrolases (only 24 GHs) compared to other members of the Bacteroidia class (Nelson et al., 2003). Again, the most plausible explanation is the high speciation of this species to dental plaque degradation that results in the restricted diversity of substrates utilized by this bacterium.

Some marine representatives of the Flavobacteria class seem to alternate between two life strategies depending on the abundance of carbon sources. This is notably the case of Polaribacter dokdonensis (strain MED152), Leeuwenhoekiella blandensis (strain MED217^T), and Dokdonia donghaensis (strain MED134). On the one hand, commonly with other Bacteroidetes, they are very well equipped to attach to surfaces and depolymerize organic matter. For example, P. dokdonensis and D. donghaensis genomes encode many enzymes to degrade proteins (93 and 120 peptidases, respectively) and polysaccharides (30 and 22 GH, respectively; Gonzalez et al., 2008; Kirchman, 2008; Woyke et al., 2009). On the other hand, when polymeric substrates become scarce, these species switch to a free-living lifestyle adapted to carbon-poor environments. Genome analysis showed the presence of proteorhodopsin, a light-dependent H⁺ pump that can drive ATP synthesis (Beja et al., 2000). This protein allows the phototrophic production of sufficient energy to maintain the population growth when the concentration of organic carbon decreases (Gomez-Consarnau et al., 2007). Additionally, P. dokdonensis is enriched in enzymes involved in anaplerotic reactions, and assimilates CO₂ faster in light conditions than in the dark (Gonzalez et al., 2008). Altogether, these results suggest that marine Bacteroidetes may cope as well with feast and famine, and complement the understanding of their role in carbon cycles (DeLong and Beja, 2010).

Beyond the prediction of numerous polysaccharide-degrading enzymes, two large paralogous families of proteins have been found in Bacteroidetes genomes, which likely participate in polysaccharide uptake. These include homologs of the outer membrane proteins SusC and SusD from B. thetaiotaomicron, which are involved in starch utilization. Seminal work of the Salyers’ group showed that this human gut symbiont degrades starch via a dedicated starch utilization system (Sus) with several proteins acting in coordination to sense, bind, and hydrolyze the substrate (Anderson and Salyers, 1989a,b; Shipman et al., 2000). The genes encoding these proteins cluster on the bacterial chromosome into typical polysaccharide utilization loci (PUL).

In B. thetaiotaomicron, the Sus locus comprises eight genes, susRABCDEFG (Martens et al., 2009). SusR is an inner membrane regulatory protein which activates the transcription of the other genes in the presence of maltose or starch (D'Elia and Salyers, 1996). SusC, SusD, SusE, and SusF are outer membrane proteins involved in the binding of the polysaccharide (Shipman et al., 2000). In fact, analyses of mutant strains have shown that SusC and SusD account together for 70% of the starch-binding capabilities of the wild type (Reeves et al., 1997). Surface-bound starch is hydrolyzed by the outer membrane α-amylase SusG, which acts endolytically and releases oligosaccharides larger than maltotriose (Shipman et al., 1999; Martens et al., 2009). These degradation products are then channeled to the periplasm through the TonB-dependent receptor, β-barrel-type SusC, where they are further cleaved by the neopullulanase SusA and α-glucosidase SusB. The atomic structures of several protein members have been resolved, namely the starch-binding SusD (Koropatkin et al., 2008), the α-glucosidase SusB (Kitamura et al., 2008), and the α-amylase SusG (Koropatkin and Smith, 2010). The Sus locus is organized into two transcriptional units under the control of SusR, one containing susA and the other containing susB to susG (D'Elia and Salyers, 1996; Reeves et al., 1997). This allows a co-regulation of the PUL (Anderson and Salyers, 1989b).

The characterization of this first starch-specific PUL was -followed by the discovery of numerous PULs in Bacteroidetes. susC-like and susD-like genes are strikingly frequent in Bacteroidetes genomes, often appearing in tandem and as the central units of substrate-specific PULs. B. thetaiotaomicron possesses 107 paralogs of susC, of which 101 are paired to a susD-like gene. 62 of these pairs are part of larger clusters, together with polysaccharide-degrading enzymes (Xu et al., 2003). In addition, some of these PULs comprise enzymes targeting glycan decorations, such as sulfatases or acetyl esterases. Thus, depending on the specificity of the predicted enzymes, one can infer the favorite substrate(s) of a given PUL. In total, PULs represent 18% of the genome of B. thetaiotaomicron (Martens et al., 2008). The closely related B. fragilis, B. vulgatus, and Parabacteroides distasonis also possess numerous PULs (Kuwahara et al., 2004; Xu et al., 2007). These PULs likely favor the success of Bacteroides spp. in the uptake of dietary and host-derived polysaccharides in the highly competitive gut habitat, and may explain their evolution as symbionts.

Interestingly, environmental species also harbor plenty of specific PULs. The annotation of the F. johnsoniae genome revealed 42 pairs of susC-like and susD-like genes, among which many were associated with CAZymes (McBride et al., 2009). The authors were notably able to predict PULs targeting starch (homologous the B. thetaiotaomicron Sus locus), chitin, and hemicelluloses. The genome of the marine Flavobacteria G. forsetii encodes 40 paralogs of SusC, and 14 clusters of susCD-like genes were detected (Bauer et al., 2006), often in the vicinity of CAZyme genes. This suggests that environmental Bacteroidetes as well as their gut-associated cousins use a unique and similar strategy to bind and degrade polymeric organic matter. Indeed, SusD homologs are only found in Bacteroidetes representatives. Thus, we speculate that the appearance of PULs including a susD-like gene in the ancestral Bacteroidetes could have at least partly driven the emergence of the phylum, and allowed its evolution as a group specialized in carbohydrate degradation.

The size of Bacteroidetes genomes varies considerably between species (Table 2). Among published sequencing projects, Chitinophaga pinensis has the largest genome (9.1 Mb) whereas Candidatus Sulcia muelleri has the smallest (0.3 Mb). This great discrepancy can be at least partly associated with the different ecological niches colonized and the biological functions played by Bacteroidetes (Figure 1). Obligate intracellular symbionts have a more reduced genome size, due to their peculiar lifestyle. These species have evolved through successive inactivation and loss of genes, affecting virtually every cellular process. The possible causes of this genome reduction are multiple, including the unusual stability and metabolic richness of the cytoplasmic compartment they inhabit (McCutcheon and Moran, 2010). Pathogenic Bacteroidetes, such as P. gingivalis and F. psychrophilum, have a small genome typically around 2 Mb. This can be linked to the dedication of their metabolic capabilities toward the infection of specific sites. Living in complex habitats and metabolizing a lot of different substrates, environmental, and intestinal species tend to have larger genomes (Figure 1), -correlating with their broader catabolic capabilities. This is also the case of the opportunistic pathogen B. fragilis, which is part of the -normal human gut microbiota but can cause infections at many other sites. To date, marine representatives harbor smaller genomes than their intestinal or terrestrial counterparts, but this will certainly progress as the number of complete genome sequences increases. For example, our group has recently annotated the genome of the marine Flavobacteria Zobellia galactanivorans. It comprises 5.5 Mb and encodes 4738 proteins, representing one of the largest genomes for a marine Bacteroidetes (Barbeyron et al., unpublished data). Interestingly, the proteorhodopsin-containing P. dokdonensis and D. donghaensis possess the smallest proteome among environmental species (2646 and 2284 proteins, respectively). Thus, the genome size fits the ecological niche. Big genomes increase the metabolic capacities, and hence broaden the spectrum of potential substrates for bacteria living in complex environments. In more stable habitats, bacteria tend to specialize toward specific functions and harbor smaller genomes. This raises the question of the nature of the ancestral Bacteroidetes genome. Indeed, the present variations in genome size could be due either to massive loss of genes from a large ancestral genome, or to successive acquisitions completing a small genome with genes representing a selective advantage.

In this respect, the sequencing era unraveled the plasticity of Bacteroidetes genomes, which evolved, and probably still evolve, through dynamic processes. The outcome of this plasticity reflects in the rapid deterioration of the global synteny between evolutionary-related species living in the same environment, as revealed for gut and rumen Bacteroidetes (Xu et al., 2007; Purushe et al., 2010). Their evolution is driven by highly frequent genetic rearrangements, gene duplications, and lateral gene transfers (LGT) between species. Genome analysis of two B. fragilis strains revealed extensive DNA inversions affecting the expression patterns of several genes (Kuwahara et al., 2004; Cerdeno-Tarraga et al., 2005). These events control the antigenic composition of bacterial surface structures and likely help B. fragilis evading the immune system and colonize novel sites. Notably, the expression of 20 SusC-like proteins, most of them coupled with SusD homologs, is regulated through DNA inversions (Kuwahara et al., 2004). This may participate in the cell adaptation to degrade specific polysaccharides found at the infection sites. Bacteroidetes evolution is also characterized by frequent gene duplications and further divergence in sequence and function, leading to considerably expanded paralogous groups. Notably, GH and SusC/SusD-like proteins are amongst the largest paralogous families in several sequenced Bacteroidetes (Xu et al., 2003, 2007; Bauer et al., 2006). In addition to this intra-strain plasticity, Bacteroidetes genomes evolve through inter-species exchange of genetic material (Thomas and Nielsen, 2005). Using a phylogenetic approach, Xu et al. (2007) showed that around 5.5% of the genes in gut Bacteroidetes genomes were laterally acquired from non gut-associated bacteria, among which glycosyltransferases (GT) where significantly over-represented. These LGT events could partly explain the niche specialization of different species. The authors suggest that acquisition of new genes from outside the gut brought novel metabolic pathways to intestinal Bacteroidetes and broaden the spectrum of digestible substrates. Furthermore, it has been shown that the convergence of GT and GH repertoires in gut Bacteroidetes sharing the same habitat is largely due to massive LGT rather than gene duplications (Lozupone et al., 2008). Conjugative LGT events are also demonstrated to be responsible for antibiotic resistance spreading in natural communities of gut Bacteroides (Shoemaker et al., 2001). The exchange of genetic material is not necessarily restricted to closely related species and can overcome phylogenetic barriers. Indeed, glyceraldehyde-3-phosphate dehydrogenase genes have been horizontally transferred from a β-proteobacteria to a Bacteroidetes (Figge et al., 1999). Another example is the transfer of genes between an Archeae and the hyperhalophilic Sphingobacteria S. ruber (Mongodin et al., 2005).

Taken together, recent analyses of Bacteroidetes genome sequences have shown that: (i) there is a gradation in the size of the genomes correlated with the functional specialization; (ii) genomes can undergo massive reorganizations; (iii) highly frequent LGT events allow spreading of novel metabolic capabilities inside Bacteroidetes populations. As already mentioned, intestinal Bacteroidetes are specialized in the degradation of plant-derived polymers, a feature shared with environmental relatives. In the last part of this review, we will discuss the potential connections between these two communities that are not prima facie obviously interacting.

Several studies have shown that the diet strongly influences the intestinal microbiota. Early research focused on the comparison of fecal microbes retrieved from individuals with different nutritional habits. Benno et al. (1986) showed significant variations in the cultivable microbiota of rural Japanese and urban Canadians. They proposed that this discrepancy relates to the contrasted diet of the two populations. Similarly, in a rat model transplanted with human microbiota, the consumption of resistant starch changed the bacterial composition compared to a sucrose diet (Silvi et al., 1999). Yet, the use of cultivable bacterial counts is a limited method (Amann et al., 1995) and no statistical difference was found when comparing the fecal microbiota of strictly vegetarians and individuals consuming a general diet (Goldberg et al., 1977). The development of culture-independent techniques to assess bacterial abundance and diversity helped testing the influence of diet on the GIT microbiota. In a mouse model reproducing the human intestinal microbiome, the bacterial community composition and the representation of metabolic pathways was strongly dependant on the nature of the diet (Turnbaugh et al., 2009). The proportion of Bacteroidetes representatives decreased drastically when animals were switched from a chow containing low levels of fat and high level of plant polysaccharides to a Western diet (high fat, high sugar). Feeding on the diet rich in plant polysaccharides resulted in an enriched set of pathways including N-glycan, glycosaminoglycan, and starch degradation that are typical for Bacteroidetes (Turnbaugh et al., 2009). An independent study on a murine model showed that a high-fat diet was associated with a decrease in more than 30 lineages within the Bacteroidetes phylum, including in the Bacteroidaceae, Prevotellaceae, and Rikenellaceae families (Hildebrandt et al., 2009). Recent studies have investigated the diet impact on the human gut microbiota. The consumption of chemically modified resistant starch (RS4) instead of normal, digestible starch led to a shift in the bacterial community (Martinez et al., 2010). Even if results varied substantially between the 10 considered subjects, RS4 consumption was notably followed by enrichment in Bacteroidetes, among which Parabacteroides distasonis increased sevenfold. The observed changes were completely reversible within 1 week, demonstrating the high population dynamics (Martinez et al., 2010). In a recent comparison of the fecal microbiota of children from Burkina Faso, and Italy, De Filippo et al. (2010) showed a significant difference in the community composition. African children showed a higher proportion of Bacteroidetes (57 vs. 22%) and a lower proportion of Firmicutes (27 vs. 63%) than Europeans. The authors explained this difference by the higher dietary fiber content of the rural African food, mainly composed of cereals, legumes, and vegetables, which would favor the development of the polysaccharide-degrading Bacteroidetes. Interestingly, the genera Prevotella, Xylanibacter, Cytophaga, and Paludibacter were found exclusively in African microbiota. This is probably due to their increased fitness to grow on polysaccharides abundant in the Burkina Faso diet, such as xylan or cellulose (De Filippo et al., 2010). The control of gut microbiome composition by the diet quality likely denotes a selection of the population that optimally degrades the available substrates. In the case of Bacteroidetes, the selection criteria would primarily be based on the ability to digest complex polymers. Some species may have acquired specific catabolic pathways that others lack. This hypothesis has been recently tested in a mouse model. Germ-free mice were inoculated with two Bacteroides thetaiotaomicron and B. caccae strains of which the latter one can grow with inulin as carbon source. When the mice where fed with an inulin rich diet the ratio of the two species changed toward B. caccae. The ability to use inulin was associated with a GH32 absent in B. -thetaiotaomicron. This clearly showed that diet selects species composition in the animal intestine (Sonnenburg et al., 2010).

The question arises if different human populations with different diets contain specific food adaptations on the genetic level of their gut microbes. Recently, our group has shown that gut Bacteroidetes were able to get gene updates from environmental species to acquire novel functions (Hehemann et al., 2010; Rebuffet et al., 2011). Indeed, in the marine flavobacterium Zobellia galactanivorans, we have discovered and characterized the first porphyranases (Hehemann et al., 2010) as well as a 1,3-α-3,6-anhydro-L-galactosidase (Rebuffet et al., 2011). These enzymes are used by marine bacteria to degrade agarocolloïds, sulfated galactans only found in the cell walls of red algae, such as Porphyra or Gracilaria. When using these new sequences as lead sequences to probe publicly available databases, homologs were identified not only in other marine bacteria, but surprisingly also in the human gut isolate Bacteroides plebeius. The genome of B. -plebeius (DSM 17135) contains a porphyran/agar degradation locus, transferred from an ancestral marine Bacteroidetes (Hehemann et al., 2010; Rebuffet et al., 2011). This PUL was identified as a result of biochemical and structural characterization of the first two porphyranases belonging to the family GH16 (PorA and PorB; Hehemann et al., 2010) and the 1,3-α-3,6-anhydro-L-galactosidase AhgA belonging to the family GH117 (Rebuffet et al., 2011). The PUL contains sequences coding for two putative β-agarases (GH86), one β-agarase (GH16), two β-galactosidases (GH2), a sulfatase, a carbohydrate-binding module, and a susD-like gene associated with its TonB-dependant receptor (Figure 2). Altogether these enzymes form a complete system of detection and degradation for porphyran and agar, which provides B. plebeius with the set of utensils to use these polysaccharides as carbon source.

Metagenomic data revealed that porphyranases and 1,3-α-3,6-anhydro-L-galactosidases are absent in North American and Danish population but present in Spanish and Japanese populations with proportions of 10 and 38% respectively (Hehemann et al., 2010; Rebuffet et al., 2011). The biological rationale of “marine” enzymes in gut microbes could be linked to the high input of sea-derived products in the diet of these two populations. Indeed, the Japanese consume about 14.2 g seaweed per day and person (Fukuda et al., 2007), and the most popular seaweed is Nori (Porphyra spp.) used to make maki-sushi (Nisizawa et al., 1987). Similarly, Spain is the second largest consuming nation of seafood in the world (Manrique and Jensen, 2001). In both populations, contact between human-associated microbes and non-sterile seaweed or seafood could have created a favorable condition for a LGT from marine bacteria to human gut Bacteroidetes. Noteworthy, Nori is the only food that contains porphyran, which allowed associating the transfer of these genes to one special food source. This first evidence of a life style-associated adaptation of the genetic repertoire of the human gut microbiome could be detected due to the unique signature of seaweed degrading CAZymes. Therefore marine CAZymes may be particularly good probes to reveal such adaptations in the human gut microbiome.

These examples of LGT show that gut bacteria are able to acquire new functions via transfer of a complete degradation pathway from food-associated environmental bacteria. Interestingly, the Firmicutes Epulopiscium sp., which lives in the gut of grazer surgeonfish (Clements and Bullivant, 1991), also possesses a putative agar degradation locus acquired from marine bacteria (Rebuffet et al., 2011). One can therefore extrapolate that similar events occurred in omnivorous and herbivorous animals. During the course of evolution, the metabolic repertoire of their gut microbes could have been influenced by contacts with food-associated environmental bacteria. Thus, acquisition of selectively advantageous genes by successive LGT events could explain how gut symbionts acquired CAZymes involved in green plant polysaccharide degradation.

In conclusion, some crucial criteria seem to have particularly favored the adaptation of Bacteroidetes to such contrasting environments, rendering their distribution/dominance close to ubiquitous. Among these, we note the presence of specific Sus-like PULs that are dedicated to polysaccharide degradation (Xu et al., 2003; Flint et al., 2008) and their extreme genome flexibility (Xu et al., 2007) that assures the capacity to gain new functions by LGT from species living in different habitats. Consequently, the recently revealed LGT linking environmental and gut Bacteroidetes (Hehemann et al., 2010) raises the question as to how the ancestral members of the intestinal microbiota might have evolved from plant-degrading species, settled in the GIT of early herbivores, such as marine protists. Evidently, gut symbionts have co-evolved with their hosts (De Filippo et al., 2010), notably in response to the consumption of specific food, and this interdependence likely shaped and still shapes the lifestyle of human populations (Ley et al., 2006a). This “food connection” points toward the fact that recurrent contacts between environmental and gut microbes can have beneficial effects (Rook and Brunet, 2005). At the same time, it underlines the potential problem of our modern lifestyle and the consumption of hyper-hygienic, extensively processed food for human health, depriving us of the environmental reservoirs of microbial genes that allow adaptation by lateral transfer (De Filippo et al., 2010; Sonnenburg, 2010). On the other hand, global travel and trade are providing contact to new types of food and to diverse populations, perhaps giving access to new microbes harboring novel genes destined for integration into our microbiome.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.